Bacillus subtilis MreB paralogues have different filament architectures and lead to shape remodelling of a heterologous cell system


E-mail; Tel. (+49) 7612032641; Fax (+49) 7612032773.


Like many bacteria, Bacillus subtilis cells contain three actin-like MreB proteins. We show that the three paralogues, MreB, Mbl and MreBH, have different filament architectures in a heterologous cell system, and form straight filaments, helices or ring structures, different from the regular helical arrangement in B. subtilis cells. However, when coexpressed, they colocalize into a single filamentous helical structure, showing that the paralogues influence each other's filament architecture. Ring-like MreBH structures can be converted into MreB-like helical filaments by a single point mutation affecting subunit contacts, showing that MreB paralogues feature flexible filament arrangements. Time-lapse and FRAP experiments show that filaments can extend as well as shrink at both ends, and also show internal rearrangement, suggesting that filaments consist of overlapping bundles of shorter filaments that continuously turn over. Upon induction in Escherichia coli cells, B. subtilis MreB (BsMreB) filaments push the cells into strikingly altered cell morphology, showing that MreB filaments can change cell shape. E. coli cells with a weakened cell wall were ruptured upon induction of BsMreB filaments, suggesting that the bacterial actin orthologue may exert force against the cell membrane and envelope, and thus possibly plays an additional mechanical role in bacteria.


The presence of actin and its prokaryotic counterpart MreB in all three kingdoms of life suggests that a predecessor protein has been present in the last common ancestor of all cells and organisms. The property to form dynamic filaments that can grow, bundle, shrink and branch has been a powerful invention in nature that has lent the ability to perform an enormous spectrum of functions in cells. Actin mediates maintenance of cell shape, motility, intracellular transport and cell division, among many other crucial cellular activities (Pollard, 2007; Kueh and Mitchison, 2009). Bacterial MreB on the other hand clearly plays a role in the maintenance of cell morphology (which in most bacteria appears to be primarily determined by the cell wall), and has also been implicated in cell cycle progression and cell polarity, as well as in cellular differentiation (Carballido-Lopez, 2006; Graumann, 2007). In Escherichia coli, Bacillus subtilis and Caulobacter crescentus, where MreB has been best studied, it is essential for viability (Jones et al., 2001; Figge et al., 2004; Kruse et al., 2005). The protein localizes as helical filaments underneath the cell membrane, which are highly dynamic with rapid subunit turnover in B. subtilis and C. crescentus (Jones et al., 2001; Defeu Soufo and Graumann, 2004; Kim et al., 2006). Filament dynamics are essential for the proper function of MreB (Defeu Soufo and Graumann, 2006). MreB interacts with membrane proteins that are also crucial for cell morphology (MreC and MreD), and also with membrane proteins that mediate the incorporation of new cell wall material into the extending murein sacculus (Figge et al., 2004; Kruse et al., 2005; Defeu Soufo and Graumann, 2006; Kawai et al., 2009a). One hypothesis for the effect of MreB on cell shape is the idea that MreB filaments direct the localization of cell wall synthetic enzymes and thus dictate the shape of the bacterial envelope. Indeed, incorporation of new cell wall material occurs in a helical pattern, supporting this idea (Daniel and Errington, 2003). It is still unclear how MreB is organized into the relatively regularly spaced helical filaments underneath the membrane, and if the interacting membrane proteins localize with similar dynamics. Also, a stringent correlation between the positioning of cell wall synthetic enzymes and the form of the murein sacculus has not yet been demonstrated.

Gram-positive bacteria have a multi-layered murein sacculus, and apparently a greater need to co-ordinate the multitude of cell wall synthetic enzymes. In contrast to E. coli and many proteobacteria, which have a single mreB gene, most Gram-positive bacteria (as well as some Gram-negatives) have two or three MreB paralogues (Graumann, 2007), which in B. subtilis are termed MreB, Mbl (MreB-like) and MreBH. While E. coli MreB interacts with MreC and MreD proteins (Kruse et al., 2005), B. subtilis Mbl (but not MreB) interacts with MreC (Defeu Soufo and Graumann, 2006); MreBH on the other hand interacts with and influences the localization of the cell wall hydrolase LytE (Carballido-Lopez et al., 2006). The deletion of mreB is lethal, and that of mbl leads to viable cells which, however, have a drastically altered cell shape, while the deletion of mreBH has a minor effect on the morphology of the cells, which become slightly wider and mildly bent (Jones et al., 2001; Defeu Soufo and Graumann, 2003). The depletion of MreB results in the formation of large, round cells, while mbl mutant cells are highly twisted, yet still retain some form of rod shape. Based on the different phenotypes of the deletion of the genes, it has been speculated that they perform different functions; however, all three paralogues colocalize to the helical filaments underneath the membrane, and interact with each other (Carballido-Lopez et al., 2006; Defeu Soufo and Graumann, 2006). Interestingly, the deletion of mreB or of mbl can be suppressed by the addition of high concentrations of magnesium and sucrose to the growth medium, such that the cells grow relatively normal (Formstone and Errington, 2005). Also, the deletions can be suppressed by an additional mutation in a gene regulating the expression of cell wall-synthesizing enzymes (Schirner and Errington, 2009a), suggesting that the absence of single MreB paralogues can be overcome by a stabilization of the cell envelope.

In this work, we characterize the different filament-formation properties of the three B. subtilis MreB paralogues and show that they localize in strikingly different architectural forms in a heterologous cell system. We also show that the expression of MreB alters cell morphology in the heterologous system and applies force against the cell membrane, which is a novel concept for the establishment of cell morphology by a bacterial actin-like protein.


B. subtilis MreB (BsMreB) and MreBH, but not Mbl, form different kinds of extended filaments in E. coli cells

To address the question of the different polymerization properties of MreB paralogues from B. subtilis, and their contribution to the cytoskeletal structure formed by them, we turned to a heterologous cell system. We chose E. coli cells, which are highly diverged from B. subtilis, but possess similar cell dimensions and allow for the fine-tuned expression of three proteins in different combinations, using two compatible high-copy-number plasmids. We expressed CFP–Mbl, YFP–BsMreB or mCherry-MreBH at low induction levels; all protein fusions are fully functional in B. subtilis cells (Carballido-Lopez and Errington, 2003; Defeu Soufo and Graumann, 2004) (Fig. 1A–C), also when two paralogues are expressed as fluorecent protein fusions as sole source of the proteins (Carballido-Lopez et al., 2006; Defeu Soufo and Graumann, 2006). We found that BsMreB and MreBH readily formed extended filamentous structures 30 min after induction of expression in E. coli cells, while Mbl was observed to form several patch-like structures along the membrane, but no clear filaments (Fig. 1D–F). Because it was important to avoid over-production artefacts, we set the levels of protein expression such that they were similar to protein levels in B. subtilis cells. Western blot experiments performed with anti-BsMreB or Mbl revealed that the levels of the proteins expression in E. coli were indeed similar as in B. subtilis wild-type cells, using 0.25 mM IPTG as inducer (Fig. 1G). These experiments show that BsMreB and MreBH can form extended filaments in the absence of any factor from B. subtilis cells. Filamentous structures of BsMreB were not well organized, and did not follow a regular helical pattern, as observed for E. coli MreB (EcMreB) (Kruse et al., 2003), or for BsMreB in B. subtilis cells (Jones et al., 2001; Defeu Soufo and Graumann, 2004), but frequently crossed each other (Fig. 1D). However, BsMreB filaments were mostly observed at the cell periphery, although, occasionally, clearly membrane-detached filaments were observed (Fig. S1C). Strikingly, MreBH did not form helical filaments, but ring structures that were relatively evenly spaced (average spacing of 0.8 ± 0.1 µm) along the length of the E. coli cells (Fig. 1F and Fig. S2A and B), which, in rare cases, were connected by a filament following a half-helical turn (Fig. S2A). Thus, MreBH filaments have a pitch of zero, and rarely open into a short helix, while BsMreB filaments have an entirely different architecture mainly involving a helical or straight arrangement along the membrane. This is in striking contrast to B. subtilis cells, in which all three MreB paralogues form helical filaments with a regular pitch of 0.5 µm (Defeu Soufo and Graumann, 2004; Carballido-Lopez et al., 2006).

Figure 1.

Expression of YFP-BsMreB, CFP–Mbl and mCherry-MreBH in B. subtilis and E. coli cells.
A–C. Subcellular localization of (A) YFP–BsMreB, (B) CFP–Mbl and (C) mCherry-MreBH in exponentially growing B. subtilis cells.
D–F. Expression in E. coli cells after 45 min induction – (D) YFP–BsMreB filaments, (E) CFP–Mbl structures, (F) mCherry-MreBH filaments.
G. Western blot (using anti-BsMreB or anti-Mbl antiserum) showing the expression levels of YFP–BsMreB (left panel) or of CFP–Mbl (right panel) in E. coli cells 30 or 60 min after induction of FP fusion proteins synthesis compared with the expression levels in B. subtilis cells.
H–K. Colocalization experiments in E. coli cells – (H) YFP–BsMreB (green in merge) and CFP–Mbl (red in merge) filaments 60 min after induction, (I) CFP–Mbl (green in merge) and mCherry-MreBH (red in merge) filaments 60 min after induction, (J) YFP–BsMreB (green in merge) and mCherry-MreBH (red in merge) filaments 60 min after induction, (K) YFP–BsMreB (green in merge), CFP–Mbl (red in merge) and mCherry-MreBH (blue in merge) filaments 60 min after induction. E. coli cells were grown in LB at room temperature to an OD600 of 0.2 and FP fusion proteins were induced by adding IPTG to the final concentration of 0.25 mM.
Scale bars 2 µm.

Mbl formed assemblies or aggregates underneath the cell membrane in a random pattern (Fig. 1E). It should be noted that Mbl patches may consist of random aggregates of Mbl or of short Mbl filaments, which cannot be determined because of the limitation of the resolution of the light microscope.

BsMreB, Mbl and MreBH filaments grow and shrink by extension/retraction at both ends and polymerize into a single structure in E. coli cells

We wished to investigate if B. subtilis MreB paralogues influence each other when coexpressed in the heterologous cell system. Therefore, we coexpressed all three different combinations of two paralogues, and all three together, each tagged with a different variant FP fusion. Although Mbl itself was not able to form extended filaments (Fig. 1E), it formed filaments when coexpressed with BsMreB (Fig. 1H) or with MreBH (Fig. 1I), and colocalized with both other paralogues. Interestingly, when BsMreB and MreBH were coexpressed, they also colocalized, but the architecture of the formed structures was somewhat intermediate between the disorganized BsMreB filaments and the regular MreBH rings (Fig. 1J). When all three paralogues were coexpressed, all colocalized to the same filamentous structures (Fig. 1K) that showed a more organized helical pattern than BsMreB expressed on its own (Fig. 1D). The facts that BsMreB, Mbl and MreBH colocalized to the same filamentous structures, that the proteins affect each other's properties to form filaments, and that all three interact with each other in vivo (Defeu Soufo and Graumann, 2006) show that the three proteins form either mixed bundles of filaments (i.e. association in parallel of single filaments of individual protein), or completely mixed filaments within the bundles (i.e. association in parallel of single filaments built by a mix of all three proteins). It is therefore clear that the proteins form a single coherent cytoskeletal structure, rather than independent colocalizing helical filaments. The finding that the mixed filaments formed by BsMreB and MreBH, or by all three paralogues more closely resembled those observed in B. subtilis cells, rather than the disorganized filaments formed by BsMreB itself or the ring-like MreBH structures, suggests that the co-assembly of the BsMreB/Mbl/MreBH filaments push the arising mixed filaments into a more regular helical arrangement. Thus, B. subtilis MreB paralogues strongly affect each other's polymerization properties and the architecture of the resulting cytoskeletal structure.

We also followed the extension of growing BsMreB and mixed filaments in time-lapse experiments. Figure 2A shows that BsMreB filaments extended at both ends (in 25 cases out of 30 extensions monitored), but could also elongate at one end (Fig. 2B and Movie S1), which is also true for mixed Mbl/MreBH filaments (Fig. 2C and Movie S2) of for mixed BsMreB/Mbl/MreBH filaments (Fig. 2D and Movie S3). To obtain more insight into the dynamics of filament assembly, we performed FRAP experiments on cells expressing YFP–BsMreB. Fluorescence intensity recovered with approximately the same kinetics from both sides relative to the centre of the bleached area, and recovered relatively uniformly within the entire bleached area (Fig. 3A–E and Movie S4) in the 12 movies captures. These observations support that YFP–BsMreB filaments lack polarity in E. coli cells, but show subunit turnover along their entire length, similar to GFP-EcMreB filaments expressed in fission yeast (Srinivasan et al., 2007). Because it was described that single GFP–MreB molecules exhibit treadmilling-like dynamic in C. crescentus (Kim et al., 2006), meaning growth at one end (+) and monomer dissociation at the other end (−) of a filament, the absence of polarity in YFP-BsMreB filaments in E. coli suggests that the visible filaments may be composed of parallel, oppositely oriented associations of individual MreB filaments. Importantly, we observed extension and shrinkage of filament ends, which appeared to occur in parallel for solely expressed YFP–BsMreB (data not shown), for two coexpressed proteins (Fig. 2C) and for all three paralogues within the joint structures formed (Fig. 2D). Thus, in a heterologous system, BsMreB/Mbl/MreBH filaments can extend and retract, although retraction is much more rarely observed than filament growth. Extension of filaments was observed in a timescale of sub-minutes with an average growth speed of 0.05 µm min−1 (maximum 0.1 µm min−1) for single-, double- or triple-expression filaments. Thus, MreB filaments can extend at both ends, similar to but much more slowly than ParM, which pushes plasmids towards opposite cell poles (Moller-Jensen et al., 2003; Garner et al., 2004). In a previous report, we suggested that in B. subtilis, MreB filaments may extend at one end and depolymerize at the other end, as judged from time-lapse microscopy (Defeu Soufo and Graumann, 2004). Based on the results from this work, MreB filaments likely also extend at both ends in B. subtilis cells, which is difficult to observe because of the helical arrangement of the filaments. In any event, it is important to note that basic bipolar polymerization and depolymerization appears to be conserved between bacterial actin-like proteins.

Figure 2.

Time-lapse of extension of filaments.
A and B. Growing YFP–BsMreB filaments, (A) at both ends (the white star indicates an arbitrary reference and white arrowheads indicate the tips of the filament), (B) at one end [one arrowhead (white) is static and the other (unfilled) indicates the tip of the extending filament].
C. Internal (apparently with no membrane connection) growth and shrinkage of CFP–Mbl filament in a strain coexpressed with mCherry-MreBH. The sample is imaged at the middle plane. The white bars are the references at time point zero min. Note that the filament grows and shrinks at both ends.
D. YFP–BsMreB (i), CFP–Mbl (ii) and mCherry-MreBH (iii) extending and retracting in the mixed polymer (the white arrowhead is static and the unfilled arrowhead indicates the tip of the filament which grows and shrinks). Cells were grown in LB at room temperature to an OD600 of 0.2 and FP fusion proteins were induced by adding IPTG to the final concentration of 0.25 mM.
Scale bars 2 µm.

Figure 3.

FRAP experiment on E. coli cells expressing YFP–BsMreB.
A. FRAP montage showing recovery with time (white arrowheads indicate the bleached area).
B–E. Typical FRAP analysis of YFP-BsMreB showing that recovery is all over the bleached area. (B) Pre-bleach cell. White bar indicates the middle reference point. (C) Bleach and (D) post-bleach. The boxes represent the areas where the florescence intensity was measured: (a) for the left and (b) for the right side of reference bar in (B). (E) FRAP profiles of (a) and (b) regions showing that the fluorescence recovery follows the same patterns from both sides. Cells expressing YFP–BsMreB were grown in LB at room temperature to an OD600 of 0.2 and the fusion protein was induced by adding IPTG to the final concentration of 0.25 mM.
Scale bars 2 µm.

Filament architecture of MreBH can be altered towards that of BsMreB by a single amino acid exchange

The strikingly different architecture of BsMreB and MreBH single filaments prompted us to address the question of how these differences can arise on molecular bases. We investigated the positioning of diverging amino acids within highly conserved regions at positions that correspond to subunit interfaces based on the structure of Thermotoga maritima MreB (van den Ent et al., 2001) between BsMreB, Mbl and MreBH proteins and MreB orthologues from different bacteria. Our attention was drawn to a small patch in which the residues are different in all MreB paralogues in B. subtilis (Fig. 4A). We therefore exchanged the A41 residue in MreBH for the corresponding S43 residue in BsMreB and expressed the mCherry-MreBHA41S protein in E. coli cells using the same expression conditions as for all other experiments described above. Strikingly, mCherry-MreBHA41S no longer formed exclusively ring structures (Fig. 4B), but much more frequently helical and highly curved filaments that were very similar by eye to those formed by YFP-BsMreB (Fig. 4C and Fig. S2C). Wild-type MreBH formed about 88% ring versus 12% non-ring, while mutant MreBH formed 69% non-ring versus 31% ring structures. The reverse experiment, changing the residue S43 in BsMreB to the corresponding A41 in MreBH, did not show a remarkable change in filament architecture from wild-type BsMreB (data not shown). Also, when the corresponding residue in Mbl, K40, was exchanged to S43 as in BsMreB, this modification did not show any difference in CFP–MblK40S structures (patches) compared with wild-type CFP–Mbl (data not shown). Thus, despite the high variability between MreB proteins in B. subtilis (BsMreB shares 55% identity with Mbl and 57% identity with MreBH), a single amino acid residue at the contact interface can alter filament architecture of one paralogue towards that of another.

Figure 4.

Alteration of MreBH filaments by a single amino acid exchange.
A. Sequences alignment of BsMreB, Mbl and MreBH. Only the N-terminal parts of the sequences are shown. Marked in grey are residues found at the interface interaction in the protofilament (van den Ent et al., 2001). The mutated residue in MreBH is boxed.
B and C. (B) mCherry-MreBH wild-type (WT) compared with (C) mCherry-MreBHA41S. Note that ring-like structures of wild-type MreBH are converted to mostly helical structures of mutant MreBH. Cells were grown in LB to an OD600 of 0.2 and FP fusion proteins were induced with IPTG to the final concentration of 0.25 mM. Scale bars 2 µm.

We further wished to investigate if MreBHA41S is able to alter the helical filament architecture in the native host cells. To this end, mCherry-MreBHA41S was expressed in B. subtilis as the sole source of protein under the control of the xylose promoter. mCherry-MreBHA41S still localized as distinct foci along the lateral cell wall; however, the localization pattern was less regular than the helical structures formed by mCherry-MreBH wild-type protein (Fig. S3A and B). Interestingly, cells showed a clear defect in shape. Ninety per cent (180 cells counted) of the cells were either wider, thinner or bent compared with wild-type cells (Fig. S3), suggesting that the mutation interferes with the proper function of the cytoskeletal element in the native host. We also observed a high frequency of minicells (17%, 240 cells counted), which – unlike minicells reported in minC or minD mutant cells (Levin et al., 1998; Rothfield et al., 2005) – often contained DNA (39%, 118 cells counted) (Fig. S3C). This phenotype has not been observed for any mreB or mbl mutations, and shows that in addition to cell shape maintenance, cell division is also perturbed in MreBHA41S mutant cells.

Expression of MreB orthologues in E. coli alters cell morphology

So far, we had induced B. subtilis MreB paralogues in E. coli cells and followed their localization during the first one hour, in which cell morphology did not change dramatically, to rule out an effect of altered cell shape on the formation of the filaments. We wondered if an extended heterologous expression of BsMreB could have an effect on cell morphology. We therefore expressed BsMreB, Mbl or MreBH in E. coli cells at a moderate level, i.e. at levels in which BsMreB and MreBH form defined filaments, and followed changes in cell morphology for several hours. Strikingly, cells assumed an increasingly altered morphology 1 or 2 h after expression of BsMreB. Both cell width and cell length continued to increase, such that 2 h after expression of BsMreB, cells were highly enlarged (up to 8 µm in length and 6 µm in width – 65 cells measured) and oddly shaped (Fig. 5B, first row, and Movie S1). Contrarily, induction of Mbl led to a mild increase in cell size, the largest cells observed were 4 µm in length and 2 µm in width (54 cells measured), and most cells were oval rather than rod shaped (Fig. 5B, second row). Expression of MreBH had a yet even different effect: E. coli cells continued to be rod shaped with a slightly increased width (1.5 µm – 72 cells measured), but were highly elongated, due to a (direct or indirect) defect on cell division, as observed by the lack of septa within the cell filaments (Fig. 5B, third row). Contrarily, expression of a second copy of EcMreB fused to RFP (EcMreB–RFPSW) (Bendezu et al., 2009) led only to a moderate cell shape defect with an average length and width of 3.4 and 1.54 µm (58 cells measured), respectively (Fig. 5B, lowest row), compared with the wild-type cells (2.74 µm in length and 1.2 µm in width – 35 cells measured). Thus, expression of BsMreB highly enlarges cell size in all dimensions, while Mbl – as expected from its inability to form extended filaments – had only a minor effect on cell morphology.

Figure 5.

Alteration of cell morphology through the expression of BsMreB.
A. E. coli wild-type (WT) cells.
B. E. coli cells expressing BsMreB (first row), Mbl (second row), MreBH (third row) or a second copy of EcMreB fused to RFP (EcMreB–RFPSW) from a plasmid (bottom row). Cells were grown in LB to an OD600 of 0.2 and samples taken for microscopy (‘− IPTG’). Cultures were then induced with 0.25 mM IPTG and samples taken every 30 min for microscopy (‘+ IPTG’). Right panel shows cells after 90–120 min.
C. Few YFP–BsMreB filaments ‘pushing’ the cell membrane resulting into almost square-shape cell.
D. Multiple YFP–BsMreB filaments interconnected ‘pushing’ the cell membrane resulting in round-shape cell. Arrows indicates the direction of filament extension.
E and F. (E) Subcellular localization of EcMreB–RFPSW and (F) dual labelling of EcMreB–RFPSW (red in overlay) with YFP–BsMreB (green in overlay). Green and red arrow heads indicate distinct signals/filaments of the two MreB orthologues. Cells were grown in LB to an OD600 of 0.2 and proteins or FP fusion proteins were induced with IPTG to the final concentration of 0.25 mM. Cells were taken for microscopy 2 h after induction.
Scale bars 2 µm.

Strikingly, the extension of YFP–BsMreB filaments had a visible effect on the extension of the cell envelope. In Fig. 5C and in Movie S1, it can be observed that the cell envelope is pushed and distorted towards the direction of the growing YFP-BsMreB filaments. The phenomenon of cell wall growth following the path of extension of YFP-BsMreB filaments was observed in all of the 83 images and in five time-lapse experiments captured for this study, suggesting that the extension of YFP-BsMreB filaments can exert force onto the cell envelope, which markedly alters cell morphology. Interestingly, elongated or straight YFP–BsMreB filaments were observed to extend the cells in a rather elongated or irregular manner (Fig. 5C, Fig. S1A and Movie S1), while a network of thinner filaments led to the generation of enlarged round cells (Fig. 5D and Fig. S1B and C), supporting the idea that extended BsMreB filaments dictate the direction of cell wall extension. Because it was recently reported that the expression of MreB from the Gram-positive bacterium Clostridium perfringens in B. subtilis affects the integrity of the cytoskeleton which in turn is critical for the cell viability (Schirner and Errington, 2009b), it is possible that the shape defect of E. coli cells observed in this study could be subsequent to the endogenous EcMreB function being perturbed by the expression of BsMreB. To address this issue, we performed YFP–BsMreB dual labelling experiments with EcMreB–RFPSW. EcMreB–RFPSW expressed from a plasmid localized indistinguishably from the pattern shown in the previous report by Bendezu et al. (2009) (Fig. 5E). Figure 5F shows that YFP–BsMreB filaments rarely colocalized with EcMreB–RFPSW filaments indicating that EcMreB–RFPSW and YFP-BsMreB form distinct polymers in E. coli. Even at later time points, when BsMreB had altered normal cell shape, EcMreB–RFPSW signals were not colocalizing with those of YFP–BsMreB (data not shown). Therefore, EcMreB function is not likely to be significantly altered through the expression of BsMreB at least at early time points of coexpression. It is also important to note that a loss of function of EcMreB leads to the formation of round cells (Wachi and Matsuhashi, 1989; Kruse et al., 2005), while the expression of BsMreB induces predominantly irregular bulged and much larger cells (Fig. 5A and C), showing that the expression of BsMreB does not mimic a loss of function of EcMreB, but directly leads to an alteration of cell shape.

It is a formal possibility that BsMreB could affect cell shape in E. coli by repositioning enzymes that mediate cell wall synthesis. However, BsMreB and EcMreB are only 55% identical, so it seems unlikely that BsMreB interacts with MreC, MreD and Pbps in E. coli, all of which are required for proper cell wall synthesis (Kruse et al., 2005; Defeu Soufo and Graumann, 2006; Kawai et al., 2009a). A possible interaction of BsMreB with the interaction partners of EcMreB would imply that BsMreB should at least partially complement the loss EcMreB. To asses this hypothesis, we used a ΔmreB E. coli strain bearing a copy of the mre operon (which contains mreBCD) on a low-copy plasmid R1 derivative. The mre operon on the plasmid can be depleted via induction of CopA antisense RNA, which will instantaneously stop the replication of plasmid R1 (Kruse et al., 2005). This way, YFP–BsMreB could be induced from an additional plasmid during depletion of EcMreB. As expected, Fig. 6B shows that E. coli cells became cocci-like shape when EcMreB was depleted compared with the normal rod cell shape in Fig. 6A. Interestingly, when YFP-BsMreB was expressed while EcMreB was depleted, the cells gradually remodelled from rod shape to enlarged oval cells (Fig. 6C–E), but cells continued to grow as poorly as mreB mutant cells, showing that BsMreB is not able to complement the loss of EcMreB. This finding supports the idea that BsMreB is unlikely to interact with E. coli proteins involved in cell morphogenesis.

Figure 6.

Complementation analysis of EcMreB-depleted cells with BsMreB.
A and B. ΔmreB (MC1000ΔmreB/pTK549/pKG339) E. coli strain bears mre operon on pTK549 plasmid (derivative of plasmid R1) and also possesses a second plasmid (pKG339) which carries an IPTG-inducible CopA antisense RNA (the regulator of the frequency of the replication of plasmid R1). Induction of extra copies of CopA stopped the replication of pTK549 that led to the depletion of the EcMreB in the offspring which turned from rod-shaped to cocci-like-shaped cells (B) while the cells grew like wild type in the absence of IPTG (A).
C and D. A third plasmid carrying YFP–BsMreB under the control arabinose promoter was introduced in MC1000ΔmreB/pTK549/pKG339 strain. Induction of YFP–BsMreB or BsMreB (data not shown) did not restore the normal cell shape in EcMreB-depleted cells. Note that at lower concentration of arabinose (+, 0.002%), the appearance of YFP–BsMreB filaments was delayed (C and D) compared with higher concentration (++, 0.02%) (E).
Scale bars 2 µm.

BsMreB induces lysis of E. coli cells having a weakened cell wall

The idea that BsMreB filaments may mechanically influence cell shape through the modification of the cell wall extension prompted us to address the question whether BsMreB filaments mechanically push E. coli cells into a different cell shape by exerting force onto the inner cell membrane (and indirectly onto the cell wall). We reasoned that the cell wall of E. coli cells can resist the force from growing MreB filaments, but would enlarge according to the direction of internal force. To investigate this possibility, we treated E. coli cells with sublethal concentrations of penicillin to weaken cell wall rigidity. These cells grew more poorly than non-treated cells, as seen by spot-titre experiments (Fig. 7A) and were highly filamentous due to an inhibition of the formation of division septa (Fig. S4), but were able to grow in liquid medium. Strikingly, when BsMreB was induced in penicillin-treated cells, cell growth was seriously compromised as cell survival sharply dropped due to cell lysis (Fig. 7A and Fig. S5A). Cell rupture could clearly be observed, at which places usually a large amount of DNA was released (Fig. 7B and C). Interestingly, YFP-BsMreB filaments were usually visible pointing towards or through the site of cell rupture (Fig. 7B and C). Conversely, expression of BsMreB in non-treated cells had a very mild effect on the doubling time of the culture (Fig. 7A and Fig. S5A), but did not induce any detectable cell rupture. A similar effect on the viability of E. coli cells was observed when BsMreB was expressed in EcMreB-depleted cells. The more BsMreB was expressed the stronger growth was negatively affected (Fig. S5B and C), suggesting that E. coli cells without EcMreB are as fragile as ampicillin-treated cells and are prone to lysis when BsMreB is expressed. Thus, weakening of the cell wall whether it is with sublethal concentrations of penicillin or with EcMreB depletion during expression of BsMreB leads to cell rupture, supporting the idea that MreB filaments may exert considerable force onto the cell envelope from underneath the cell membrane. However it remains possible that the observed cell lysis could be due to or exacerbated by the malfunction of any E. coli morphogenic protein consequent to cell deformation.

Figure 7.

Effect of BsMreB induction on cells with weakened cell wall.
A. Spot-titre experiment showing that the growth of E. coli cells treated with ampicillin is abolished during expression of BsMreB. Overnight E. coli cells that contained YFP–BsMreB expressing plasmid were diluted as indicated (3.5 × 10−1 to 10−6 at OD600) and 10 µl were spotted on LB agar supplemented with appropriate concentration of antibiotics and/or inducer as indicated. Addition of IPTG leads to the depletion of EcMreB, while arabinose induces BsMreB. The plates were incubated at 37°C overnight.
B and C. Cells releasing DNA due to the rupture of the cell envelope (white arrowheads). Green arrowheads indicate the remains of YFP–BsMreB filaments oriented towards the site of cell rupture. Cells were grown in LB broth until OD600 of 0.2 at 37°C (from a single colony) and further supplemented with 0.5 µg ml−1 ampicillin or 0.25 mM IPTG. Pictures were taken after an additional 1 h growth. [Note that cells expressing IPTG-inducible YFP–BsMreB were more sensitive to ampicilin and IPTG on LB agar plate compared with LB broth (see Fig. S4)]. Scale bars 2 µm.

We have found that overexpression of EcMreB also caused cell deformation and led to the lysis of cells with weakened cell wall, as seen by spot-titre experiments and fluorescence microscopy (data not shown). This observation implies a possible mechanical effect of EcMreB against the cell membrane and suggests that while a normal amount of EcMreB is important for the mechanical integrity of the cell (Wang et al., 2010), an overdose of the protein alters the filamentous structures to push the cells into an enlarged form.


The discovery that bacteria contain an actin-like cytoskeleton has been one of the most unexpected discoveries in recent years. It has become clear that the actin-like MreB proteins form helical filamentous structures underneath the cell membrane in many rod-shaped bacteria, and are essential for cell shape maintenance and for viability in a wide variety of bacteria (Carballido-Lopez, 2006; Pichoff and Lutkenhaus, 2007; Graumann, 2009). Key questions that need to be answered to understand how MreB can achieve its function at a molecular basis are how the protein obtains its helical localization, and what the functions of the cytoskeletal structure are. Our work provides several conceptually novel insights into BsMreB, and its two paralogues Mbl and MreBH (many bacteria – like B. subtilis– contain several MreB paralogues). BsMreB and its paralogues Mbl and MreBH differentially affect cell morphology (Jones et al., 2001; Defeu Soufo and Graumann, 2003), while all three proteins colocalize to a helical structure underneath the cell membrane that continuously remodels in a timescale of seconds to minutes (Carballido-Lopez and Errington, 2003; Defeu Soufo and Graumann, 2004; 2006; Carballido-Lopez et al., 2006). Our finding that the three MreB paralogues form strikingly different filamentous structures by themselves, and that filament architectures change when two paralogues or three are coexpressed in a heterologous cell system, provides an explanation of why cell shape changes in different way in mreB, mbl or mreBH mutants in B. subtilis. Our observations also suggest that B. subtilis mreB mutant cells can grow in a special magnesium and sucrose-complemented medium because Mbl and MreBH suffice to form filamentous structures, whereas the depletion of Mbl in mreB mutant cells is lethal, most likely because MreBH does not form helical filaments by itself.

It has become clear that the BsMreB/Mbl/MreBH cytoskeletal structure positions enzymes that are involved in the de novo synthesis of the cell wall, suggesting that the incorporation of new cell wall material along the lateral cell wall is mediated – at least in part – by the MreB helix (Figge et al., 2004; Dye et al., 2005; Kawai et al., 2009a). Indeed, incorporation of new cell wall material occurs along a helical path along the length of the cell, as well as in the cell middle once cell division has commenced. Although MreB is essential for viability under standard growth conditions (i.e. in LB medium or minimal medium), an mreB null mutant is viable under conditions of high magnesium ions and high osmotic pressure in the medium (Formstone and Errington, 2005). Under these conditions, cells grow relatively well, but are sensitive to osmotic salt stress (Kawai et al., 2009b), and depend on the presence of Mbl (Defeu Soufo and Graumann, 2006). These experiments suggest that the MreB/Mbl/MreBH helix may also provide a mechanical function in stabilizing the cell envelope. Indeed, we provide evidence in support of the idea that MreB filaments exert force onto the cell envelope when expressed in E. coli cells, leading to a remodelling of the shape of the cells. Thus, in addition to cresentin (C. crescentus intermediate filament-like protein), which has been reported to mechanically induce bacterial cell curvature (Cabeen et al., 2009), a bacterial actin-like protein can also mechanically distort E. coli cells. This idea suggests that the ability of actin and intermediate filaments to generate mechanical strength within cells could be conserved from prokaryotic to eukaryotic cells. Moderate expression of BsMreB greatly increases the width and length of E. coli cells, and strongly alters cell morphology into oddly shaped giant cells. The finding that weakening of the cells leads to cell rupture after induction of BsMreB filaments suggests that these structures can push the cell envelope into a different architecture and even rupture a weakened cell wall. We speculate that in Gram-positive bacteria, which have a much higher internal turgor than Gram-negative bacteria, yet are generally much more osmotically resistant, need an intact MreB cytoskeletal structure to mechanically stabilize the cell envelope.

We further provide evidence that MreB paralogues from B. subtilis form filaments with strikingly different architectures. While MreB from T. maritima forms rather straight filaments or rings in vitro (van den Ent et al., 2001), MreB from B. subtilis polymerizes into dynamically remodelling helical filaments underneath the cell membrane (Jones et al., 2001; Defeu Soufo and Graumann, 2004), as does MreB from C. crescentus (Figge et al., 2004; Gitai et al., 2004). MreB from E. coli also forms helical filaments, which appear to be static, though (Thanedar and Margolin, 2004). However, when expressed in E. coli cells, BsMreB forms irregular helices and often rather straight filamentous stretches, showing that cellular factors must exist in B. subtilis cells to convert MreB filaments into a regular helix. One such factor appears to be the other two B. subtilis MreB paralogues. In striking contrast to any MreB-like protein characterized so far, MreBH forms ring structures in the heterologous cell system. Thus, B. subtilis MreB paralogues can have drastically different architectures when expressed on their own. However, when coexpressed, BsMreB and MreBH form much more regular helical structures than BsMreB by itself. When Mbl, the third paralogue, is also expressed concomitantly, the three proteins colocalize into a single filamentous structure that most closely resembles filaments formed in B. subtilis, although Mbl does not form extended filaments when expressed by itself. Evidently, the three paralogues co-polymerize and influence the tripartite structure in a way that it assumes a relatively regular helical arrangement. Our results also demonstrate that B. subtilis MreB paralogues not only colocalize and interact with each other, but indeed form a coherent structure that is influenced by each individual MreB protein.

A further striking observation is our finding that a single amino acid exchange in MreBH that alters a subunit interface contact towards that of BsMreB can drastically alter MreBH filament architecture and convert the ring structures into BsMreB-like helices and straight filaments. Thus, we not only show the first visible proof for strikingly different polymerization properties of bacterial MreB proteins, but also that they are convertible. These data demonstrate that MreB proteins are ideally suited for a synthetic biology approach and for rational design of filamentous structures.

Like ParM (Garner et al., 2004) or EcMreB [expressed in yeast cells (Srinivasan et al., 2007)], BsMreB or Mbl or MreBH filaments grow by extension at both ends, showing that this is likely a general property for prokaryotic actin-like proteins. Our FRAP experiments show that BsMreB filaments also turn over along their length, in a non-polar fashion. These data suggest that MreB structures consist of several overlapping shorter filaments that form a bundle, and that the smaller filaments continuously turn over. Interestingly, the extension of filaments of BsMreB expressed in E. coli cells was slower than in B. subtilis cells, suggesting that regulatory factors may exist in B. subtilis cells that increase filament extension kinetics. We also found that ends of BsMreB filaments can retract after a period of extension, reminiscent of dynamic instability of microtubules (Kueh and Mitchison, 2009) and of ParM protein (Garner et al., 2004). However, it is still not clear if BsMreB filaments consist of parallel or antiparallel bundles of filaments, so MreB and ParM cannot be directly compared.

In toto, although we cannot rule out an indirect, non-mechanical effect of expression of BsMreB in a heterologous cell system, our results suggest that the polymerization of MreB into filaments underneath the cell membrane provides an internal force against the cell, in agreement with the recent report that MreB contributes to the stiffness of the E. coli cell (Wang et al., 2010). We show that B. subtilis MreB paralogues have different filament architectures, but form a single filamentous structure together, such that the loss of one paralogue affects the structure of the mixed polymer formed by the other paralogues. In vitro investigation of the dynamics and physical properties of MreB will further help to understand better the cellular function of this class of proteins.

Experimental procedures

Strain and plasmid construction

All Bacillus strains and plasmids used in this work are listed in Tables S1 and S2 respectively. E. coli XL1 blue or DH5α were used for propagation and maintenance of plasmids while E. coli Bl21 Star™ (DE3) from Invitrogen was used as heterologous system for protein expression. E. coli strains that expressed BsMreB or YFP–BsMreB were obtained by transforming E. coli Bl21 Star™ (DE3) competent cells with plasmids pJS63 (sm Strep-tag-mreB) or pJS64 (sm Strep-tag-yfp-mreB), respectively, selecting for streptomycin. Mbl, CFP–Mbl, MreBH, mCherry-MreBH or mCherry-MreBHA41S E. coli-expressing strains were generated by transforming E. coli Bl21 Star™ (DE3) competent cells with plasmids pJS68 (bla Strep-tag-mbl), pJS69 (bla Strep-tag-mCer-mbl), pJS70 (bla Strep-tag-mreBH) or pJS72 (bla Strep-tag-mCherry-mreBH) or pJS74 (bla Strep-tag-mCherry-mreBHA41S), respectively, selecting for ampicillin. To generate the strains expressing a second copy of E. coli MreB fused to RFP, plasmid pJS75 (bla Strep-tag-EcmreB–RFPSW) or pJS76 (sm Strep-tag-EcmreB–RFPSW) was used to transform E. coli Bl21 Star™ (DE3) competent cells having, respectively, an ampicillin- or a streptomycin-resistant strains.

For dual- and triple-expression strains we used the advantage offered by pETduet-1 and pCDFDuet™-1 of Duet system vectors (Invitrogen). Both vectors have two open reading frames (ORFs) or different compatible replicons and different drug resistance that can allow when using together, cloning and expression of up to four proteins in a single E. coli strain. For dual-expression strains: YFP–BsMreB/CFP–Mbl, YFP–BsMreB/mCherry-MreBH or YFP–BsMreB/EcMreB–RFPSW, plasmids pJS64 (sm Strep-tag-yfp-mreB in pCDFDuet™-1) and pJS69 (bla Strep-tag-mCer-mbl in pETduet-1), pJS64 (sm Strep-tag-yfp-mreB in pCDFDuet™-1) and pJS72 (bla Strep-tag-mCherry-mreBH in pETduet-1) or pJS64 (sm Strep-tag-yfp-mreB in pCDFDuet™-1) and pJS75 (bla Strep-tag-EcmreB–RFPSW in pETduet-1) were simultaneously used to transformed E. coli Bl21 Star™ (DE3) competent cells selecting for streptomycin and ampicillin. CFP–Mbl/mCherry-MreBH dual-expression strain was generated by transforming E. coli Bl21 Star™ (DE3) competent cells with pJS73 (bla Strep-tag-mCherry-mreBH and Strep-tag-mCer-mbl) bearing both cfp-mbl (in the first ORF) and mCherry-MreBH (in the second ORF) constructs selecting with ampicillin. Triple-expression YFP–BsMreB/CFP–Mbl/mCherry-MreBH strain was generated by transforming at once E. coli Bl21 Star™ (DE3) competent cells with pJS64 (sm Strep-tag-yfp-mreB) and pJS73 (bla Strep-tag-mCherry-mreBH and Strep-tag-mCer-mbl) selecting for streptomycin and ampicillin.

For complementation experiments, pJS78 (cat PBAD -mreB) or pJS79 (cat PBAD-yfp-mreB) plasmids were used to transformed E. coli MC1000ΔmreB/pTK549/pKG339 strain (Kruse et al., 2005) competent cells selecting for kanamycin (pTK549), tetracycline (pKG339) and chloramphenicol (pJS78 or pJS79). Beforehand, the cloramphenicol resistance cassette (inserted in the E. coli chromosome during mreB deletion) was removed from MC1000ΔmreB/pTK549/pKG339 strain using plasmid pCP20 as described previously (Datsenko and Wanner, 2000). IPTG leads to the repression of EcMreB expression, while arabinose induces BsMreB or YFP-BsMreB expression.

To investigate the MreBHA41S mutant in B. subtilis, pJS77 (Pxyl-NterComGA-mCherry-mreBHA41S) was used to transform B. subtilis wild-type competent cells selecting for chloramphenicol. The resulted strain JS100 (Pxyl-mCherry-MreBHA41S) owned mCherry-mreBHA41S under the control of the inducible xylose promoter at the original locus, as verified by sequencing.

Growth conditions

Escherichia coli transformants were selected on LB agar supplemented with 100 µg ml−1 ampicillin, 50 µg ml−1 streptomycin, 25 µg ml−1 chloramphenicol, 50 µg ml−1 kanamycin and/or 10 µg ml−1 tetracycline. Strains were grown in LB medium supplemented, where required, with 1% glucose (to induce catabolite repression), 0.25 µM IPTG or different concentrations of arabinose (to induce the protein expression of interest). Experiments were always carried out with fresh transformants.

Restricted experimental conditions are stated in the figure legends.


Fluorescence microscopy was performed using a Zeiss Axio Imager A1 equipped with a Cool SNAP HQ camera and a TIRF objective with an aperture of 1.45. For FRAP experiments, we used a Zeiss Axio Observer Z1 (inverted microscope) equipped with a Cascade II 512 camera and an external lasers source. The specimen was bleached and the fluorescence recovery was monitor with lasers beams of 405 and 488 nm respectively.

Picture acquisition and analysis were performed with Metamorph 6.5 (MDS Analytical Technologies). DNA was stained with 4′,6-diamidino-1-phenylindole (DAPI; final concentration 0.2 ng ml−1) and membranes were stained with FM4-64 (final concentration 1 nM).


We thank Piet de Boer of Case Western Reserve University and Kenn Gerdes of Newcastle University for the generous gift of the E. coli strains FB76 and MC1000ΔmreB/pTK549/pKG339, MC1000/pBAD33 and pCP20 BT340 ApR pSC101 repA(ts) 30°C respectively. We also thank Obaidur Rahman of Newcastle University for kindly providing us with the information about the strains from Kenn Gerdes Lab. This work was supported by FOR929 from the Deutsche Forschungsgemeinschaft.