|Home | About | Journals | Submit | Contact Us | Français|
Cytokinesis in Gram-negative bacteria is mediated by a multiprotein machine (the divisome) that invaginates and remodels the inner membrane, peptidoglycan, and outer membrane. Understanding the order of divisome assembly would inform models of the interactions among its components and their respective functions. We leveraged the ability to isolate synchronous populations of Caulobacter crescentus cells to investigate assembly of the divisome and place the arrival of each component into functional context. Additionally, we investigated the genetic dependency of localization among divisome proteins and the cell cycle regulation of their transcript and protein levels to gain insight into the control mechanisms underlying their assembly. Our results revealed a picture of divisome assembly with unprecedented temporal resolution. Specifically, we observed 1) initial establishment of the division site, 2) recruitment of early FtsZ-binding proteins, 3) arrival of proteins involved in peptidoglycan remodeling, 4) arrival of FtsA, 5) assembly of core divisome components, 6) initiation of envelope invagination, 7) recruitment of polar markers and cytoplasmic compartmentalization, and 8) cell separation. Our analysis revealed differences in divisome assembly among Caulobacter and other bacteria that establish a framework for identifying aspects of bacterial cytokinesis that are widely conserved from those that are more variable.
Progression of the cell cycle culminates in the physical separation of the mother cell into two daughters through the process of cytokinesis. In the vast majority of bacteria, cytokinesis is mediated by the divisome, a multiprotein complex that assembles near midcell and drives invagination of the inner membrane (IM), synthesis and remodeling of the peptidoglycan (PG) cell wall and, in Gram-negative organisms, invagination of the outer membrane (OM). The tubulin-like GTPase, FtsZ, forms the structural basis of the divisome, acting as a scaffold for assembly of the rest of the division machinery (Goehring and Beckwith, 2005, Margolin, 2005) and generating constrictive force (Osawa et al., 2008, 2009). Over a dozen proteins assemble downstream of FtsZ and are essential for cytokinesis. The precise roles of these factors in executing division are mostly unknown, however there is evidence for the following general functions: interaction with and stabilization of the FtsZ ring (FtsA, ZapA, FzlA) (Addinall and Lutkenhaus, 1996, Beall and Lutkenhaus, 1992, Goley et al., 2010b, Gueiros-Filho and Losick, 2002, Martin et al., 2004, Ohta et al., 1997, Sackett et al., 1998), synthesis and remodeling of peptidoglycan (DipM, FtsW, FtsI/Pbp2B, AmiC) (Bernhardt and de Boer, 2003, Boyle et al., 1997, Costa et al., 2008, Daniel et al., 2000, Goley et al., 2010a, Henriques et al., 1992, Moll et al., 2010, Poggio et al., 2010, Weiss et al., 1997), coordination of division with chromosome segregation (FtsK/ SpoIIIE) (Wang et al., 2006, Wu and Errington, 1994, Yu et al., 1998), outer membrane invagination (Tol-Pal complex) (Gerding et al., 2007, Yeh et al., 2010a), and stabilization of interactions within the divisome (FtsQ/DivIB, FtsL, FtsB/DivIC, FtsN) (Addinall et al., 1997, Buddelmeijer et al., 2002, Chen et al., 1999, Ghigo and Beckwith, 2000, Katis et al., 1997, Martin et al., 2004, Moll and Thanbichler, 2009, Rowland et al., 1997, Sackett et al., 1998, Sievers and Errington, 2000).
The physical process of cell division must be tightly coordinated in time and space with other cell cycle events, such as cell growth, chromosome segregation, and cell differentiation. The need for this coordination is particularly evident in the dimorphic α-proteobacterium Caulobacter crescentus, in which the cell division site has emerged as a critical spatial landmark that receives signals from the cell poles that effect downstream cell cycle events. In newborn Caulobacter swarmer cells, FtsZ is localized at the new cell pole opposite the chromosomal centromere that is anchored at the old, flagellated pole (Thanbichler and Shapiro, 2006). The swarmer cell differentiates into a stalked cell by shedding its polar flagellum, building a stalk in its place, and initiating replication of the single, circular chromosome. Upon duplication, one copy of the parS chromosomal centromere is quickly segregated to the opposite cell pole, bringing with it a bound complex of the partitioning protein, ParB, and the inhibitor of FtsZ polymerization, MipZ (Thanbichler and Shapiro, 2006, Toro et al., 2008, Viollier et al., 2004). The completion of centromere segregation results in bipolar localization of MipZ. MipZ promotes depolymerization of FtsZ at the cell pole, resulting in the assembly of the FtsZ polymeric ring structure (the Z ring) at the site of lowest MipZ concentration in the cell: roughly at midcell (Thanbichler and Shapiro, 2006). In this way, the site of Z ring assembly, and therefore the future division site, is specified in coordination with chromosome segregation.
In addition to its central role in cell division, Caulobacter FtsZ also recruits proteins that direct cell elongation and cellular polarity. Prior to directing inward growth of PG during division, FtsZ organizes midcell-localized PG synthesis for the elongation phase of growth (Aaron et al., 2007). This is mediated, at least in part, by the midcell recruitment of MurG, which catalyzes the last step in lipid II (PG precursor) synthesis. The actin homolog, MreB, is also localized to midcell in an FtsZ-dependent manner, and is functionally implicated in PG synthesis and cell shape maintenance (Figge et al., 2004b, Gitai et al., 2004). Subsequently, through an unknown mechanism, there is a switch from elongation phase PG synthesis to division. IM fission and the resulting compartmentalization of the cytoplasm late in the division process is essential to specification of the developmental fate of the two daughters, since it produces biochemically distinct compartments that contain different concentrations and/or post-translational modifications of critical regulatory proteins (Judd et al., 2003, Matroule et al., 2004). Very late in the cell cycle, the new pole marker, TipN, is recruited to the division site, and TipN remains at the new pole after cell separation to direct polar development (Huitema et al., 2006, Lam et al., 2006). Outer membrane invagination and fission, which happens significantly later than inner membrane fission (Judd et al., 2005) and is dependent on the Tol-Pal complex (Yeh et al., 2010b), marks the completion of cytokinesis.
Understanding how the divisome functions to implement cell division and how its activity is regulated and integrated into the cell cycle requires detailed knowledge of the timing of its assembly relative to cell cycle progression. Comprehensive genetic analyses in Escherichia coli suggested a linear hierarchy of divisome assembly that might indicate a simple series of binary interactions that lead to formation of a functional divisome (Goehring and Beckwith, 2005). However, studies aimed at identifying protein-protein interactions among divisome components in both E. coli and Bacillus subtilis suggest a more complicated network of interactions (Buddelmeijer and Beckwith, 2004, Di Lallo et al., 2003, Goehring et al., 2005). Microscopy experiments that followed midcell localization of a subset of divisome proteins over time in E. coli and B. subtilis revealed a two-step process: FtsZ and its associated proteins localize to midcell first, and late divisome proteins assemble after a substantial maturation period (Aarsman et al., 2005, Gamba et al., 2009). Not all divisome proteins were included in those studies, however, and they were not performed using synchronized vegetatively growing cells.
Caulobacter is more easily synchronized than E. coli or B. subtilis, and is thus ideally suited to high-resolution temporal studies. Moreover, its cell cycle and developmental programs are well characterized, allowing us to place the arrival of each protein into functional context with respect to cell cycle events. Several cell division proteins, including FtsZ, FtsA, FtsQ, DipM, and FzlA, have been reported to be cell cycle regulated at the transcript and/or protein levels in Caulobacter, indicating that divisome assembly and function is tightly regulated in time (Goley et al., 2010a, 2010b, Martin et al., 2004, Quardokus et al., 1996). Scattered reports of the cell cycle dependent localization of a number of Caulobacter divisome proteins have been published and are mostly consistent with the two-step assembly pathway reported in other organisms (Aaron et al., 2007, Costa et al., 2008, Goley et al., 2010a, 2010b, Thanbichler and Shapiro, 2006, Moll and Thanbichler, 2009, Moll et al., 2010, Poggio et al., 2010, Wang et al., 2006, Yeh et al., 2010a). To date, however, a comprehensive analysis of the assembly of all divisome and division-site localized proteins has not been reported. Our goal in the current study was to systematically analyze the temporal regulation of 19 divisome and division-site-localized proteins over the course of the cell cycle in Caulobacter. To this end, we analyzed their temporal localization patterns and performed extensive analysis of the genetic dependency of their subcellular localization. Transcript levels and protein levels over the course of the cell cycle were determined for a subset of the division proteins. Our results indicate that proteins assemble at the Caulobacter division site in a series of at least 7 modules that can be temporally correlated with cellular events like chromosome segregation, cell elongation, initiation of invagination, and cell separation, leading to insights into the functions of individual divisome components. Surprisingly few proteins rely solely on any other protein for localization, consistent with a complex network of interactions underlying divisome structure.
A number of proteins have been shown previously to localize to the cell division site in Caulobacter and to participate in cell growth and morphology specification (MurG, MreB, DipM (Figge et al., 2004a, Gitai et al., 2004) Goley et al., 2010a, Moll et al., 2010, Poggio et al., 2010)), cell division (FzlA, DipM, FtsZ, FtsA, FtsK, FtsQ, FtsI, FtsN, Tol-Pal complex (Quardokus et al., 1996, Martin et al., 2004, Thanbichler and Shapiro, 2006, Wang et al., 2006, Moll and Thanbichler, 2009, Goley et al., 2010b, Yeh et al., 2010)), or cell polarity specification (TipN, MreB (Huitema et al., 2006, Lam et al., 2006)), or to have unknown functions (FzlC, KidO (Goley et al., 2010b, Radhakrishnan et al., 2010)) (Fig. 1A). In addition to these previously characterized proteins, we identified and cloned the Caulobacter homologs of the ZapA, FtsE/X, FtsL, FtsB, and FtsW cell division proteins (Table S1) and determined that fluorescent fusions of each localize to the division site.
For each of the 19 division-site localized proteins listed above, we aimed to determine the time of appearance of each fluorescently tagged protein at the incipient division site in synchronized cell populations. In order to compare localization timing between strains, we integrated an inducible copy of each fluorescent fusion at the chromosomal vanA or xylX locus for vanillate or xylose-inducible expression, respectively, in a strain background bearing mipZ-cerulean as the only copy of mipZ at its native chromosomal locus (Fig. 1B). The dynamics of MipZ localization are well characterized and provide an accurate marker for the timing of the initiation of chromosome segregation (Thanbichler and Shapiro, 2006). Thus, the inclusion of MipZ-Cerulean in each strain enabled comparison of cell cycle progression between strains (Fig. 1C). We synchronized each strain (Evinger and Agabian, 1977), suspended the isolated swarmer cells in rich PYE liquid media for growth at 28°C, withdrew samples at 10-minute intervals, and imaged phase contrast, MipZ-Cerulean, and the fluorescent fusion protein of interest at each time point. Under these conditions, the cell cycle was completed in ~90 minutes. At each time point in each strain we quantified: 1) the localization of MipZ (percentage of cells with monopolar vs bipolar MipZ), 2) the localization of the protein in question (percentage of cells with polar, diffuse, midcell, or other localization) 3) and invagination of the cell envelope (both the percentage of cells with an invagination at the incipient division site visible by manual inspection of phase contrast images and the degree of envelope invagination in pixels) (Fig. 1C). Each strain was independently synchronized and analyzed at least twice.
Representative images from the synchrony of strain EG490, bearing mipZ-cerulean at the mipZ locus and ftsZ-yfp at the vanA locus are presented in Fig. 1D and quantification of the synchronies of that strain is presented in Fig. 1E. By 10 minutes post-synchrony, most cells exhibited bipolar MipZ localization, indicating that the chromosomal centromere was duplicated and segregated, and by 30 minutes essentially all cells had bipolar MipZ. MipZ remained bipolar until cell separation at ~90 minutes, when we observed a marked decrease in bipolar MipZ localization owing to the presence of daughter cells that had not yet duplicated and/or segregated their centromeres. We frequently observed very late predivisional cells that had 3 MipZ foci: one each at the extreme cell poles, and one near midcell (18.7 ± 3.9% of cells at t = 80 min, n = 6 synchronies, 210-557 cells each). We interpret these to be cells wherein the cytoplasm has been compartmentalized, allowing DNA replication and centromere segregation in the stalked compartment prior to the completion of splitting of the PG and OM fission. We observed visible invagination of the envelope, marking the switch from the elongation mode of growth to the division mode, in the majority of cells by 60 minutes post-synchrony. This, too, decreased upon cell separation at ~90 minutes, when daughter swarmer and stalked cells lacking visible invaginations were plentiful.
In newborn swarmer cells of strain EG490, FtsZ-YFP was observed at the new cell pole opposite the MipZ-Cerulean focus, and assembled into a loose band at midcell shortly after bipolarization of MipZ, as reported previously (Fig. 1D, E and (Thanbichler and Shapiro, 2006)). The Z ring became more focused by 30 minutes post-synchrony, and FtsZ remained at midcell until cell separation. At 90 minutes, we observed cells with monopolar MipZ and an FtsZ focus at the opposite pole, as well as daughter stalked cells that exhibited bipolar MipZ and a loose band of FtsZ at midcell. As described here for FtsZ, our strain construction strategy allowed us to assess the timing of localization of all 19 division-site localized proteins relative to chromosome segregation and envelope invagination (Fig. 2).
We repeated the analysis described above for 18 division-site localized proteins in addition to FtsZ. Quantification of the timing of bipolar localization of MipZ, initiation of invagination (Fig. 1F, ,4B)4B) and half-maximal invagination (Fig. 4B, S1) in each strain analyzed revealed remarkably constant timing of these events. This indicated that the timing of cell cycle progression of each strain was similar and enabled direct comparison of the timing of midcell localization of division-site localized proteins across strains.
Each division-site localized protein exhibited a characteristic time-dependent localization pattern (Fig. 2). Of the 19 proteins, 14 began the cell cycle at the new cell pole, while the rest (FzlA, FtsE, MurG, FtsA, and KidO) were diffuse in swarmer cells (Fig. 2, ,3).3). Of the polar proteins, 8 were displaced from the cell pole prior to accumulating at the division plane and 6 localized to both the pole(s) and midcell (Fig. 2, ,3).3). Each protein arrived at midcell with characteristic timing (ranging from ~10 to 70 minutes into a 90 minute cell cycle), and most remained at midcell until the completion of cell division. A clear exception, however, is MreB, which was dispersed from midcell into a punctate distribution ~25 minutes before cell separation (Fig. 2).
To assign an order of assembly of division-site localized proteins, we calculated the percentage of cells with a midcell band or focus of each protein at each time point (for representative data, see Fig. 4A, top). We then fit a curve to those data and found the time at which half-maximal midcell localization was achieved for each (Fig. 4A, bottom). We performed similar curve fitting to calculate the timing of bipolar MipZ localization, initiation of invagination, and half-maximal degree of invagination. This analysis revealed a series of stages and transitions in divisome assembly and associated events (Fig. 4B). First, MipZ becomes bipolar, followed closely by the assembly of a loose band of FtsZ near midcell. Ten to sixteen minutes later, the FtsZ binding proteins ZapA, FzlC, FtsE, and FzlA first appear at midcell. Concomitant with FzlA localization at midcell, 22 to 25 minutes into the cell cycle, proteins involved in cell growth and morphology specification (DipM, MurG, and MreB) localize at midcell. Next, an apparent transition to divisome assembly occurs wherein TolQ (representing the Tol-Pal complex) and FtsA are recruited to the division site at 30 and 36 minutes post-synchrony, respectively. Subsequently, core divisome proteins (FtsN, FtsQ, FtsI, FtsK, and FtsL) arrive. FtsW and FtsB arrive 5-10 minutes after the core set of divisome proteins, immediately preceding initial invagination of the cell envelope. KidO then arrives at midcell, followed by dispersal of MreB into a patchy distribution and midcell localization of TipN.
The analysis described above suggests that each division site localized protein has a prescribed time of arrival at midcell. There are several unavoidable caveats to our experimental approach, however, that are important to keep in mind. First, inherent to the use of fluorescent fusions is the probability that some cleavage of the fluorescent moiety will occur, producing diffuse background signal that could conceivably obscure a weak localized signal. This could lead to defining artificially late arrival times for those strains in which a high background signal is observed. Although it is not possible to rule this out completely, most of the proteins we imaged showed robust localized signal once they arrived at midcell, indicating that the fusion was intact and capable of localizing. Moreover, we took care to define proteins as localized to midcell when even the slightest enrichment there was observed. Second, although each of the fluorescent fusions analyzed localizes to the cell division site with reproducible timing, many are not fully functional, suggesting that the addition of a bulky tag interferes with one or more essential functions of the protein and could affect its timing of localization. Finally, the use of inducible fluorescent fusions (as opposed to fusions expressed from the native gene locus) can result in non-native protein levels and/or timing of expression, either of which could affect the time of recruitment to the division site. However, using inducible fusions allowed us to ensure normal cell morphology and equivalent timing of cell cycle progression from strain to strain. In the following two sections we address these issues and obtain further evidence to support the assembly pathway outlined above.
As introduced above, our experimental approach circumvents transcriptional regulation of divisome assembly since the genes encoding the fluorescent fusions were expressed constitutively from inducible promoters. Transcript levels of hundreds of genes in Caulobacter are known to vary over the course of the cell cycle and transcriptional regulation can be critical in specifying the timing of cell cycle events (Laub et al., 2000). To ensure that our expression strategy did not alter the normal timing of division site protein localization, we attempted to generate strains bearing fluorescent fusions to a subset of the genes encoding division-site localized proteins at their chromosomal loci under the control of their own promoters in the mipZ-cerulean background. We selected proteins representing each stage of divisome assembly for analysis (FtsA, ZapA, FzlA, FzlC, FtsE, DipM, MurG, FtsK, FtsQ, FtsL, FtsI, and FtsW). Of these, we were unable to recover strains bearing fluorescent fusions to FtsA, FtsL, or FtsQ, indicating that the tags interfere with the essential functions of these proteins. However, we obtained strains producing native-tagged versions of ZapA, FzlA, FzlC, FtsE, DipM, MurG, FtsK, FtsI, and FtsW. We assessed cell cycle progression of each of these strains by following MipZ-Cerulean localization as described above and found that, although viable, the strains with mCherry-FzlC, DipM-mCherry, FtsK-mCherry and Venus-FtsI produced as the only copy from the native gene locus progressed more slowly through the cell cycle than the others, so we did not include them in our analysis. These results indicate that most cell division proteins in Caulobacter do not tolerate large protein fusions without adverse effects on cell division and viability, perhaps reflecting the importance of protein-protein interactions within the division machinery.
Nevertheless, strains bearing ZapA-mCherry, mCherry-FzlA, Venus-FtsE, MurG-mCherry, or mCherry-FtsW produced from their native loci were viable and progressed through the cell cycle with normal timing. We therefore determined the timing of midcell localization of these proteins and compared it to their timing of localization when expressed from the inducible xylX promoter. The cell cycle dependent localization and arrival times of the native and inducible versions were indistinguishable for MurG and FtsW (Fig. 5). Native Venus-FtsE localized to the division site with the same timing as the inducible version, however the protein produced by native expression localized to the cell pole in swarmer cells, whereas the induced version did not (Fig. 3, ,5).5). The native tagged versions of FzlA and ZapA localized to midcell slightly earlier than the inducible versions, but in the same time window (i.e. between initial assembly of FtsZ at midcell and recruitment of proteins involved in PG synthesis). Thus, for the proteins tested, expressing division-site localized proteins using an inducible promoter did not appreciably change the timing of their arrival at midcell.
To further assess the robustness of the divisome assembly order determined above and to investigate unanticipated results from our analysis, we selected pairs of proteins to image in the same cells. First, we selected three pairs of proteins that our results indicate arrive simultaneously: ZapA with FtsE, ZapA with FzlA, and FtsI with FtsK. We generated strains bearing native tagged versions of each, synchronized them, and determined the percentage of cells with neither protein at midcell, both proteins at midcell, and either one or the other at midcell at a time point when about half the cells had both proteins at midcell. For ZapA/FtsE, we observed a small percentage of cells (1.5%) with an apparent enrichment of FtsE at midcell but no ZapA, and a slightly higher percentage (5%) with ZapA at midcell but no FtsE. This is consistent with our determination that native tagged ZapA arrives at midcell slightly ahead of native tagged FtsE (Fig. 6A). In the case of ZapA/FzlA double labeling, we never observed cells that had only one of these proteins at midcell (Fig. 6B), indicating that they do indeed arrive simultaneously. The same was true of FtsI/FtsK, although in that strain we found low percentages of cells that had FtsI at midcell but not FtsK (4.8%), and FtsK at midcell but not FtsI (3.3%) (Fig. 6C).
Our temporal analysis revealed a few surprises in light of previous work. Specifically, in E. coli, FtsW was reported to be required for recruitment of FtsI to the division site, but we find that FtsI arrives first of the two in Caulobacter. We imaged native Venus-FtsI with vanillate-induced mCherry-FtsW (the double native-tagged strains was viable, but slightly filamentous, so it was not used) and found that, as suggested by the analysis in separate strains, FtsI was often observed at midcell when FtsW was absent (39.8% of cells 40 minutes post-synchrony) (Fig. 6D). FtsW was never observed to arrive prior to FtsI, confirming our previous result. Also contrary to studies in E. coli and B. subtilis, which found that FtsA arrives at midcell at the same time as FtsZ and other FtsZ-binding proteins, we observed a significant delay in FtsA midcell recruitment in Caulobacter, as described previously (Moll and Thanbichler, 2009). To further investigate this observation, we imaged double labeled strains bearing native ZapA-mCherry (representing the early-arriving FtsZ-binding proteins) with xylose-induced YFP-FtsA and mCherry-FtsW (a late-arriving divisome protein) with xylose-induced YFP-FtsA. We observed a midcell band of ZapA without midcell FtsA in a significant fraction (17.5%) of cells, and never observed FtsA at midcell before ZapA, indicating that it is recruited well after initial Z ring assembly (Fig. 6E). Additionally, we found that 33.1% of cells had FtsA at midcell before FtsW, whereas FtsW was observed at midcell before FtsA in fewer than 1% of cells (Fig. 6F). Collectively, these data provide further support for the order of assembly established by our initial analysis.
The temporal order of divisome protein localization we observed is consistent with a series of protein-protein interactions wherein each protein recruits the next to arrive in time. Extensive genetic probing of the dependency relationships among division-site localized proteins in E. coli support the existence of such a serial recruitment mechanism (Buddelmeijer and Beckwith, 2002). For example, in E. coli, FtsZ is required to recruit FtsA, which is required to recruit FtsK, which is required to recruit FtsQ, etc. To test if this is the case in Caulobacter, we used a set of strains with deletions of genes encoding non-essential divisome proteins and inducer-dependent expression of genes encoding essential divisome proteins (Table SI). We then introduced fluorescent fusions of other division-site localized proteins into these deletion or depletion strains and asked if the fluorescent fusion was still localized to midcell foci in the absence of the deleted or depleted protein. When possible, we used depletion times guided by previously published studies that determined the time at which the depleted protein was completely undetectable by western blotting (e.g. for FtsZ, FtsA, FzlA, DipM, TolA, Pal, and FtsN). The ΔzapA and ΔftsB deletion strains were confirmed by PCR and, in the case of ΔzapA, western blotting with ZapA antisera (data not shown). For the depletion strains for which we lack antibodies against the depleted protein (FtsE, FtsX, FtsL, FtsK, and FtsW), we analyzed protein localization at a depletion time when we observed a severe cell division defect, but before we detected significant cell death. Our analysis of the depletion strains for which antibodies are available to confirm depletion indicates that it is generally safe to assume that the depleted protein is undetectable in this phenotypic time window.
We used conservative criteria to define protein “localization”. If the fluorescent fusion was observed even weakly in midcell foci in the absence of the depleted or deleted protein, we defined it as localized. However, proteins that weakly localized in a given depletion strain are indicated as such in Table 1. On the other hand, if the fluorescent fusion was diffuse or dispersed throughout the cytoplasm or cell envelope, it was defined as delocalized. Our results therefore indicate cases where there is absolute requirement for localization of a given protein on another, and do not address in detail those cases in which localization is only partially disrupted. Table 1 summarizes these results along with those of previously published experiments addressing the genetic dependency of division-site protein localization in Caulobacter.
We were surprised to find that, in contrast to the reported situation in E. coli, very few Caulobacter proteins are strictly required for the localization of any others. The exceptions were 1) FtsZ, which was absolutely required for the midcell localization of every protein we tested (Table 1 and Fig. 7A), 2) components of the Tol/Pal complex, which were previously shown to be required for normal localization of TipN, 3) FtsX, which was required for the localization of the FtsE protein with which it forms a heterodimeric ABC transporter complex (Table 1 and Fig. 7D), and 4) FtsL, which was absolutely required for localization of FtsB and FtsQ and without which FtsI and FtsW were localized only weakly to midcell foci (Table 1 and Fig. 7E-I). None of the fluorescently tagged division-site localized proteins we tested required the presence of FtsA to localize to midcell foci, (Table 1 and Fig. 7B), in contrast to E. coli, where FtsA is very early in the hierarchical dependency for divisome protein localization. Moreover, FtsW was not absolutely required for FtsI to localize to midcell foci in Caulobacter (Table 1 and Fig. 7J). This differs from the scenario in E. coli where FtsW is required for localization of FtsI to the division plane, but is consistent with our observation that FtsI localized prior to FtsW in Caulobacter. Collectively these results indicate that, of the cell division proteins we were able to delete or deplete, FtsZ and FtsL perform key roles in recruiting other factors to midcell and that the assembly of the division machinery does not involve a strictly linear array of protein-protein interactions.
Previous work has shown that the transcript and protein levels of FtsZ (Kelly et al., 1998, Sackett et al., 1998), FzlA (Goley et al., 2010b), FtsA, FtsQ (Martin et al., 2004, Sackett et al., 1998), DipM (Goley et al., 2010a, Moll et al., 2010), and KidO (Radhakrishnan et al., 2010) vary over the cell cycle. Such cell cycle regulation of the concentrations of key components of the divisome could provide a robust mechanism for ensuring that each protein arrives at midcell at the functionally appropriate time. To further address the potential importance of this type of regulation for the observed temporal order of assembly of the divisome, we analyzed the transcript levels of the 19 division-site localized proteins included in our study. Analysis of Affymetrix data from synchronized cells grown in minimal M2G media (McGrath et al., 2007) indicated that transcript levels of 10 of the 19 genes are very strongly cell cycle regulated (Fig. 8A). ftsZ transcript levels peak in stalked cells, whereas ftsW, murG, ftsI, ftsQ, fzlA, ftsB, ftsA, ftsK, and kidO peak in pre-divisional cells (Fig. 8A and (Kelly et al., 1998, Sackett et al., 1998, Goley et al., 2010b, Radhakrishnan et al., 2010)). FtsZ, FtsQ, FzlA, FtsA, and KidO protein levels were previously shown to be cell cycle regulated, with the peak in concentration of each protein shortly after the peak in concentration of its transcript (Kelly et al., 1998, Martin et al., 2004, Goley et al., 2010b, Radhakrishnan et al., 2010). We attempted to follow protein levels of FtsW, FtsI, FtsB, and FtsK by Flag-M2 epitope tagging the native copy of each gene to determine if the observed cell cycle variation in transcript levels resulted in variation of protein levels. Given the apparent importance of FtsL in recruiting other factors to midcell (Table 1), we also followed the levels of FtsL-M2 protein over the course of the cell cycle, even though its transcript was not strongly cell cycle regulated. Unfortunately we were unable to detect FtsW-M2, FtsI-M2 or FtsB-M2 by immunoblotting of whole cell lysates. However, immunoblot analysis of FtsK-M2 and FtsL-M2 over the course of the cell cycle showed that they are mildly cell cycle regulated, reaching their highest levels in late pre-divisional cells (Fig. 8B). We conclude that a subset of division-site localized proteins are strongly regulated at the transcript and protein levels, notably the previously described cases of FtsZ, FtsA, FtsQ, and KidO. FtsL, FtsK, FzlA, and DipM exhibit weak, but reproducible, cell cycle variation in protein levels.
By analyzing the dynamic subcellular localization of 19 divisome components as a function of the cell cycle, we have shown that the assembly of divisome and division-site localized proteins in Caulobacter occurs in a series of stages, shown schematically in Figure 9. In combination with the genetic localization dependency we determined for cytokinetic ring assembly, we now have a framework to investigate the underlying molecular mechanisms of midcell recruitment and functions of divisome proteins during cellular growth and division.
Previous studies of a subset of cell division proteins in E. coli and B. subtilis showed that their midcell assembly follows a two-step pathway (Aarsman et al., 2005, Gamba et al., 2009). Our comprehensive analysis of synchronized cell populations has allowed us to define seven stages of recruitment of proteins to the incipient division site in Caulobacter (Fig. 9). First, the site of cell division is specified by segregation of MipZ with the chromosomal centromere and its associated ParB to the new cell pole. FtsZ-YFP localizes in a loose midcell assembly just after bipolarization of MipZ. Five to 10 minutes later, early FtsZ binding proteins appear at midcell (FzlA, ZapA, FtsE, and FzlC) and FtsZ localization is stabilized at the incipient division site (Fig. 9 and Thanbichler and Shapiro, 2006, Costa et al., 2008). The underlying cause for the observed delay between initial midcell assembly of FtsZ and recruitment of early interacting partners is unclear since at least ZapA and FzlA are present in swarmer cells (YCY, HHM, and LS, unpublished observation, and (Goley et al., 2010b)). It is possible that induction of FtsZ-YFP from the vanA locus in newborn swarmer cells increases the FtsZ concentration enough above native levels to prematurely stimulate midcell FtsZ polymerization. Thanbichler and Shapiro (2006) demonstrated that, using induction conditions similar to those used here, FtsZ-YFP comprises only a small fraction of the total FtsZ in the cell. However, even a moderate increase in FtsZ concentration could tip the scales in favor of premature FtsZ polymerization. If that is the case, the timing of midcell localization of ZapA (the earliest-localizing FtsZ-binding protein) likely reflects the true time of initial FtsZ assembly at midcell. When imaged in the same cells, we rarely observe cells that have a midcell focus of FtsZ, but no focus of ZapA or FzlA (EDG, YCY, and LS, unpublished observations), suggesting that FtsZ and its early binding proteins arrive at the division site close in time. Alternatively, native FtsZ may, indeed, assemble at midcell before its early interacting partners. If that is true, FtsZ protein levels may need to reach a critical threshhold to form enough polymeric FtsZ to enrich ZapA and other FtsZ-binding proteins at midcell. ftsZ transcription is regulated by at least two master regulators: CtrA, which is present in swarmer and late pre-divisional cells, represses ftsZ expression (Kelly et al., 1998), and DnaA, which transiently accumulates in stalked cells, activates ftsZ transcription (Hottes et al., 2005). Thus, FtsZ levels begin to increase dramatically in early stalked cells (Quardokus et al., 1996), and might reach a critical threshold thereafter for recruitment of the FtsZ-binding proteins, ZapA, FzlA, FtsE, and FzlC. Although we cannot conclusively state that native FtsZ assembles at midcell before its early binding partners, it is clear that the cell is competent to initiate assembly of the Z ring from the time of bipolarization of MipZ, provided enough FtsZ is present.
After recruitment of early FtsZ binding proteins, factors involved in peptidoglycan remodeling are recruited to midcell (Fig. 9). Their appearance parallels the timing of medial peptidoglycan elongation (Aaron et al., 2007) and requires FtsZ (Table 1). We did not observe a requirement for any of the early FtsZ-binding proteins for the recruitment of MurG, MreB, or DipM to midcell (Table 1). Moreover, these proteins are recruited efficiently to the extended constrictions of cells overexpressing GTPase defective FtsZ, indicating that their localization is mediated by direct interaction with either FtsZ or a highly abundant, possibly non-protein, factor (Goley et al., 2010b). We favor the second possiblility, as we failed to detect a direct interaction between MurG or DipM and FtsZ using purified proteins in in vitro FtsZ co-sedimentation assays (EDG and LS, unpublished). Moreover, DipM resides in the periplasm and is recruited to midcell via its peptidoglycan-binding LysM motifs (Goley et al., 2010a, Moll et al., 2010, Poggio et al., 2010), suggesting the presence of a positive-feedback mechanism for recruitment of these factors to midcell via the PG synthetic machinery, its substrate, or the cell wall, itself. Perhaps MreB and FtsZ interact directly, however this is technically challenging to assess using purified components. Even if MreB and FtsZ do interact, an alternative mechanism must be in place to recruit MurG and DipM to the division site since both localize to midcell independently of MreB (Aaron et al., 2007, Goley et al., 2010a, Poggio et al., 2010).
The recruitment of TolQ to the division plane followed the arrival of peptidoglycan remodeling proteins (Fig. 9). This suggests that local changes in the cell envelope resulting from peptidoglycan synthesis at that site support midcell enrichment of Tol-Pal components. Recruitment of the Tol-Pal complex during the elongation phase of growth is consistent with its role in establishing and maintaining dynamic contacts between the outer membrane, inner membrane, and cell wall during cellular growth (Yeh et al., 2010, Anwari et al., 2010). It is interesting to note that the Tol-Pal complex in E. coli is thought to arrive at the division site very late, as it requires the presence of the late-arriving FtsN division protein (Gerding et al., 2007). Caulobacter TolQ does not require FtsN for localization (Yeh et al., 2010) and both TolQ and FtsN arrive earlier in the pathway of divisome assembly in Caulobacter than would be predicted from E. coli studies.
In contrast to other bacteria (Aarsman et al., 2005, Gamba et al., 2009), there is a significant delay between Z ring assembly and arrival of FtsA to the Caulobacter division plane (Fig. 2, ,44 and (Moll and Thanbichler, 2009)). FtsA binds to FtsZ (Wang et al., 1997, Din et al., 1998) and is thought to play a key role in tethering FtsZ to the membrane (Pichoff and Lutkenhaus, 2005), so this observation leads to several questions. First, how is its recruitment delayed 20 minutes behind arrival of other FtsZ-binding proteins to the division site? One trivial explanation is that the fluorescent fusions used (which cannot complement loss of FtsA and are therefore not fully functional) lead to artificially late recruitment by interfering with its interaction with FtsZ. However, we observed the same late arrival time with N- and C-terminal fluorescent fusions to FtsA (Fig. 4B), making it less likely that the tag is specifically interfering with its interaction with FtsZ. Moreover, studies aimed at determining the time of midcell arrival of E. coli or B. subtilis FtsA used similar fluorescent fusions, but found that it arrives early. Since the interaction mechanism of FtsA with FtsZ is likely to be conserved, this also argues for a specific mechanism at work in Caulobacter to delay FtsA assembly at midcell. If that is the case, transcriptional and proteolytic regulation of FtsA may ensure that it is not present in sufficient quantities to interact with FtsZ and localize to the division site until just before the core divisome proteins are recruited. Our experimental design circumvented transcriptional regulation of ftsA, since the fluorescent fusion was expressed constitutively from the xylose-inducible xylX locus. However, it is possible that the fluorescent FtsA fusion protein was unstable until the time of FtsA recruitment. The stability of the FtsA protein varies over the cell cycle, with it being stabilized in stalked and pre-divisional cells as compared to swarmers (Martin et al., 2004). Alternatively, it may be that FtsA has lower affinity for FtsZ than the early FtsZ-binding proteins and requires a higher concentration of FtsZ filaments at midcell to be recruited there, particularly in the presence of competing factors.
The second question that the late FtsA recruitment raises is, assuming FtsZ must be tethered to the inner membrane to form a stable Z ring, how is this accomplished prior to the arrival of FtsA? The only obvious candidate for an additional membrane-tethering factor is the FtsE/X heterodimer, since in E. coli FtsE binds to FtsZ while FtsX is imbedded in the inner membrane (Corbin et al., 2007). FtsE/X is highly conserved and may play a redundant membrane tethering function with FtsA and (in γ-proteobacteria) ZipA. Alternatively, an unidentified membrane-tethering protein may serve to bring FtsZ to the membrane early in the cell cycle. It is interesting to note that, in contrast to E. coli, Caulobacter cells depleted of FtsA show substantial shallow constriction of the cell envelope (Fig. 7B and (Martin et al., 2004)). This indicates that the divisome can still assemble and operate to initiate constriction in the absence of FtsA, albeit imperfectly, and further supports the existence of one or more additional membrane-tethering factors for FtsZ.
Subsequent to the arrival of FtsA, five core divisome proteins (FtsQ, FtsK, FtsI, FtsL, and FtsN) appear enriched at midcell (Fig. 9). The delay between early FtsZ-binding protein recruitment and core divisome assembly was 25% of the cell cycle (Fig. 4). The time difference between these two clusters is comparable to those observed in E. coli and B. subtilis (Aarsman et al., 2005, Gamba et al., 2009), indicating that the assembly of core divisome components is triggered by a conserved mechanism. Although it was tempting to speculate that the arrival of FtsA at the division site triggers recruitment of these late factors, FtsA was surprisingly not required for the localization of any other factor to midcell foci (Table 1). FtsQ transcript and protein levels are also highly cell cycle regulated, however, so FtsQ could play a critical role in stabilizing the core divisome in a temporally regulated manner (Sackett et al., 1998, Martin et al., 2004). In E. coli and B. subtilis, FtsQ has been shown to form a complex with FtsL and FtsB that is important for divisome stability (Daniel and Errington, 2000, Buddelmeijer et al., 2002, Buddelmeijer and Beckwith, 2004, Goehring et al., 2005, Daniel et al., 2006, Goehring et al., 2006, Gonzalez and Beckwith, 2009). Unlike E. coli or B. subtilis FtsB/DivIC, we found that the Caulobacter FtsB homolog (encoded by CC1725) is not essential for viability (Table 1). FtsB/DivIC and FtsL are not essential for viability in Streptomyces coelicolor (Bennett et al., 2007), nor is FtsQ/DivIB essential in Streptococcus pneumoniae (Le Gouellec et al., 2008). Components of the FtsQLB complex are, therefore, not universally required for viability in organisms possessing homologs of these proteins. Instead, their functions may be more or less important depending on the growth conditions. Nevertheless, our finding that FtsL is required for efficiently recruiting FtsQ, FtsB, FtsI, and FtsW to the division site supports the idea that a complex of at least FtsL and FtsQ is central to divisome integrity in Caulobacter.
Visible invagination of the envelope, reflecting the switch from the elongation to constriction phase of growth, occurs just after assembly of the core divisome, coincident with the arrival of FtsW and FtsB. Since FtsB is not essential in Caulobacter, we hypothesize that FtsW plays a critical role in triggering constriction via activation of the peptidoglyan remodeling machinery. In E. coli, FtsW is important for localizing the FtsI transpeptidase to the division site. However, we find that FtsI arrives prior to FtsW at the division site in Caulobacter and that FtsI does not require FtsW to localize to midcell (Fig. 4B, ,6D,6D, ,7J).7J). FtsW has also been postulated to be a lipid II flippase (Ikeda et al., 1989, Boyle et al., 1997), an activity that is expected to be essential for implementing the division mode of PG synthesis.
The arrival of KidO at midcell occurs slightly later than the first visible invagination of the envelope, consistent with the proposed role of KidO in promoting Z-ring disassembly during constriction (Radhakrishnan et al., 2010). KidO is presumably recruited via direct interaction with FtsZ, but how KidO localization is delayed until after constriction occurs is unclear. It may be that changes in the levels of FtsZ and its regulators over the course of the cell cycle lead to differing prevalence of distinct FtsZ structures with different interacting partners. TipN, a cell polarity factor, is the last known protein to be recruited to the division site. Notably, the arrival of TipN at midcell is concomitant with the dispersal of MreB into a patchy distribution. Previous work showed that TipN is required for localization of MreB to the cell division site (Lam et al., 2006). Dynamic localization of TipN might, therefore, regulate the switch in MreB localization from a midcell ring to a dispersed pattern. In addition, it was recently reported that TipN affects the timing and position of Z ring formation by affecting the dynamics of the MipZ-associated parS/ParB complex (Schofield et al., 2010). TipN plays a key role in maintaining the directionality of chromosome segregation by interacting with ParA at the cell pole (Ptacin et al., 2010, Schofield et al., 2010). Thus, the timing of chromosome segregation and cell division is tightly coordinated by both MipZ and TipN.
Our data indicate that Caulobacter executes a defined pathway for divisome assembly that is not strictly linear, as it was initially thought to be in E. coli (Buddelmeijer and Beckwith, 2002). Rather, Caulobacter proteins are recruited to midcell as a series of functional modules, with proteins that are predicted to function coordinately arriving approximately simultaneously. This assembly pathway is consistent with the concerted mode of divisome assembly proposed for the E. coli divisome from experimental approaches aimed at dissecting interactions among division proteins (Goehring et al., 2005, Goehring et al., 2006). The differences in divisome assembly among Caulobacter, E. coli, and B. subtilis may represent functionally important distinctions between distantly related species. The assembly and operation of the conserved core division apparatus is fine-tuned through the actions of less highly conserved regulatory proteins and cell cycle control mechanisms to allow integration with distinct developmental processes and morphogenetic events. Continued investigation of the mechanisms underlying divisome function in diverse species will distinguish those aspects of cell division that are universal from those that represent adaptations to specific lifestyles.
All Caulobacter strains were derived from CB15N and grown at 28°C in peptone yeast extract (PYE) or M2-glucose minimal medium (M2G) with antibiotics when required (antibiotic concentrations in liquid (solid) PYE media: 1 (1) μg/mL chloramphenicol, 1 (5) μg/mL gentamycin, 5 (25) μg/mL kanamycin, 1 (2) μg/mL oxytetracycline, 25 (100) μg/mL spectinomycin, (5) μg/mL streptomycin). All experiments were performed with cells in log phase of growth in PYE with the exception of synchronization for determination of transcript or protein levels over the course of the cell cycle, for which cells were grown in M2G. For small-scale synchrony, cells were grown in 20 mL PYE with antibiotics to OD660 ~0.3, harvested by pelleting at 6000xg, resuspended in 750 μL ice-cold M2 salts, and combined with 750 μL Percoll. The cell suspension was centrifuged for 20 min at 4°C and 10000×g. The bottom swarmer band was isolated, washed 3 times in 1 mL ice-cold M2 salts, and resuspended in warm PYE media for growth at 28°C. For the depletion strains, cells were grown in PYE medium containing 0.3% xylose, washed with plain PYE medium three times, and then resuspended in PYE medium containing 0.2% glucose or 0.3% xylose. Samples were analyzed by phase contrast and fluorescence microscopy. The plasmids and strains used are detailed in Tables SII and SIII.
For localization studies, 0.3% xylose or 0.5 mM vanillic acid (pH 7.5) were added to induce expression of protein fusions when indicated. For time-course experiments, cells were grown in PYE medium, synchronized, and viewed on 1% agarose in M2G media. Phase contrast and fluorescence microscopy images were obtained using a Leica DM 6000 B microscope with a HCX PL APO 100x/1.40 Oil PH3 CS objective, Hamamatsu EM-CCD C9100 camera, and custom-designed KAMS microscope control and image analysis software (Christen et al., 2010). Images were processed with Adobe Photoshop.
Quantitative measurements of the localization patterns of division-site localized proteins were performed manually in ImageJ by determining the percentage of cells of each strain and each time point with polar, midcell, or diffuse localization. If a cell had both polar and midcell localization of the fusion in question, it was counted as midcell, since the primary objective was to follow arrival of each protein at midcell. The quantitative measurements of cell size, membrane boundary and degree of invagination as a function of time were performed automatically in MATLAB. Degree of invagination was determined by subtracting the cell width in pixels at the “waist” (i.e. at the incipient division site) from the average cell width in pixels at the “shoulders” (i.e maximum width of the cell body near the quarter positions). Measurement of the location and number of MipZ peaks was performed automatically in MATLAB. Cells were located by applying a calibrated threshold to the phase contrast images. Then, the corresponding cell regions of fluorescence images were segmented out. The cells were aligned along their long axis by the regionprops function in MATLAB. We modified pkfnd from SPtrack to analyze the number of MipZ peaks. Only cells with unipolar or bipolar peaks were included to determine the percentage of cells with bipolar MipZ. We manually counted a subset of images to determine the percentage of cells with 3 MipZ peaks: 2 at the cell poles and one at midcell or in the process of segregation.
The half-maximum times of the appearance of fusion proteins at midcell were determined by using the curve fit function in Prism (GraphPad, San Diego, CA). All data were fit with a sigmoidal dose response (variable slope) curve with the exception of the bipolar MipZ data, which was best fit with a one-phase exponential curve.
For all 19 proteins analyzed for timing of arrival at midcell, we examined their corresponding mRNA profiles as published by McGrath et al. (2007). We computed a “cell cycle regulation index” for each one of these mRNA profiles, as proposed by de Lichtenberg et al. (2005). Of the 19 profiles studied,10 ranked among the 20% most cell cycle regulated transcripts in the genome and were classified as such (represented in the vertical axis).
For a given experiment, equivalent OD units of cell lysate were run on SDS-PAGE. Western blotting to monitor Flag-M2 fusions, HU2 and FtsZ protein levels was performed with standard procedures. Samples were probed with primary antibodies at the following concentrations: α-Flag-M2 (Sigma-Aldrich) (1:1000), α-FtsZ (1:10000) (Mohl et. al, 2001), and α-HU2 (1:5000) (Goley et. al, 2010a). Secondary horse radish peroxidase-conjugated goat anti-rabbit or donkey anti-mouse antibodies were used at a 1:10000 dilution and chemiluminescent substrate was applied prior to exposure to film.
We thank Martin Thanbichler, Grant Bowman, Jerod Ptacin, Nathan Hillson, and James Gober for sharing strains, plasmids, and antibodies. We are especially grateful to members of the Shapiro and McAdams laboratories for helpful discussions. EDG was funded by the Helen Hay Whitney Foundation. This work was supported in part by the US Department of Energy under Contracts No. DE-FG02-05ER64136 (LS and HHM) and by the NIH under grant R01 GM 32506 (LS).