In Caulobacter crescentus, intact cables of the actin homologue, MreB, are required for the proper spatial positioning of MurG which catalyses the final step in peptidoglycan precursor synthesis. Similarly, in the periplasm, MreC controls the spatial orientation of the penicillin binding proteins and a lytic transglycosylase. We have now found that MreB cables are required for the organization of several other cytosolic murein biosynthetic enzymes such as MraY, MurB, MurC, MurE and MurF. We also show these proteins adopt a subcellular pattern of localization comparable to MurG, suggesting the existence of cytoskeletal-dependent interactions. Through extensive two-hybrid analyses, we have now generated a comprehensive interaction map of components of the bacterial morphogenetic complex. In the cytosol, this complex contains both murein biosynthetic enzymes and morphogenetic proteins, including RodA, RodZ and MreD. We show that the integral membrane protein, MreD, is essential for lateral peptidoglycan synthesis, interacts with the precursor synthesizing enzymes MurG and MraY, and additionally, determines MreB localization. Our results suggest that the interdependent localization of MreB and MreD functions to spatially organize a complex of peptidoglycan precursor synthesis proteins, which is required for propagation of a uniform cell shape and catalytically efficient peptidoglycan synthesis.
The morphology of bacterial cells is ultimately determined by a peptidoglycan cell wall comprised of a repeating disaccharide polymer of glycan strands cross-linked by peptide bridges. In Gram-negative bacteria, cell wall synthesis occurs in two distinct subcellular compartments, the cytosol and periplasm. The initial biosynthetic stage within the cytosol generates soluble cell wall precursors, UDP-N-acetylglucosamine (UDP-GlcNAc) and UDP-N-acetylmuramic acid (MurNAc)-pentapeptide, catalyzed sequentially by MurA, MurB, MurC, MurD, MurE and MurF (Barreteau et al., 2008) (Fig. 1A). In the subsequent cytoplasmic membrane-associated stage, UDP-MurNAc-pentapeptide is ligated to an undecaprenyl phosphate carrier lipid by the integral membrane protein MraY, forming what is commonly referred to as Lipid I (Fig. 1A). In the final stage of precursor synthesis, catalysed by MurG, GlcNAc is ligated to Lipid I to form Lipid II. Lipid II is then translocated by an undefined mechanism across the inner membrane and into the periplasm (Suzuki et al., 1980; van Heijenoort, 2007; Bouhss et al., 2008). There, a class of membrane-anchored enzymes possessing transglycoslyase and/or transpeptidase activities, collectively known as penicillin-binding proteins (PBPs), incorporate peptidoglycan precursors into pre-existing cell wall material.
Recent experiments have made significant progress towards illuminating the temporal and spatial regulation underlying bacterial cell morphogenesis. Collectively, these studies revealed mutations resulting in cell shape defects that have generally fallen into two classes. The first class comprises genes encoding enzymes that directly produce peptidoglycan precursors or catalyse its assembly. For example, mutations in genes encoding PBPs convert rod-shaped Escherichia coli into round or branched cells (Nilsen et al., 2004). The second class of cell shape mutations produce a similar rounded cell phenotype, but occur in genes encoding proteins without apparent peptidoglycan synthesizing activity (Wachi et al., 1989, Wachi et al., 1987). The best understood of these, mreB, encodes a prokaryotic homologue of filamentous actin (van den Ent et al., 2001; Carballido-López et al., 2006;). Subcellular localization experiments have shown that B. subtilis MreB and its homologues form helical structures wrapping around the circumference of the cell (Jones et al., 2001; Carballido-López and Errington, 2003). Similar patterns of MreB localization have been observed in diverse organisms including Caulobacter crescentus (Figge et al., 2004; Gitai et al., 2004), E. coli (Shih et al. 2003) and Rhodobacter sphaeroides (Slovak et al., 2006). Notably in Caulobacter crescentus, these helical structures dynamically re-locate to the midcell just prior to cytokinesis (Figge et al., 2004; Gitai et al., 2004). Cells depleted of MreB, in general, eventually transform into round or lemon-shaped cells with apparent cell wall defects, suggesting that MreB plays a critical role in co-ordinating cell wall biosynthesis. In support of this idea, labelling of newly incorporated cell wall precursors using a fluorescent derivation of vancomycin (Van-FL) revealed a helical pattern of labelling that was dependent on Mbl, a homologue of MreB in B. subtilis (Daniel and Errington, 2003). Additionally, in Caulobacter, PBP2 and the murein precursor synthesis enzyme, MurG, exhibit subcellular localization patterns that are strikingly similar to that adopted by MreB (Divakaruni et al., 2007). MurG localization is dependent on intact helical cables of MreB as treatment with A22, an inhibitor of MreB polymerization, disrupts the positioning of MurG (Divakaruni et al., 2007).
The PBP-catalysed insertion of precursors into pre-existing cell wall is thought to occur at sites within the peptidoglycan that have been hydrolysed by lytic transglycosylases (Holtje, 1998; Vollmer and Holtje, 2001). Biochemical experiments have indicated that PBPs, lytic transglycosylases and endopeptidases exist in complex as a peptidoglycan synthesizing holoenzyme in order to properly co-ordinate opposing lytic and biosynthetic activities (Romeis and Holtje, 1994; von Rechenberg et al., 1996; Vollmer et al., 1999). MreC, a cell shape-determining protein with a large periplasmic domain, localizes in a helical pattern, and is hypothesized to have an important role in the spatial positioning of this peptidoglycan synthesizing complex (Divakaruni et al., 2005; 2007). Affinity chromatography experiments have shown that MreC interacts with several PBPs in Caulobacter cell extracts (Divakaruni et al., 2005), and bacterial two-hybrid experiments with B. subtilis proteins demonstrated that MreC could interact with a number of high molecular PBPs (van den Ent et al., 2006). Additionally, in Caulobacter, PBP2, the outer membrane-anchored lytic transglycosylase, MltA, and its interacting partner, MipA, localize in a helical-like or banded pattern that is dependent on MreC (Divakaruni et al., 2007). This localization pattern is likely to be required for efficient cell wall synthesis as depletion of MreC almost completely eliminates peptidoglycan synthesis when assayed using Van-FL labelling (Divakaruni et al., 2007).
These experiments suggest that MreB and MreC may function to spatially organize cell wall assembly by positioning enzymes involved in peptidoglycan biosynthesis on either side of the cytoplasmic membrane. The composition of these cytosolic and periplasmic complexes involved in cell morphogenesis remains to be defined. Do the cell shape determining proteins interact with, and/or position a large complex of cell wall biosynthetic enzymes? Are there connections between multi-protein complexes present in the cytoplasm and periplasm? Are other morphogenetic proteins present in these complexes? In this manuscript we address these questions.
We show that MreB plays a crucial role in spatially organizing several cytosolic peptidoglycan precursor biosynthetic enzymes. We also demonstrate that MreD plays an important role in maintenance of rod shape that the spatial localization of MreD and MreB cables is interdependent and that these two proteins may position peptidoglycan synthesizing enzymes through direct protein–protein interactions. We performed an extensive two-hybrid analysis assaying for interactions between proteins located in the cytosol, periplasm and the inner membrane. Novel interactions were discovered between peptidoglycan synthesizing enzymes and the key morphogenetic proteins, MreB, MreC and MreD. We hypothesize that MreD, an integral membrane protein, serves to connect cytosolic peptidoglycan precursor biosynthesis to periplasmic-localized cell wall assembly. We propose that large protein interaction networks function to spatially organize and activate compartmentalized peptidoglycan biosynthetic activities from the cytosol, across the inner membrane, to the periplasm.
Cytosolic peptidoglycan precursor synthesis enzymes are positioned by morphogenetic proteins
Previous experiments have shown that the enzyme catalysing the final step in peptidoglycan precursor synthesis, MurG, when fused to the fluorescent protein, mCherry, localizes in a banded pattern along the length of the cell that is dependent on MreB (Divakaruni et al., 2007). We tested whether other enzymes in this biosynthetic pathway (Fig. 1A) exhibited a similar subcellular pattern of localization. This was accomplished by fusing mCherry to the carboxyl-terminus of the cytosolic proteins in the murein precursor synthesis pathway MurB, MurC, MurD, MurE and MurF, as well as to the integral membrane protein, MraY (Fig. 1A and Fig. S1). Each fusion was integrated by single cross-over, homologous recombination at its native locus. All of these fusions exhibited a remarkably similar subcellular localization pattern, appearing as a series of repeating bands or extended foci perpendicular to the long axis of the cell (Fig. 1B–F). Treatment with A22, an inhibitor of MreB polymerization, for as little as 60 min resulted in the redistribution of each fusion protein to the midcell or poles (Fig. 1B–F). Normal localization throughout the cell length was restored by washing the cells and suspending in fresh media (not shown). These fusions also exhibited this pattern of localization, before and after treatment with A22, when DAPI was omitted (for example see Fig. S2). Additionally, immunoblot analysis demonstrated the majority of the fusion protein in the cell, remained intact upon A22 treatment (Fig. S3). Cells harbouring these fusions also possess a wild-type copy of the encoding gene and exhibited normal morphology and growth rate (not shown). Interestingly, however, when cells expressing a MurD–mCherry fusion were grown in liquid culture, growth was severely impaired and the majority of the cells were puffy or spherical in shape, indicating that this fusion protein interfered with the spatial co-ordination of peptidoglycan synthesis (Fig. 1G). In these rounded cells, the MurD–mCherry fusion protein did exhibit a banded localization pattern similar to that observed with the other peptidoglycan precursor synthesis enzymes (Fig. 1G). We also examined the localization pattern of, MurA, the first enzyme in the murein precursor synthesis pathway and found that it did not adopt the characteristic pattern of localization exhibited by the other enzymes in the pathway (Fig. S1). Next, we determined whether these fusions were functional by constructing strains in which the sole copy of the murein synthetic enzyme was fused to mCherry. In all cases the fusions were expressed and exhibited a pattern of localization that was similar to that observed when a wild-type copy of the gene was present (Fig. S1). In summary, these experiments demonstrate that MurB, MurC, MurD, MurE, MurF, MurG and MraY, six enzymes in the peptidoglycan precursor synthesis pathway, have a similar localization pattern that is dependent on intact MreB cables.
These MreB-dependent localization patterns were similar to that observed previously with a MurG–mCherry fusion (Divakaruni et al., 2007). In order to assay localization of MurG using an independent method, we raised antiserum to purified MurG, affinity purified the anti-MurG antibodies and used them in immunofluorescence microscopy experiments with wild-type cells. The localization of native MurG observed in this experiment (Fig. 1H) was almost identical to that observed previously using living cells expressing a MurG–mCherry fusion. Also consistent with previous results, treatment of the cells with A22 caused MurG localization to collapse to distinct foci either at the pole, midcell or both locations.
Two-hybrid assay of interactions between morphogenetic proteins and peptidoglycan biosynthetic enzymes
The similar localization patterns and the distinct re-positioning of these peptidoglycan biosynthetic enzymes suggest that they may form a complex with each other and/or cytoskeletal proteins. Therefore, we employed a bacterial two-hybrid system in order to assay interactions between cell-shape determining proteins and peptidoglycan-synthesizing enzymes. With this E. coli-based two-hybrid assay, the T25 and the T18 fragments of the catalytic domain of Bordetella pertussis adenylate cyclase (cya) are fused to full-length copies of each protein. If two proteins of interest interact, a functional complementation will occur between T25 and T18 subunits, producing cAMP, and triggering the transcriptional activation of catabolic operons (e.g. lactose and maltose). A potential advantage of this system is that it does not require fusion proteins to be associated with DNA for an interaction to be detected and thus is well suited for assessing interactions of membrane proteins (Karimova et al., 2005).
We chose three representative murein biosynthetic enzymes to assay interactions with cell shape proteins: MurF, a predicted cytosolic protein, MurG, a membrane-associated cytosolic protein (Bupp and van Heijenoort, 1993), and the integral membrane protein, MraY. In all cases, fusions were constructed using both the T25 and T18 domains of adenylate cyclase at the amino- or carboxyl-terminus of each protein. These fusions were tested for interactions with MreB, MreC, MreD, RodZ and RodA, as well as the cytokinetic, tubulin homologue, FtsZ. Apparent protein interactions, as indicated by significant β-galactosidase activity (≥ 200 units), were found between a number of cell shape proteins and the peptidoglycan biosynthetic enzymes (Fig. 2A and B). In this system, we found that MurG interacted with MreB, RodA and MreD, the latter generating relatively high β-galactosidase activity (MreD-T25/MurG-T18, 1243 ± 354 units), indicative of a particularly strong interaction. It should be noted that these interactions are only detected when the adenylate cyclase portion of the fusion is cytosolic. For example, membrane topology prediction of MreD indicates that its carboxyl-terminus resides in the cytosol, rendering only carboxyl-terminal fusions suitable for this assay. Analysis of MraY interactions yielded results that were similar to those observed with MurG. MraY generated relatively strong positive two-hybrid activity with MreB and MreD, and relatively weaker, but significant activity with RodA (Fig. 2B). In contrast to MurG and MraY, MurF did not exhibit strongly positive interactions with any of the cell shape proteins we tested. However, MurF fusions did generate significant positive assay results with MraY (Fig. 2B). MraY also apparently interacted with MurG, suggesting that these peptidoglycan biosynthetic enzymes may form a complex that also interacts with MreB, MreD and RodA (Fig. 3).
We also tested for interactions between the cell shape proteins using this two-hybrid assay. We found significant interaction between MreB and MreD fusions, as well as between MreB and RodA fusions. The recently described morphogenetic protein (Shiomi et al., 2008; Alyahya et al., 2009; Bendezu et al., 2009), RodZ, also exhibited apparent strong interaction with both MreD and RodA (Fig. 2B). However, no significant interaction between RodA and MreD, or MreB and RodZ, was evident using this assay. Lastly, FtsZ and MreB fusions generated weak but significant β-galactosidase activity (335 ± 34 and 270 ± 41 units), suggesting that these two cytoskeletal proteins may also interact with each other. These two-hybrid assay results suggest that the bacterial cell shape proteins likely form an integrated structural complex, which directly interacts with murein biosynthesis enzymes, an idea consistent with their dependence on MreB for proper localization.
In previous experiments, we found that MreC associated with PBPs in detergent-solubilized cell extracts (Divakaruni et al., 2005) and that MreC was required to position PBP2, the lytic transglycosylase, MltA and its interacting protein, MipA (Divakaruni et al., 2007). Here we wanted to determine whether the bacterial two-hybrid system could be used to assay interactions between the periplasmically localized domains of these proteins. Accordingly, we fused either the T25 or T18 domains of adenylate cyclase to the periplasmic domains of MreC, PBP2, PBP3, MltA, MipA and Rod (see Fig. 2C for sequence co-ordinates included in these fusions). Since this system only detects interactions occurring in the cytoplasm or on the cytoplasmic side of the membrane, these fusions lacked either the membrane spanning or periplasmic targeting sequences of these proteins. Notably, we found significant positive interactions of PBP3, PBP2, MipA and MltA with MreC. We could also detect apparent interactions between MipA and both PBP3 and MltA (Fig. 2C). However, we could not detect a positive interaction between the RodZ predicted periplasmic domain and any of the morphogenetic proteins. Taken together, these results suggest that MreC may co-ordinate the spatial organization of these proteins through direct protein–protein interactions. As a negative control we used the periplasmic domain of MreC in interaction assays with the cytosolic morphogenetic proteins as well as MreB in experiments with periplasmically localized proteins. In both cases, as expected, we were unable to detect any positive interactions.
Our previous biochemical fractionation experiments have indicated that MreC from C. crescentus is a periplasmic protein (Divakaruni et al., 2005). In contrast, experiments indicate that MreC in other organisms likely possesses a cytoplasmic membrane spanning domain. For example, experiments with B. subtilis (van den Ent et al., 2006) and E. coli (Kruse et al., 2005) have demonstrated that full-length MreC can interact with, among other proteins, MreD and several PBPs, through it membrane spanning sequences. Given the limitations of biochemical fractionation experiments, we wanted to test whether full-length MreC from Caulobacter could interact with other cell shape proteins. Accordingly, we created adenylate cyclase fusions the amino terminus of MreC (i.e. T18- and T24-MreC1-372) and employed these in the two-hybrid assay. In contrast to previous observations with MreC from other organisms, we were unable to demonstrate any significant, positive interactions between full-length T18- and T24-MreC1-372 and MreD, PBP2, PBP3, or MreB (data not shown), a result consistent with the idea that Caulobacter MreC may be a periplasmic protein.
In summary, the results of this comprehensive set of bacterial two-hybrid assays suggest that the cell shape proteins, MreB, MreD and RodA form a large network of interactions with each other and enzymes involved in peptidoglycan precursor synthesis (Fig. 3). Additionally, these experiments show that MreC interacts with the periplasmic domains of a number of proteins involved in cell wall assembly. By applying this to results from previous biochemical, cell biology and genetic studies in Caulobacter, B. subtilis and E. coli, we constructed an interaction map between the Mre proteins and enzymes in the peptidoglycan precursor synthesis pathway (Fig. 3). These studies suggest that MreB and the integral membrane protein, MreD, play critical roles in positioning both cytosolic and membrane proteins. Likewise, we propose that MreC functions as an important spatial organizer of proteins with large domains residing in the periplasmic space. These two compartmentalized complexes are likely to be connected to each other by interactions within the cytoplasmic membrane. In support of this idea, MreC from both B. subtilis (van den Ent et al., 2006) and E. coli (Kruse et al., 2005) has been shown to interact with MreD in a bacterial two-hybrid assay.
Role of MreD in directing cell morphogenesis
The bacterial two-hybrid experiments presented here indicate that MreD may function in organizing complexes consisting of other cell shape proteins and peptidoglycan precursor synthesis enzymes. We generated a conditional mutant of mreD in C. crescentus by creating an in-frame deletion of mreD in strains containing an epitope-tagged, but otherwise wild-type copy of mreD (mreD-M2) under control of either a xylose- or vanillate-inducible promoter. Both mutant strains could not grow on plates in the absence of inducer, suggesting that mreD is an essential gene in Caulobacter. When grown in liquid medium in the presence of inducer, both strains exhibited a wild-type cell morphology (Fig. 4). In order to examine the effect of MreD depletion on cell morphology, mid-logarithmic phase cells grown in medium with either xylose or vanillate were harvested, washed and suspended in medium lacking inducer. We monitored steady-state protein levels of MreD-M2 over time during this depletion experiment by immunoblot analysis with anti-M2 antibody (i.e. anti-FLAG) (Fig. S5). We observed the accumulation of significantly rounded cells in culture following approximately 8 h of depletion (Fig. 4). After 10 h incubation in the absence of inducer, the majority of the cells (86%; n = 200) were almost completely spherical (Fig. 4). Interestingly, this morphological phenotype, although similar to that observed in E. coli strains depleted for MreD (Kruse et al., 2005), is different than the lemon-shape observed in Caulobacter cells depleted of either MreB or MreC (Figge et al., 2004; Divakaruni et al., 2007). One possibility is that MreD-depleted cells have undergone a significant loss of cell wall integrity, since, upon extended depletion, the majority of the cells eventually lysed (data not shown).
Subcellular localization interdependence of MreB and MreD
The results of the bacterial two-hybrid assay indicated that MreD and MreB may directly interact with each other. We wanted to test whether depletion of MreD affected the localization pattern of MreB and vice versa. First, in order to determine whether MreB localization was influenced by MreD, cells were depleted of MreD as described above, and after various time intervals following removal of inducer, they were fixed and processed for immunofluorescence microscopy using affinity-purified anti-MreB antibodies (Fig. 5). Following 6 h incubation in medium lacking inducer, before gross morphological defects were observed, the helical MreB localization seen in wild-type cells switched to concentrated foci at the midcell (63%) and/or the poles (35% poles and midcell) (Fig. 5 and Fig. S7). Following 8 h of depletion, the majority of the cells exhibited midcell (74%) and a significant fraction (20%) had weak or barely foci (Fig. 5). These foci remained intact in rounded cells, depleted of MreD for extended time periods (Fig. 5 and Fig. S7). Steady-state protein levels of MreB remained relatively constant during the course of MreD depletion (Fig. S5), suggesting that the observed alteration in MreB localization was the result of decreased levels of MreD. In contrast, MreC localization remained mostly unchanged upon depletion of MreD (Fig. S7), indicating that MreC may be positioned by an independent process.
Previous experiments with Caulobacter have demonstrated that MreD adopted a banded-like pattern of localization similar to that observed for other cell shape-determining proteins (Fig. S6) (Divakaruni et al., 2007). Therefore, we wanted to also determine whether MreB cables influenced the positioning of MreD. In order to accomplish this, we assayed the localization pattern of carboxyl-terminal mCherry fusion to MreD that was expressed from nitrate-inducible promoter on a low-copy plasmid, in Caulobacter cells treated with A22 (Fig. 6A). Following exposure to A22 for 90 min, MreD–mCherry localization shifted from a predominantly banded pattern to strong localization at the poles and/or midcell in the majority of cells (83%; n = 200). These results suggest that the banded pattern of MreD–mCherry localization was dependent of the presence of intact MreB cables. The re-positioning of MreD localization to poles and midcell upon A22 treatment was similar to that previously observed for the murein biosynthetic enzyme, MurG. These observations, as well as a positive interaction in the bacterial two-hybrid assay, indicate MreD and MurG might be in a complex that is spatially positioned by the same mechanism. In order to test this idea, we assayed for the colocalization of MreD–mCherry and MurG–GFP fusions in living cells. When cells harbouring both fusions were grown to mid-logarithmic phase, both MurG–GFP and MreD–mCherry exhibited a banded pattern throughout the length of the cell, which when overlaid upon each other showed significant regions of apparent colocalization (Fig. 6B). We then tested the effect of A22 treatment on colocalization of MreD and MurG fusions. A22 treatment resulted in the collapse of bands of MreD–mCherry and MurG–GFP into discrete foci (Fig. 6C). Remarkably, midcell and/or polar foci of these fusions were found to colocalize at the same location in almost all of cells, suggesting that in the absence of MreB cables, the redistribution of MreD and MurG is driven by the same localization mechanism.
MreD is required for the spatial organization of peptidoglycan precursor synthesis enzymes
Next we tested whether MreD was required for the localization of peptidoglycan precursor synthetic proteins since MurG and MraY showed positive interactions with MreD in the bacterial two-hybrid assay (Fig. 2B). Cells containing MreD-M2 under control of a xylose-inducible promoter where grown to mid-logarithmic phase in the presence of xylose, washed and then suspended in fresh medium lacking inducer. At various time points thereafter the cells were harvested, fixed and assayed for MurG localization by immunofluorescence microscopy using anti-MurG antibodies (Fig. 7A). Following 4 h of incubation in the absence of inducer, MurG exhibited a banded pattern throughout the cell (100%; n = 150) that was indistinguishable from cells grown in the presence of xylose. Additionally, in some cells, focal points at the midcell and poles were also observed (26%; n = 150). Following 6 h of MreD depletion, the majority of MurG was found concentrated near the middle of the cells and/or the poles (82%; n = 150). This pattern of localization persisted upon continued depletion (Fig. 7A). Steady-state protein levels of MurG during this depletion experiment, as indicated by immunoblot analysis, remained relatively constant throughout the time-course (Fig. S5). Similar effects of MreD depletion were also observed using a MurG–mCherry fusion, which redistributed from a banded to midcell/polar localization pattern after 6 h of MreD depletion (Fig. 7C). Likewise, depletion of MreD shifted the MraY–mCherry fusion localization pattern from bands distributed along the entire length of the cell, to discrete foci at the poles or midcell (Fig. 7B). These experimental results suggest that MreD has a critical role in maintaining the spatial localization pattern of both MraY and MurG.
Since MreD spans the inner membrane, we next wanted to test whether MreD was involved in organizing the localization of morphogenetic proteins that function in the periplasmic space. We examined the effect of MreD depletion on the localization of two different proteins: PBP2, which possesses an inner membrane-spanning domain connected to a catalytic periplasmic domain, and the periplasmic MipA. Both of these proteins adopt a banded pattern of localization that wraps around the circumference of the cell along its entire length. The localization pattern of PBP2, as assayed by immunofluorescence microscopy, remained intact even after 10 h of MreD depletion (Fig. 7E). Likewise, MreD depletion had little effect on the pattern of MipA–mCherry localization (Fig. 7F). These results suggest that MreD probably does not direct the subcellular localization pattern of morphogenetic proteins that function in the periplasm, but rather functions in the positioning of cytoplasmic cell wall synthesis proteins.
MreD is required for spatially organizing nascent peptidoglycan synthesis
These experiments suggest that MreD is required for the positioning of complexes of cell wall synthetic enzymes at the cytoplasmic membrane. One possibility is that the MreB/MreD mediated spatial distribution of these precursor synthesis proteins may be required for spatially co-ordinated cell wall synthesis. In order to test this idea, we labelled cells with a fluorescent derivative of the antibiotic ramoplanin (Ramo-FL) (Tiyanont et al., 2006) to visualize sites of cell wall synthetic activity. Ramoplanin binds to the reducing end of the glycan strand present only at sites of nascent peptidoglycan synthesis and on lipid II (Walker et al., 2005). Additionally, Ramo-FL has been shown to bind to regions of B. subtilis cells engaging in cell wall synthesis (Tiyanont et al., 2006). We found that Ramo-FL efficiently labelled living, unfixed Caulobacter cells. Cells were grown to mid-logarithmic phase, labelled with Ramo-FL (0.2 µg ml−1) for 5 min, and then immediately visualized by fluorescence microscopy (within less than 10 min). Prolonged exposure of the cells to Ramo-FL for longer than 30 min resulted in cell lysis, indicating that Caulobacter is sensitive to this antibiotic (not shown). The majority of the cells exhibited labelling along their long axis in a helical and/or banded pattern (100%; n = 150), as well as, in some cases, at the cell poles (26%; n = 150) (Fig. 8A). An intense staining at the midcell of predivisional cells was also observed. To test the effect of MreD on nascent peptidoglycan synthesis, Ramo-FL labelling was performed on cells depleted of MreD for various periods (Fig. 8B). Upon initial incubation of the MreD depletion strain with Ramo-FL, the labelling pattern was comparable to wild-type cells with both helical and mid-cell/polar labelling. However, beginning at 4 h of MreD depletion, and more prominently at 6 and 8 h time points, the cells displayed discrete foci of Ramo-FL labelling that was located at either the midcell or poles with a less intense or diffuse staining over the rest of the cell (Fig. 8B). Following 10 h, the cells had become rounded and the distinctive positioning of Ramo-FL labelling was completely abolished, evident as weak, diffuse fluorescence present throughout the cytoplasm. It was possible that the weak labelling observed upon extended depletion of MreD was the result of a significant fraction of dead or metabolically inactive cells present in the population. In order to test this, we monitored the incorporation of [35S]-methionine into new protein over the course of the depletion experiment and found no significant difference in the rate of proteins synthesis following the removal of inducer (Fig. S9), indicating that that the diffuse Ramo-FL labelling upon MreD depletion represents spatially disorganized de novo peptidoglycan synthesis.
In previous experiments, we labelled Caulobacter cells with a fluorescent derivative of vancomycin (Van-FL) in order to visualize nascent cell wall synthesis (Divakaruni et al., 2007). In contrast to Ramo-FL, Van-FL binds to cell wall precursors that have been newly incorporated into pre-existing cell wall. We determined whether depletion of MreD would also have an effect on Van-FL labelling. In wild-type cells, as well as those at the start of the depletion experiment, Van-FL labelling was similar to that observed for Ramo-FL, with cells exhibiting a relatively intense banded pattern of Van-FL binding (Fig. S8). Over the course of the depletion experiment, this banded pattern was replaced by a more diffuse staining and, as was the case with Ramo-FL labelling, accompanied by the gradual appearance of discrete foci. These labelling experiments suggest that MreD functions in spatially organizing MreB and/or peptidoglycan biosynthetic complexes in the cytosol.
We next wanted to determine whether the concentrated foci of Ramo-FL that appeared upon A22 treatment was colocalized with murein precursor enzymes. In order to test this, we examined the colocalization patterns of MurG– and MurB–GFP fusions with that of Ramo-FL labelling (Fig. 9). In untreated cells, the localization of Ramo-FL labelling with that of both MurG- and MurB–GFP, for the most part, overlapped with each other throughout the length of the cell (Fig. 9). Colocalization of these proteins with precursor synthesis was even more evident in A22-treated cells. The disruption of MreB cables with A22 resulted in the diffuse localization of MurG- and MurB–GFP throughout the cytoplasm with a region of one or two concentrated foci of fluorescent fusion protein localization. The location of intense Ramo-FL labelled foci in these A22-treated cells colocalized at high frequency with MurG- (88%) and MurB–GFP (73%) foci (Fig. 9). This observation suggests that regions containing localized murein synthetic enzymes, in A22-treated cells, are actively engaged in peptidoglycan synthesis.
In order for bacteria to reproduce with nearly identical physical dimensions, they have evolved mechanisms to control temporal and spatial peptidoglycan synthesis. The experiments described here indicate that a large morphogenetic complex consisting of proteins performing peptidoglycan precursor synthesis in the cytosol is positioned in the cell by the morphogenetic proteins, MreB and MreD. Additionally, these experiments indicate a localization interdependence of MreD and MreB, suggesting that interaction between these two proteins, and complex formation with cell wall synthesis enzymes, is required for spatial positioning within the cell.
The bacterial actin homologue, MreB, plays an essential role in cell-shape maintenance in rod-shaped bacteria by guiding lateral cell wall growth. Thus, in the absence of MreB, cells transform into a rounded, lemon-shaped morphology. MreB, which adopts a helical-like localization pattern, is thought to guide new cell wall synthesis by spatially positioning proteins required for peptidoglycan biosynthesis (Daniel and Errington, 2003; Carballido-López et al., 2006; Divakaruni et al., 2007; Kawai et al., 2009a,b). For example, recent experiments have shown that MreB homologues of B. subtilis physically interact with several PBPs (Kawai et al., 2009a,b). Likewise, in Caulobacter, MreB has been shown to be required for the subcellular localization of the murein precursor synthesis protein, MurG (Divakaruni et al., 2007). This inner membrane-associated protein catalyses the final cytoplasmic step in synthesis of the lipid-bound disaccharide-pentapeptide peptidoglycan precursor (See Fig. 1). Since it was possible that MurG may be in complex with other proteins in this biosynthetic pathway, we determined whether MurB, MurC, MurE, MurF and MraY also adopted a localization pattern that was dependent on MreB. Not only did these proteins localize in a pattern that was similar to that of MurG, but strikingly, disruption of MreB cables delocalized MurB, MurC, MurE, MurF and MraY to identical polar and/midcell regions of the cell. These observations suggest that MurB, MurC, MurE, MurF, MurG and MraY form a complex within the cell, an idea partially supported by the bacterial two-hybrid assay that showed an apparent interaction between MurF and MraY, and between MraY and MurG. In this pathway, MurF catalyses the step immediately before MraY, and the MraY-catalysed reaction directly precedes MurG. We hypothesize that these interactions may result in metabolic channeling where tripartite complex formation permits the product of the MurF reaction to be efficiently utilized as a substrate in the MraY catalysed reaction, and so forth, thus restricting the diffusion of pathway intermediates. Eliminating bulk diffusion through organized peptidoglycan biosynthesis may be advantageous with regard to catalytic efficiency since the substrate of the subsequent reaction in the pathway (i.e. the cell wall) is relatively fixed, and does not diffuse. Interestingly, expression of a MurD–mCherry fusion resulted in swollen or rounded cells perhaps indicating that disruption of these complexes may be sufficient to alter cell morphology.
We also performed a comprehensive examination of protein interactions among cell morphogenetic proteins using the bacterial two-hybrid assay. We found positive interactions with MreB and biosynthetic enzymes MurG and MraY, a finding consistent with previous co-immunoprecipitation experiments with E. coli that indicated an interaction between these three proteins (Mohammadi et al., 2007). MreD and MreB were also found to interact with these two proteins, suggesting that MreD and MreB play a central role in organizing the spatial positioning of cytosolic murein precursor synthesis (Fig. 9). RodA and RodZ are also probably found in this morphogenetic complex as they apparently interacted with each other. Consistent with these findings are previous experiments implicating RodA in lateral cell growth and peptidoglycan synthesis (de Pedro et al., 2001). Similarly, RodZ has been shown to be required for maintenance of a rod shape both in E. coli and in Caulobacter (Shiomi et al., 2008; Alyahya et al., 2009; Bendezu et al., 2009), and two-hybrid experiments with E. coli proteins also indicated that RodZ and MreB interact.
Additionally, we were interested in determining whether the cytokinetic tubulin homologue, FtsZ, could interact with any of the cell shape proteins. In Caulobacter, MreB localization is dynamic, with lateral spirals reorganizing at the midcell in a cell cycle- and FtsZ-dependent fashion (Vollmer et al., 1999; Figge et al., 2004; Gitai et al., 2004). We found that FtsZ apparently interacted with MreB in the bacterial two-hybrid assay. With the exception of RodA, FtsZ did not exhibit positive interaction with other cell shape proteins. These results suggest that cell shape proteins may specifically direct lateral cell wall expansion during growth, and do not function with the proteins comprising the divisome. It will be interesting to define the function of MreB at the midcell, and whether it, as well as RodA, interacts with other proteins involved in cytokinesis.
We also tested for interactions among proteins possessing functional domains mainly located in the periplasmic space, including MreC, PBP2, PBP3, MltA and MipA. Fusions of periplasmic domains of these proteins to either of the two halves of adenylate cyclase were expressed in the cytoplasm in the bacterial two-hybrid assay. We found that MreC interacted with both truncated versions of PBP2 and PBP3 lacking their membrane-spanning helices. This observation indicates that PBP2 and PBP3 can interact with MreC through their catalytic periplasmic domains. While assaying interactions between periplasmic proteins in the very different environment of the cytosol could result in assigning false positive interactions, our findings are consistent with previous experiments with Caulobacter, showing that PBPs present in detergent-solubilized cell extracts could be affinity purified using MreC immobilized to sepharose beads (Divakaruni et al., 2005). Interestingly, two-hybrid experiments with B. subtilis proteins also showed a positive interaction of several PBPs with MreC (van den Ent et al., 2006).
Previous affinity chromatography experiments with E. coli extracts demonstrated that the outer membrane-anchored lytic transglycosylase, MltA, interacted with PBP1B and a periplasmic protein, MipA (Vollmer et al., 1999). Complexes between PBPs and murein hydrolases are thought to be required to co-ordinate the degradation of pre-existing cell wall that is required for the insertion of peptidoglycan precursor subunits (Holtje, 1996). In Caulobacter, MltA and MipA localize in a characteristic banded pattern that is dependent on MreC (Divakaruni et al., 2007). Consistent with these previous studies, we found that Caulobacter MltA and MipA interact with each other, as well as with MreC, in the bacterial two-hybrid assay. Thus, MreC may function to organize a peptidoglycan synthesizing holoenzyme complex consisting of PBPs and lytic transglycosylases in the periplasm, while MreB probably functions to spatially organize peptidoglycan precursor synthesis in the cytosol (Fig. 10).
Are these two complexes somehow connected to each other such that the localization pattern of MreB ultimately directs the positioning of the PBP holoenzyme? Or, are these cytosolic and periplasmic complexes independently, spatially organized? Experiments with Caulobacter have shown that MreC and PBP2 maintain their distinct patterns of localization upon relatively brief treatment with the MreB polymerization inhibitor, A22 (Divakaruni et al., 2005; Dye et al., 2005), supporting the idea that there exist two independent spirals of MreB and MreC. Though, it is conceivable that MreC and PBP2 may remain localized under these conditions due to their likely associations with the cell wall. We envision that the integral membrane MreD protein, directly or indirectly, may serve as a bridge connecting cytosolic precursor synthesis with periplasmic peptidoglycan assembly. Consistent with this idea, two-hybrid interaction experiments with B. subtilis proteins showed a strong apparent interaction between MreD and MreC (van den Ent et al., 2006). However, we did not find any significant interactions between these proteins. Thus, this model of morphogenetic protein interactions across the cytoplasmic membrane is likely oversimplified. For example, recent experiments with B. subtilis have indicated the MreB also interacts with a number of PBPs (Kawai et al., 2009a,b).
A novel finding from the experiments presented here is the observation that MreD was required for MreB to adopt its characteristic helical localization pattern. MreB cables were lost during MreD depletion before there was a transformation into rounded cells, indicating that the loss of localization was attributable to an absence of MreD, and not an artifact associated with an alteration in cell shape. Conversely, helical localization of MreD was dependent on MreB, suggesting that both proteins likely form a complex that adopts a helical localization pattern. A critical function of MreB and MreD may be to distribute cell wall assembly complexes along the length of the cell, thus promoting lateral cell wall growth. Accordingly, we tested the effect of MreD depletion on cell wall synthesis using fluorescent derivatives of the antibiotics ramoplanin (Ramo-FL) and vancomycin (Van-FL), both of which bind to peptidoglycan precursors. In previous studies, Ramo-FL and Van-FL labelling of B. subtilis cells yielded similar results, indicating that the incorporation of newly synthesized precursors into pre-existing cell occurred in a helical fashion (Daniel and Errington, 2003; Tiyanont et al., 2006). Here we demonstrated that Ramo-FL can be used to track new cell wall synthesis in Caulobacter. Ramoplanin likely inhibits cell wall synthesis in Caulobacter since prolonged exposure resulted in the cells becoming rounded and eventually lysing. Labelling with Ramo-FL revealed a banded or helical pattern along the length of the cell similar to that obtained when the cells were labelled with Van-FL. Importantly, depletion of MreD, using both labelling methods, shifted the pattern from lateral to discrete, relatively intense labelled foci near poles and midcell. Depletion of MreD caused a remarkably similar shift in the localization patterns of the murein precursor synthesis proteins, suggesting that these foci of peptidoglycan synthesis may be attributable to a concentration of precursor enzymes at these same locations. In contrast, MreC depletion results in the disappearance of complexes of PBP2 and MltA and a cessation of new cell wall synthesis (Divakaruni et al., 2007).
Thus, in the presence of MreB cables and MreD, a lateral distribution of cell wall synthetic enzymes is imposed, resulting in co-ordinated lateral cell wall synthesis. We propose that the MreD/MreB positioning of precursor enzymes ‘activates’ cell wall assembly by efficiently providing a supply of substrate for the similarly localized complex of PBPs and murein hydrolases in the periplasm. It has been shown that the abundance of undecaprenyl-phosphate lipid carrier is the rate-limiting factor in precursor availability for PBPs (Kohlrausch et al., 1989), perhaps indicating that MreD/MreB localization of precursor synthetic complexes is employed as a strategy to overcome this kinetic obstacle during cell wall synthesis.
Strains, media and DNA manipulations
Bacterial strains and plasmids used in this work are listed in Table S1. Transduction using bacteriophage ΦCR30 was performed as described by Ely and Johnson (1977). Plasmids were introduced into C. crescentus by either bacterial conjugation using E. coli S17-1 (Simon et al., 1983) as a donor or by electroporation. Caulobacter crescentus wild-type strain LS107 (syn-1000 Δbla6) (Alley et al., 1992) and its derivatives were grown in peptone yeast extract medium (PYE), M2-glucose (M2G), M2G-Nitrate (M2N) medium with glucose and/or xylose (0.2%)/vanillate (0.5 mM) when needed, nitrogen source (ammonium chloride or sodium nitrate), and supplemented with appropriate antibiotics when required. A22 was used at a concentration of 10 µg ml−1. In order to deplete MreD, cells were grown in PYE medium containing inducer (xylose or vanillate) to mid-logarithmic phase (OD600 = 0.5–0.8), washed three times with media lacking inducer and grown in PYE substituted with 0.2% glucose. For cloning and analysis of DNA, PCR and transformation, standard procedures were used (Sambrook et al., 1989). Bacterial two-hybrid plasmids were generated containing in-frame amino- or carboxyl-terminal T25 fusions (pKT25 or pKNT25) and amino- or carboxyl-terminal T18 fusions (pUT18C and pUT18) to each gene of interest. For 3′ primer of each gene, a stop codon was included or excluded depending on its destination for amino- or carboxyl-terminal fusion respectively. Detailed description of the construction of each gene fusion to T25 and/orT18 is provided in Supplemental Materials and Methods.
All plasmids containing fluorescent protein fusions were introduced into C. crescentus wild-type strain LS107 by conjugation using host E. coli strain S17-1 or electroporation using LS107 electro-competent cells. A xylose (pX.rbs.mreD.M2) or vanillate (pBV-DEST.rbs.mreD) inducible complementing copy of mreD was introduced by electroporation generating JG5045 and JG5040 respectively. In order to create an mreD depletion strain, pNPTS128-ΔmreD was introduced into C. crescentus for allelic exchange by conjugation and selected for kanamycin resistant and sucrose sensitive isolates. Complementing copies of epitope-tagged mreD-M2 were then introduced by transduction using ΦCR30 transducing lysates of JG5045 and JG5040. Allelic replacement of endogenous mreD with ΔmreD was performed by growing cells overnight in PYE supplemented with 0.3% xylose or 0.5 mM vanillate to induce mreD-M2, and plated on 3% sucrose-PYE. Sucrose-resistant colonies were screened for kanamycin sensitivity. The resulting depletion strains were either xylose (JG5041) or vanillate (JG5043) dependent. Transducing lysates of murG–mCherry (JG5027), mraY–mCherry (JG5044) and mipA-mCherry (JG5029) were used to transfer each fluorescent fusion into the MreD depletion strain (JG5041), producing strains JG5046, JG5047 and JG5048 respectively.
Purification of MurG
Escherichia coli expression strain BL21(DE3) containing His-MurG (pDEST-HIS.murG) was grown to an OD600 of 0.3 mM, isopropyl-β-d-thio-galactopyranoside (IPTG) was then added to a final concentration of 1 mM and growth was continued at 37°C with shaking for 2 h. Cells were lysed by French press and His-MurG was purified from lysates by chromatography using Ni2+-NTA agarose. Purified protein was used for antiserum production, prepared by a commercial source (Cocalico).
Bacterial two-hybrid interaction assays
Bacterial two-hybrid analysis was performed as described previously (Karimova et al., 1998). Briefly, recombinant pKT25 (or pKNT25) and pUT18C carrying genes of interest were used in various combinations to co-transform into competent BTH101 E. coli cells, which were then plated onto indicator plates containing LB-X-gal-(5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside, 20 mg ml−1)-IPTG (0.5 mM) and incubated for 48–72 h at 30°C. Functional complementation of adenylate cyclase activity between the hybrid proteins was quantified by measuring the β-galactosidase activities in chloroform-treated E. coli cells harbouring the plasmids as indicated. Each value represents the mean value ± SD from triplicate assays. Empty T18 and T25 fusion vectors were transformed as a negative control, and pT18-zip and pT25-zip (two halves of the leucine zipper) were used as a positive control.
In order to visualize cells during MreD depletion, cultures were sampled, treated with the fluorescent membrane dye, FM4-64 (4 µg ml−1; Molecular Probes) and covered with a poly-l-lysine-treated coverslip. Images were captured and analysed using a Deltavision deconvolution microscope employing a Resolve3D image-acquisition software package (Applied Precision). At least 15 and up to 20, 0.1 µm optical sections were obtained and deconvolved. Images were sampled at either 1000× or 1600× magnification as indicated. For immunofluorescence microscopy, cells were fixed at mid-logarithmic phase and incubated with appropriate antibodies as described previously (Figge et al., 2004). To quantify localization patterns, at least three independent fields per culture were chosen at random. For fluorescent ramoplanin labelling, cells were treated with 1:1 ratio of labelled and unlabelled Ramo-FL at a final concentration of 0.2 µg ml−1 for 5 min, mounted on poly-L-lysine-treated slides, and immediately (less than 10 min) examined using fluorescence microscopy. Fluorescent vancomycin labelling was performed as described previously (Divakaruni et al., 2007).
Ramoplanin and Ramoplanin-FL was a generous gift of M. Lazarus and S. Walker and A22 from J. Wagner and Y. Brun. We are grateful to members of the laboratory, C. Baida, M. Galustyan, M. Graf, R. Lewis, M. Llewellyn, and to A. Divakaruni for critical reading of the manuscript. Additionally we thank Josh Beck and Peter Bradley for helpful assistance and mCherry antisera. C.L.W. was supported by USPHS National Research Service Award GM07104. This work was supported by a grant from the National Science Foundation MCB-0641333 (to J.W.G.).