In Escherichia coli, at least 12 proteins colocalize to the cell midpoint, assembling into a membrane-associated protein machine that forms the division septum. Many of these proteins, including FtsK, are essential for viability but their functions in cell division are unknown. Here we show that the essential function of FtsK in cell division can be partially bypassed. Cells containing either the ftsA R286W mutation or a plasmid carrying the ftsQAZ genes suppressed a ftsK44(ts) allele efficiently. Moreover, ftsA R286W or multicopy ftsQAZ, which can largely bypass the requirement for the essential cell division gene zipA, allowed cells with a complete deletion of ftsK to survive and divide, although many of these ftsK null cells formed multiseptate chains. Green fluorescent protein (GFP) fusions to FtsI and FtsN, which normally depend on FtsK to localize to division sites, localized to division sites in the absence of FtsK, indicating that FtsK is not directly involved in their recruitment. Cells expressing additional ftsQ, and to a lesser extent ftsB and ftsN, were able to survive and divide in the absence of ftsK, although cell chains were often formed. Surprisingly, the cytoplasmic and transmembrane domains of FtsQ, while not sufficient to complement an ftsQ null mutant, conferred viability and septum formation in the absence of ftsK. These findings suggest that the N-terminal domain of FtsK is normally involved in stability of the division protein machine and shares functional overlap with FtsQ, FtsB, FtsA, ZipA and FtsN.
Cell division in Escherichia coli requires the co-ordinated assembly of at least 12 proteins into a molecular machine, the divisome (Errington et al., 2003; Weiss, 2004). The first known step of this process involves the tubulin homologue FtsZ (Erickson, 1997), which polymerizes into a ring structure, called the Z ring, at the potential site of division (Bi and Lutkenhaus, 1991). Once formed, the Z ring recruits a number of additional division septum-synthesizing proteins either directly or indirectly. The first of these, ZipA and FtsA, have overlapping roles in cell division (Pichoff and Lutkenhaus, 2002). A gain of function mutation in FtsA, FtsA R286W or FtsA*, can completely bypass the requirement for ZipA in cell division, indicating that ZipA plays an accessory role and does not directly recruit downstream proteins (Geissler et al., 2003). A linear recruitment pathway for the divisome proteins has been proposed based on the ability of each protein to localize to the division site upon inactivation or removal of other divisome proteins (Buddelmeijer and Beckwith, 2002) (Fig. 1).
The recruitment of FtsK to the Z ring is dependent on FtsZ, FtsA, ZipA, and FtsEX, but is independent of FtsQ and subsequent proteins in the recruitment pathway (Wang and Lutkenhaus, 1998; Hale and de Boer, 2002; Pichoff and Lutkenhaus, 2002; Schmidt et al., 2004). Thermoinactivation of an ftsK44(ts) allele results in the formation of multiseptate filamentous cells and loss of viability, suggesting that FtsK acts at a late stage of septum formation. However, depletion of FtsK results in smooth filaments lacking any sign of septal invagination, suggesting that the thermoinactivated FtsK44 is partially functional and that FtsK has an early role in septum formation (Draper et al., 1998; Wang and Lutkenhaus, 1998; Chen and Beckwith, 2001). Despite all that is known about FtsK localization and the function of its C terminus, the precise role of FtsK in cell division remains elusive.
The protein immediately after FtsK in the recruitment pathway, FtsQ, is a bitopic membrane protein with a 24 amino acid cytoplasmic domain, a 25 amino acid transmembrane domain, and a 227 amino acid periplasmic domain (Carson et al., 1991; Guzman et al., 1997). The periplasmic domain of FtsQ is sufficient for mid-cell localization when fused with a non-FtsQ cytoplasmic and transmembrane domain and green fluorescent protein (GFP) (Chen et al., 1999). This construct is also able to complement ftsQ(ts) and ftsQ null strains, indicating that the periplasmic domain of FtsQ is responsible for the required function of FtsQ in cell division. Like FtsK, FtsQ is essential for cell division and is conserved in a wide variety of bacteria (Margolin, 2000).
Although much evidence supports the linear localization dependency pathway, recent findings suggest that various protein subassemblies act together to form the complete divisome. The interaction of essential division proteins ZipA and FtsA with the C terminus of FtsZ is well established (Ma and Margolin, 1999; Hale et al., 2000; Mosyak et al., 2000; Yan et al., 2000). Recently, a complex of FtsQ, FtsL and FtsB has been detected by immunoprecipitation (Buddelmeijer and Beckwith, 2004). This complex forms independently of FtsK, the previous component in the recruitment pathway. In support of this uncoupling, a fusion of FtsQ to ZapA, which interacts directly with FtsZ (Low et al., 2004), is able to recruit FtsL and FtsI to the Z ring independently of FtsK (Goehring et al., 2005). Intriguingly, ZapA-FtsQ cannot recruit GFP-FtsN to the division site, but can back-recruit GFP-FtsK. A potential subassembly between FtsA, FtsI and FtsN has been inferred using two different types of two-hybrid assays in E. coli (Corbin et al., 2004; Karimova et al., 2005). Similar assays indicate that there is a web of interactions among cell division proteins that complicate the linear recruitment model (Di Lallo et al., 2003; Karimova et al., 2005).
It has been proposed that the bitopic membrane proteins which are essential for cell division, including FtsQ, FtsL, FtsB and FtsN, serve to link cytoplasmic and periplasmic divisome components (Guzman et al., 1997; Ghigo and Beckwith, 2000; Yang et al., 2004). Because the periplasmic portions of FtsQ and FtsN are sufficient for the function of either protein in cell division (Dai et al., 1996; Guzman et al., 1997; Ursinus et al., 2004), it suggests that their short cytoplasmic and transmembrane domains either serve solely as membrane anchors or have as yet unknown accessory functions. Herein we show that the cell division function of the N terminus of FtsK can be replaced to a significant degree by several factors, including FtsA*, increased coexpression of FtsQ, FtsA and FtsZ, and increased FtsQ alone. Plasmid-borne FtsN or FtsB could also compensate for the loss of FtsK to a lesser extent. Although cells lacking FtsK under any of the above compensatory conditions were not as proficient at dividing as wild-type cells, the evidence presented here suggests that FtsK does not have a unique function in cell division.
Suppression of ftsK(ts) and ftsQ(ts) alleles by ftsA*and pZAQ
The ability of FtsA R286W (FtsA*) to fully suppress the requirement for ZipA in E. coli cell division prompted us to ask whether FtsA* could suppress the inactivation of other division proteins. We transformed a plasmid expressing either wild-type FtsA (pBAD-FtsA) or FtsA* (pBAD-FtsA*) into various strains containing temperature-sensitive alleles of cell division proteins. Uninduced expression of ftsA or ftsA* from these plasmids is sufficient to complement an ftsA mutant (Fig. 2A) or, in the case of ftsA*, to suppress the requirement for zipA (Geissler et al., 2003). Each strain was streaked onto Luria-Bertani (LB) agar at the non-permissive (42°C) or permissive (30°C) temperature and incubated overnight. As shown in Fig. 2A, pBAD-FtsA* was able to suppress the thermosensitivity of the ftsQ1 and ftsK44 alleles. In contrast, neither pBAD-FtsA* nor pBAD-FtsA were able to suppress the ftsI2158(ts) allele (Fig. 2A and data not shown). When grown in broth, cells of ftsK44 strains containing pBAD-FtsA* exhibited a nearly wild-type appearance at 30°C or 42°C whereas those containing pBAD-FtsA or pBAD33 formed non-dividing filaments at 42°C (Fig. 2B–D). In comparison, cells of ftsQ1 strains containing pBAD-FtsA* were somewhat longer than the corresponding ftsK44/pBAD-FtsA* cells, but were still significantly shorter than the non-dividing filaments seen with pBAD-FtsA or pBAD33 at 42°C (Fig. 2E–G). Therefore, pBAD-FtsA* suppressed the thermosensitive cell division defects of ftsK44 and ftsQ1, although the suppression of the latter was not as complete. The ftsI2158 strain, as expected, remained filamentous in the presence of FtsA* (Fig. 2H–J). In dilution plating experiments, ftsK44 or ftsQ1 strains grown at 42°C for 7 h produced at least 100-fold more colony-forming units with pBAD-FtsA* as compared with those with pBAD-FtsA (data not shown). These data are consistent with the ability of FtsA* to suppress the cell division defects of ftsK44 and ftsQ1 at the non-permissive temperature.
Previously, we showed that the presence of a plasmid, pZAQ, which contains the ftsQAZ gene operon (Bi and Lutkenhaus, 1990), could also partially suppress the loss of ZipA (Geissler et al., 2003). As a result, we tested whether extra ftsQAZ expression could suppress the requirement for other cell division proteins by repeating the temperature shift assays described above with pZAQ instead of pBAD-FtsA*. Thermosensitive strains of ftsZ, ftsA, ftsK and ftsQ (ftsZ84, ftsA12, ftsK44 and ftsQ1) containing pZAQ formed colonies on plates incubated at 42°C, indicating that pZAQ not only complemented ftsZ84, ftsA12 and ftsQ1 as expected, but also ftsK44 (data not shown). To examine the effect of pZAQ in these thermosensitive mutants on individual cells in broth, we grew cells to early logarithmic growth (OD600∼0.1), then shifted them to pre-warmed medium at 42°C and grew them for ∼4 h. Similar to our plate results, pZAQ was able to suppress the thermosensitivity of ftsZ84, ftsA12, ftsK44 and ftsQ1 mutants in liquid broth (Fig. 2L–M and data not shown). Despite its ability to grow on plates at the non-permissive temperature, the ftsK44 strain containing pZAQ formed a mixture of filaments and normal sized cells at 42°C (Fig. 2M), but elongated cells were also observed upon expression of GFP-FtsK1-859 in the ftsK44 strain (Fig. 2K).
Suppression of ftsK44 was not dependent on production of all three proteins from pZAQ, as strains containing either pZA′Q or pZAQ′, which inactivate ftsA or ftsQ respectively, but leave the other two genes functional, could form colonies on plates and appeared similar to pZAQ in broth at 42°C (data not shown). Therefore, extra FtsA+FtsZ or extra FtsQ+FtsZ could also serve to suppress the thermosensitivity of ftsK44. Importantly, however, expression of extra FtsZ alone from pWM404 could complement ftsZ84 but could not suppress the thermosensitivity of ftsA12, ftsK44, ftsQ1 or ftsI2158 (data not shown), suggesting that ftsA and/or ftsQ are the genes expressed from pZAQ that are important for suppression. As expected from the lack of complementation of the ftsI2158 mutant by pBAD-FtsA*, pZAQ was also unable to suppress the thermosensitivity of ftsI2158 either on plates or in broth (data not shown).
Suppression of a null allele of ftsK
We next determined if the null alleles of ftsK and ftsQ could also be suppressed. We first constructed a set of chromosomal deletions, replacing ftsK, ftsQ, ftsL, ftsB or ftsW with the aph gene that confers kanamycin resistance (see Experimental procedures). As they are normally lethal, these null alleles were complemented by plasmids that express the appropriate gene. Because ftsK44 was efficiently suppressed by FtsA* and pZAQ, we first tested the ftsK null mutant. When grown under FtsK depletion conditions, the ΔftsK::kan strains (WM2109 or WM2294) produced filamentous cells that initially contained multiple septa, but over time became largely non-septate (Fig. 3A–C and data not shown). This phenotype is similar to those previously described for other FtsK depletion strains (Draper et al., 1998; Wang and Lutkenhaus, 1998; Chen and Beckwith, 2001).
To test whether the ftsK null allele could be suppressed, the ΔftsK::kan marker was transferred into the wild-type E. coli strain W3110 carrying either pZAQ or pBAD-FtsA* by phage P1 transduction, selecting for kanR. A similar method was used to show previously that the ftsA* allele, and pZAQ to a lesser extent, can permit ΔzipA::kan cells to form colonies (Geissler et al., 2003). We were able to transduce ΔftsK::kan into both pZAQ and pBAD-FtsA* containing cells at high frequency. Various combinations of polymerase chain reaction (PCR) primers were used to confirm the complete deletion of ftsK from the chromosome as well as the proper location of aph within ftsK (Fig. 3D).
We next determined if the proteins downstream of ftsK in the septal recruitment pathway (Fig. 1) were still essential for viability in the presence of FtsA*. Because FtsA* was able to suppress the thermosensitivity of the ftsQ1 allele, we first attempted to introduce a ΔftsQ::kan allele into recipient cells carrying pBAD-FtsA*. However, we were unable to introduce this null allele unless the recipient contained ftsQ on a plasmid (data not shown). As a result, we conclude that FtsA* was not able to compensate for the loss of FtsQ. To test whether other downstream cell division genes could be suppressed, we transduced W3110 containing pZAQ or pBAD-FtsA* with P1 lysates from ΔftsB::kan,ΔftsL::kan, or ΔftsW::kan deletion strains, but were unable to isolate any transductants containing allele replacements (data not shown). These results suggest that of the cell division genes tested, only ftsK could be completely deleted and remain viable in the presence of ftsA* or pZAQ.
Morphology ofΔftsK::kan cells suppressed by ftsA*or extra ftsQAZ
Although pZAQ and pBAD-FtsA* could partially compensate for the loss of ftsK in terms of viability, there was still a significant alteration in cellular morphology. Importantly, unlike the smooth filamentation seen during FtsK depletion in otherwise wild-type cells (Fig. 3C), both pZAQ/ΔftsK::kan and pBAD-FtsA*/ΔftsK::kan cells formed a mixture of normal-length and multiseptate chains (Fig. 4F and data not shown). At 30°C, the W3110 parent, pZAQ/ΔftsK::kan and ftsA*/ΔftsK::kan strains grew with doubling times of 75, 81 and 120 min respectively. The percentage of cells that were of normal length was 29–39% (Table 1 and data not shown), with the remainder present as minicells, short filaments or cell chains. These latter cell division defects are consistent with the ∼10X loss in viable colony-forming units measured for the pZAQ/ΔftsK::kan strain as compared with the pZAQ/ftsK+strain at 37°C (Fig. 4A). Interestingly, pZA*Q/ΔftsK::kan cells are even less viable than pZAQ ΔftsK::kan cells (B. Geissler and W. Margolin, unpubl. results), suggesting that Z-ring dynamics may be severely altered when ZipA is present and both FtsA* and FtsZ are overexpressed (Geissler et al., 2003) regardless of the absence of FtsK. Anucleate minicells were often observed with the pZAQ ΔftsK::kan cells (Fig. 4F–G), indicating that the extra FtsA and FtsZ from pZAQ was capable of triggering extra polar divisions (Ward and Lutkenhaus, 1985; Begg et al., 1998) (Fig. 4D–E) even in the absence of FtsK.
Table 1. Summary of the suppression of ftsK or ftsQ null alleles by various plasmids.
. Three different classes of cellular morphology were observed: WT, similar to wild-type cells; CH, chains of cells, some short cells; CHX, chains and many dead cells; NT, no transductants were isolated; ND, no data. Numbers in parentheses indicate the percentage of cells with wild-type length (< 4 µm).
. Viability measurements are in relation to wild-type W3110 cells: +++, similar to wild-type cells; ++, decreased viability; +, culturable, but very sick; ND, no data.
Despite the role of the C terminus of FtsK in chromosome dimer resolution and DNA transport away from the closing septum, expressing only the N-terminal cell division domain of FtsK disrupts nucleoid separation in only ∼10% of cells (Liu et al., 1998). Depletion of FtsK has no detectable effect on chromosome segregation (Chen and Beckwith, 2001), although it could be argued that trace amounts of FtsK present after depletion might still be functional for DNA transport. Because we introduced a complete deletion of ftsK into strains contaning pBAD-FtsA* or pZAQ, the effects of trace levels of FtsK can be ruled out. 4′,6-diamidino-2-phenylindole (DAPI) staining of the ΔftsK::kan cells containing pZAQ (WM2267) indicated that the vast majority of chromosomes segregated normally (Fig. 4F–G), similar to ftsK+ cells (Fig. 4B–E and H–I). Therefore, in agreement with previous results with FtsK depletion, it appears that chromosomes can segregate mostly normally in the complete absence of FtsK.
Cells of the pZAQ/ΔftsK::kan strain were 3.2 times longer on average than cells of the pZAQ/ftsK+ strain (8.2 µm vs. 2.6 µm), resulting mostly from formation of cell chains (Table 2). The extent of chain formation was measured by counting the number of clearly defined division septa per cell; there were nearly five times as many septa per cell in the ΔftsK::kan strain compared with the ftsK+ strain (1.52 vs. 0.31). To examine the pattern of Z rings in the absence of ftsK, we performed immunofluorescence microscopy on fixed cells. Despite the chaining, anti-FtsZ immunofluorescence showed the average space between Z rings was only 1.7X longer without ftsK (5.88 µm vs. 3.31 µm), which is similar to previous data from FtsK depletion experiments (Chen and Beckwith, 2001). This suggests that although there was a significant alteration in the ability of WM2267 cells to properly separate, FtsZ was still able to assemble into Z rings at regular intervals.
Table 2. Quantification of anti-FtsZ immunofluorescence data from cells from logarithmically growing cultures.
m/ring. Calculated by dividing the total length of all cells by the number of rings.
. Calculated by dividing the total number of visible septa by the number of cells.
2.6 ± 0.8
WM2267 (pZAQ ΔftsK)
8.2 ± 6.7
WM2268 (pZAQ ΔzipA)
4.1 ± 2.4
2.6 ± 0.5
Potential causes for the cell separation defect in the absence of FtsK
The cell chain phenotype observed with WM2267 suggested that although the cells could form division septa and divide, there was a defect in daughter cell separation. E. coli normally divides to form individual daughter cells, a process mediated by cell wall amidases. One such amidase, AmiC, was recently found to localize to the division site in an FtsN-dependent manner. Overexpressing AmiC-GFP is capable of relieving the cell chaining phenotype caused by removing other amidases (Bernhardt and de Boer, 2003). In an attempt to drive cell separation within these ΔftsK::kan cell chains, we transformed a plasmid with an isopropyl-β-D-thiogalactoside (IPTG)-dependent AmiC-GFP fusion into WM2267. With or without induction of AmiC-GFP expression with 1 mM IPTG, there was no difference in cell morphology (data not shown), suggesting that additional AmiC was not sufficient to separate chains caused by deleting ftsK.
Another possible explanation for formation of cell chains is that although there was a division septum formed between each cell, complete septum closure was impeded by an improperly segregated nucleoid. Such nucleoid segregation defects are often caused by chromosome dimers generated by RecA-mediated recombination. Although the normal nucleoid distribution described above argued against this possibility, we tested the role of RecA in causing chromosome dimers and potentially cell chains in the ΔftsK::kan cells by chromosomally inactivating recA and observing the cells microscopically. There was no obvious difference in the morphology of pBAD-FtsA*/ΔftsK::kan (WM2197) cells vs. pBAD-FtsA*/ΔftsK::kan recA (WM2204) cells (data not shown). These data indicate that failure to resolve chromosome dimers is not responsible for inhibiting cell separation in strains lacking ftsK.
Localization of later cell division proteins in the absence of FtsK
The appearance of septa, minicells, and cells of normal length shows that the ΔftsK::kan strains containing pZAQ or FtsA* are capable of dividing. However, later septation proteins such as FtsI and FtsN normally require FtsK for their recruitment to the Z ring (Chen and Beckwith, 2001). Therefore, we wanted to determine if these components were still recruited to division septa in the absence of FtsK. We examined the localization of GFP fusions to FtsI or FtsN in cells lacking FtsK and containing either FtsA* (Fig. 5) or pZAQ (data not shown). As shown in ftsA*/ΔftsK::kan cells, each fusion produced fluorescent bands at potential cell division sites prior to septum formation (arrows) or at constrictions within cell chains. These data indicate that, despite the established linear dependency pathway for septal protein recruitment, FtsI and FtsN can localize to division sites independently of FtsK. In accord with our inability to delete ftsQ, ftsB, ftsL or ftsW in the presence of FtsA*, these results suggest that the downstream divisome components are still essential in the absence of FtsK.
Extra ftsQAZ can partially compensate for the loss of ZipA
We showed previously that introducing a ΔzipA::kan allele into an E. coli MG1655 derivative containing pZAQ resulted in colony formation, but the cells formed long aseptate filaments that grew very poorly in broth (Geissler et al., 2003). In the W3110 strain background, pZAQ was able to suppress the ftsK null allele as efficiently as does ftsA*, whereas in the MG1655 background, only ftsA* but not pZAQ could suppress the ftsK null. Therefore, we were interested in exploring whether pZAQ might suppress ΔzipA::kan more efficiently in the W3110 strain background, which seems to be more permissive for this type of suppression. This would help to address whether the mechanism behind suppression by FtsA* is similar to that of expression of extra FtsQ, FtsA and FtsZ. When the ΔzipA::kan allele was introduced into W3110 containing pZAQ, cells became longer than normal (Fig. 4H–I), but only about 1.6-fold compared with W3110 containing pZAQ (Table 2). This suggested that pZAQ could significantly compensate for the lack of ZipA in this strain background.
To determine whether this increase in cell length resulted from destabilization of Z rings in the absence of ZipA, we examined Z rings in these cells by immunofluorescence microscopy. Removing zipA from pZAQ-containing cells decreased the percentage of cells containing at least one Z ring from 63% to 40% (Table 2). Intriguingly, despite the increased levels of FtsZ because of the presence of pZAQ, the average spacing between Z rings was increased by approximately threefold in the absence of ZipA (Table 2). This is consistent with the role of ZipA in stabilizing Z rings (Hale and de Boer, 1997; Pichoff and Lutkenhaus, 2002; Raychaudhuri, 1999). Despite this decreased ability to form Z rings in cells with pZAQ and ΔzipA, these rings, once assembled, seemed to be able to support cell division. The number of septa per pZAQ/ΔzipA::kan cell was nearly identical to that of pZAQ/zipA+cells, suggesting that the increased cell length in the absence of zipA was not a result of aborted division attempts, but most likely because fewer Z rings were assembled. In pZAQ-containing W3110 cells, the cost of losing zipA was minimal (Fig. 4A).
To test if pZAQ or FtsA* could alleviate the requirement for two division proteins simultaneously, we attempted to remove both zipA and ftsK from W3110 cells carrying pZAQ or ftsA*. We first constructed a chloramphenicol-resistant version of an ftsK null allele, ΔftsK::cat. However, attempts to introduce ΔzipA::kan into W3110 containing pZAQ/ΔftsK::cat or ftsA*/ΔftsK::cat, or, conversely, ΔftsK::cat into W3110 containing pZAQ zipA::kan or ftsA*/zipA::kan by phage P1 transduction were unsuccessful. This suggests that the effects of ZipA and FtsK are additive, and that at least one of them is required for cell division even in the presence of FtsA* or extra FtsQ, FtsA and FtsZ from pZAQ. In support of this idea, we found that overproduction of ZipA from pWM1703 permitted the ΔftsK::kan strain to form colonies (data not shown).
Extra FtsQ, and to a lesser extent FtsB or FtsN, can compensate for the loss of FtsK
Previously, Draper et al. (1998) showed that supplying FtsN on a plasmid was able to partially suppress the requirement for ftsK in cell division. Although filaments and chains were reported, these cells were apparently still viable. We confirmed this result using two different ftsN expression plasmids. Upon induction with arabinose or IPTG respectively, pWM2022 expresses native FtsN from the PBAD promoter or pWM1152 expresses functional GFP-FtsN from the PTrc promoter. The ΔftsK::kan allele could be readily introduced into W3110 cells containing either plasmid in the presence of inducer. However, the cell population consisted mainly of long filaments and lysed cells (Fig. 6B, Table 1, and data not shown). A plating viability experiment indicated that in the absence of FtsK, extra FtsN from pWM2022 supports ∼1000X fewer viable cells than extra GFP-FtsK1-859 from pWM1747 (Fig. 6A).
Because extra FtsN or ZipA were able to allow viability of the ΔftsK::kan strain, we then tested whether overexpression of other divisome components could compensate for the loss of FtsK. We were not able to introduce ΔftsK::kan by transduction into W3110 carrying plasmids that synthesized GFP alone or GFP fusions to FtsQ, FtsL, FtsB, FtsW or AmiC. All plasmids were able to complement deletions of their cognate chromosomal alleles, indicating that those GFP fusions were functional. Unexpectedly, however, we found that ΔftsK::kan could be introduced efficiently into W3110 containing plasmids expressing FtsQ (pJC10) or FtsB (pWM2058) without GFP tags. When grown in broth, both strains displayed cell chains and misshapen cells in addition to many normal sized cells (Fig. 6C–D and Table 1). The ΔftsK::kan strain containing pWM2058 (FtsB) also exhibited a large percentage of dead cells when grown overnight, similar to the ΔftsK::kan strain containing pWM2022 (FtsN) (Fig. 6B and D, arrows). We validated this result using two additional ftsQ-expressing plasmids, pLMG161 and pLD137; W3110 derivatives containing each of these plasmids were also able to be transduced with ΔftsK::kan (data not shown). W3110 containing ΔftsK::kan and pJC10 (FtsQ) yielded ∼100× more viable cells than W3110/ΔftsK::kan containing pWM2022 (FtsN), but still ∼10 × fewer viable cells than W3110/ΔftsK::kan containing pWM1747 expressing GFP-FtsK1-859 (WM2294) (Fig. 6A). This partial suppression by FtsQ is consistent with the presence of cell chains and filaments observed in broth (Fig. 6C).
To follow up on the ability of extra FtsQ to compensate for the loss of FtsK, we attempted to introduce ΔftsK::kan by transduction into W3110 containing pZAQ′, which expresses ftsA and ftsZ but is expected to express only the first 24 amino acids of ftsQ containing the cytoplasmic tail. We consistently isolated 30% or fewer transductants with pZAQ′ as compared with W3110/pZAQ. An increase in the number of cells in chains as well as filament lengths (data not shown) also indicated that pZAQ′ was less efficient at suppressing ΔftsK::kan than was pZAQ. We attempted to introduce ΔftsK::kan into W3110 containing either pZA′Q or pMK4, which expresses ftsZ alone, but were unable to isolate any transductants, suggesting that extra FtsA is important for compensating for the loss of FtsK. In addition to suggesting that FtsQ plays a role in ΔftsK::kan suppression, these results suggest that extra FtsA and FtsZ together can partially compensate for the loss of FtsK. However, as the cytoplasmic tail of FtsQ may be synthesized from pZAQ′, we cannot rule out the possibility that the cytoplasmic domain of FtsQ also contributes to the suppression.
Domains necessary for FtsQ-mediated compensation for the loss of FtsK
Because FtsA* is localized at the cytoplasmic side of the membrane and can partially compensate for the loss of FtsK, we reasoned that the short N-terminal domain of FtsQ, the only part of FtsQ that resides in the cytoplasm, might be the domain responsible for partial suppression of the ΔftsK::kan allele. To test this hypothesis, we introduced ΔftsK::kan into W3110 containing various FtsQ domain swap plasmids (Guzman et al., 1997) and measured the ability of these strains to form colonies. We isolated and confirmed by PCR many ΔftsK::kan transductants of W3110 containing plasmid pLD92, which expresses the cytoplasmic and transmembrane domains of FtsQ (FtsQCT) fused to the periplasmic domain of FtsL (pBAD18-QQL). When grown in broth, these cells displayed a mixture of cell chains and normal-length cells (Fig. 6E). This phenotype is similar to ΔftsK::kan derivatives containing plasmids pJC10 (pBAD-FtsQ; Fig. 6C) or pLD137 (pBAD18-QQQ) (data not shown and Table 1). Extra FtsL from pBAD18-LLL (pWM2406), normally sufficient to complement an ftsL null mutant, did not permit introduction of the ΔftsK::kan allele (data not shown). This indicates that extra FtsL cannot compensate for the loss of FtsK, and is consistent with the idea that the suppression of ΔftsK::kan by pLD92 (pBAD18-QQL) is mediated by the FtsQCT portion and not by the periplasmic domain of FtsL expressed from pLD92.
In further support of our hypothesis, pLD104, which expresses the FtsQ periplasmic domain fused to the cytoplasmic and transmembrane domains of MalF (pBAD18-FFQ), did not permit introduction of ΔftsK::kan by transduction. As mentioned in the previous section, ΔftsK::kan also could not be introduced into a strain expressing GFP-FtsQ, under the same conditions which the fusion protein could complement ΔftsQ. This suggests that the GFP tag on FtsQ might be able to inhibit the suppression of the ftsK null allele. This may be because the GFP tag inhibited full function of the cytoplasmic domain of FtsQ, whereas the transmembrane and periplasmic domains remained unaffected by the tag. As mentioned above, we were not able to introduce ΔftsK::kan into a strain carrying GFP-FtsB under conditions that normally complement a null allele of ftsB, further suggesting that the GFP tags may interfere with the ability of FtsQ and FtsB to compensate for the loss of FtsK.
The N-terminal domain sequence of FtsK is conserved only in proteobacteria
The ability to compensate for the loss of FtsK prompted us to ask whether the minimal N-terminal domain of FtsK that is essential for cell division is well conserved in bacteria. We performed a blast search on all completed prokaryotic genomes using either the N-terminal 210 residues or the C-terminal 1110 residues of FtsK as query sequences. Although the C terminus of FtsK containing the DNA translocation domain is conserved throughout most bacterial species, we found that the N-terminal cell division domain is only conserved among the proteobacteria. Within the currently sequenced prokaryotic genomes, there are significantly homologous orthologs (> 0.001 e-value) among four of the five proteobacterial subclasses, with the most highly orthologous proteins present within the enterobacteria. Notably, the SpoIIIE protein from Bacillus subtilis, a functional homologue of FtsK that localizes to polar division septa in B. subtilis, does not appear to contain an FtsK-like N-terminal domain.
In this study, we have shown that E. coli strains deleted for the essential cell division gene ftsK can remain viable when genes for several other cell division proteins are expressed from multicopy plasmids (ftsAZ, ftsQ, zipA, ftsB or ftsN) or altered (ftsA R286W, otherwise known as ftsA*). Efficiently suppressed strains lacking FtsK exhibited many cells of normal length, but also contained cell chains, indicating that no suppressing condition was able to restore a wild-type septation phenotype to the entire cell population. This is in contrast to the nearly wild-type septation phenotype conferred on a ΔzipA::kan mutant by ftsA*. However, the ability to form colonies in the absence of FtsK was likely a result of the proportion of cells that were able to divide and separate. The ability to partially bypass the essential cell division function of FtsK has important implications for its function, the proteins able to suppress its requirement, and the assembly of the divisome itself.
Because FtsK is in the middle of the linear recruitment pathway of cell division proteins (Fig. 1), the question arose as to whether the pathway remained intact after removal of FtsK in the suppressed strains. Using GFP fusions, we showed that FtsI and FtsN, which lie downstream of FtsK in the linear dependency pathway and thus normally depend upon FtsK for their recruitment to the Z ring, localized proficiently to division sites in the absence of FtsK. We therefore postulate that the recruitment of these downstream proteins probably does not depend directly upon FtsK in wild-type cells. The paradigm for this is the deletion of zipA in ftsA* cells. ZipA is normally required for recruitment of downstream proteins (Hale and de Boer, 2002; Pichoff and Lutkenhaus, 2002), but when zipA is removed in the presence of ftsA*, the downstream proteins still are recruited (Geissler et al., 2003). This can be explained if ZipA does not recruit downstream proteins directly in wild-type cells, but instead acts indirectly to enable another factor, such as the Z ring, to recruit them.
By analogy, we propose that in wild-type cells, FtsK helps indirectly to stabilize the Z ring and its other components, which in turn allows for the recruitment of FtsQ and downstream proteins. This potential general stabilization function is consistent with the ability of other proteins to compensate for the loss of FtsK, as these proteins probably contribute to the stabilization of the assembling divisome via mechanisms similar to those achieved by FtsK. FtsK may share a stabilization function with ZipA, for example, because there is increased spacing between Z rings in the absence of FtsK or ZipA, and extra ZipA can partially compensate for the absence of FtsK. Another potential explanation for the results, however, is that the proteins normally found downstream of FtsK are recruited to the divisome in a different order than in ftsK+ strains. Using separate assays, several authors have reported that FtsA is capable of interacting with FtsI (Di Lallo et al., 2003; Corbin et al., 2004; Karimova et al., 2005) and FtsN (Corbin et al., 2004; Karimova et al., 2005). The excess FtsA in ΔftsK::kan cells carrying pZAQ might be capable of directly recruiting FtsI and FtsN in the absence of FtsK. Moreover, FtsQ has been shown to interact with FtsI, which could back-recruit FtsQ to the division ring after FtsI recruitment by FtsA (Goehring et al., 2005). These possibilities and others are currently being investigated.
A function for FtsK in Z-ring stabilization may also explain why cell division is largely restored by FtsA* in ftsK44 and ftsK null mutants. The precise mechanism by which FtsA* acts is not yet known, but because it can suppress the loss of ZipA, which has been shown to stabilize the Z ring (Raychaudhuri, 1999; Hale et al., 2000), it is likely that FtsA* enhances FtsA′s normal Z-ring stabilization function (Pichoff and Lutkenhaus, 2002). By suppressing the loss or thermosensitivity of FtsK, it is reasonable to hypothesize that FtsK has an analogous overlapping stabilization function that can be compensated by FtsA* or excess ZipA. Additional evidence for a role of FtsK in Z-ring stabilization is that pZAQ itself can mimic the suppressing effects of FtsA*, both for zipA and ftsK mutants. In this case, simply increasing the levels of FtsQ, FtsA and FtsZ at their normal stoichiometry stabilizes Z rings, possibly by increasing the number or bundling of FtsZ protofilaments at the division site, or by antagonizing disassembly factors such as MinC. It is less clear why the ftsQ1 thermosensitive mutant was suppressed by FtsA*, but not the ftsQ null mutant. One possibility is that many of the thermosensitive cell division mutants such as ftsQ1 are defective in maintaining interactions with other septal proteins, and that such defects can be overcome by putative stabilizing factors such as FtsA*. However, FtsQ probably has another essential function in septation, such as the recruitment of FtsL and FtsB, which cannot be compensated for by stabilizing the Z ring.
Because increased expression of ftsQ, ftsB or ftsN can suppress the requirement for ftsK in cell division to varying degrees, it is reasonable to propose that some common structural element among these three proteins may be responsible for their functional overlap. One potential similarity is the presence of a short cytoplasmic domain linked to a transmembrane domain. In the case of FtsQ, we demonstrated that its cytoplasmic and transmembrane domains were sufficient to permit viability in the absence of ftsK. This was unexpected, as domain-swapping experiments have shown that the periplasmic domain of FtsQ is sufficient to complement an ftsQ null allele (Chen et al., 2002). As a result, the cytoplasmic and transmembrane domains of FtsQ had, up to now, no potential role other than to serve as a membrane anchor for the periplasmic domain. However, our findings suggest that these two domains, while dispensable for FtsQ-specific function, have roles, possibly in Z-ring stabilization, that overlap those of the N-terminal cell division domain of FtsK. These overlapping functions are consistent with FtsK–FtsQ interactions implicated in previous studies (Di Lallo et al., 2003; Goehring et al., 2005; Karimova et al., 2005). By analogy, the data for FtsQ suggest that the cytoplasmic and transmembrane domains of FtsB and FtsN may also have functions that overlap with FtsK and FtsQ. FtsL, however, was not able to permit growth of ΔftsK::kan colonies although it was sufficient to complement an ftsL null mutant. Although there are several potential explanations for this, we favour the idea that the cytoplasmic tail and/or transmembrane domains of FtsL are not involved in Z ring stabilization and hence have functions distinct from the analogous domains of FtsQ and FtsN. This idea is supported by the fact that the cytoplasmic tail of FtsL, unlike the cytoplasmic domains of FtsQ or FtsN, is required for function (Ghigo and Beckwith, 2000).
It is not yet clear whether both cytoplasmic and transmembrane segments of FtsQ are needed to compensate for the loss of FtsK, or whether only one of the segments is involved. One clue that the cytoplasmic domain may be important is that expression of a GFP-FtsQ fusion was not able to permit viability of the ftsK null strain although it could complement an ftsQ mutant. Although there may be other potential explanations, one idea is that the cytoplasmic domain of FtsQ may help to stabilize the Z ring, possibly by directly contacting FtsZ or FtsA (Karimova et al., 2005), and that the N-terminal GFP tag may sterically interfere with this function. A similar model may explain how FtsB, but not GFP-FtsB, could partially compensate for the loss of FtsK.
Because of their small size and presence together at the Z ring, we speculate that the cytoplasmic and/or transmembrane domains of FtsK, FtsQ, FtsB and FtsN, and potentially FtsL and FtsW, stabilize the interactions among divisome components. Their potential importance in stabilization is evidenced by the presence of these domains in the majority of the 12 known division proteins. Although pZAQ or ftsA* were able to compensate for the loss of ftsK and zipA, they were not able to suppress an ftsI(ts) mutant or null alleles of ftsB, ftsL, or ftsW. This can be explained if these later proteins have unique functions in septal wall synthesis in addition to their possible overlapping roles in stabilization of the protein complex at the Z ring.
We therefore propose that there are two main roles for cell division proteins: (i) to stabilize the interactions between divisome components and (ii) to generate the division septum. Furthermore, we propose that the stabilization of the divisome is a result of multiple protein components sharing a short cytoplasmic domain inserted into the periplasmic membrane, which assemble in a cooperative manner. For example, the N-terminal domain of FtsK may function to increase the local concentration of FtsQ and FtsB, thus explaining why increasing FtsQ and FtsB expression in the absence of FtsK permits cell division to occur. This model also fits nicely with a recent study of the timing of septal protein recruitment, which suggests that the divisome is formed in two distinct steps (Aarsman et al., 2005). Finally, the model is consistent with the lack of conservation of many of these proteins throughout bacteria (Margolin, 2000). In species lacking one or more of the stabilizing components, other proteins with similar domains would be expected to compensate.
What is the function of FtsK in cell division? Clearly it is not essential for viability when there are compensatory changes. However, many of these cells lacking FtsK form chains, indicating that the Z ring contracts and the septum forms, but a late step in cell separation is partially defective. Our experiments with a recA derivative ruled out the possibility that this separation defect was caused by chromosome dimers interfering with septation, making it more likely that a septation-specific activity of FtsK is needed for optimum cell separation. Because the DNA transfer domain of SpoIIIE of B. subtilis is homologous to that of FtsK, and because SpoIIIE is required for membrane fusion during spore engulfment (Errington et al., 2001; Sharp and Pogliano, 2003), it is attractive to propose that the N-terminal domain of FtsK may facilitate membrane fusion during cytokinesis. Like SpoIIIE, this domain of FtsK is predicted to span the membrane four times (Dorazi and Dewar, 2000), but our blast analysis indicated that these two proteins do not share significant sequence similarity in their N-terminal domains. In the absence of FtsK, membranes probably are able to fuse in many cells by some other mechanism.
Bacterial strains, plasmids and media
Escherichia coli strains and plasmids used in this study are listed in Table 3. Bacteria were grown in LB media (0.5% NaCl) containing antibiotics at 30°C, unless otherwise indicated. Antibiotics were added at the following concentrations: chloramphenicol (cat) at 20 µg ml−1, kanamycin (kan) at 25 µg ml−1, ampicillin (Amp) at 100 µg ml−1, or tetracycline (Tet) at 10 µg ml−1 as needed. Thymine was added to the medium whenever the thyA WM640 (ftsK44) and WM2149 (ftsQ1) strains were used. l-arabinose (ara) was added to a final concentration of 0.2% (or indicated concentrations) to induce expression from PBAD promoters (Guzman et al., 1995). To induce expression of GFP-constructs, IPTG was added to a final concentration of 100 µM or as indicated. Anti-FtsZ immunoflourescence, microscopy, cell fixation, and staining with DAPI were done essentially as described previously (Geissler et al., 2003).
Oligonucleotide primers used in plasmid construction are listed in Table 4. These primers were used to amplify fragments from TX3772 genomic DNA using Hi-Fidelity Polymerase (Roche Scientific). To construct pWM1747, the first 859 residues of FtsK were fused to GFP in pBAD30 using primers ftsK1 and ftsK3, digested with SacI and XbaI, and ligated into same cut pAG (pBAD30-GFP-FtsA). To construct pWM1845, primers ftsL1 and ftsL2 were used to amplify ftsL. This product was digested with EcoRI and HindIII and ligated into pDSW207 cleaved with the same enzymes. Plasmid pWM2058 contains ftsB amplified with ftsB1 and ftsB2, cut with SacI and XbaI, and ligated into pDSW209 cleaved with the same enzymes. Plasmid pDSW208 was used to construct AmiC-GFP. The amiC gene was amplified with amiC1 and amiC2 primers, digested with SacI and HindIII, and ligated into pDSW208 cleaved with the same enzymes. The ftsN gene was amplified with primers ftsN1 and ftsN2, cleaved with SacI and HindIII, and ligated into pBAD30 cleaved with the same enzymes to create pWM2022. Plasmids pWM2248 (pZA′Q) and pWM2255 (pZAQ′) were constructed as described previously (Begg et al., 1998). Briefly, pZAQ was digested with either BglII or MluI, the overhangs were filled in with the Klenow fragment of DNA polymerase and ligated to yield frameshift mutations in ftsA or ftsQ respectively. These plasmids were then confirmed by their ability to complement thermosensitive alleles of ftsQ, ftsA or ftsZ.
Table 4. Oligonucleotide primers used in plasmid and deletion constructions.
A non-GFP-fusion derivative of pDSW210 was made (pWM2060) by replacing the ScaI-EcoRI fragment on pDSW210 with the ScaI-EcoRI fragment from pDSW209; this removes the GFP gene, while the bla gene and the multiple cloning site on pDSW210 remain intact. The ftsB gene was amplified with primers ftsB3 and ftsB2 and digested with SmaI and XbaI, then ligated into pWM2060 cleaved with SmaI and XbaI to create pWM2356.
Plasmid pBAD18-LLL was constructed by replacing the cytoplasmic and transmembrane domains of FtsQ in pBAD18-QQL (pLD92) with the cytoplasmic and transmembrane domains of FtsL that correspond to the first 57 amino acids of FtsL. FtsL1-57 was amplified using primers FtsL57-N and FtsL57-C; this product and pLD92 were digested with EcoRI and MscI and ligated to generate pWM2406. Plasmid pWM2356, expressing ftsB, and pBAD18-LLL (pWM2406) were shown to be functional by the ability to introduce either ΔftsB:kan or ΔftsL:kan alleles respectively, into W3110 derivatives containing them.
Complementation of thermosensitive mutants
Thermosensitive mutant strains were transformed with the indicated plasmids and grown at the permissive temperature, 30°C. These strains were grown in LB + antibiotics overnight, and diluted 1:200 into fresh medium and grown to an OD600 of ∼0.15 at 30°C. Cultures were then split; one aliquot was diluted into fresh medium pre-warmed to the non-permissive temperature of 42°C, while the other was diluted into medium at 30°C. These cultures continued growing for ∼3 h, and then cells were immobilized in 1% LB agarose and observed microscopically as previously described (Geissler et al., 2003).
Construction and confirmation of ΔftsK:kan
To construct a depletion strain of FtsK, we used recombineering (Yu et al., 2000) to directly replace portions of the E. coli chromosome with an antibiotic resistance cassette. We constructed oligonucleotides that contained 40 bp of the region directly surrounding ftsK fused to portions of pUC4K upstream of the kanamycin resistance gene (aph) or of the end of aph containing the stop codon (primers Kkan1 and Kkan2). These oligonucleotides were used to amplify portions of the lrp and lolA genes flanking ftsK, such that the final PCR product contained lrp-aph-lolA. This product was then digested with DpnI for 6 h to eliminate the pUC4K plasmid template from the reaction mixture.
After digestion, this fragment was used to transform recombination-competent DY329 (Yu et al., 2000) containing pBAD30-GFP-FtsK1-859 (pWM1747). By supplying the N-terminal portion of ftsK, which is sufficient to complement a null mutant of ftsK for cell division and viability, from an inducible/repressible promoter, we were able to deplete FtsK by switching from broth containing arabinose to glucose. Upon plating on LB + Amp + Kan + 0.1% arabinose at 30°C, we isolated 50 colonies, five of which were inoculated separately into LB broth + Kan + 0.1% arabinose. Overnight cultures were diluted 1:1000 into fresh medium containing arabinose and grown to early logarithmic phase. These cells were pelleted, washed once in LB only, and used to inoculate fresh LB broth containing either 0.1% ara or 1% glucose. After 4 h further incubation, the cells were examined microscopically. All five DY329 + pWM1747 ΔftsK::kan cultures grown in arabinose appeared similar to the control DY329 + pWM1747 cells grown under the same conditions. However, all five DY329 + pWM1747 ΔftsK::kan cultures grown in glucose were highly filamentous, indicating that FtsK had been successfully depleted. One of the five was saved as WM2109.
The absence of ftsK in these strains was confirmed by PCR, using primers within ftsK (K-105 and K-585), primers upstream of ftsK (K-C-up) and inside aph (Aph4 or Aph721), or primers inside aph (Aph3) and downstream of ftsK (K-C-down). After PCR confirmation, we grew P1 phage on the strain and used the resulting lysate to transduce various other strains to KanR. After each subsequent transduction, we used ftsK-specific PCR primers to confirm the presence of the ΔftsK::kan allele.
To make a chloramphenicol-resistant version of ΔftsK, we replaced the aph cassette in WM2109 with the cat cassette from pBC(SK+) using the protocol described above. Primers Kcat1 and Kcat2 were used to generate lrp-cat-lolA, which was electroporated into recombination-competent WM2109 and plated on LB + Amp + cat + 0.1% arabinose. Colonies were then tested for kanamycin sensitivity, arabinose dependence, and confirmed by PCR as described above; the final strain was saved as WM2264.
Constructing FtsQ, FtsB, FtsL, and FtsW conditional null strains
We constructed FtsQ and FtsW depletion strains as described previously (Chen et al., 2002; Mercer and Weiss, 2002). To construct genomic deletions of ftsB and ftsL, we followed essentially the same protocol outlined above for ftsK, except we used DY329 containing plasmids expressing each protein under constitutive promoters (pWM2058 and pWM1845 respectively). To confirm each deletion, PCR primers were used to amplify the region surrounding each aph insertion. P1 phage was grown on each strain to transduce the deletions into other recipient strains by selection for kanR. As for ΔftsK::kan, we tested each kanR transductant by PCR using X-C-up and X-C-down primers to confirm that each was deleted for ftsQ, ftsB, ftsL or ftsW.
Complementation of null mutants by transduction
Plasmids were tested for their ability to complement null mutants by transduction as previously described (Chen et al., 2002). P1 phage was grown on the donor strains carrying an fts gene replaced by a kan cassette and used to transduce various strains containing plasmids to kanR. Arabinose (0.2%) was used to induce expression of the fts genes in pBAD plasmid constructs, and 0.1 mM IPTG was used to express zipA from pWM1703. Fresh transductants were restreaked and screened for allele replacement by PCR, which demonstrated linkage of the antibiotic resistance cassette to the region surrounding the deleted gene of interest.
The viability of each strain was determined by plating 5 µl drops from 10-fold serial dilutions of overnight cultures onto LB plates; the plates were then incubated at 30°C. The relative plating efficiency was determined for each mutant strain based on comparison with its parent strain on the same plate.
We thank Christophe Bernard, Arielle Kagan, Swapna Thanedar, and Brian Corbin for help with plasmid constructions and Melanie Hargrove for helpful experiments not included herein. We are grateful for Nathan Goehring for helpful discussions and Jonathan Beckwith for providing ftsQ plasmids used in this work. This work was supported by National Institutes of Health Grant R01-GM61074.