FtsA mutants impaired for self-interaction bypass ZipA suggesting a model in which FtsA's self-interaction competes with its ability to recruit downstream division proteins

Summary

Z- ring assembly requires polymers of the tubulin homolog FtsZ to be tethered to the membrane. Although either ZipA or FtsA is sufficient to do this, both of these are required for recruitment of downstream proteins to form a functional cytokinetic ring. Gain of function mutations in ftsA, such as ftsA* (ftsA-R286W), bypass the requirement for ZipA suggesting that this atypical, well conserved, actin homolog, has a more critical role in Z-ring function. FtsA forms multimers both in vitro and in vivo, but little is known about the role of FtsA polymerization. In this study we identify FtsA mutants impaired for self-interaction. Such mutants are able to support Z-ring assembly and are also able to bypass the requirement for ZipA. These mutants, including FtsA*, have reduced ability to self-interact but interact normally with FtsZ and are less toxic if overexpressed. These results do not support a model in which FtsA monomers antagonize FtsZ polymers. Instead, we propose a new model in which FtsA self-interaction competes with its ability to recruit downstream proteins. In this model FtsA self-interaction at the Z ring is antagonized by ZipA, allowing unpolymerized FtsA to recruit downstream proteins such as FtsN.

Introduction

Cytokinesis in E. coli requires 10 essential proteins and the involvement of many more non-essential proteins, which assemble into a functional cytokinetic ring in two temporally distinct steps (de Boer, 2010). In the first step, a Z ring is assembled that consists of polymers of FtsZ tethered to the membrane by two proteins, FtsA and ZipA. Although either of these proteins is sufficient to support Z ring formation, downstream division proteins are not recruited unless both are present (Pichoff and Lutkenhaus, 2002; Hale and de Boer, 2002). Of the two, FtsA appears more important since mutations in ftsA such as ftsA* (ftsA-R286W) can bypass the requirement for ZipA (Geissler et al., 2003). Also, ZipA is less conserved than FtsA in evolution, and FtsA is thought to directly contact one or more of the downstream proteins (Di Lallo et al., 2003; Karimova et al., 2005; Corbin et al., 2004). ZapA and ZapC interact directly with FtsZ and are also present at this stage but are not essential (Hale et al., 2011; Durand-Heredia et al., 2011; Gueiros-Filho and Losick, 2002). Another non-essential protein, ZapB, is recruited to the Z ring through ZapA (Galli and Gerdes, 2010). The roles of ZapA, ZapB and ZapC are not yet clear, but their presence seems to increase the integrity of the Z ring (Dajkovic et al., 2010; Monahan et al., 2009; Hale et al., 2011; Durand-Heredia et al., 2011). In a second step occurring later in the cell division cycle, the remaining cell division proteins are recruited to the Z ring to constitute a fully functional cytokinetic ring that divides the cell (Aarsman et al., 2005).

FtsA is a member of the actin family although it is structurally different than typical actins (van den Ent and Lowe, 2000). One feature of FtsA that distinguishes it from actin is the presence of an extended C-terminal tail that includes an amphipathic helix (MTS) at the extreme C-terminus responsible for the binding of FtsA to the membrane (Lowe and van den Ent, 2001; Pichoff and Lutkenhaus, 2005). When this tail is removed the truncated FtsA binds poorly to Z rings and instead assembles into rod-like structures in the cytoplasm. The failure of the truncated FtsA to be recruited to the Z ring is its inability to compete with full length FtsA. In the absence of full length FtsA the truncated FtsA localizes efficiently to a Z ring (formed with the aid of ZipA) (Pichoff and Lutkenhaus, 2005; Shiomi and Margolin, 2008).

The major structural difference between FtsA and actin is the domain structure. Typical actins have 4 domains IA, IB, IIA and IIB and these are all involved in actin polymerization. In contrast, FtsA is missing domain IB but has an additional domain IC inserted elsewhere in the structure (van den Ent and Lowe, 2000). Despite this altered structure, FtsA has been reported to polymerize in vitro and a C-terminally truncated version of FtsA forms cytoplasmic rods in vivo suggesting it polymerizes (Lara et al., 2005; Feucht et al., 2001; Yim et al., 2000; Pichoff and Lutkenhaus, 2005). Questions remain, however, about the significance of FtsA's polymerization since the in vitro assembly is not associated with ATPase activity (the observed polymers are not dynamic) (Lara et al., 2005), and the in vivo assembly is only observed with FtsA truncated for its membrane targeting sequence (MTS) or with mutations in this MTS (Pichoff and Lutkenhaus, 2005). Evidence indicates domain IC interacts with FtsN and FtsI, late recruits to the cytokinetic ring (Corbin et al., 2004; Bernard et al., 2007; Rico et al., 2010). A mutant lacking domain 1C is unable to self-interact but is recruited to the Z ring however it is nonfunctional, as it is unable to recruit the downstream proteins and (Rico et al., 2004; Corbin et al., 2004).

Mutations in ftsA can bypass the need for ZipA and other cell division proteins such as FtsK and FtsN (Bernard et al., 2007; Geissler et al., 2007; Goehring et al., 2007; Geissler and Margolin, 2005; Geissler et al., 2003). The original report found a single mutation, ftsA-R286W designated ftsA*, that arose at low frequency and bypassed the requirement for ZipA (Geissler et al., 2003). It is located in the domain IIB. Additional phenotypes of this allele include a 20% shorter average cell length and increased resistance to conditions that are disruptive to Z ring assembly including increased MinC levels or excess ZipA (Geissler et al., 2007). Another ftsA mutation (ftsA-I143L) affects domain IC and was isolated as a suppressor of FtsQ deficiency (Goehring et alM., 2007). This mutation is also able to bypass ZipA but not FtsQ. An additional allele of ftsA was isolated as a suppressor of FtsN deficiency (Bernard et al., 2007). This allele contained three mutations but subsequently it was shown that a single mutation E124A is responsible for most of the activity. However, this allele can only bypass FtsN if it is overproduced (Gerding et al., 2009). It also appears that this allele can bypass ZipA. Thus, a mutation affecting a residue in domain IIB or mutations affecting residues in domain IC can bypass ZipA. The mechanism(s) of this bypass is not clear but it appears that these ftsA alleles increase the integrity of the Z ring by promoting stabilizing interactions (Geissler et al., 2007).

In addition to the observed polymerization of FtsA, self-interaction has been observed with bacterial and yeast 2-hybrid systems (Pichoff and Lutkenhaus, 2007; Shiomi and Margolin, 2007; Pichoff and Lutkenhaus, 2005; Rico et al., 2004; Di Lallo et al., 2003; Yim et al., 2000). A model for the dimerization of FtsA was proposed based in part on the analysis of covariation of residues during evolution of FtsA orthologs (Carettoni et al., 2003). In this dimer model domain IC of one FtsA molecule contacts domain IA of the other FtsA molecule. In this model domain IC occupies the position of the IB domain of actin. In such a model the interaction could be propagated but each FtsA subunit would be rotated 180° with respect to adjacent subunits. To explore the model Shiomi and Margolin (Shiomi and Margolin, 2007) introduced mutations in domain IA, and found that a mutation, M71A, resulted in an FtsA that was nonfunctional but was more toxic than wild type FtsA when overexpressed and had reduced self-interaction in a bacterial two-hybrid test. This is in contrast to the hypermorphic FtsA-R286W (FtsA*) that is functional but less toxic when overexpressed and has increased self-interaction in the same bacterial two-hybrid test (Shiomi and Margolin, 2007; Geissler et al., 2007; Geissler et al., 2003). These results led to a model in which FtsA monomers inhibit FtsZ assembly. In addition, it was subsequently shown in vitro that FtsA-R286W reduces FtsZ polymer mass in the presence of ATP suggesting it reduces the length of FtsZ polymers (Beuria et al., 2009). This effect was observed at a physiological ratio but in these in vitro experiments the multimer status of FtsA-R286W is not clear (and does not seem to change with addition of ATP). Interestingly, domain IA of FtsA has modest structural homology to the N-terminal domain of MinC, which antagonizes FtsZ assembly, and it was suggested that FtsA and MinC have similar mechanisms (Cordell et al., 2001; Beuria et al., 2009).

In this study, we used the ability of a C-terminally truncated FtsA to assemble into rods as a visible screen to isolate mutants impaired for self-interaction. Our subsequent analysis of these mutants and others revealed that all mutants displaying decreased self-interaction and retaining function allowed the bypass of ZipA suggesting that they actually support Z ring assembly rather than antagonize assembly.

Results

FtsA mutants impaired for rod formation bypass ZipA

Previously, we observed that a GFP-FtsAΔMTS fusion assembled into rods in the cytoplasm when expressed in wild type cells (Pichoff and Lutkenhaus, 2005). Since our previous results showed that FtsA mutants that self-interact in the Y2H form rods when expressed as GFP-FtsAΔMTS fusions while the ones that do not self-interact in Y2H do not form rods (Pichoff and Lutkenhaus, 2005; Pichoff and Lutkenhaus, 2007), we assume that this ability to form rods reflects the ability of FtsA to self-interact. Since little is known about how FtsA self-interacts we decided to use this as a screen to see if mutants that were affected in self-interaction could be obtained. To do this the coding sequence of ftsAΔMTS was mutagenized by PCR random mutagenesis and cloned upstream of GFP in an inducible vector to give pSEB385M. This plasmid library was then introduced into XL1-Blue cells and plated in the presence of 0.0001% arabinose to induce expression of the FtsAΔMTS-GFP fusion. Rod formation with pSEB385 is not as robust as observed previously with pSEB294 (GFP-FtsAΔMTS) (Pichoff and Lutkenhaus, 2005). This is likely due to the positioning of the GFP at the C-terminus of FtsA instead of at the N-terminus. Nonetheless, it was sufficient to allow us to screen for mutations abolishing rod formation. This lower efficiency of making rods may even have been helpful in the sense that a mutation that would have had only a moderate effect on GFP-FtsA, might cause a diffuse localization of FtsA-GFP. In addition, by using FtsA-GFP any mutations introducing a stop codon in the ftsA gene should be eliminated from the screen since GFP would not be expressed. Following the visual screen, 5 mutants (out of a 1000 screened) were selected that failed to form the characteristic rods observed with the wild type fusion (Fig. 1). In addition to displaying a diffuse cytoplasmic fluorescence, two of these formed a few aggregates (spots) instead of rods, two formed a markedly reduced number of rods and one failed completely to form rods or aggregates.

Fig. 1.

Screening for FtsAΔMTS mutants that fail to assemble into cytoplasmic rods. ftsAΔMTS-gfp fusions under the control of the P_BAD promoter (pSEB385 derivatives) were introduced into XL1-Blue. Colonies were streaked on 0.0001% arabinose plates and examined by fluorescence microscopy after 6 h at 37°C. Each ftsA mutant is indentified on the top left corner of each panel. One consequence of using the arabinose system to regulate expression is that there is some variation in expression between cells in the population (Siegele and Hu, 1997). Arrow heads indicate rods in WT panel, wide arrows indicate GFP localization at the Z-ring in R286W panel and the slim arrows indicate examples of the rare aggregates in D210G and G236D panels.

Sequence analysis of these mutants revealed that the two that formed a few aggregates had the mutations D210G and G236D. In previous work (Pichoff and Lutkenhaus, 2007) we observed that mutations expected to affect ATP binding, including D210A, produced aggregates (distinct from rods) in the cytoplasm and eliminated the ability of FtsA to interact in the yeast two-hybrid system with both itself and FtsZ. Since D210G forms aggregates, it is consistent with D210 being required for ATP binding and that rod formation requires ATP binding. Also, in a previous screen for mutants that affected the interaction between FtsA and FtsZ we isolated the G236C mutation. The G236C mutation caused a loss of interaction with both FtsZ and FtsA in the yeast two-hybrid test and when incorporated into GFP-FtsAΔMTS produced aggregates (Pichoff and Lutkenhaus, 2007). The G236 residue is buried in the structure and we surmised that altering this residue causes localized denaturation of this region of FtsA and aggregation. The G236D allele that we isolated in this screen would likely be similar. In addition, ftsA-D210G and ftsA-G236C were unable to complement an FtsA depletion strain (data not shown), indicating they encoded non-functional FtsA proteins.

The mutant that displayed no rods and only cytoplasmic fluorescence was T249M and the mutants that displayed markedly reduced ability to form rods were R177C and R286W [R286W also shows some septal localization as observed before (Shiomi and Margolin, 2008)]. Obtaining R286W in our screen was very surprising since this is the FtsA* mutant which was reported to display stronger self-interaction and had gained the ability to bypass the requirement of ZipA for cell division (Shiomi and Margolin, 2007; Geissler et al., 2003). To check this seeming discrepancy we analyzed R286W and R177C for self-interaction using the yeast two-hybrid system. Although both R286W and R177C displayed self-interaction in this system it was less than observed with wild type FtsA (Table 1). This surprising result led us to determine if R177C and T249M, like R286W, could bypass the requirement for ZipA. In an initial test, the mutations were introduced into pSEB306, which expresses ftsA under the control of IPTG. The resulting plasmids were then transformed into PS234 (a double mutant strain carrying thermosensitive alleles for both ftsA and zipA) and the cells were plated at 42°C at increasing concentrations of IPTG to check the ability of the different ftsA alleles to suppress the zipA1 (Ts) mutation (Fig. 2). Wild type ftsA was used as a negative control since it does not bypass ZipA nor allow the double mutant to grow at 42°C. The three alleles tested (R286W [positive control], R177C and T249M) allowed the growth of PS234 indicating they were able to suppress the zipA1 (Ts) mutation. To make sure that they were able to completely bypass the ZipA requirement for cell division, both R177C and T249M were recombined onto the chromosome of W3110. The resultant two strains were then tested for their ability to accept zipA∷kan by P1 transduction with W3110 ftsA-R286W as a positive control. zipA∷kan transductants were obtained with all three strains at the same high frequency confirming that the new ftsA alleles could also bypass zipA (data not shown). Both W3110 ftsA-R177C zipA∷kan and W3110 ftsA-T249M zipA∷kan displayed slightly elongated cells when compared to their isogenic zipA⁺ strains (Fig. 3), but this is also true, even if it is less marked, for the W3110 ftsA-R286W strain.

Table 1.

Summary of the ability of various ftsA alleles to complement an FtsA-null strain, of their toxicity by overexpression and interactions in yeast two-hybrid.

Mutations in FtsA	Bang Mutants	Minimum [IPTG] to complement P163 (μM)	[IPTG] to cause toxicity in P163 (μM)	Self–Interaction by Y2H	Interaction with FtsZ by Y2H
WT		15	30	+++	++
G49D	14, 15, 16	15	30	−	++
G49S	2	15	60	+	++
C90Y	10, 31	7.5	not toxic	++	++
D138E	6	30	60	++*	+++
I143F	9	30	60	− *	++
A156V	42, 43	N/A	N/A	+++	+
H159Y	25	30	not toxic	++	++
M167I	46	15	not toxic	−	+++
R177C	41, 44, 45	0	N/A	++*	++
T215A	22	15	not toxic	++*	++
P250L	37	15	30	+++	+/−
A254T	11	30	N/A	+	++
V275A	24	15	250	+	+
V277M	1, 21, 23, 47	15	125	− **	+/−**
R283H	5	15	125	+/−	++
R286Q	17	15	125	+	++
R286W	13	7.5	not toxic	++*	+++
T292A	3	15	250	++	+++
A329V	26	7.5	125	++	+++
A330V	31.2	30	not toxic	+	++
R357A	8	30	not toxic	+++	++

The ability of different FtsA mutants (as indicated in column 1) to complement the FtsA depletion strain P163 [CH2 recA ftsA⁰/pDB280 (rep (Ts) ftsA⁺)] when expressed from the pSEB306 derivative (pBang - column 2) was measured by spot assay (as described Fig. 2). For each allele, the minimal IPTG concentration required to obtain individual colonies at 42°C is reported in column 3. The toxicity by overexpression of these mutants was also measured in the P163 strain when expressed from pSEB306+ derivative plasmids (higher expression levels than pSEB306) as described Fig. 5 and is reported in column 4. These FtsA mutants deleted of their MTS were fused with the DNA binding domain of GAL4 and tested against FtsAΔMTS or full length FtsZ fused with the activating domain of GAL4 (fifth and sixth columns respectively). For each protein interaction tested, the β-galactosidase assay was done multiple times on colonies obtained from several independent transformations using the filter lift assay as described in Clontech manual and gave reproducible results.

+++, indicates a blue color developing in 2 h or less;

++, blue color in 4 h or less;

+, blue color in 8 h or less;

+/−, indicates light blue color after overnight incubation;

−, indicates no blue color after overnight incubation.

^*indicates when known, that mutations at these residues also affect rod formation (data not shown)

^**note that results were not reproducible suggesting some protein stability issue for that mutant in yeast.

Fig. 2.

Ability of various ftsA alleles defective in rod formation to suppress zipA1 (Ts). Plasmids expressing FtsA or FtsA mutants (as annotated) under the control of an IPTG-inducible promoter (pSEB306 series) were transformed into the double mutant strain PS234 [ftsA12 (Ts) zipA1 (Ts)] at 30°C. Colonies of each strain were resuspended in LB, serially diluted 10 fold, spotted onto plates containing increasing IPTG concentrations and incubated at 42°C.

Fig. 3.

Phenotypes of W3110 and W3110 zipA∷kan cells carrying different ftsA alleles. The ftsA mutations (indicated on the top of each double panel) that allow the bypass of ZipA were placed at the native chromosomal ftsA locus by recombineering (see Material and Methods). For each strain, representative pictures of the cells were then taken by phase contrast microscopy. For each panel, the W3110 background is at the top and W3110 zipA∷kan at the bottom. With FtsA and FtsA-G45A only a picture of the W3110 background is shown since the introduction of zipA∷kan into these strains was not possible. A higher resolution version of that figure is provided as Fig. S1.

Surprisingly, employing a screen looking for FtsA mutants with decreased self-interaction we found three mutations R177C, T249M and R286W that allow the bypass of ZipA. This finding indicates that the basis for bypassing ZipA is not increased self-interaction as suggested in earlier studies based only on R286W, but instead may correlate with decreased self-interaction.

Many mutations in ftsA bypass ZipA

The original mutation that bypassed ZipA arose at low frequency suggesting that it was the only change in ftsA that could result in the bypass of zipA (Geissler et al., 2003). However, more recently two mutations in ftsA, one reported to bypass ftsN (E124A) and one reported to suppress some ftsQ defects (I143L) were observed to suppress the loss of zipA (Villanelo et al., 2011; Bernard et al., 2007; Goehring et al., 2007). The determination here that R177C and T249M allow introduction of a zipA deletion confirms that additional alleles of ftsA can be isolated that completely bypass the requirement of zipA for cell growth. Although, T249 is in the same domain as R286 the other three mutations alter residues that map in different domains of FtsA.

To determine if more bypass mutations could be isolated, a plasmid library expressing randomly mutagenized ftsA under control of an IPTG-inducible promoter (pSEB306M) was transformed into PS234 (ftsA12 [Ts] zipA1 [Ts]) and colonies selected at 42°C on plates containing 30 μM IPTG. This concentration of IPTG was chosen since it produces approximately the WT level of FtsA (this is the minimum IPTG concentration that allows pSEB306 to rescue an isogenic strain PS236 carrying only the ftsA12 (Ts) allele). Plasmids (annotated pBANG1 to 46) were isolated from these PS234/pSEB306M colonies and retested for suppression by transforming fresh PS234 cells at 30°C and then testing transformants on increasing concentrations of IPTG at 42°C (Fig. 4). Thirty passed this test and sequencing these revealed twenty one different alleles of ftsA (several were found multiple times, Table 1) that suppressed zipA1 (Ts). Included among these suppressors are the alleles known to bypass ZipA [ftsA-R286W (Bang13) and ftsA-R177C (Bang41, 44 and 45)] and alleles affecting the same residues as known ftsA alleles that bypass ZipA [ftsA-I143F (Bang9) and ftsA-R286Q (Bang17)] (Fig. 4 and Table 1). As expected, because of the condition under which they were obtained, all the mutants allowed PS234 to grow at 42°C on plates with 30 μM IPTG (Fig. 4).

Fig. 4.

Suppression of ftsA12 (Ts) zipA1 (Ts) by the pBang plasmids. pBang plasmids (pSEB306 derivatives carrying ftsA mutations) were isolated as described in the text and transformed into PS234 [ftsA12 (Ts) zipA1 (Ts)] at 30°C. Colonies were then spotted on plates containing increasing IPTG concentrations as described in Fig. 2 and incubated at 42°C. The ftsA allele carried by each plasmid is indicated.

It is known that overexpression of ftsA-R286W is less toxic than overexpression of WT ftsA. Consistent with this, expression of ftsA-R286W from pSEB306 at 1 mM IPTG did not inhibit colony formation (Fig. 4). Many of the ftsA alleles that suppress zipA1 (Ts) behaved similarly. However, a few of these FtsA mutants are toxic when overexpressed, especially FtsA-D138E, A156V and A329V, at least in the absence of functional ZipA. This result indicates that not all alleles of ftsA with the ability to suppress zipA1 (Ts) have reduced toxicity when overexpressed (see later; Fig. 5).

Fig. 5.

Comparison of the toxicity due to overexpression of the different ftsA alleles. pSEB306+ derivatives (providing higher expression levels than pSEB306 used in Fig. 2 and Fig. 4) expressing the different FtsA mutants (as indicated on the right) were transformed into the FtsA depletion strain P163 [CH2 recA ftsA⁰/pDB280 (rep (Ts) ftsA)] at 30°C. Colonies of each strain were resuspended in LB, serially diluted, spotted onto plates containing increasing IPTG concentrations and incubated at 42°C. Note that for PSEB306+ the ftsA alleles R177C, A254T and A156V show some toxicity in P163 (even at 30°C on glucose plates transformants grow very slow and the colonies are flat) and do not grow very well in the spot test.

It is also interesting to note that most of the ftsA mutations allow growth of PS234 (ftsA12 [Ts] zipA1 [Ts]) at a level of induction much lower than 30 μM IPTG suggesting that these ftsA alleles are more efficient than WT ftsA in cell division. This result was confirmed by transforming plasmids carrying the different mutations into an FtsA depletion strain (P163) and testing for complementation at 42°C with increasing IPTG concentrations (Table 1 shows the minimum IPTG concentration required for full complementation). Many of these mutations were able to complement the FtsA depletion strain at the same or a lower induction level than WT FtsA and some mutants such as R286W and R177C complemented with very little or no IPTG. R286W was shown in earlier reports to be as stable as WT FtsA. Also the ability of a FtsA mutant that bypasses ZipA to complement an FtsA depletion strain at lower expression than WT FtsA was previously shown for E124A [which is at least as stable as WT FtsA] (Shiomi and Margolin, 2007; Shiomi and Margolin, 2008) implying that these FtsA mutants might be more efficient for cell division than WT FtsA.

Many of these ftsA alleles are a little more efficient for cell division in the absence of zipA [PS234] than in a zipA⁺ [P163] background. This is also illustrated by the fact that pSEB306 carrying ftsA-A156V or ftsA-A254T are a little toxic in a zipA⁺ background. These plasmids can be transformed into PS234 (ftsA12 [Ts] zipA1 [Ts]) [although at 30°C on glucose plates they don't grow as well as at 42°C (Fig. 4)], however, pSEB306-A156V only gives a few flat and very slow growing colonies in P163 (ftsA⁰zipA⁺/pDB280 (rep (Ts) ftsA⁺) at 42°C. When cloned on the higher expression level plasmid pSEB306+, A156V, R177C and A254T could not be readily introduced into P163 at 30°C (on glucose plates), only giving a few slow growing colonies that did grow well when re-streaked (Fig.5).

To confirm that these 21 suppressors of zipA1 (Ts) were also able to completely bypass ZipA, the ftsA alleles were recombined onto the chromosome of W3110 and the resultant strains tested for acceptance of zipA∷kan by P1 transduction (Fig. 3). Twenty out of the 21 alleles gave similar transduction frequencies as W3110 ftsA-R286W (used as a control; aka 500 to 1000 transductants per experiment) confirming that these alleles bypassed ZipA for cell growth. When the only exception ftsA-D138E (Bang6) was cloned in pSEB306, it was able to suppress zipA1 (Ts) over a very narrow window of expression level (Fig. 4). This result suggests that the physiological level of expression obtained by recombining this allele at the ftsA locus might not be compatible with the level needed to bypass ZipA.

During these transduction experiments we noticed a very low number (between 0 and 4 per transduction) of zipA∷kan transductants with W3110 or W3110 strains containing ftsA alleles unable to bypass ZipA (used as negative controls). These transductants apparently acquired suppressor mutations that allowed ZipA to be bypassed. For a few of these ZipA bypass mutants, we transduced the ftsA alleles (linked to leu∷Tn10) into naïve W3110 and tested the resultant strains for the ability to bypass ZipA. Since none of the transductants showed an increase in the ability to accept zipA∷kan we concluded that these acquired mutations were not at the ftsA locus. Also, these P1 transduction experiments suggest that there are at least two chromosomal loci (in addition to ftsA), which when mutated, allow transduction of zipA∷kan and the bypass of ZipA (one linked to zipA; data not shown). This result suggests that there are additional mechanisms other than mutations in ftsA that allow the ZipA requirement for cell division to be bypassed. These other ZipA bypass mutations, not in ftsA, were not examined further in this study.

When we located the various residues that were altered in these FtsA mutants that allow ZipA to be bypassed (Fig. 6), we noticed that the residues were not grouped in one particular area of the structure but were located in a few areas on all domains of FtsA. This result is more compatible with the bypass mutations impairing the ability of FtsA to self-interact, similar to our earlier experimental results, rather than affecting an interaction with a specific unknown protein partner. We did not get any mutations that affect residues in the area of the FtsA structure that we identified in previous published studies as important for interaction with FtsZ or targeting of FtsA to the membrane suggesting that our search for FtsA mutants that allow the bypass of ZipA was pretty specific despite the numerous mutations obtained.

Fig. 6.

Location of residues in FtsA that allow the bypass of ZipA when altered. The residues altered in the E. coli ftsA mutants that bypass ZipA (Bang mutants) are represented on the Thermotoga maritima FtsA (PDB-1E4G) structure in red if the mutations also affect self-interaction, in blue if self interaction is not affected and in orange when there is no self-interaction data. Previously reported residues essential for interaction with FtsZ are colored black. The colored residues are (E. coli number followed by T. maritima number): in red G45 – G44, G49 – G48, C90 – R90, D138 – D140, Y139 – D141, I143 – V145,, H159 – T161, M167 – V169, K173 – N175, R177 – D179, L204 – R206, G214 – N216, T215 – F217, F247 – L249, M249 – T241, A254 – S256, V275 – I276, V277 – Y278, G281 – D282, R283 – N284, R286 – K287, T292 – K293, A329 – P329, A330 – G330, in blue D47 – D46, E124 - G126, E141 – R143, A156 – I158, P250 – S252, R357 – R357, in orange N170 – M172, C178 – T180, and in black I232 –V234, P233 - P235, Y234 – V236, G236 - G238, V238 – H240, D242 – D244, R300 – R301, E303 – E304, L307 – K308.

Most of the ftsA alleles that bypass ZipA are also impaired for self-interaction

Since the mutations we isolated initially that impaired the ability of FtsA to self-interact (as judged by decreased rod formation and confirmed by the yeast two-hybrid system) bypassed ZipA, we decided to test these newly isolated bypass (“Bang”) mutations for their ability to self interact using the yeast two-hybrid system. As a positive control for expression and stability of our constructs in yeast cells we also tested these mutations for their effect on the FtsA-FtsZ interaction. As illustrated in Table 1, very few of the FtsA mutants seem to be affected for their interaction with FtsZ indicating that the mutant proteins are stable in yeast. In contrast to the lack of effect on the FtsA-FtsZ interaction, most of these mutants (18 out 21) showed reduced ability to self-interact when compared to wild type FtsA.

In work to be described elsewhere, we systematically mutagenized the most conserved residues on the surface of FtsA and used these mutants in a battery of tests to check their properties, including the ability to form rods. This approach allowed us to identify, in addition to the mutants described above, other FtsA mutants that are able to complement an ftsA-null strain and have reduced ability to form rods when expressed as GFP-FtsAΔMTS fusions (from pSEB294 derivatives) suggesting that they are also impaired for self-interaction. These mutations include G45A, Y139A (also tested in yeast two-hybrid system; see Table S1), K173A, L204A, G214D, F247A and G281A. Each of these mutations is able to suppress zipA1 (Ts) when expressed from the pSEB306 plasmid (data not shown and Table S1) and all of them, except G45A and F247A, allow ZipA to be bypassed when recombined onto the W3110 chromosome (Fig. 3). However, when G45A and F247A are expressed from pSEB306 they allow PS234 (ftsA12 [Ts] zipA1 [Ts]) to grow at non-permissive temperature at 60 μM IPTG, indicating that they can also suppress zipA1 (Ts) if expressed at a higher level. This requirement for a higher level of expression than provided by the chromosomal locus (equivalent to 30 μM IPTG for that strain) explains why these two alleles are not able to bypass the requirement for ZipA when recombined onto the chromosome (data not shown).

In summary, every single time we find an ftsA mutation that decreases FtsA self-interaction but is still functional, it is also able to bypass ZipA.

FtsA mutants more efficient at functioning in septation bypass ZipA

During our systematic mutagenesis of conserved surface residues on FtsA (to be reported elsewhere) we isolated four mutants (FtsA-D47K, -E141K, -N170E, and -C178A) that allow formation of individual colonies in the complementation test of the FtsA depletion strain (P163) at a lower expression level than WT FtsA (Table S1). In the experimental work presented above we noticed that some ftsA alleles that allow bypass of ZipA required less expression to complement an ftsA null mutant and form individual colonies than cells expressing WT ftsA (Table 1 and Table S1) implying that these FtsA mutants are more efficient in cell division than the wild type protein. We therefore tested the four mutants from the systematic mutagenesis for their ability to bypass ZipA. All four of them allow PS234 strain to grow at 42°C suggesting that these alleles could bypass ZipA (Supplemental Table 2). This was confirmed when these alleles of ftsA were recombined onto the chromosome at the native locus and we observed that zipA∷kan could be readily introduced into these strains by P1 transduction (Fig. 3). Thus, it appears that ftsA alleles that function more efficiently in cell division than the wild type are likely to allow ZipA to be bypassed.

We also observed that ftsA alleles that are very efficient for cell division have variable toxicity levels when overexpressed (Fig. 5). This is also true when we compare the overexpression toxicity of different alleles of ftsA that affect residue R286. These alleles have very different degrees of toxicity when overexpressed (Fig. 5) even though all of them reduce FtsA's self-interaction (to varying degrees) and all, besides ftsA-R286E, are equally or more efficient than WT ftsA in supporting cell division (Table S1). These results show that there is no apparent relationship between the ability of an FtsA mutant to support cell division at a lower expression level than WT and its overexpression toxicity.

Finally, it is also interesting to note (Fig. 3) that for each mutant, cells are always shorter in the presence of zipA (top panels) than in its absence (zipA∷kan cells in bottom panels). This suggests that ZipA performs a non-essential function that improves the process of cell division and that none of these FtsA mutants are able to provide it.

Residues in FtsA's H1 helix are essential for function but not because of a role in self-interaction

A recent study suggested that residues E69 and M71 are very important for FtsA's self-interaction based upon results obtained using the bacterial two-hybrid system. The study showed that mutations altering these residues decreased FtsA's self-interaction (Shiomi and Margolin, 2007). These residues were examined because they were implicated as important for self-interaction in an earlier computer based study examining the co-variation of FtsA's conserved residues during evolution (Carettoni et al., 2003). This latter analysis suggested that FtsA's alpha helix H1, in which both E69 and M71 are located, is important for self-interaction. Neither our screen for FtsA mutants impaired in rod formation nor our screen for bypassing the requirement of ZipA in cell division identified mutations that affected amino acids in helix H1. This was surprising because if this part of the protein is important for self-interaction, and functional mutants that are affected for self-interaction gain the ability to bypass ZipA, we might have obtained mutations affecting this helix. A possible explanation from that study is that FtsA's self-interaction is essential to prevent FtsA monomer buildup which is proposed be deleterious to Z ring formation (Shiomi and Margolin, 2007). However, this hypothesis is based upon: 1) the FtsA-R286W mutant having increased self-interaction and less toxicity than WT FtsA when overexpressed, and 2) the FtsA-M71A mutant having decreased self-interaction and more toxicity than WT FtsA when overexpressed. It was proposed that FtsA monomers cause deleterious interactions with FtsZ and that increasing oligomerization of FtsA enhances Z ring integrity (Beuria et al., 2009; Shiomi and Margolin, 2007).

In contrast to the above hypothesis, our results obtained so far indicate that decreasing the ability of FtsA to self-interact does not usually impair FtsA function or Z ring integrity, but in many cases it even causes a gain of function since it allows FtsA to bypass ZipA. Because of this, we decided to re-investigate the effect of mutations that alter residues E69 and M71 in helix H1 for their effect on FtsA self-interaction, FtsA function and the ability of FtsA to bypass ZipA for cell growth. We therefore made mutations that altered these two residues by site directed mutagenesis and then tested the alleles for self-interaction by the yeast two-hybrid assay (Table S2). We also expressed the alleles in P163 (ftsA depletion strain) to check if they were functional and in PS234 (ftsA12 [Ts] zipA1 [Ts]) to check their ability to bypass the ZipA requirement for cell division.

As controls for the yeast two-hybrid test we also used FtsA variants with a complete deletion of domain 1C or lacking the S12-S13 region. Previous results show that these mutants have lost FtsA self-interaction while retaining the ability to interact well with FtsZ (Rico et al., 2004; Corbin et al., 2004). As expected, these two constructs showed a loss of self-interaction but no change in interaction with FtsZ further validating the yeast 2-hybrid system (Table S2). Even though FtsAΔS12-13 shows no self-interaction in our yeast two-hybrid assay, it is still able to form, but only rarely, a few rods when tested in our rod formation assay, suggesting that its ability to self interact is dramatically impaired but not completely abolished. With FtsAΔ1C we never observed any rods in our assay (data not shown). It was previously shown that these two proteins, despite being monomeric or at least very impaired for self-interaction, still localize very efficiently to the Z-ring even though they are not fully functional for cell division (Rico et al., 2004; Corbin et al., 2004). Overexpression of FtsAΔ1C did not appear to be more toxic than WT FtsA and overexpression of FtsAΔS12-13 caused a phenotype consistent with a stimulation of Z ring assembly, such as shorter cell length and the localization of FtsZ to the cell poles, suggesting that Z ring formation is more resistant to MinC than with WT FtsA (Rico et al., 2004). This is despite FtsAΔS12-13 probably being mostly monomeric, still interacting with FtsZ as well as WT FtsA and having the IA domain that is thought to act like MinC to destabilize the FtsZ rings when FtsA is a monomer. These results do not support the proposal that FtsA monomers cause deleterious interactions with FtsZ and that increasing the oligomerization of FtsA enhances Z ring integrity.

Most of the mutants we tested with altered residues in helix H1 had a moderate loss of interaction with both FtsA and FtsZ suggesting that protein expression or stability in yeast was slightly affected. Significantly though, the M71A mutation (also M71E) did not show a significant loss of self-interaction in contrast to what was observed with the bacterial two-hybrid system (Shiomi and Margolin, 2007). Consistent with the earlier study, FtsA-M71A and FtsA-M71E are not functional since they do not complement an FtsA depletion strain (data not shown). Also, as reported in that study, overexpression of FtsA-E69P and FtsA-E69R cause very different phenotypes (one more toxic and the other less toxic than WT FtsA, respectively). This difference led to the prediction that these two mutants have very different ability to self-interact. However, in the yeast two-hybrid assay, these mutants behave identically with a moderate loss of self-interaction when compared to WT FtsA. In addition, it was reported that FtsA-E69R was able to complement the ftsA12 (Ts) mutation (Shiomi and Margolin, 2007), but we found it was not able to complement the ftsA null mutation (P163-ftsA depletion strain), indicating that this allele is not fully functional. The complementation of ftsA12 by ftsA-E69R shows that FtsA12 and FtsA-E69R (each non functional) are able to complement each other to supply FtsA function, suggesting that they probably interact with each other.

FtsA-E69V is the only mutant we ever tested that caused an increase in FtsA self-interaction in the yeast two-hybrid assay (Table S2). This mutant was described in a previous study as functional and consistent with this, we found that this mutant is able to complement an FtsA depletion strain (Fig. 5). From the previous study one would predict that this mutant should also stimulate Z ring assembly, potentially allowing the bypass of ZipA, and should be less toxic than WT FtsA when overexpressed. Our results show that this is not the case. The E69V mutant does not suppress zipA1 (Ts) (data not shown), does not cause any minicelling nor produce shorter cells suggesting it is not more resistant to MinC than WT FtsA (data not shown) and it is at least as toxic as WT FtsA when overexpressed (Fig. 5).

In summary, it appears that helix H1 is very important for FtsA function since most of the mutations we made that affect it were not functional. However, the reason for loss of function is probably not due to their effect on FtsA's ability to self-interact.

The results obtained above with FtsAΔS12-13 are consistent with what we observed with FtsA-R286W (R286 is in the S12-13 region of FtsA) raising the possibility that the results concerning self-interaction obtained in the earlier studies (Shiomi and Margolin, 2007; Shiomi and Margolin, 2008) using the bacterial two-hybrid system might be problematic. As an additional test we made more mutations that altered R286 and tested them in the yeast two-hybrid assay (Table S1). Every single mutation altering this residue caused a loss of FtsA self-interaction and every single mutant gained the ability to bypass ZipA requirement for cell division (Table S1 and Fig. 3). We also tested other FtsA mutants (E124A, E124Q and I143L) described in earlier studies, which allow bypass of ZipA (Bernard et al., 2007; Goehring et al., 2007). The assumption before this study was that they would act like FtsA-R286W and increase FtsA's self-interaction and as a consequence stimulate Z ring assembly. However, and consistent with what we found with FtsA-I143F (Bang9), the I143L substitution decreased the ability of FtsA to self-interact. FtsA-E124A and FtsA-E124Q behaved similar to each other, showing qualitatively no difference in self-interaction compared to WT FtsA. As a control we also tested all of these constructs for interaction with FtsZ in the yeast two-hybrid. All substitutions at R286 and I143 had little effect on the interaction with FtsZ (some actually enhanced interaction compared to WT FtsA), whereas changes at residue E124 cause a moderate loss in the ability of FtsA to interact with FtsZ.

Finally, helix H1 does not seem to be involved in FtsA's self-interaction. Also, the location of residues on the FtsA molecule (Fig. 6) altered in our mutants affected for self-interaction do not support the proposed model for FtsA polymerization based on the study of co-variation of conserved residues in evolution (Carettoni et al., 2003).

Discussion

In the results presented here we describe many alleles of ftsA that can efficiently bypass the requirement for ZipA in cytokinesis. We also show that most of these FtsA mutants have decreased ability to self-interact while their interaction with FtsZ is largely unaffected. The most striking observation we made in this report is that each time we encounter a functional FtsA mutant displaying reduced self-interaction, it is able to bypass the requirement for ZipA.

In contrast to earlier studies that proposed that FtsA monomers are deleterious to Z-ring formation (Beuria et al., 2009; Shiomi and Margolin, 2007), our results demonstrate that FtsA mutants impaired for self-interaction are able to support Z-ring formation and are often more efficient than wild type FtsA. Thus, decreasing the ability of FtsA to self-interact is clearly the basis of one of the mechanisms that allow cells to bypass the ZipA requirement for cytokinesis in E coli.

The difference in the self interaction results obtained between the yeast two-hybrid assay used here and the bacterial two-hybrid assay used in earlier studies leads to a different interpretation about how self-interaction influences FtsA's function in cell division. We believe the yeast two-hybrid results are a better reflection of FtsA's self-interaction behavior because, in every case it was examined, the results are consistent with the behavior of these mutant proteins in the rod forming assay [30 times, 8 in this work and the 22 in earlier reports (Pichoff and Lutkenhaus, 2007; Pichoff and Lutkenhaus, 2005)]. We also think that measuring the interaction between E. coli proteins that are normally involved within the same protein complex using the bacterial two-hybrid system [which is done in E. coli (Karimova et al., 1998)] could lead to some artefactual results. For example, an interaction detected between FtsZ and ZapB was later found to be indirect as it depended upon a third intermediate partner ZapA (Galli and Gerdes, 2010). In the same way, FtsA mutants that interact with FtsZ better than the WT FtsA (such as FtsA-R286W in both yeast and bacterial two-hybrid experiments) could artificially display a stronger self-interaction in the bacterial two-hybrid system because they might be recruited more efficiently to the Z ring causing an accumulation of the mutant protein that leads to an artificially enhanced signal. On the other hand, alleles that encode mutants affecting the interaction with FtsZ could show lower self-interaction than WT FtsA because they are not as efficiently recruited to the Z ring (they may not compete with the endogenous FtsA present in the cells). Also, along the same line, mutants that are more toxic than WT FtsA may accumulate additional mutations lowering their ability to interact with FtsZ and/or FtsA during the actual bacterial two-hybrid assay which uses conditions were the proteins to be tested are highly expressed for a long period of time (at least overnight) (Karimova et al., 1998; Shiomi and Margolin, 2007).

A new model for the role of FtsA to explain the ability of FtsA mutants with decreased self-interaction to bypass ZipA

In an earlier study we showed that either FtsA or ZipA is able to attach FtsZ polymers to the membrane enabling Z-ring formation but both of them are required for the recruitment of the later cell division proteins so that constriction can ensue (Pichoff and Lutkenhaus, 2002; Hale and de Boer, 2002). The ability to bypass the requirement for ZipA in cell division means that these FtsA mutants are able to recruit the later components of the divisome and allow constriction. These FtsA mutants are able to do this independently, and the presence or absence of ZipA in the cells does not seem to matter too much, although the presence of ZipA is beneficial because cells with ZipA are always shorter than cells without it (Fig. 3). Because FtsA is more conserved in evolution than ZipA, it is likely that FtsA is responsible for the recruitment of the other conserved cell division proteins and that ZipA is only facilitating the process. Also, there is evidence that FtsA interacts with later cell division proteins.

Recent publications suggest that most proteins that are components of the division machinery downstream of FtsA rely on a complex network of interactions to be recruited to the divisome, instead of a single upstream protein (Goley et al., 2011; Gamba et al., 2009; Aarsman et al., 2005; de Boer, 2010). This lack of a linear hierarchical pathway is consistent with previous observations (Goehring et al., 2005) using a ZapA fusion to late cell division proteins showing that artificially recruiting a later acting cell division protein to the Z ring allows the back recruiting of the earlier acting proteins. That study showed that if FtsA is not present FtsN cannot be recruited to the Z ring even if all other essential components of the divisome are present. This suggests that FtsA, in addition to interacting with earlier cell division proteins, also interacts with FtsN, the last known essential protein recruited to the Z ring just before constriction commences. There is also evidence that domain IC of FtsA interacts with FtsN (Corbin et al., 2004). Interestingly, in Caulobacter crescentus, FtsN is recruited immediately after FtsA and before the other “conserved” components such as FtsL, FtsI or FtsW, which in E. coli, are recruited to the Z ring before FtsN (Goley et al., 2011).

From the results presented here, it appears that lowering FtsA's ability to self-interact helps with its ability to recruit one or more of these later cell division proteins (an earlier component of the divisome or FtsN or maybe both). This could also explain why at least a few of these FtsA mutants are also able to bypass the requirement for other cell division proteins like FtsK or correct deficiencies of some FtsQ mutants (Bernard et al., 2007; Goehring et al., 2007; Geissler and Margolin, 2005). The domain IC of FtsA as shown here and in earlier studies is not only involved in self-interaction of FtsA (Rico et al., 2004; Shiomi and Margolin, 2007) but is also involved in recruiting later cell division proteins such as FtsN and FtsI (Rico et al., 2004; Corbin et al., 2004). It has also been suggested that domain IC of FtsA is flexible in relation to the whole molecule (van den Ent and Lowe, 2000). Thus, it is likely that domain IC is involved in self-interaction and recruitment of downstream proteins. An interesting possibility is that these two interactions compete with each other and when domain IC is locked in position in an FtsA polymer it is less able to interact with the late cell division proteins.

Since both FtsA and ZipA compete for FtsZ interaction (they both interact with the same region of the carboxy tail of FtsZ) it is not surprising that any change in FtsZ/ZipA or FtsZ/FtsA ratios causes inhibition of cell division (Haney et al., 2001; Pichoff and Lutkenhaus, 2002). One attractive possibility is that ZipA on a Z ring modulates the self-interaction of FtsA on the ring and by doing this ZipA regulates the ability of FtsA to recruit the latter cell division proteins.

In a model in which FtsA monomers interact better with one or more of the later cell division proteins than FtsA multimers we could make the following predictions. In a situation where there are mostly FtsA monomers on the Z ring, even in the absence of ZipA, these FtsA molecules are able to attach FtsZ polymers to the membrane to form a Z ring and are also able to recruit the later cell division proteins to the ring efficiently to allow cell division. If ZipA is present, cell division might be even more efficient because ZipA could help with Z-ring formation and the FtsA mutants could recruit the late division proteins earlier in the cell cycle, leading to division even earlier in the cell cycle resulting in shorter cells. Another prediction is that too much WT FtsA on the ring leads to too much FtsA self-interaction (and displacement of ZipA) causing inefficient recruitment of the later cell division proteins, ultimately blocking cell division. In contrast, an excess of an FtsA mutant with decreased self-interaction on the ring would not be too damaging since it would still be able to recruit the later cell division proteins and allow constriction (displacement of ZipA won't matter since its essential function is no longer required), suggesting that such a mutant would be less toxic than the WT FtsA upon overexpression. We have found mutants in our study that behave according to these predictions as does the original FtsA-R286W. Even if not all mutants that bypass ZipA and have decreased self-interaction follow these predictions, it is interesting to note that 30 out of 38 are not as toxic as WT FtsA when overexpressed and 12 out of 38 complement an FtsA-depletion strain at a lower induction level than WT FtsA.

It is not clear if lowering the ability to self-interact is the only mechanism by which the FtsA mutants can bypass the ZipA requirement for cell division. We did obtain 7 out 38 FtsA mutants that do not seem to be affected for self-interaction and are still able to bypass ZipA. It is possible that these mutants have decreased self-interaction but it is too modest to be detected in our tests. This is likely because the residues affected in these mutants map on the same area of the FtsA molecule (Fig. 6) as residues involved in self interaction. Another possibility though is that these mutants bind a late recruited protein with a higher affinity and this leads to the bypass of ZipA.

Finally, we have mutants (not shown in this study) that seem to have lost the ability to self-interact and are unable to bypass ZipA. However, they are unable to complement an FtsA depletion mutant and this lack of function would explain why they are not able to bypass the requirement for ZipA. We do not know if these mutants are not functional because they cannot self-interact, which would suggest that some FtsA self-interaction is essential for its function, or if the mutations affect some other property of FtsA (for example, its ability to interact with an essential component of the cell division machinery). Before this can be answered more work needs to be carried out to characterize these mutants in detail.

Bypassing the requirement for ZipA is relatively easy so why is ZipA essential even though it is not conserved?

Many different mutations in ftsA are able to completely bypass the ZipA requirement for cell division. We also noticed during our P1 transduction experiments of zipA∷kan that bypass mutations arose at a frequency of 10⁻⁴. In the cases we examined these bypass mutations are not in ftsA but occur in at least two other loci (one linked to the zipA locus). Finally in our earlier work we determined that the spontaneous survival at 42°C of our zipA1 (Ts) mutant strain was quite high with about 10⁻⁴ survivors (these could be intergenic) (Pichoff and Lutkenhaus, 2002). These results suggest that the ZipA requirement for cell division is relatively easy to bypass and makes one wonder why ZipA is essential despite not being conserved in evolution. One hint is that the cell length of all the FtsA mutants that bypass ZipA are longer in the absence of ZipA than in a zipA⁺ background. Also, cells of the FtsA mutants, in the absence of ZipA, are almost never as short as WT cells. So ZipA appears to enhance cytokinesis even when it is not needed for Z-ring assembly or recruitment of downstream proteins. ZipA could enhance cytokinesis by itself (there is evidence that ZipA promotes FtsZ polymerization in vivo and in vitro) or because ZipA directly recruits the MinC/DicB inhibitor of cell division to the Z ring, it is also possible that ZipA recruits a protein with a positive, but not an essential, role in the cytokinesis process.

It is interesting to note that in Gram positive bacteria, and especially in Bacillus subtilis, there are three proteins, FtsA, EzrA and SepF (YmlF) that interact with the conserved C-terminal tail of FtsZ (Singh et al., 2008; Singh et al., 2007) and are involved in the regulation of Z-ring formation. FtsA and SepF promote Z-ring formation while EzrA is considered a negative regulator, since ezrA null mutants have extra Z rings (Singh et al., 2008; Hamoen et al., 2006; Haeusser et al., 2004; Gueiros-Filho and Losick, 2002). None of these genes are considered absolutely essential, since in their absence cells still grow although they have different phenotypic defects. While sepF or ezrA mutants are not dramatically affected for cell division, ftsA mutants are very filamentous. While a sepF mutant is synthetic lethal with ftsA or ezrA mutants, sepF overexpression suppresses the filamentation phenotype of an ftsA mutant (Singh et al., 2008; Hamoen et al., 2006). FtsA and EzrA are able to interact with multiple late divisome components, in addition to FtsZ, whereas SepF appears to interact exclusively with FtsZ or EzrA [at least in Staphylococcus aureus, (Steele et al., 2011)]. It is possible that what is happening in these different mutants of B. subtilis is related to what we think is happening with FtsA mutants that bypass ZipA. In the absence of FtsA, overexpression of SepF could counteract the deleterious effect that EzrA has on Z-ring formation by regulating EzrA interaction with the Z ring or by interacting directly with EzrA and blocking its activity. This could allow EzrA to function as a membrane tether for FtsZ polymers leading to the formation of the Z ring, and since EzrA interacts with other cell division proteins a fully functional cytokinetic ring can form. In the absence of EzrA, SepF could allow FtsA to behave like some of our FtsA mutants that cause the E. coli Z ring to be more resistant to MinC, thereby allowing formation of extra Z rings inside the cells. When both EzrA and SepF are absent, this would phenocopy too much FtsA self-interaction (overexpression of FtsA in E. coli may compete with ZipA binding to the FtsZ's C-terminal sequence) in which the later cell division proteins cannot be efficiently recruited to form a complete septal ring.

Based on the above hypothesis, we suggest that one essential role of ZipA (in addition to its non-essential role in enhancing the cytokinesis process) is to regulate FtsA's self-interaction at the Z ring, ultimately allowing formation of a complete septal ring by promoting FtsA's ability to recruit downstream division proteins.

Materials and Methods

Media, culture and phenotypic observations

All bacterial cultures and microscopy techniques (phase contrast and fluorescence) were done as described earlier (Pichoff and Lutkenhaus, 2005). LB media or M9 minimal media supplemented with glucose were used in these studies. Additions to media were: glucose at 0.2%, arabinose at 0.0001%, ampicilin at 100 μg/ml, tetracycline at 12 μg/ml, kanamycin at 25 μg/ml, and variable concentrations of IPTG as indicated in the text.

Plasmids and strains

XL1-Blue (Clonetech), W3110 (WT strain, lab collection), PS234 [W3110 ftsA12 (Ts) zipA1 (Ts)], PS236 [W3110 ftsA12 (Ts)], P163 [CH2 recA ftsA⁰/pDB280 (rep (Ts) ftsA)] were described earlier (Pichoff and Lutkenhaus, 2007; Pichoff and Lutkenhaus, 2005; Pichoff and Lutkenhaus, 2002) and used for mutant selection, overexpression and complementation assays as described in the text. Plasmids pSEB306, pSEB306+ (express ftsA under an IPTG inducible promoter, derivatives of pDSW210 [lower expression] and pDSW208 [higher expression] respectively) and pSEB294 (pBAD18 derivative that express GFP-FtsAΔMTS) were also described in these previous reports.

Construction of W3110 ftsA12

P1 phage was grown on W3110 and the leu⁺ gene was transduced into PS236 [W3110 ftsA12 (Ts) leu∷Tn10] recipient. Transductants were selected on M9 minimal + 8 mM sodium citrate plates at 30°C (to select for the gain of leu⁺ gene). 20 of these colonies were then subcloned on fresh plates of the same media at 30°C and 42°C (to check that about 50 % of the cells were still thermosensitive [frequency of co-transduction between the ftsA and leu]). Finally the loss of tetracycline resistance (carried by Tn10) of these W3310 ftsA12 (Ts) leu⁺ colonies was confirmed.

Construction of pSEB385

A PCR fragment carrying ftsAΔMTS (encoding for FtsA deleted of its last 15 amino acids) was obtained by PCR using W3110 DNA as template and the oligos 5′ FtsAdel EcoRI and XbaI 3'fus. The PCR fragment was then cut by EcoRI and XbaI and cloned into pJC104 [a pBAD18 derivative carrying ftsZ-GFPmut2 (Mukherjee et al., 2001)] in order to replace the ftsZ gene (cloned between the same restriction sites) by ftsAΔMTS giving pSEB385.

Recombineering of ftsA alleles onto the chromosome

The recombineering technique used here using the pKD46 plasmid was adapted from (Datsenko and Wanner, 2000). Each ftsA allele was amplified by PCR using the Pfu polymerase (Stratagene) and a plasmid template carrying the desired mutation. PCR fragments were then purified by QIAquick (Qiagen) columns according to manufacturer's guidelines and resuspended in 50 μl of water. 2.5 μl of these PCR fragments were electroporated into 100 μl of PS234 (W3110 ftsA12leu∷Tn10zipA1)/pKD46 cells prepared as described below. Transformants were then plated at 42°C on LB agar plates and grown overnight. P1 phage was grown on one transductant and the ftsA allele it carries was then transduced into a W3110 ftsA12 recipient. Transductants were selected on LB agar + 8 mM sodium citrate plates at 42°C. 20 of these colonies were then subcloned on fresh plates of the same media at 37°C with or without tetracycline (to check that co-transduction between the ftsA allele and leu∷Tn10 was around the expected frequency of 50%). One colony of each, tetracycline resistant or sensitive, was picked and the chromosomal ftsA allele was amplified by PCR using the Pfu polymerase (Stratagene) and checked for the presence of the desired mutation by sequencing of the PCR fragment. Sometimes, electroporation into the PS234 derivative did not yield any transformants so the allele was introduced into PS236 (W3110 ftsA12leu∷Tn10)/pKD46. Using this procedure, the only ftsA alleles we were not able to recombine onto the chromosome were the ones that required higher expression levels than wild type ftsA to suppress an ftsA12 or a ΔftsA strains at the nonpermissive conditions.

For the electroporation experiments, strains were grown at 30°C over night in LB+ ampicillin, then diluted 1/100 in fresh media. At OD₅₄₀ about 0.1, 0.04% arabinose was added to the culture and incubated at 30°C for an additional 3 hours. Cultures where then placed on ice and cells prepared as follows: the chilled cultures were centrifuged for 5 min at 7K rpm and washed 3 times using cold filtered solution of 20% glycerol in H₂0 (100 ml, 50 ml and 25 ml washes for each 100 ml culture). The final pellets were resuspended in 400 μl of 20% glycerol and 50 μl of cells were used per electroporation. Unused cells were stored at −80°C. Electroporations were done using Biorad Gene Pulse with the EC1 factory setting. Following the pulse, 600 μl LB was added and the suspension was incubated at 37°C for 60 minutes and then 150 μl of suspension was spread per plate. Electroporation without DNA gave no colonies while the one with 1.5 μl of purified PCR fragments gave about 100 colonies per plate.

Mutagenesis (random and site directed)

As a general rule, plasmids carrying a mutated allele of ftsA had the residue change following the plasmid such as pSEBXXX-R286W while the plasmid carrying wild type ftsA would be called pSEBXXX. Also, pSEBXXX-M represents the library of plasmids with randomly mutagenized ftsA alleles. Each mutant was confirmed by sequencing.

Random PCR mutagenesis of ftsA or ftsAΔMTS was done using Stratagene's Gene Morph II kit with similar conditions as described earlier (Pichoff and Lutkenhaus, 2007). Mutagenized PCR fragments were cloned into the plasmids using the same restriction sites as during the construction of the original non mutated plasmids pSEB306 or pSEB385.

Site directed mutagenesis of ftsA was done using the Stratagene's Quickchange kit following the manufacturer guidelines. The sequence of the mutagenic primers will be provided upon request.

Y2H protocol and plasmids

The yeast two-hybrid assay was done exactly as described earlier (Pichoff and Lutkenhaus, 2007; Pichoff and Lutkenhaus, 2002) using the same yeast strain (SFY524) and pairs of plasmids for the interaction with FtsA (pSEB347[Gal4bd-FtsAΔMTS]/pSEB304[Gal4act-FtsAΔMTS]) or with FtsZ (pSEB347[Gal4bd-FtsAΔMTS]/pSEB135[Gal4act-FtsZ]). Each ftsA mutation was introduced into the pSEB347 plasmid by site directed mutagenesis.