Bacterial actin-like proteins play a key role in cell morphology and in chromosome segregation. Many bacteria, like Bacillus subtilis, contain three genes encoding actin-like proteins, called mreB, mbl and mreBH in B. subtilis. We show that MreB and Mbl colocalize extensively within live cells, and that all three B. subtilis actin paralogues interact with each other underneath the cell membrane. A mutation in the phosphate 2 motif of MreB had a dominant negative effect on cell morphology and on chromosome segregation. Expression of this mutant allele of MreB interfered with the dynamic localization of Mbl. These experiments show that the interaction between MreB and Mbl has physiological significance. An mreB deletion strain can grow under special media conditions, however, depletion of Mbl in this mutant background abolished growth, indicating that actin paralogues can partially complement each other. The membrane protein MreC was found to interact with Mbl, but not with MreB, revealing a clear distinction between the function of the two paralogues. The phosphate 2 mutant MreB protein allowed for filament formation of mutant or wild-type MreB, but abolished the dynamic reorganization of the filaments. The latter mutation led to a strong reduction, but not complete loss, of function of MreB, both in terms of chromosome segregation and of cell morphology. Our work shows that that the dynamic localization of MreB is essential for the proper activity of the actin-like protein and that the interactions between MreB paralogues have important physiological significance.
Actin is a cytoskeletal element that provides vital functions in eukaryotic and in prokaryotic cells. In eukaryotes, actin filaments give mechanical strength to cells, but also supply dynamic properties, and serve as structural fibres in muscle contraction. Additionally, actin proteins have motor-like functions (Marx, 2003; Mogilner and Oster, 2003; Upadhyaya and van Oudenaarden, 2003), most notably in cell migration through pushing of membranes, but also in intracellular movement of vesicles. In vitro, actin filaments can grow and shrink, and are able to deform vesicles and thus push membranes, providing the force to elongate cellular extensions such as pseudopods (Lauffenburger and Horwitz, 1996; Mitchison and Cramer, 1996). Many bacteria also possess actin-like proteins (van den Ent et al., 2001), a single copy (in proteobacteria, e.g. Escherichia coli or Caulobacter crescentus), two paralogues (e.g. Thermotoga maritima, Fusobacterium nucleatum) or three paralogues (in Bacillus subtilis and in many Gram positives, and in some cyanobacteria, e.g. Gloeobacter violaceus). Actin-like protein MreB is essential for cell viability (Jones et al., 2001), and appears to perform a dual function, both in chromosome segregation and in maintenance of proper cell shape (Graumann, 2004). However, the true function and molecular mechanism of bacterial actins are still unclear.
MreB (and its paralogues in B. subtilis, Mbl and MreBH) forms helical filaments underneath the cell membrane in E. coli, C. crescentus and B. subtilis, which are highly dynamic (Jones et al., 2001; Carballido-Lopez and Errington, 2003; Shih et al., 2003; Defeu Soufo and Graumann, 2004; Figge et al., 2004). MreB and Mbl appear to move along helical tracks, with a speed of about 0.1 μm s−1, providing a potential motor-like force. Actin polymerizes into a two-stranded right handed helix through addition of ATP-bound actin monomers. Actin movement most likely arises through growth at the barbed end of the filament, while actin is released from the pointed end following ATP hydrolysis (a process termed treadmilling). Active pushing is thought to occur through binding of actin monomers to the tip of the filament when the object moves away, thus preventing backward movement, such that the object is driven by Brownian diffusion, with the actin filament dictating a single direction (polymerization rachett) (Mogilner and Oster, 2003).
During the depletion of B. subtilis MreB or of Mbl, or of C. crescentus MreB, origin regions on the chromosomes fail to separate properly, leading to a severe (or in case of Mbl moderate) segregation defect (Defeu Soufo and Graumann, 2003; Gitai et al., 2005), likewise to overproduction of a dominant negative mreB allele in E. coli (Kruse et al., 2003). In support of an active role in segregation, MreB appears to be associated with the nucleoids in B. subtilis (Defeu Soufo and Graumann, 2004), and has been shown to directly or indirectly interact with the origin regions on the chromosome in C. crescentus (Gitai et al., 2005). Interestingly, E. coli MreB interacts with RNA polymerase (Kruse et al., 2006), which has also been implicated as an important factor mediating movement of origin regions towards opposite cell poles (Dworkin and Losick, 2002). Because the general direction of transcription is away from the origin region on both arms of the chromosome, MreB-anchored RNA polymerase may push origin regions towards the cell poles, providing the force and direction for pole-ward DNA segregation. The function of bacterial actin orthologues is best understood in case of an E. coli plasmid segregation system. Plasmid-encoded E. coli ParR protein binds to a specific cis site on the duplicated plasmids, which are located close to the cell centre, and induces polymerization of the ParM actin homologue (Moller-Jensen et al., 2002). ParM filaments contain plasmids at their pole ward ends, so ParM filaments appear to push plasmids towards each cell pole (Moller-Jensen et al., 2003). On the other hand, B. subtilis MreB and Mbl, E. coli MreB and C. crescentus MreB strongly affect the formation of proper cell shape (Jones et al., 2001; Figge et al., 2004; Gitai et al., 2004), while B. subtilis MreBH has a mild effect on cell shape (Defeu Soufo and Graumann, 2003). Interestingly, Mbl (but not MreB) was implicated in the proper insertion of new cell wall material into the growing peptidoglycan layer, which also apparently follows a helical pattern (Daniel and Errington, 2003), although this function of Mbl has recently been questioned (Tiyanont et al., 2006). Thus, it appears that MreB orthologues provide two essential but highly distinct functions. A direct role for MreB-like proteins in the insertion of new cell wall material is suggested by recent finding showing that: (i) E. coli MreB interacts with the MreC membrane protein (Kruse et al., 2005), which is equally essential for viability and the formation of rod cell shape (Lee and Stewart, 2003), and that the depletion of MreC affects the formation of MreB helical structures (Defeu Soufo and Graumann, 2005); (ii) that C. crescentus MreC interacts with Pbps, which extend the peptidoglycan layer and which also appear to localize in a helical pattern in C. crescentus and in B. subtilis (Figge et al., 2004; Scheffers et al., 2004; Divakaruni et al., 2005); (iii) that MreC also appears to localize in a helical pattern within the membrane (Divakaruni et al., 2005; Leaver and Errington, 2005); and (iv) that the specific localization of C. crescentus Pbps is lost in the absence of MreC or of MreB, although it is retained after inhibition of filament formation of MreB (Dye et al., 2005; Figge et al., 2004). However, C. crescentus MreB does not strictly colocalize with MreC, and the specific localization of MreB or of MreC is not perturbed in the absence of each other (Dye et al., 2005), and the specific localization of B. subtilis Pbps is not perturbed in the absence of MreB or of Mbl (Scheffers et al., 2004), so it remains unclear how bacterial actin-like proteins influence the organization of cell wall synthesis and seemingly in parallel, organize or drive chromosome segregation.
We have found that the localization of B. subtilis MreB is affected by Mbl and MreBH, as well as by MreC and MreD (two membrane-bound cell shape determinants) (Defeu Soufo and Graumann, 2005), showing that an interplay exists between actin paralogues and MreCD membrane proteins. We have sought to gain further insight into the connection between B. subtilis actin paralogues, and also to address the question if the apparent movement of MreB is important for its function. We found that MreB and Mbl frequently colocalize and interact at many sites along the membrane, revealing a physical interaction between the two proteins. We also found that a mutant form of MreB that is unable to localize dynamically is strongly defective in chromosome segregation and in cell morphology, showing that the formation of dynamic filaments is required for proper function of MreB.
A mutation in the phosphate 2 motif interferes with the dynamics of MreB filaments
We wished to investigate if mutations affecting ATPase activity in eukaryotic actin interfere with the localization of MreB. A precedent for this is a mutation in the highly conserved phosphate 2 motif, which strongly reduces ATPase activity, but not ATP binding, in actin (Kabsch and Holmes, 1995; Posern et al., 2002), and results in a dominant negative phenotype in E. coli (Kruse et al., 2003). We introduced the corresponding mutation, D158A, into MreB, and expressed MreB D158A ectopically, driven by the hyperspank promoter that is IPTG-inducible. While wild-type GFP-MreB [which is fully functional, i.e. cells expressing the GFP fusions as sole copy of the corresponding protein grew with indistinguishable doubling time and with indistinguishable cell morphology compared with wild-type cells (Defeu Soufo and Graumann, 2004)] expressed in a similar way, or from its original position (as a sole source of MreB) forms helical structures underneath the cell membrane and along the nucleoids (Fig. 1A, note that GFP-MreB is generally absent from the cell poles), GFP-MreBD158A formed filamentous structures that were much more irregular than those formed by wild-type GFP-MreB (Fig. 1B, > 300 cells analysed). That is, while the maximum pitch of the filaments (or spacing between distinct GFP-MreB signals, in case of discontinuous filaments) along the membrane within the cell was 0.6 μm (± 0.14 μm), spacing between GFP-MreBD158A signals was up to 1.3 μm (± 0.25 μm, n = 80). In addition, cells expressing GFP-MreBD158A were considerably wider and larger than wild-type cells [average width of 1.6 μm (± 0.23 μm) versus 1.1 μm (± 0.1 μm, n = 220), maximum length of 15 versus 4.4 μm, respectively], and were frequently curved and twisted (Fig. 1B). GFP-MreBD158A also formed filamentous structures in anucleate cells (Fig. 1B, grey arrowhead), in contrast to wild-type MreB under our experimental conditions (Defeu Soufo and Graumann, 2004). To test if GFP-MreBD158A also forms filamentous structures in the absence of wild-type MreB, we introduced an in frame deletion at the mreB locus (kind gift from J. Errington, Newcastle, UK) into strains expressing GFP-MreB or mutant GFP-MreB from the ectopic chromosomal site. The pattern of localization of GFP-MreB or of GFP-MreBD158A was very similar in the presence or absence of MreB expressed from the original locus (compare Fig. 1C with A, and D with B respectively). These experiments show that a mutation in the phosphate 2 motif of MreB allows for the formation of helical structures, but interferes with the formation of proper helical filaments.
Because the D158A-mutation may stabilize MreB filaments, due to its potentially lower ATPase activity, we performed time-lapse microscopy on cells expressing wild type or D158A-mutant GFP-MreB. Within 10 s time intervals, GFP-MreB filaments changed their localization underneath the membrane, they appeared to move along the membrane in a helical path (Fig. 2A and Movies S1 and S2). In striking contrast to this, D158A-mutant GFP-MreB filaments did not change their localization during the course of the experiment (Fig. 2B, Movies S3 and S4). Solely back and forth movement caused by focal drift is apparent in Fig. 2B and in Movies S3 and S4, but the position of GFP-MreBD158A signals is constant between the first and last time interval (Fig. 2B). D158A mutant GFP-MreB did not show any dynamic localization in the presence or absence of wild-type MreB (Fig. 2B and data not shown).
To assess the dynamics of GFP-MreB filaments through another technique, we performed fluorescence recovery after photobleaching (FRAP), using an Argon laser that is focused on a circular spot of approximately 1 μm diameter. This way, a fraction of a B. subtilis cell can be bleached in an epifluorescence set-up, and fluorescence recovery can be monitored by time-lapse epifluorescence. GFP-MreB rapidly regained fluorescence, 1 min after bleaching, discrete GFP-MreB signals were apparent within the bleached area, and after 2 min, a clear helical pattern of localization is visible in Fig. 3A. On average, 2.2 (± 0.2) min were required until clear GFP-MreB filaments could be seen within a bleached cell area with 1 μm diameter (with eight experiments performed). Contrary to wild-type GFP-MreB, GFP-MreBD158A did not recover fluorescence for an extended time. In Fig. 3B, a stretch along the cell membrane is bleached, and no fluorescence recovery is apparent after 6 min; full fluorescence is visible after 60 min, and on average, 50 (± 5) min were required for D158A-mutant GFP-MreB to regain full fluorescence (with six experiments performed). Similarly long recovery of fluorescence was observed for mutant GFP-MreB in the absence of wild-type MreB (data not shown). Therefore, mutant GFP-MreB filaments were rigid and did not show dynamic localization like wild-type MreB. Thus, residue D158 is essential for dynamic localization of MreB, possibly affecting polymer turnover and/or disassembly.
D158A-mutant MreB is dominant negative and affects the localization of wild-type MreB and of Mbl
When D158A-mutant GFP-MreB was ectopically expressed, a strong influence on cell morphology was observed. During expression of mutant GFP-MreB, cells were much larger and longer (indicating a retardation of cell division) than wild-type cells (see above), and 78% were highly curved (Fig. 1C). Identical observations were made after expression of mutant GFP-MreB or mutant MreB in the absence of wild-type MreB (Fig. 1D, Fig. S2). Mostly large single nucleoids were observed in normal-sized cells (< 4 μm) expressing mutant MreB, and only 4% of these cells had two separated nucleoids (> 350 cells analysed). In contrast, 34% of wild-type cells (and thus of cells up to 4 μm) contained two nucleoids (Fig. 1A, compare with B). Two separated nucleoids were usually found in very large mutant cells (> 6 μm), and 1% anucleate cells were observed. Thus, D158A-mutant MreB displays a strikingly similar phenotype to that of the corresponding mutation in E. coli (Kruse et al., 2003), it is dominant negative, shows a strong block in chromosome segregation, and a defect in the formation of proper cell shape.
To test if mutant MreB affects the localization of wild-type MreB, or forms independent, aberrant structures, we induced D158A-mutant MreB ectopically in a strain expressing GFP-MreB from the original mreB locus. In the absence of mutant MreB, GFP-MreB localized normally (Fig. 4A), but in the presence of the mutant version, GFP-MreB displayed aberrant filamentous structures: these were highly similar to those formed by D158A-mutant GFP-MreB in 75% of the cells (Fig. 4B, left panel), while in 25% of the cells, more dot-like GFP-MreB signals were apparent (Fig. 4B, right panel, > 300 cells analysed). Both kinds of cells showed aberrant nucleoid morphology and cell morphology. Time-lapse microscopy revealed that the dynamic localization of GFP-MreB was markedly reduced in the presence of mutant MreB (data not shown), showing that D158A-mutant MreB strongly influences the formation of wild-type MreB filaments, and suggesting that wild type and mutant MreB form mixed filaments in which the mutant protein blocks the dynamics of assembly/disassembly reactions.
To test if mutant MreB affects the localization of Mbl, we expressed D158A-mutant MreB in a strain expressing GFP-Mbl as sole source of Mbl [GFP-Mbl is functional (Defeu Soufo and Graumann, 2004)]. Induction of MreBD158A had a strong effect on GFP-Mbl, whose localization changed from a more dot-like appearance in wild-type cells (Fig. 4C) to a clear band-like pattern (Fig. 4D), and now rather resembled that of D158A-mutant GFP-MreB (Fig. 1B). Even more strikingly, induction of mutant MreB abolished dynamic localization of GFP-Mbl filaments. In the absence of mutant MreB, Mbl dots localized dynamically along the membrane in a helical pattern (Fig. 4E, Movie S5), however, after induction of mutant MreB, the filaments formed by GFP-Mbl were rather static. Only focal drift is apparent in Movie S6, the position of filaments remains relatively constant over the time of the experiment (Fig. 4F). Thus, a mutation in MreB affects the localization and kinetics of Mbl filaments.
Inhibition of MreB dynamics strongly interferes with its function, but allows for cell survival
A strain carrying an in frame mreB deletion can be propagated in a special medium (PAB-SMM) containing high magnesium and sucrose concentrations (Formstone and Errington, 2005). Although it was reported that a strain carrying a mreB deletion grows like wild-type cells in PAB-SMM medium, we found a high number of mreB mutant cells (about 50%, > 350 cells analysed) display cell shape and chromosome condensation defects in PAB-SMM medium (Fig. S1). We have noted that survival of mreB mutant cells in PAB-SMM medium is not solely due to high magnesium and sucrose concentrations, because rich [Luria–Bertani (LB)] or minimal media supplemented with identically high magnesium and sucrose concentrations do not sustain growth of the mutant cells. Therefore, additional media effects in PAB-SMM medium must exist that allow for growth of B. subtilis cells lacking MreB, and subtle differences in PAB medium may explain the discrepancy of observations of the mreB mutant phenotypes. In any event, because an mreB deletion is lethal in normal growth medium (Jones et al., 2001; Defeu Soufo and Graumann, 2003), the ability of the mreB deletion strain to grow in PAB-SMM medium is a powerful tool to investigate the functionality of mutations in MreB. To test if the D158A mutation in MreB can support MreB function, we transformed the mreB deletion strain grown in PAB-SMM medium with wild type or mutant mreB alleles expressed ectopically, and plated the cells on normal growth medium. To ensure that wild type and mutant MreB are expressed at equal level, we performed Western blot analysis, which showed that GFP-MreB and GFP-MreBD158A are expressed at similar levels (Fig. 5A, lanes 2 and 3). While transformation with IPTG-driven wild-type mreB resulted in normally growing colonies, only poorly growing colonies were obtained using D158A-mutant mreB, and no transformants were obtained in the absence of added DNA (Fig. 5B). More than 85% of all cells expressing D158A-mutant MreB only showed highly abnormal cell morphology, similar to cells expressing wild type and mutant MreB (of which 50% show a defect in morphology, see Fig. 1B). All of the 15% of cells showing only a mild cell shape defect (only wider and curved cells) contained highly abnormal nucleoids (Fig. S2), showing that dysfunctional MreB does not support proper chromosome segregation. To test if MreB is overexpressed in our experiments, we grew ΔmreB cells expressing MreB (or GFP-MreB) with different concentrations of IPTG. Only cells grown with 1 mM IPTG (maximum induction) grew like wild-type cells, lower concentrations of IPTG led to the generation of cells having cell shape and chromosome segregation defects (data not shown), showing that our experiments were performed under MreB levels that reach physiological levels. The fact that GFP-MreB complemented for growth, cell shape and chromosome segregation in ΔmreB cells like MreB underscores that the GFP fusion is functional. These results show that a mutation in the phosphate 2 motif interferes with the function of MreB. Importantly, the above results show that dynamic localization of MreB is important for the proper function of the protein, because D158A-mutant MreB can still form filamentous structures, but these are rigid rather than dynamic. However, dynamic reorganization of MreB filaments is not essential for cell viability, because the D158A-mutant allele can sustain cell growth, albeit very poorly.
MreB and Mbl colocalize underneath the cell membrane
Because MreB and Mbl influence each other in terms of the pattern of localization and dynamics, we wished to investigate whether MreB and Mbl form distinct intracellular filamentous structures, or if a spatial connection exists between the two major actin-like proteins. Attempts to generate YFP or CFP fusions of MreB or of Mbl failed, when GFP variants with eukaryotic codon usage were employed. We mutagenized GFP to YFP, and, as suggested by Veening et al. (2004), added a 12 aa sequence from a well-translated B. subtilis protein to the N-terminus of a variant of eCFP (Rizzo et al., 2004), which resulted in fully functional fusion proteins, indicating that indeed, codon usage caused a problem for the expression of the GFP variants, when present at the N-terminus of proteins (eYFP and eCFP have been fully functional with various C-terminal fusions generated in our laboratory). YFP-MreB and CFP-Mbl showed somewhat distinct fluorescent signals (there was no crossbleeding between channels, Fig. S3), while CFP-Mbl localized mostly in dot-like structures and rarely as clear filament, YFP-MreB formed both filamentous and dotted structures (Fig. 6A). Although a considerable number of signals were well separated from each other (35%), cells contained a high number of signals that were adjacent (< 0.25 μm distance, 20%) or coincident (45%, 180 cells analysed with the region measurement tool in Metamorph 6.0). However, it is important to include in these studies the fact that both, MreB and Mbl filaments show dynamic localization, and movement or extension of filaments in a second-time scale, such that during switching of CFP and YFP filters, filaments have moved to some degree. Therefore, colocalization may be even more extensive than apparent from Fig. 6A. These experiments show that MreB and Mbl colocalize at many places along the lateral cell membrane, possibly using similar tracks for the formation of dynamic filaments, or even forming mixed polymers within the filaments.
MreB, Mbl and MreBH interact underneath the cell membrane
We employed the BiFC technique to investigate if MreB, Mbl and MreBH physically interact within live cells. BiFC is a highly specific and sensitive method to detect protein–protein interactions within live cells (Hu et al., 2002). Two separate parts of YFP form a fluorescent YFP molecule only when they are attached to two proteins that strongly interact with each other, but not when they are expressed independent of interacting factors. When YN-MreB or YC-Mbl were expressed individually, cells grew indistinguishably from wild-type cells, and only background fluorescence was visible throughout the cells (Fig. 6D and E). Likewise, when the YN fragment was expressed in a strain expressing YC-Mbl, only background fluorescence was detectable (Fig. 6F). Upon induction of YN-MreB from an ectopic site on the chromosome in cells expressing YC-Mbl, clear fluorescent signals were observed at many places along the lateral cell membrane (Fig. 6G). Solely fluorescent dots were observed, but no clear filaments comparable to those seen with YFP-MreB. Cells contained between seven and 11 clear BiFC signals (on average eight signals), showing that MreB and Mbl extensively interact within B. subtilis cells. Likewise, YN-MreB or YC-MreBH did not show any signal above background fluorescence when expressed individually (data not shown), but clear BiFC signals could be monitored during simultaneous expression of both fusion proteins (Fig. 6H). BiFC signals were much weaker compared with the YN-MreB/YC-Mbl strain, consistent with our finding that GFP-MreBH signals are much weaker than those of GFP-MreB or of GFP-Mbl, in spite of the same promoter driving the fusions. Additionally, only 2–6 BiFC signals were detectable in YN-MreB/YC-MreBH cells. Individual expression of YC-MreBH and of YN-Mbl did not result in any fluorescence signal above background (data not shown), but induction of both fusion constructs yielded from four up to seven signals within 97% of the cells analysed (> 200, Fig. 6I). Due to the resolution limit of light microscopy and our stringent definition of BiFC signals (only discrete foci that were > 8% above background fluorescence were scored), the number of BiFC foci is certainly a strong under-estimation of the extent of interaction of the actin paralogues. Additionally, formation of a full YFP protein is subject to sterical constraints, and depends on the orientation of the tags. These experiments show that all three B. subtilis actin-like proteins interact with each other, with a major interaction between MreB and Mbl, a less pronounced interaction between Mbl and MreBH, and an only moderate interaction between MreB and MreBH.
Mbl confers viability in the absence of MreB in special medium
To find out which other cell function may complement the mreB deletion in PAB-SMM medium, and thus why the deletion is viable under these conditions, we transformed the deletion strain with a construct in which transcription of the mbl gene is under control of the xylose promoter. Colonies were obtained in the presence of xylose, but not in the absence of the inducer (data not shown). In the presence of xylose, strain JS68 (ΔmreB, pxyl-mbl) grew similar to strain 3725 (ΔmreB), but removal of xylose led to a cessation of growth and to cell lysis (data not shown). During depletion of Mbl, cells entirely lost their rod-like cell morphology and clear chromosome segregation defects became apparent (Fig. 7B, compare with A). In the presence of inducer, strain JS68 formed ∼ 1% anucleate cells, while two doubling times after depletion of Mbl, 5% anucleate cells were observed (150 cells analysed). The experiments show that the presence of Mbl is essential in the mreB deletion strain, although the mbl gene itself is not essential (Abhayawardhane and Stewart, 1995), and that the defect in chromosome segregation and in cell shape occurring in ΔmreB cells growing in PAB-SMM medium is strongly exacerbated during loss of Mbl function. Therefore, Mbl can compensate for the loss of MreB under special medium conditions, suggesting that the two proteins have partially overlapping functions.
Mbl, but not MreB, colocalizes with MreC
MreC localizes to the cell membrane (Defeu Soufo and Graumann, 2004), and also displays a non-uniform pattern of localization that has been interpreted as a helical arrangement (Leaver and Errington, 2005). Because MreC has been shown to interact with Pbps in C. crescentus (Divakaruni et al., 2005), and because Mbl may be required for the helical insertion of new cell wall material, while MreB appears dispensable for this process (Daniel and Errington, 2003), we investigated if MreB or Mbl colocalize with MreC. We generated CFP-MreB or CFP-Mbl fusions and combined them with a YFP-MreC fusion that is functional (Defeu Soufo and Graumann, 2004). It is apparent from Fig. 6B that CFP-Mbl foci frequently coincided with discrete signals formed by YFP-MreC (in 85% of the cells). It should be noted that in contrast to cells expressing CFP-Mbl or YFP-MreC individually, cells expressing both fusions were frequently wider (1.4 ± 0.2 μm vs. 1.1 ± 0.15 μm on average) and shorter than wild-type cells, suggesting that GFP tags on both proteins interfere with the function of the proteins. We have observed this phenomenon before with proteins that interact with each other (Volkov et al., 2003; Kidane et al., 2004). Contrarily, CFP-MreB did not reveal any considerable colocalization with YFP-MreC, the pattern of foci of MreC that are apparent within the membrane coincided with YFP-MreB signals in only 5% of all cells analysed (n > 250).
To further investigate the link between Mbl and MreC, we tested for a direct interaction between the proteins employing BiFC. When YC-MreC was expressed by itself, no fluorescence above background was detectable (Fig. 6J), likewise to the sole expression of YN-MreB (Fig. 6E) or of YN-Mbl (data not shown). Simultaneous expression of YC-MreC and of YN-Mbl resulted in the formation of two to five BiFC signals within 96% of the cells (> 250 cells analysed, Fig. 6K), whereas BiFC signals in the strain expressing YC-MreC and YN-MreB were detectable in only 1% of all cells, and in these, only a single signal was observed (Fig. 6L). We consider this as random background BiFC signal and thus as a ‘no interaction’ between MreB and MreC. These findings show that Mbl directly interacts with MreC at the cell membrane.
To support the detected protein interactions, and to determine, if the lack of BiFC signal in the YC-MreC/YN-MreB strain was due to sterical hindrance, or indeed due to a lack of physical closeness, we performed FRET analysis with YFP-MreC, CFP-Mbl and CFP-MreB. FRET images are taken by using specific excitation of CFP, and monitoring of YFP emission, whose level is determined in arbitrary units by the imaging software. These experiments were facilitated by the fact that YFP-MreB and YFP-MreC did not show any fluorescence above background in the FRET filter (average of 79 ± 2 units, wild-type cells devoid of a YFP fusion showed 79 ± 1 units), and were only visible in the YFP filter (Fig. 8A and data not shown). CFP-Mbl and CFP-MreB showed weak signals in the FRET filter, on average 91 (± 3) units or 87 (± 2) units, versus 70 (± 1) units for background noise outside of cells (Fig. 8B, and data not shown, at least 150 cells were scored for each channel). The signals in the FRET filter in the CFP-Mbl/YFP-MreB strain were on average 102 (± 4) units (Fig. 8C), and thus 12.1% higher than the signals of the CFP-Mbl strain. Figure 8C shows the signal intensity in the FRET filter of the CFP-Mbl/YFP-MreB strain after subtraction of the signal from the CFP-Mbl strain (netFRET). Concomitantly, the CFP signal of the dually labelled strain decreased by 8% compared with the single CFP strain (Fig. 8C, compare ‘CFP’ panel with Fig. 8B). As expected, the FRET signals were mostly found at the lateral side of the membrane, but rarely at the cell poles (Fig. 8C). In agreement with the colocalization studies showing that MreB and Mbl do not colocalize at all positions along the membrane, FRET signals were not apparent at all positions of YFP signals (Fig. 8C, compare ‘YFP’ with ‘netFRET’). It should be noted that the FRET interaction between MreB and Mbl appears to be slightly different from the BiFC interaction, because the signal to noise ratio is much lower in FRET images, and because FRET can also occur at positions where tagged proteins are merely close to each other but do not necessarily interact. Average signal intensity was even higher in the CFP-Mbl/YFP-MreC strain [110 (± 3) units, corresponding to 20.9% increase, Fig. 8D], and netFRET signals were observed along the lateral sides as well as at the cell poles (Fig. 8D). In contrast to this, no increase in FRET signal intensity was found in the CFP-MreB/YFP-MreC strain compared with the CFP-MreB strain (Fig. 8E), showing that MreC and Mbl, but not MreC and MreB, are in close proximity at the B. subtilis cell membrane.
Bacillus subtilis and many other bacteria from distinct branches contain two or three actin paralogues, while E. coli and other bacteria contain only a single recognizable mreB gene. All three B. subtilis actin paralogues have been implicated in both chromosome segregation [with MreB being most important, and MreBH the least important (Defeu Soufo and Graumann, 2003)] and in establishment of proper cell morphology (Jones et al., 2001), but the reasons for the gene multiplicity and for the seemingly overlapping functions have been unclear. Our work shows that contrarily to earlier data using single fluorescent labels, which suggested that MreB and Mbl form-independent helical structures (Jones et al., 2001), MreB and Mbl largely colocalize within the cell, although MreB and Mbl did not colocalize at all places along the membrane. Most importantly, we found that the three B. subtilis actin-like proteins interact with each other in vivo and affect each other's localization. We have adapted the BiFC technique to B. subtilis, which revealed that MreB, Mbl and MreBH interact at many places along the lateral cell wall, in agreement with differential GFP labelling of MreB and Mbl. Thus, MreB and Mbl frequently use similar tracks underneath the membrane, where they are in physical contact, or may even form mixed filaments. Unfortunately, purified B. subtilis MreB is only partially functional (H.J. Defeu Soufo and P.L. Graumann, unpublished data), so biochemical investigation of the interaction has not yet been possible. However, interaction could also be shown using FRET. MreB showed the strongest BiFC interaction with Mbl and to a much lesser extent with MreBH, while the interaction of MreBH and Mbl appeared to be more extensive than that of MreB and MreBH. These experiments suggest a certain hierarchy of interaction between the MreB paralogues.
The interaction of MreB and of Mbl is relevant, because the induction of a mutant version of MreB that has an aberrant pattern of localization and is not fully functional had a strong effect on the localization of Mbl (even in cells that displayed normal cell morphology, ruling out a non-specific effect via perturbation of cell shape). The mutant MreB inhibited dynamic localization of Mbl filaments, showing that a defect (most likely in polymer turnover or disassembly) in MreB affects the dynamics of Mbl filaments. Conversely, depletion of Mbl strongly interferes with the formation of GFP-MreB filaments (Defeu Soufo and Graumann, 2005), showing that both actin paralogues affect each other's mode of filament formation. The depletion of MreBH has a much weaker effect on the localization of GFP-MreB than the depletion of Mbl (Defeu Soufo and Graumann, 2005), in agreement with the apparently lower degree of interaction of MreB and MreBH as shown in this work. Therefore, if the proper function of one actin paralogue is perturbed, the effect will translate onto the localization and function of the other actin-like proteins. These findings support the idea that MreB paralogues provide structural support for the other paralogues, or even form overlapping polymers within the visible filaments. Our findings have important implications for the analysis of the function of MreB-like proteins in bacteria, and can explain why the depletion of Mbl also affects chromosome segregation, possibly through an effect onto MreB. Conversely, depletion of MreB may affect cell shape through an effect on Mbl, although MreB also plays a major role in maintenance of cell morphology itself (Formstone and Errington, 2005).
Dual labelling and interaction studies between MreB, Mbl and MreC strongly support the notion that Mbl may primarily act on the formation of proper cell shape. CFP-Mbl showed clear colocalization and interaction with YFP-MreC, while CFP-MreB filaments rarely colocalized with YFP-MreC. Rather, CFP-MreB and YFP-MreC formed distinct subcellular structures, and showed neither BiFC nor FRET interaction. These data show that Mbl, but not MreB, is in close contact with MreC, and may affect the incorporation of cell wall material via MreC. In C. crescentus, MreC has been shown to interact with Pbps (Divakaruni et al., 2005), indicating that MreC may direct the helical localization of Pbps. However, C. crescentus Pbps mislocalize in the absence of functional MreB or of MreC, although MreB and MreC form-independent helical structures underneath and within the cell membrane, suggesting that C. crescentus MreB affects the localization of Pbps via a MreC-independent path, but that both pathways are necessary for proper arrangement of cell wall synthesis.
The difference in interaction with MreC, the different phenotypes with regard to maintenance of rod shape and to chromosome segregation suggest that MreB paralogues have somewhat distinct functions in B. subtilis. However, in agreement with our findings suggesting the formation of mixed polymers, our genetic experiments suggest that MreB and Mbl confer overlapping tasks. It is possible to obtain a mreB in-frame deletion in a special medium containing high magnesium and sucrose concentrations, which most likely stabilizes the cell wall, strongly suggesting that MreB also confers a crucial function in the direction of cell wall synthesis (besides its role in chromosome segregation), as was suggested by Errington and coworkers (Formstone and Errington, 2005). Under these special circumstances, mbl is essential, because the depletion of mbl is lethal in the mreB deletion strain, while it is not lethal in wild-type strains. These experiments suggest that Mbl can partially substitute for the function of MreB under special conditions, i.e. when different ion concentrations can compensate the specific function of MreB in cell morphology. Possibly, MreB exerts its role on cell morphology through an interaction with MreD, or with RodA, two other membrane proteins that play a central role in the formation of a rod shaped cell (Iwaya et al., 1978; Defeu Soufo and Graumann, 2003).
A further major finding of our work is the proof that the dynamic localization of MreB and of Mbl filaments is important for the function of both proteins. To gain insight into the role of MreB filaments, and their dynamics, we generated a mutation in MreB within the phosphate 2 motif that forms part of the ATPase pocket in MreB. The mutation in the phosphate 2 motif strongly reduces ATPase activity in eukaryotic actin, and may have a similar effect in MreB, but it should be noted that experimental evidence for this is missing. Expression of mutant GFP-MreB in B. subtilis resulted in the formation of helical filaments, similar to E. coli (Kruse et al., 2003), but the mutant protein was only partially functional. In the absence of wild-type MreB, mutant MreB caused a severe defect in cell morphology and in chromosome segregation, and even led to similar defects when expressed in the presence of wild-type MreB. Strikingly, filaments formed by mutant GFP-MreB did not show any dynamic reorganization like wild-type GFP-MreB, but were highly rigid. Moreover, expression of mutant MreB strongly reduced the dynamics of wild-type GFP-MreB filaments, revealing that mutant MreB directly affects the localization of wild-type MreB. Likewise, induction of mutant MreB arrested the dynamics of GFP-Mbl filaments, showing a similarly strong effect on the dynamics of the other actin-like protein. Because cells expressing mutant MreB only grew very poorly, these finding show that dynamic localization of MreB filaments is essential for their proper function.
MreB orthologues are present in round cyanobacteria and in round planctomyces species, and MreB is essential for growth of round Rhodobacter sphaeroides cells (Slovak et al., 2005), in which clearly, MreB has a function different from the regulation of rod cell shape. In E. coli and in C. crescentus, MreB is involved in the formation of the proper cell shape, and in the separation of duplicated chromosomes towards opposite cell poles. Many Gram-positive bacteria appear to have opted for the distribution of aspects of the two distinct functions to different MreB paralogues, which, however, tightly interact with each other and affect each other's function, and provide important structural support to each other.
Escherichia coli XL1-Blue (Stratagene) or B. subtilis strains were grown in LB rich medium supplemented with 50 μg ml−1 ampicillin or other antibiotics [5 μg ml−1 chloramphenicol (Cm), 50 μg ml−1 spectinomycin (Spec), or 5 μg ml−1 kanamycin (Kan)], where appropriate, in S750 medium, or in Difco antibiotic medium 3 (PAB). 2 × concentrated SMM (1 M sucrose, 33.7 mM maleic acid and 40 mM MgCl2, pH 7.0) was use to dilute 2 × PAB in a 50:50 ratio for the use of PAB/SMM medium. For induction of the hyperspank promoter, the culture media were supplemented with 0.1–1 mM isopropyl-β-d-thiogalactopyranoside (IPTG). For induction of xylose promoter, glucose in S750 medium was exchanged for 0.5% fructose and xylose was added up to 0.5%.
Constructions of plasmids
All primers and plasmids are listed in the Supplementary material. YFP of CFP fusions to MreB were non-functional when eYFP or eCFP were used, apparently because of a problem of codon usage. For the generation of a functional YFP variant of MreB, the gfp mut1 gene in pSJ17 (Defeu Soufo and Graumann, 2005) was mutagenized by polymerase chain reaction (PCR) to obtain a yfp allele, by changing tyrosine 203 to threonine (T203Y), using primers 1 and 2, generating plasmid pJS24. YFP-MreC was obtained through excision of mreB from pJS24 using EcoRI and ApaI enzymes, and through insertion of the entire mreC gene, amplified from chromosomal DNA by PCR, resulting in pJS26 plasmid. For the generation of a CFP N-terminal fusion, Cerulean cfp was amplified from plasmid mCerulean-C1 (Rizzo et al., 2004) with the addition of the N-terminal 12 amino acids from the comGA gene (Veening et al., 2004), using primers 3 and 4, and was inserted into PstI and ApaI sites within pHJDS1 (Defeu Soufo and Graumann, 2004), giving pHJDS2 for cfp N-terminal fusion at the original locus, driven by the xylose promoter. To obtain CFP variants of MreB, Mbl or MreC, the 5′ part of mreB, mbl or mreC (about 500 bp) were PCR-amplified from chromosomal DNA and inserted between EcoRI and ApaI sites in pHJDS2. This established pJS29, pJS28 or pSJ30 respectively.
To generate a mutant allele of mreB, plasmid pJS17 was mutagenized by site directed PCR mutagenesis. Aspartate was exchanged to alanine at position 158 in the phosphate 2 B box (D158A), using primers 7 and 8. This way, plamid pJS34was generated.
For the expression of mreB and mutant mreB alleles without fusion gene, gfpmut1 was excised from pSG1729 (Lewis and Marston, 1999) using KpnI and ApaI sites, and full length mreB, or mreBD158 genes amplified from pJS17, or pJS34, respectively, using primers 9 and 10, and were inserted into pSG1729, yielding pJS39, or pJS41 respectively.
For mutant mreB allele without fusion driven by the hyper-spank promoter, non-fused genes were amplified using primers 11 and 12 from pSG1729 derivatives pJS39, or pJS41, and were cloned into pDR111 (kind gift from D. Rudner, Harvard Medical School) using HindIII and SphI restriction sites. The following plamids were generated: pJS42 and pJS44.
To employ the BiFC strategy, yfp was split into two fragments, the N-terminal part (YN) made of the first 154 amino acids, and the C-terminal part (YC) made of the last 86 amino acids (Hu et al., 2002). Both parts were PCR-amplified from pSG1187 (Feucht and Lewis, 2001), using primers 13 and 14, or 15 and 16, respectively, and were introduced separately into pSG1729 using KpnI and ApaI sites, or into pHJDS1 using PstI and ApaI sites (each time exchanging the gfpmut1 gene) to generate pHJDS4 (yn part, integration at amy locus), or pHJDS5 (yc part, for integration at original locus) respectively. The N-terminal part of ComGA (12 aa) was added upstream of each part when designing the forward primers. A long 15 aa linker was also kept for the reverse primers. Full length mreB and mbl genes were inserted into pHJDS4, generating plasmids pJS45 and pJS47, while the 5′ regions of mbl, mreC or mreBH were inserted into pHJDS5 between ApaI and EcoRI sites, generating plasmids pJS51, pJS52 or pJS53 respectively. For YN fusion to be induced by the hyperspank promoter at the amy locus, the fusion gene was amplified (using primers 11 and 12 for yn-mreB and, or 17 and 18 for yn-mbl, which include the ribosome binding site from pHJDS4) from pJS45 and pJS47, and were introduced in pDR111 using HindIII and SphI for yn-mreB, or SalI and NheI for yn-mbl. pJS46 and pJS48 were established this way respectively.
To generate a strain for dual visualization of MreB and Mbl, B. subtilis PY79 was transformed with plasmid pJS24, selecting for Spec resistance, and the resultant strain JS36 (yfp-mreB::amy) was transformed with chromosomal DNA from JS40 (cfp-mbl) strain, selecting for Cm and spec resistance, generating strain JS45 (yfp-mreB::amy/cfp-mbl). For the generation of a strain for dual localization of MreB and MreC, PY79 was transformed with plasmid pJS26, selecting for Spec resistance, and the resultant strain JS38 (yfp-mreC::amy) was transformed with chromosomal DNA from JS41 (cfp-mreB) strain, selecting for Cm and Spec resistance, generating strain JS48 (yfp-mreC::amy/cfp-mreB). For simultaneous visualization of Mbl and MreC, strain JS38 (yfp-mreC::amy) was transformed with chromosomal DNA from JS40 (cfp-mbl) strain, selecting for Cm and Spec resistance, generating strain JS47 (yfp-mreC::amy/cfp-mbl). To generate a strain for the visualization of mreB mutant allele, PY79 was transformed with plasmid pJS34, selecting for Spec resistance, generating strain or JS51 (gfp-mreBD158A::amy).
For the induction of mreB mutant alleles in strains expressing GFP-MreB or GFP-Mbl, competent cells from JS59 (mreBD158A::amy) were transformed with chromosomal DNA from JS12 (gfp-mreB) or JS13 (gfp-mbl), selecting for Cm and Spec resistance, generating strains JS61 (mreBD158A::amy/gfp-mreB) or JS63 (mreBD158A::amy/gfp-mbl).
To evaluate the complementation efficiency of wild-type mreB or of mutant alleles in the ΔmreB strain, strain 3725 (Formstone and Errington, 2005) was transformed with chromosomal DNA from JS57 (mreB::amy), or JS59 (mreBD158A::amy) (for non-fused genes), or from JS25 (gfp-mreB::amy), or JS56 (gfp-mreBD158A::amy) (for gfp-fused genes), selecting for Kan and Spec resistance, producing JS65 (ΔmreB/mreB::amy), and JS64 (ΔmreB/mreB1D158A::amy) or JS67 (ΔmreB/gfp-mreB::amy) and JS66 (ΔmreB/gfp-mreBD158A::amy) respectively. To investigate if an mreB-deleted strain can still growth in the absence of Mbl, strain 3725 (Formstone and Errington, 2005) was transformed with chromosomal DNA from JS2 (Pxyl-mbl) (Defeu Soufo and Graumann, 2004), selecting for Kan and Spec resistance, producing JS68 (ΔmreB/Pxyl-mbl).
For the generation of BiFC strains, strain JS69 (yn::amy) was transformed with plasmid pJS51, strain JS71 (yn-mreB::amy) was transformed with plasmids pJS51, pJS52 or pJS53, and strain JS72 (yn-mbl::amy) was transformed with plasmids pJS52 or pJS53, selecting for Cm and Spec resistance, generating strains JS70 (yn::amy/yc-mbl), JS74 (yn-mreB/yc-mbl), JS75 (yn-mreB/yc-mreBH), or JS76 (yn-mreB/yc-mreC) and JS78 (yn-mbl/yc-mreBH) or JS79 (yn-mbl/yc-mreC) respectively.
Serum raised against 6-His-GFP was used for detection after gel electrophoresis (equal numbers of cells were applied, and equal amounts of protein were verified by Bradford tests) and blotting onto nitrocellulose membranes.
For microscopic analysis, Bacillus strains were grown in S750 defined medium (Jaacks et al., 1989) complemented with 0.1% casamino acids, or in or PAB/SMM medium. Fluorescence microscopy was performed on an Olympus AX70 microscope, or on a Zeiss Axioimager, using TIRF objectives with an aperture of 1.45. Cells were mounted on agarose gel pads containing S750 growth medium on object slides. Images were acquired with a digital CCD camera; signal intensities and cell length were measured using the Metamorph 4.6 program (Universal Imaging, USA). Image deconvolution was performed on Z-stacks using Autodeblur X1.4.1 software for measurement of distances between distinct GFP-MreB signals/pitch of GFP-MreB filaments.
For FRAP studies, an argon laser was mirrored onto the specimen from above the filter plane, using the side port of the Zeiss microscope. The size of the laser beam (50 μm) was generated by a pinhole inserted into the laser beam within the module that incorporates the optical wire into the side port (custom made by A&S, Munich, Germany).
BiFC signals were defined as being more than 8% above background fluorescence, which was measured by using an area of the cell that was devoid of any discrete signal as determined by eye as background fluorescence, and each fluorescent signal was analysed for fluorescence using the Metamorph region scaling tool.
FRET studies were performed using a CFP filter cube (exciter/dichroic) lacking the emission filter, and a microimager that splits the emission light into a CFP emitter (CFP channel) and a YFP emitter (FRET) channel; a YFP exciter/dichroic cube was used to for the corresponding YFP images. To determine the actual FRET signal, average fluorescence in the FRET channel (corresponding to YFP emission) from a strain expressing the corresponding CFP fusion only (> 150 cells analysed) was subtracted from the average FRET signal from the strain expressing both corresponding CFP and YFP fusions (> 150 cells analysed); fluorescence intensity of least 150 cells was determined in the Metamorph 6.0 program to obtain the value for average fluorescence. YFP fusions did not emit any signal above background in the FRET filter, so the YFP contribution to the FRET signal was 0. Average CFP fluorescence intensity was determined from 100 to 120 cells, but the increase in YFP fluorescence was more significant than the decrease in CFP fluorescence in all cases of FRET interaction. DNA was stained with 4',6-diamidino-2-phenylindole (DAPI; final concentration 0.2 ng ml−1) and membranes were stained with FM4-64 (final concentration 1 nM).
We thank Astrid Steindorf for technical assistance, and Jeffrey Errington of Oxford University for the generous gift of mutant strains. This work was supported through the Deutsche Forschungsgemeinschaft and the University of Freiburg.