The bacterial MreB actin cytoskeleton is required for cell shape maintenance in most non‐spherical organisms. In rod‐shaped cells such as Escherichia coli, it typically assembles along the long axis in a spiral‐like configuration just underneath the cytoplasmic membrane. How this configuration is controlled and how it helps dictate cell shape is unclear. In a new genetic screen for cell shape mutants, we identified RodZ (YfgA) as an important transmembrane component of the cytoskeleton. Loss of RodZ leads to misassembly of MreB into non‐spiral structures, and a consequent loss of cell shape. A juxta‐membrane domain of RodZ is essential to maintain rod shape, whereas other domains on either side of the membrane have critical, but partially redundant, functions. Though one of these domains resembles a DNA‐binding motif, our evidence indicates that it is primarily responsible for association of RodZ with the cytoskeleton.
There is a Have you seen ...? (February 2009) associated with this Article.
Bacterial MreB actin has been implicated in cell shape maintenance, chromosome segregation and cell polarization events in a variety of rod‐shaped species, including the well‐studied model organisms Bacillus subtilis, Caulobacter crescentus and Escherichia coli. The shape of most bacterial cells is dictated by the shape of the murein (peptidoglycan) sacculus, in essence a giant and dynamic cell‐shaped molecule that surrounds the entire cytoplasmic membrane. How, despite often considerable turgor pressure, non‐coccal organisms manage to mould and maintain this molecule in a particular shape is an unsolved and intensely studied issue (for reviews, see Shih and Rothfield, 2006; Cabeen and Jacobs‐Wagner, 2007; den Blaauwen et al, 2008).
The MreB protein of E. coli is the only known actin in the cell and, as in other species, accumulates just underneath the cytoplasmic membrane in a spiral/banded‐like pattern along the long axis of the cell (Jones et al, 2001; van den Ent et al, 2001; Kruse et al, 2003; Shih et al, 2003; Figge et al, 2004; Gitai et al, 2005). Clear evidence for an important role of MreB in cell shape maintenance came from the isolation of spherical E. coli mreB mutants more than 20 years ago (Wachi et al, 1987), well before the protein was recognized as an actin. About that time the four additional proteins that are currently known to be critical in determining cell shape, MreC, MreD, penicillin‐binding protein 2 (PBP2) and RodA were identified as well (Tamaki et al, 1980; Wachi et al, 1989). The genes for MreC and MreD reside with mreB in the mreBCD operon, whereas those for PBP2 (MrdA) and RodA (MrdB) reside in the unlinked mrd operon. Although MreB is cytoplasmic, MreC and PBP2 are bitopic and MreD and RodA are polytopic cytoplasmic membrane species. PBP2 is the only murein synthase in E. coli that is specifically required for extension of the cylindrical portion of the sacculus during cell elongation (Spratt, 1975; de Pedro et al, 2001; Vollmer and Bertsche, 2008). RodA is likely needed for proper PBP2 function (Ishino et al, 1986; de Pedro et al, 2001).
MreC forms a dimer and is thought to interact with MreB, MreD and several of the high molecular weight murein synthases (PBPs), including PBP2 (Divakaruni et al, 2005, 2007; Dye et al, 2005; Kruse et al, 2005; van den Ent et al, 2006). MreC, MreD and PBP2 accumulate in a spotty or helical manner along the cell envelope in E. coli, B. subtilis and/or C. crescentus (den Blaauwen et al, 2003; Figge et al, 2004; Divakaruni et al, 2005; Dye et al, 2005; Leaver and Errington, 2005). These localization patterns are reminiscent of that of MreB, as well as of the helical patterns of new murein insertion that have been observed along the cylindrical portions of rod‐shaped cells. (Daniel and Errington, 2003; Tiyanont et al, 2006; Divakaruni et al, 2007; Varma et al, 2007). Hence, it is proposed that the helical actin fibres function as cytoplasmic tracks for murein synthase and/or hydrolase activities in the periplasm. This would topologically constrain these activities, resulting in helical insertion of new murein and elongation of the cell (Daniel and Errington, 2003; Figge et al, 2004; Carballido‐Lopez et al, 2006).
The MreB cytoskeleton has also been implicated in chromosome segregation in several organisms (Kruse et al, 2003, 2006; Soufo and Graumann, 2003; Gitai et al, 2005; Srivastava et al, 2007). However, such a role is not evident in all bacteria (Hu et al, 2007), and additional studies have cast doubt on a critical role of MreB in chromosome segregation in B. subtilis and E. coli (Formstone and Errington, 2005; Karczmarek et al, 2007).
To help elucidate the assembly and/or functions of the MreB cytoskeleton in bacteria, we sought to identify additional proteins required for maintaining the normal rod shape of E. coli cells. Although mre and mrd mutants of E. coli can propagate as small spheres on poor medium, they succumb to a lethal division defect on rich medium, unless they are supplied with an extra source of the FtsZ cytokinetic protein (Vinella et al, 1993; Kruse et al, 2005; Bendezu and de Boer, 2008).
We took advantage of this property in a genetic screen for mutants that require extra FtsZ for good growth on rich medium. This led us to identify a well‐conserved bitopic membrane protein of unknown function (YfgA) as a new cell shape protein that we named RodZ. Our evidence shows that RodZ is a component of the MreB cytoskeleton, and is required for its normal spiral‐like configuration. Analyses of domain deletion/substitution variants identify a juxta‐membrane portion of RodZ as essential to its function, and indicate that other domains engage the cell shape machinery in distinct ways. Curiously, although one of these domains resembles a DNA‐binding motif, our evidence indicates that it has a dominant function in the association of RodZ with the MreB cytoskeleton.
RodZ (YfgA) is a new cell shape factor in E. coli
In E. coli, the five known cell shape maintenance proteins MreBCD and MrdAB are all conditionally essential for growth. Cells lacking any of these proteins propagate stably as small spheres at low mass doubling rates, but form giant non‐dividing spheroids at higher ones. An extra supply of the FtsZ division protein, however, suppresses lethality by allowing the shape mutants to propagate as small dividing spheres at higher rates as well (Bendezu and de Boer, 2008). We made use of this latter property in a screen for additional cell shape factors by selecting for transposon mutants that (i) required additional FtsZ for survival or good growth on rich medium and (ii) showed a cell shape defect (see Supplementary data).
Mutant strain Rod2352 was especially interesting as it propagated as spheroids and EZTnkan‐2 had inserted in yfgA, a gene of unknown function. As mutants both lost rod shape and require an overdose of FtsZ for propagation on LB at or below 30°C (see below), we renamed it rodZ. The gene lies immediately upstream of ispG (gcpE), the product of which catalyses an essential step in isoprenoid biosynthesis (Hecht et al, 2001). We next created ΔrodZ strains (Figure 1A; Supplementary Table S2). Similar to the original insertion mutant, ΔrodZ cells failed to maintain rod shape. The growth properties of ΔrodZ cells were complex, and cell shapes depended to some extent on growth conditions. Compared with the wt parent TB28, the mass doubling time of FB60 [ΔrodZ] in M9‐mal minimal medium increased markedly (from ∼64 to >250 min; Tables I and III), but the mutant still formed (small) colonies relatively efficiently on M9 agar (Figure 2A). Cells propagated as spheroids in M9 medium, with few cells displaying more complicated shapes (Figure 2B). Mass doubling time was reduced less in LB medium (∼54 versus ∼35 min at 37°C; Table I). However, although the mutant still formed colonies on LB agar at 42°C and 37°C, it failed to do so at or below 30°C (Figure 2A, right panels). Cell shapes in LB at these lower temperatures were the most complex, with many cells growing into very large non‐dividing spheroids bearing one or more cone‐like protrusions (Figure 2B). Comparatively, cell shapes were the least abnormal on LB at 42°C, although virtually all cells were still misshapen, with many resembling ellipsoids, lemons and/or very wide rods (Figure 2B).
Both the growth and shape defects of ΔrodZ cells were fully corrected by pFB290 [Plac::rodZ] or other rodZ constructs, but not by pFB234 [Plac::ispG] (Figure 1; Supplementary Figure S1A; Table I, and see below), showing that neither defect was due to polar effects of the rodZ lesion on ispG or other genes.
Recovery of the rodZ::EZTnkan‐2 allele in our screen suggested that moderate overexpression of ftsZ is sufficient to suppress lethality of rodZ mutants. Accordingly, FB60 [ΔrodZ] cells that carried pDR3 [Plac::ftsZ] efficiently formed (small) colonies on LB agar at 30°C, provided IPTG was included in the medium (Figure 2C). As with Δmre and Δmrd cells, extra FtsZ allowed ΔrodZ cells to propagate as smaller cells without restoring their shape defects. In M9, this resulted in the formation of smaller spheroids (not shown). In LB, under otherwise non‐permissive conditions, extra FtsZ gave rise to heterogeneous populations of severely mis‐shaped cells that frequently displayed branches, bulges and oddly placed and/or angled constrictions (Figure 2D).
We conclude that rodZ is critical in maintaining the rod shape of E. coli. Similar to the mreBCD and mrdAB genes (Bendezu and de Boer, 2008), rodZ is conditionally essential, and RodZ− lethality under non‐permissive conditions can be suppressed by an elevated level of FtsZ. Additionally, ΔrodZ cells displayed a strong medium‐dependent growth defect that is not observed in Δmre and Δmrd cells (Bendezu and de Boer, 2008). This defect was not suppressed by elevated levels of FtsZ, by the presence of amino‐acid mixtures, or by changing the carbon source (data not shown). Though the reason for this growth phenotype is still unclear, it suggests that RodZ, besides cell shape maintenance, has additional significant function(s) in proliferation.
RodZ is a conserved, moderately abundant, transmembrane protein with a putative DNA‐binding domain
RodZ is a type II (N‐in) bitopic membrane protein (Newitt et al, 1999) of 337 residues, and is predicted to possess a Cro/CI‐type DNA‐binding domain near its N terminus (HTH, residues ∼1–84). This is followed by a highly basic juxta‐membrane region (JM, ∼85–110, net charge of +8), a transmembrane domain (TM, ∼111–133) and a substantial periplasmic domain (P, ∼134–337) consisting of a region rich in proline and threonine (∼134–253), and a C‐terminal domain that may be rich in β‐strands (∼254–337) (Figure 1B and C). Database searches suggest that RodZ‐related membrane proteins with a cytoplasmic Cro/CI (Xre)‐type DNA‐binding domain are present in many Gram‐negative as well as Gram‐positive organisms from most bacterial phyla (see COG1426; Tatusov et al (2003) and data not shown). The function of none of these has yet been established.
We raised antisera against the purified protein and estimated its cellular abundance by quantitative immunoblotting. The results indicated that RodZ is present at ∼650 copies per average exponentially growing cell in LB medium and that its cellular concentration in LB and M9 is about equal (results not shown).
As RodZ resembles a transmembrane transcription factor, we initially considered the possibility that it might be needed for expression of one or more of the other known shape proteins. The compatible plasmids pFB174 [PBAD::mreBCD] and pTB59 [Plac::mrdAB] can restore rod shape to ΔmreBCD and ΔmrdAB cells, respectively (Bendezu and de Boer, 2008). However, neither plasmid by itself, or in combination, affected the shape defect of ΔrodZ cells, suggesting that this defect was not due to a lack of any of these proteins (Supplementary Figure S1B). In addition, MreB levels in ΔrodZ cells were close to that in wt cells (Supplementary Figure S1C). Together with the finding that the HTH domain is not strictly needed for imposing rod shape per se (see below), these observations render it unlikely that RodZ is required for rod shape as a transcription factor.
RodZ localizes in a spiral‐type manner along the membrane
Strain FB60(iFB273) [ΔrodZ(Plac::gfp‐rodZ)] expresses GFP–RodZ in an IPTG‐dependent manner from a construct (iFB273) that was integrated at the chromosomal attHK022 site using the CRIM system (Haldimann and Wanner, 2001). Cells of this strain showed the typical RodZ− morphology upon growth in medium without inducer (Figure 2E1). However, cell morphology became indistinguishable from wt cells in the presence of IPTG at 250 μM or higher, showing that GFP–RodZ is fully capable of restoring rod shape. Fluorescence microscopy showed a spiral‐like distribution of the fusion along the length of cells (Figure 2E3), reminiscent of that previously seen for the MreB cytoskeleton (Kruse et al, 2003; Shih et al, 2003).
RodZ is part of the spiral‐like MreB cytoskeleton
Obtaining a fully functional fluorescent version of E. coli MreB by appending fluorescent tags at either terminus proved fruitless (not shown). However, we were able to construct a functional sandwich fusion (MreB–RFPSW) by inserting mCherry RFP between helices 6 and 7 of the protein (van den Ent et al, 2001). We then created strains that produce the MreB–RFPSW sandwich under native regulatory control and as the sole actin in the cell. Substitution of mreB with mreB‐rfpsw in these strains was verified by PCR and western blot analyses (Figure 3A and C). Cell morphology and growth rates of mreB‐rfpsw strains were indistinguishable from wt controls, and MreB–RFPSW was localized in a banded/spiral‐like manner along the long axis of cells (Figure 3B and D).
We next co‐visualized MreB–RFPSW and GFP–RodZ in cells of strain FB101(iFB273) [mreB‐rfpsw ΔrodZ (Plac::gfp‐rodZ)] in which production of the former is under native control, that of the latter is IPTG dependent, and the fusions are the sole sources of MreB and RodZ. Cells had a completely normal appearance in the presence of inducer (Figure 3E), implying that both fusions were also functional under these conditions. Notably, the two proteins appeared to colocalize perfectly at all stages of the cell cycle and in all cells examined (Figure 3E, and not shown).
This suggested that the MreB cytoskeleton might function as a scaffold for proper localization of RodZ. As any of the other cell shape proteins may associate with this scaffold as well (Kruse et al, 2005), we explored the possibility that they mediate the colocalization of RodZ and MreB. To this end, we co‐visualized MreB–RFPSW and GFP–RodZ in strains that completely lack MreC and MreD [ΔmreCD], or PBP2 and RodA [ΔmrdAB]. These strains also carried pTB63 [ftsQAZ], allowing the spherical cells to propagate readily (Bendezu and de Boer, 2008). MreB–RFPSW and GFP–RodZ still colocalized perfectly in either strain, even though the proteins co‐accumulated in mostly peripheral clusters/foci rather than in obvious spiral‐like patterns (Figure 3F and G). To assess whether such clustering of GFP–RodZ depended on MreB, we examined cells of strain FB30(λFB273)/pTB63 [ΔmreBCD(Plac::gfp‐rodZ)/ftsQAZ], which lacks all three Mre proteins, as well as in a related strain that lacks MreB specifically. As illustrated in Figure 3H and I, GFP–RodZ distributed evenly along the membrane in spheroids of either strain, showing that MreB indeed directs the cellular location of RodZ.
As their colocalization suggested that RodZ and MreB might interact, we assayed for in vivo interactions between RodZ and the other cell shape proteins using the BACTH bacterial two‐hybrid system (Table II; Supplementary Figure S2). This system is based on reconstitution of adenylate cyclase activity when the T18 and T25 domains of Bordetella pertussis CyaA are brought close together through interactions between fusion partners (Karimova et al, 1998).
Each CyaA domain was appended to the cytoplasmic N terminus of RodZ, MreC, MreD, PBP2 or RodA, while it was inserted in between G228 and D229 of MreB to generate MreB–T18SW and MreB–T25SW sandwich fusions. Pairs and inverse pairs of T18 and T25 fusions were co‐produced in indicator strain BTH101 [cya‐99], and putative interactions were assessed by a qualitative plate assay for β‐Gal activity (van den Ent et al, 2006). Notably, pairing RodZ with MreB produced very strong signals, supporting the notion that the two proteins reside in close proximity in vivo. Self‐pairing yielded the next strongest signals, suggesting a considerable level of RodZ self‐interaction. Interestingly, pairing RodZ with MreC in either genetic configuration also produced clear, but weaker, positive signals. Pairings with MreD or PBP2 yielded even weaker signals at best, and pairings with RodA yielded essentially no signals above those obtained with unfused T18 or T25 controls (Table 2; Supplementary Figure S2).
We conclude that RodZ is an integral component of the MreB cytoskeleton.
Aberrant assembly of MreB in RodZ− cells
To assess how RodZ affects MreB localization, we first examined the distribution of MreB–RFPSW in spheroids of strain FB85/pTB63 [mreB‐rfpSWΔrodZ /ftsQAZ], that lack RodZ entirely. As noted above, MreB–RFPSW accumulated in numerous small patches/foci of comparable intensity along the membrane of spheroids that are devoid of MreCD or MrdAB (Figures 3F, G and 4A). In comparison, MreB–RFPSW concentrated in notably fewer and larger peripheral clusters in RodZ− spheroids. This was especially apparent in LB‐grown ΔrodZ spheroids, in which most of MreB–RFPSW accumulated in a small number (∼1–4) of intensely stained clusters (Figure 4B and C).
To explore whether this aberrant assembly of MreB is more likely a cause or a consequence of the loss of cell shape in ΔrodZ cells, we monitored MreB–RFPSW during depletion of RodZ from cells of strain FB81 [mreB‐rfpSWPBAD::rodZ] in which chromosomal rodZ is under control of the ara regulatory region. Although virtually all cells were still rod‐shaped after growth in the absence of arabinose for 3.3 mass doublings, about half of them already displayed a small number of intense MreB–RFPSW accumulations, and the typical spiral‐like distribution of the protein was not or poorly discernible in these cells (Figure 4E). Even after an additional 2.3 doublings, cells were still rod‐shaped (though wider than normal), but the majority now displayed intense and aberrant MreB–RFPSW assemblies (Figure 4F).
These results indicate that RodZ has an important function in the formation of the extended spiral‐like configuration of the MreB cytoskeleton typically seen in rod‐shaped cells. Furthermore, as aberrant assembly of MreB precedes the loss of cell shape during depletion of RodZ it is likely that the shape phenotype of RodZ− cells is, at least partly, caused by the inability of MreB to form a proper cytoskeleton.
Rod shape depends on a proper MreB to RodZ ratio
Overexpression of MreB was previously reported to cause a cell division defect (Wachi and Matsuhashi, 1989; Kruse et al, 2003). We noticed, however, that the effects of MreB or RodZ overexpression in either TB28 [wt] (not shown) or FB83 [mreB‐rfpsw] (Figure 5) cells depend strongly on the growth medium. When cells harbouring pFB216 [Ptac::mreB] were grown in LB with IPTG (250 μM), they became filamentous, as expected, though the filaments were also notably wider than wt rods (average diameter D=1.7 μm versus D=1.0 μm; see Supplementary Table S1). In contrast, when MreB was overexpressed in M9 medium (∼2.5‐fold; Figure 5F), cells lost shape and formed large non‐dividing spheroids (Figure 5C). Similarly, overexpression of RodZ from the compatible plasmid pFB291[Ptac::rodZ] caused some elongation and widening of cells in LB (Supplementary Table S1), but overexpression in M9 (∼6.5‐fold; Figure 5F) led to a complete loss of rod shape. In such M9‐grown spheroids of strain FB83/pFB291 [mreB‐rfpsw/Ptac::rodZ], MreB–RFPSW accumulated in large clusters (Figure 5B), indicating that both the lack (Figure 4B–F) and overabundance of RodZ can cause MreB to form large structures that are ineffective in directing proper cell elongation.
Interestingly, co‐overexpression of MreB and RodZ in FB83/pFB216/pFB291 prevented the loss of cell shape seen when either protein is overexpressed alone. Instead, cells formed long and narrow (D=0.8–0.9 μm; Supplementary Table S1) filaments in either medium (Figure 5D). Thus, maintenance of a cylindrical cell shape depends critically on a proper MreB to RodZ ratio. In addition, it appears that MreB and RodZ significantly reinforce each other's ability to inhibit cell division when present at elevated levels.
Unique nucleoid and MreB/RodZ distribution patterns upon co‐overexpression of MreB and RodZ
In addition to the division defect, cells that co‐overexpressed RodZ and MreB showed other notable features. First, filaments of either TB28 [wt] (not shown) or FB83 [mreB‐rfpsw] showed unusually large nucleoid‐free gaps, suggesting a defect in chromosome integrity, replication or segregation (Figure 5D3 and E3). Second, inspection of MreB–RFPSW spirals in such filaments of strain FB83 revealed that they were not evenly distributed along the long axis of the filaments, but became enriched in zones that almost always corresponded to a nucleoid‐free gap (Figure 5D2 and E4, arrowheads). Third, over a third (39/105) of the filaments showed one or two prominent bulges, and these were always enriched with MreB–RFPSW. Curiously, all such bulges also contained DAPI‐stained material, and often a substantial amount, as if these were sites where nucleoids had invaded an MreB‐enriched zone, or vice versa (Figure 5D, arrow).
We expected RodZ to co‐enrich with MreB under these conditions as well. To verify this, we imaged cells of strain FB83/pFB299/pFB309 [mreB‐rfpsw/Ptac::mreB rodZ/Psyn35::gfp‐rodZ], in which MreB and RodZ can be overexpressed from a single plasmid, and production of GFP–RodZ is under control of a weak constitutive promotor on a compatible plasmid. Figure 5E shows an example of filaments of this strain after growth in the presence of IPTG, and when they are still relatively short. Note that they still contain an even number of fairly well separated (pairs of) nucleoids, suggesting that chromosome replication had proceeded apace. However, nucleoids failed to distribute normally towards the poles of the filaments, leaving large nucleoid‐free zones at each cell end, and these were also the regions where GFP–RodZ had indeed preferentially co‐accumulated with MreB–RFPSW. Moreover, filament ends at which an MreB/RodZ‐enriched zone substantially overlapped the nearest nucleoid, often already showed some local widening of the cell cylinder (Figure 5E, arrows). On further growth, filaments of this strain showed phenotypes as that described for FB83/pFB216/pFB291 (Figure 5D), with both proteins enriched at both internal and polar nucleoid‐free zones, as well as at more pronounced bulges (not shown).
Depletion of RodZ does not greatly affect chromosome segregation
The MreB cytoskeleton has been implicated in chromosome segregation (Kruse et al, 2003, 2006; Soufo and Graumann, 2003; Gitai et al, 2005). Given its properties described above, RodZ would be an attractive candidate for mediating MreB‐directed nucleoid dynamics. However, inspection of DAPI‐treated FB60 [ΔrodZ] cells failed to reveal DNA‐less cells (0/200) or any other gross nucleoid segregation defects (Supplementary Figure S3D and E). Moreover, we failed to detect any obvious defects in segregation of the origin and terminus regions of the chromosome during depletion of RodZ (Supplementary Figure S3). These results indicate that a lack of RodZ has little, if any, direct effect on chromosome dynamics.
Roles of RodZ domains in cell growth, cell shape and RodZ localization
In genetic footprinting analyses of the E. coli genome, rodZ (yfgA) was classified as non‐essential because a transposon insertion in codon 156 was apparently well tolerated by cells growing in rich medium (Gerdes et al, 2003). This prompted us to engineer a strain bearing a translation stop after codon 155 of chromosomal rodZ (Figure 1A). Indeed, FB61[rodZ1−155] cells grew well and were still rod‐shaped in both minimal and rich medium, though they were up to 45% wider on average than cells of the wt parent TB28 (Table I; Supplementary Figure S4). Thus, the periplasmic part of RodZ is largely dispensable, although it may help to fine‐tune the precise dimensions of cells.
We next studied the roles of each of the HTH, basic JM, TM and P domains in RodZ function. To this end, we created a set of derivatives of the integrative CRIM construct pFB273 [Plac::gfp‐rodZ], encoding variants of GFP–RodZ in which one or more of the domains is missing or replaced. To replace the JM and/or TM domain, we used corresponding domains of MalF (TM1; Guzman et al, 1997), and mCherry replaced the P domain in some variants.TB28 [wt] and FB60 [ΔrodZ] cells producing each variant were then examined for growth rate, cell morphology and localization of the mutant protein (Figure 6; Table 3). Western blot analyses indicated that none of the variants was subject to excessive degradation (Supplementary Figure S5).
Notably, the results in toto (Figure 6; Table III) revealed that both normal growth rate and cell shape is strictly dependent on: (i) the JM domain of RodZ (Figure 6, compare A with I, D with E and F with G), (ii) a TM domain to anchor it to the membrane (Figure 6, compare B and D with C), and (iii) the JM–TM part of RodZ being either preceded by its HTH domain, or followed by its P domain (Figure 6, compare B and F with H). Furthermore, although substitution of RodZ's TM with TM1 of MalF did not abolish RodZ function, the comparable variant with RodZ's native TM was clearly more effective in restoring rod shape to ΔrodZ cells (Figure 6B and D; Table III).
Surprisingly, the requirements for a normal localization pattern of RodZ were not the same as those for good growth and rod shape of cells. Thus, a membrane‐tethered version of the HTH domain was both required (Figure 6C and F–H) and sufficient (Figure 6B, D, E and I) for the protein to sharply accumulate spiral‐like in rods or in peripheral foci in non‐rods.
The properties of a fusion lacking the HTH domain (GFP–RodZ83−337) were particularly revealing as it still restored rod shape to a majority of ΔrodZ cells, even though many rods appeared wider than normal and the rest of the population (∼25%) showed irregularities such as uneven width, branching and/or a spheroidal shape (Figure 6F; Table III). However, the protein distributed far more evenly along the membrane in both corrected ΔrodZ rods and wt cells (Figure 6F) than HTH+ variants (Figure 6A, B, D, E and I). The distribution of GFP–RodZ83−337 was not completely homogeneous, however, suggesting that some of the protein still accumulated in a more organized manner (Figure 6F3 and F4).
These results showed that the HTH domain is not strictly required for RodZ function per se, but that it is responsible for the typically sharp spiral‐like distribution of the protein. In turn, this indicated that the HTH domain contributes significantly to the association of RodZ with the MreB cytoskeleton, and this was supported by BACTH assays. A membrane‐tethered fusion containing just the HTH domain (T18–RodZ1−84–MalF1−39–RFP) still showed a very strong interaction with MreB (T25–MreBSW) as well as a weaker one with full‐length RodZ (T25–RodZ). In contrast, interactions with MreC or the other shape proteins were no longer obvious with this fusion, suggesting that these require other parts of RodZ (Table II; Supplementary Figure S2). Conversely, a fusion lacking just the HTH domain (T18–RodZ83−337) still showed interactions with MreB, RodZ, as well as MreC, but all three appeared distinctly weaker than those seen with the full‐length T18–RodZ fusion (Table II; Supplementary Figure S2).
As the JM domain of RodZ is essential for its function in cell shape maintenance, we also explored whether its substitution affected its interactions with cell shape proteins in the BACTH assay. However, the interaction pattern of a T18–RodZ1−84–MalF1−16–RodZ111−337 fusion was very similar to that of T18–RodZ (Table II; Supplementary Figure S2). This suggests that, if any interaction between the JM domain and other shape proteins occurs, these may be relatively weak and/or short‐lived compared with those involving other domains of RodZ.
A novel genetic screen for cell shape mutants of E. coli led us to uncover RodZ as a new cell shape factor in bacteria. Like the MreBCD and MrdAB proteins, RodZ has a critical function in the maintenance of rod shape and is conditionally essential for viability. Though rodZ (yfgA) had no known function, we note that the ill‐defined divD lesions of classical round‐cell mutants of Salmonella typhimurium mapped very near the location of rodZ (Wyche et al, 1974), suggesting that these may have been allelic.
The phenotypes of ΔrodZ cells showed similarities as well as interesting differences with mre and mrd shape mutants (Bendezu and de Boer, 2008). Similar to the latter, ΔrodZ cells showed division defects and could grow into giant spheroids with elaborate intra‐cytoplasmic membrane systems (e.g. Figure 6C, G and H). In addition, ΔrodZ cells showed conditional lethality on rich medium, which was suppressible by extra FtsZ (Figure 2). Unlike mre and mrd mutants, however, lethality on rich medium of ΔrodZ cells was also partially suppressed by growth at 42°C, and cell shape appeared the least abnormal then as well. This suggests that some other factor can partially compensate for the lack of RodZ under these conditions. It is unclear what this factor might be, but it is unlikely to be any of the other cell shape proteins (Supplementary Figure S1). Furthermore, although mre‐ and mrd‐null mutants form spheroids exclusively, ΔrodZ cells often displayed more complex shapes, especially under non‐permissive conditions (LB, <37°C). This might be a direct consequence of the misassembly of MreB into the large clusters seen in these cells. If such non‐spiral assemblies still were to direct some new murein synthesis, it is easy to imagine this leading to bizarre cell shapes.
Another notable difference with mre and mrd mutants is the large reduction of mass doubling rates of ΔrodZ versus wt cells in M9 medium. This growth defect likely helps ΔrodZ cells to survive on this medium, as forced reduction of growth rates helps mre and mrd mutants to survive as well (Bendezu and de Boer, 2008). Our mutational analyses so far failed to separate the roles of RodZ in growth and cell shape (Table III). Further work may do so, or a single RodZ activity may control both cell shape and growth rate. It will be interesting to learn the precise cause of this growth defect, as it suggests that RodZ links the cell shape machinery with some cellular activity that is especially important for cell proliferation on poor medium.
Apart from the cell shape phenotypes of ΔrodZ cells, our conclusion that RodZ is a component of the MreB cytoskeleton is supported by the colocalization of RodZ and MreB (Figures 3 and 5), by BACTH assays indicating that RodZ interacts with MreB and MreC in vivo (Table II; Supplementary Figure S2), and by evidence that the ratio of MreB to RodZ must be kept within limits to ensure an elongated cell shape (Figure 5).
Furthermore, the atypical accumulation of MreB in dense clusters in cells lacking RodZ (Figure 4) implies that the protein has a critical function in ensuring assembly of the cytoskeleton into its typical spiral‐like configuration. In eukaryotes, the dynamics and architecture of F‐actin are modulated by a plethora of actin‐binding proteins that function at a variety of steps in assembly/disassembly reactions (Pollard, 2007). RodZ may similarly modulate MreB assembly directly. Robust in vitro systems will be needed to address whether it does so and, if so, at what step(s).
Our domain analyses (Figure 6; Table III) show that the basic juxta‐membrane (JM) domain of RodZ is the only one that is strictly required for the maintenance of rod shape. Hence, this domain is likely the most directly involved in ensuring a proper configuration of the MreB cytoskeleton. The JM domain is clearly not sufficient for RodZ function, however, as the domain also needs to be membrane‐tethered and, additionally, needs to be accompanied by either the cytoplasmic HTH or the periplasmic P domain (Figure 6; Table III). These results are inconsistent with a simple binary interaction between RodZ and the cytoskeleton, but rather argue that RodZ engages cytoskeletal partners on both sides of the membrane.
The most parsimonious scenario that is consistent with all our data has the following features: (i) the JM domain of RodZ directly or indirectly coaxes MreB fibres into a spiral‐like configuration, ensuring that cells establish/maintain a proper long axis. (ii) The JM domain itself has insufficient affinity for the cytoskeleton to stably associate with it. (iii) The HTH and P domains each contribute separately, and to different extents, to the association of RodZ with the cytoskeletal apparatus. This ensures that, even when one of these domains is missing, the local concentration of JM domains near MreB fibres in the cytoplasm remains sufficiently high to stimulate rod shape in a majority of cells. (iv) The HTH domain is the dominant localization determinant of RodZ, and directly or indirectly engages MreB in the cytoplasm (Figure 6; Tables II and III). As RodZ and MreB still colocalized sharply in the absence of the other cell shape proteins (Figure 3), it is unlikely that they mediate this interaction. The simplest interpretation is that the HTH domain binds MreB directly, but this needs to be confirmed with purified components. (v) In the periplasm, the P domain engages another transmembrane component of the cell shape machinery that itself associates with the MreB cytoskeleton. In rod‐shaped cells producing only the ΔHTH variant of RodZ, some of the protein could still be detected in weak spots or bands, but it localized far less sharply than HTH+ variants and was clearly less effective in imposing rod shape than the ΔP variant (Figure 6; Table III). This suggests that the P domain alone is only moderately effective in keeping RodZ incorporated into the cytoskeleton. What the P domain partners with remains to be established firmly, but MreC is an attractive candidate as BACTH assays now suggest that MreC interacts with MreB and MreD (Kruse et al, 2005; our unpublished results), as well as with RodZ (Table II; Supplementary Figure S2).
Though the primary structure of the HTH domain suggests a DNA‐binding function, depletion of RodZ had little, if any, effect on nucleoid segregation (Supplementary Figure S3). In addition, our evidence strongly indicates that the domain is actually used to engage the cytoskeleton. Still, interactions between RodZ and the chromosome remain possible. The unique inverse distribution patterns of nucleoids and MreB/RodZ‐enriched zones in filaments that co‐overexpress the proteins suggest that nucleoids somehow determine the positioning of these zones or vice versa (Figure 5). In turn, this suggests the possibility that cytoskeletal function may somehow be modulated by the underlying nucleoid. This will require further investigation, as will a full understanding of other aspects of the complex phenotype caused by increased levels of MreB and RodZ.
Very recently, we learned (personal communication) that two other groups independently identified RodZ as an important cell shape factor in E. coli (Shiomi et al, 2008) as well as in C. crescentus (Christine Jacobs‐Wagner). The latter emphasizes the likelihood that RodZ function has been widely conserved among the bacteria.
In addition to the discovery and initial characterization of RodZ, we also described the creation of a functional MreB–RFPSW sandwich fusion, which we exploited throughout this study and should prove useful in future studies on the cytoskeleton. Evidently, the presence of an extra domain between helices 6 and 7 can be remarkably well tolerated by E. coli MreB. It will be worth exploring if this is true for other (bacterial) actins as well.
Materials and methods
E. coli strains, plasmids, phages, growth conditions, BACTH assays and rod screen
Details on all strains (Supplementary Table S2), genetic constructs (Supplementary Table S3) and their source or construction are provided as Supplementary data. Relevant genotypes [in brackets] are also given in the text.
Unless stated otherwise, cells were grown at 30°C in LB (0.5% NaCl) or in M9 minimal medium supplemented with 0.2% maltose, 0.2% casamino acids and 50 μg/ml l‐tryptophan (M9‐mal). When appropriate, the medium was supplemented with 50 μg/ml ampicillin (Amp), 50 μg/ml spectinomycin (Spec), 25 μg/ml kanamycin (Kan), 25 μg/ml chloramphenicol (Cam) or 12.5 μg/ml tetracycline (Tet). Amp and Cam concentrations were reduced to 15 and 10 μg/ml, respectively, when cells carried bla or cat integrated into the chromosome. Other details are specified in the text.
For BACTH analyses (Karimova et al, 1998), plasmid pairs encoding the indicated T18 and T25 fusions were co‐transformed into BTH101 [cya‐99], and individual colonies were patched on M9 agar containing 0.2% glucose, 50 μg/ml Amp, 25 μg/ml Kan, 40 μg/ml X‐Gal and 250 μM IPTG. Plates were incubated at 30°C and inspected after 24, 30 and 36 h.
EZTnkan‐2 (Epicentre) mutagenesis of strain TB28/pFB184 [ΔlacIZYA/Plac::sdiA::lacZ], the subsequent screen for rod mutants and the mapping of insertions were carried out essentially as described (Bernhardt and de Boer, 2004). Complete details are provided as Supplementary data.
Live cells were imaged on 1.2% agarose pads made with either 0.5% NaCl (for LB cultures) or M9 salts (for M9 cultures). When indicated, cells were chemically fixed as described earlier (Bendezu and de Boer, 2008) and DAPI was added to 0.25 μg/ml, 2 min prior to imaging.
Microscopy set‐ups are detailed further in Supplementary data.
Measurements of cellular parameters and quantitative western blot analyses were carried out as before (Bendezu and de Boer, 2008). Details on the purification of antigens for antibody production are given as Supplementary data.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Supplementary Figure S1
Supplementary Figure S2
Supplementary Figure S3
Supplementary Figure S4
Supplementary Figure S5
We thank Yu‐Ting Su and Elizabeth Zeng for help in plasmid construction; Stuart Austin, Daniel Ladant, Gouzel Karimova, Kenneth Marians, David McPheeters and Steven Sandler for materials, and Hironori Niki and Christine Jacobs‐Wagner for communicating their independent discoveries of RodZ and for agreeing on its name. This study was supported by a Human Frontiers Science Program award (RGP0001/2003) and NIH GM57059 (to PAJdB), and NIH NRSA Institutional Training Grant T32GM08056 (to FOB). TGB holds a Career Award in the Biomedical Sciences from the Burroughs Wellcome Fund.
- Copyright © 2009 European Molecular Biology Organization