Purification of monomeric MreB
MreB has been purified under various ionic and buffer condition, and is known to be prone to spontaneous polymerization or aggregation. Addition of magnesium can efficiently induce filament formation of MreB from various bacteria [10, 11, 39–41], and YFP-MreB from B. subtilis polymerizes as efficiently as non-fused MreB [42]. MreB from C. crescentus and T. maritima have recently been purified and imaged on vesicles or on a flat membrane, where they form antiparallel double filaments as smallest unit [11]. We wished to obtain a better understanding of MreB filaments from a Gram-positive bacterium. In order to obtain monomeric Strep-YFP-MreB from B. subtilis, we tested various previously described as well as novel buffer and growth conditions, but MreB was predominantly present close to the void volume of the gel filtration columns (Fig. 1A), especially under salt concentrations below 300 mM NaCl, only a small amount of monomeric MreB could be obtained (Fig. 1A). Concentrations of over 300 mM NaCl in the purification buffer resulted in flocculent aggregation of MreB (data not shown). Overexpression in media with increased salt or sugar concentrations to induce weak osmotic pressure and the addition of betaine as an osmoprotectant has been shown to reduce the occurrence of aggregation for proteins that are difficult to purify [43]. A combination of overnight expression at low temperatures under weak osmotic pressure with 500 mM sorbitol (plus 1 mM betaine) and purification of the obtained cell pellets using a buffer containing 300 mM salt (Purification buffer: 100 mM Tris HCl, 300 mM NaCl, 1 mM EDTA, 0.2 mM ATP, 5% glycerol pH 7.5) resulted in a peak containing monomeric MreB (Fig. 1A) that could be isolated via size-exclusion chromatography. The protein was dialyzed and stored in low salt polymerization buffer (5 mM TRIS-HCl, 0.1 mM CaCl2, 0.2 mM ATP, pH 7.5).
Although the yield could not be further increased by any additional measure that was tried, the amounts were sufficient for further experiments. Of note, monomeric MreB from fraction A6 (Fig. 1A,B) remained monomeric for one week of storage at 4 °C (data not shown) and retained its polymerization activity. Like Strep-YFP-MreB, Strep-CFP-Mbl and Strep-mCherry-MreBH could also be purified as monomers (Fig. 1B). For simplicity, the “Strep” tag is no longer mentioned in the following text. To verify that the isolated fractions were truly monomeric, firstly yields of the respective proteins from the monomer peaks were loaded onto a 5–15% sucrose density gradient and separated via ultracentrifugation. Gel-filtration standard from Biorad was used as a reference (Fig. 1C). The marker proteins were observed in lane 1: Myoglobin (17 kDa), 2: Ovalbumin (44 kDa), 4: Gamma-Globulin (158 kDa), 10: Thyroglobulin (670 kDa), whereas YFP-MreB, CFP-Mbl, mCherry-MreB (~ 65 kDa) all appeared starting in lane 2, indicating low molecular mass (Fig. 1C). There were also visible bands in lane 2 and weak bands in lane 3 for the three fluorophore-tagged paralogs, but no visible band for higher molecular masses. To further test if the peaks isolated from gel filtration were truly monomeric, we performed a photometric mass analysis of the proteins (Fig. 1D) [44]. For all three proteins only a single peak was observed. YFP-MreB showed a monomeric peak at 68 ± 23 kDa, CFP-Mbl at 76 ± 37 and mCherry-MreBH at 101 ± 29 (Fig. 1D). Overall, the photometric approach offered high precision, even though the peak for mCherry-MreBH was slightly higher than the expected size of a monomer. Taken together with other data previously detailed, a monomeric form of the three MreB paralogs could be successfully obtained (Fig. 1A-D). 300 mM NaCl during the purification procedure, together with expression under osmotic pressure, appears to be the most stable condition to avoid polymerization of MreB (which likely occurs at lower salt concentration) or aggregation, which we assume to happen at higher salt conditions.
Mreb Forms Predominantly Bundles Of Filaments In Solution
We used transmission EM and negative staining with uranyl acetate, and visualized monomeric purified MreB and YFP-MreB in low-salt polymerization buffer (Fig. 2A and 2D), and polymers after induction of polymerization with 10 mM MgCl2 (Fig. 2B/C/E/F). Without the addition of MgCl2 no filamentous structures were observed, only small accumulations, likely monomers of MreB and YFP-MreB, were present on the grid (Fig. 2A + D). After induction of polymerization, extended filamentous structures had formed that could branch or fuse (Fig. 2B,C,E,F). The width of these filaments was in most cases below 200 nm, but occasionally intertwined macrostructures were observed, that could be over 200 nm wide (Supp. Figure 2A-B). In some cases these structures appeared to be twisted bundles of protofilaments (Additional file 1C-D). Individual sheets of filaments could branch off or fuse (Additional file 1C, E), with the smallest number of protofilaments observed being two (e.g. Figure 2B). This is in agreement with earlier reports that the minimal unit of MreB filaments are two (anti-) parallel protofilaments [10, 11]. Importantly, filaments formed by YFP-MreB (Fig. 2E/F) were visually indistinguishable from those of MreB (Fig. 2B/C). Interestingly, most structures were highly branched, such that many filamentous structures had an overall fuzzy appearance (Additional file 1A). Although these analyses are in general agreement with our measurements of a preferred width of YFP-MreB filaments of 75 nm in vivo [29], we do not believe that the observed structures are close matches of in vivo filaments, which have a much smoother appearance [45, 46]. Nevertheless, the observed structures appear to be filamentous, and thus disordered structures rather than aggregates of MreB monomers. We interpret our findings as consequences of MreB polymerizing away from the membrane, and of the absence of regulatory mechanisms and cellular interactors such as RodZ [47, 48] and EF-Tu [42].
MreB monomers have membrane affinity and form extended filaments after addition of magnesium or of calcium
We next moved to imaging of YFP-MreB by fluorescence microscopy (FM). MreB has been shown to form polymers on various surfaces, including mica [10, 49, 50] and membranes [11], to which MreB from B. subtilis and from E. coli have intrinsic affinity via an internal hydrophobic loop or via an amphipathic N-terminal helix respectively [34, 38]. MreB filaments from C. crescentus have been visualized on their natural interaction surface, where they form flat sheet-like structures, with double filaments being the smallest visible form of individual filaments [11]. Visualization of MreB from a Gram positive bacterium on membranes has been missing so far. We therefore adapted a planar lipid bilayer system devised by the Schwille group [51], such that the formation of MreB filaments can be visualized on a biological membrane. Addition of calcium (2 mM) to vesicles composed of lipids from E. coli cells (Fig. 3A) led to the formation of a planar membrane (Fig. 3B, C). The membrane was fluid as verified by FRAP analysis (Additional file 2A). The planar membrane could be stained with e.g. FM4-64, yielding homogeneous red fluorescence, and no fluorescence in the yellow channel (Fig. 3B). When monomeric YFP-MreB in low-salt polymerization buffer was added to the membrane, and the solution was subsequently washed with several volumes of polymerization buffer lacking YFP-MreB, a homogeneous staining of the membrane was observed (Fig. 3C). These experiments show that even non-polymerized YFP-MreB has membrane affinity, which strongly increases the local concentration of the protein at the cell membrane, and will enhance the efficiency of polymerization. In this case, 2D diffusion could be employed to find binding places at existing filaments, or to form nucleation centers. Addition of 10 mM magnesium to purified YFP-MreB (2 µM) induced the formation of a network of filaments that were attached to the membrane, and remained attached even after washing of the reaction chamber (Fig. 3D). Occasionally, the planar membrane contained holes, at which filaments were never observed (suppl. Figure 2B), revealing that MreB forms membrane-associated polymers in our experimental system. Induction of MreB filament formation led to a depletion of the homogeneous fluorescence at the membrane (Fig. 3D), indicating that membrane-attached MreB is efficiently incorporated into the filaments.
Using STED microscopy, we could visualize YFP-MreB filaments with a resolution of below 80 nm. Figure 3E reveals that filaments were not straight, but curved or even helical. Figure 3E shows that filaments could branch and fuse, or twist, similar to what was observed using electron microscopy. Therefore, YFP-MreB filaments observed by FM appear to match the structures seen by EM.
The extent of filament formation was dependent on MreB concentration. At or below 0.1 µM YFP-MreB, we did not observe any elongated filamentous structures; instead, only focal structures were observed (Fig. 3F). These were distinct from the uniform distribution of MreB in the absence of magnesium (Fig. 3C), so it is likely that the fluorescent foci correspond to small MreB assemblies that could serve as nucleation centers. A minimal concentration of 0.2 µM was required to generate visible filaments (Fig. 3G), whose number and length increased with rising concentration; at 1 µM YFP-MreB, filaments of a length of up to 6.3 µm could be detected at the membrane (Fig. 3H, suppl. Figure 3C). However, for many filaments, their ends were no longer detectable at the plane of the membrane, but extended away from the membrane. This became more pronounced with higher protein concentrations (Fig. 3I). Above 2 µM YFP-MreB, filaments were mostly present in form of a network, where filaments formed non-productive interactions due to branching. Even though the filaments extended away from the membrane they were still bound to it at the base of the structure: when we washed the reaction chamber with different buffers, the filamentous structures stayed attached to the membrane. This enabled us to change conditions during later experiments. As the intracellular concentration of MreB has been determined to be around 5 µM [13], our data suggest that intracellular conditions must exist, or regulators, which prevent the formation of non-productive filament interactions as seen in vitro.
We employed two strategies to investigate if the observed filaments are an artifact of the fluorescent protein fusion. Firstly, we added purified MreB to a concentration of YFP-MreB, which by itself does not lead to the formation of extended filaments. This approach also allows us to investigate if the architecture of filaments is altered when most of the structures were made up of non-tagged MreB. Addition of 10 µM MreB to 0.1 µM YFP-MreB and induction via magnesium or calcium resulted in the formation of highly extended filaments (Fig. 3L) that were visibly indistinguishable from those formed by 10 µM YFP-MreB by itself (Fig. 3J). Secondly, we incorporated a cysteine residue between the Strep-tag and the N-terminal residue of MreB; the N-terminus can be modified to carry a GFP and still be a functional fusion protein in vivo [29, 31]. Purified Cys-MreB was labelled with a fluorescein chromophore, and was added to the membrane, which led to the generation of filamentous structures after addition of magnesium (Fig. 3J); these structures had similar dimensions as YFP-MreB filaments. Additionally, when 0.1 µM YFP-MreB was mixed with non-stained Cys-MreB at 1 µM concentration, visible filaments arose that would not be seen at such a low concentration of YFP-MreB itself (Fig. 3K), suggesting that the stain does not influence the architecture of the filaments significantly. These experiments show that purified MreB can form extended filamentous structures on a flat membrane system in vitro, whose architecture is adequately reflected by YFP-MreB.
MreB filaments have an average width of 90 nm and frequently exhibit a curved architecture
We took advantage of the fact that at or below 1 µM concentration of MreB, filaments formed a less extensive network, and attempted to gain information on their length, with the caveat that we had to use the length of extension away from the cell membrane as an approximation for filament length. At 0.2 µM concentration, filaments could reach up to 2.3 µm, and had an average length of 1 µm, which increased to 1.5 µm at 0.5 µM, and to 1.85 µm at 1 µM YFP-MreB (Additional file 3A-C). At the latter concentration, filaments could extend to over 6 µM. Note that as will be explained below, monovalent ions strongly reduce filament size, and the above experiments were performed without the addition of K+. Nevertheless, they show that at given protein concentrations, MreB appears to polymerize into filaments whose length has a Gaussian rather than an arbitrary distribution.
When Z-stacks of membrane-polymerized MreB were captured, the filaments had the appearance of helical structures (Fig. 4A, Additional movie). Similar to MreB, CFP-Mbl filaments frequently showed curved and helical appearance (Fig. 4C). Extended and curved filaments were also observed under high potassium (100 mM) concentrations (Fig. 4D).
We measured the width of YFP-MreB filaments, using STED microscopy. For smaller filaments (< 1 µm), the width was 91.7 ± 35 nm (n = 100), suggesting that the structures contain many MreB protofilaments, because the width of an individual MreB double protofilament would be about 8 to 9 nm. For larger filaments, the average width was determined to be 178 ± 95 nm (n = 100). In vivo, filament width was determined as 75 nm [29]. Our data have to be viewed with caution, keeping in mind that the observed networks of filaments are visibly quite dissimilar from structures formed in vivo.
Divalent cations promote filament formation, while monovalent ion have an adverse effect
MreB filament formation has been shown to be affected by ion concentrations in sedimentation and light scattering experiments [39, 40], which have been widely used to analyze polymerization of actin and actin-like proteins [52, 53]. Addition of increasing concentrations of magnesium to purified MreB resulted in rapid increase in light scattering (Additional file 4A), indicative of rapid polymerization into filaments. Addition of potassium to the reaction strongly decreased the amount of scattering (Additional file 4B). We tested different ions to visualize the effect on the polymerization of MreB, employing 2 µM monomeric YFP-MreB in low-salt polymerization buffer. The addition of different concentrations of magnesium or of calcium visibly affected the formation of YFP-MreB filaments nucleating at the membrane (Fig. 5A). The amount of filaments increased in a magnesium dose-dependent manner, with visible saturation occurring at 10 mM magnesium (Fig. 5A). Additionally, the amount of protein used had a pronounced effect on the filament architecture (Fig. 5A). Using 10 mM MgCl2 and 5 µM YFP-MreB resulted in a level of filament formation that led to extended networks, similar to that of 10 mM CaCl2 and 5 µM YFP-MreB (Fig. 5A), showing that MreB filaments respond to both divalent ions in a similar fashion. Addition of calcium showed visibly indistinguishable degrees of filament formation (Fig. 5A) compared with magnesium. Therefore, membrane-associated MreB reacts to both divalent ions in a similar manner as MreB in solution.
Monovalent ions have been described to have a negative effect on the polymerization of MreB [40]. We used 10 mM MgCl2 to induce filament formation of MreB, in a solution containing different amounts of KCl. Employing the membrane system, we also observed an inhibitory effect of potassium on filament formation. Considerable inhibition was seen starting at concentrations of 50 mM KCl (Fig. 5B), and at 100 mM concentration, only short MreB filaments were visible by light microscopy (Fig. 5B). At 300 mM KCl, mostly foci and few filaments of YFP-MreB were detectable, indicating that only small assemblies of MreB exist at this concentration, but no longer extended filaments (Fig. 5B, see Additional file 5 for a quantitative analysis). Importantly, starting at 50 mM K+ concentration, no more branched filaments were observed (Fig. 5B), revealing that potassium counteracts non-productive filament interactions of MreB. The effect of 50, 100 or 300 mM NaCl was very similar to that of KCl (data not shown). Thus, monovalent ions were effective in their inhibitory activity at roughly 10 fold higher concentrations than divalent ions. In the cell, the higher concentration of potassium compared to magnesium and calcium will therefore reduce the length of MreB filaments, and counteract filament branching and thus non-productive meshwork formation.
We also tested if monovalent cations can induce the dissociation of preformed MreB filaments. The addition of 50, 100 or of 300 mM KCl did not show any effect on preformed YFP-MreB filaments (data not shown), revealing that once formed, monovalent cations no longer show an effect of MreB filaments, possibly, because their putative specific binding sites are now buried within the subunit interaction surface.
MreB, Mbl and MreBH form mixed polymers that can laterally associate to preexisting filaments
We wished to gain further insight into the architecture of MreB filaments and to study the relation of the three MreB paralogs in vitro.
All three protein fusions exhibited affinity to the planar membrane (Fig. 6A). After inducing polymerization with 10 mM MgCl2, CFP-Mbl and mCherry-MreBH formed visible nucleation foci at 0.1 µM, and visible filamentous structures at 0.5 µM, whose length and number on the surface area increased with increasing protein concentration (Fig. 5C, Additional file 6). At 15 µM concentration, mCherry-MreBH formed filaments to a lesser degree than YFP-MreB or CFP-Mbl (Fig. 5C), but in general, all three MreB paralogs behaved very similarly with regard to polymerization on the membrane.
To assay any direct interaction between the paralogs, we mixed 1 µM of each, YFP-MreB, CFP-Mbl and mCherry-MreBH, which individually form only short filaments (Fig. 3H, Fig. 5C). When mixed, the joint formation of ion-induced, extended filaments was observed (Fig. 6D), showing that MreB paralogs cooperatively form single polymeric structures, in an additive manner. YFP-MreB also formed co-polymers with CFP-Mbl or mCherry-MreBH alone (Fig. 6C and data not shown), and likewise did CFP-Mbl and mCherry-MreBH (Fig. 6B).
The 100% overlay exemplified in Fig. 6D suggests that of all three proteins are closely associated within the joint molecules, when present at the initial stage of polymerization. These findings raise the interesting questions if MreB paralogs can associate with a preformed polymer, and if mixed polymers form by the addition of monomers to the ends of preexisting filaments, or in a lateral manner, or both. In the first case, we would expect that preformed filaments of one colour would contain extensions of another colour, in the latter case, preexisting filaments would be labeled with a second colour along their entire length. Figure 6E shows that CFP-Mbl was able to assemble at sites of previously formed YFP-MreB filaments, in addition to its independent filament formation. Likewise, mCherry-MreBH formed filaments along the length of pre-existing YFP-MreB filaments (data not shown), and also at preexisting Mbl filaments (Fig. 6F), while YFP-MreB could also laterally attach to preformed CFP-Mbl filaments (data not shown). Of note, all co-polymers had an identical appearance, but we did not observe that an end of a co-filament had only a single colour (i.e. that of the lastly added paralog). These data show that filaments formed by one MreB paralog can be laterally extended by a second and third paralog, but are not extended at the end to a detectable degree, supporting the formation of lateral sheets of filaments.
Mreb Forms Filaments Form Between Multilayered Vesicles
MreB filaments can also form in solution, i.e. independent of a supporting bilayer. We wished to investigate if the proximity of a membrane favors the formation of filaments over that of in solution. We therefore generated lipid vesicles in the presence of non-polymerized YFP-MreB, such that frequently, multilayered vesicles would form. YFP-MreB was added to a planar membrane within a reaction chamber, additional lipid vesicles were added, and the mixture was removed from the chamber, vortexed and imaged after addition of calcium. This way, added vesicles (lacking YFP-MreB) were encircled by larger vesicles derived from the planar membrane and by non-polymerized YFP-MreB, which was then induced to form filaments. We observed the formation of YFP-MreB filaments at vesicle interfaces (Fig. 7A), especially where two vesicles were tightly packed (Fig. 7B). Interestingly, YFP-MreB filaments extended between the two juxtaposed biological membranes, and bound to both membranes, in an apparently helical pattern (Fig. 7C). These experiments show that the presence of a membrane, and especially two neighboring membranes favours the formation of filaments that are curved, similarly as observed on a planar membrane system, where filaments will extend away from the membrane.