Through a series of genetic manipulations and adjustments to growth conditions, detailed below, we have been able to generate Min system oscillation in B. subtilis. This phenomenon is observed in the majority, if not all, of the cells in the population and occurs with an oscillation cycle time similar to that observed in E. coli (Fig. 1a, Movie S1).

MinD_Ec does not oscillate in the absence of MinE in E. coli (^{Hu & Lutkenhaus, 2001}), nor does it do so when introduced into B. subtilis (^{Pavlendová et al., 2010}). We therefore examined the effect of introducing MinD_Ec together with MinE into B. subtilis by constructing strains expressing yfp–minD_Ec and minE in a wild-type (IB1229) and a minD_Bs deletion (IB1230) background. In many cells, we observed YFP–MinD_Ec foci close to the cell membrane, especially in strain IB1229. Movement of these ‘dots’ was generally confined to a small local region (Movie S2), and occasionally the dots relocalized towards one of the cell poles. In IB1230 cells, YFP–MinD_Ec movement reminiscent of oscillation in E. coli was visible, especially in shorter cells (Fig. 1b, Movie S3). Since overexpression of Min proteins causes cell elongation (^{Marston & Errington, 1999}; ^{Pavlendová et al., 2010}), the longer cells exhibiting the brightly fluorescing dots are likely to have higher YFP–MinD_Ec concentrations.

Higher concentrations of MinD_Ec and MinE may interfere with the function of the Min system by biasing the proportions of the complexes formed. In addition, interaction among E. coli and B. subtilis Min system components may cause slower movement of YFP–MinD_Ec. In E. coli, the period of the Min oscillation cycle is 20–50 s (^{Raskin & de Boer, 1999a}; ^{Touhami et al., 2006}). To compare the oscillation times in E. coli and B. subtilis, we timed the YFP–MinD_Ec oscillation cycle in E. coli strain ΔminCDE P_lac : : yfp–minD_Ec : : minE–cfp (YLS1 : : pYLS68) (^{Shih et al., 2002}). In our hands, oscillation was observed with a period of about 1 min at room temperature. In contrast, the oscillation of YFP–MinD_Ec in B. subtilis ΔminD_Bs yfp–minD_Ec minE (IB1230) cells was slower at 1.5–3.5 min per cycle. Increasing the temperature to 30 °C, a change that in E. coli results in faster oscillation (from a cycle time of 20 s at 22 °C to 8 s at 30 °C; ^{Touhami et al., 2006}), did not significantly enhance the oscillation frequency of YFP–MinD_Ec in B. subtilis. We reasoned that the presence of B. subtilis DivIVA or MinJ might be limiting the mobility of YFP–MinD_Ec. To test this idea, we produced YFP–MinD_Ec and MinE in a B. subtilis strain in which either minD and divIVA (ΔminD_Bs ΔdivIVA yfp–minD_Ec minE, IB1242) or minD and minJ (ΔminD_Bs ΔminJ yfp–minD_Ec minE, IB1363) were deleted. In these cells, the period of the oscillation cycle was essentially unchanged (1.5–3 min), but oscillation was observed in almost all cells (Fig. 1a, Movie S1).

Next we explored the possibility that the lower frequency of YFP–MinD_Ec oscillation in the B. subtilis system was caused by perturbations in the concentration ratios of the Min proteins. In the B. subtilis strains described here, YFP–MinD_Ec and MinE were expressed from the P_hyperspank and P_xyl promoters, respectively, while in E. coli YLS1 : : pYLS68, both genes were transcribed from the P_lac promoter. To compare MinDE expression levels in B. subtilis and in E. coli, we performed Western blot analysis. It is possible to visualize both MinD_Ec and MinE on one blot using a monoclonal anti-GFP antibody, in a strain where both MinD_Ec and MinE are in fusion with fluorescent proteins (ΔminD_Bs yfp–minD_Ec minE–gfp, IB1155). Under induction conditions similar to those used for the microscopy experiments (0.5 mM IPTG and 0.1 % xylose), it can be seen in Fig. 1(c) that while the concentrations of YFP–MinD_Ec (upper bands in lanes 1, 2, 3, 5 and 6) are similar in both systems, the concentration of MinE–CFP (lower band, lane 6) in E. coli strain ΔminCDE P_lac : : yfp-minD_Ec : : minE–cfp (YLS1 : : pYLS68) is higher than the concentration of MinE–GFP in B. subtilis strain expressing both YFP–MinD_Ec and MinE–GFP (IB1155) (lower band, lanes 1, 2 and 3). Although significant differences in the MinE–GFP expression levels under the three induction conditions tested (Fig. 1c, lanes 1, 2 and 3) were not observed, induction with 0.1 % xylose led to the highest YFP–MinD_Ec oscillation frequency, which approached one oscillation period per minute in many cells of ΔminD_Bs (IB1230) and ΔminD_Bs ΔdivIVA (IB1242) B. subtilis strains. These experiments show that in the presence of MinE, YFP–MinD_Ec oscillates in B. subtilis and that the characteristics of the oscillation process closely reproduce the oscillation behaviour of the Min system observed in E. coli.

Over several days on DSM agar plates, colonies formed by strain ΔminD_Bs yfp–minD_Ec minE (IB1230) remained brighter coloured than those formed by wild-type B. subtilis cells, which became darker coloured as the cells sporulated. This suggested that IB1230 cells were impaired in sporulation. We therefore measured the sporulation efficiency of B. subtilis cells expressing the E. coli Min proteins. Interestingly, the sporulation efficiency of strain IB1230 is 10-fold lower (9 %) than that of wild-type cells (Table 2), suggesting that pole-to-pole oscillation of MinD inhibits spore formation. The sporulation efficiency of the strain ΔminD_Bs ΔdivIVA yfp–minD_Ec minE (IB1242), which gives the highest YFP–MinD_Ec oscillation frequency, was not tested, as divIVA mutants are already impaired in sporulation (^{Thomaides et al., 2001}).

Next, we inspected cells that retained wild-type minD_Bs (yfp–minD_Ec minE, IB1229). The sporulation efficiency of these cells was lower (53 %) than that of the wild-type but significantly higher than that of the ΔminD_Bs yfp–minD_Ec minE (IB1230) strain (Table 2). It is possible that higher levels of MinD (MinD_Bs plus MinD_Ec) lead to less efficient oscillation and thus to higher sporulation efficiency than that of strain IB1230. To test this possibility, we reinvestigated the sporulation efficiency of strain ΔminD_Bs yfp–minD_Ec minE (IB1230) under conditions in which the expression of YFP–MinD_Ec was increased by addition of 1 mM IPTG. Increased expression of MinD_Ec had no effect on the sporulation efficiency, which remained at 9 %. It seems therefore that it is the presence of MinD_Bs per se that causes the partial rescue of sporulation in strain yfp–minD_Ec minE (IB1229). When we examined the localization of YFP–MinD_Ec in this strain, we found that in most cells the YFP fluorescence appeared in the form of spots close to the membrane. In comparison with strain ΔminD_Bs yfp–minD_Ec minE (IB1230), clear YFP–MinD_Ec oscillation was visible in fewer cells. This observation implies that MinD_Bs binds to MinD_Ec and inhibits its MinE-induced oscillation.

We confirmed the implied interaction between the E. coli and B. subtilis MinD proteins using the bacterial two-hybrid system (Fig. 2). Overall, it seems that oscillation of MinD_Ec correlates with the lower sporulation frequency. Supporting this assertion, in a control ΔminD_Bs strain expressing MinD_Ec in the absence of MinE (ΔminD_Bs yfp–minD_Ec, IB1111), where MinD_Ec does not oscillate, sporulation is unimpaired (Table 2). This result excludes the possibility that the mere presence of MinD_Ec inhibits sporulation. In a control strain deleted for MinD_Bs (ΔminD_Bs, IB1056), the sporulation efficiency is only slightly decreased. In a strain expressing MinE alone (ΔminD_Bs minE–gfp, IB1107), the sporulation efficiency decreased to 56 %.

The simplest explanation for the decreased sporulation efficiency of strain IB1230 (ΔminD_Bs yfp–minD_Ec minE) is that interactions with MinD_Ec (Fig. 2) induce oscillation of MinC_Bs, leading to increased MinC concentrations at the cell poles which prevent asymmetrical septation. However, the sporulation efficiency of strain IB1370 (ΔminC_Bs ΔminD_Bs yfp–minD_Ec minE), in which both MinD_Bs and MinC_Bs are deleted, is even lower (2 %) than that observed in strain IB1230 (ΔminD_Bs yfp–minD_Ec minE). Thus, MinC oscillation does not explain the observed lowering of the sporulation efficiency of strain ΔminD_Bs yfp–minD_Ec minE (IB1230), in which MinD oscillation takes place. It is important to note that in MinD-deficient B. subtilis cells, the sporulation septum is often misplaced closer to mid-cell (^{Barák et al., 1998}; ^{Thomaides et al., 2001}). In addition, MinCD depletion causes loss of polarity in SpoIIIE-mediated chromosome translocation (^{Sharp & Pogliano, 2002}). However, neither of these two phenotypes is associated with as obvious a reduction in the sporulation efficiency as that observed in strains IB1230 or IB1370 (Table 2). Thus, we assume that the heterologous, oscillating Min system has an additional inhibitory effect on the complex process of sporulation either during asymmetrical septum formation or in the later stages.

The master regulator of sporulation initiation is Spo0A, a response regulator that is phosphorylated by a multi-component phosphorelay (^{Hoch, 1993}; ^{Perego & Hoch, 2002}). Phosphorylated Spo0A binds to specific promoter regions (‘0A boxes’), and activates or represses the expression of scores of genes required for sporulation (reviewed by ^{Piggot & Losick, 2002}; ^{Barák et al., 2005}). To test whether strains exhibiting oscillation of MinD_Ec are defective in sporulation initiation, we examined Spo0A expression levels by Western blotting (Fig. 3a). Spo0A is present at similar levels in strain ΔminD_Bs yfp–minD_Ec minE (IB1230), which exhibits Min system oscillation, and in the wild-type strain (PY79), which does not. Indeed in all strains examined, Spo0A was detected at normal levels, the exception being the control strain in which spo0A has been deleted (IB220; ^{Schmeisser et al., 2000}). Thus the reduced sporulation efficiency associated with the oscillating Min system is not caused by perturbations in the level of Spo0A. Since spo0A expression is positively autoregulated (^{Molle et al., 2003}), normal Spo0A levels indicate that the activity of Spo0A, and the system of proteins that activate it, is unaffected by the oscillating Min system.

A more likely explanation for the lowered sporulation efficiency is a defect in polar cell division. Oscillating MinD_Ec is expected to bind to MinC_Bs, thus conferring pole-to-pole oscillation on the cell division inhibitor, which would prevent polar septum formation. This hypothesis was tested by membrane staining. Cells of the wild-type strain (PY79) and strain ΔminD_Bs yfp–minD_Ec minE (IB1230) were grown until hours 2 and 4 of sporulation, and the membranes were stained with the dye FM4-64. The pattern of staining defined three discernible cell classes: (i) cells with a polar septum (stage II), (ii) cells in the later stages of sporulation (stage III and later), and (iii) vegetative cells.

For the wild-type strain, after 2 h, 44 % of the cells had not entered into sporulation, 49 % of cells showed a clear polar septum and the remaining cells were in stage III or later (Fig. 3b). Cells of strains ΔminD_Bs yfp–minD_Ec minE (IB1230) and ΔminC_Bs ΔminD_Bs yfp–minD_Ec minE (IB1370), which harbour the oscillating E. coli Min system components, were noticeably impaired in the formation of asymmetrical septa. In the second hour of sporulation, forespores in stage III or later were not observed, and an asymmetrical septum was observed in only about 27 % of the cells (IB1230 and IB1370). As mentioned previously, the sporulation efficiency of strain ΔminD_Bs yfp–minD_Ec minE (IB1230) is around 9 %. This indicates that even though polar septa are forming in 27 % of these cells at hour 2 of sporulation, only one-third of these give rise to resistant spores. In summary, B. subtilis cells, in which the E. coli Min system proteins oscillate, initiate sporulation normally but are impaired in sporulation septum formation.

The evolutionary implications of these observations are that bacteria which form endospores will have DivIVA/MinJ rather than MinE as the auxiliary component(s) of MinCD. Until recently, sporulation was thought to be restricted to species of Gram-positive bacteria. As shown in Table S3, the genomes of all Gram-positive endospore-forming bacteria encode a DivIVA homologue and most also encode a MinJ homologue. Interestingly, most of the sporulating Clostrideae sp. also possess a MinE homologue (Table S3). However, it is not known whether these MinE proteins are functional, if they are part of Min systems which oscillate, and what, if any, interplay there is with the DivIVA/MinJ system during vegetative growth and sporulation.

The chromosomes of almost all rod-shaped Gram-negative bacteria encode a MinE homologue, and some encode homologues of DivIVA (^{Rothfield et al., 2005}; Table S3). Gram-negative bacteria have hitherto been considered to be non-sporulating, with the possible exception of a sparsely documented example in Thermus. In addition, Myxococcus forms spores by converting the rod-shaped vegetative cell into a spherical spore without prior asymmetrical division (^{Kaiser, 2003}).

From the available data it is hard to deduce which Min system evolved from which, just as we do not know whether the common ancestor of Gram-positive and Gram-negative bacteria possessed these different characteristics. We can speculate that the Min systems either evolved separately or, more likely, evolved together in Gram-positive bacteria for the alternate life cycles of vegetative growth and sporulation, as MinE and DivIVA/MinJ are present in most Clostrideae sp. If this assumption is true, then most probably Gram-negative bacteria evolved from a Gram-positive bacterium. This notion is supported by the recent fascinating description of the cell membrane structures of Acetonoma longum (evolutionarily a close relative of Clostrideae sp.) during sporulation and spore outgrowth (^{Tocheva et al., 2011}). Those authors show that during sporulation the inner membrane of the mother cell is inverted and transformed to become an outer membrane of the germinating cell. Their results point to sporulation as a mechanism by which the bacterial outer membrane could have arisen. If A. longum is the missing link between single- and double-membraned bacteria, it is not surprising that it possesses the two cell-division regulatory systems that characterize Gram-positive and Gram-negative micro-organisms. Further work is needed to address whether and how these two systems function together in the same cell.