A new FtsZ-interacting protein, YlmF, complements the activity of FtsA during progression of cell division in Bacillus subtilis

Authors


*E-mail nogasawa@bs.naist.jp; Tel. (+81) 743 72 5430; Fax (+81) 743 72 5439.

Summary

The assembly of ring-like structures, composed of FtsZ proteins (i.e. the Z ring), is the earliest and most essential process in bacterial cytokinesis. It has been shown that this process is directly regulated by the FtsZ-binding proteins, FtsA, ZapA, and EzrA, in Bacillus subtilis. In this study, protein complexes that are involved in Z-ring formation were chemically cross-linked in vivo, purified by affinity chromatography, and analysed by mass spectrometry. Analysis of the results identified YlmF as a new component of the FtsZ complex. Yeast two-hybrid analysis and fluorescence microscopy of YFP–YlmF in B. subtilis cells indicated YlmF localizes to the division site in an FtsZ-dependent manner. A single disruption of YlmF resulted in a slight elongation of cells; however, simultaneous inactivation of both YlmF and FtsA showed synthetic lethality caused by complete blockage of cell division due to the defect in Z-ring formation. In contrast, the ftsA-null mutant phenotype, caused by inefficient Z-ring formation, could be complemented by overexpression of YlmF. These results suggest that YlmF has an overlapping function with FtsA in stimulating the formation of Z rings in B. subtilis.

Introduction

Cell division is a tightly regulated process that ensures accurate and efficient segregation of chromosomal material and maintains both cell size and shape. The FtsZ protein is a bacterial tubulin homologue that has the ability to polymerize and form a ring-like structure (the Z ring) at sites of cell division (Errington et al., 2003). Z-ring formation is the earliest event identified in the progression of bacterial cell division. The Z ring is required for the recruitment of other cell division proteins and serves as a framework for assembly of the cell division machinery (Errington et al., 2003; Romberg and Levin, 2003). In Escherichia coli and Bacillus subtilis, the Z ring has been shown to recruit 13 and 10 cell division proteins, respectively, to the sites of cell division directly or indirectly (Gueiros-Filho and Losick, 2002; Errington et al., 2003; Haeusser et al., 2004; Aarsman et al., 2005).

In E. coli, both FtsA and ZipA have been shown to interact directly with FtsZ and to be the first proteins recruited to the Z ring (Errington et al., 2003; Romberg and Levin, 2003). They are essential for cellular growth and recruitment of cell division proteins to cell division sites at later stages of cell division (Lutkenhaus and Donachie, 1979; Hale and de Boer, 1997; Wang et al., 1997; Pichoff and Lutkenhaus, 2002). Both proteins localize to the Z ring independently in an FtsZ-dependent manner; Z-ring formation also occurs in cells in which either protein is lacking but results in an inactive structure. However, in the absence of both proteins, FtsZ fails to form the Z ring and instead appears as spots on fluorescence microscopy at expected sites of cell division (Pichoff and Lutkenhaus, 2002). These observations suggest that FtsA and ZipA have overlapping roles in the maintenance of Z-ring integrity (Pichoff and Lutkenhaus, 2002). Moreover, a single residue change in the FtsA protein bypasses the requirement for ZipA without changing the cellular growth rate, supporting the premise that these proteins have overlapping roles (Geissler et al., 2003).

On the other hand, B. subtilis has no ZipA homologue but possesses three FtsZ-interacting proteins, FtsA, ZapA and EzrA (Errington et al., 2003). The B. subtilis FtsA protein has been shown to interact directly with FtsZ (Wang et al., 1997; Jensen et al., 2005) and to localize to sites of cell division (Feucht et al., 2001). However, in contrast to E. coli, ftsA can be disrupted in B. subtilis, even though ftsA-null mutants showed a severe deficiency in cell division (Beall and Lutkenhaus, 1992; Jensen et al., 2005). In the absence of FtsA, FtsZ appears primarily as diffuse bands at regular intervals and the proportion of normal Z-ring formation is reduced to about 10% of wild-type cells, suggesting a critical role for FtsA in Z-ring formation in B. subtilis (Jensen et al., 2005). ZapA was identified as a factor that suppressed the lethality of the overexpression of the cell division inhibitor, MinD (Gueiros-Filho and Losick, 2002). ZapA protein binds directly to FtsZ and promotes bundling of either the FtsZ monomer or FtsZ protofilaments in vitro (Gueiros-Filho and Losick, 2002). EzrA has been proposed to be as a negative regulator of Z-ring formation. Cells lacking EzrA formed Z rings at the cell poles as well as medial sites, and lowered the critical concentration of FtsZ required for ring formation (Levin et al., 1999). EzrA protein directly interacts with FtsZ to prevent its assembly in vitro, possibly by blocking subunit addition or inhibiting lateral interactions of FtsZ filaments (Haeusser et al., 2004). In addition, we have reported recently that EzrA and FtsL, an essential late-stage cell division protein, promote Z-ring dissociation and constriction at medial division sites with a partly complementary function (Kawai and Ogasawara, 2006). Although ZapA and EzrA are both dispensable for cell growth and cell division in B. subtilis, the simultaneous depletion of both proteins results in a severe block of cell division, indicative of their participation in the B. subtilis cell division process (Gueiros-Filho and Losick, 2002).

The assembly of cell division proteins at sites of cell division has been studied primarily by examining the dependence of protein localization on other cell division proteins as well as by two-hybrid analysis. In order to examine the interactions between cell division proteins directly in living cells, in the present study we developed a method to isolate protein complexes in vivo using histidine tags fused to the gene of interest and to identify the proteins in these isolated complexes by mass spectrometry. By adopting this method, we demonstrated the association of FtsZ, FtsA, ZapA and EzrA in vivo. Furthermore, we identified the YlmF protein as a new component of this complex. It was suggested that YlmF is involved in cell division because a YlmF mutant resulted in the abnormal morphology of Streptococcus pneumoniae cells (Fadda et al., 2003). Interestingly, YlmF is conserved in bacteria that have no FtsA homologue and has been demonstrated to be essential for cellular growth in Synechococcus elongatus PCC 7942 (Miyagishima et al., 2005) without FtsA homologue, suggesting a functional relationship between FtsA and YlmF. Indeed, we found that deletion of YlmF in FtsA-depleted B. subtilis cells resulted in a complete inhibition of Z-ring formation and cell division. Additionally, overexpression of YlmF in FtsA-deleted cells suppressed the division defect associated with deletion of FtsA. Thus, we demonstrated that YlmF has an overlapping role with FtsA in the formation of Z rings in B. subtilis. Based on these new findings, the mechanism of early-stage cell division will be discussed.

During preparing this manuscript, it has been suggested that YlmF (renamed SepF) is required for proper execution of septum synthesis in B. subtilis (Hamoen et al., 2005). We also discussed relevance of our findings to their results.

Results

Identification of YlmF as a component of the FtsZ complex

To detect protein–protein interactions in which FtsZ and the FtsZ-associated proteins, FtsA, ZapA and EzrA, are involved in vivo, we constructed B. subtilis strains that had histidine tags fused to the 3′ end of each gene on the chromosome. Construction was performed by integrating the pMUTinHis derivatives as shown in Fig. 1A. The resultant strains, 168ezrAHis, 168ftsZHis, 168ftsAHis and 168zapAHis, grew at the similar rate with that of parental wild-type strain, although partial impairment of the function was observed for FtsA-His as described in Experimental procedures. To examine the expression of histidine-tagged proteins in the mutant strains, Western blotting analysis was performed using anti-His-tag antibody, and the amount of FtsZ-, FtsA- and EzrA-His molecules per cell were estimated to be 4400, 970 and 6100 respectively (Fig. S1). Cellular amount of FtsZ and FtsA has been reported to be 5000 and 1000 molecules respectively (Feucht et al., 2001), indicating that His-tag fusion did not affect the expression level significantly for FtsZ and FtsA.

Figure 1.

Isolation and analysis of the protein complexes involved in Z-ring formation.
A. Schematic representation of the ftsAZ region of the FtsA-His expression strain (168ftsAHis) used for purification of the FtsA complex. Note that this strain was cultured in the presence of 1 mM IPTG to induce ftsZ expression from the spac promoter (Pspac). His12: coding sequence of 12 histidines. lacI: repressor gene for Pspac. bla: ampicillin-resistance gene (beta-lactamase). emr: erythromycin-resistance gene. ter: E. coli terminator rrnB. AC: C-terminal region of ftsA.
B. SDS-PAGE of FtsA and DanC complexes. The FtsA and DnaC complexes, purified from 30 OD600 units of 168ftsAHis and 168dnaCHis cells, respectively, were separated by 4–20% gradient SDS-PAGE and visualized by silver staining. All distinguishable bands were analysed by mass spectrometry as described in Experimental procedures. The width of gel slices used for the analysis is indicated by bars at the right of gel image, and the proteins from which more than two peptides were detected by mass spectrometry are indicated. M, protein standards; ND, no protein was detected.
C. The EzrA, FtsZ, FtsA, YlmF and ZapA complexes, purified from 4, 4, 36, 56 and 54 OD600 units of 168ezrAHis, 168ftsZHis, 168ftsAHis, 168ylmFHis and 168zapAHis cells, respectively, were separated by 4–20% gradient SDS-PAGE after heating and visualized by silver staining. Positions, corresponding to native and histidine-tagged proteins, are indicated by filled and open arrowheads on the gel respectively. M, protein standards.
D. FtsZ in the EzrA, FtsZ, FtsA, YlmF and ZapA complexes were detected by Western blotting using anti-FtsZ serum. The same amount of proteins as in panel C was analysed.

To avoid dissociation of interactions during the purification process, protein complexes were cross-linked using a membrane-permeable cross-linker, formaldehyde, as described previously, with modifications (Layh-Schmitt et al., 2000; Vasilescu et al., 2004). Protein complexes were then purified using Ni resin, followed by the passage through a Microcon-100 column to remove proteins with molecular masses lower than 100 kDa, and the high molecular complexes were separated using SDS-PAGE after heating to disrupt cross-linked proteins. Although both purity and quantity of the isolated complexes were dependent on the bait proteins, the protein complex purified using FtsA as bait showed, in addition to background bands, clear specific bands by SDS-PAGE (Fig. 1B). Thus, we used the complex for mass spectrometric analysis to identify proteins specifically involved in the complex of FtsZ and the FtsZ-associated proteins. In order to distinguish the specific protein from non-specific backgrounds, we analysed in parallel the complex isolated by using DnaC [a component of replisome, interacting with DnaI (Imai et al., 2000; Bruand et al., 2001)] as bait (Fig. 1B). All visible bands from the two complexes were subjected to in gel trypsin digestion, followed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis, and we could assign specific proteins in most of them. Several proteins abundant in cells were simultaneously identified in both samples, and they were also copurified in other protein complexes [e.g. RpoC (beta subunit of RNA polymerase), SigA (housekeeping sigma factor) and OdhB (small subunit of 2-oxoglutarate dehydrogenase (S. Ishikawa and N. Ogasawara, unpubl. data)], indicating that they were non-specific backgrounds. Thus, we concluded that FtsA, FtsZ, EzrA, YlmF and Mbl would form a complex in vivo. YlmF is a 17 kDa protein with unknown function in B. subtilis, and Mbl is a cell shape determining protein (Jones et al., 2001; Daniel and Errington, 2003). Their interaction with cell division proteins has not been reported yet.

Furthermore, specific bands corresponding to FtsA, FtsZ, EzrA and YlmF, but not to Mbl, were observed in protein complexes that were isolated using FtsZ, ZapA and EzrA as baits, although some of the bands were not clear and the confirmation by mass spectrometry has not been performed yet (Fig. 1C). Detection of possible EzrA, FtsA, FtsZ and YlmF bands in the ZapA complex, strongly suggested the existence of ZapA in the in vivo complex of early cell division proteins, although ZapA could not be detected by mass spectrometry. Additionally, we constructed the strain expressing YlmF-His (168ylmFHis) and detected the bands corresponding to EzrA, FtsA and FtsZ in the YlmF complex (Fig. 1C). Finally, Western blotting analysis using anti-FtsZ antibody showed that all complexes contained FtsZ (Fig. 1D).

In sum, our complex analysis demonstrated that FtsZ, FtsA, EzrA and ZapA could be isolated as a complex in vivo. In addition, we also identified YlmF as a new component of the complex. Identification of Mbl in the complex using FtsA as bait is intriguing, but we did not further validate the result in this study.

Direct interactions among FtsA, EzrA and ZapA have not been demonstrated. Thus, in the complex we isolated, they will bind to FtsZ filaments without contact between them. To determine the protein YlmF directly interacts within the FtsZ complex, we performed yeast two-hybrid analyses. For this purpose, we fused each protein-encoding gene to the DNA-binding and activator domains of the GAL4 gene in pGBT9 and pAD424 respectively. Interactions were assessed by expression of the HIS3 reporter gene as described in Experimental procedures (Fig. 2). In our assay, the interaction between FtsZ and EzrA, which has been demonstrated by in vitro cross-linking experiments (Haeusser et al., 2004), and the dimmer formation of FtsA, which has been shown in vitro and in vivo (Feucht et al., 2001), could not be detected. We, however, demonstrated the following known interactions: FtsZ–FtsZ, FtsZ–FtsA and FtsZ–ZapA (Wang et al., 1997; Gueiros-Filho and Losick, 2002), although the self interaction of FtsZ seemed to be quite week in our assay comparing with a previous yeast two-hybrid analysis (Wang et al., 1997). Furthermore, we found that YlmF has the ability to interact with FtsZ and to form a YlmF–YlmF dimer.

Figure 2.

Yeast two-hybrid analysis of interactions between FtsZ and FtsZ-associated proteins. Diploid strains were obtained by mating PJ69-4A-containing pGBT9 derivatives with PJ69-4a-containing pGAD424 derivatives. Interactions between proteins were detected on selection medium (SC-LWH) supplemented with 5 mM 3-AT after 5 day incubation.

FtsZ-dependent localization of YlmF at sites of cell division

To determine the localization of YlmF to sites of cell division, we fused the yfp gene, under the control of the IPTG-inducible spac promoter (Pspac), to the 5′-terminus of the ylmF gene and integrated this construct into the aprE locus of B. subtilis CRK6000. Fluorescence microscopy of YFP–YlmF-expressing cells, YK203 (aprE::Pspac-yfp-ylmF), revealed transverse bands in the middle of the cells (Fig. 3A), indicating that YlmF indeed localizes to sites of cell division. To examine the dependence of YlmF localization on FtsZ, we introduced yfp-ylmF into a temperature-sensitive mutant of ftsZ, ts1(ftsZ1) (Callister et al., 1983), in which FtsZ localization became diffused between nucleoids at the non-permissive temperature (Michie et al., 2006), and the resultant strain [YK199; ts1(ftsZ1) aprE::Pspac-yfp-ylmF] was examined at both the permissive (30°C) and non-permissive (48°C) temperatures. The results revealed that localization of YFP–YlmF, observed at sites of cell division at the permissive temperature, was dispersed in filamentous cells during the 90 min period after a shift to the non-permissive temperature (Fig. 3B). Thus, we concluded that YlmF localizes to sites of cell division in an FtsZ-dependent manner.

Figure 3.

FtsZ-dependent localization of YFP–YlmF fusion proteins.
A. Phase-contrast, DAPI-stained nucleoids and YFP fluorescence images of YK203 cells (aprE::Pspac-yfp-ylmF) from exponentially growing cultures in PAB medium with 1 mM IPTG at 37°C.
B. YK199 cells [ts1(ftsZ1) aprE::Pspac-yfp-ylmF] were grown to exponential phase (OD600 = 0.1) in LB medium supplemented with 1 mM IPTG at 30°C. Phase-contrast and YFP fluorescence images were captured at 0 and 90 min after temperature shift up from 30°C to 48°C.

Partial impairment of cell division in the ylmF-null mutant

To investigate the role of YlmF in cell division, we constructed a deletion mutant of ylmF by replacing ylmF with the spectinomycin-resistance gene to create strain YK204 (ylmF::spec). The growth rate of YK204 cells appeared normal under the growth conditions examined [minimal or PAB medium at 25, 30, 37 and 45°C (data not shown)], although the cells were slightly elongated compared with parental cells (CRK6000). The average cell length of the YK204 and parental cells was 5.4 ± 1.3 µm (average of 504 cells) and 4.1 ± 1.0 µm (average of 503 cells) respectively.

The sequence of the B. subtilis genome suggested that ylmF is the third gene in an operon whose order is ylmD, ylmE, ylmF, ylmG, ylmH and divIVA. DivIVA is responsible for the localization of the MinCD complex to the poles of the cell, to inhibit Z-ring formation there (Marston et al., 1998); its absence impairs cell division (Cha and Stewart, 1997; Edwards and Errington, 1997). To eliminate the possibility that the slight cell elongation caused by the disruption of ylmF was due to an effect on downstream gene expression, we introduced a xylose-inducible copy of ylmF, Pxyl-ylmF, into the amyE locus of YK204. When the resulting YK205 cells (ylmF::spec amyE::Pxyl-ylmF) were cultured in the absence of xylose, the average cell length was 5.4 ± 1.2 µm (average of 503 cells). In contrast, in the presence of 1% xylose, the average cell length was reduced to 4.1 ± 0.8 µm (average of 503 cells), comparable to that of wild-type cells. In agreement with this observation, Affymetrix DNA-chip analysis revealed that the expression level of ylmF from the xyl promoter became comparable with that from the native ylmF promoter after 60 min incubation in the presence of 1% xylose (data not shown). These results suggest that inactivation of YlmF indeed results in a partial impairment of cell division.

Effect of the ylmF-null mutation in cells with reduced concentrations of FtsA

In E. coli, FtsA is essential for cell division (Lutkenhaus and Donachie, 1979). In contrast, B. subtilis ftsA-null strains were viable, although they grew slowly and became extremely filamentous during vegetative growth (Beall and Lutkenhaus, 1992; Jensen et al., 2005). It was reported that disruption of ylmF results in a partial cell division defect in affected strains of S. pneumoniae (Fadda et al., 2003), belonging to Firmicutes possessing both ftsA and ylmF. Recently, a YlmF homologue present in the Cyanobacterium, S. elongatus PCC 7942, was found to be essential for cell division and viability (Miyagishima et al., 2005). Interestingly, no ftsA homologue has been identified in any Cyanobacterial genome to date, although both ylmF and ftsZ have been found in all Cyanobacteria. These results prompted us to investigate the functional relationships between YlmF and FtsA in B. subtilis.

We therefore examined the effects of the absence of YlmF on cell division under conditions where the cellular FtsA were depleted. In addition, we examined the effects of the inactivation of either ZapA or EzrA under those same conditions. These experiments were accomplished by introducing disrupted ylmF, zapA or ezrA genes into strain MD133 (ftsA::cat aprE::Pspac-ftsA) in which ftsA at native position had been replaced with promoter- and terminator-less cat gene to ensure the FtsZ expression from the native promoter and the ectopic FtsA expression is dependent on the spac promoter (M. Yoshimura and N. Ogasawara, unpublished). In agreement with the viability of ftsA-deleted strains, MD133 cells were able to grow on Luria–Bertani (LB) plate in the absence of IPTG (i.e. in the absence of FtsA), although they formed long filaments (data not shown), as shown in Fig. 4A. In contrast, the ylmF-disrupted derivative (YK200; ftsA::cat aprE::Pspac-ftsA ylmF::spec) became unable to grow in the absence of IPTG. The inactivation of zapA and ezrA did not affect growth at the same condition respectively.

Figure 4.

Effects of reduced levels of FtsA on growth of ylmF-, ezrA- and zapA-inactivated cells.
A. Growth of ylmF, ezrA and zapA deletion mutants was assayed when expression of FtsA was reduced. Control, ΔylmF, ΔezrA and ΔzapA represent MD133 (ftsA::cat aprE::Pspac-ftsA), YK200 (ftsA::cat aprE::Pspac-ftsA ylmF::spec), YK201 (ftsA::cat aprE::Pspac-ftsA ezrA::spec), and YK202 (ftsA::cat aprE::Pspac-ftsA zapA::spec) cells respectively. The strains were grown on LB supplemented with 1 mM IPTG, inoculated onto LB plates with 0.5 mM or without IPTG, and subsequently incubated overnight at 30°C before photography.
B. MD133 (control) and YK200 (ΔylmF) cells, grown to mid-exponential phase (OD600 = 0.3) in LB medium supplemented with 1 mM IPTG at 30°C, were inoculated into LB medium without IPTG to a get a final OD600 of 0.03 and were subsequently cultured at 30°C for 90 min. Phase-contrast images were captured at 0 and 90 min after the removal of IPTG.

To examine cell division in the ftsA ylmF double mutant cells, YK200 cells were first cultured in LB liquid medium containing 1 mM IPTG, transferred into medium without IPTG, and changes in cell morphology after 90 min incubation in the absence of IPTG were examined. In the presence of IPTG, the cell length of control MD133 cells and YK200 cells was comparable (Fig. 4B). In contrast, after 90 min incubation in the absence of IPTG, the YK200 cells appeared as long filaments with lysed cells, even though FtsA was not completely depleted in this condition in MD133 cells, judging from their mild cell elongation compared with that of the ftsA-depleted cells (see Fig. 6).

Figure 6.

Effects of YlmF overexpression on the ftsA-null phenotype.
A. Growth of YK207 cells (ftsA::erm Pspac-ftsZ amyE::Pxyl-ylmF) in PAB medium in the presence of various xylose concentrations at 30°C is shown. Growth of CRK6000 (wild-type) and YK206 (ftsA::erm Pspac-ftsZ) cells in PAB medium are also shown as controls.
B. Average length of 500 cells for each strain at exponential growth phase (OD600 = 0.3) is indicated by error bars. Expression levels of the ylmF mRNA were measured by quantitative RT-PCR as described in Experimental procedures and gel images of RT-PCR products in each condition are shown in the bottom.
C. Phase-contrast (a, c, e and g) and GFP-FtsZ (b, d, f and h) images of exponential phase YK209 cells (ftsA::erm Pspac-ftsZ amyE::Pxyl-ylmF aprE::gfp-ftsZ) grown in the presence (a–f) and absence (g and h) of 2% xylose, which induces YlmF. ab, cd and ef were different fields of the same preparation. Mini-cells and FtsZ-spirals are depicted by arrowheads.
D. FtsZ levels in CRK6000 (wild-type) cells cultured in PAB medium and YK207 (ftsA::erm Pspac-ftsZ amyE::Pxyl-ylmF) cells grown in PAB medium in the absence and presence of 2% xylose were analysed by Western blotting using anti-FtsZ serum. Total cellular proteins from equal amount of cells were analysed.

Although we have not as yet determined the cellular amount of FtsA in these mutants because of unavailability of appropriate antibody, our results suggested that the absence of YlmF causes severe division defect and lethality in cells lacking or producing a limited amount of FtsA.

Z-ring formation in ftsA and ylmF double mutant

It has been suggested that FtsA is critical for the efficient formation of functional Z rings (Jensen et al., 2005). Our results suggest that YlmF also acts in Z-ring formation. To investigate this possibility, we carried out immunofluorescence microscopy with antibodies against FtsZ in the ftsA ylmF double mutant (YK200, ftsA::cat aprE::Pspac-ftsA ylmF::spec). In the control strain MD133 (ftsA::cat aprE::Pspac-ftsA), after 90 min incubation in the absence of IPTG, typical Z ring (sharp band) was still observed in the cells (Fig. 5A, d–f). In contrast, in YK200 cells cultivated at the same condition, FtsZ localization became dispersed in nucleoid free regions of filamentous cells and the remaining Z rings became abnormal spiral-like structure (Fig. 5A, j–l).

Figure 5.

Subcellular localizations of FtsZ in the double mutants of ftsA and ylmF.
A. Immunolocalization of FtsZ in MD133 (ftsA::cat aprE::Pspac-ftsA; a–f) and YK200 (ftsA::cat aprE::Pspac-ftsA ylmF::spec; g–l) cells. The panels show Cy2 immunofluorescence (a, d, g and j), DAPI (b, e, h and k) and merged images (c, f, i and l), indicating Cy2 immunofluorescence in green and DAPI in red. MD133 cells and YK200 were collected for analysis at 0 (1 mM IPTG: a–c and g–i) and 90 min (no IPTG: d–f and j–l) after the removal of xylose from medium. Enlarged image of an example of a spiral-like structure, indicated by circle, is included.
B. Immunolocalization of FtsZ in YK210 cells (ftsA::erm Pspac-ftsZ ylmF::spec amyE::Pxyl-ylmF). The panels show Cy2 immunofluorescence (a, d and g), DAPI (b, e and h) and merged images (c, f and i), with Cy2 immunofluorescence in green and DAPI in red. YK210 cells were collected for analysis at 0 (a–c) and 90 min [d–i; two sets of images (d–f and g–I) of different fields in the same preparations are shown] after the removal of xylose from medium.
C. FtsZ levels in MD133, YK200 and YK210 cells, and their parental wild-type cells, 168 and CRK6000, were analysed by Western blotting using anti-FtsZ serum. Each strain was cultured in the same conditions as described in panels A or B, and total cellular proteins from equal amount of cells were analysed.

These results indicated that simultaneous absence of FtsA and YlmF indeed resulted in a severe inhibition of Z-ring formation. To further examine this possibility, we constructed a new double mutant, whose viability is maintained by inducible copy of YlmF. To this end, we constructed a new ftsA-null strain in which the entire ftsA sequence was substituted by erm (erythromycin-resistance gene) and the spac promoter ensuring expression of the downstream ftsZ gene (YK206, ftsA::erm Pspac-ftsZ). YK206 grew in the absence of IPTG because the lacI gene was not present in this strain, and thus, ftsZ is expressed constitutively. Then we introduced an inducible copy of ylmF, Pxyl-ylmF, into the amyE locus of YK206 and replaced the ylmF gene with a spectinomycin-resistance gene. The resulting strain YK210 (ylmF::spec ftsA::erm Pspac-ftsZ amyE::Pxyl-ylmF) grew in the presence of xylose (i.e. in the presence of YlmF). Furthermore, we found that the impairment of cell division due to the FtsA inactivation was partially corrected (Fig. 5B, a–c), as discussed in the next section. In contrast, when xylose was withdrawn (i.e. when YlmF was depleted), cell division was inhibited and long filaments developed. Furthermore, strong inhibition of Z-ring formation was observed in the long filaments of YK210 cells and no typical Z ring (sharp band) became detectable (Fig. 5B, d–j).

We examined FtsZ levels in MD133, YK200 and YK210 cells, and they were not affected by withdrawal of IPTG or xylose, although somewhat changed compared with parental wild-type cells (Fig. 5C). These observations strongly suggest that monomers or protofilaments are not organized into the functional Z-ring structure at division sites in the absence of both FtsA and YlmF in B. subtilis cells, resulting in complete inhibition of cell division. Similar FtsZ localization pattern has been observed in E. coli cells in the absence of both ZipA and FtsA, which have overlapping roles in the formation and stabilization of the Z ring (Pichoff and Lutkenhaus, 2002).

Complementation of the ftsA-null phenotype by overexpression of YlmF

Complete inhibition of Z-ring formation by simultaneous inactivation of FtsA and YlmF suggests their overlapping roles in the progression of cell division. Furthermore, the filamentous phenotype due to the ftsA inactivation was apparently suppressed by the overexpression of YlmF in YK210 cells (Fig. 5B). Therefore, we further examined the ability of YlmF to complement impairment of cell growth and division in the ftsA-null mutant. As observed in the ftsA-null mutants, YK206 cells (ftsA::erm Pspac-ftsZ), grew at half the parental rate (increase in OD600, Fig. 6A), were approximately three times the length of parental cells (Fig. 6B), and were more frequently lysed in the absence of xylose (data not shown). The phenotype of YK207 cells (ftsA::erm Pspac-ftsZ amyE::Pxyl-ylmF), having additional xylose-inducible copy of ylmF at the amyE locus, was similar to that of parental YK206 cells when grown in the absence of xylose (Fig. 6A and B). Intriguingly, the deficiencies of growth rate and division caused by the disruption of FtsA were corrected depending on the xylose concentration in the growth medium, i.e. depending on the level of additional expression of YlmF from Pxyl-ylmF (Fig. 6A and B). In addition, a lysis of cells, which resulted in decreased growth rate, at least partly, in the ftsA-null strain (Jensen et al., 2005) was significantly suppressed by the overproduction of YlmF (data not shown). In the presence of 2% xylose, both growth rate and cell length of YK207 cells (ftsA::erm Pspac-ftsZ amyE::Pxyl-ylmF) resembled those of wild-type cells. FtsZ level were not significantly affected in YK207 cells in the absence and the presence of xylose as shown in Fig. 6D. These results demonstrated clearly that the defect in cell division caused by the ftsA inactivation was compensated with the YlmF overproduction. Consistent with this complementation, the reduced efficiency of Z-ring formation in the absence of FtsA was corrected by the overproduction of YlmF (Fig. 6C).

We observed anomalous, higher-order structures of FtsZ that look like spirals in the middle of cells in which YlmF is overexpressed in the absence of FtsA (Fig. 6C). Mini-cell formation at the poles was also observed. In the strain YK209 (ftsA::erm Pspac-ftsZ amyE::Pxyl-ylmF aprE::PftsAZ-gfp-ftsZ), an additional copy of ftsZ at the aprE locus expressed GFP-fusion proteins to approximately the same level as native proteins, as assessed by Western blotting (data not shown). However, the same amount of GFP-FtsZ did not result in the same phenotype in a wild-type, YlmF-overexpressing, or ftsA-null (ftsA::erm Pspac-ftsZ aprE::PftsAZ-gfp-ftsZ) background (data not shown). Thus, although GFP-FtsZ may contribute to mini-cell formation to a certain extent, the majority of this minor but significant phenotype is caused by the combined factors of the absence of FtsA and the overexpression of YlmF, resulting in the stabilization and/or inhibition of proper Z-ring disassembly.

Discussion

Large-scale analysis of protein complexes in E. coli, through the isolation of complexes using tandem affinity purification or sequential peptide affinity tags followed by mass spectrometry, has allowed the identification of many protein–protein interactions but has failed to detect the known interaction between FtsZ and FtsA. This failure may be because detectable interactions have been limited to the identification of relatively stable complexes (Butland et al., 2005). In the present study, in order to prevent the dissociation of in vivo complexes during purification, we applied a cross-linking method and successfully demonstrated that EzrA and ZapA are indeed involved in a protein complex with FtsZ in vivo, as well as with FtsA, whose in vivo interaction with FtsZ was previously described (Jensen et al., 2005). More importantly, we identified YlmF as a new component of the FtsZ complex. Thus, the procedure presented here not only provided important clues to understand the mechanism of Z-ring formation but also provides a useful strategy for finding new factors present in unstable complexes.

Functional analysis of the B. subtilis YlmF protein in this study demonstrated that YlmF is indeed one of the cell division proteins that localizes to sites of cell division probably through its direct interaction with FtsZ. Furthermore, we demonstrated synthetic lethality after simultaneous depletion of both YlmF and FtsA due to complete inhibition of Z-ring formation and thus cell division, and the complementation of a ftsA-null phenotype by the overexpression of YlmF. Comparative analysis of complete, available, bacterial genome sequences using the Microbial Genome Database (Uchiyama, 2003) showed that YlmF is conserved in Firmicutes (i.e. low G+C content, Gram-positive bacteria, including B. subtilis), Symbiobacteria, Fusobacteria, Actinobacteria (i.e. high G+C content, Gram-positive bacteria), and Cyanobacteria. On the other hand, FtsA is conserved in Firmicutes, Symbiobacteria, Proteobacteria and some additional phyla. Mollicutes, commonly called mycoplasmas in a class of Firmicutes, is exceptional as some of them lack even FtsZ as well as FtsA and/or YlmF. Interestingly, YlmF was recently shown to be an essential gene in the Cyanobacterium, S. elongatus PCC 7942 (Miyagishima et al., 2005), which lacks FtsA. Conversely, FtsA is an essential gene in both E. coli and Caulobacter crescentus (Lutkenhaus and Donachie, 1979; Sackett et al., 1998), Proteobacteria that lack YlmF. These results are consistent with our finding that both FtsA and YlmF have overlapping roles in B. subtilis cell division and that expression of either protein is essential for B. subtilis cell division and growth. Thus, it also suggested that YlmF would play a critical role in cell division in FtsA-lacking bacteria, as FtsA does in YlmF-lacking bacteria, and that both FtsA and YlmF would work additively with FtsZ in bacteria that possess both proteins. In B. subtilis, the disruption of ylmF partially impairs cell division without affecting growth rate. Meanwhile, the ftsA-null mutation causes a severe defect in cell division, suggesting that FtsA contributes more strongly to cell division than does YlmF. Although unexpected, it has been reported that disruption of ylmF caused a partial cell division defect in affected strains (Fadda et al., 2003) but ftsA remained an essential gene in S. pneumoniae (Lara et al., 2005). However, the absolute requirement for FtsA was determined by its unsuccessful gene replacement within the cat cassette by a double-crossing over event (Lara et al., 2005). As ftsA is present in an operon with ftsZ in S. pneumoniae, we speculated that the replacement of ftsA with cat would result in a synthetic lethal effect, both because of the absence of FtsA and the consequently changed expression levels of FtsZ. This effect was in fact observed in B. subtilis when ftsA::cat from MD133 (ftsA::cat aprE:: Pspac-ftsA) was introduced into wild-type cells.

In E. coli, ZipA was shown to have overlapping functions with FtsA for the formation and/or stabilization of the Z ring, similar to the role for YlmF in B. subtilis reported here, although both FtsA and ZipA are essential for E. coli growth. In spite of the similar functions of FtsA, ZipA and YlmF, their masses (molecular weights of YlmF, ZipA and FtsA are 17, 37 and 48 kDa respectively) and overall amino acid sequences are quite different. How then can such different proteins promote Z-ring formation? The cross-linking of FtsZ protofilaments by ZapA dimer or tetramer bridges has been proposed as a possible mechanism based on in vitro studies (Gueiros-Filho and Losick, 2002; Low et al., 2004). As for FtsA, a proposed mechanism has not yet been directly demonstrated, but evidence in E. coli suggests that destabilization of either FtsA–FtsZ or FtsA–FtsA interactions results in severe defects in cell division (Ma and Margolin, 1999; Yim et al., 2000; Rico et al., 2004). In agreement with these observations, we found that YlmF interacted with both FtsZ and itself. Thus, YlmF may also be able to promote Z-ring formation by cross-linking FtsZ protofilaments.

The highly conserved C-terminal region of FtsZ has been determined to be essential for interactions with both FtsA and ZipA but not for localization or multimerization of FtsZ at sites of cell division (Wang et al., 1997; Ma and Margolin, 1999). Furthermore, the crystal structure of the ZipA–FtsZ C-terminal peptide (amino acids 367–383) complex was solved and showed that Ile374, Phe377 and Leu378 were all important amino acids for binding. This complex forms by hydrophobic contacts that are independent of amino acid side-chain specificity, implying that there are less steric constraints on binding partners (Mosyak et al., 2000). Interestingly, this FtsZ peptide is highly conserved in both Actinobacteria and Cyanobacteria, which have no FtsA homologue. Thus, these facts suggest that this peptide may be responsible for binding to YlmF in certain bacteria, including B. subtilis.

It should be noted that when YlmF was overexpressed in a ftsA-null background, not only was the reduced efficiency of Z-ring formation corrected but cell division also recovered to normal levels, indicating that an excess amount of YlmF could compensate for both the recruitment of cell division proteins during later stages of cell division and the constriction of Z rings. Thus, our results may suggest that some structure in the Z ring promoted by FtsA and/or YlmF recruits downstream cell division proteins, but that neither FtsA nor YlmF recruit them through direct interaction. On the other hand, FtsA has been shown to be essential for the recruitment of downstream cell division proteins; there are evidences for a direct interaction of FtsA with FtsI in E. coli (Di Lallo et al., 2003; Corbin et al., 2004; Karimova et al., 2005). Thus, it is possible that B. subtilis FtsA interacts with PbpB, counterpart of E. coli FtsI, and if so, this might explain how FtsA contributes more strongly to cell division than does YlmF. Although interactions between FtsA and downstream proteins including PbpB could not be specifically detected in our study, this fact does not preclude the presence of direct interactions because purification of membrane-spanning proteins and detection of small proteins by mass spectrometry would be technically difficult. It is also possible that the use of FtsA-His, which is not fully functional in cell division, may be the reason why the downstream cell division proteins could not be detected in the complex.

During preparing this manuscript, Hamoen et al. (2005) reported that YlmF (SepF) acts in cell division. They showed that YlmF binds to FtsZ and itself, YlmF localizes to division sites depending on FtsZ, and the disruption results in partial defect in cell division, as reported in this study. They further demonstrated that combination of a ylmF and ezrA mutation resulted in a synthetic-lethal division defect, although Z rings and a late assembling division protein, Pbp2B, still assembled in the filamentous cells. Therefore, Hamoen et al. proposed that YlmF is not involved in Z-ring formation and divisome assembly, but is active in a later step of cell division process; the constriction of the divisome and/or synthesis of the septum wall. However, similar synergistic effect has been reported when EzrA and ZapA are simultaneously underexpressed, even though ZapA acts in an early step of cell division by promoting the assembly of FtsZ (Gueiros-Filho and Losick, 2002). Moreover, we recently reported that expression of YneA, which suppresses cell division during the SOS response, and disruption of noc, which acts as an effector of nucleoid occlusion, in ezrA mutant cells resulted in accumulation of multiple non-constricting Z rings and lethal division defect (Kawai and Ogasawara, 2006). EzrA has been proposed as a negative regulator of Z-ring formation, and the inactivation stabilizes Z ring and it leads to the limiting amount of FtsL (Levin et al., 2001; Kawai and Ogasawara, 2006), which is essential for the complete assembly of late division complex and the following constriction process (Daniel et al., 1998). Interestingly, FtsL overproduction could suppress the synthetic defect caused by the expression of YneA and disruption of noc or zapA in ezrA background (Kawai and Ogasawara, 2006). The synthetic lethality of simultaneous deletion of ylmF and ezrA was also suppressed by the FtsL overproduction (data not shown).

As demonstrated in this study, B. subtilis contains two proteins, FtsA and YlmF, which differ in both size and homology but can independently promote the same early-stage processes in cell division. At this point, B. subtilis is a unique bacterial model that will be useful for further studies of bacterial cytokinesis. Additional functional analyses of both proteins will provide a better understanding of early-stage cytokinetic events.

Experimental procedures

Bacterial strains and cell growth

The B. subtilis strains used in this study are listed in Table 1. B. subtilis 168, CRK6000, and their derivatives were grown in either LB medium or antibiotic medium 3 (PAB; Difco Laboratories), supplemented with adenine and guanosine (20 µg ml−1 each) at 37°C, unless otherwise stated. When appropriate, erythromycin, chloramphenicol, tetracycline and spectinomycin were added to the medium at final concentrations of 0.5, 5, 10 and 100 µg ml−1 respectively. The E. coli strains DH5α, C600, and XL10-Gold (Stratagene) were used for plasmid construction and propagation.

Table 1. B. subtilis strains and mutants used in this study.
StrainRelevant genotypeaSource or referenceb
  • a. 

    Antibiotic resistance genes are expressed as follows: cat, chloramphenicol; kan, kanamycin; spec, spectinomycin; tet, tetracycline; emr, erythromycin.

  • b. 

    Arrows indicate construction by transformation using plasmid DNA, chromosomal DNA, or PCR products.

168trpC2Pasteur stock (Vagner et al., 1998)
168ftsAHis168 ftsA::pMUTinHisΔftsApMUTinHisΔftsA → 168
168ftsZHis168 ftsZ::pMUTinHisΔftsZpMUTinHisΔftsZ → 168
168zapAHis168 zapA::pMUTinHisΔzapApMUTinHisΔzapA → 168
168ezrAHis168 ezrA::pMUTinHisΔezrApMUTinHisΔezrA → 168
168ylmFHis168 ylmF::pMUTinHisΔylmFpMUTinHisΔylmF → 168
168dnaCHis168 dnaC::pMUTinHisΔdnaCpMUTinHisΔdnaC → 168
ASK600168 aprE::PftsAZ-gfp-ftsZ (spec)Kei Asai (unpublished)
ts1168 ts1(ftsZ1)Callister et al. (1983)
YK199168 ts1(ftsZ1) aprE::Pspac-yfp-ylmF (spec)YK203 → ts1
MD133168 ftsA::cat aprE::Pspac-ftsA (kan)M. Yoshimura and N. Ogasawara (unpublished)
YK200168 ftsA::cat aprE::Pspac-ftsA (kan) ylmF::specYK204 → MD133
YK201168 ftsA::cat aprE::Pspac-ftsA (kan) ezrA::specYK012 → MD133
YK202168 ftsA::cat aprE:: Pspac-ftsA (kan) zapA::specYK013 → MD133
CRK6000purA16 metB5 hisA3 guaBLaboratory stock (Moriya et al., 1990)
YK012CRK6000ezrA::specKawai and Ogasawara (2006)
YK013CRK6000zapA::specKawai and Ogasawara (2006)
YK203CRK6000aprE::Pspac-yfp-ylmF(spec)pAPNC213-yfp-ylmF → CRK6000
YK204CRK6000ylmF::specPCR product → CRK6000
YK205CRK6000ylmF::spec amyE::Pxyl-ylmF (tet)pXTylmF → YK204
YK206CRK6000ftsA::erm Pspac-ftsZPCR product → CRK6000
YK207CRK6000ftsA::erm Pspac-ftsZ amyE::Pxyl-ylmF (tet)pXTylmF → YK206
YK208CRK6000aprE::PftsAZ-gfp-ftsZ (spec)ASK600 → CRK6000
YK209CRK6000ftsA::erm Pspac-ftsZ amyE::Pxyl-ylmF (tet) aprE::PftsAZ-gfp-ftsZ (spec)ASK600 → YK207
YK210CRK6000ftsA::erm Pspac-ftsZ amyE::Pxyl-ylmF (tet) ylmF::specYK204 → YK207

Plasmid construction

All primers used for PCR are listed in Table S1. Structures of plasmids and strains constructed in this study were examined by DNA sequencing to avoid PCR mutations.

The pMUTinHis plasmid was employed in the fusion of a histidine tag to the 3′-terminus of target genes. Fragment that encoded 12 histidines with cloning sites at both ends was amplified by PCR using 2HC-F and 2H-3Stop primers from pSI2HC (S. Ishikawa and N. Ogasawara, unpublished) and then inserted between the SalI and SacI sites of pMutinNC (Morimoto et al., 2002). To construct pMUTinHis-ΔftsA, -ΔftsZ, -ΔzapA, -ΔezrA, -ΔylmF and -ΔdnaC, 260–500 bp fragments, encoding the 3′ portion of each gene except for the stop codon, were amplified by PCR using the following primer sets: ftsA.f–ftsA.r, ftsZ.f–ftsZ.r, zapA.f–zapA.r, ezrA.f–ezrA.r, ylmF.f–ylmF.r and dnaC.f–dnaC.r, respectively, from B. subtilis 168 genomic DNA and were subsequently cloned in the HindIII or EcoRI and XhoI sites of pMUTinHis.

Plasmid manipulation of the B. subtilis genes was accomplished using the Gateway cloning technology, as per the manufacturer’s instructions (Invitrogen). The coding sequences were amplified from the wild-type 168 chromosome by PCR using primers containing the attB sequence at both ends and were cloned into the pDONR201 plasmid by BP clonase to create entry clones. Entry clones were used to construct expression clones through recombination reactions between the entry clone and the destination vector.

To attach either the entire attB1 or attB2 sequence at either end of the PCR product, we used two-step adaptor PCR. For the first step, PCR was performed for six cycles using specific forward and reverse primers that contained half the sequence of attB1 (5′-AAAAAGCAGGCTCG-3′) or attB2 (5′-AGAAAGCTGGGTC-3′) at the 5′ end respectively. Then, the first reaction mixture was used as a template for the second reaction using adaptor-attB1 and adaptor-attB2 primers that contained the entire attB1 and attB2 sequences respectively. In this study, ftsA, ftsZ, zapA, ezrA and ylmF sequences were amplified using the following primer sets: ftsAgwF–ftsAgwR, ftsZgwF–ftsZgwR, zapAgwF–zapAgwR, ezrAgwF–ezrAgwR and ylmFgwF–ylmFgwR respectively, and were subsequently cloned into pDONR201 (Invitrogen) to create entry clones (i.e. pENTR-ftsA, -ftsZ, -zapA, -ezrA and -ylmF).

The pGBT9 and pGAD424 plasmids for yeast two-hybrid experiments were obtained from Clontech. To convert these plasmids to Gateway system destination vectors, pGBT9 and pGAD424 were doubly digested with EcoRI and BamHI, Mung Bean nuclease was added to create blunt ends (Takara Bio), and the plasmids were ligated into the rfC cassette (Invitrogen). The resulting destination vectors were designated pDBGW and pADGW respectively. The pGBT9 derivatives (i.e. pGBT9-ftsA, -ftsZ, -zapA, -ezrA and -ylmF) and the pGAD424 derivatives (i.e. pGAD424-ftsA, -ftsZ, -zapA, -ezrA and -ylmF) were constructed by recombination reactions between the entry clones (i.e. pENTR-ftsA, -ftsZ, -zapA, -ezrA and -ylmF) and the destination vectors, pDBGW and pADGW.

To construct the destination vector for the integration of the IPTG-regulated yfp-fusion gene at the aprE locus on the B. subtilis chromosome, the yfp gene was amplified from pMyfp (Lindow et al., 2002) using primers fp-SD-HBg-F and fp-rfA-R [fp-rfA-R has a region that is homologous to the attR1 sequence of the rfA cassette (Invitrogen) at its 5′ end]. The yfp-rfA fragment was created by recombinant PCR using both the amplified yfp fragment and the rfA cassette as templates, and using primers fp-HBg-F and attR2-EVBg-R (anneals to the attR2 sequence). The resulting PCR product was digested with BglII and EcoRV and subsequently cloned into the BamHI and SmaI sites of pAPNC213 (Morimoto et al., 2002). The resulting destination vector was designated pAPNC213-yfp-GW. The plasmid pAPNC213-yfp-ylmF was constructed by a recombination reaction between pENTR-ylmF and pAPNC213-yfp-GW. To construct pXT, the tet cassette, which was amplified from pBEST307 (Itaya, 1992) using the tetF2-tetR2 primer set (Kawai et al., 2003), was cloned into the EcoRV site of pX, an integration vector, into the amyE locus for xylose-inducible gene expression (Kim et al., 1996). The ylmF fragment, containing both its ribosome binding site and stop codon, was amplified using SDylmF-SpeI and ylmF-BR primers, digested with SpeI and BglII, and cloned into the SpeI and BamHI sites in pXT. The resulting plasmid (pXTylmF) was integrated into the amyE locus of the B. subtilis chromosome by two crossing-over events to create a mutant in which ylmF expression could be induced by xylose.

Construction of strains expressing histidine-tagged fusions

Bacillus subtilis strains that had histidine tags fused to the 3′ end of each gene on the chromosome was constructed by integrating the pMUTinHis derivatives (Fig. 1A and Table 1). Strain 168ftsZHis grew normally, but 168ftsAHis cells become elongated, in the presence of 1 mM IPTG to ensure downstream ftsZ expression, although growth rate was not affected, indicating that FtsA-His was not fully functional. Growth rate of 168ezrAHis and 168zapAHis was also normal. In addition, null mutation of ezrA (ezrA::spec) could be introduced into strain 168zapAHis and zapA::spec into 168ezrAHis, although ezrA zapA double disruption resulted in synthetic defect in cell division (Gueiros-Filho and Losick, 2002), indicating that the both fusions were functional. Synthetic lethality of the simultaneous deletion of ylmF and ezrA has also been reported (Hamoen et al., 2005) and null mutation of ezrA (ezrA::spec) could not be introduced into strain 168ylmFHis, suggesting that YlmF-His was not fully functional, although growth rate of 168ylmFHis cells was normal.

Construction of ylmF and ftsA deletion mutants

To construct the ylmF deletion mutant, the 5′ upstream and 3′ downstream regions of ylmF were amplified from B. subtilis 168 chromosomal DNA using primer sets ylmDHF–ylmFdel1R (complementary rPCR1 sequence, GGATA GACTCCACCAGAAGAGC, at its 3′ end) and ylmFdel2F (with the rPCR2 sequence, CGGAAGGATACTACATCCTGG, at its 3′ end)–ylmHXR respectively. The spec gene with the promoter region was amplified from pJL62 (LeDeaux and Grossman, 1995) using rPCR–SpecF2 (with the rPCR1 sequence at its 3′ end) and rPCR–SpecR2 (complementary rPCR2 sequence at its 3′ end). The resulting three fragments were joined by recombinant PCR using the ylmDHF–ylmHXR primer set and integrated into the B. subtilis chromosome by two crossing-over events.

To construct the ftsA-null mutant, the upstream and downstream regions of ftsA were amplified from B. subtilis 168 chromosomal DNA using primer sets fAd1F–fAd1R and fAd2F–fAd2R respectively, and a 1.9 kb sequence, containing both the erm gene and Pspac, was amplified from pMutinNC (Morimoto et al., 2002) using spacFS (complementary fAd1R sequence at its 5′ end) and spacR (complementary fAd2F sequence at its 5′ end). The resulting three fragments were joined by recombinant PCR using the fAd1F–fAd2R primer set and integrated into the B. subtilis chromosome by two crossing-over events.

Purification of protein complexes

An overnight liquid culture of B. subtilis cells expressing histidine-tagged protein in LB medium containing 0.5 µg ml−1 erythromycin and 1 mM IPTG at 30°C was inoculated into 200 ml of the same medium at an initial OD600 of 0.01. When the cells, grown at 37°C, reached an OD600 of 0.5–0.6, the culture was treated with formaldehyde (0.4% final concentration) for 30 min. Tris-HCl (2 M, pH 9.0) was added at a final concentration of 20 mM and the culture was incubated for an additional 10 min. Cells were washed twice with 20 mM Tris-HCl (pH 7.5) and stored at −80°C. Collected cells were disrupted by sonication on ice in U buffer (100 mM Hepes, 50 mM imidazole, 8 M urea, 0.5 M NaCl, 1 mM DTT and 1 mM PMSF, pH 7.4). After centrifugation at 8000 rpm for 10 min, MagneHis beads (Promega) were added to the supernatant, followed by incubation at room temperature for 2 h with gentle shaking. MagneHis beads were washed three times with U buffer and bound proteins were eluted with elution buffer [100 mM Tris-HCl (pH 7.5), 0.5 M imidazole, 1% SDS, and 1 mM DTT]. To remove non-specifically bound, uncross-linked proteins, the eluate was passed through a Microcon-100 column (Millipore) to remove proteins with molecular masses lower than 100 kDa. Protein complexes that were retained on the membrane were washed twice with elution buffer and recovered by addition of 20 µl of the same buffer. Cross-linked proteins were dissociated by heating overnight at 65°C. To identify proteins in the complex by mass spectrometry, eluates were separated by SDS-PAGE after heating; separated proteins were visualized by Coomassie brilliant blue R-250 staining. In-gel enzymatic digestion, peptide analysis, and protein identification by LC-MS/MS were carried out as described by Kuwana et al. (2002).

Fluorescence microscopy

Live cells were examined on agarose slides (Price and Losick, 1999). Cell morphology and nucleoid distribution were examined by fluorescence microscopy after DAPI (4′,6-diamino-2phenyl indole) staining, as described previously (Hassan et al., 1997). Fluorescence was detected using fluorescence microscopy (DMRE-HC, Leica), a cooled digital CCD camera (1300Y, Roper Scientific), and a YFP filter set (Omega). Images were uploaded and analysed to determine the intracellular locations of the signals using MetaMorph software (Universal Imaging). Immunofluorescence microscopy was performed as by Kawai et al. (2003).

Yeast two-hybrid analysis

Yeast strains, PJ69-4 A (MATa, trp1-901, leu2-3, 112ura3-52, his-200, gal4Δ, gal80Δ, LYS2::GAL1-HIS3, GAL2-ADE2, met2::GAL7-lacZ) and PJ69-4α (MATα, trp1-901, leu2-3, 112ura3-52, his-200, gal4Δ, gal80Δ, LYS2::GAL1-HIS3, GAL2-ADE2, met2::GAL7-lacZ) were kindly provided by Philip James (James et al., 1996).

The pGBT9 derivative (TRP1) and the pGAD424 derivative (LEU2) were used to transform PJ69-4A and PJ69-4α respectively, using the method described by Bongiorni et al. (2005) with minor modifications. Transformants were mated in the appropriate liquid medium using flat-bottomed, 96 well plates. After mating, the cultures were collected, washed with sterilized water, and spotted onto synthetic complete (SC) agar plates that lacked leucine and tryptophan (SC-LW) for selection of LEU2 and TRP1 diploid cells. Selected cells were cultured in liquid SC-LW for 1 day and were then replica-plated on selection agar medium that lacked histidine (SC-LEH) and contained 5 mM 3-aminotrizaole (3-AT) to inhibit auto-activation of the HIS3-reporter gene.

Quantitative RT-PCR

Total cellular RNA was extracted from B. subtilis cells as described previously (Kawai et al., 2003). First-strand cDNA was synthesized from 15 µg of RNA using the SuperScript III reverse transcriptase and random primers following the manufacturer’s instructions (Invitrogen). In the second PCR reaction, a 108 bp internal sequence of ylmF was amplified by iProof High-Fidelity DNA polymerase (BIO-RAD) using ylmFmeasureF and ylmFmeasureR primers, which were designed by Primer3 software (Rozen and Skaletsky, 2000). Various concentrations of the cDNA products were used as templates to determine the appropriate condition for quantitative amplification. After the optimization, quantitative RT-PCR was performed 30 cycles using 2 µl of 3.2 × 104 times diluted first cDNA reaction mixture as template in 20 µl of total volume, and then 5 µl of the products was separated in a 2% agarose gel, followed by ethidium bromide staining. Gel images were photographed on a UV illuminator and the amount of DNA was quantified by NIH Image (http://rsb.info.nih.gov/nih-image/).

Acknowledgements

We thank Mika Yoshimura and Kei Asai for generous gifts of mutant strains, Taku Oshima for valuable discussions. This work was supported by a Grant-in-Aid for Scientific Research on Priority Areas, Life System Informatics, from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

Ancillary