Bacterial actin MreB forms antiparallel double filaments

1Cell shape is a characteristic and hereditary feature that is crucial for existence and its regulation is a common challenge for organisms across all biological kingdoms. Shape in eukaryotic cells is provided by the cytoskeleton that consists of actin, tubulin, and intermediate filaments. Prokaryotes also have a dynamic, filamentous network of proteins, which are homologous to the eukaryotic cytoskeletal elements. In non-spherical bacteria, the actin homologue MreB is essential for shape maintenance as depletion of MreB through genetic knockouts or MreB-targeted drug treatment results in misshapen cells that eventually lyse. Incorrect localisation of MreB adversely affects polar targeting of protein complexes and, in some organisms, chromosome segregation. Vital for MreB's role in cell shape determination is the interaction with the elongasome, the protein complex orchestrating peptidoglycan cell wall morphogenesis. The elongasome consists of MreC and MreD (MreB's operon partners), RodA, PBP1A, PBP2, RodZ, as well as MurF, MurG and MraY. The elongasome ensures that new glycan strands cross-linked by short peptides are inserted along the peptidoglycan scaffold during cell elongation. The peptidoglycan layer provides structural integrity to the cell wall and counteracts turgor-induced lysis.

2The existence of helical, cytomotive bundles of MreB originally observed along the length of rod-shaped cells has been questioned by recent light microscopy studies that suggest MreB forms discontinuous fibres underneath the cell membrane that are driven by the elongasome itself. The MreB fibres are attached to the cell membrane via a hydrophobic loop in Thermotoga maritima as well as Gram+ bacteria and via an additional N-terminal amphipathic helix in Gram− bacteria. The tendency of purified MreB from Gram− bacteria to aggregate can be attributed to the presence of these hydrophobic sequences. Hence, previous biochemical and structural work has focussed on MreB from T. maritima (TmMreB) that can be obtained in much higher quantities than E. coli MreB or B. subtilis MreB. MreB assembles into straight, double protofilaments in the presence of ATP or GTP and these can gather laterally into sheets. Although the architecture of a single protofilament is surprisingly conserved within the actin family of proteins, the assembly of protofilaments into higher order structures differs greatly. The overall architecture has implications for filament polarity, which affects differential growth of the filament ends and directional movement. Force-generating motors that push plasmids apart all contain pairs of parallel helical protofilaments as their basic building block. Presumably, the membrane-binding ability of MreB and FtsA constrains their filament architecture to straight (or slightly bent), rather than helical, pairs of protofilaments. But how are these filaments arranged?

3Here, we demonstrate that MreB from Caulobacter crescentus forms pairs of antiparallel protofilaments as revealed by crystal structures, an architecture unprecedented among actin-like proteins. The antiparallel arrangement ensures that the amphipathic helix of each protofilament binds to membranes as shown by cryo-electron microscopy (cryo-EM). Pairs of protofilaments are essential for cell shape maintenance in Escherichia coli and these are shown to be oriented in an antiparallel fashion by in vivo cross-linking. The mechanism for protofilament formation of C. crescentus MreB is elucidated by crystallography and shown to be reminiscent of that of other actin-like proteins, revealing the typical domain closure upon polymerisation. Finally, co-crystallisation studies of Caulobacter MreB with the antimicrobial reagents A22 and MP265 explain their inhibitory mechanism and shed light on the activation mechanism of ATP hydrolysis during polymerisation.

4Structural insight into MreB from mesophilic bacteria has been hampered by the tendency of the proteins to aggregate upon purification, but through biochemical and bioengineering tools sufficient quantities of stable versions of Caulobacter crescentus MreB were obtained (Table 1) that were amenable to structural studies. Deleting the N-terminal amphipathic helix and mutating the residues in the hydrophobic loop (Figure 1—figure supplement 1) make the protein (ΔCcMreBdh) less prone to aggregation. The crystal structure of Caulobacter MreB (ΔCcMreBdh) reveals the typical actin fold, containing two domains on either side of a nucleotide binding cleft (Figure 1A; ). Both domains are separated into two subdomains, A and B. Domain II is structurally the most preserved domain within the actin family of proteins, whereas domain IB is the most diverse. But even this domain is very similar in both Caulobacter MreB and Thermotoga maritima MreB (TmMreBh, ), superimposing with an RMSD of 0.97 Å (Figure 1A). The protofilament architecture is also very well conserved, with an identical subunit repeat distance of 51.1 Å and a similar arrangement of the longitudinal interface (Figure 1C). However, only single protofilaments were present in the Thermotoga MreB crystals, despite the fact that electron microscopy images clearly show doublets. In contrast, the crystals of Caulobacter MreB contain doublets that, remarkably, crystallised in an antiparallel arrangement (Figure 1D). The electrostatic surface potential of the double protofilaments reveals the hydrophobic loop and the basic charges surrounding the membrane-binding site (Figure 1B).

5Full-length, untagged MreB from C. crescentus (CcMreB) binds to a lipid monolayer as a double protofilament as shown by negative stain electron microscopy (EM, Figure 2A), similarly to TmMreB. 2D averaging of images of negatively stained filaments shows double protofilaments interacting with their flat sides (Figure 2B), which coincides with other, double-helical actin-like protein filaments. Although the resolution is low, the entire double protofilament structure of ΔCcMreBdh fits nicely into the 2D averaged EM images with a longitudinal subunit repeat of 5.1 nm and the widths of the doublet encompassing 6.5 nm (Figure 2C). The previously reported placement of two single protofilaments of TmMreB into 2D averaged filaments, though merely an educated guess at that time, turned out to be right. By introducing a single point mutation in the interprotofilament interface of CcMreB (V118E), single protofilaments are formed as shown by negative stain EM (Figure 2D). Pairs of protofilaments of CcMreB distort liposomes as seen by cryo-EM (Figure 2E; Video 1) and coat the membrane extensively (compare Figure 2G to uncoated membranes in Figure 2F). The irregular arrangement of CcMreB filaments on liposomes makes it unsuitable for subtomogram averaging. Hence, we turned to rigid lipid tubes made of E. coli total lipid extract mixed with C24:1 β-D galactosyl ceramide. These tubes remain straight in the presence of MreB and are suitable for obtaining higher resolution reconstructions revealing a side view that shows how the protein coats the membrane (Figure 3A–D; Video 2). Although overall appearances are very similar, TmMreB forms much straighter protofilaments on lipid tubes than CcMreB (Figure 3C), making TmMreB more amenable to subtomogram averaging (Figure 3E–G). Both 2D and 3D averaging unambiguously show how close an MreB filament sits on the tubes, extending the thickness of the density including the lipid bilayer to 12.5 nm (Figure 3E). As shown in Figure 3E and Figure 3G, the bilayer is resolved well and the longitudinal repeat distance of the columns protruding from the outer leaflet is in agreement with the expected subunit repeat of 5.1 nm in TmMreB protofilaments. Also, subtomogram averaging clearly demonstrates that membrane binding of MreB distorts the outer, but not the inner leaflet of the lipid bilayer, which is in agreement with the proposed binding mode of TmMreB via a hydrophobic loop. It is not very surprising to us that MreB has not been seen in cells by cryo-EM, as tomograms of cells are usually taken at rather low magnification and high defocus, leading to a resolution lower than what is needed to resolve the leaflets of the lipid bilayer, let alone, an MreB filament from the protein-embedded membrane.

6A first indication that a doublet of protofilaments is required for MreB's function comes from genetic studies in E. coli (Figure 4). An MreBCD knockout strain (FB17), otherwise not viable, survives if complemented with extra FtsZ and a copy of a plasmid expressing MreC and MreD (pλ::mreCD; pFB124). Additional FtsZ is provided indirectly from a second plasmid (pFB112) that constitutively expresses the transcription factor SdiA, ensuring elevated levels of FtsQAZ. Cells become rod-shaped and lose their dependency on elevated levels of FtsZ if complemented with a third plasmid carrying a functional copy of mreB under the lac promoter (pFB209) (Figure 4A). This complex genetic set up enables investigation of the phenotype of MreB containing mutations that might render the protein non-functional, since MreB is essential in E. coli. Altered versions of MreB were introduced that carried a mutation disrupting either the intra- or interprotofilament interface (EcMreB-S284D or EcMreB-V121E, respectively, Figure 4D), as deduced from the crystal structure. Indeed, a single mutation in the intra-protofilamant interface renders the protein monomeric as shown by structural analysis of the equivalent mutant S283D in CcMreB (presented below), whereas a single mutation in the interprotofilament interface results in single protofilaments as discussed above (CcMreB-V118E, Figure 2D). In the absence of MreB, cells are spherical (first column Figure 4B), but they revert back to rod shape when expressing wild-type MreB (second column Figure 4B). When either the intra- (S284D) or inter-protofilament (V121E) interface is disrupted, cells remain spherical (third and fourth column Figure 4B, Figure 4C). These results demonstrate firstly that MreB protofilament formation is essential for cell shape maintenance and secondly that these protofilaments need to interact with their flat sides to make functional MreB filaments.

7Thus far we have established that pairs of protofilaments are required for functional MreB in vivo and that these protofilaments interact with their flat sides (Figure 4). To verify that the antiparallel arrangement of protofilaments observed both in EM and in the crystal structure reflects the functional state of MreB in cells, site-specific cross-linking was performed. Cysteine residues were introduced at specific sites in and outside the protofilament interface and mutated MreB was expressed in the MreB knockout strain as described above. But before that, cysteine residues present in native MreB had to be mutated to serine residues to ensure that any cross-linked product would arise from the newly introduced cysteine(s). As shown in Figure 5—figure supplement 1, the cysteine-free form of MreB fully complemented an mreB knockout strain. At late log phase, the cysteine mutants were subjected to the thiol-reactive cross-linker bismaleimidoethane (BMOE). Thiol-reactive cross-linkers work by forming covalent bonds between two cysteine residues. As the protofilaments in the antiparallel arrangement are symmetry-related, certain single cysteine locations may produce a cross-linked dimer as long as the distance between the mutated residues is within reach of the cross-linker arm (∼8 Å). As shown in Figure 5, cross-linked dimers were obtained when cysteine residues were introduced in the protofilament interface (Figure 5B, lanes 1–7, 9–10, 12, 13, 14), but not when cysteine residues were located outside this interface (Figure 5B, lanes 16–19), nor in the absence of any cysteine (Figure 5B, lane 15). When a pair of cysteine residues was introduced on opposite sides of the antiparallel protofilament interface, stronger bands for the cross-linked dimers were repeatedly obtained than those of the sum of each mutant alone, providing further proof of the antiparallel arrangement (Figure 5B, lanes 8 and 11, compared to lanes 6–7 and 9–10, respectively).

8The phenotypes of the mutants were analysed at the same time as the cross-linking was performed—at late log phase—a time point previously shown to give good complementation with wild-type MreB. The cysteine-free mutant of MreB shows good complementation, as do some, but not all of the cysteine mutants (Figure 5—figure supplement 1). The lack of complementation of some of the mutants does not correlate with their ability to form filaments, indicating that analysis of the phenotype, which is known to be exquisitely sensitive, does probe more features than can be attributed to MreB polymerisation alone, such as filament dynamics and the interaction with host proteins.

9To form a detailed picture of the requirements of MreB to polymerise, the full polymerisation cycle of MreB was investigated by X-ray crystallography. Crystal structures of CcMreB in different nucleotide states, either as a monomer or as a polymer, were determined to a resolution better than 2 Å (Figure 6A). Most crystal structures were obtained using ΔCcMreBdh, which behaved biochemically better than the full-length protein, yet was still showing correct filament formation as presented in Figure 6—figure supplement 1. A single mutation in the intra-protofilament interface (S283D, ΔCcMreBdmh) renders the protein monomeric, as shown both by crystallography (Figure 6) and by EM (Figure 6—figure supplement 1). This mutant turned out to be most useful in studying the polymerisation cycle. From looking at the crystal structures of MreB in different nucleotide states, it immediately becomes apparent that surprisingly small structural changes can trigger MreB polymerisation. Monomeric CcMreB shows little structural movement between the ADP- and the ATP analogue AMPPNP-bound states (Figure 6D) with domains I and II overlapping with an RMSD of 1.58 and 0.57 Å, respectively (Figure 6B). An initial domain closure is observed upon binding of AMMPNP (Figure 6D) by the monomer. This movement progresses into a full propeller twist once it reaches the double protofilament state (Figure 6E). Upon polymerisation, domain IB shows a rotation of 21.7° towards the nucleotide binding cleft as analysed by Dyndom (Figure 6—figure supplement 2A; ). This domain closure is accompanied by a twist between domains IA and IIA, as can be appreciated from the top view in Figure 6—figure supplement 2D and from the schematic in Figure 6C. The relatively small conformational change upon polymerisation has a dramatic effect on the chemistry of the active site. In the AMPPNP protofilament conformation, the active site residues E140 and T167 coordinate a water molecule that is in line for the hydrophilic attack on the gamma phosphate (Figure 6F). In the monomeric state, the active site residue E140 of MreB has moved away from the nucleotide by just 1 Å. But consequently, it can no longer coordinate the attacking water, which is absent in this structure, and thus renders the monomer inactive (Figure 6F). Once MreB is in the polymeric state, the presence or absence of nucleotide has very little effect on the overall structure, which is presumably stabilised by inter- and intra-protofilament contacts. The mechanism of disintegration of the polymer remains to be elucidated.

10The MreB inhibitor S-(3, 4-dichlorobenzyl) isothiourea (A22) and its less cytotoxic and much more water-soluble derivative MP265 have been used extensively in cell shape studies as they perturb cell morphology reminiscent of an MreB knock-out. Analyses of revertants that are resistant to A22 suggested that the inhibitor targets the active site of MreB. Here, we report the mechanism of the inhibitory action using biochemical and structural methods.

11Isothermal titration calorimetry experiments were used to acquire binding constants for the inhibitor MP265 and nucleotides that bind to the active site of ΔCcMreBdh. It revealed that ΔCcMreBdh has the highest affinity for ATP (Kd ∼1 µM), followed by ADP (Kd ∼4 µM) and ATPγS (Kd ∼4.6 µM) and finally by AMPPNP (Kd ∼65 µM). The inhibitor MP265 binds to ΔCcMreBdh with a Kd ∼27 µM and, somewhat surprisingly, its binding affinity increases 20- to 30-fold in the presence of di- and trinucleotide phosphates (Figure 7A–C). This is in contrast to results published earlier, which reported that binding of A22 to TmMreB seems to be sterically incompatible with simultaneous binding of ATP, as was determined by modelling.

12Co-crystallisation of A22 and ΔCcMreBdh shows unambiguously that the inhibitor binds in the presence of ADP (Figure 7D). The anomalous difference map reveals the positions of the chlorine atoms of A22 and the di-phosphate of ADP (Figure 7D). At 1.5 Å resolution, it absolutely defines the exact position of a small molecule like A22. Because MP265 is much more manageable, co-crystals of MP265 and ΔCcMreBdh were obtained in the presence of both ADP and AMPPNP (Figure 6A,B, Figure 7E,F). The crystal structures of the complexes show that the inhibitor most likely prevents phosphate release in the MreB protofilament: MP265 blocks the exit channel for the phosphate and binds directly to the active site residue E140, preventing it from coordinating the catalytic water. In addition, it binds to the gamma phosphate (Figure 7E). Interestingly, crystals containing the MreB-inhibitor complex always contain single, rather than double protofilaments (Figure 6B; Table 2). A closer look at the inter-protofilament interface explains why double protofilaments can no longer form: as the inhibitor prevents the main dimerisation helix (Ala117-Ala130) from reaching into the interface, it causes weakening of the inter-protofilament contacts and likely reduces stability of the double protofilament (Figure 7F).

13The antiparallel arrangement of protofilaments of C. crescentus MreB as revealed by the crystal structure (Figure 1D) and by EM (Figure 2A–C) is unprecedented within the actin family of proteins. It ensures that the amphipathic helix of both protofilaments can bind to the membrane and that both protofilaments are able to interact with bitopic proteins, such as RodZ. In vivo evidence shows that the antiparallel arrangement is not an in vitro artefact and does indeed reflect the filament architecture in E. coli (Figures 4 and 5). Filament formation could have a synergistic effect on the elongasome by bringing together components that reach a higher local concentration through the interaction with polymeric MreB. Given the almost surprising high degree of similarity between the structures of MreB from mesophilic and thermophilic organisms (Figure 1A), it is expected that the antiparallel arrangement of MreB protofilaments is conserved among the bacterial kingdom. This has far reaching consequences for the molecular mechanism of MreB's function. Actin, MamK, ParM, and the other plasmid-segregating, actin-like proteins form pairs of protofilaments with subunits facing the same direction. In F-actin, the polar ends exhibit different growth and shrinkage rates and selectively interact with binding partners that affect filament protection or nucleation. Pairs of parallel ParM protofilaments extend from the ParRC-interacting end. Discrimination of either end has to be absent in the antiparallel arrangement found in the doublets of MreB protofilaments (Figure 1—figure supplement 2) and hence MreB filaments must grow and shrink from both ends. The bidirectional movement of MreB filaments in vivo is in agreement with the above hypothesis. However, previously reported treadmilling of CcMreB molecules through filaments assumes an overall polarity that, as we show, is absent in the filament. It has to be determined whether MreB filaments fluctuate at a steady state, growing and shrinking from both ends, or exhibit dynamic instability leading to catastrophic disassembly, as described in detail for ParM.

14All MreB proteins studied to date by EM appear in pairs rather than isolated single protofilaments, indicating that double protofilaments could be more stable and/or that filament nucleation requires inter-protofilament contacts. In vitro, single CcMreB protofilaments can only be formed upon disruption of the protofilament interface (Figure 2D). Furthermore, the equivalent mutation in vivo renders the protein non-functional (Figure 4). In the presence of the antimicrobial agents A22 and MP265, MreB forms single but no double protofilaments (Figure 7). By dislocating one of the dimerisation helices and thereby weakening the inter-protofilament interface, the inhibitor might destabilise the MreB filament. This would explain the disruption of MreB filaments by A22 as reported previously.

15Association of protofilaments into filaments is unique for each actin-like protein and appropriate for their function. F-actin forms parallel, two-stranded helical filaments with a right-handed twist, plasmid segregating protein ParM forms parallel helical filaments with a left-handed twist, that arrange in an antiparallel fashion, whereas cell division protein FtsA forms pairs of straight filaments with unknown orientation. Low-resolution images of actin-like protein Alp12 suggest that two pairs of parallel filaments twist around each other in an antiparallel fashion, whereas MamK and Crenactin form a two-stranded, non-staggered helix. Despite these variations in overall filament architecture, the head-to-tail arrangement of subunits in a single protofilament is surprisingly conserved, given the low sequence similarity (∼15%). In addition, all protofilaments interact with each other using their flat sides, irrespective of the filament architecture. Moreover, the propeller twist required for filament dynamics in response to nucleotide binding and hydrolysis is preserved. The result of the propeller twist is the closure of domains I and II and the flattening of one side of the molecule that makes the inter-protofilament interface (Figure 6C, Figure 6—figure supplement 2). The recently published molecular dynamics simulations of MreB are in agreement with this observation. Furthermore, detailed structural studies of the different nucleotide states of the best studied actin-like proteins MreB and ParM show that the polymerisation cycle is fairly conserved, with the largest domain movement taking place upon the transition from the monomeric state to the polymeric form (Figure 6,). In both proteins, the active site geometry in the filament conformation favours nucleotide hydrolysis, which is driven by a catalytic glutamate (E140 in CcMreB and E148 in ParM) that positions the catalytic water in line with the gamma phosphate. The co-crystal structures of AMPPNP-bound CcMreB with the inhibitor confirm the critical role of this catalytic residue E140 (Figure 7). In the AMPPNP-bound monomer conformation, no catalytic water is present in CcMreB (Figure 6F), whereas the catalytic water in monomeric AMPPNP-bound ParM has moved away from the ideal geometry, thus hampering hydrolysis.

16It has been reported previously that MreB can polymerise in the presence of either ATP or GTP. From the ΔCcMreBdh co-crystal structures with ADP and AMPPNP, it is clear that the active site could well accommodate a guanidine instead of an adenine as the major contacts are with the ribose and phosphate moieties of the nucleotide, a feature, again, shared with ParM.

17Taken together, our findings describe a novel architecture of actin-like protofilament pairs that is unprecedented among the actin family of proteins, whether of eukaryotic or prokaryotic origin. The straight, antiparallel arrangement of MreB protofilaments ensures that each subunit can interact with the membrane and membrane-bound components of the elongasome. The non-polar nature of MreB filaments explains common assembly and disassembly kinetics for both filament ends. Future experiments will explore the implications of the antiparallel architecture of MreB on cell wall synthesis and investigate whether MreB dictates bidirectional growth of the cell wall.

18Genes of Caulobacter crescentus MreB (GenBank: ACL95077.1) and Thermotoga maritima MreB1 (GenBank: AAD35673.1) were cloned into the small T7-vector pHis17 (Bruno Miroux, personal communication) by means of PCR and restriction cloning, encoding for the his-tagged proteins: ΔCcMreBh (CCM2), ΔCcMreBdh (pFE397), and ΔCcMreBdmh (CCM3). Full-length, non-tagged CcMreB was cloned as an N-terminal His-SUMO fusion into pET28a (KanR) using In-Fusion technology (Clontech). The vector encoding MGSSHHHHHH-SUMO from pET28a (POPINS, http://www.oppf.rc-harwell.ac.uk/OPPF/protocols/cloning.jsp) was linearised with HindIII and KpnI, purified and mixed with the cloning enhancer treated insert encoding full-length CcMreB for the In-Fusion reaction, resulting in plasmid pFE403. Protofilament interface mutations were introduced by site-directed mutagenesis, essentially as described below for the mutants used in genetic complementation. Full-length non-tagged TmMreB was cloned as a C-terminal Intein–Chitin binding domain fusion into NdeI/SapI linearised pTXB1 (NEB, ampR) using In-Fusion (Clontech), resulting in plasmid pFE349. The gene encoding for the C-terminal catalytic domain of human SUMO1/sentrin specific peptidase (SENP) was cloned into BamHI/NotI sites of pGex-6p-1 (ampR) and was a gift from David Kommander (MRC-LMB). All plasmids were verified by sequencing and are listed in Supplementary file 1.

19Full-length CcMreB was expressed from pFE403 (KanR) and the inter-protofilament mutant V118E in full-length CcMreB was expressed from pFE542 (KanR) in C41 cells (Lucigen). Both proteins are fused to an N-terminal His-SUMO tag. Protein expression was induced with 1 mM IPTG and cells were grown in 2xTY medium at 18°C for 8 hr. Bacteria were pelleted by centrifugation, resuspended in buffer A (50 mM Tris–HCl, 300 mM NaCl, 10% glycerol, 5 mM TCEP, pH 8.0), supplemented with DNase (Sigma), RNase (Sigma), and protease inhibitor tablets (Roche) and lysed using a cell disruptor (Constant Systems) at 20 kpsi. The lysate was cleared by centrifugation in a 45Ti rotor (Beckman) at 40k rpm for 25 min and loaded at 2 ml/min on two 5 ml HisTrap columns (GE healthcare). The columns were washed with 5% buffer B (1M imidazole, in buffer A, pH 8.0) and non-tagged CcMreB was eluted from the column after overnight incubation with Sumo protease (GST-SENP). The eluted protein was passed through a GST column to catch away the Sumo protease. The protein was purified from higher order oligomers using size-exclusion (S300HR, GE Healthcare) equilibrated in buffer C (20 mM CHES, 300 mM NaCl, 10% glycerol, 2 mM TCEP, pH 9.5) and concentrated to ∼10 mg/ml using Centriprep YM-10 concentrators before being frozen into small aliquots. Electrospray masspec for full-length WT protein gave 36,691 Da (36,682 Da expected).

20The protein was expressed overnight at 25°C in E. coli C41(DE3) cells (Lucigen) with 1 mM IPTG. Cells were resuspended in buffer A (50 mM Tris, 300 mM NaCl, 10% wt/vol glycerol, 5 mM TCEP, pH 8.5) and then opened using a Constant Systems cell disruptor, set at 30k psi pressure. After clearing the lysate in a 45Ti rotor (Beckman) at 35,000 rpm it was loaded at 2 ml/min on two 5 ml HisTrap columns (GE Healthcare). Increasing amounts of buffer B (1 M imidazole, pH 8.0) were applied in steps of 2, 5, 10, 30 and 100%, with most of the protein eluting in pure form at 30% B. The pooled fractions were concentrated to around 2–5 ml with Centriprep YM-10 concentrators (Millipore) and applied to a Sephacryl S300HR column, equilibrated in buffer A. Fractions from the size exclusion column were checked by SDS-PAGE, pooled and re-concentrated to around 20–30 mg/ml using Centriprep YM-10 concentrators. Typically, around 15 mg of final product were obtained from 12 L culture. Electrospray masspec gave 36,916 Da (36,911 Da expected).

21Expression and purification followed that of ΔCcMreBh, with the following alterations. Expression was performed at 15°C overnight. Elution from the HisTrap nickel resin peaked at 10% B. The protein elutes with bound ADP from the final S300HR column if no EDTA is used in buffer A. (This was apparent from the 260/280 nm ratio and the crystals obtained from the material, structures: s_2, s_2_a, and s_2_m, see Table 2). When 5 mM EDTA was included in buffer A for the Sephacryl S300HR run, the protein eluted without bound nucleotide. The protein becomes more problematic when free of nucleotide and needs to be kept cold in order to avoid precipitation, certainly at higher concentrations. Electrospray masspec gave 36,813 Da (36,809 Da expected).

22Expression and purification followed that of ΔCcMreBh, with the following alterations. Expression was performed at 15°C overnight. Elution from the HisTrap nickel resin peaked at 30% B. When 5 mM EDTA was included in buffer A for the Sephacryl S300HR run, the protein eluted without bound nucleotide. When free of nucleotide, the protein becomes more problematic and needed to be kept cold at all times in order to avoid precipitation, certainly at high concentrations.

23Full-length TmMreB was expressed as an Intein-CBD-tagged protein fusion from plasmid pFE349 and purified as described previously.

24GST-SENP was expressed in C41 cells (Lucigen) overnight at 15°C. Cells were collected by centrifugation and lysed in buffer A (50 mM Tris, 150 mM NaCl, 1 mM EDTA, 5% glycerol, 2 mM DTT, pH 8.5), supplemented with DNase (Sigma), RNase (Sigma), and protease inhibitor tablets (Roche) using a cell disruptor (Constant Systems, at 20 kpsi). Lysate was cleared by centrifugation for 25 min, at 40k rpm, in a Ti45 rotor (Beckman). The supernatant was bound to 20 ml GST sepharose 4B beads (GE Healthcare) in batch, rotating for 2 hr, at 4°C. The beads were thoroughly washed in buffer A, including a wash in buffer A + 500 mM NaCl and the fusion protein eluted in buffer A supplemented with 10 mM reduced glutathione. After concentrating with a Vivaspin 20 concentrator (30 MWCO), the protein was further purified by size-exclusion chromatography on a Sephacryl S200HR column (GE Healthcare), equilibrated in CcMreB buffer A (50 mM Tris, 300 mM NaCl, 10% glycerol, 5 mM TCEP, pH 8.0). The protein was concentrated to 10–15 mg/ml and frozen in aliquots. Typical yields were ∼80 mg of protein from 10 L culture.

25Initial crystallisation conditions were found using our in house nanolitre crystallisation facility. Final crystals were grown in 96-well MRC crystallisation plates (SWISSCI AG), combining 100 nl of protein solutions at 10–30 mg/ml with 100 nl of reservoir solutions, which are listed in Supplementary file 2. Crystals were harvested in loops using a dedicated cryoprotecting solution (Supplementary file 2) and snap-frozen in liquid nitrogen. In house X-ray diffraction data were collected on an Fr-E Superbright rotating anode generator (Rigaku), equipped with a MarDTB image plate detector (marresearch GmbH). Synchrotron data were collected on beamlines I02, I03, I04 and I04-1 at Diamond Light Source (Harwell, UK) and beamlines id23eh1 and id14eh4 at ESRF (Grenoble, France), as indicated in Table 2. Diffraction data were indexed and integrated with XDS and scaled and merged with SCALA. All structures were solved by molecular replacement using PHASER, either with T. maritima MreB as the search model or a related structure from within this study. Structures were refined using PHENIX.refine using recommended settings and standard restraint libraries, alternating with manual checking and rebuilding in MAIN. Final models were deposited in the Protein Data Bank (PDB) with codes listed in Table 2.

262D lipid monolayers were prepared following a modified version of the protocol of. A carbon-coated electron microscopy nickel grid was deposited on a drop of E. coli total lipid extract (Avanti Polar Lipids) floating on CcMreB polymerisation buffer (12.5 mM Tris-HCl, 7.5, 2 mM ATP, 4 mM MgCl2) in a teflon well. The grid was removed and after the excess of liquid had been gently blotted away, pre-assembled CcMreB filaments were added for 60 s before staining with 2% uranyl acetate. CcMreB polymerisation (0.25 mg/ml) was carried out in a low salt buffer (12.5 mM Tris–HCl, 7.5, 2 mM ATP, 4 mM MgCl2). Imaging was performed using a 120 kV Tecnai 12 electron microscope (FEI) and the data processed with EMAN2.

27Liposomes were prepared from E. coli total lipid extract (Avanti Polar Lipids) and lipid tubes were formed by mixing E. coli total lipid extract with 30% of D-Galactosyl-β1-1’-N-Nervonoyl-D-erythro-sphingosine (C24:1 β-D Galactosyl-Ceramide, Avanti Polar Lipids). 1 mg/ml of lipid was included in polymerisation buffers for both CcMreB and TmMreB (1 mg/ml); CcMreB polymerization buffer: 12.5 mM Tris-HCl, 7.5, 2 mM ATP, 4 mM MgCl2 and TmMreB polymerisation buffer: 100 mM Tris-HCl, 7.5, 200 mM NaCl, 2 mM ATP, 4 mM MgCl2. MreB-coated lipid tubes samples were applied to a thin continuous carbon film grid (Lacey carbon, Cu, 400 mesh) whereas MreB-liposome samples were applied to holey carbon grids (Quantifoil R2/2, Cu/Rh, 200 mesh) both from Agar Scientific and subsequently frozen in liquid ethane using a FEI Vitrobot. Grids were then transferred to either a FEI Tecnai G2 Polara or a FEI Titan Krios operated at 300 kV. Single frame images were acquired on the G2 Polara with a back-thinned FEI Falcon II detector. Cryo-electron tomography experiments were carried out on a FEI Titan Krios microscope equipped with a Gatan Quantum GIF and K2 Summit direct electron detector operated using SerialEM. Tomograms were recorded at a nominal magnification of 26000x (4.5 Å/pixel) with a 3° increment and a tilt range of 60 to −60°, keeping the total electron dose below 120e−Å−2. Tilt series data were reconstructed using the IMOD package. Two-dimensional cryoEM data were processed using EMAN2. Subtomogram averaging reconstructions were conducted for 16 tubes from 10 different tomograms independently and ab initio using the AV3 package. The overall arrangement of the MreB filaments in all analysed tubes was the same. The subtomogram averaging reconstruction from one representative tube is displayed in Figure 3G.

28Mutated versions of MreB were examined when expressed from a lac promoter on a low copy number plasmid in an MreBCD knock-out strain (FB17, ). In addition, the MreBCD knock-out strain carried plasmid pFB112 (tetR), constitutively expressing transcription factor SdiA and plasmid pFB124 (specR), expressing MreCD at 37°C. Mutant versions were created following a modified protocol of site-directed mutagenesis using desalted, partially overlapping primers. Instead of Pfu polymerase, Q5 High fidelity polymerase (NEB) was used that possesses a higher fidelity and processivity than Pfu polymerase. The mutation was included in the overlapping region of the primer pair that has a Tm of 51–57°C. The Tm for the non-overlapping region was 60–63°C (6–10°C higher than the primer pair overlap). A typical PCR reaction contained 25 ng template plasmid and 0.5 µM of each primer, and cycled 12 x, with an annealing's temperature 3°C above the lowest non-overlapping Tm. The PCR products were treated with DpnI for 2 hr and transformed into chemically competent DH5α cells. Sequencing confirmed a 100% success rate. Mutant MreB was transformed into the strain FB17/pFB112/pFB124 and grown for 6 hr in LB supplemented with 250 µM IPTG, 50 µg/ml ampicillin and 50 µg/ml spectinomycin at 37°C. Cells were stained with FM4-64 and examined with a Zeiss laser-scanning confocal microscope (LSM510) equipped with a 63x oil immersion objective lens. Images were processed using ImageJ.

29On the basis of the gained structural information a single or pair of cysteine residues was introduced into a cysteine-free form of E. coli MreB as the only source of MreB as described above. Over night cultures were diluted 1:100 into 10 ml LB supplemented with 50 µg/ml spectinomycin, 50 µg/ml ampicillin and 250 µM IPTG and grown for 6 hr at 37°C. Cultures were cooled down by adding 30% ice and spun at 1000×g for 10 min at 4°C. The pellet washed in ice-cold 10 ml PBSG (PBS, supplemented with 0.1% glycerol), spun at 1000×g for 10 min at 4°C. The cell pellet was resuspended in 2 ml PBSG, supplemented with 5 mM EDTA. Half of the resuspended cells were subjected to the thiol-reactive cross-linker bismaleimidoethane (BMOE, Pierce) at a concentration of 100 µM from a 20 mM stock in DMSO, whereas the same amount of DMSO was added to the other half of the cells as a control. After 10 min on ice, the reaction was quenched with 28 mM β-mercaptoethanol and spun for 10 min at 14 krpm at 4°C (Eppendorf table top centrifuge). The pellets were resuspended in 600 µl B-PER (Thermo Scientific), supplemented with 25 mM DTT, 2.5 mM MgCl2, 4 mM CaCl, lysozyme, DNAse, RNAse, and incubated at RT for 10 min. The lysed cells were mixed with NuPAGE LDS sample buffer (Life technologies) supplemented with 8 M urea and 5 mM buffered TCEP (pH 7) and heated for 10 min 70°C. Cross-linked products were separated from the monomers on a NuPAGE 4–12% Bis-Tris gel (Life technologies) and detected by Western blot using affinity purified α-MreB antibodies.

30All ITC measurements were performed at 25°C using an auto-iTC 200 instrument (GE Healthcare) in 50 mM Tris, 300 mM NaCl, 10% vol/vol glycerol, 1 mM TCEP, 5 mM MgCl2, pH 8.0. Samples were stored by the instrument in 96-well microtitre plates at 5°C prior to loading and performing the ITC titrations. Standard experiments used 19 × 2 µl injections of nucleotide or MP265 into ΔCcMreBdh protein preceded by a single 0.5 µl pre-injection. Heat from the pre-injection was not used during fitting. Data were analysed manually in the Origin software package provided by the manufacturer and fit to a single set of binding sites model. All measurements of binding were corrected using control dilution ITC experiments in which the nucleotide or MP265 was injected into appropriate buffer alone. The small endothermic heats of each injection in these experiments were fitted to a simple linear function that was subsequently subtracted from the equivalent integrated heats of the experiment when protein was present. The concentration of ΔCcMreBdh in the cell was typically ∼70 µM while the concentration of nucleotide or MP265 used in the syringe was ∼2 mM.