Introduction

Top Abstract Introduction Materials and Methods Results and Discussion References

Cell division in Escherichia coli requires a number of conserved essential proteins. The earliest protein to arrive at the division site is FtsZ, which polymerizes to form the Z ring (1, 2). FtsA, ZipA, FtsK, FtsQ, FtsL, YgbQ, FtsW, FtsI, and FtsN are then recruited to the Z ring in sequential order (3-6). ZipA and FtsA localize to the Z ring independently (7, 8), whereas FtsL and YgbQ localization is codependent (3). The colocalization of all of these proteins presumably reflects the assembly of a protein machine at the membrane that orchestrates the growth of the division septum. FtsZ and FtsA are cytoplasmic proteins, and ZipA has a cytoplasmic C terminus and a membrane-bound N terminus (9). The others are integral membrane proteins with extensive periplasmic domains. Some key questions in bacterial cell division are how does this membrane-spanning machine assemble and how many subassemblies exist. Because FtsZ, FtsA, and ZipA all are required for recruitment of the later proteins (10, 11), understanding how these three proteins interact will shed light on the assembly of the entire machine.

ZipA was originally found in a biochemical screen for FtsZ-binding proteins (9) and has several proposed functions, one of which is to bundle FtsZ protofilaments in vitro (12, 13). This interaction between ZipA and FtsZ is mediated via the C termini of both proteins (14-16) and the bundling observed in vitro may translate into increased Z-ring stability in vivo (10, 12). Another proposed function of ZipA is to serve as a membrane anchor for the Z ring (17). This anchor function may be essential, because substitution of another domain with similar topology for the ZipA N-terminal membrane domain did not restore ZipA function (13). The third proposed function for ZipA is the recruitment of downstream proteins, as mentioned above.

FtsA is structurally similar to actin and other members of the ATPase superfamily (18). FtsA, like ZipA, interacts with the C terminus of FtsZ (14, 15, 19, 20), but its function in cell division is unknown. FtsA lacks a membrane-spanning domain but is membrane-associated (21, 22). Whereas Z rings assemble proficiently in the absence of either FtsA or ZipA, Z rings fail to form if both are absent (10). Therefore, it appears that FtsA and ZipA have overlapping roles, both in recruitment of downstream proteins and stabilizing the Z ring. The relative stoichiometry of FtsZ, ZipA, and FtsA is 150:10:1, which is crucial for cell division. For example, excess FtsA or ZipA inhibits cell division unless there is a corresponding increase in FtsZ (9, 23, 24). Interestingly, Bacillus subtilis lacks an obvious ZipA homolog, and FtsA is far more abundant, with the FtsZ/FtsA ratio at 5:1 (25).

Why is ZipA so crucial for E. coli cell division, yet apparently absent in most species outside the enteric bacteria (4)? One possibility is that other proteins fulfill the same role. Another is that FtsA, which is much more widely conserved, might be able to substitute for ZipA if FtsA were at a higher concentration or in a different conformation. This idea prompted us to test whether E. coli could find a way to survive without ZipA. Our hypothesis was that if most bacteria can divide without ZipA, then there might be a way to bypass the need for it in E. coli and that this bypass pathway might shed important light on the functions of FtsA, ZipA, and FtsZ. Here, we show that cell division in E. coli can indeed occur without ZipA, provided there is a corresponding change in the function of FtsA.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results and Discussion References

Media, Chemicals, and Other Reagents.

Antibiotics were added to LB medium at the following concentrations: 40-50 µg/ml kanamycin (kan), 20 µg/ml chloramphenicol, 50 µg/ml ampicillin, 50 µg/ml spectinomycin (spc), and 10 µg/ml tetracycline. L-Arabinose and isopropyl--D-galactopyranoside (IPTG) were added at the concentrations indicated in the text. M9 glucose medium and LB medium have been described (26). Protein markers were from Invitrogen.

Strains and Plasmids.

E. coli host strains for physiological experiments include TX3772 [MG1655 lacU169 (27)], WM1115 [ftsA12 (Ts) in TX3722], and WM1282 (see below). For further details of all strains and plasmids see Table 2, which is published as supporting information on the PNAS web site, www.pnas.org. The ftsAZ region from a zipA suppressor strain (WM1520) was first cloned by replacing the BglII-ClaI fragment in pZAQ (28) with the corresponding genomic fragment from WM1520 by PCR. The resulting plasmid was called pZAQsup, then pZA*Q once the mutation in ftsA was confirmed to be the suppressor. Plasmid pET28-FtsA was made by cloning ftsA into NdeI-BamHI-digested pET28a (Novagen); the NdeI site contains the ftsA start codon. Plasmid pET28-FtsA* was made by replacing an internal KpnI-AscI fragment with the KpnI-AscI fragment from pZA*Q. XbaI-PstI fragments of pET28a-FtsA and pET28a-FtsA* containing ftsA fused to the hexahistidine tag were then cloned into XbaI-PstI-digested pBAD33 (29), creating pBAD-FtsA and pBAD-FtsA*, respectively. To create an IPTG-inducible zipA-GFP fusion (pWM796), the zipA gene lacking a stop codon was PCR-amplified and cloned as a SacI-XbaI fragment to replace ftsZ in pZG (15). Plasmid pWM1703 was made by filling in the XbaI site of pWM796, thus introducing a stop codon after zipA. Plasmid pWM1802 expresses ZipA from the Ptrc promoter of pDSW208 (30). To construct the IPTG-inducible minC expression plasmid pWM1509, minC was cloned into the XbaI site of pDSW209 (30) to make a GFP-minC fusion. To make a GFP-FtsN fusion, ftsN was cloned between the SacI and XbaI sites of pDSW207 (30).

Isolation and Mapping of ftsA R286W.

The ZipA depletion strain containing zipA::aph and a temperature-sensitive plasmid carrying zipA and ftsZ (CH5/pCH32) was obtained from Piet de Boer (Case Western Reserve University, Cleveland) and has been described (9). We renamed this strain WM1282. To isolate suppressors of the zipA null mutation, WM1282 cells were grown at 28°C in LB kan to early logarithmic phase, then shifted to 42°C, diluted once, and incubated at 42°C for 3 days. After allowing for settling of the long filaments, a small aliquot from the top was removed and plated on prewarmed LB kan plates at 42°C. Four colonies appeared, which were purified on LB kan and LB spc at 42°C. The loss of the temperature-sensitive zipA covering plasmid pCH32 was confirmed by the lack of growth on LB spc. P1 phage lysates were made from all four survivor strains as well as WM1282 and used to transduce TX3772 (WT) or TX3772 containing pCH32 (WM1302) to Kan^R at 32°C. All five of the lysates resulted in 50-100 Kan^R transductants with TX3772/pCH32 recipients but no transductants with TX3772 recipients, indicating that the suppressor was not linked to zipA::aph.

To determine linkage to ftsAZ, WM1302 was transduced to tetracycline^R at 30°C with a P1 phage lysate from a leu::Tn10 strain, WM1109. The Leu phenotype was verified by the failure of the cells to grow on M9 glucose plates lacking leucine. WM1302 leu::Tn10 was then transduced to kan^R at 30°C with a phage P1 lysate from WM1282, introducing zipA::aph into the chromosome and making the cells thermosensitive. This strain in turn was then transduced to Leu⁺ (growth on M9 glucose plates) with a P1 lysate made from the zipA suppressor strain WM1520. This process allowed 50% cotransduction of the ftsQAZ region, and thus the suppressor, into this strain. One of the many transductants able to form colonies at 42°C was saved and grown further at 42°C to remove pCH32; it was then confirmed to be Spc^S. WM1659 (ftsA*) was constructed by transducing WM1109 (TX3772 leu::Tn10) to Leu⁺ with a P1 lysate from WM1520. This process allowed cotransduction of ftsA* into this strain at 50% efficiency. WM1659 was then transduced to Kan^R with the zipA::aph allele from WM1282 to make WM1657.

Antibody Production and Immunoblot Analysis.

To obtain anti-ZipA antiserum, we purified the C-terminal cytoplasmic domain of ZipA as was done in ref. 13. This segment of zipA, corresponding to residues 186-328, was cloned between the BamHI and XhoI sites of pET28a+ to create pWM1610. This plasmid carries the zipA segment fused to T7 and hexahistidine tags at the N terminus and a hexahistidine tag at the C terminus [HT-ZipA(186-328)-H]. The 18.4-kDa HT-ZipA(186-328)-H protein was expressed in BL21(DE3)/plysS cells carrying pWM1610 after induction with 1 mM IPTG for 3 h and affinity purification as described (13). The purified protein was used to raise rabbit anti-ZipA polyclonal antiserum (Capralogics, Hardwick, MA). To detect GFP-MinC and FtsZ, we used affinity-purified polyclonal anti-GFP and anti-FtsZ, as described (27). For immunoblot analysis, 15 µg of total protein from each sample was separated by SDS/PAGE and transferred to a nitrocellulose membrane. The membrane was blocked with gelatin and probed with primary antibody (1:5,000 for anti-ZipA) followed by goat-anti-rabbit-conjugated horseradish peroxidase secondary antibody, and developed with ECL reagents (Amersham Pharmacia). A Fluorichem Imager was used to quantitate the chemiluminescence of the resulting protein bands.

Microscopy and Analysis of Cells.

In most cases, phase-contrast images were taken of wet mounts of cells. For determination of cell lengths, cells were fixed on coverslips before microscopic analysis. Immunofluorescence microscopy with anti-FtsZ antibody was performed as described (27). Cell lengths were measured with PHOTOSHOP (Adobe Systems, Mountain View, CA) and statistics were compiled in EXCEL (Microsoft). Growth rates were followed by optical density with a Klett-Summerson photometer. Viable counts were determined by counting the number of colonies after plating dilutions of cells on LB agar.

Molecular Modeling and Protein Alignments.

The FtsA crystal structure was imported from the Protein Data Bank and manipulated graphically with WEBLAB VIEWERLITE 3.20 (Molecular Simulations, Waltham, MA). Protein alignments were obtained by using BLASTP at the National Center for Biotechnology Information web site (www.ncbi.nlm.nih.gov), and then optimized manually.

	Results and Discussion

Top Abstract Introduction Materials and Methods Results and Discussion References

Isolation of Suppressors of a zipA Null Mutant.

Because ZipA is not highly conserved, we hypothesized that it might be possible to isolate mutants of E. coli that could bypass the requirement for ZipA in cell division. To select for such mutants, we used WM1282, a recombination-deficient strain harboring a zipA::aph (Kan^R) disruption and a Spc^R zipA-containing plasmid (pCH32) that is thermosensitive for replication (9). The cells were grown at 30°C and then shifted to 42°C for many generations to promote loss of the plasmid and select for suppressors. Plating at 42°C yielded four Spc^S Kan^R colonies that lacked the zipA-covering plasmid. All four of these strains, which may have been siblings as they were from the same culture, were able to transfer Kan^R via P1 transduction to a WT recipient strain carrying pCH32 but not to a strain lacking the plasmid, suggesting that zipA::aph was still intact in these survivor strains and that the suppressor was not located within cotransduction distance. The presence of the chromosomal zipA::aph disruption was confirmed by PCR, sequencing, and Southern blotting (data not shown). These results suggested that we had isolated one or more extragenic suppressor mutations that could bypass the requirement for zipA.

The absence of detectable ZipA protein in the zipA::aph suppressor strains was confirmed by immunoblotting with an antibody raised against a C-terminal domain of ZipA (Fig. 1, lane 3, and data not shown). As has been observed (9), ZipA protein migrated with an apparent molecular mass of 50 kDa despite its predicted size of 36.5 kDa (Fig. 1, lane 1). Specificity of our antibody for ZipA was demonstrated in three ways. First, overproduction of ZipA resulted in a more intense 50-kDa band compared with the control (Fig. 1, lane 4, compare with lane 1). Second, overproduction of ZipA-GFP resulted in a new band at 80 kDa, the predicted size of the 50-kDa ZipA fused to 27-kDa GFP (Fig. 1, lanes 2 and 5). Third, the 50-kDa band became significantly less intense after depletion of ZipA by growth of WM1282 at 42°C for 4 h (Fig. 1, compare lanes 6 and 7).

View larger version (21K): [in this window] [in a new window]   Fig. 1.   Immunoblot showing cellular ZipA in strains used in this study. Lane 1, TX3772 (WT); lane 2, pWM796 (ZipA-GFP) in WM1659 (ftsA*) + 1 mM IPTG; lane 3, WM1657 (ftsA* zipA); lane 4, pWM1703 (ZipA) in WM1657 (ftsA* zipA) + 1 mM IPTG; lane 5, pWM796 (ZipA-GFP) in WM1657 (ftsA* zipA) + 1 mM IPTG; lane 6, WM1282 grown at 42°C for 4 h to deplete ZipA; lane 7, WM1282 grown at 30°C. ftsA* is the suppressor allele. Positions of 50-and 81-kDa protein markers are shown on the left.

A Single Residue Change in FtsA Is Necessary to Confer ZipA Independence.

Because ZipA, FtsA, and FtsZ are known to interact, we reasoned that the suppressor mutations might map to the ftsAZ region. To determine linkage to ftsAZ, we used a recipient strain for phage P1 transduction that contained a leu::Tn10 marker, which is 50% linked to ftsZ by cotransduction, zipA::aph from WM1282, and the thermosensitive zipA-covering plasmid (pCH32). Using one of the zipA::aph suppressor strains (WM1520) as a donor, this thermosensitive recipient was transduced to Leu+. Approximately 50% of the transductants became thermoresistant, suggesting that the zipA suppressor mutation was successfully cotransduced and that it mapped in or near ftsAZ.

To verify that the suppressor was in ftsAZ and to map it more precisely, we PCR-amplified a BglII-ClaI fragment containing most of ftsA and all of ftsZ from the original suppressor strain WM1520 and replaced the native ftsAZ in pZAQ with these ftsAZ sequences. WT TX3772 cells containing this new plasmid (pZAQsup) or the original pZAQ were then transduced with zipA::aph. Surprisingly, Kan^R colonies were isolated at a similar frequency in both recipient strains (see below). However, pZAQ-containing cells were highly filamentous, whereas the pZAQsup-containing cells were short (see below). Therefore, the suppressor mapped to the ftsAZ region downstream from the BglII site in ftsA. DNA sequencing of the entire ftsAZ region containing the suppressor revealed a single base change from the WT sequence, a C  T transition at base pair 856 of ftsA, corresponding to an arginine-to-tryptophan change at position 286 in the FtsA protein. All four original suppressor strains contained the same mutation, which we will refer to hereafter as ftsA*. Likewise, pZAQsup was renamed pZA*Q. R286W may be the only ftsA mutation that can cause the bypass, because when we deleted zipA by an independent method and selected independent surviving colonies, all of those tested contained the R286W change (B.G. and W.M., unpublished results).

Mostly Normal Cell Division in Cells Containing ftsA* in Place of WT ftsA.

Although ftsA* allowed survival of zipA cells, we were interested to know what effects ftsA*, either alone or with zipA, had on cell division and growth. We constructed two derivatives of TX3772 that contained ftsA* instead of native ftsA; one was zipA+ (WM1659) and the other zipA (WM1657) (see Materials and Methods). The strains grew normally, as doubling times for TX3772, WM1659, and WM1657 in LB were 29, 30, and 31 min, respectively, at 37°C and 115, 111, and 111 min, respectively, at 28°C. Viable counts were very similar under both conditions. The morphology of most ftsA* cells, whether zipA+ or zipA, was similar to those of WT cells; the same was true for Z rings as examined by immunofluorescence (Fig. 2 A-F). The most notable differences were that (i) a small proportion of ftsA* cells exhibited twisting at the septum that was made more visible after elongating the cells with cephalexin (Fig. 2 G-J), and (ii) some minicells were observed for WM1659 but not the others (data not shown). Minicells may arise because of the ftsA*-mediated resistance to the MinC division inhibitor (see below). We conclude that the cell division deficiency of cells lacking ZipA is suppressed by ftsA*, although cell division is not completely normal in cells with ftsA*.

View larger version (73K): [in this window] [in a new window]   Fig. 2.   Mostly normal cell division with ftsA* replacing WT ftsA. Shown are WT strain TX3772 (A and D), ftsA* zipA+ strain WM1659 (B and E), and ftsA* zipA strain WM1657 (C and F). (A-C) Phase-contrast images. (D-F) Immunofluorescence microscopy for FtsZ (black staining) of the cells in A-C, respectively. (G-J) Phase-contrast images of nondividing WM1657 cells after growth in cephalexin for 2 h to accentuate the twisting at the potential septum. (Scale bar = 5 µm.)

The ftsA* Mutation Is Sufficient to Bypass the Requirement for ZipA.

To confirm that mutations other than ftsA* and coexpression of ftsZ or ftsQ were not necessary for ZipA-independent cell division, we introduced plasmids pBAD-FtsA and pBAD-FtsA* into a thermosensitive ftsA mutant (WM1115) and confirmed that either ftsA or ftsA* could complement the mutant without addition of arabinose inducer. With arabinose, WT cells with either plasmid became filamentous, which is the expected FtsA oveproduction phenotype (23, 24). However, when the pBAD derivatives were introduced into the zipA depletion strain WM1282, only pBAD-FtsA* allowed colony formation at 42°C with or without arabinose (Fig. 3 and data not shown). As expected, WM1282 cells containing pBAD-FtsA became filamentous upon shifting to 42°C, whereas cells containing pBAD-FtsA* remained short (data not shown). Likewise, transduction of TX3772 containing pBAD-FtsA* with zipA::aph yielded viable colonies but TX3772/pBAD-FtsA did not, either with or without arabinose (data not shown). We conclude from these data that synthesis of FtsA* alone, but not FtsA alone, is sufficient to bypass the requirement for ZipA, and that no secondary mutations are necessary. In support of this conclusion, FtsZ levels normalized to total cell protein as measured by Western blots were similar in ftsA* and ftsA+ cells (data not shown), ruling out the possibility that ftsA* mutation significantly increased ftsZ expression.

View larger version (37K): [in this window] [in a new window]   Fig. 3.   FtsA* is sufficient to bypass the need for ZipA. Colony growth is shown at 32°C or 42°C on LB chloramphenicol plates without arabinose for the ZipA depletion (WM1282) strains containing the plasmids indicated.

FtsA* Eliminates the Requirement for ZipA in Recruitment of Downstream Cell Division Proteins.

The bypass of the ZipA requirement for cell division suggests that all functions of ZipA, including recruitment of later cell division proteins, are now taken care of by FtsA*. However, another possibility, albeit unlikely, was that in the presence of FtsA*, later cell division proteins were no longer recruited nor required for division. We addressed this question in two ways. First, when WM1657 and WM1659 cells were treated with cephalexin, which inhibits the late cell division protein FtsI, the cells became filamentous (Fig. 2 G-J). This finding indicates that FtsI remains essential. Second, a GFP fusion to the last known cell division protein recruited to the division site, FtsN, localized correctly to the division site independently of ZipA as long as FtsA* was present (data not shown). These results suggest that in the absence of ZipA the functions of later cell division proteins remain intact.

This conclusion has important implications for the assembly of the cell division protein complex. ZipA is required for recruitment of late cell division proteins (11). Our results indicate that it is unlikely that ZipA recruits these proteins directly, but instead normally acts via stimulation of recruitment by FtsZ, FtsA, or another as yet unknown factor.

Effects of Multicopy Coexpression of ftsQA*Z and ftsQAZ.

Because ftsA* does not appear to have a significant phenotype when in single copy other than some minicells and twisted septa with cephalexin, we investigated the effects on cell division when ftsA* was expressed in multiple copy along with ftsQ and ftsZ. In zipA+ cells, plasmid pZAQ, which carries native ftsQAZ and some flanking sequences, produced a mixture of minicells and cells of mostly normal length in zipA+ cells (Fig. 4B and data not shown). This finding is consistent with previous reports (31, 32). In contrast, a significant proportion of cells containing pZA*Q (pZAQ with the single-residue change to make ftsA*) displayed pronounced morphological abnormalities in zipA+ cells (Fig. 4A) whereas pZA*Q/zipA- cells looked much like pZAQ/zipA+ cells (Fig. 4C). More than half of pZA*Q/zipA+ cells exhibited twisted septa that were much more extensive than in cells with single-copy ftsA*, abnormal polar morphology, cell chains, and other morphological aberrations. These results suggested that the septa might be abnormally stable, so FtsZ staining was examined by immunofluorescence microscopy. In contrast to pZAQ/zipA+ cells, which had Z rings at potential division sites and cell poles (Fig. 4 G and J), FtsZ staining in pZA*Q/zipA+ cells was much more irregular, varied greatly in intensity among cells, and often exhibited helical patterns (Fig. 4 E, F, H, and I, arrows in H and I). These results suggested that Z rings might be unusually stable in zipA+/pZA*Q cells, but we could not always correlate FtsZ staining with any particular type of morphological aberration. For example, some obvious twisted septa had FtsZ staining, but others had no detectable FtsZ staining, perhaps because the cells were already dead. The abnormal morphology was not observed with pZAQ (Fig. 4C) or after overproduction of FtsA or FtsA* alone (data not shown), suggesting that coexpression of ftsQ and/or ftsZ with ftsA* in multicopy is required for the phenotype.

View larger version (79K): [in this window] [in a new window]   Fig. 4.   Effects of overproduction of FtsA or FtsA* on cell morphology. (A-D) Representative phase-contrast images of pZA*Q in TX3772 (A), pZAQ in TX3772 (B), pZA*Q in TX3772 transduced with zipA::aph (C), and pZAQ in TX3772 transduced with zipA::aph (D). (E-J) Nomarski (E-G) and FtsZ immunofluorescence microscopy (H-J) of pZA*Q in TX3772 (E, F, H, and I) and pZAQ in TX3772 (G and J). Arrows in H and I highlight helical FtsZ patterns. (Scale bar = 5 µm.)

We found that pZAQ allowed growth of colonies transduced with zipA::aph, although the cells were very filamentous (Fig. 4D). This finding suggested that perhaps overproduced WT FtsA could mimic the chromosomal ftsA* allele and partially suppress the absence of zipA. However, as mentioned above, overproduction of FtsA by arabinose induction of pBAD-FtsA did not suppress zipA::aph or increase viability of WM1282 at 42°C. This result suggests that coexpression of ftsQ and/or ftsZ on pZAQ is probably necessary for this weak suppression.

FtsA* Suppresses Cell Division Inhibition Caused by Overproduction of MinC and ZipA.

Overproduction of MinC destabilizes Z rings, resulting in cell filamentation (33). The potential stabilization of septa by excess FtsA* from pZA*Q (Fig. 4) prompted us to test whether Z rings in FtsA*-containing cells were more resistant to MinC than WT cells. We used a GFP-MinC fusion (pWM1509) because it has been shown to be functional (34) and we could use anti-GFP antibodies to quantitate its cellular levels. IPTG-mediated induction of GFP-MinC overproduction for 2 h from pWM1509 resulted in a 4-fold increase in GFP-MinC levels in TX3772 and WM1659 (Table 1 and Fig. 6, which is published as supporting information on the PNAS web site). As expected, induced TX3772 cells elongated >10-fold on average compared with uninduced cells (Table 1). In contrast, induced cells of the ftsA* strain WM1659 were only twice as long as uninduced ftsA* cells on average. GFP-MinC induction resulted in an 1,000-fold loss of viability of TX3772 cells, as compared with WM1659 overproducing GFP-MinC, uninduced TX3772, or uninduced WM1659 (Table 1).

Although Z-ring stability can be enhanced by doubling zipA gene dosage and possibly increasing ZipA protein level (12), overproduction of ZipA several-fold also inhibits cell division, which can be reversed by increasing FtsZ levels (9, 13). Because FtsA* renders ZipA unnecessary, we investigated whether FtsA* might confer resistance to the negative effects of excess ZipA. IPTG-mediated induction of ZipA overproduction from pWM1802 for 2 h in either TX3772 or WM1659 (ftsA*) resulted in a 2-fold increase in ZipA levels over uninduced levels in either strain (Table 1 and Fig. 7, which is published as supporting information on the PNAS web site). WT TX3772 cells overproducing ZipA became highly filamentous, elongating 7-fold on average compared with uninduced cells. In contrast, WM1659 (ftsA*) cells with 2-fold excess ZipA did not elongate at all compared with uninduced cells (Table 1). ZipA overproduction in TX3772 also resulted in 10-fold less viability as compared with induced WM1659 or uninduced TX3772 and WM1659 (Table 1). Similar results were obtained in experiments with pWM1703, another plasmid that expresses ZipA (data not shown). Previously, at least 6-fold higher levels of ZipA or several tagged and deletion derivatives of ZipA were needed to observe significant cell filamentation (13); however, our results with inhibition of cell division after only 2-fold overproduction of ZipA are reproducible and may result from differences in strain backgrounds or induction conditions.

These data demonstrate that the presence of ftsA* resists cell division inhibition and killing caused by excess ZipA or MinC (see also Figs. 6 and 7). Whereas the mechanism by which excess ZipA inhibits septation is not clear, the MinC resistance of ftsA* suggests that FtsA* acts to stabilize Z rings. Because overproduction of WT FtsA can suppress MinC-mediated cell filamentation (35), we suggest that ftsA* may mimic overproduction of FtsA, at least for anti-MinC activity. Moreover, our results indicate that ftsA* not only bypasses the requirement for ZipA, but also antagonizes the negative effects of excess ZipA.

Conservation of R286 and Its Proposed Location in the FtsA Molecule.

An alignment of FtsA proteins shows that R286 is well conserved among the -proteobacteria, which is the same group of bacteria that contain ZipA homologs (Fig. 5B). Of the species known to have ZipA, all have arginine or lysine at this position except for Pasteurella multocida, which has a glutamate. However, many species lacking ZipA also often have arginine at this position. With one exception, tryptophan, phenylalanine, and tyrosine are absent at this position, ruling out the possibility that W286 is required for viability in species lacking a ZipA homolog.

View larger version (43K): [in this window] [in a new window]   Fig. 5.   Molecular modeling of FtsA* (R286W) on the FtsA crystal structure and conservation of R286 across species. (A) A ribbon diagram of FtsA from T. maritima, with K287 highlighted as a ball-and-stick representation within the S13 domain. (B) An amino acid alignment of the region surrounding the FtsA S13 -sheet (IKTTT in T. maritima) based on the presence or absence of the conserved arginine. From the top down, the groupings are , , , and  Gram-negative proteobacteria, other Gram-negative proteobacteria, Gram-positive bacteria, and the most divergent species. Species in all caps have ZipA homologs; species underlined have the conserved arginine.

Residue R286 of E. coli FtsA appears to be equivalent to K287 of Thermotoga maritima FtsA, which has been crystallized (Fig. 5B). K287 lies within the S13 -sheet of subdomain 2B, at the top of the monomer structure (18) (Fig. 5A). This subdomain has high sequence conservation among FtsA homologs (Fig. 5B) and some conservation within other members of the actin family. However, the sequence similarity in this region between E. coli and T. maritima is too low to be certain of its precise position in the structure or whether R286 faces in or out of the E. coli FtsA protein. It is intriguing that (i) K287 in T. maritima can form a salt bridge with D250 within the same subunit (data not shown), and (ii) residues within the similar S13 loop of MreB, another actin family member, contact residues in helix H8 of the previous subunit during polymerization of MreB protofilaments (36). No similar salt bridge occurs in E. coli FtsA, as there are no acidic residues in the vicinity of the D250 equivalent. However, the MreB functional model suggests that perhaps the R286W change causes a significant alteration in the ability of FtsA monomers to self-interact (37). Another possibility is that this region is involved in contacting FtsZ and the R286W change alters the interaction. Clearly, future biochemical characterization of the FtsA and FtsA* proteins and their interactions with FtsZ is needed, as well as more information about the function of specific domains and residues in FtsA.

Implications for ZipA and FtsA Function.

One possible explanation for why bacteria outside the -proteobacterial family apparently lack ZipA homologs is that they contain ZipA-like homologs too divergent to discern by sequence comparisons. However, it is also possible that these other species contain an FtsA with more FtsA*-like properties and/or have significantly higher intracellular concentrations of FtsA, therefore naturally bypassing the requirement for ZipA. The possibility that higher FtsA levels might be sufficient in some situations stems from our finding that elevated levels of WT FtsA from pZAQ can partially compensate for the lack of ZipA. B. subtilis, whose FtsA is present at significantly higher concentrations than that of E. coli (25), may be an example of this strategy.

Nevertheless, it remains surprising that in E. coli ftsA* can bypass the need for ZipA without a significant cost. The normal growth and mostly normal cell division phenotype in the absence of ZipA indicates that FtsA* can perform all tasks attributable to ZipA, including protein recruitment, tethering FtsZ to the membrane, and stabilizing assembled FtsZ. These conclusions support the idea that ZipA primarily acts as an accessory factor that regulates FtsZ assembly. Other candidates for this type of accessory role are the nonessential B. subtilis proteins ZapA and EzrA, which positively and negatively modulate FtsZ assembly, respectively (38, 39). Interestingly, although EzrA and ZipA have opposing effects on FtsZ assembly, both have similar topology, with a cytoplasmic C terminus and a membrane-bound N terminus. It is likely that these accessory factors function to maintain a balance between FtsZ assembly and FtsZ disassembly, and that the added putative stabilization function of FtsA* just happens to compensate nearly perfectly for the loss of ZipA stabilization activity. Just as new factors involved in cell division are being discovered, our results here demonstrate that some factors may be found to be replaceable under the right conditions. This approach should help to define the roles of these factors in E. coli and diverse organisms and may also help to minimize the number of purified proteins needed for future in vitro studies of cell division.