A novel component of the division-site selection system of Bacillus subtilis and a new mode of action for the division inhibitor MinCD


*E-mail jeff.errington@ncl.ac.uk; Tel. (+44) 191 222 8126; Fax (+44) 191 222 7424;

**E-mail marc.bramkamp@uni-koeln.de; Tel. (+49) 221 4706472; Fax (+49) 221 4705091.


Cell division in bacteria is governed by a complex cytokinetic machinery in which the key player is a tubulin homologue, FtsZ. Most rod-shaped bacteria divide precisely at mid-cell between segregated sister chromosomes. Selection of the correct site for cell division is thought to be determined by two negative regulatory systems: the nucleoid occlusion system, which prevents division in the vicinity of the chromosomes, and the Min system, which prevents inappropriate division at the cell poles. In Bacillus subtilis recruitment of the division inhibitor MinCD to cell poles depends on DivIVA, and these proteins were thought to be sufficient for Min function. We have now identified a novel component of the division-site selection system, MinJ, which bridges DivIVA and MinD. minJ mutants are impaired in division because MinCD activity is no longer restricted to cell poles. Although MinCD was thought to act specifically on FtsZ assembly, analysis of minJ and divIVA mutants showed that their block in division occurs downstream of FtsZ. The results support a model in which the main function of the Min system lies in allowing only a single round of division per cell cycle, and that MinCD acts at multiple levels to prevent inappropriate division.


Bacterial cell division is governed by a cytokinetic ring made up of more than a dozen largely conserved proteins (Errington et al., 2003; Harry et al., 2006). The central player is a tubulin homologue, FtsZ, which polymerizes into linear oligomers that assemble into a higher-order ring structure, the Z ring, to which other division proteins are recruited. The dynamics of Z ring assembly are regulated by GTP binding and hydrolysis, just as for tubulin, and by a number of FtsZ-interacting proteins. In Bacillus subtilis, FtsA (Jensen et al., 2005; Pichoff and Lutkenhaus, 2005), EzrA (Levin et al., 1999; Haeusser et al., 2004), ZapA (Gueiros-Filho and Losick, 2002) and SepF (Hamoen et al., 2006; Ishikawa et al., 2006) have all been reported to associate directly with the Z ring. However, much remains to be discovered about the molecular details of these interactions and their functions. Some time after the assembly of this ‘inner ring’ of mainly cytoplasmic proteins, a set of transmembrane proteins mostly with major domains outside the cell (the ‘outer ring’ proteins) are recruited, including FtsL (Sievers and Errington, 2000), DivIB (Harry and Wake, 1997), DivIC (Katis et al., 1997), FtsW (Wang et al., 1998) and PBP2B (Scheffers et al., 2004). The functions of these proteins are also poorly understood, with the exception of PBP2B, which is a penicillin-binding protein essential for the synthesis of cell wall peptidoglycan in the division septum.

In rod-shaped bacteria like B. subtilis cell division is accurately targeted to the mid-point of the long axis of the cell by the combined action of two negative regulatory systems. The first system, nucleoid occlusion, depends on a protein called Noc, which is a DNA-binding protein that associates with sites widely distributed over the chromosome. Noc is also an inhibitor of Z ring assembly in the vicinity of the nucleoid (Wu and Errington, 2004). Escherichia coli has an unrelated but functionally equivalent protein called SlmA (Bernhardt and de Boer, 2005). Although the nucleoid occlusion system ensures that division does not cut through the nucleoid it leaves open the possibility of division occurring inappropriately in the DNA-free regions at the cell poles. Inappropriate polar division is prevented by the action of the MinCD division inhibitor, which is widely conserved throughout bacteria and even in some eukaryotic organelles (Szeto et al., 2002; Aldridge et al., 2005). In the absence of minC or minD cells divide incorrectly close to the cell pole generating anucleate minicells. In vivo evidence, mainly from overexpression studies, shows that MinC is the actual inhibitor of division and that it acts by interfering with FtsZ assembly (Bi and Lutkenhaus, 1993; Hu et al., 1999; Justice et al., 2000; Shiomi and Margolin, 2007). Indeed, recent work has shown that the N- and C-terminal domains of MinC inhibit FtsZ assembly via at least two distinct in vitro activities (Hu and Lutkenhaus, 2000; 2003; Shiomi and Margolin, 2007). MinD seems to stimulate MinC activity, by concentrating it close to the membrane, where FtsZ ring assembly occurs, and it is also used to target the division inhibitor to cell poles.

Although MinCD is highly conserved, targeting its activity to cell poles is achieved in quite different ways in E. coli and B. subtilis. In E. coli, the MinE protein drives a remarkable pole-to-pole oscillation of MinCD by binding to MinD and stimulating ATP hydrolysis. The net effect of this oscillation is that FtsZ tends to assemble only at mid-cell, where the time-averaged concentration of MinCD is the lowest (de Boer et al., 1992; Raskin and de Boer, 1999; Lutkenhaus, 2007). In B. subtilis, MinE is not present, and MinCD resides simultaneously at both cell poles via the action of a polar-targeting protein DivIVA. Homologues of DivIVA are found in many Gram-positive bacteria where they seem to play a variety of roles in controlling polar morphogenesis (Kang et al., 2008; Letek et al., 2008). The current view of DivIVA function would be that it simply recruits MinCD to the cell poles, via a direct interaction with MinD, and thereby sequesters it away from mid-cell until division initiates (Marston et al., 1998; Marston and Errington, 1999). However, despite repeated attempts using a variety of different methods, we have not been able to detect a direct interaction between DivIVA and MinD.

Here we describe the discovery of MinJ protein, a novel component of the division-site selection system of B. subtilis that bridges DivIVA and MinD. MinJ is targeted to cell poles by interaction with DivIVA, and MinJ in turn recruits MinD. In the course of characterizing the phenotype of minJ mutants, we unexpectedly discovered that the B. subtilis MinCD protein can act downstream of FtsZ assembly, even when expressed at normal levels, suggesting that this protein controls selection of the division site in multiple ways. Finally, our data are described in the light of a model in which the main function of the Min system is to deactivate the division machine after use, thereby preventing a second round of division at newly formed cell poles. While this work was being revised for publication we became aware that Patrick and Kearns (2008) had undertaken a similar study with largely similar results and conclusions.


A novel division-associated gene minJ (yvjD)

We previously found that the essential FtsL protein is controlled by regulated intramembrane proteolysis (RIP) (Makinoshima and Glickman, 2006) and identified an S2P protease responsible for FtsL turnover (Bramkamp et al., 2006). Many examples of RIP are now known and in most cases at least two proteases cooperate in the regulation. In an attempt to identify other proteases involved in FtsL regulation we searched for possible protease-encoding genes in the complete B. subtilis genome. Among the candidates was a gene called yvjD or swrAB, which is apparently co-regulated with swrAA, a key regulator of swarming motility in B. subtilis (Kearns et al., 2004; Calvio et al., 2005). Previous work had hinted at an essential role for swrAB and an in vitro proteolytic activity (Calvio et al., 2005). We propose to rename the swrAB gene minJ in accordance with the paper by Patrick and Kearns (2008). We disrupted minJ by insertion of an integrating plasmid (pMUTIN4; Vagner et al., 1998) and found that the resulting strain (3865), though viable, was markedly impaired for cell division, particularly during logarithmic growth. The cells grew as long, aseptate filaments with irregular and rare division septa (Fig. 1B; compare with the wild type, Fig. 1A). The average cell length (measured on cells stained with a membrane dye) was around 14 μm, with a large range of lengths (Fig. 1B). Very short anucleate minicells were also formed at a significant frequency (around 8%; Table 1). Apparently, when cells succeeded in dividing once they often went on to divide again close to the original site. Chromosome segregation appeared normal in the mutant (based on DAPI staining), showing that the effects on cell division were not secondary to a replication or segregation defect (data not shown). To check that disruption of minJ was responsible for this phenotype (rather than being a polar effect on downstream gene expression), a wild-type copy of minJ, controlled by a xylose-inducible promoter, was introduced into the mutant at an ectopic locus (amyE) (strain 3878). This strain exhibited a normal cell length distribution in the presence of 0.05–0.5% xylose (Table 1), excluding the possibility that the phenotype was due to a polar effect on downstream gene expression.

Figure 1.

Effects of minJ mutations on cell morphology. Phase-contrast images of typical fields of cells of (A) wild-type B. subtilis 168, (B) strain 3865 (ΔminJ), (C) strain 3309 (ΔminCD) and (D) strain MB012 (ΔminCDJ) are shown. Corresponding histograms showing cell length distributions are shown to the right. Scale bar 5 μm.

Table 1.  Cell length and minicell phenotypes.
StrainRelevant genotype (and growth condition)Length ± SD μm (No. cells counted)% Minicells
  1. Cells were grown in CH medium at 37°C. For estimation of minicell formation ≥ 200 cells were counted for each strain. ND, not done.

168Wild type3.83 ± 1.07 (201)0.25
3865ΔminJ14.12 ± 13.3 (100)7.97
MB003gfp–minJΔminJ3.40 ± 1.0 (316)3.1
MB004minJ–gfpΔminJ2.72 ± 0.63 (240)< 0.1
3381ΔminC4.65 ± 1.34 (205)27.4
3309ΔminCD6.23 ± 2.17 (186)22.0
MB013ΔminCΔminJ5.51 ± 1.74 (265)24.9
MB012ΔminCDΔminJ5.93 ± 2.72 (321)21.7
 (+Xylose)4.46 ± 1.37 (433)ND
 (−Xylose)4.56 ± 1.47 (538)ND
 (+Xylose)4.6 ± 2.49 (260)ND
 (−Xylose)6.07 ± 3.28 (269)ND

Expression of minJ was examined by β-galactosidase activity, taking advantage of the lacZ fusion to the native minJ promoter generated by pMUTIN4 insertion. Expression from the minJ promoter was readily detected during vegetative growth (∼30 Miller units) and not detectable during sporulation (data not shown).

minJ disruption prevents maturation of the Z ring

To test which step in cell division was affected by the minJ mutation we introduced GFP fusions to several different division proteins and examined whether their localization pattern was affected. Here we used a membrane stain (left panels in Fig. 2A–D) to detect rare division septa in the minP mutant and assess the extent to which division proteins were recruited to intermediate sites in the filaments. As shown in Fig. 2A, GFP–FtsZ formed fairly regular discrete structures at regular intervals along the cell filaments. These structures frequently appeared to be short helices, rather than the usual rings (enlarged images shown in Fig. 2E). Co-staining for DNA revealed that these helices were generally localized in the spaces between nucleoids, consistent with them being positioned by the nucleoid occlusion system (data not shown) (Wu and Errington, 2004). Occasionally, GFP–FtsZ assembled close to cell poles, apparently at sites where minicells were being formed (arrow in Fig. 2E). Similar patterns of localization were seen for YFP–FtsA (Fig. 2B and F), GFP–EzrA (Fig. 2C) and YFP–ZapA (not shown), showing that minJ mutation does not prevent the recruitment of these proteins to Z assemblies. Figure 2G–J shows the normal localization patterns for the various fusions, in non-filamentous wild-type cells.

Figure 2.

Subcellular localization of various division proteins in a ΔminJ background.
A–D. The strains examined contained the following fluorescent reporters: (A) FtsZ–GFP (strain 3869); (B) YFP–FtsA (strain MB009); (C) EzrA–GFP (strain 3874); (D) GFP–PBP2B (strain 3868). The panels on the left show fluorescence microscopic images of a membrane stain (Nile Red) and the corresponding GFP/YFP images are to the right.
E and F. Enlargements of the tips of representative filamentous cells of the strains in (A) and (B) respectively. Fluorescent structures probably located between nucleoids are indicated by white bars, polar assemblies with arrows and structures that could be helical with arrowheads.
G–J. Controls with the fusion proteins in wild-type background: (G) strain 2020 (GFP–FtsZ); (H) strain MB007 (YFP–FtsA); (I) strain 4057 (EzrA–GFP); (J) strain 3122 (GFP–PBP2B).
The left panel shows a phase contrast and the right panel the GFP/YPF fluorescence. Scale bar 2 μm.

In contrast, the remaining proteins that we examined were all impaired in recruitment to the FtsZ assemblies, including GFP–PBP2B (Fig. 2D) and GFP–FtsL (not shown). A few discrete bands of GFP localization were evident but these were located at rare division sites or the sites of minicell formation. Elsewhere, GFP–PBP2B was dispersed, although still apparently membrane-associated. It therefore appears that disruption of minJ interferes with the recruitment of the ‘late’ division proteins to FtsZ assemblies.

MinJ localizes to division sites and is retained at mature cell poles

To examine the subcellular distribution of the MinJ protein we constructed GFP fusions to its N- and C-termini (strains MB001 and MB002). Both fusions were inserted ectopically (at the amyE locus) and expressed from the Pxyl promoter. No overexpression phenotype was observed when these fusions were induced by addition of xylose (up to 0.5%). To test whether the fusions were functional they were introduced into cells bearing the insertionally inactivated allele of minJ. The wild-type cell phenotype of the parent strain was completely restored by expression of either fusion, using anywhere between 0.05% and 0.5% xylose. Curiously, the GFP–MinJ fusion did not fully complement the null mutation, as a few minicells were still made (∼3%; Table 1, Fig. S1). We cannot exclude the possibility that the fusion proteins are less active or less stable than wild-type MinJ.

Typical examples of the subcellular localization of GFP–MinJ and MinJ–GFP are shown in Fig. 3. The patterns of localization were indistinguishable, with both proteins localizing at the cell poles (arrowheads) and in mid-cell bands (stars) likely to correspond to division sites (Table 2). Deconvolution of representative images confirmed that at least some of the bands correspond to rings, as is typical of most division-associated proteins (Movie S1). However, the retention of the fusion proteins at mature cell poles (i.e. the poles of daughter cells after they have completely separated) was unusual, as most division proteins are released from the division site as the Z ring constricts or the pole matures.

Figure 3.

Subcellular localization of MinJ. B. subtilis strains expressing MinJ–GFP (MB002; left) or GFP–MinJ (MB001; right) were grown in CH medium supplemented with 0.5% xylose. Arrowheads point to examples of GFP fluorescence at cell poles, and stars to MinJ at nascent division sites. Scale bars 2 μm.

Table 2.  Subcellular localization of MinJ.
StrainMinJ localization (% of cells counted)
PolarPolar and septalSeptal
  1. Cells were grown in CH medium at 37°C.

MB001 (MinJ–GFP) (n = 187)334226
MB002 (GFP–MinJ) (n = 168)423621

Targeting of MinJ to division sites depends on both early and late division proteins

To investigate which division proteins were needed for recruitment of MinJ to the cell division apparatus we began by using a recently described inhibitor of FtsZ, PC58538 (Stokes et al., 2005). This compound destabilizes FtsZ polymers, leading to a rapid dispersion of FtsZ protein throughout the cell. After culturing cells for about two generations (90 min) in the presence of PC58538, cells were clearly elongated and septa were almost absent (as determined by membrane staining). Within these cells fluorescence from a GFP–MinJ fusion protein was completely dispersed throughout the cells (Fig. 4B). Control cells treated with solvent (DMSO) only exhibited the expected GFP–MinJ localization at division sites and cell poles (Fig. 4A). The apparently dispersed localization of GFP–MinJ in treated cells was unexpected for an integral membrane protein. Apparently, MinJ protein is degraded when it is delocalized or no longer associated with FtsZ (S. van Baarle and M. Bramkamp, unpublished). Further work is needed to determine the role, if any, of this degradation. We then examined the localization of GFP–MinJ in a strain in which PBP2B could be depleted. Again, MinJ did not localize to internucleoid spaces and was relatively dispersed, but it was still able to target to the infrequent division sites that occur under Pbp2B depletion conditions, and to mature cell poles (Fig. 4D). Therefore, localization and or stability of MinJ probably depends on assembly of most of the division machinery.

Figure 4.

Localization of GFP–MinJ in cells affected in cell division. Wild-type (strain MB001; A and B) or PBP2B-depletable (MB019; C and D) cells were grown to mid-exponential phase and treated with an inhibitor of FtsZ assembly (B), or in the presence (C) or absence (D) of IPTG. Phase-contrast (left) or GFP images of typical cells (right) are shown. Arrowheads point to examples of MinJ localizations.

Polar localization of MinJ depends on DivIVA and is required for MinD localization

The unusual localization of MinJ at both division sites and cell poles was reminiscent of the division-site selection proteins DivIVA, MinC and MinD. To test whether these proteins were needed for polar localization of MinJ the GFP–MinJ fusion was introduced into cells with null mutations in divIVA and minD. Strikingly, the polar (and division-site) localization of MinJ was eliminated in the divIVA mutant (Fig. 5A) but localization was normal in the minCD mutant (taking account of the minicell phenotype) (Fig. 5B). We then made reciprocal tests to see whether polar localization of DivIVA and MinD were dependent on MinJ. DivIVA localization has two distinct elements, polar and mid-cell (Edwards and Errington, 1997; Marston et al., 1998). As shown in Fig. 5C, in the absence of MinJ, DivIVA–GFP did not form bands or helices at intermediate positions along the filaments but strong fluorescence was evident at the poles of all of the filaments. In striking contrast, polar localization of GFP–MinD (as well as division-site localization) was abolished in the minJ mutant (Fig. 5D). Instead, the protein was dispersed throughout the cytoplasm or localized in foci that were scattered throughout the filaments. Therefore, MinJ appears to act as a bridge between DivIVA and MinD: MinJ requires DivIVA to be retained at cell poles, and MinJ needs to be present at the poles for MinD to be retained. We will return to the effects on localization to mid-cell sites below.

Figure 5.

Interdependence tests for localization of the division-site selection proteins.
A and B. Effect of (A) divIVA and (B) minCD mutation on MinJ–GFP localization (strains MB015 and MB017 respectively).
C and D. Effect of minJ mutation on localization of (C) DivIVA–GFP (strain 3872) and (D) MinD–GFP (MB006).
Examples of localizations to nascent division sites (star) and to the cell poles (arrowheads) are indicated. Cells were grown in CH medium at 37°C. Membranes were stained with Nile Red. Scale bar 2 μm.

Suppression of the minJ division block by inactivation of the MinCD division inhibitor

The results described above suggested that MinJ might be a component of the division-site selection system, in which case the filamentous phenotype of minJ mutants might be due to uncontrolled activity of the MinCD division inhibitor. If so, filamentation should be dependent on MinCD function. We therefore constructed double mutants of minJ with either ΔminC or ΔminCD mutations. The resultant strains were examined by phase-contrast and fluorescence microscopy (Fig. 1C and D and Table 1). In both cases the double mutants were indistinguishable from the respective minC or minCD parent strains – they produced abundant minicells but their average cell length (excluding minicells) was much less than that of the minJ single mutant (Table 1). Therefore, the impaired division of minJ mutants is likely to be caused by abnormal activity of the MinCD inhibitor.

Synthetic lethal interaction of minJ and noc

We previously showed that loss of both the nucleoid system and the Min system gives a synthetic lethal division defect (Wu and Errington, 2004). If minJ is important for Min function it should also be synthetic lethal with noc. Accordingly, repeated attempts to construct a minJ noc double mutant by simple transformation crosses failed. To obtain quantitative evidence that the double mutant is lethal, we took DNA from a noc::spc insertional mutant and transformed this into the minJ::pMUTIN4 mutant (strain 3865) and into wild-type cells as a control. Selection for spectinomycin resistance gave abundant colonies with the wild-type recipient (1 × 103 ml−1), but only rare transformants with the 3865 recipient (2 × 101 ml−1). This difference in frequency was not due to reduced competence for transformation in the minJ recipient because the same competent cultures were transformed with roughly equal (high) frequency by a control DNA (6 × 103 ml−1). We then examined the rare transformants that were obtained on the plates that should have generated noc minJ double mutants. In all (of 16) colonies tested, the cells proved to be erythromycin-sensitive, indicating that they had lost the pMUTIN4 insertion. This result provided strong evidence that combination of noc and minJ mutations is lethal. To examine the phenotype of doubly defective cells we constructed a minJ conditional (xylose-dependent) mutant (strain 3878) and introduced the noc disruption in the presence of 0.05% xylose. Although the removal of inducer did not lead to cell death, probably due to leakiness of repression of the Pxyl promoter, in the absence of inducer cells showed an increase in cell length (Table 1) consistent with the double mutant being affected in cell division.

MinJ interacts directly with DivIVA, MinD and other division proteins

The subcellular localization of MinJ and its apparent function as a bridge between DivIVA and MinD led to predictions about protein–protein interactions that we tested by bacterial two-hybrid analysis (Karimova et al., 2005). In a series of experiments multiple interactions were detected, and Fig. 6 shows a single plate that collates the interactions of significance. MinJ showed strong interactions with itself, with division proteins FtsA, EzrA, FtsL and PBP2B and with both DivIVA and MinD, but not MinC. The multiplicity of interactions with the various division proteins is not unusual and similar results have been described previously (Daniel et al., 2006; Claessen et al., 2008). The interaction with both DivIVA and MinD was significant as it supported the notion that MinJ is a bridge between these proteins. Equivalent tests with MinC revealed only weak interactions with itself and MinD. MinD interacted reciprocally with MinC, as well as showing weak interactions with two division proteins (FtsL and PBP2B), but importantly, it did not detectably interact with DivIVA. DivIVA showed only self-interaction. Therefore, these data provided further support for the idea that MinJ could act as an intermediate in delivering MinD to DivIVA at cell poles.

Figure 6.

Protein interactions detected by bacterial two-hybrid tests. E. coli strain BTH101 was co-transformed with two-hybrid vector plasmids (pUT18C, ‘T-18’; and pUT25C, ‘T-25’) expressing fusions to various cell division protein genes as indicated. Transformants were spotted onto selective plates containing X-Gal and incubated at 30°C for 40 h. Blue coloration indicates a positive interaction.

Inhibition of division by the Min system occurs downstream of FtsZ assembly

The above data strongly implicated MinJ as a novel and critical player in the division-site selection system of B. subtilis. According to this model, the filamentous phenotype of minJ mutants is due to unrestricted activity of the MinCD division inhibitor. A considerable amount of work has been done on the division inhibitor in E. coli, and a long-standing assumption emerging from this work has been that the effect is manifested at the level of FtsZ assembly (de Boer et al., 1990; Pichoff and Lutkenhaus, 2001; Dajkovic et al., 2008) (although there has been one report of a later effect; Justice et al., 2000). We previously reported that divIVA mutants of B. subtilis are defective in Z ring formation, again consistent with MinCD acting on FtsZ assembly (Marston et al., 1998). However, this work was done many years ago, and it was based on immunofluorescence methods, which tend to give relatively poor resolution and are technically demanding and much less reproducible than GFP methods. Given that the minJ mutant appeared to form regular FtsZ structures, albeit sometimes helical or malformed, and that these structures recruited FtsA and ZapA, we decided to revisit the phenotype of divIVA mutants. As shown in Fig. 7, the FtsZ–GFP pattern of fluorescence in divIVA null mutant (Fig. 7B) was very similar to the pattern observed in the minJ mutant (Fig. 7A). Regular bands and helices were formed, at sites corresponding to internucleoid spaces. This provides additional support for the notion that minJ mutants are affected in MinCD regulation and provides a new view of MinCD action, on a critical event downstream of FtsZ.

Figure 7.

FtsZ structures assemble frequently in cells lacking MinJ (A, strain RD022) or DivIVA (B, strain RD023). Arrowheads point to FtsZ structures with a helix-like appearance. Cells were grown in CH medium at 37°C. Scale bar 5 μm.


MinJ is a novel component of the division-site selection system that bridges DivIVA and MinD

A key element of the long-standing model for functioning of the Min system of B. subtilis is that DivIVA recruits MinCD to the cell poles, thereby preventing minicell formation and sequestering MinCD away from mid-cell until division is underway (Cha and Stewart, 1997; Edwards and Errington, 1997; Marston et al., 1998; Karoui and Errington, 2001). An obvious prediction of this model was that DivIVA should interact with MinD. However, numerous attempts to detect such an interaction by a variety of methods have been unsuccessful (L. Hamoen et al., unpublished). The discovery of MinJ represents a crucial step forward by providing a protein that can bridge DivIVA and MinD. Several lines of evidence support this idea. First, MinJ localizes in a manner that is highly similar to the localization of other division-site selection proteins, DivIVA (Marston et al., 1998), MinD and MinC (Marston and Errington, 1999) (although fine details of the various localization patterns vary, particularly in terms of the relative amounts of protein at mid-cell versus the cell poles). Based on the data presented above and previously (Marston et al., 1998; Autret and Errington, 2001) the pattern of dependence for recruitment to the cell poles can be summarized in a simple linear hierarchy (Fig. 8A). Second, minJ mutants are division-defective but, in common with divIVA mutants and unlike other division mutants, the phenotype is suppressed by inactivation of minC or minD (Cha and Stewart, 1997). This suppression presumably occurs because the cell division arrest of minJ mutants is due to impaired control over the MinCD division inhibitor (Cha and Stewart, 1997; Marston and Errington, 1999). Third, as summarized in Fig. 8B, the protein interaction tests (Fig. 6) demonstrated clear interactions between all of the adjacent pair-wise combinations of proteins in Fig. 8A, but not between non-adjacent proteins.

Figure 8.

Summary of protein interactions detected and a model for Min function.
A. Interdependences revealed by localization.
B. Interactions detected by bacterial two-hybrid analysis.
C–G. Schematic model for MinCD function during cell cycle progression. The horizontal bar represents the division site, which is initially occupied by early proteins, particularly FtsZ (C). At this stage the division machine is unable to mature (indicated by red shading) due to inhibition by MinCD (blue boxes). Away from MinCD inhibition (e.g. at the mid-point of long cells), the early division proteins become active (green) (D) and the late proteins are recruited (green hatching), resulting in commitment to division (E). This allows recruitment of DivIVA and MinJ (not shown) and in turn, MinCD (F). Constriction results in disassembly of the division machinery and MinCD is on hand to prevent reassembly or continued activity of the division machine (G).

minJ is predicted to encode a protein of 397 amino acids (44 kDa). The N-terminal portion of the protein is predicted to have multiple (most likely six) transmembrane spans. On the basis of topology predictions and preliminary protease accessibility data (M. Bramkamp, unpublished), the N- and C-termini are likely to be intracellular. The C-terminal domain of about 110 amino acids comprises a PDZ domain, which is a structure that is frequently involved in protein–protein interactions (Jemth and Gianni, 2007). Although the PDZ domain is very widespread throughout the three kingdoms of life, proteins with an N-terminal transmembrane domain similar in organization to MinJ are strikingly present only in the Firmicute group of the Gram-positive bacteria, and specifically in rod-shaped organisms, such as Bacillus, Clostridium, Listeria and Lactobacillus. They appear absent from closely related spherical organisms, such as Staphylococcus, Streptococcus and Lactococcus. This distribution closely mirrors the distribution of MinC and MinD in Gram-positive bacteria, again consistent with a common function. It should be noted, however, that homologues of the MinC and MinD proteins are found in many diverse groups of bacteria outside the Gram-positives. In E. coli at least, the Min system is regulated in a remarkably different way, involving oscillation of the MinCD proteins from pole to pole, governed by MinE protein (Lutkenhaus, 2007).

Returning to the B. subtilis system, it seems that DivIVA provides a static marker for the cell poles (based on a polar-targeting mechanism that remains to be elucidated), which recruits the integral membrane protein, MinJ. MinJ in turn recruits the MinCD complex, possibly by interaction between its PDZ domain and MinD. We think that the proximity of minJ to swrAA is of limited consequence because the swrAA gene appears to be restricted to B. subtilis and its close relatives (Calvio et al., 2005).

The B. subtilis Min system acts downstream of FtsZ ring formation

The most surprising result emerging from our analysis of MinJ was that the arrest of cell division in minJ mutants occurs post FtsZ ring formation. We then showed that DivIVA mutants also arrest at this later step. Our previous report that divIVA mutants are affected at the earlier step of FtsZ assembly was based on immunofluorescence microscopy (Marston et al., 1998). Immunofluorescence experiments require damaging fixation and permeabilization treatments, which may have resulted in reduced visibility of FtsZ structures, or perhaps the rings formed in divIVA mutants are more fragile. In any case, a block in Z ring assembly was also anticipated based on numerous studies of the Min systems of both B. subtilis and E. coli. There is one report of a block in recruitment of FtsA to the Z ring in E. coli (Justice et al., 2000) but all other studies, both in vivo and in vitro, have supported an effect on Z ring assembly (Bi and Lutkenhaus, 1990; de Boer et al., 1990; Hu and Lutkenhaus, 1999; Levin et al., 2001; Pichoff and Lutkenhaus, 2001; Shiomi and Margolin, 2007; Dajkovic et al., 2008). In all of the in vivo studies the effect of MinCD was assessed in cells overproducing MinC or a tagged derivative. These experiments and the in vitro data demonstrate overwhelmingly that MinC can act on Z assembly. Furthermore, recent results show that E. coli MinC, at least, has two different activities affecting FtsZ assembly dynamics (Dajkovic et al., 2008). However, in the light of our new data, we suggest that, at least in B. subtilis, and when expressed at normal physiological levels, the MinCD inhibitor mainly acts downstream of FtsZ ring assembly, preventing the recruitment of the late group of transmembrane division proteins. In support of this idea we note that there are at least two other situations in which the natively expressed MinCD system of B. subtilis seems not to inhibit Z ring assembly. The first is in ezrA mutants, which make extra Z rings close to the cell poles, despite the presence and activity of MinCD (Levin et al., 1999). Again, these rings do not mature into active division sites presumably because of a downstream action of MinCD. Second, during sporulation, in which polar septa are formed, again, despite the continued presence of the MinCD inhibitor in its normal polar zones (Barak et al., 1998; Sharp and Pogliano, 2002). More work is needed to determine whether physiological concentrations of MinCD act by specifically blocking the recruitment of one or more late division proteins, or whether they simply reduce some feature of FtsZ assembly to the point where the late proteins cannot efficiently be recruited. Nevertheless, it is clear that several proteins, including FtsA, EzrA, SepF and ZapA, do associate with FtsZ and that specifically positioned ring- or helix-like structures can be formed.

A model for commitment to division in B. subtilis

Our results highlight the distinction between the time of recruitment of the division-site selection proteins (including DivIVA, MinJ and MinCD) to the division site, which is a late event dependent on all of the essential division proteins, and the point of action of MinCD inhibition, which is earlier, between recruitment of the early and late division proteins. The arrival of the MinCD inhibitor at the division site late in its development, perhaps after initiation of constriction, suggests a model for the mechanism of commitment to division in wild-type B. subtilis (Fig. 8C–G). In cells that are not ready to divide, or when FtsZ begins to assemble close to a cell pole (Fig. 8C), the MinCD inhibitor prevents further development of the Z structure by preventing recruitment of later proteins. FtsA, ZapA and EzrA (at least) are permitted by this system. EzrA additionally suppresses Z ring formation near the cell poles (Levin et al., 1999; 2001). At the correct mid-cell site, diminished inhibition by MinCD (Fig. 8D) allows maturation of the Z ring and recruitment of the late division proteins (Fig. 8E). The cell becomes committed to division at this point because we postulate that MinCD cannot reverse the recruitment of these proteins once assembled. Completion of divisome assembly allows constriction to begin. At about this point DivIVA, MinJ and MinCD arrive at the division site (Fig. 8F) (so they belong to the group of late proteins that are susceptible to the MinCD block in recruitment) but they are not competent to inhibit ongoing division. As division proceeds and the division machinery starts to disassemble (Fig. 8G), localized MinCD, with its ability to act at multiple levels to inhibit FtsZ assembly and the recruitment of late proteins, could be crucial for preventing immediate reassembly of a division complex that would otherwise generate a minicell. Therefore the recruitment of MinCD to mid-cell could primarily be a device for resetting the division system, allowing it to occur only once at any given site.

Experimental procedures

Bacterial strains, plasmids and oligonucleotides

All bacterial strains and plasmids used in this study are listed in Table S1. Oligonucleotides are listed in Table S2.

General methods

Liquid cultures of B. subtilis strains were routinely grown in casein hydrolysate (CH) medium (Sterlini and Mandelstam, 1969). Transformations were plated on nutrient agar plates (Oxoid) supplemented with antibiotics as required [5 μg ml−1 chloramphenicol, 5 μg ml−1 kanamycin, 50 μg ml−1 spectinomycin, 0.3 μg ml−1 erythromycin, 50 μg ml−1 xylose (0.5% w/v, unless otherwise stated) and isopropyl-β-d-thiogalactopyranoside (IPTG) (1 mM)].

Generally, all molecular biology experiments were performed according to standard protocols.

FtsZ dependency experiment

To test the dependence of MinJ localization on FtsZ polymerization strain MB001 (Pxyl-gfp-minJ) was grown in CH medium (30°C) supplemented with 0.2% xylose. Logarithmically growing cells (OD ∼0.5) were diluted in fresh, pre-warmed medium (1/10) and grown again to OD 0.5. This procedure was repeated three times. After the last dilution cells were incubated for 30 min. Subsequently the culture was split in half (5 ml, each). DMSO (2 μl) was added to the control culture and 2 μl of the FtsZ inhibitor PC58538 (Stokes et al., 2005) dissolved in DMSO was added to the other half (final concentration 50 μg ml−1). Localization of GFP–MinJ was analysed by fluorescence microscopy after 0, 30 and 90 min. Results shown are from the 90 min time point.

Depletion experiments

To deplete PBP2B cells were grown in 10 ml of CH media and 1/10 diluted twice in fresh media containing the appropriate inducer concentration (0.5% xylose and 1 mM IPTG). Logarithmically growing cells were collected by centrifugation (at 30°C) and cell pellets were washed in 10 vols of pre-warmed CH medium without IPTG. Cells were finally re-suspended in 10 ml of CH medium without IPTG and allowed to continue growth at 30°C, thereby depleting PBP2B. Samples were taken and analysed microscopically after 150 min.

Enzyme assay

β-Galactosidase activity was assayed as described previously (Daniel et al., 1996). Briefly, 200 μl of samples were flash frozen in liquid nitrogen and stored at −80°C.

Samples were thawed by addition of 600 μl of Z buffer (50 mM sodium phosphate buffer, pH 7.0, 10 mM KCl, 1 mM MgSO4, 50 mM β-mercaptoethanol). Lysozyme (200 μg ml−1) and DNase (100 μg ml−1) were added and the samples were incubated at 30°C for 20 min. To start the reaction 200 μl of ONPG (o-nitrophenyl-β-d-galactopyranoside; 4 mg ml−1) dissolved in Z buffer was added. After the characteristic yellow colour developed the reaction was stopped by addition of 400 μl of 1 M Na2CO3. The production of p-nitrophenolate was determined at 420 nm.

Bacterial two-hybrid analysis

To screen for interactions of MinJ with other proteins, the minJ coding sequence was amplified by PCR using primers MinJ B2H For and MinJ B2H Rev, which introduced XbaI and KpnI restriction sites at either end of the gene. The minJ PCR fragment was then cloned into the bacterial two-hybrid vectors pUT18C and p25-N (Karimova et al., 2005; Claessen et al., 2008), to give fusions of the T18 adenylate cyclase truncation to the N-terminus of MinJ and a C-terminal fusion of the T25 truncation to MinJ respectively. Interactions were tested as described previously (Daniel et al., 2006) with the exception that 12.5 μl of the co-transformations were spotted onto nutrient agar plates supplemented with 100 μg of ampicillin ml−1, 25 μg of kanamycin ml−1, 0.004% X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) and 0.1 mM IPTG. Colony colour was noted following incubation at 30°C for 40 h.

Microscopic imaging

For phase–contrast and fluorescence microscopy 1–3 μl of a culture sample was placed on a microscope slide coated with a thin agarose (1%) layer and covered with a coverslip. For membrane or DNA staining a 10 μl culture sample was mixed with 1 μl of Nile Red (Molecular Probes) (12.5 μg ml−1) or 1 μl of DAPI (Sigma) solution (1 μg ml−1 in 50% glycerol) respectively. Images were taken on a Zeiss Axiovert 200M microscope equipped with a Sony CoolSnap HQ cooled CCD camera (Roper Scientific) or on a Zeiss AxioImager M1 equipped with a Zeiss AxioCam HRm camera. Generally, an EC Plan-Neofluar 100×/1.3 Oil Ph3 objective was used. Digital images were acquired with the AxioVision (Zeiss) software and analysed using the METAMORPH 4.6.9 software (Universal Imaging). Final image preparation was performed in Adobe Photoshop 6.0 (Adobe Systems Incorporated).

Three-dimensional reconstruction

For three-dimensional reconstruction of GFP-labelled proteins the DeltaVision®RT (Applied Precision) system was used. Briefly, the system is based on a Olympus IX71 Inverted Microscope equipped with a Sony Interline ICX285 ER monochrome camera. Images were taken with a PLAN-APO 100X Oil, 1.4NA, 0.10 mm WD objective and analysed using the softWoRx® imaging workstation and Resolve3D image acquisition software. Preparation of the images was identical to normal fluorescent microscopy.


We thank Drs Yoshi Kawai, Leendert Hamoen and Ling Juan Wu for providing strains. Furthermore, we would like to thank Mrs Anja Stankowski for performing β-Gal assays. M.B. was supported with a postdoctoral fellowship by the Deutsche Forschungsgemeinschaft (DFG). This work was supported by a grant from the UK Biotechnology and Biological Sciences Research Council to J.E.