PomZ, a ParA-like protein, regulates Z-ring formation and cell division in Myxococcus xanthus

Authors


Summary

Accurate positioning of the division site is essential to generate appropriately sized daughter cells with the correct chromosome number. In bacteria, division generally depends on assembly of the tubulin homologue FtsZ into the Z-ring at the division site. Here, we show that lack of the ParA-like protein PomZ in Myxococcus xanthus resulted in division defects with the formation of chromosome-free minicells and filamentous cells. Lack of PomZ also caused reduced formation of Z-rings and incorrect positioning of the few Z-rings formed. PomZ localization is cell cycle regulated, and PomZ accumulates at the division site at midcell after chromosome segregation but prior to FtsZ as well as in the absence of FtsZ. FtsZ displayed cooperative GTP hydrolysis in vitro but did not form detectable filaments in vitro. PomZ interacted with FtsZ in M. xanthus cell extracts. These data show that PomZ is important for Z-ring formation and is a spatial regulator of Z-ring formation and cell division. The cell cycle-dependent localization of PomZ at midcell provides a mechanism for coupling cell cycle progression and Z-ring formation. Moreover, the data suggest that PomZ is part of a system that recruits FtsZ to midcell, thereby, restricting Z-ring formation to this position.

Introduction

In all organisms, accurate positioning of the division site is essential for generating appropriately sized daughter cells with a correct chromosome complement. In growing bacteria, cell division generally occurs by binary fission at midcell and depends on assembly of the tubulin GTPase homologue FtsZ into a ring-like structure, the Z-ring, at the incipient division site (Adams and Errington, 2009). Hence, the position of Z-ring formation dictates the position of the division site. Accordingly, all known systems that regulate positioning of the division site control placement of the Z-ring (Rothfield et al., 2005). The spatial regulation of Z-ring positioning is well-studied in Bacillus subtilis, Caulobacter crescentus and Escherichia coli and in all three organisms involves negatively acting systems that inhibit Z-ring formation at the cell poles and over the nucleoid in that way leaving only midcell free for Z-ring formation (Rothfield et al., 2005). It has been speculated that positioning of the Z-ring may be positively regulated in growing B. subtilis (Moriya et al., 2010; Rodrigues and Harry, 2012) and it was recently demonstrated that during spore differentiation in aerial hyphae in Streptomyces coelicolor positioning of FtsZ at the division sites is positively regulated (Willemse et al., 2011).

FtsZ is essential for cell division in most bacteria and is thought to have two functions. First, subsequent to Z-ring formation, FtsZ functions as a scaffold to directly or indirectly recruit the remaining components of the division machinery, the divisome, required to carry out cytokinesis (Adams and Errington, 2009). Second, membrane-associated FtsZ may generate a force that result in constriction of the membrane (Osawa et al., 2008). In vivo, the Z-ring is thought to consist of short, overlapping protofilaments that associate by lateral interactions (Li et al., 2007). The filaments are highly dynamic with subunits exchanging with a cytoplasmic pool on a timescale of seconds (Stricker et al., 2002). Several proteins have been identified in E. coli, B. subtilis and C. crescentus that interact directly with FtsZ in vitro and stabilize the Z-ring in vivo once the Z-ring has formed at the correct position (Adams and Errington, 2009). While FtsZ is highly conserved in bacteria and presumably has the same function in cell division in different species, these FtsZ interacting proteins are not conserved with the exception of the FtsA protein that links the Z-ring to the membrane (Pichoff and Lutkenhaus, 2005).

FtsZ is a cytoplasmic protein and consists of an N-terminal GTP-binding domain and a C-terminal domain (Löwe and Amos, 1998). Several FtsZ proteins have been shown to polymerize to form filaments in a GTP-dependent manner in vitro (Bramhill and Thompson, 1994; Lu et al., 1998; White et al., 2000; Thanbichler and Shapiro, 2006; Willemse et al., 2011). Polymerization depends on the head-to-tail association of adjacent FtsZ molecules and involves the GTP-bound N-terminal domain of one subunit interacting with the C-terminal domain of the preceding subunit (Oliva et al., 2004). Importantly, FtsZ GTPase activity is switched on by this interaction because a key residue for catalysis in the C-terminal domain is inserted into the GTP-binding pocket of the adjacent N-terminal domain creating a complete GTPase active site (Oliva et al., 2004). Thus, the polymeric form of FtsZ is the active GTPase. GTP hydrolysis causes the disassembly of FtsZ filaments (Bramhill and Thompson, 1994; Mukherjee and Lutkenhaus, 1994). Consistent with the central role of FtsZ polymerization in cell division, the inhibitors of Z-ring formation prevent FtsZ polymerization as in the case of SlmA, MipZ and MinC (Hu et al., 1999; Thanbichler and Shapiro, 2006; Cho et al., 2011) while the positive regulator SsgB stimulates FtsZ polymerization (Willemse et al., 2011).

The systems that spatially regulate positioning of the Z-ring are not conserved. In growing bacteria the components of these systems localize to the cell poles, oscillate between the cell poles or localize over the nucleoid. Polarly localized systems include the Min-system in B. subtilis that consists of the MinC, MinD, MinJ and DivIVA proteins (Rothfield et al., 2005; Bramkamp et al., 2008; Patrick and Kearns, 2008), and the ParB/MipZ system in C. crescentus (Thanbichler and Shapiro, 2006). In the B. subtilis Min-system, MinC is the inhibitor of Z-ring formation and forms a complex with MinD, which is a member of the ParA/MinD/Soj superfamily of P-loop ATPases, while MinJ is the adaptor between the MinCD complex and the polar targeting protein DivIVA. In the MipZ/ParB system, the MipZ ATPase, which is also a ParA/MinD/Soj superfamily member, is the inhibitor of Z-ring formation and ParB is a DNA-binding protein that forms a complex with MipZ and binds to a region close to the polarly localized origins of replication. Oscillating spatial regulators of Z-ring formation include the Min-system in E. coli that consists of MinC and MinD with functions similar to those in B. subtilis and MinE that drives the oscillations of the MinCD complex (Rothfield et al., 2005). Spatial regulators that localize over the nucleoids include the non-homologous Noc (Wu and Errington, 2004) and SlmA (Bernhardt and de Boer, 2005) DNA-binding proteins in B. subtilis and E. coli respectively. In total, these negatively acting systems function to inhibit Z-ring formation at the cell poles in the case of the MinCDJ/DivIVA, the MinCDE and MipZ/ParB systems, and over the nucleoid – also referred to as nucleoid occlusion – in the case of Noc and SlmA. In sporulating S. coelicolor cells, the SsgAB proteins localize between the chromosomes at the incipient division sites prior to FtsZ and then recruit FtsZ by direct interactions (Willemse et al., 2011).

Neither the MinCD core components of the two Min-systems (Rothfield et al., 2005), MipZ of the MipZ/ParB system (Thanbichler and Shapiro, 2006) nor the SsgAB proteins (Willemse et al., 2011) are universally conserved and there are bacteria that lack homologues of all these systems. Among these bacteria is the δ-proteobacterium Myxococcus xanthus, which has mostly been studied to understand social behaviour in bacteria (Konovalova et al., 2010): In the presence of nutrients, the rod-shaped M. xanthus cells grow and divide by binary fission while in the absence of nutrients growth ceases and the rod-shaped cells aggregate to form multicellular fruiting bodies inside which cells differentiate to spherical spores. Here, we report the identification of the ParA/MinD/Soj P-loop ATPase PomZ. Our data show that PomZ is important for cell division as well as for the correct positioning of the division site at midcell in growing cells of M. xanthus. Moreover, our data show that PomZ is important for Z-ring formation and its correct positioning at the incipient division site. Based on the localization pattern of PomZ, we speculate that PomZ is part of a system that recruits FtsZ to the division site in that way positively regulating Z-ring formation and cell division.

Results

PomZ is important for cell division and division site positioning

While studying agmE (MXAN0635; from here on PomZ (Positioning at midcell of FtsZ) for its proposed function in cell motility (Youderian et al., 2003), we made several observations suggesting that PomZ is important for cell division. A ΔpomZ mutation results in formation of filamentous cells and chromosome-free minicells (Fig. 1A–C), which are division defects typical of mutations in genes involved in the spatial regulation of Z-ring formation. Wild-type (WT) cells vary in length from 2.9 to 13.8 μm (average 5.3 ± 1.9 μm) whereas ΔpomZ cells vary in length from 1.4 to 38.4 μm [average 10.7 ± 6.1 μm (excluding minicells)]. Moreover, 9% of WT cells had a cell division constriction at midcell, while only 3% of ΔpomZ cells had division constrictions (Fig. 1D). Importantly, the constrictions in the ΔpomZ cells were observed along the cell length and not only at midcell as in WT (Fig. 1C and D). In both WT and ΔpomZ cells, division constrictions were only observed in nucleoid-free gaps (Fig. 1C). The cell division defects were corrected by ectopic expression of pomZ+ at native levels (Fig. 1A and B; Fig. S1). Similar to WT, the ΔpomZ mutant had a doubling time of 5 h in rich medium and, thus, had no growth defect. From these observations we conclude that PomZ is important for cell division and for restricting and positioning the division site to midcell.

Figure 1.

PomZ is important for cell division site positioning.

A. Morphological changes in cells of the indicated genotypes. Minicell marked by arrow. Scale bar, 2 μm.

B. Histogram of cell length distribution of cells of the indicated strains (n > 250 for each strain). Numbers in brackets indicate mean cell length ± standard deviation (SD) (excluding minicells), percentage of minicells (cells shorter than 2.5 μm). +++ indicates overexpression of ftsZ.

C. Aberrant positioning of constrictions (arrows) leads to formation of chromosome-free minicells in ΔpomZ cells. Nucleoids were stained with DAPI. Scale bar, 2 μm.

D. Constrictions are aberrantly localized in ΔpomZ cells. Each triangle indicates the localization of a constriction as a function of cell length (WT n = 500, ΔpomZ n = 1100).

Lack of PomZ does not interfere with chromosome replication and segregation

Mutants with a defect in chromosome replication or segregation may generate filamentous cells and minicells (Niki et al., 1991). To test whether chromosome replication or segregation were affected by the ΔpomZ mutation, we used two approaches. First, the number of nucleoids per cell length was enumerated and the position of the nucleoid midpoints determined in WT and ΔpomZ cells after staining with DAPI (4′,6-diamidin-2-phenylindol). Because ΔpomZ cells are longer than WT cells, nucleoid number and position were also determined in WT after treatment with cephalexin, which in E. coli inhibits division without interfering with replication (Pogliano et al., 1997). WT and ΔpomZ cells generally contained one, two, four or eight nucleoids (Fig. 2A–C). Several ΔpomZ cells contained six nucleoids (Fig. 2B and C). Presumably these cells arise from a cell with eight nucleoids dividing between nucleoids six and seven. Moreover, WT and ΔpomZ cells contained the same number of nucleoids per cell length suggesting that initiation of replication is normal in the absence of PomZ (Fig. 2C). To obtain more precise information about nucleoid positioning, the positions of the nucleoid midpoints in cells with one, two or four nucleoids were determined. As shown in Fig. 2D and E, the distributions of nucleoid midpoints positions were similar in WT and ΔpomZ cells suggesting that chromosome segregation is normal in the absence of PomZ.

Figure 2.

Chromosome replication and segregation are similar in WT and ΔpomZ cells.

A and B. Representative cells of the indicated genotypes with different numbers of nucleoids. WT and ΔpomZ cells were stained with DAPI and in the case of WT also after cephalexin treatment for 8 h and visualized by DIC and fluorescence microscopy. Scale bar, 2 μm. The schematics above the micrographs indicate the number of nucleoids per cell. Cells had a doubling time of 5 h before addition of cephalexin.

C. WT and ΔpomZ cells contain the same number of nucleoids per cell length. The number of nucleoids per cell plotted as a function of cell length. n = 100 for both strains.

D and E. Histogram showing the distribution of nucleoid midpoints along the long axis in WT (D) and ΔpomZ (E) cells with one (black), two (brown) or four (orange) nucleoids. Numbers indicate mean ± SD for the localization of nucleoid midpoints as percentage of cell length. n = 100 for both strains.

F and G. Representative cells of the indicated genotypes containing different numbers of ParB–YFP clusters in WT (F) and ΔpomZ (G) cells. WT before and after cephalexin treatment for 6 h and ΔpomZ cells were visualized by DIC and fluorescence microscopy. Scale bar, 2 μm. The schematics above the micrographs indicate the number of ParB–YFP clusters in a cell. Cells had a doubling time of 5 h before addition of cephalexin.

H. WT and ΔpomZ cells contain the same number of ParB–YFP clusters per cell length. The number of ParB–YFP clusters per cell is plotted as a function of cell length. n = 200 for both strains.

I and J. Histogram showing the distribution of ParB–YFP clusters along the long axis of WT (J) and ΔpomZ (M) cells with one (black), two (brown) or four (orange) clusters. Numbers indicate mean ± SD for the localization of ParB–YFP clusters as percentage of cell length. n = 200 for both strains.

In the second approach to determine whether chromosome replication and segregation were affected by the absence of PomZ, we followed the copy number of a chromosomal locus as well as its location. To this end, we focused on the ParA and ParB proteins, which have been shown to be involved in chromosome segregation in several species (Gerdes et al., 2010). In these systems, ParB binds to conserved parS sequences that are typically localized close to the origin of replication and the parAB operon (Livny et al., 2007). Bioinformatic analyses identified MXAN7477 and MXAN7476 in the M. xanthus genome as homologues of ParA and ParB respectively. Using 12 parS sequences previously identified bioinformatically in the M. xanthus genome (Livny et al., 2007) as seeds and allowing for one mismatch, we identified a set of 24 parS sequences in the region upstream of parAB and one remote parS sequence in the M. xanthus genome (Fig. S2). The 24 parS sequences localize approximately 30 kb from the origin. In vitro analyses using purified M. xanthus ParB confirmed that ParB binds to the consensus parS sequence (A. Harms et al., in preparation).

To follow ParB localization, we constructed a ParB–YFP fusion and determined its localization in WT and ΔpomZ cells. As described above, WT was treated with cephalexin to obtain longer WT cells with multiple chromosomes. WT as well as ΔpomZ cells mostly contained between one and eight ParB–YFP clusters (Fig. 2F–H). Presumably cells with three, five and six ParB–YFP clusters reflect asynchrony in the initiation of replication or in origin segregation in cells with multiple chromosomes. WT and ΔpomZ cells contained the same number of ParB–YFP clusters per cell length (Fig. 2H). Also, a detailed analysis of the localization of ParB–YFP revealed that it localized similarly in WT cells and ΔpomZ cells with one, two or four clusters (Fig. 2I and J). In total, these data strongly suggest that PomZ is not required for chromosome replication and segregation.

PomZ is required for Z-ring formation and positioning

Next, we sought to determine the function of PomZ in cell division. Therefore, the effect of PomZ on FtsZ (MXAN5597) accumulation and localization was determined. In the ΔpomZ mutant, FtsZ accumulated at levels similar to that observed in WT (Fig. S3). As in other bacteria (Ma et al., 1996; Thanbichler and Shapiro, 2006), we found that a FtsZ–mCherry did not support growth in the absence of native FtsZ. Therefore, we constructed a merodiploid ftsZ+/ftsZ–mcherry strain. Cells of this strain were slightly longer than WT (Fig. 1B) as reported for E. coli ftsZ+/ftsZ–gfp cells (Ma et al., 1996). In 51% of pomZ+ cells, FtsZ–mCherry localized diffusely in the cytoplasm or in clusters that did not span the cell width and were distributed along the cell length (from here on, diffuse/speckled pattern); the remaining 49% of cells contained FtsZ–mCherry at midcell in a single structure with the appearance of a transverse band spanning the cell width (Fig. 3A and B). The two patterns correlated with cell length (Fig. 3B) and cell cycle progression (Fig. 3C), i.e. cells containing one nucleoid or two incompletely segregated nucleoids had FtsZ–mCherry in the diffuse/speckled pattern, and cells with segregated nucleoids had FtsZ–mCherry at midcell and localizing between the two segregated nucleoids. To confirm that FtsZ–mCherry localization reflects the localization of native FtsZ, immunofluorescence microscopy using α-FtsZ antibodies was carried out. As shown in Fig. 3D, native FtsZ localized similarly to FtsZ–mCherry with 49% of cells having FtsZ at midcell. These observations suggest that the transverse band at midcell corresponds to the Z-ring and demonstrate that FtsZ–mCherry is incorporated into the Z-ring together with native FtsZ. As expected, it was also observed by time-lapse microscopy that FtsZ–mCherry localization is dynamic and changes over time from diffuse/speckled to midcell localization (Fig. S4).

Figure 3.

PomZ is important for formation and correct positioning of the Z-ring.

A. FtsZ–mCherry localizes in two patterns in WT. ftsZ+/ftsZmcherry cells were analysed. Left and right panels show diffuse/speckled and midcell localization of FtsZ–mCherry respectively. Insets, structures indicated by arrows. Scale bar, 5 μm.

B. FtsZ–mCherry localization correlates with cell length. ftsZ+/ftsZmcherry cells were analysed. Box plots summarize cell length of cells with a diffuse/speckled (left) and midcell (right) FtsZ–mCherry localization (n = 100). Boxes enclose the 25th and 75th percentile with the horizontal black line representing the median, the horizontal red line the mean, whiskers the 10th and 90th percentile, and circles outliers.

C. FtsZ–mCherry localization correlates with cell cycle progression. DAPI stained ftsZ+/ftsZmcherry cells were analysed. Upper panels, cell with one nucleoid, middle panels, cell with incompletely segregated nucleoids, lower panels, cell with fully segregated nucleoids. Numbers indicate percentage of cells with that localization (n = 100). Scale bar, 5 μm. Fluorescence intensity profiles of FtsZ–mCherry (red) and DAPI (green) in arbitrary unit (a.u.) along the cell length are shown on the right.

D. FtsZ localizes in two patterns in immunofluorescence microscopy in WT. Left and right panels show diffuse/speckled and midcell localization respectively. Numbers indicate percentage of cells with that localization (n = 117). Scale bar, 5 μm.

E. Localization of FtsZ–mCherry in ΔpomZ cells is abnormal. ΔpomZ, ftsZ+/ftsZmcherry cells were analysed. Upper panel, diffuse/speckled FtsZ–mCherry localization, middle panels, Z-ring close to pole, lower panel, Z-ring in midcell region. Insets, structures indicated by arrows. Scale bar, 5 μm. Fluorescence intensity profiles of FtsZ–mCherry (red) and DAPI (green) are shown as in (C).

F. Quantification of FtsZ–mCherry localization patterns in ΔpomZ cells as a function of cell length (n = 800). Box plot shown as in (B).

G. Z-rings are aberrantly localized in ΔpomZ cells. Each triangle indicates the localization of a Z-ring as percentage of cell length and as a function of cell length (WT n = 100, ΔpomZ n = 800).

H. FtsZ is not limiting for Z-ring formation. Localization of overproduced FtsZ–mCherry in WT (n = 100) and ΔpomZ (n = 100) cells. Numbers indicate percentage of cells with that localization. Panels 1 and 3 show diffuse/speckled FtsZ–mCherry localization and panels 2 and 4 show Z-rings. Insets, structures indicated by arrows. Scale bar, 5 μm.

Strikingly, in ΔpomZ cells, FtsZ–mCherry was observed in the diffuse/speckled pattern in 98% of cells and only 2% contained a Z-ring (Fig. 3E and F). These Z-rings localized to nucleoid-free gaps along the entire cell length and independently of cell length (Fig. 3E and G), thus, mirroring the localization of cell division constrictions in ΔpomZ cells (Fig. 1D). In agreement with these observations, we observed by immunofluorescence microscopy using α-FtsZ antibodies that 3% of the ΔpomZ cells contained a Z-ring and that these Z-rings localized along the entire cell length (data not shown). Cells containing more than one Z-ring were not observed. Thus, PomZ is important for Z-ring formation and for restricting and positioning the Z-ring at midcell.

In E. coli lack of the MinCDE system results in division defects similar to those caused by ΔpomZ; however, a Min mutant forms Z-rings at an increased frequency (Yu and Margolin, 1999; Bernhardt and de Boer, 2005). Two- to sevenfold FtsZ overproduction corrects the division defect in a min mutant arguing that this defect is caused by multiple Z-rings competing for FtsZ (Bi and Lutkenhaus, 1990; Bernhardt and de Boer, 2005). To test whether the cell division defect in the ΔpomZ mutant was corrected by FtsZ overproduction, FtsZ or FtsZ–mCherry were overproduced approximately fivefold in the ΔpomZ mutant (Fig. S3). This overproduction resulted in a minor increase in Z-ring formation from 2% to 8% of cells (Fig. 3H); however, no correction of the cell division defect was observed and filamentous cells and chromosome-free minicells still formed (Fig. 1B). Thus, even after overproduction of FtsZ, the severe defect in Z-ring formation and cell division in the ΔpomZ mutant is not corrected suggesting that a factor important for Z-ring formation and cell division is missing in the ΔpomZ mutant.

Two- to sevenfold overproduction of FtsZ in WT E. coli causes an increase in the division frequency with the formation of chromosome-free minicells and higher levels of FtsZ overproduction cause inhibition of cell division (Ward and Lutkenhaus, 1985). In WT M. xanthus, fivefold overproduction of FtsZ or FtsZ–mCherry caused an insignificant increase in Z-ring formation from 49% to 54% and these Z-rings were all localized at midcell (Fig. 3H) suggesting that FtsZ is not limiting for Z-ring formation in WT M. xanthus. Moreover, the overproduction of FtsZ or FtsZ–mCherry also resulted in the formation of longer cells than in WT but not of minicells (Fig. 1B) suggesting that the stoichiometry of the divisome components in M. xanthus is important for cell division as is the case of the divisome in E. coli.

PomZ localization correlates with cell cycle progression

To dissect how PomZ affects Z-ring formation, we determined the localization of PomZ using an active PomZ–mCherry fusion expressed at native levels (Fig. 1B; Fig. S1). PomZ localizes in three patterns that correlate with cell length (Fig. 4A and B) and cell cycle progression (Fig. 4C). Short cells with one nucleoid had PomZ–mCherry in a patchy pattern over the nucleoid; longer cells with one nucleoid or two incompletely segregated nucleoids contained PomZ–mCherry in a single cluster over the nucleoid positioned at 37.1 ± 7.1% of the cell length (from here on, off-centre cluster), and the longest cells with two completely segregated nucleoids had PomZ–mCherry at midcell in the gap between the two nucleoids and forming a structure that spanned the cell width. PomZ–mCherry was also present at division constrictions (Fig. 4A) suggesting that PomZ is an integral part of the divisome. Time-lapse microscopy with 5 min time intervals demonstrated that PomZ–mCherry localization is dynamic and changes from patchy to off-centre to midcell (Fig. S5). Time-lapse recordings with shorter time intervals did not reveal additional PomZ dynamics.

Figure 4.

PomZ localization correlates with cell cycle progression.

A. PomZ–mCherry localizes in three patterns. PomZ–mCherry in ΔpomZ cells was analysed. Left panel, patchy pattern; second panel, off-centre cluster; third panel, midcell cluster; fourth panel, PomZ–mCherry at constriction (yellow arrow). Scale bar, 5 μm. Insets, structures indicated by white arrows.

B. Quantification of PomZ–mCherry localization as a function of cell length. Box plots summarize cell length of cells with a patchy, off-centre and midcell PomZ–mCherry localization (n = 100). Box plot is shown as in Fig. 3.

C. PomZ–mCherry localization correlates with cell cycle progression. DAPI stained ΔpomZ cells with PomZ–mCherry were analysed. Two upper panels, cells with one nucleoid, third panel, cell with incompletely segregated nucleoids, lower panel, cell with fully segregated nucleoids. Numbers indicate percentage of cells with that localization (n = 113). Scale bar, 5 μm. Fluorescence intensity profiles of PomZ–mCherry (red) and DAPI (green) in a.u. are shown as in Fig. 3.

D. Effect of rifampicin and cephalexin on PomZ–mCherry localization. ΔpomZ cells with PomZ–mCherry were grown in the absence of rifampicin (Rif.) and cephalexin (Ceph.) and rifampicin or cephalexin added at t = 0 h. Histogram shows % of cells with the indicated PomZ–mCherry localization (n = 100). Cells had a doubling time of 5 h before addition of rifampicin or cephalexin.

Because the PomZ–mCherry localization pattern correlates with cell cycle progression, we tested whether cell cycle progression regulates PomZ localization. Therefore, the effect of rifampicin, which inhibits initiation of replication as well as transcription (Kimchi and Rosenberg, 1976), and cephalexin, which inhibits cell division, on PomZ–mCherry localization was determined. After 8 h of rifampicin treatment, 88% of cells had PomZ–mCherry in the patchy localization pattern (Fig. 4D). In parallel, the levels of PomZ–mCherry only decreased slightly (data not shown) suggesting that initiation of replication or a linked event such as chromosome segregation is required for the patchy to off-centre and midcell relocation. Similarly, after 8 h cephalexin treatment PomZ–mCherry was found at midcell in 71% of cells (Fig. 4D). In the cephalexin-treated cells, additional PomZ clusters were not assembled between the newly replicated chromosomes suggesting that cell division is required for midcell to patchy relocation of PomZ.

The observations that the ΔpomZ mutant has a reduced frequency of division constrictions as well as of Z-rings, that the few Z-rings formed in the absence of PomZ are localized away from midcell, that PomZ localizes to midcell and appear to be part of the divisome and that FtsZ is not limiting for Z-ring formation support a hypothesis in which PomZ is involved in recruiting FtsZ to midcell and stabilizing the Z-ring. A prediction of this model is that overproduction of PomZ or PomZ–mCherry could lead to an increase in the frequency of Z-ring formation. Strains overproducing PomZ and PomZ–mCherry phenocopied the ΔpomZ mutant with the formation of chromosome-free minicells and long filamentous cells and only 2% of cells contained a Z-ring as determined by immunofluorescence microscopy (data not shown). However, overproduced PomZ–mCherry localized aberrantly and was only observed in the patchy pattern over nucleoids and the nucleoids were abnormally shaped (data not shown) suggesting that the effect of overproduction of PomZ or PomZ–mCherry on Z-ring formation and cell division is indirect and caused by the abnormal localization of PomZ. Therefore, the strains overproducing PomZ or PomZ–mCherry were not considered further.

PomZ localizes to the division site before FtsZ

Cells containing PomZ at midcell match those with FtsZ at midcell, i.e. they are long and contain two completely segregated nucleoids. If PomZ is involved in recruiting FtsZ to midcell, then PomZ should localize to midcell before FtsZ. To test this prediction, we determined PomZ and FtsZ localization in the same cells. For this purpose, we generated a FtsZ–GFP fusion. This fusion localizes similarly to native FtsZ and FtsZ–mCherry with 47% of cells having the protein at midcell (data not shown). In a strain containing FtsZ–GFP and PomZ–mCherry, four combinations of localization patterns were observed that correlated with cell-length (Fig. 5A and B). In 45% of cells the two proteins colocalized at midcell. Notably, 12% of cells contained PomZ–mCherry at midcell and FtsZ–GFP in a diffuse/speckled pattern whereas FtsZ–GFP was never observed at midcell without PomZ–mCherry at midcell. Similarly, in a strain expressing PomZ–mCherry and native FtsZ, PomZ–mCherry and FtsZ (as determined using immunofluorescence microscopy) colocalized at midcell and PomZ localized at midcell while FtsZ was in a diffuse/speckled pattern (data not shown). Importantly, FtsZ was never observed at midcell without PomZ–mCherry at midcell. Thus, PomZ localizes before FtsZ at midcell.

Figure 5.

PomZ localizes to midcell before and in the absence of FtsZ.

A. PomZ–mCherry and FtsZ–GFP colocalize at midcell and PomZ–mCherry ‘arrives’ first. pomZ+, ftsZ+ cells expressing PomZ–mCherry and FtsZ–GFP were analysed. Four localization patterns were observed with the indicated percentages in (B). Scale bar, 5 μm. Fluorescence intensity profiles of PomZ–mCherry (red) and FtsZ–GFP (green) are shown as in Fig. 3.

B. Box plots summarize cell length of cells with the indicated patterns of PomZ–mCherry and FtsZ–GFP localization (n = 800). Box plot is shown as in Fig. 3.

C. PomZ–mCherry and FtsZ–GFP remain at midcell if division is blocked. Cells expressing PomZ–mCherry and FtsZ–GFP were treated with cephalexin for 8 h. Cells had a doubling time of 5 h before addition of cephalexin. Scale bar, 5 μm. Fluorescence intensity profiles of PomZ–mCherry (red) and DAPI (green) in a.u. are shown as in Fig. 3.

D. PomZ–mCherry localizes to midcell in the absence of FtsZ. Upper panel, level of FtsZ during FtsZ depletion in SA4107 (pomZ+, ΔftsZ/PcuoA-ftsZ, pomZ–mCherry). Cells were transferred to copper-free medium at t = 0 h. Upper, middle and lower panels, levels of FtsZ, PilC and PomZ/PomZ–mCherry during FtsZ depletion. PilC is involved in type IV pili function (Bulyha et al., 2009) and is used as a loading control. Lower panel, % of cells with indicated PomZ–mCherry localization (n = 100 at each time point). Cells had a doubling time of 5 h in absence of copper until 12 h.

In cephalexin treated WT E. coli cells, Z-rings often assemble between newly replicated chromosomes (Pogliano et al., 1997). As shown in Fig. 4D, in WT M. xanthus cells treated with cephalexin, PomZ accumulated at midcell and not between newly replicated chromosomes. Consistent with the idea that PomZ is important for Z-ring formation, we observed that after 8 h of treatment with cephalexin of the strain containing FtsZ–GFP and PomZ–mCherry 81% of cells contained the two proteins at midcell (Fig. 5C) and no additional Z-rings were assembled between the newly replicated chromosomes. Also, the observations that the single Z-ring formed in the long cephalexin treated WT cells is at midcell whereas the rare Z-rings formed in the long ΔpomZ mutant are typically localized away from midcell demonstrate that the abnormal localization of Z-rings in the ΔpomZ mutant is caused by lack of PomZ and is not due to the long cell length.

PomZ localizes to the division site in the absence of FtsZ

To test whether PomZ localizes to midcell in the absence of FtsZ, PomZ–mCherry localization was followed in a ΔftsZ strain with a copy of ftsZ integrated at the cuoA locus and expressed under control of the copper activated cuoA promoter (Gómez-Santos et al., 2012). After removal of copper, cells continued to grow with a normal growth rate for 12 h and from then on had a reduced growth rate. FtsZ levels had decreased threefold by 4 h and were undetectable from 8 h by immunoblot analysis whereas PomZ levels remained constant (Fig. 5D). From 0 to 10 h of the depletion, corresponding to two doubling times, cell length increased approximately twofold and constrictions were visible until 8 h (Fig. 5D) suggesting that some cells still divided during the early stages of the FtsZ depletion. Importantly, the frequency of cells with PomZ–mCherry at midcell increased during FtsZ depletion and reached a maximum of 86% at 10 h; in parallel the frequency of cells with PomZ–mCherry in a patchy or off-centre pattern decreased (Fig. 5D). Notably, most of the cells in which FtsZ was not detected (from 8 to 12 h) had PomZ at midcell (Fig. S6) and the frequency of cells with PomZ–mCherry at midcell increased from 8 to 10 h when no FtsZ was detectable. Upon further growth in the absence of FtsZ, PomZ–mCherry localized aberrantly throughout cells suggesting that the divisome eventually disassembles in the absence of FtsZ. We conclude that PomZ localizes to midcell in the absence of FtsZ suggesting that PomZ localizes to midcell independently of FtsZ.

PomZ binds ATP and GTP

PomZ is a member of the ParA/Soj/MinD superfamily of P-loop ATPases (Fig. S7) that dimerize and in some cases also form filaments in an ATP-dependent manner (Leonard et al., 2005; Gerdes et al., 2010). To characterize PomZ in vitro, an N-terminally His6-tagged PomZ protein was purified (Fig. 6A). Control experiments showed that His6-PomZ expressed at native levels corrected the cell division defects caused by the ΔpomZ mutation (data not shown), thus, demonstrating that the protein is active. His6-PomZ (37.9 kDa) eluted corresponding to a mass of 128 kDa in gel filtration experiments suggesting that it was purified in a higher oligomeric state (Fig. 6B).

Figure 6.

FtsZ has cooperative GTPase activity in vitro.

A. Purification of FtsZ and His6-PomZ. SDS-PAGE analysis of FtsZ (lane 1) and His6-PomZ (lane 2) after staining with Coomassie blue. Single asterisk (*) indicates ClpB and double asterisk (**) Hsp70 from E. coli that co-purified with FtsZ. Positions of molecular markers are indicated.

B. Overlaid size gel filtration elution profiles of FtsZ and His6-PomZ. Elution patterns were measured at 280 nm for FtsZ (green) and His6-PomZ (red) from a Superdex 200 5/150 GL size exclusion column. Arrows indicate the elution maxima of marker proteins.

C. TNP-ATP binding to His6-PomZ. The TNP-ATP concentration was held constant at 500 nM with increasing concentrations of His6-PomZ. Data shown were corrected for TNP-ATP background fluorescence. Notice that His6-PomZ shows no fluorescence at 540 nm.

D. Competition of TNP-ATP binding to His6-PomZ with GTP or ATP. TNP-ATP at a concentration of 500 nM was bound to 11.5 μM His6-PomZ. The decrease in the fraction of bound TNP-ATP upon competition with ATP or GTP is shown as a function of increasing GTP or ATP concentrations. Data shown were corrected for TNP-ATP background fluorescence.

E. Mant-GTP and mant-ATP binding to FtsZ. The increase in anisotropy upon binding of mant-labelled nucleotides to increasing concentrations of FtsZ was determined for binding of 25 nM mant-GTP or 100 nM mant-ATP.

F. Competition of mant-GTP binding to FtsZ with GTP and ATP. Mant-GTP (100 nM) was bound to 5 μM FtsZ and competed with increasing concentrations of GTP or ATP.

G. NTPase activity of FtsZ. The specific NTPase activity of increasing concentrations of FtsZ was measured for 500 μM GTP or 500 μM ATP.

H. FtsZ and His6-PomZ do not form filaments as observed by right angle light scattering. Experiments were performed with 10 μM FtsZ at 8°C and pH 7.2 in the presence or absence of 10 μM His6-PomZ as indicated. Nucleotides were added at 100 s as indicated by the colour code at final concentrations of 500 μM or 500 μM GTP with 10 mM CaCl2. The increased signal in the presence of both FtsZ and His6-PomZ is the sum of the background signals caused by the two individual proteins.

I. FtsZ interacts with His6-PomZ in pull-down experiments. WT M. xanthus cell extract was applied to a Ni2+-NTA-agarose column with or without bound His6-PomZ. Eluted proteins were separated by SDS-PAGE and visualized by Coomassie Brilliant Blue R-250 staining. Proteins in visible bands were identified by mass spectrometry. Positions of His6-PomZ, FtsZ and molecular markers are indicated.

J. NTPase activity of FtsZ, His6-PomZ and the two proteins together. Specific NTPase activities in the presence of 500 μM GTP and 500 μM ATP as a function of increasing concentrations of FtsZ alone, His6-PomZ alone or in the presence of equimolar amounts of FtsZ and His6-PomZ. Note that the FtsZ curve for GTP hydrolysis is not sigmoidal because data points at low FtsZ concentrations were not included.

His6-PomZ was purified essentially in a nucleotide free form (OD260/OD280 ratio of 0.55). His6-PomZ bound mant (N-methylanthraniloyl)-ATP and mant-GTP with low affinities (data not shown) but bound TNP-ATP (trinitrophenyl-ATP) with a Kd of 4.3 μM (Fig. 6C). Reverse titration of PomZ with increasing concentrations of TNP-ATP resulted in a Kd of 5.0 μM. Bound TNP-ATP was competed with ATP and GTP with half maximal effective concentration (EC50) values of 113 and 265 μM respectively (Fig. 6D). Neither ATP nor GTP hydrolysis was detected, even at His6-PomZ concentrations up to 10 μM (Fig. 6J; data not shown). Under all conditions tested and at a PomZ concentration up to 10 μM, filament formation was observed neither by right angle light scattering (Fig. 6H) nor by negative stain electron microscopy (EM) (data not shown).

Because we were unable to detect nucleotide hydrolysis by His6-PomZ in vitro, we tested genetically the relevance of nucleotide binding and hydrolysis by PomZ by expressing the PomZK66Q and PomZD90A variants in WT and ΔpomZ cells (Fig. S7B). The corresponding substitutions in Soj of B. subtilis and MipZ and ParA of C. crescentus block ATP binding (ParAK20Q) and ATP hydrolysis (SojD40A, MipZD42A, ParAD44A) (Leonard et al., 2005; Thanbichler and Shapiro, 2006; Ptacin et al., 2010). PomZK66Q with or without fused mCherry was highly unstable precluding in vivo analyses. Importantly, PomZD90A and PomZD90A–mCherry accumulated at levels similar to PomZ and PomZ–mCherry (Fig. S1). Both proteins were unable to complement the ΔpomZ mutation and were negative dominant in WT (Fig. 1B; data not shown). PomZD90A–mCherry formed a single cluster that localized randomly over a nucleoid independently of cell length (Fig. 7A) and the PomZD90A–mCherry cluster rarely colocalized with a Z-ring (Fig. 7B).

Figure 7.

ATPase activity is required for PomZ function and localization.

A. PomZD90A–mCherry localizes along the cell length. ΔpomZ cells expressing PomZD90A–mCherry were DAPI stained. The panel shows the PomZD90A–mCherry cluster over a nucleoid. Fluorescence intensity profiles of PomZ–mCherry (red) and DAPI (green) (a.u., arbitrary units) along the cell length are shown on the right. Scale bar, 5 μm. Inset, structure indicated by arrow.

B. PomZD90A–mCherry and FtsZ rarely colocalize. ΔpomZ cells expressing PomZD90A–mCherry and FtsZ–GFP were analysed. Scale bar, 5 μm. Fluorescence intensity profiles of PomZ–mCherry (red) and FtsZ–GFP (green) (a.u., arbitrary units) along the cell length are shown on the right. In cell ‘1’, the PomZD90A–mCherry cluster is close to a Z-ring-like structure and in cell ‘2’ apart from a Z-ring-like structure. Insets show structures indicated by arrows.

FtsZ hydrolyses GTP cooperatively

To characterize M. xanthus FtsZ in vitro, native FtsZ was purified (Fig. 6A and B). FtsZ (44.7 kDa) eluted corresponding to a mass of 174 kDa in gel filtration experiments suggesting that it was also purified in a higher oligomeric state (Fig. 6B). FtsZ was purified essentially in a nucleotide free form (OD260/OD280 ratio of 0.55). FtsZ bound mant-GTP and mant-ATP with a Kd of 0.07 and 0.4 μM respectively (Fig. 6E). In comparison, FtsZ of E. coli (Dajkovic et al., 2008), Methanocooccus jannaschi (Huecas et al., 2007) and C. crescentus (see below) bind mant-GTP with a Kd of 9 μM, 0.25 μM and 2.7 μM respectively. Titration of 5 μM FtsZ with increasing concentrations of mant-GTP resulted in a linear increase in fluorescence up to 4.5 μM mant-GTP demonstrating that at least 90% of FtsZ was functional in mant-GTP binding. Mant-GTP binding was competed with GTP or ATP with half maximal effective concentration (EC50) values of 0.3 and 67 μM, respectively (Fig. 6F), demonstrating that FtsZ, as expected, has a higher affinity for GTP than for ATP. Similarly to other FtsZ proteins, FtsZ hydrolysed GTP in a highly concentration-dependent and cooperative manner (Hill coefficient of 3.4) with a maximal activity of 2.4 GTP min−1 at 10 μM FtsZ (Fig. 6G) while no hydrolysis of ATP was observed. A similar concentration-dependent and cooperative GTPase activity is observed for other FtsZ proteins although the concentration at which maximal GTPase activity is reached differs, e.g. for FtsZ from E. coli (Dajkovic et al., 2008), B. subtilis (Wang and Lutkenhaus, 1993) and C. crescentus (see below) this concentration is 3 μM, 5 μM and 1 μM respectively.

The highly concentration-dependent and cooperative GTP hydrolysis suggested that FtsZ undergoes GTP-dependent polymerization. However, under all conditions tested and with FtsZ concentrations up to 10 μM, at which the highest specific GTPase activity was reached, filament formation was detected neither by right angle light scattering (Fig. 6H) nor by EM (data not shown). FtsZ filaments can be stabilized by Ca2+ (Yu and Margolin, 1997). However, FtsZ filament formation was not observed after inclusion of 10 mM CaCl2 (Fig. 6H). Non-hydrolysable GTP analogues may (Scheffers et al., 2000; Mingorance et al., 2005; Goley et al., 2010) or may not (Bramhill and Thompson, 1994; Yu and Margolin, 1997; White et al., 2000) induce or stabilize FtsZ filaments. FtsZ filament formation was also not observed after incubation with the non-hydrolysable GTP analogues GMP-PNP (β, γ-imidoguanosine 5′-triphosphate) or GTP-γ-S (guanosine-5′-O-3-thiotriphosphate) (Fig. 6H). Finally, filaments were observed neither at pH 6.5, which stimulates lateral association of FtsZ protofilaments (Mukherjee and Lutkenhaus, 1999; Gundogdu et al., 2011), nor at different temperatures (8°C, 20°C or 30°C) (data not shown).

Because FtsZ from M. xanthus behaves unlike other FtsZ proteins characterized in vitro with respect to detectable filament formation, we characterized native FtsZ from C. crescentus (FtsZcc), which forms filaments in a GTP-dependent manner (Thanbichler and Shapiro, 2006), under our experimental conditions for comparison (Supporting information and Fig. S8; Table S1). Indeed FtsZcc rapidly formed filaments as shown using right angle light scattering and EM (Supporting information and Fig. S8; Table S1).

FtsZ and PomZ interact in pull-down experiments

To test whether PomZ and FtsZ interact, we performed pull-down experiments using His6-PomZ in combination with total-cell extracts of WT M. xanthus. In these pull-down experiments, FtsZ in M. xanthus cell extracts bound as the most abundant protein to His6-PomZ (Fig. 6I; Table S2).

To test for a direct interaction between the purified His6-PomZ and FtsZ several experiments were performed. First, in pull-down experiments in which His6-PomZ was bound to a Ni2+-NTA-agarose matrix, binding of purified FtsZ was detected neither in the absence nor in the presence of ATP and GTP (data not shown). Second, in experiments in which equimolar amounts of His6-PomZ were added to FtsZ in the presence of ATP and GTP no effect on the total nucleotide hydrolysis rate was observed (Fig. 6J). Finally, in experiments in which equimolar amounts of His6-PomZ were added to FtsZ in the presence of ATP and GTP filament formation was observed neither by right angle light scattering (Fig. 6H) nor by EM (data not shown). Similar results were obtained when the two proteins were mixed in 1:2, 1:1 and 2:1 ratios in the presence of ATP and GTP (data not shown). Moreover, filament formation was observed neither in the presence of GTP and the non-hydrolysable ATP analogue AMP-PNP (β, γ-imidoadenosine 5′-triphosphate) nor in the presence of non-hydrolysable GTP analogues and ATP (data not shown).

Discussion

The data presented demonstrate that PomZ is important for cell division as well as for Z-ring formation and positioning in growing M. xanthus cells. Two lines of evidence demonstrate that PomZ is important for cell division: First, in the absence of PomZ, cells contain threefold fewer division constriction sites than WT suggesting that ΔpomZ cells divide threefold less frequently than WT. Second, in the absence of PomZ, Z-ring formation is compromised and the frequency of Z-ring formation strongly reduced demonstrating that PomZ is important for Z-ring formation. Two lines of evidence demonstrate that PomZ is important for correct positioning of the cell division site. First, the few division constrictions observed in the absence of PomZ are not restricted to midcell as they are in WT. Second, the few Z-rings formed in ΔpomZ cells are also not restricted to midcell.

In E. coli and/or B. subtilis FtsA, ZipA, ZapA, ZapB, ZapC and SepF have structural functions and act to stabilize the Z-ring (Adams and Errington, 2009; Hale et al., 2011). With the exception of FtsA from B. subtilis (Beall and Lutkenhaus, 1992), in the absence of any one of these proteins, Z-rings or Z-ring-like structures still form at essentially WT frequencies and at the correct location (Jensen et al., 2005; Adams and Errington, 2009; Hale et al., 2011). Moreover, in the absence of FtsZ, none of these six proteins localize to midcell (Adams and Errington, 2009). Among FtsA, ZipA, ZapA, ZapB, ZapC and SepF, M. xanthus only encodes a homologue of FtsA (MXAN5599). Our data suggest that PomZ is not a simple functional analogue of any of the five remaining proteins because PomZ in addition to being important for Z-ring formation is also important for the correct positioning of the Z-ring. Moreover, unlike ZipA, ZapA, ZapB, ZapC and SepF, PomZ localizes to midcell before and in the absence of FtsZ. Thus, PomZ is functionally distinct from these five structural proteins.

The observations that PomZ is important for Z-ring formation as well as for its correct positioning suggest that PomZ is involved in spatially regulating Z-ring formation. Among proteins of regulatory systems that act as spatial regulators to inhibit Z-ring formation in other bacteria, i.e. MinCDE, MinJ, MipZ, DivIVA, M. xanthus only contains a DivIVA homologue (MXAN3112). Deletion of MXAN3112 does not cause cell division defects (data not shown). In E. coli, lack of MinE causes a reduction in Z-ring formation (Bi and Lutkenhaus, 1993). However, MinE does not localize to the divisome but rather oscillates between the cell poles with a period of 1–2 min (Fu et al., 2001; Hale et al., 2001) and is indirectly required for Z-ring formation by acting as a topological determinant that restricts the activity of MinC in the MinCD complex to the cell poles (Rothfield et al., 2005). Thus, PomZ and MinE display two different localization patterns suggesting that they function differently in the regulation of Z-ring formation.

Five lines of evidence suggest that PomZ may function as a positive regulator of Z-ring formation. First, in the absence of PomZ, Z-ring formation is strongly reduced. Second, PomZ localizes to the incipient division site before FtsZ. Third, PomZ localizes to the division site in the absence of FtsZ. Fourth, the localization pattern of PomZ is unlike that of other spatial regulators of Z-ring formation, which localize to the cell poles or oscillate between the cell poles, and inhibit Z-ring formation. Finally, FtsZ is not rate-limiting for Z-ring formation and cell division. Based on these in vivo data we suggest that PomZ is a spatial regulator of cell division that has two functions: First, PomZ is involved in identifying the incipient division site and localizes to this site before and in the absence of FtsZ. Second, PomZ is involved in recruiting FtsZ to the incipient division site as well as in stabilizing the Z-ring. PomZ interacts with FtsZ in pull-down experiments with M. xanthus cell extracts; however, we did not detect a direct interaction between PomZ and FtsZ in vitro using purified proteins. Our inability to detect a direct interaction between PomZ and FtsZ strongly suggests that proteins other than PomZ are involved in recruiting FtsZ to midcell and in stabilizing the Z-ring. In this model, PomZ would function as a positive spatial regulator of Z-ring formation. It was recently reported that in B. subtilis cells that lack the Min-system as well as the Noc-based nucleoid occlusion system, the Z-ring localizes correctly to midcell suggesting the existence of an additional mechanism for targeting FtsZ to the division site (Moriya et al., 2010; Rodrigues and Harry, 2012). It has been suggested that this mechanism may involve positively acting factor(s) (Moriya et al., 2010). These observations taken together with the recent observation that during spore differentiation in S. coelicolor positioning of FtsZ at the division sites is positively regulated by the SsgAB proteins (Willemse et al., 2011) suggest that positive regulation of Z-ring formation and cell division site positioning in bacteria is not restricted to differentiating cells but may be a more widespread regulatory mechanism also in growing bacteria.

FtsZ from M. xanthus displays highly concentration-dependent and cooperative GTP hydrolysis strongly suggesting that GTPase activity depends on oligomerization as in the case of other FtsZ proteins analysed. Consistently, FtsZ contains the residues involved in monomer interactions and GTP hydrolysis (Fig. S9). However, we were unable to detect FtsZ filament formation using standard methods for FtsZ filament detection under all conditions tested. Thus, FtsZ likely assembles filaments that are too short to be detected by EM and right angle light scattering. In the case of FtsZ of Thermobifida fusca, which was used as a proxy for FtsZ of S. coelicolor in in vitro work, the protein forms short filaments in the absence of SsgB and filament length is significantly increased in the presence of the positive spatial regulator SsgB (Willemse et al., 2011). In E. coli, B. subtilis and C. crescentus the known spatial regulators of Z-ring positioning act negatively to inhibit Z-ring formation throughout cells except at midcell. Interestingly, in these three bacteria, FtsZ alone in a GTP-dependent manner polymerizes in vitro to form long filaments. Consistently, in the absence of the negative spatial regulators extra Z-rings are formed (Wu and Errington, 2004; Bernhardt and de Boer, 2005; Thanbichler and Shapiro, 2006). For a system to function in which Z-ring positioning is positively regulated, Z-ring formation must be coupled to the correct localization of the positively acting spatial regulator(s). If that were not the case, FtsZ would form Z-rings independently of the positive regulator and the system would not function. Therefore, the apparent inability of purified M. xanthus FtsZ alone to assemble into long filaments in vitro are consistent with the genetic and cytological data suggesting that PomZ is involved in positively specifying the site of Z-ring formation.

PomZ localization is dynamic and cell cycle regulated. The cell cycle events and the mechanisms communicating these events to PomZ are not known. But our data suggest that initiation of chromosome replication/segregation causes the relocation of PomZ from patchy to off-centre localization; completion of chromosome replication/segregation causes PomZ relocation from off-centre location to midcell; and, completion of cytokinesis causes PomZ relocation from midcell to patchy. Several ParA proteins bind non-specifically to DNA in that way localizing over the nucleoid (Gerdes et al., 2010). Similarly, we suggest that PomZ localizes in the patchy pattern by binding non-specifically to DNA. We do not know how PomZ localizes to the off-centre; however, it is interesting to note that the cluster consistently localizes over the nucleoid suggesting that it may connect to a particular chromosomal region. Similarly, it is not known how PomZ localizes to the incipient division site.

Z-ring formation is spatially and temporally regulated to ensure that division gives rise to daughter cells of correct size and chromosome number. The cell cycle-dependent localization of PomZ suggests a mechanism for coupling cell cycle progression to Z-ring formation and cell division. Soj in B. subtilis is a nucleotide-dependent molecular switch with different properties and localization depending on whether ADP or ATP is bound and in which the nucleotide-bound state is cell cycle regulated (Murray and Errington, 2008; Scholefield et al., 2011). This information taken together with the indications that ATP hydrolysis is important for PomZ function and localization, lead us to propose that the nucleotide bound state of PomZ and, therefore, localization is cell cycle regulated. According to this model, only PomZ at midcell would be involved in stimulating Z-ring formation. Future challenges will be to determine how PomZ localization and activity is regulated in M. xanthus

Neither in WT nor in ΔpomZ cells were Z-ring formation and constrictions observed over the nucleoid suggesting that M. xanthus contains a system for nucleoid occlusion that functions independently of PomZ. In E. coli and B. subtilis nucleoid occlusion depends on the non-homologous DNA-binding proteins SlmA and Noc respectively. Noc is homologous to ParB proteins. The M. xanthus genome only encodes the ParB protein involved in chromosome segregation, suggesting that this protein is not involved in nucleoid occlusion. The M. xanthus genome does not encode a SlmA homologue. The protein(s) involved in nucleoid occlusion in M. xanthus, therefore, remains to be identified.

While SsgA and SsgB are only found in certain Actinobacteria, the Min-systems in E. coli and B. subtilis, the MipZ/ParB system in C. crescentus and the PomZ system in M. xanthus all share in common a P-loop ATPase of the ParA/MinD/Soj superfamily (Fig. S7B and C), i.e. MinD, MipZ and PomZ, suggesting that such an ATPase could have been at the heart of an ancient system for the spatial regulation of Z-ring positioning in a primordial organism that divided by binary fission. Several members of the ParA/MinD/Soj ATPases have been shown to be determinants of macromolecular localization in bacteria including, in addition to the Z-ring, chromosomes and plasmids (Gerdes et al., 2010), cytoplasmic protein complexes (Thompson et al., 2006), polar protein complexes (Ringgaard et al., 2011) and carboxysomes (Savage et al., 2010). Thus, in an evolutionary scheme an ancient ParA/MinD/Soj ATPase could have been a general spatial regulator of macromolecular localization, which was subsequently incorporated into systems regulating Z-ring positioning.

Experimental procedures

Bacterial strains, cell growth and strain construction

Myxococcus xanthus strains and plasmids are listed in Tables 1 and 2 respectively. E. coli strains were grown in LB broth or 2× YT medium for protein overexpression (Sambrook and Russell, 2001). Plasmids were propagated in E. coli TOP10 (F-, mcrA, Δ(mrr-hsdRMS-mcrBC), ϕ80lacZΔM15, ΔlacX74, deoR, recA1, araD139, Δ(ara, leu) 7679, galU, galK, rpsL, endA1, nupG) unless otherwise stated. M. xanthus strains were grown at 32°C in CTT media or on CTT agar plates with kanamycin and oxytetracycline (40 and 10 μg ml−1 respectively) (Søgaard-Andersen et al., 1996) unless otherwise stated. Cephalexin and rifampicin were added to a final concentration of 100 and 16 μg ml−1 respectively. The in-frame deletion of pomZ in SA3108 was generated as described (Shi et al., 2008) (position 66–930 deleted, co-ordinates relative to the start codon in pomZ) (Fig. S7A). The ftsZ-depletion strain was constructed by homologous recombination of pMR3165, which contains a construct for the in-frame deletion of ftsZ (in-frame deletion extends from position 3 to 1215 relative to the start codon of ftsZ), into the ftsZ gene by a single cross-over in the WT DZF1. Plasmid pNG10AftsZ, which contains ftsZ downstream of the cuoA promoter, was integrated by a single homologous recombination event at the cuoA promoter. To obtain the in-frame deletion of ftsZ, cells were grown in CTT medium supplemented with tetracycline to select for maintenance of pNG10AftsZ and 300 μM copper sulphate to induce expression of ftsZ from the cuoA promoter (Sanchez-Sutil et al., 2007), but in the absence of kanamycin to allow recombination within pMR3165. Correct clones were identified by PCR. JMCuftsZ accumulates FtsZ at WT levels in the presence of 300 μM copper sulphate. Construction of other plasmids is described in Supplementary Materials and Table S3. Plasmids containing Pnat or PpilA pomZ(–mCherry) or ftsZ(–mCherry/GFP) were integrated by site-specific recombination at the Mx8 attB site except for pKA53 and pKA58, which were integrated by homologous recombination at the carS locus. Strains containing plasmids integrated at the attB site, the carS or cuoA locus were constructed by electroporation of plasmids into the relevant strains. All strains were verified by PCR.

Table 1. Myxococcus xanthus strains used in this work
StrainRelevant characteristicsaReference
  1. aPlasmids mentioned in parentheses contain the indicated alleles and were integrated at the chromosomal Mx8 attachment site, at the carS locus in the case of pKA53 and pKA58, or at the cuoA locus in the case of pNG10AftsZ. In Pnat and PpilA constructs, pomZ and ftsZ alleles were transcribed from the relevant native promoter or the pilA promoter respectively. +++ indicates that the relevant protein is overproduced.
DK1622Wild typeKaiser (1979)
DZF1pilQ1Morrison and Zusman (1979)
SA3108ΔpomZThis study
SA3119ΔpomZ/Pnat-pomZ+ (pKA26)This study
SA3121ΔpomZ/PpilA-pomZ+++ (pKA19)This study
SA3131ΔpomZ/Pnat-pomZ–mCherry (pKA28)This study
SA3139ftsZ+/Pnat-ftsZ–mCherry (pKA32)This study
SA3142ΔpomZ, ftsZ+/Pnat-ftsZ–mCherry (pKA32)This study
SA3146ΔpomZ/Pnat-pomZD90A–mCherry (pKA43)This study
SA3179ΔpomZ, ftsZ+/Pnat-pomZD90A–mCherry (pKA58), Pnat-ftsZ–GFP (pKA51)This study
SA3159pomZ+, ftsZ+/Pnat-pomZ–mCherry (pKA53), Pnat-ftsZ–GFP (pKA51)This study
SA3165ftsZ+/PpilA-ftsZ+++ (pKA48)This study
SA3166ftsZ+/PpilA-ftsZ–mCherry+++ (pKA61)This study
SA3167ΔpomZ, ftsZ+/PpilA-ftsZ+++ (pKA48)This study
SA3168ΔpomZ, ftsZ+/PpilA-ftsZ–mCherry+++ (pKA61)This study
JMCuftsZpilQ1, ΔftsZ/PcuoA-ftsZ (pNG10AftsZ)This study
SA4107pilQ1, ΔftsZ/PcuoA-ftsZ (pNG10AftsZ), Pnat-pomZ–mCherry (pKA53)This study
SA4202pomZ+, Pnat-parB–YFP (pAH07)This study
SA4224ΔpomZ, Pnat-parB–YFP (pAH07)This study
SA4275ΔpomZ, Pnat-His6-pomZ (pAH71)This study
Table 2. Plasmids used in this work
PlasmidRelevant characteristicsaReference/source
  1. aPromoters for the expression of a particular construct are indicated together with the integration site on the chromosome.
  2. bIn-frame deletion extends from position 66 to 930 relative to the start codon of pomZ.
  3. cIn-frame deletion extends from position 3 to 1215 relative to the start codon of ftsZ.
pKA1Construct for in-frame deletion of pomZbThis study
pKA3Overproduction of His6-PomZThis study
pKA19PpilA-pomZ, Mx8 attBThis study
pKA26Pnat-pomZ, Mx8 attBThis study
pKA28Pnat-pomZ–mCherry, Mx8 attBThis study
pKA32Pnat-ftsZ–mCherry, Mx8 attBThis study
pKA43Pnat-pomZD90A–mCherry, Mx8 attPThis study
pKA48PpilA-ftsZ, Mx8 attBThis study
pKA51Pnat-ftsZ–GFP, Mx8 attBThis study
pKA53Pnat-pomZ–mCherry, carSThis study
pKA58Pnat-pomZD90A–mCherry, carSThis study
pKA61PpilA-FtsZ–mCherry, Mx8 attBThis study
pKA70Overproduction of FtsZThis study
pMM1Overproduction of His6-FtsZThis study
pMR3165Construct for in-frame deletion of ftsZcGarcia-Moreno et al. (2009)
pMT219Overproduction of FtsZccThanbichler and Shapiro (2006)
pNG10AftsZPcuoA-ftsZ, copper-dependent expression of ftsZ, cuoAThis study
pAH07Pnat-parB–YFP, Mx8 attBThis study
pAH71Pnat-His6-pomZ, Mx8 attBThis study

FtsZ depletion experiments

Strain SA4107 was grown at 32°C in CTT containing 300 μM copper sulphate and appropriate antibiotics. At an optical density at 550 nm of ∼ 0.5 cells were harvested (4000 r.p.m., 15 min, RT) and resuspended in copper-free CTT medium. Samples for Western blot and microscopy analysis were taken immediately before and during the depletion experiment. The culture was grown with continuous dilutions in pre-warmed copper-free CTT medium to maintain exponential growth.

Immunoblot analysis

Rabbit antiserum against His6-PomZ or His6-FtsZ were generated using standard procedures (Sambrook and Russell, 2001). PilC antibodies were described previously (Bulyha et al., 2009). Immunoblotting was performed using standard procedures (Sambrook and Russell, 2001) with polyclonal rabbit α-PomZ, α-FtsZ and α-mCherry antibodies and peroxidase-conjugated goat α-rabbit immunoglobulin G secondary antibodies as recommended by the manufacturer (Roche). For detection of GFP-tagged proteins, monoclonal α-GFP mouse antibody (Roche) and peroxidase-conjugated rabbit α-mouse immunoglobulin G secondary antibody (DakoCytomation) were used. Blots were developed using the Supersignal West Pico chemiluminescence reagent (Pierce). All gels were loaded with samples containing equal amounts of total protein.

Microscopy and data analysis

Cells from exponentially growing cultures were transferred to a slide with a 1.0% agarose pad buffered with TPM (10 mM Tris-HCl pH 7.6, 1 mM KHPO4 pH 7.6, 8 mM MgSO4) and covered with a coverslip. For DAPI staining, cells were incubated with DAPI (1 μg ml−1) for 10 min prior to microscopy. Cells were observed using a Zeiss AxioImager M1 fluorescence microscope equipped with a Cascade 1 K CCD camera (Photometrics). Images were processed using Metamorph (Molecular Devices). For time-lapse recordings, cells were treated as described except that the agarose pad was buffered with 0.25% CTT. Images were processed as described. Cell length (mean ± standard deviation) and fluorescence signals were quantified using Metamorph. Immunofluorescence microscopy was performed as described (Leonardy et al., 2007).

Analytical gel filtration

Experiments were carried out in buffer DB (50 mM HEPES/NaOH, pH 7.2, 50 mM KCl, 0.1 mM EDTA, 10% glycerol, 1 mM β-mercaptoethanol) (Thanbichler and Shapiro, 2006). Purified proteins were applied separately to a Superdex 200 5/150 GL size exclusion column equilibrated with buffer DB. The column was calibrated with ferritin (440 kDa), aldolase (158 kDa), ovalbumin (44 kDa) and ribonuclease (13.7 kDa) using the same buffer and conditions.

Nucleotide binding assays

Experiments were carried out in buffer DB on a temperature-controlled ISS PC1 spectrofluorometer (ISS, USA) with a cooled photomultiplier. Samples were allowed to equilibrate for 90 s between measurements. For mant-GTP and mant-ATP binding, the increase in anisotropy was determined at an excitation wavelength of 350 nm and an emission wavelength of 450 nm at 8°C in the presence of a constant mant-GTP or mant-ATP concentration and increasing protein concentrations. The data were fitted to a one site-saturating ligand binding model using Sigmaplot. To determine the binding stoichiometry of mant-GTP to FtsZ, 5 μM FtsZ was titrated with increasing concentrations of mant-GTP and anisotropy determined as described. To determine the half maximal effective concentration (EC50) with which mant-NTPs or TNP-ATP was competed by ATP or GTP, the NTP analogue was bound to a protein, and titrated with increasing concentrations of ATP or GTP. The EC50 value was calculated by fitting the data to a sigmoidal dose response model using Sigmaplot. To measure TNP-ATP binding, the increase in fluorescence upon protein binding was determined at the excitation wavelength 409 nm and emission wavelength 540 nm. The binding affinity was determined by fitting the data to a model for one site-saturating ligand binding using Sigmaplot. To approximate the binding stoichiometry of TNP-ATP to His6-PomZ, titration of 500 nM His6-PomZ with increasing concentrations of TNP-ATP was compared with the titration of 500 nM TNP-ATP with increasing concentrations of His6-PomZ.

ATP/GTP hydrolysis assays

Reactions were carried out in buffer DB supplemented with 10 mM MgCl2 and initiated by addition of 500 μM GTP and/or ATP and incubated at 30°C. Pi released was determined using a Malachite Green assay (Lanzetta et al., 1979) and the data fitted to the Hill equation using Sigmaplot.

Right angle light scattering experiments

Experiments were performed at 8°C, 20°C and 30°C with 10 μM FtsZ in buffer DB supplemented with 10 mM MgCl2 on the temperature-controlled ISS PC1 spectrofluorometer with a cooled photomultiplier. The excitation and emission wavelengths were set to 350 nm. Filament formation was started by the addition of nucleotides to a final concentration of 500 μM and the increase in light scattering was followed for 500 s. As indicated, 10 mM CaCl2 or 10 μM His6-PomZ with or without 500 μM ATP were included. For measurements at pH 6.5, a buffer with final concentrations [50 mM MES (pH 6.5), 50 mM KCl, 10 mM MgCl2, 1 mM β-mercaptoethanol, 7.5% glycerol] was used.

Negative-stain electron microscopy

Samples were prepared as described (Thanbichler and Shapiro, 2006). Briefly, proteins were pre-centrifuged for 15 min at 10 000 g, 4°C. M. xanthus FtsZ (3–10 μM), FtsZcc (3–10 μM) and His6-PomZ (1.5–10 μM) were incubated alone or together in 1:1, 2:1 or 1:2 ratios and in the absence or presence of 2 mM GTP and 1 mM ATP, 2 mM GMP-PNP or GTP-γ-S and 1 mM ATP, 2 mM GTP and 1 mM AMP-PNP for 20 min in buffer P (50 mM HEPES/NaOH, pH 7.2, 50 mM KCl, 10 mM MgCl2, 1 mM β-mercaptoethanol). Aliquots (10 μl) were withdrawn and applied to carbon-coated grids. After 1 min, the grids were washed with distilled water, negative stained as described with 2% (w/v) uranyl acetate as staining solution (Valentine et al., 1968; Hoppert and Holzenburg, 1998). Electron micrographs were taken with a Zeiss EM 902a transmission electron microscope (Zeiss, Oberkochen) operated in the conventional bright-field mode. Images were recorded with a 1 K digital camera (Proscan, Lagerlechfeld).

Pull-down experiments, SDS/PAGE and MALDI-MS/MS analysis

Three milligrams of purified His6-PomZ was applied to a Ni2+-NTA-agarose column (Qiagen). M. xanthus cell lysate was prepared as follows: 50 ml of exponentially growing WT cells were harvested, resuspended in buffer D (50 mM NaHPO4, 300 mM NaCl, pH 8.0) supplemented with 10 mM imidazole in the presence of proteases inhibitors (Roche) and lysed by sonication. Cell debris was removed by centrifugation at 5000 g for 10 min at 4°C and the cell-free supernatant applied to the Ni2+-NTA-agarose column with or without bound His6-PomZ. Bound proteins were eluted with buffer D supplemented with 500 mM imidazole. Proteins eluted from the columns were identified using two methods. In one method, proteins were separated by SDS-PAGE and gels stained with Coomassie Brilliant Blue R-250. Protein bands were excised from the gel and identified by MALDI-MS/MS. The data from this approach are shown in Fig. 6I. In the second method to identify proteins eluted from the Ni2+-NTA-agarose column with or without bound His6-PomZ, the eluate was directly exposed to trypsin digestion for 10 h at 22°C without prior separation by SDS-PAGE. Subsequently, samples were treated as described. The data from this approach are shown in Table S2. As an additional negative control, proteins from M. xanthus WT cell extract that bound to a Ni2+-NTA-agarose column with bound His6-RomR protein were identified. Bound proteins did not include FtsZ (data not shown)

Acknowledgements

We thank Monserrat Elías-Arnanz for plasmids and Kristin Wuichet for helpful discussions. The German Research Council within the framework of the Graduate School ‘Intra- and Intercellular Transport and Communication’, the Max Planck Society, the LOEWE Research Center for Synthetic Microbiology and the Spanish Government (grants to J.M.-D. CSD2009-00006 and BFU2009-07565) supported this work.

Ancillary