The FtsA protein is a member of the actin superfamily that localizes to the bacterial septal ring during cell division. Deletions of domain 1C or the S12 and S13 β-strands in domain 2B of the Escherichia coli FtsA, previously postulated to be involved in dimerization, result in partially active proteins that do not allow the normal progression of septation. The truncated FtsA protein lacking domain 1C (FtsAΔ1C) localizes in correctly placed division rings, together with FtsZ and ZipA, but does not interact with other FtsA molecules in the yeast two-hybrid assay, and fails to recruit FtsQ and FtsN into the division ring. The rings containing FtsAΔ1C are therefore incomplete and do not support division. The production of high levels of FtsAΔ1C causes filamentation, an effect that has been reported to result as well from the imbalance between FtsA+ and FtsZ+ molecules. These data indicate that the domain 1C of FtsA participates in the interaction of the protein with other FtsA molecules and with the other proteins that are incorporated at later stages of ring assembly, and is not involved in the interaction with FtsZ and the localization of FtsA to the septal ring. The deletion of the S12–S13 strands of domain 2B generates a protein (FtsAΔS12–13) that retains the ability to interact with FtsA+. When the mutated protein is expressed at wild-type levels, it localizes into division rings and recruits FtsQ and FtsN, but it fails to sustain septation at normal levels resulting in filamentation. A fivefold overexpression of FtsAΔS12–13 produces short cells that have normal division rings, but also cells with polar localization of the mutated protein, and cells with rings at abnormal positions that result in the production of a fraction (15%) of small nucleoid-free cells. The S12–S13 strands of domain 2B are not essential for septation, but affect the localization of the division ring.
Structurally FtsA belongs to the superfamily of ATP-binding proteins represented by actin, Hsp70 and hexokinase (Bork et al., 1992). The structure of FtsA from Thermotoga maritima shows that it strikingly differs from actin and other members of the family (including the bacterial actin homologue MreB) in the orientation of the whole 1C domain (van den Ent and Löwe, 2000). The E. coli FtsA protein is able to bind ATP (Sánchez et al., 1994) and to interact with other division proteins (FtsN, FtsQ and FtsZ) and with itself (Hale and de Boer, 1997; Wang et al., 1997; Yim et al., 2000). FtsA obtained from Bacillus subtilis membranes has been reported to have a strong ATPase activity (Feucht et al., 2001), although this has not been found in FtsA from E. coli or T. maritima (J. Mingorance and M. Vicente, unpubl. results).
Phe420, the very last carboxy-terminal residue, is dispensable both for the role of FtsA in septation and for the correct interaction of the protein with itself (Yim et al., 2000). However, deletions that eliminate the last five carboxy-terminal residues of FtsA cause the loss of both the septation activity and the interaction, but they have no effect on the correct localization of the truncated protein in the septal ring. Extending the deletion into W415, the sixth residue from the carboxy end, results in a protein that is unable to localize into the septal ring. Therefore, the carboxy-terminal end is involved in the interaction between the E. coli FtsA molecules, and moreover, the ability to establish these interactions correctly is required for septation.
A predictive model to describe the dimerization of E. coli FtsA (Carettoni et al., 2003) suggests that domains 1C and 2B include some regions that are crucial to establish the interaction between the E. coli FtsA molecules. A non-lethal gain-of-function point mutation, R286W, in the FtsA S13 β-strand (domain 2B) has already been isolated and found to bypass the requirement for a functional ZipA protein to complete septation (Geissler et al., 2003).
We have constructed variants of the E. coli FtsA protein in which either the 1C domain or the S12 and S13 β-strands in domain 2B have been deleted. Although none of the resultant proteins is fully functional in septation, both retain some partial activity. The absence of domain 1C abolishes the ability of the truncated protein to interact with other FtsA molecules, but does not prevent the localization of the otherwise inert truncated protein at unproductive potential septation sites correctly distributed along the length of the cell. On the contrary, the deletion of the S12–S13 β-strands generates a protein that still retains some ability to interact with FtsA+ and provokes the misplacement of active septation sites in the cell, yielding DNA-less and other aberrant cells.
FtsA lacking domain 1C fails to sustain division in E. coli
Comparison of the structure of ATP-binding proteins of the actin family reveals that in the structure of T. maritima FtsA domain 1C occupies a remarkably different position relative to that occupied by the equivalent domain 1B in other members of the family (van den Ent and Löwe, 2000). A predictive model for the structure of the E. coli FtsA protein, based on the crystal structure of FtsA from T. maritima, has been published (Carettoni et al., 2003). The model suggested that domain 1C may play a role in the interaction between FtsA molecules and this might be important for the activity of the protein in septation. To test these possibilities, we constructed a deletion mutant of FtsA lacking domain 1C. This domain is continuous and protrudes from the core of the molecule, suggesting that it might be removed without altering the structure of the remaining protein fragment. The deletion was designed using as a blueprint the three-dimensional (3D) model of the E. coli FtsA derived from the crystal structure of T. maritima FtsA (van den Ent and Löwe, 2000), and giving particular attention to two factors: the proximity of the residues located at the beginning and the end of the deletion and the localization of the initial residue in a flexible loop that after the deletion might allow the connection to the residue to prevent alterations in the 3D structure of the remainder of the molecule. According to these criteria a deletion was constructed spanning residues Gly84 to Ile161 (Fig. 1).
The deletion was introduced into pMSV20 (Sánchez et al., 1994), and into pLYV30, where it is under the control of the inducible Ptac promoter (Yim et al., 2000), yielding the plasmids pARV1 and pARV11 respectively. The levels of the wild-type and mutated proteins produced by these plasmids in the E. coli strain OV16 were checked by Western blotting using an affinity-purified antibody raised against E. coli FtsA. OV16 contains an ftsA16 amber allele in a temperature-sensitive suppressor background, and is therefore a thermonull mutant (Donachie et al., 1979; Lutkenhaus and Donachie, 1979). At permissive temperature, 30°C, the amount of FtsA in OV16 is ≈ 10% that of a wild-type strain grown under similar conditions. Transformation of OV16 with the plasmid pMSV20 increased the amount of FtsA to nearly wild-type levels. In the absence of IPTG, the cells transformed with pLYV30 had five times more FtsA than pMSV20 transformants, while they had about 50 times more after 2 h of induction with 50 µM of IPTG. The expression of the deletion gene from pARV1 and pARV11 produced truncated proteins with a molecular weight of 34 kDa, as predicted from the sequence. The levels of these proteins were similar to those of the wild-type protein (43 kDa) produced from pMSV20 or pLYV30 either at 30°C or at 42°C, showing that the deletion does not alter the stability of the remaining protein fragment.
To test the functionality of the mutant FtsAΔ1C protein OV16 was transformed with pARV1 and pMSV20 plasmids. The ability of the transformants to grow at 30°C or 42°C was tested on NBT plates. The ftsAΔ1C allele at 42°C did not complement ftsA16 in OV16, while the control ftsA+ allele did. The same results were obtained when we tested transformants yielding higher expression levels, resulting from the leakiness of the non-induced Ptac promoter in pJF119 derivatives: pARV11 (ftsAΔ1C) and pLYV30 (ftsA+).
In agreement with the lack of complementation of ftsA16 by ftsAΔ1C, OV16/pARV1 (ftsAΔ1C) cells formed long filaments at 42°C (Fig. 2), while the length of the control strain OV16/pMSV20 (ftsA+) was similar to the wild type (OV2; Yim et al., 2000) and slightly shorter than the length of OV16 at 30°C. We conclude that FtsAΔ1C is stable but non-functional in vivo, suggesting that domain 1C is needed for FtsA to be active in septation.
Lack of interaction between FtsAΔ1C and FtsA+
We have analysed the interacting properties of FtsAΔ1C and its ability to localize into the septal ring because both features have been already related to the activity of FtsA in cell division (Yim et al., 2000; Carettoni et al., 2003). Using the yeast two-hybrid system, we analysed the ability of FtsAΔ1C to interact with FtsA+. The ftsA+ and ftsAΔ1C genes were each fused to the GAL4-binding and to the GAL4-activation domain sequences, and the resulting plasmids were transformed into the yeast reporter strain Y187. The expression of these four fusions in yeast was checked by Western blotting, confirming that all of them produced similar amounts of fusion proteins. None of the single transformants producing either FtsA+ or the mutated protein activated the expression of the reporter gene, while in the transformants containing the two ftsA+ fusions the expression of the reporter was induced as previously reported (Yim et al., 2000). The combination of the FtsAΔ1C, either at the GAL4 binding or activation domain, with the complete FtsA+ did not activate the expression of the reporter, indicating that the absence of domain 1C in any of the two FtsA fusions impaired the interaction between two FtsA molecules. Accordingly, the combination of the two fusions with FtsAΔ1C did not activate the expression of the reporter gene either.
Localization of FtsAΔ1C into septal rings
The carboxy-terminal end of FtsA, which is required for the interaction between FtsA molecules, also plays a role in the localization of the protein into the septal ring (Yim et al., 2000). Although deletions that eliminate the last five carboxy-terminal residues of FtsA have no effect on the correct localization of the truncated protein, extending the deletion into W415, the sixth residue from the carboxy end, resulted in a protein that was unable to localize properly. Using immunofluorescence with specific anti-FtsA serum, we have examined the intracellular localization of FtsAΔ1C under conditions (OV16 background at 42°C) in which the amount of FtsA+ is negligible (less than 1% of the wild-type contents; Fig. 3A).
Under these conditions, rings of the FtsAΔ1C protein that did not progress into active septa were detected at the potential division sites in OV16/pARV1 (ftsAΔ1C) grown at 42°C (Fig. 3F), similarly FtsA+ localized correctly in OV16/pMSV20 (Fig. 3P), but in this case the rings proceeded normally into dividing septa. We can conclude then that deletion of domain 1C does not impair the ability of FtsA to be recruited into the Z-ring. Moreover, domain 1C properties, being required for septation and interaction but not for localization into the ring, are similar to those reported for the region comprising the last five carboxy-end residues of the protein (Yim et al., 2000). Taken together, these observations indicate that in the FtsA molecule the ability to interact with other FtsA molecules can be dissociated from the ability to localize in the division ring, and that both are required for the protein to be fully functional in septation.
Some cell division proteins fail to be recruited into the division ring in the presence of FtsAΔ1C
The imbalance between FtsZ and FtsA levels and their ability to interact with each other have been invoked to explain cell filamentation caused by high levels of FtsA+ (Dai and Lutkenhaus, 1992; Dewar et al., 1992). The expression of both ftsAΔ1C and ftsA+ at high levels (2 h of induction with 50 µM IPTG in OV16/pARV11 and OV16/pLYV30 respectively) caused filamentation. In both strains, the filaments were smooth and straight and lacked FtsZ or FtsA rings. The same morphology and absence of rings were observed when the experiment was performed in strain MC1061 (data not shown). On the contrary, when FtsAΔ1C was expressed at wild-type levels (OV16/pARV1), the FtsZ protein localized normally into Z-rings at potential division sites, as it did in the absence of FtsA (Fig. 3B and G). These results suggest that FtsAΔ1C retains the ability to interact with FtsZ, and that the localization of FtsAΔ1C in the division ring (Fig. 3F) might be directed by its interaction with FtsZ (Addinall and Lutkenhaus, 1996a; Ma et al., 1996).
In the absence of FtsA, only FtsZ (Fig. 3B) and ZipA (Fig. 3C) were recruited into rings. This has been interpreted as being a consequence of the disruption in the normal sequence of incorporation of the division proteins into the ring (Addinall and Lutkenhaus, 1996a; Hale and de Boer, 1999; Chen and Beckwith, 2001). As expected, ZipA was correctly localized into rings at potential septation sites in OV16 transformants expressing ftsAΔ1C at 42°C (Fig. 3H). Immunolocalization of other ring components showed that both FtsQ and FtsN fail to localize into rings in cells containing the truncated FtsAΔ1C protein (Fig. 3I and J respectively) as in those altogether devoid of FtsA (Fig. 3D and E). FtsQ has been reported to be incorporated into the ring after FtsA, ZipA and FtsK (Chen and Beckwith, 2001), and FtsN is a multicopy suppressor of a temperature-sensitive ftsA mutation that is the last protein being recruited into the ring (Dai et al., 1993; Addinall et al., 1997). We conclude that FtsAΔ1C, although able to be recruited itself into the division ring, fails to promote the recruitment of further components of the ring (except FtsZ and ZipA that do not depend on FtsA for recruitment), and therefore aborts the progression of the ring into the formation of a functional division septum. Consequently, as expected from the complementation results already described, FtsAΔ1C is not able to sustain cell division.
A mutant lacking the S12 and S13 β-strands of domain 2B of FtsA is weakly active in septation
The FtsA dimerization model proposed by Carettoni et al. (2003) suggests that the S12 and S13 β-strands of domain 2B might be involved in the interaction between two FtsA molecules, but contrary to domain 1C, this domain is not continuous, and probably it cannot be deleted without altering the structure of the remaining protein fragment. Therefore, to test whether these β-strands are actually involved in the biological role of FtsA, we have constructed deletion mutants that express proteins in which both of them are absent but maintain most part of domain 2B. The constructions were engineered to minimize any other alteration in the structure of the resultant protein (Fig. 1). The two chosen residues located at the beginning and the end of the deletion, Ser274 and Gln289, respectively, are close in the E. coli 3D model. The serine 274 residue is in a loop, allowing therefore the connection of the two ends of the deletion without further alterations in the structure of the remainder of the molecule. Because of the generation of an additional unsought mutation, another deletion was also obtained spanning from Val275 to Arg290 (see Experimental procedures). All the experiments were performed with the two constructions, and in all cases their behaviour was identical, therefore we will simplify their nomenclature in this report referring to them collectively as FtsAΔS12–13.
Complementation analyses of ftsA16 were conducted for ftsAΔS12–13 in a way similar to that described above for ftsAΔ1C. The mutations were introduced into pMSV20 (Sánchez et al., 1994): pARV31 (FtsAΔ275–290) and pARV32 (FtsAE273D/Δ274–289). OV16 was transformed with these two plasmids and the production of the mutated proteins was checked by Western blot analysis. In all of them, a protein of the expected molecular weight, 41.4 kDa, was produced at levels similar to those of the wild-type protein in the control strains. For complementation analysis, the cells were plated on NBT plates and incubated at 30°C or 42°C. None of them could grow at 42°C at low expression levels. It is concluded therefore that FtsAΔS12–13 cannot substitute for the wild-type FtsA protein at these levels.
The effect of expression of FtsAΔS12–13 at wild-type levels on cell morphology was analysed in the presence or absence of the wild-type FtsA (OV16 cells at 30°C and 42°C respectively). In agreement with the lack of complementation of ftsA16 by ftsAΔS12–13, the cells containing FtsAΔS12–13 but devoid of FtsA+ formed filaments (Fig. 2). However, the length of these filaments was three- to fourfold shorter than the OV16 filaments at 42°C, indicating that FtsAΔS12–13 might retain a weak activity, enough to support infrequent divisions, but not to maintain cell viability.
Even at permissive temperature, the amount of FtsA in OV16 is about 10% that of the wild type, probably because of the low efficiency of the temperature-sensitive tyrT suppressor (supFA81(Ts); Celis et al., 1973). As a result, OV16 cells at 30°C are nearly 25% longer than the wild-type MC1061 (Fig. 2). Expression of the ftsAΔS12–13 allele at the temperature permissive for the suppressor had the effect of shortening the length of the OV16 transformants containing pARV31 or pARV32 (ftsAΔS12–13) to values close to those found for the ftsA+ OV16 transformants (Fig. 2). These results are consistent with the idea that the FtsAΔS12–13 proteins retain some activity at division.
A point mutation R286W in the FtsA S13 β-strand has been recently reported to be able to bypass the requirement for the ZipA protein in cell division (Geissler et al., 2003). The possibility that the FtsAΔS12–13 proteins could equally rescue CH5/pCH32, a zipA thermonull mutant strain (Hale and de Boer, 1999), was tested. The CH5/pCH32 strain was transformed with pMSV20 or, pARV31 and pARV32, and also pLYV30 (ftsA+ expressed from a Ptac promoter) and its derivatives pARV33 (ftsAΔ275–290) and pARV34 (ftsAE273D/Δ274–289). The ability of the transformants to grow at the restrictive temperature, when ZipA is not synthetized, was checked. Neither FtsA+ nor the FtsAΔS12–13 deletions, even if overproduced from Ptac, could rescue CH5/pCH32.
The FtsAΔS12–13 proteins interact with FtsA+, localize at division rings and recruit other components into them
If, as stated above, the FtsAΔS12–13 proteins still have some cell division activity, they should retain those activities that have been proven essential for the role of FtsA in septation, among them they might be able to interact with the FtsA+ protein. The ability of the FtsAΔS12–13 proteins to interact with FtsA+ was assayed using the yeast two-hybrid system. We have found that FtsA–GAL4 fusions are often difficult to clone in E. coli, and once cloned they inhibit cell division and growth poorly, although they can be propagated in yeast. The FtsAΔS12–13 deletion could be fused to the GAL4-DNA-binding domain, but after several attempts the fusion with the GAL4 activation domain could not be obtained neither with the FtsAΔ274-289 nor with the FtsAΔ275–290 deletions, therefore the interaction could only be tested in one sense. The two deletions fused to the GAL4-DNA-binding domain were probed against the wild-type FtsA fused to the GAL4-activation domain. As with the FtsAΔ1C mutants, the expression of the fusion in yeast was confirmed by Western blotting. The two deletion mutants produced blue colonies in colony-lift filter assays that were indistinguishable from the assay with the wild-type gene fused to both GAL4 domains, indicating that the FtsAΔS12–13 protein interacts with wild-type FtsA. The self-interaction of the mutant could not be tested.
Immunofluorescence labelling of filaments formed at 42°C by OV16/pARV31 and OV16/pARV32 showed localization into septal rings of both FtsZ (Fig. 3L) and the mutant FtsAΔS12–13 proteins (Fig. 3K). Contrary to the case of OV16 filaments containing FtsAΔ1C, if the rings formed with the truncated FtsAΔS12–13 proteins retained some cell division activity, they should be proficient at localizing the rest of the cell division proteins. Immunofluorescence microscopy using sera against ZipA, FtsQ or FtsN revealed that these proteins did also localize at rings within the filaments (Fig. 3M–O). As both FtsQ and FtsN are dependent for their recruitment into the division ring on the previous localization of FtsA in the septal rings, we conclude that the truncated FtsAΔS12–13 proteins retain the activity of FtsA+ that is necessary for the recruitment of other division proteins into the septum.
Septation sites containing the FtsAΔS12–13 proteins are frequently misplaced
When OV16 transformants that harbour the deleted ftsAΔS12–13 genes under the control of Ptac in pARV33 and pARV34 were grown at 42°C in the absence of inducer, they expressed the mutant gene at about five times the wild-type levels. Under these conditions the cells did not filament. After 2 h of exponential growth at 42°C, the average length of the cells was 3.9 µm for OV16/pARV33 and 4.1 µm for OV16/pARV34, while the control strain transformed with pLYV30 (FtsA+) had an average of 4.3 µm, and the strain with the vector alone had 31.6 µm. In spite of this, the mutated proteins gave a partial complementation of OV16. Immunofluorescence analysis showed that some cells had rings in anomalous, non-central, positions and others had polar foci (Fig. 4). Non-central division septa were observed either when FtsA+ was present at normal (MC1061), low (OV16 at 30°C; Fig. 5) or very low (OV16 at 42°C; Fig. 5) levels. Even in the absence of inducer, these cultures contained a small but noticeable proportion (15% at 30°C) of short, round or almost round nucleoid-free cells. These small cells measured from 1 to 1.5 µm, which is longer than the minicells observed in min mutants (for comparison see de Boer et al., 1988; Addinall and Lutkenhaus, 1996b; Yu and Margolin, 1999), suggesting that they were not formed by polar divisions, but more likely by misplaced septa. The misplaced rings were localized in DNA-free regions, conforming to the nucleoid occlusion model for the placement of the division ring (Woldringh et al., 1990). Immunofluorescence performed with anti-FtsZ serum revealed that in OV16/pARV33 cells, grown at 42°C for 2 h, the FtsZ protein was also localized at abnormal sites (not shown).
Overproduction of the FtsAΔS12–13 truncated proteins in these strains (50 µM IPTG during 2 h) led to the loss of cell viability and resulted in the production of aberrant cell morphologies (C-shaped cells in Fig. 6) similar to those reported for the overproduction of some FtsA proteins truncated in the carboxy-terminus (Gayda et al., 1992; Yim et al., 2000) and in cell lysis.
Our results allow us to propose that two distinct regions of the cell division protein FtsA, domain 1C and the S12 and S13 β-strands of domain 2B, are associated to two functions that are essential for its role in septum formation. Domain 1C is necessary for the interaction with other FtsA molecules and the S12 and S13 strands of domain 2B affect the selection of the cell centre to form a division ring.
The role of domain 1C in the interaction between FtsA molecules and their activity at the division ring.
Carettoni et al. (2003) have proposed that the residues 126–133 in the domain 1C are involved in the homodimerization of FtsA and in its biological function. Our results using the yeast two-hybrid system support the involvement of domain 1C in the interaction between FtsA molecules, and suggests, as deduced from the absence of complementation, that the lack of self-interaction is associated to the loss of the FtsA activity in septation, failing to sustain the sequential incorporation into the division ring of proteins that are recruited at later stages.
FtsA interacts with the carboxy-end of FtsZ and this interaction is essential for the proper functioning of the division ring (Din et al., 1998; Ma and Margolin, 1999; Yan et al., 2000). Increasing the amount of FtsA or FtsZ induces filamentation, an effect that is thought to result from the imbalance between them (Dai and Lutkenhaus, 1992; Dewar et al., 1992). We have found that FtsAΔ1C localizes correctly at the potential division sites, and that the production of high levels of FtsAΔ1C provokes filamentation and loss of Z-rings, suggesting that the absence of domain 1C in FtsAΔ1C does not prevent the interaction with FtsZ and indicating that domain 1C would not be directly involved in the FtsA–FtsZ interaction.
The FtsAΔ5 protein, which lacks the last five carboxy-end residues of E. coli FtsA, did not interact with FtsA+ in the yeast two-hybrid system, had no activity in septation and was able to localize at division rings, although larger deletions did not localize (Yim et al., 2000). Together with our results, this supports the idea that the FtsA self-interaction and the correct cellular localization of the protein can be dissociated, and that both are needed for the role of this protein in cell division.
The S12 and S13 β-strands in domain 2B of FtsA are dispensable for the interaction with FtsA+ and for the completion of septa.
The analysis of strains producing the FtsAΔS12–13 proteins showed that the S12 and S13 β-strands in domain 2B are dispensable for the interaction with FtsA+. The model of Carettoni et al. (2003) postulated an interaction between FtsA molecules involving the domains 1C and 2B, and within domain 2B the S12 and S13 β-strands were proposed to interact with the residues 126–133 of domain 1C. According to our data, this last interaction would not be essential for dimerization. A possible explanation might be that even in the absence of the S12 and S13 β-strands there is still a large contact surface between domains 1C and 2B that might be enough to maintain an interaction, although the complex formed would not be fully functional. The FtsAΔS12–13 proteins colocalize with FtsZ at septation sites where they can recruit other components of the division ring, among them we have specifically detected FtsQ and FtsN (Fig. 3N and O). At 42°C, OV16 transformants that produce FtsAΔS12–13 proteins at wild-type levels (pARV31) formed filaments shorter than those of the parental strain (Fig. 2). Moreover, increasing five times the amount of mutated protein decreased the length of the cells to the wild-type length. These results indicate that the rings containing the FtsAΔS12–13 proteins as their only form of FtsA are weakly functional in septation. Further increasing the amount of mutated protein produces severe morphological alterations similar to those produced by the overexpression of FtsA carboxy-terminal deletion mutants (Gayda et al., 1992; Yim et al., 2000).
The distribution of division rings becomes altered in the presence of the FtsAΔS12–13 proteins
Our results indicated that the production of these proteins was accompanied by the presence of an unusually high proportion (15%) of short cells deprived of DNA in the population and of misplaced division rings (Fig. 4). The precise localization of a septum along the cell length seems to be determined by several factors, the absence of the MinCD inhibitor of FtsZ polymerization and the simultaneous local absence of a nucleoid being the best documented ones (Bi and Lutkenhaus, 1993; Sun et al., 1998; Yu and Margolin, 1999; Harry, 2001). Our results do not allow us to unequivocally define a causative sequence for the observed abnormal morphologies. A role for FtsA and ZipA on the stabilization of the E. coli FtsZ ring has been suggested (Hale and de Boer, 1999; RayChaudhuri, 1999; Pichoff and Lutkenhaus, 2002; Geissler et al., 2003). A point mutation in the S13 β-strand of FtsA has been reported to bypass the need for ZipA in division and to counteract the inhibition of Z-ring formation by MinC (Geissler et al., 2003). A plausible interpretation of our results would be that the putative FtsZ stabilizing activity of the FtsAΔS12–13 proteins is increased relative to that of the wild type, in this case nucleation of Z-rings could occur at abnormal locations, including places that are normally inhibited by MinC. The FtsA and FtsZ polar foci found in a fraction of the cells might be remnants of previous division rings that remain after cell division because of an increase in the stability of ring structures (Addinall and Lutkenhaus, 1996b).
These and previous results (Yim et al., 2000; Carettoni et al., 2003; Geissler et al., 2003) suggest that domain 2B of FtsA might affect Z-ring localization while domain 1C might be involved both in the FtsA–FtsA self-interaction and in the recruitment of downstream division proteins. The role of the carboxy-end in septation is less clear, because a deletion of five residues affects the self-interaction of FtsA, but not the localization at the septal ring, while longer deletions affect both. The position of the carboxy-terminal end within the 3D structure of the FtsA molecule is not known, because being flexible, it was not rendered in the crystal structure of T. maritima FtsA, and the model shown in Fig. 1 ends therefore 35 residues before the end of the protein sequence. The involvement of both, domain 1C and the carboxy-terminus, in the FtsA self-interaction supports the model of Carettoni et al. (2003), while the fact that the cap of domain 2B is not essential for the FtsA self-interaction in the yeast two-hybrid assay does not discard a role for the whole domain. The model of Carettoni et al. (2003) proposed a head-to-tail interaction between FtsA molecules to form a dimer (or even higher oligomeric or polymeric structures). Therefore, although domains 2B and 1C are involved in the FtsA self-interaction, there would be a free domain 2B, and a free domain 1C in each extreme that might be available for interaction with other division proteins.
Escherichia coli and yeast strains and growth conditions
The strains used in this study and their relevant characteristics are described in Table 1.
Luria–Bertani (LB) broth (Sambrook et al., 1989) and LB agar, supplemented with antibiotics when required (ampicillin, 100 µg ml−1, kanamycin, 50 µg ml−1 and/or chloramphenicol, 50 µg ml−1) were used for routine cultures of E. coli at 37°C. Thermosensitive strain OV16 used for complementation assays was grown in nutrient broth medium (Oxoid no. 2 nutrient broth) supplemented with thymine (50 µg ml−1) (NBT) and antibiotics when required and incubated at the permissive (30°C) or restrictive (42°C) temperatures. Complementation tests were performed as described by Sánchez et al. (1994).
Yeast strains were grown in YEPD or SD medium, supplemented with required amino acids and glucose as described in the Matchmaker GAL4 Two-Hybrid System 3 manual (Clontech Laboratories, Inc.).
Cell parameter measurements, photography and immunofluorescence microscopy
Cultures were grown at the permissive temperature in liquid media in a shaking water bath so that balanced growth was maintained for several doublings (not less than four) before the beginning of the experiment. For observing the cell morphology at different conditions, as 42°C or/and overexpression, the samples were taken after 2 h in the desired conditions. The optical density at 600 nm (OD600) was measured using a Shimadzu UV-1203 spectrophotometer or a CO8000 Cell Density Meter from WPA biowave; the optical density was always kept below 0.40–0.50 by appropriate dilution with prewarmed medium. Particles were fixed in 0.75% formaldehyde. Cells were spread on thin agar layers and photographed under phase-contrast optics using a Sensys charge-coupled device camera (Photometrics) coupled to a Zeiss Axiolab HBO 50 microscope with a 100× immersion oil lens. The software used for image capture was IPLab Spectrum, and Adobe Photoshop 7.0 software was used for processing. Cell lengths were measured with Object-Image version 1.62 (N. Vischer, University of Amsterdam). One hundred cells were analysed for each sample.
For immunofluorescence microscopy, exponentially growing cells were prepared as described by Addinall and Lutkenhaus (1996a). The cells were immobilized in poly l-lysine-pretreated slides. The final lysozyme concentration used was 4 µg ml−1, and the permeabilization time was 1 min. The samples were incubated overnight at 4°C with primary antibodies. Polyclonal anti-sera used for the FtsA, FtsZ, ZipA and FtsQ immunolocalizations were MVM1 (Sánchez et al., 1994), MVJ9 (Pla et al., 1991), MVC1 (Rueda et al., 2003) and MVJ11 (Dopazo et al., 1992) respectively. The primary antibody used for FtsN immunolocalization was MVG1, a polyclonal anti-serum obtained from rabbits (Charles River Laboratories) after their immunization with purified His-FtsN. The specificity of this anti-serum was checked by Western blot with cellular extracts of JOE565, a strain containing the chromosomic ftsN gene interruped and a plasmid containing the ftsN gene under the control of the PBAD promoter (Chen and Beckwith, 2001). All of them were previously purified by membrane affinity. Alexa594-conjugated anti-rabbit serum (Molecular Probes Inc.) was used at a 1 : 100 dilution as the secondary antibody. Cells were observed by fluorescence microscopy using the Zeiss Axiolab HBO 50 microscope fitted with a HQ: Cy3 filter (excitation, 545/30 nm; emission, 610/75 nm; beamsplitter, 565LP) and 100× immersion oil lens.
When indicated, 4′,6-diamidino-2-phenylindole (DAPI) was added to observe nucleoid segregation.
Plasmid constructions and DNA manipulation
Plasmid DNA isolation, cloning techniques and transformation procedures were performed as described by Sambrook et al. (1989). Restriction endonucleases and other enzymes were purchased from and used as recommended by Roche molecular Biochemicals. Two-hybrid system cloning vectors (pGADT7 and pGBKT7) were obtained from Clontech.
Plasmids encoding the deleted FtsA proteins for the complementation assays were obtained by polymerase chain reaction (PCR)-based method of site-directed mutagenesis as described by Weiner and Costa (1994), but using the Expand High Fidelity (EHF) and the Pwo polymerases (both supplied by Boehringer Mannheim). Plasmid pMSV20 is a pBR322 derivative containing the ftsA wild-type gene with its upstream region that was used as target for PCR-directed mutagenesis (Sánchez et al., 1994). This sequence context, working together with any potential readthrough transcription from the plasmid promoters, at the copy number of the vector produces levels of each protein that fall within the level range found in strains expressing the wild-type ftsA+ from the chromosome. Primers AR1 (5′-CGACAGCGCCAGATATACCG AAG-3′) and AR2 (5′-ACATGTCACAACGATATGGCG-3′), phosphorylated by T4 polynucleotide kinase from Promega, were used to amplify the sequence of pMSV20, except the sequence encoding the domain 1C of FtsA from Gly84 to Ile161. A clone containing the desired deletion of the entire domain 1C was obtained, although by sequencing we checked that it also contains a change in the residue after the deletion site, 162Thr to Ala. The plasmid was named pARV1, which produces a truncated FtsA lacking the entire domain 1C, named FtsAΔ1C, at levels close to the one obtained from the chromosomal ftsA+ copy. In the same way, the phosphorylated primers AR5 (5′-CTCATCTTTTCCAAC GATGG-3′) and AR6 (5′-CGTCAGACACTGGCAGAGGTG-3′) were used to amplify the sequence of pMSV20, except the sequence encoding the FtsA region comprising the β-strands S12–S13 and the connecting loop between them. Two clones, pARV31 and pARV32, encoding two nearly identical FtsA proteins lacking the two β-chains S12 and S13, have been obtained. Both contain an additional point mutation next to the deletion site checked by sequencing. As a result, we have two different forms of this deletion, one in which residues 275–290 have been deleted and another in which the deletion spans residues 274–289 and contains an additional point mutation in the residue before the deletion site, 273Glu to Asp.
These truncated ftsA genes were cloned under an IPTG-inducible promoter in pJF119EH (Fürste et al., 1986) for overexpressing the deleted proteins. The truncated genes were obtained by PCR amplification from the plasmids pARV1, pARV31 and pARV32. We used the upstream primer SR1 (5′-GGGCGAATTCATGATCAAGGCGACGGAC-3′), which introduces an EcoRI restriction site immediately upstream of the start codon, and the downstream primer LY20 (5′-CAG GTCGACCGTAATCATCGTCGGCCTC-3′), which introduces a SalI restriction site immediately downstream of the stop codon of the truncated ftsA gene. The resulting plasmids were named pARV11, pARV33 and pARV34 respectively. The control plasmid pLYV30 (Yim et al., 2000) contains the wild-type ftsA gene. The presence of all the mutations was confirmed by DNA sequencing.
Plasmids for the GAL4 two-hybrid assays were constructed as follows. The ftsA+ coding sequence was obtained from pZAQ by PCR amplification with the upstream and downstream primers MF24 (5′-AATCGCATATGATCAAGGC GACGGACA-3′) and SR2 (5′-CCCGGAATTCAAACTCTTTT CGCAGCCAA-3′), which incorporate the restriction sites NdeI and EcoRI respectively. The amplified fragment was cloned into pGADT7, yielding plasmid pARV5. Plasmid pARV6 was obtained by subcloning with EcoRI and PstI the ftsA+ coding sequence from pLYV44 (Yim et al., 2000) into the same sites in pGBKT7. The coding sequence for the domain 1C of FtsA in isolation was obtained by PCR amplification with the upstream and downstream primers AR10 (5′-GGAATTCGGTAAGCACATCAGCTGCC-3′) and AR11 (5′-CGGGATCCCGGATCAGGTGCACTTTTGCC-3′), which incorporate the restriction sites EcoRI and BamHI respectively. The amplified fragment was cloned into pGADT7 and pGBKT7, yielding plasmids pARV8 and pARV9 respectively. The ftsAΔ1C coding sequence was obtained from pARV1 by PCR amplification with the upstream primer SR1 incorporating the restriction site EcoRI, and the downstream primers AR14 (5′-CCATCGATGTTAAAACTCTTT TCGCAGCC-3′) or LY20, which incorporate the restriction sites ClaI and SalI respectively. The amplified fragment flanked by the restriction sites EcoRI and ClaI was cloned into pGADT7, yielding plasmid pARV13, and the one flanked by the restriction sites EcoRI and SalI was cloned into pGBKT7, yielding plasmid pARV14. The ftsAΔ275–290 and ftsAΔ274–289 coding sequences were obtained from pARV31 and pARV32, respectively, by using PCR amplification with the upstream and the downstream primers SR1 and LY20. The amplified fragments were cloned into the restriction sites EcoRI and SalI of vector pGBKT7, yielding plasmids pARV29 and pARV30 respectively. All these constructions were also checked by sequencing.
To construct the His-FtsN fusion protein for the obtention of anti-FtsN antibodies, the coding sequence of ftsN was amplified from pJC83 (Chen and Beckwith, 2001) by PCR with the upstream and downstream primers MG1 (5′-CGGAATTCATGGTCGCTATTGCTGCC-3′) and MG2 (5′-CCCAAGCTTTCAACCCCCGGCGGCGAG-3′), which incorporate the restriction sites EcoRI and HinDIII respectively. The amplified fragment was cloned into pET28a(+) (Novagen) yielding the plasmid pMGV1, which was checked by sequencing.
Structural modelling of E. coli FtsA
The structural model for E. coli FtsA (Carettoni et al., 2003) was made using homology modelling procedures based on the multiple alignment of the members of the FtsA family of proteins including the known 3D structure of the T. maritima FtsA (Apo form structure: Protein Data Bank accession number 1E4F; ATP bound form: 1E4G; van den Ent and Löwe, 2000) The model for E. coli FtsA residues 7–385 was built using the program Swiss-Pdb Viewer and the SWISS-MODEL server facilities (http://http:www.expasy.chswissmodSWISS-MODEL.html;Guex and Peitsch, 1997). The quality of the model was checked using the WHAT-CHECK routines from the What if program (Vriend, 1990) and the Procheck validation program from the SWISS-MODEL server facilities; briefly, the quality values of the model are within the expected region for protein structural models.
Expression and purification of His-FtsN
Cultures of BL21(DE3)pLysS containing pMGV1 were grown at 37°C in LB medium supplemented with antibiotics (kanamycin and chloramphenicol) to an OD600 of 0.4. Overexpression of the proteins was induced with 2 mM IPTG (isopropyl-β-d-thiogalactopyranoside). Two hours after addition of the inducer, cells were harvested by centrifugation and resuspended in Binding Buffer [5 mM Imidazole; 0.5 M NaCl; 20 mM Tris-HCl (pH 7.9)] supplemented with 1% Triton X100. The bacteria were lysed by sonic disruption and centrifuged for 15 min at 11 000 g at 4°C. Pellets containing the fusion protein in the pellet were resuspended in Binding Buffer supplemented with 1% Triton X-100 and 6 M Urea. The denatured His-tagged proteins recovered in the supernatant were purified by metal affinity chromatography on a nickel column (His·Bind Resin; Novagen) eluting with a step gradient of imidazole ranging from 20 to 200 mM in a 0.5 M NaCl; 1% Triton X100; 6 M Urea and 20 mM Tris-HCl (pH 7.9) buffer. Integrity and purity of proteins were checked by SDS–PAGE and quantified by the Bradford method with a commercial assay (Bio-Rad).
Yeast two-hybrid assays
Yeast strains were transformed using the Li acetate method (Gietz et al., 1992) and selected in SD medium supplemented with the required amino acids and glucose at 30°C. Qualitative β-galactosidase assays were performed using Whatman no. 5 filters as described in the Matchmaker GAL4 Two-Hybrid System 3 manual (Clontech Laboratories, Inc.). For colour development, filters were incubated up to 16 h at room temperature (although no changes were usually observed after 3 h). Cell lysis was achieved by two freeze (liquid N2) and thaw (37°C) cycles. The expression of the fusion proteins was checked by Western blotting using the affinity-purified anti-FtsA anti-serum and yeast protein extracts obtained following the protocol described in the Matchmaker manual.
We are grateful to Piet de Boer for the CH5/pCH32 ZipA depletion strain and to Jon Beckwith for the JOE565 strain and the pJC83 plasmid. We thank Pilar Palacios and Mercedes Casanova for their excellent technical assistance and Paulino Gómez-Puertas and Alfonso Valencia for helpful discussions on the E. coli FtsA 3D structure. This work was financed by Grant QLK3-2000-00079 from the European Commission, BIO2000-0451-P4-02 and BIO2001-1542 from Ministerio de Ciencia y Tecnología.