Creation of an mreB alanine-scanning mutagenesis library

MreB is an important protein in the regulation of cell shape. While multiple studies have been performed to understand how MreB functions, we still have little idea of how MreB regulates cell shape and the role that individual residues play in this regulation. Past experiments have used antibiotics or plasmid-based systems to make random MreB point mutations [29,30,34,46,47,50,55,56,58,59,63–65]. These experiments have shown that different changes at a single amino acid site can tune cell size; however, there has been a lack of a systematic approach to understanding the role of each amino acid residue in cell shape regulation.

To systematically understand how MreB functions, we constructed a native-site GFP-tagged alanine-scanning mutagenesis library of mreB via lambda red recombination. All amino acids were individually changed to an alanine except for the native alanines, which were changed to leucine to maintain similar chemical properties and avoid adding flexibility with a glycine, and the initiating methionine, which was left unchanged. The construct used to make the gene replacements has msfGFP fused to mreB [46] in an internal non-conserved loop (sandwich fusion) with flanking regions of homology on both sides of mreB and a kanamycin resistance cassette (S1A Fig). The strains with mutations downstream of the GFP were sequenced to ensure the desired recombination occurred. Out of the 346 amino acids of MreB, we were unable to make nine amino acid substitutions (S1 Table and S1B Fig) leaving us with a library of 337 mutants. Only two out of nine mutations that we could not make occur in suspected self-interaction regions of MreB which were thought to be important for MreB assembly and function. Our inability to successfully transform strains with the mutant mreB alleles suggests these residues are important for viability. In support of this, these residues are highly conserved among both Gram-negative and Gram-positive species (S1 Table).

E. coli with mreB mutations can stably grow as spheres

To determine the role of each residue of mreB on cell shape we imaged the entire library grown in LB at 37^oC arrayed in 96-well plates with a GFP-tagged wild-type (WT) strain in each plate and fixed the cells before imaging. A ΔrodZ mutant was included as a control for a non-rod shape phenotype [17,27]. Surprisingly, this initial screening revealed a wide variety of cell shape phenotypes, with some mutants having disrupted shapes similar to ΔrodZ cells.

Because MreB is known to be involved in rod shape determination we expected many mutations to affect rod shape. However, based on past experiments we did not expect to find a complete loss of rod shape as this would suggest MreB is no longer functional and should therefore be lethal. To quantify how rod-like cells are, we measured the coefficient of variation of the intracellular diameter deviation (IDD) of all cells in the library. This metric measures the standard deviation of the cell diameter across the long axis of the cell [26]. The more spherical a cell is, the larger the IDD, as a rod will only have changes to its diameter at the poles. Due to the number of mutants observed we grouped cells into a histogram based on IDD values. Most of the MreB mutants have little to no change in rod shape (left side of the histogram); however, we observed mutants with shape defects beyond the rod-like spectrum (Fig 1A-1B). From left to right the WT replicates appear in bins 1, 2, 3 and 5, with an average IDD of 0.3. The cells that have the least rod-like shape (the last four bins in Fig 1A) were reimaged using live cell imaging (Figs 1, S2A, and S3B). The length of a rod will determine what percent of the cell consists of the poles effecting the IDD value (S3A Fig); therefore, to ensure our IDD measurements were not being skewed by cell length we measured the aspect ratio of the cells (long axis:short axis) in the last four bins from the IDD histogram and found that the average aspect ratio for each bin was 3.03 ± 0.43, 2.70 ± 0.47, 2.67 ± 0.45 and 2.14 ± 0.12 respectively. Using an aspect ratio of 3 as a cutoff, the last three bins (black and gray bars) will be considered non-rod-like and excluded from future cell shape analysis when applicable. Interestingly, the ∆rodZ mutant has the highest IDD value in the penultimate bin (0.4) indicating that the strains in the black bin have completely lost rod shape (IDD range from 0.4-0.42).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Alanine-scanning mutagenesis library shows non-rod shape phenotypes.

A) Histogram of 337 point-mutants (n > 100 cells for each strain) based on the coefficient of variation of intracellular diameter deviation (IDD). The black bar indicates the 18 mutants with the highest IDD, while the gray bars indicate the mutants considered to have lost rod shape. B-C) Live cell imaging and quantification of WT and the eighteen mutants (n > 600 cells for each strain) from the black bin with extreme shape defects. Data is pooled from three independent experiments. B) Coefficient of Intracellular Diameter Deviation IDD) measurements of cells from live imaging. C) Average cell area of 18 round mutants of cells from live cell imaging. D) Representative phase contrast and fluorescence images of representative non-rod-shaped mutants and WT (see S2 Fig for all 18 mutants). Single arrowheads point to the cells with MreB in foci, circular heads point to cells with cytoplasmic MreB. E) The MreB-GFP fluorescence intensity values of eighteen mutants and the WT. Error bars are the standard deviations from three independent experiments. Tukey’s multiple comparison test was used to determine statistical significance (ns = p > 0.05, * =<0.05, ** = p < 0.01, *** = p < 0.001, **** = p < .0001).

https://doi.org/10.1371/journal.pgen.1012070.g001

We further analyzed the non-rod cells by measuring their growth rates. As we expected all the strains from the last bin grow slower than WT cells (S2A and S4A Figs), while many of the strains from the other two bins show reduced growth rates (S3B and S4B Figs). The fact that we have spherical cells that grow stably at 37^oC in rich media suggests that MreB has two functions: 1) rod shape determination and 2) viability. If this is true than there should be little correlation between growth rate and IDD. We found an R² value of only 0.4 (S4C Fig) suggesting that growth rate and shape are uncoupled.

The loss of rod-shape maybe due to a destabilizing effect of the GFP sandwich fusion rather than the point mutation itself. We made five of the mreB mutations without a GFP tagged mreB and observed a similar spherical phenotype as the strains with GFP (S5A Fig), suggesting that the shape phenotype is due to the change in the MreB amino acid residue and not the addition of GFP.

Spherical cells containing mreB point mutations do not have known suppressor mutations

Because it was unexpected to be able to stably grow spherical E. coli with a non-functional MreB, we performed whole genome sequencing (WGS) to check for the presence of known suppressor mutations. We chose ten of the spherical mutants for WGS and found no mutations in the PG synthesis machinery (elongasome and divisome components), their known regulators, nor their promoter regions. However, we did find that our parental MG1655 WT strain used to make the library contains a point mutation in the bifunctional ppGpp hydrolase/synthetase, spoT (spoT_A26E), while the WT fluorescent strain (NO50) does not (S2 Table). SpoT is the major hydrolase for ppGpp with weak synthetase activity [66,67]. The levels of ppGpp have been shown to regulate cell size and FtsZ levels and may be linked to mecillinam resistance in E. coli [68–75]. While this mutation is outside of the known synthetase activity region (amino acids 67–374) it falls within the hydrolase region and may be affecting intracellular levels of ppGpp and leading to increased levels of ftsZAQ [76]. To test if this mutation affects FtsZ levels, we performed Western blot analysis on a strain with and without the spoT_A26E mutation and a strain expressing ftsZAQ from a constitutive promoter on a plasmid to compare FtsZ protein levels (S6A Fig). The spoT mutation does not lead to an increase in FtsZ levels while ectopic expression of ftsZAQ results in a 2.5-fold increase. This suggests that changes in FtsZ levels due to a spoT mutation are not supporting growth in the spherical mutants.

To further confirm that the spoT mutation was not responsible for the viability of the round mutants, we moved seven of the mreB mutations into either an MG1655 background or an MC4100 background which do not have the spoT_A26E mutation. The introduction of mreB mutations into both backgrounds leads to stably growing spherical cells (S5 Fig), further supporting the Western blot results that the spoT_A26E mutation is not responsible for stable spherical growth.

Increased levels of FtsZ are not necessary for survival in spherical cells harboring mreB mutations

It is surprising that E. coli can grow without an apparently functioning MreB. We therefore tested whether the spherical mutants over produce FtsZ. Western blots performed on the 18 most spherical mutants (black bin Fig 1A) show that half of the mutants do not exhibit a statistical difference in FtsZ levels compared to WT cells (Table 1 and S6B Fig). It is interesting that some of the mutants do overproduce FtsZ, although most are below the 2.5-fold increase seen with the overexpression plasmid, which could suggest a feedback mechanism between MreB or cell shape and ftsZ expression. In total, these data suggests that E. coli can grow as spherical cells without the overexpression of ftsZ.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Relative FtsZ levels in round mutants. Statistical significance based on Dunnett’s multiple comparison test (*** = p < 0.001, ** = p < 0.01, * = p < 0.05 and ns = p > 0.05; cells without asterisks). ^aAll mutants were compared to the WT strain. Fold-change values are the average of three independent experiments (Mean ± std).

https://doi.org/10.1371/journal.pgen.1012070.t001

Spherical strains with mreB mutations are produced through distinct mechanisms

A point mutation in mreB could lead to loss of rod shape due to changes in the levels of MreB, loss of interaction with an elongasome component, loss of self-interactions, changes to polymer characteristics, changes to ATP binding/hydrolysis, or the ability to bind the membrane. To further characterize the shapes of the round mutants, we measured the cell area (Fig 1C and S3 Table). While WT cells show a tightly regulated cell size, the spherical mutants exhibit a wide range of mean cell area and a considerable population variation. This suggests that the loss of rod shape may be happening for different reasons and that these mutations lead to cells being unable to regulate cell size.

As the entire library was made with GFP-tagged MreB, we used the fluorescence intensity to measure MreB levels in the round mutants. Two mutants do not show any difference compared to WT cells, while other mutants have either more or less MreB (Figs 1D and S2B). This further supports the hypothesis that rod shape is lost through different mechanisms and shows that rod shape can be lost without altering MreB levels. Additionally, these data reinforce MreB having distinct roles for shape regulation and viability as mutants that express lower levels of MreB may not have enough MreB to regulate rod shape but must have a sufficient MreB concentration to support growth.

In addition to measuring fluorescence intensity, we observed that the round mutants have different MreB localization patterns. WT MreB forms multiple foci across the cell body. We hypothesize that a mutation that inhibits monomer-monomer interactions, protofilament interactions, or membrane binding will result in cytoplasmic localization of MreB, which is seen in three mutants (Figs 1E and S2A, circle head). However, neither of these three have mutations in the predicted monomer-monomer interface, protofilament interface, or membrane interacting amphipathic helix [29,77]. It is possible that a single alanine substitution is not sufficient to disrupt polymer formation. The WT localization phenotype is seen in multiple mutants suggesting that polymerization is not disrupted in these strains (Figs 1B and S2A arrowhead), of which two, R65A and V181A lie in the predicted monomer-monomer interface. The presence of foci suggests the MreB in these mutants has lost the ability to interact with an essential partner resulting in filaments (foci) but not rod shape. Alternatively, foci could form if mutations lead to altered biophysical characteristics of the polymers so that they no longer function properly. It is unclear if polymerization can occur in the absence of membrane binding. These data show that MreB’s ability to form visible foci is not the sole factor on whether it is able to regulate rod shape.

MreB residues involved in shape and viability are highly conserved

We compared the conservation of the residues that lead to the 27 non-rod-shaped (S1C Fig) mutants in a variety of Gram-negative and positive organisms. 69.4% of the residues that result in nonviable or misshapen cells (25/36) are identical among the tested species compared to only 36.3% of the total proteins (S1 Table). The increased conservation of these specific residues over the total protein further supports the conclusion that they are important for MreB functions.

Cell width is tightly controlled by MreB

MreB has been linked to cell width or rod diameter regulation in C. crescentus and E. coli [46,47,78,79]. To determine which residues are important for width regulation we measured the cell width of all the mutants (S1 Data). Due to plate-to-plate variances in cell growth and size, we calculated the foldchange of the cell width of each mutant to the WT strain in each individual plate and represented the data as a histogram (Fig 2A). 60 mutants represented by the black bar (Fig 2A) have a low fold change indicating these mutants are thinner than WT. We performed live cell imaging on the 20 mutants with the smallest ratio at early exponential phase (OD 0.2-0.25). 19 (95%) are significantly thinner than WT cells with a decrease of cell width by 6.10% to 22.54% (p < 0.0001) with respect to WT cells indicating that our initial fixed-cell screen accurately measures thinner cells (Fig 2B only shows the 19 that were thinner). Previous results found a weak correlation between the width of cells and the coefficient of variation of width; therefore, we measured the coefficient of variation of width (CV_width) of these 19 thin mutants and found that almost all of them (18/19) have a smaller coefficient than WT cells (Fig 2B) suggesting that as cells get thinner, they have a more consistent width across the population [56].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Cell width is tightly controlled by MreB.

A) Histogram representing the cell width fold change of 337 point-mutants (n > 100 cells for each strain) compared to WT from fixed cell imaging. The black bar represents the thinnest mutants, and the gray bars represent the widest mutants. B) Left – The 20 thinnest mutants underwent live cell imaging. The mean cell width of the significantly thinner mutants (19/20) (black bar from A) and WT cells (n > 400 cells for each strain, p < 0.0001). Error bars are 95% CI. The secondary Y-axis represents the coefficient of variation of width among the respective thin mutants and WT. Data is pooled from three independent experiments. Right- Phase contrast images of the thin mutants and WT. C) Left- Average cell width of wider mutants (gray bars from A) and WT (n > 600 cells for each strain, p < 0.0001). Error bars are 95% CI. The secondary Y-axis represents the coefficient of variation of width among the wider mutants and the WT. Right- Phase contrast images of the wide mutants and WT. B-C) Data is pooled from three independent experiments. Tukey’s multiple comparison test was used to determine statistical significance (ns = p > 0.05, * = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < .0001).D) Scatter plot of the coefficient of variation of cell width versus average cell width of the entire fixed cell library. Orange dots represent the round mutants, and black dots represent WT cells. r = Pearson’s coefficient.

https://doi.org/10.1371/journal.pgen.1012070.g002

Similarly, 40 strains distributed over six bins (grey) (Fig 2A) have a high foldchange indicating they may be wider than WT cells. These bins include the round mutants with higher IDDs that we are excluding from this analysis and ΔrodZ (bin 7) (Fig 1A). Live cell imaging was performed on the 13 remaining strains from these upper bins at an early exponential phase (OD 0.2-0.25) and all were found to be wider than WT (p < 0.0001) in agreement with the initial screen (Fig 2C). Therefore, when including the round mutants, all 40 strains are wider than WT suggesting that the initial screen can accurately measure cells with larger cell widths. We measured the CV_width and found that all 13 wider cells have a larger CV_width (Fig 2C) suggesting that the width of large cells varies more across the population. Since both the thinnest and widest cells seem to correlate with the CV_width we compared the CV_width to width across the entire fixed cell library. In contrast to previous results, this analysis found a high correlation, implying that as cell width increases, there is more variation in the width across the population (Pearson’s r = 0.8290) (Fig 2D).

We performed a second analysis on the cell width by calculating the width foldchange against the plate-wise average cell width to account for intraplate variation. WT cells cluster in bins 2–4 while ∆rodZ cells are in bin 7. Among the 63 thinnest mutants comprising the black bins (bins 1–2 S7A Fig), 38 overlapped with the previously calculated thin group (Fig 2). We quantified the width of cells after live-cell imaging for the four mutants with the smallest ratio among the remaining strains unique to this analysis. Three of the four mutants show a significant decrease of 5.75% to 16.81% (p < 0.0001, p > 0.05 (V187A)) of cell width compared to WT cells (S7B Fig). Similarly, 60 mutants in the six dark grey bins (S7A Fig) have a large ratio, suggesting they may be wider than the plate average. After excluding the spherical strains, ΔrodZ, and overlaps from the initial analysis, we were left with 16 new strains. We imaged eight of these mutants and found that all of them were significantly wider (p < 0.0001) than WT cells showing a total increase of 10.5% to 35.88% increase in cell width (S7C Fig). Between these two methods, we have demonstrated that large-scale imaging of fixed cells accurately represents changes in cell width in both directions, as previously shown [80].

mreB mutations can affect cell length

The activity of the elongasome adds cell length during the pre-division stage of the cell cycle. The length of E. coli has been linked to division time and hence the activity of the division machinery (divisome) [81,82]. It is unclear how MreB interacts with the divisome, although in C. crescentus MreB has been shown to condense to the division plane and in E. coli some labs have reported an increase of MreB at the division plane [83,84]. In E. coli, RodZ localizes at division sites in an FtsZ-dependent manner and RodZ also interacts with MreB; however, many other labs fail to see MreB localization at the division site in E. coli [27,33,83]. Additionally, MreB can directly interact with FtsZ, which is needed to shift from elongation to division in some bacteria [38,85,86]. Hence, there may be transient crosstalk between these two complexes. If MreB is involved in regulating division, one would expect to see mutants that are affected for cell length and possibly filamentous cells in our library.

As we did for cell width, we measured the cell length foldchange of the mutants to the WT from each plate and the plate-wise average and presented the data as histograms (Fig 3A and 3B). The first analysis yields 66 mutants (black bar, Fig 3A) that have a smaller ratio than WT cells suggesting they may be shorter than WT. We quantified the cell length of the 27 mutants with the smallest ratio from this group using live cell imaging and found 21 (77.78%) show significantly decreased cell length by 5.33% to 27.33% (p < 0.0001) with most of the strains showing a smaller coefficient of variation of length (CV_length) than WT (Fig 3C), suggesting these point mutants have more control over their cell length than WT. Similarly, 16 mutants (grey bars, Fig 3A) have a high fold change, indicating they may be longer than WT, with the ΔrodZ control appearing in bin 4. Seven of these were excluded from future analysis because they are spherical, leaving only 9 high ratio mutants. To further test our accuracy on measuring long cells, we looked at the next 4 rod-shaped strains with the largest foldchange. Of these 13 mutants, 5 (29.4%) were significantly longer than WT showing a 3.95% to 35.56% increase in average cell length with a higher CV_length than WT after live cell imaging (Fig 3D), although none formed filamentous cells.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Cell length is minimally affected by MreB mutations.

A) Histogram representing the cell width fold change of 337 point-mutants (n > 100 cells for each strain) compared to WT from fixed cell imaging. The black bar represents the shortest mutants, and the gray bars represent the longest. B) Histogram representing the cell width fold change of 337 point-mutants (n > 100 cells for each strain) compared to the plate-wise average from fixed cell imaging. The black bar represents the shortest mutants, and the gray bars represent the longest. C) Left-Average cell length of the twenty-one shortest mutants (black bar from A) and WT cells (n > 400 cells for each strain, p < 0.0001, S48A and N252A p < 0.001). The secondary Y-axis shows the coefficient of variation of length. Right-Phase contrast image of cells of the shortest mutants. D) Left- Average cell length of the five longest mutants (gray bars from A) and WT cells (n > 400 cells for each strain, p < 0.0001). The secondary Y-axis shows the coefficient of variation of length. Right- Phase contrast image of cells of the longest mutants from (A). E) Average cell length of the three longest mutants (gray bars from B) and WT cells (n > 400 cells for each strain, p < 0.0001, L11A p < 0.05). The secondary Y-axis shows the coefficient of variation of length. Left- Phase contrast image of cells of the shortest mutants. F) Scatter plot of the coefficient of variation of cell length versus average cell length of the entire fixed cell library. Orange dots represent the round mutants, and black dots represent WT cells. r = Pearson’s coefficient. G) Scatter plot of the average cell length versus average cell width of the entire fixed cell library. Orange dots represent the round mutants, and black dots represent WT cells. r = Pearson’s coefficient. Dunnett’s multiple comparison test was used to determine statistical significance.

https://doi.org/10.1371/journal.pgen.1012070.g003

We calculated the foldchange against the plate-wise average to account for the intraplate variances. WT cells show a large distribution of length ratios appearing in bins 3–7. As this analysis is based on the plate-wise average of cell length it is possible that different plates have a larger number of short or long cells affecting where the WT cells fall. The rodZ mutant is in bin 4. Among the 12 shortest mutants (black bar Fig 3B), 11 overlapped with the previous short mutant list, leaving only one mutant; therefore, we didn’t examine short mutants from this analysis. Among the 19 longest mutants from the four bins (grey bars, Fig 3B), 6 were round mutants and three overlapped with the list from the previous analysis, leaving one WT and 10 possibly long mutants. Among the ten, 3 (30%) mutants were significantly longer showing 7.30% to 26.04% increase in average cell length with higher variability than WT (Fig 3E) after live cell imaging. Collectively, these results suggest longer mutants have less control over regulating cell length. We measured the correlation between the cell length and the CV_length in the entire fixed cell library, revealing only a weak correlation (Fig 3F). While the initial large-scale screen and live imaging both accurately report on cell length, it is not as robust as measuring cell width (S7 Fig). Similarly, there is only a mild correlation between cell length and width as we do not observe long and thin mutants (Fig 3G). Additionally, the fact that we do not observe any filamentous cells suggests the effects of MreB on FtsZ function is minimal.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Round mutants do not require elongasome activity for PG synthesis.

A-B) Cells were grown for six hours in LB or LB + A22 (10 µg/mL). The ratio of the OD₆₀₀ between the two conditions was determined as a growth ratio. A) A22 growth ratios of the 18 mutants with the most extreme cell shape defects. B) Growth ratios of the rod-shaped suppressors and parental strains. C) Minimum inhibitory concentration to mecillinam of the suppressor mutants and parental strains. All data shown are the average values from three independent experiments with standard deviation.

https://doi.org/10.1371/journal.pgen.1012070.g004

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Deletion of PBP1A or PBP1B is tolerated in most round mutants.

A) Spot assays of ΔmrcB round mutants and WT. B) Spot assays of ΔmrcA round mutants and WT. Overnight cultures were diluted to an equal number (0.1OD₆₀₀) and serially ten-fold diluted, and 4μL suspensions were spot inoculated into Lennox LB agar plates (5g/L NaCl). Plates were incubated for 16h at 37°C.

https://doi.org/10.1371/journal.pgen.1012070.g005

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Only intragenic mutations in mreB can suppress the round shape phenotype.

A) Left-Representative phase contrast of suppressor mutants, WT, and parental single mutant strains grown to exponential phase. The numbers represent the average max doubling time (mean (min) ± std). Right- The parent mutant P108A and suppressor sites F84V and G168A are mapped on an E. coli MreB homology model. B) IDD measurements of WT and mutants (n > 700 cells for each strain). C) Representative growth curves of WT (black line), suppressors, and parental strains. Cells were grown overnight in LB and sub-cultured in equal numbers into a 96-well plate with ten replicates for each strain. Error bars are standard deviations from those 10 replicates.

https://doi.org/10.1371/journal.pgen.1012070.g006

Spherical strains with mreB mutations no longer use the elongasome

Round E. coli, such as when rodZ is deleted or our spherical mutants, must have a cell wall, as wall less cells (L-forms) would not be able to grow in these conditions. It is unclear how spherical E. coli build their wall. ∆rodZ cells have been hypothesized to be round due to the mislocalization of MreB and therefore the elongasome, while A22-treated cells are thought to lack elongasome activity altogether due to the depolymerization of MreB. As we can stably grow spherical mreB mutants, we hypothesize that they must be producing a cell wall in an MreB-independent manner. To confirm MreB is no longer active in these round mutants we grew the 18 spherical strains with and without A22 (10 µg/mL) for six hours and determined the growth ratio between A22 and LB only conditions. All round mutants grow better than the WT strain in A22 indicating that these mutants do not require a functional MreB to build their cell wall during growth (Fig 4A). To ensure that the construction of the library did not lead to A22 resistance for all strains we randomly chose three rod shaped mutants to test for A22 sensitivity (Fig 4A). These three strains show the same sensitivity to WT cells supporting the idea that cell wall synthesis is different in the spherical mutants.

The A22 resistance of the spherical mutants suggests the elongasome is no longer functioning. To further test this we examined the MIC of these mutants to mecillinam, which targets the elongasome component PBP2. All round mutants are resistant to mecillinam by > 30-fold over WT cells clearly suggesting the elongasome is no longer necessary in these strains (Table 2).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. MIC of 18 round mutants against different antibiotics targeting PBPs.

https://doi.org/10.1371/journal.pgen.1012070.t002

To further test if the elongasome is functioning in these mutants we moved the 18 spherical causing mutations into a PBP2 temperature sensitive (PBP2_ts) mutant background [87]. As expected, when grown in the non-permissive temperature (42^oC) WT cells show a severe growth reduction (S9 Fig). Almost all of the mreB mutants (13/18) grow better than WT cells at 42^oC and some even grow better at 42^oC than they do at the permissive temperature (30^oC). This indicates the spherical mutants can grow without a functional PBP2 supporting the mecillinam data. Interestingly, four of the mutants (S37A, L91A, A225L, I198A) show an increase in growth when PBP2 is not functional, suggesting that PBP2 function may be inhibiting growth in these strains.

RodA is the transglycosylase partner of PBP2 [16,88,89]. We attempted to move the mutations into a RodA_ts strain, but were unable to make all 18, suggesting that some of the mreB mutations make cells more sensitive to disruption of RodA. In addition, three of the mutants (I198A, M90A, A225L) we were able to construct show no growth at 42^oC. However, similar to the PBP2_ts strains, many of the RodA_ts strains grow better when paired with a sphere causing mreB mutation (S10 Fig). The elongasome accounts for most of the synthesis of new PG strands during cylindrical growth; however, there are other components like the aPBPs, and the divisome that can contribute to the upkeep of the PG layer. These results can be explained if a different PG synthesis machinery is being used in these strains although why these mutants are more sensitive to disruption of RodA than PBP2 is unclear.

aPBPs are not responsible for growth in the spherical mutants

The aPBPs are bifunctional cell wall synthases that have been shown to work semi-autonomously from the elongation or division complexes [90–92]. The major aPBPs in E. coli are PBP1A and PBP1B which can be deleted individually but not together. These proteins have been shown to produce glycan strands in the absence of MreB or other cytoskeletal proteins yet cannot normally support growth alone as deletion of mreB, rodA, or mrdA (PBP2) is lethal [50,51,87].

To determine if these aPBPs or the divisome substitute for the elongasome in the spherical mutants, we measured their MICs against PBP-targeting antibiotics: cefsulodin (aPBPs), ampicillin (multiple PBPs), and cephalexin (PBP3). All the round mutants show increased susceptibility to cefsulodin and cephalexin, except for I15A, while all are more sensitive to ampicillin compared to WT (Table 2). These results imply that other PBPs may have a greater role in PG synthesis independent of MreB or the elongasome in these spherical mutants.

To further test if the aPBPs can compensate for the loss of MreB activity we attempted to move our mreB mutations into a ∆mrcB (PBP1B) or ∆mrcA (PBP1A) background. If either PBP1A or PBP1B is compensating for an impaired elongasome we would expect the deletion of mrcB or mrcA to be lethal in this background. We moved all 18 spherical mutations into the ∆mrcB strain and performed spot assays to assess their growth. The ability to make these strains shows that PBP1B is not the sole cell wall synthesizer in these strains. Interestingly, while there are a few spherical mutants that grow worse in the ∆mrcB strain, most strains show no difference in growth in the presence or absence of PBP1B and some grow better when PBP1B is absent (Fig 5A). We moved 11 of the mutants into a ∆mrcA background and saw similar results as with deletion of mrcB (Fig 5B). These results suggest that either both PBP1A and PBP1B are needed for cell wall synthesis in these strains or that there is a different PG synthesis system working. It is possible that FtsZ is driving this synthesis as the spherical mutants are sensitive to the inhibition of PBP3 which is part of the division machinery; however, there may also be a novel mechanism for PG upkeep that is independent of both MreB and FtsZ.

MreB is absolutely necessary for rod shape

When the main proteins of the elongasome are inhibited, cells become spherical, suggesting each of these proteins is important for rod shape. Previous studies using the spherical rodZ mutant identified suppressor mutations in mreB and other elongasome proteins that rescue rod shape, indicating that RodZ is not necessary for rod shape [18,26,28]. Due to the essential nature of MreB, it has been difficult to do similar experiments to test if MreB is required for rod shape. We found 18 mutants that stably grow as spheres, allowing us to test the hypothesis that, unlike RodZ, MreB is required for rod shape.

To determine if MreB is in fact necessary for rod shape determination, we performed a suppressor screen on six of the spherical mutants to see if rod shape can be restored. We used a sublethal concentration of the PBP3 targeting antibiotic cephalexin to select for rod-shape suppressors [18]. We found rod-shaped mutants for four of the round-shaped parents: I15A, M90A, P108A, and V181A (Table 3). Sequencing of mreB from a selection of the rod-shaped mutants revealed that almost all strains have reversions back to the WT mreB sequence. The two mutants of the P108A spherical parent have intragenic mutations in mreB: F84V and G168A (Fig 6A). To further confirm that these mutations are the cause of the restoration of rod shape we transduced the mreB double mutants into a clean WT background. Transductants result in rod-shaped cells. Additionally, we performed WGS on the two original suppressor strains and did not find any additional mutations. Together these results suggest the suppression comes from this second mreB mutation, indicating that MreB is absolutely required for rod shape.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Suppressor screen of round mutants with sub-lethal level of cephalexin.

https://doi.org/10.1371/journal.pgen.1012070.t003

To quantify the suppressive effects of the F84V and G168A mutations we measured the IDD of both single and double mutants compared to WT (Fig 6B). We found that G168A has a minimal effect on rod shape while F84V forms a misshapen rod, as previously reported [46]. Both double mutant strains show a decrease in IDD, but G168A restores the rod shape to P108A better than F84V, most likely due to the shape of the G168A parent being more rod-like than F84V. As the spherical P108A was shown to have a growth defect (S2A Fig), we also measured the growth rate of the suppressors and found that they have faster doubling times than the parent P108A strain (Fig 6A and 6C).

As the spherical P108A parental strain is resistant to both A22 and mecillinam we wanted to determine if the restoration of rod shape through the suppressor mutations restores sensitivity to these antibiotics. Both the G168A and F84V mutations lead to A22 resistance (Fig 4B); therefore, the double mutants are also resistant. While both G168A and F84V mutations lead to a slight increase in resistance to mecillinam over WT cells (Fig 4C), these strains are more sensitive than the parent P108A. The P108A G168A (rod shape) double mutant has a sensitivity to mecillinam similar to G168A, suggesting that elongasome activity may be restored in this strain. However, the P108A F84V strain has the same resistance to mecillinam as P108A alone, suggesting that the elongasome is still non-functional. Collectively, all these observations suggest mreB is central in determining and maintaining rod shape and cannot be bypassed.

MreB is an actin homolog

The actin superfamily is defined by conserved structural motifs needed for ATP binding/hydrolysis and polymerization. MreB has all of the structural features one would expect in an actin, including the ability to bind ATP with the nucleotide binding pocket or cleft between domains I and II [29,41,77,93,94].

The different nucleotide states modulate actin structural transitions by affecting filament dynamics and have been shown to be important in MreB polymerization in molecular dynamic simulations [95]. Mutations in the ATP hydrolytic or phosphate interacting motifs of actin lead to cells with poor or no growth [30,96–98]. Mutations in the ATP binding pocket in both B. subtilis and C. crescentus affect cell shape [57,58]. Some of the spherical mutants we observed are in or around this ATP-binding cleft or pocket. For instance, the I166A and G167A mutations are in a conserved phosphate-binding loop and lead to a round phenotype, while A298L and L300A are in the same cleft and result in similar shape defects. Taken together, this indicates the significance of the proposed ATP-binding cleft in affecting nucleotide activity and therefore cell shape.

Similarly, the release of inorganic phosphate is a crucial event for actin filament depolymerization, as different actin-binding proteins bind to an ADP versus an ADP-Pi-bound subunit [99,100]. In actin, the phosphate escape channel (N111-R177) is gated by asparagine and arginine residues. An N111S mutation causes rapid release of P_i possibly due to the absence of steric hindrance by the serine substitution [99]. Interestingly, in EcMreB, the structural equivalent to N111 is A118, suggesting a lack of gating. We observed that an A118L mutation results in a round shape-phenotype with a slow doubling time. If EcMreB has a similar P_i escape channel as actin, substitution to leucine may introduce steric hindrance that affects P_i release.

Cations associate with ATP and influence actin filament polymerization and stability [101]. In yeast, mutating some of these charged residues in combination was shown to be lethal, again suggesting that single mutants might not be sufficient to explore important residues [97]. In multiple bacterial species, it has been shown that MreB assembly and stability respond to cation (Ca⁺⁺ and Mg⁺⁺; Na⁺ and K⁺) concentration in vitro [31,102–105]. Mutations in these proposed EcMreB cation-binding residues have minimal effect on cell shape, suggesting these MreB variants retain elongasome function. Overall, our results indicate that there may be structural differences between members of the actin superfamily that are only revealed upon genetic perturbations. Additionally, there is a need to test predicted residues with multiple amino acid changes or to make multiple mutations in important domains to fully examine the importance of specific residues.

Predicted residues for MreB-MreB interactions are not important for cell shape regulation

In our growth conditions an mreB deletion would be lethal; therefore, we expected to observe lethality for residues critical for MreB assembly and function. As MreB forms antiparallel filaments it would have been expected that mutations at the predicted monomer interfaces or at the protofilament interfaces would have a large effect on cell shape or growth [106]. Unlike a previous study, where a V121E (protofilament interface) and S284D (monomer-monomer interface) mutations create extreme shape defects hypothesized to be due to defects in MreB polymerization [29], the corresponding alanine substitutions in our library maintain rod shape. In fact, most of the mutations in these predicted interfaces do not affect rod shape formation. This previous study expressed the mutant mreB from a plasmid in a strain constitutively expressing sdiA. The fact that we don’t see large cell shape changes from most of the mutations in these or other predicted residues could be due to the nature of the alanine itself, the fact that our mutants are single copy chromosomal insertions, or because multiple residues are involved in such interfaces, and the effect of a single substitution is minimal.

mreB point mutations can separate its viability and shape control functions

Out of nine residues that could not be mutated, a previous study has reported different amino acid substitution(s) for seven: G134, L293, L303, P304, A332, L307 and M335 [56], indicating that while alanine substitution may not be viable, other classes of amino acids are. In total, these results suggest at least two residues, T119 and E281, may be critical for viability while the specific amino acid change is vital for MreB function in others.

Some point mutations appear to cause cells to significantly lose rod shape, while still being able to grow. These residues must be important for MreB’s function in regulating rod shape, although the exact role that each residue plays is still unknown. The fact that these mutants can lose rod shape but still grow suggests that MreB independently regulates rod shape and viability.

In previous studies it was shown elevated FtsZAQ levels can suppress lethality of spherical E. coli cells in an mre^- background without restoring rod shape [50,51]. This has been thought to be due to the change in surface area or volume of the cells, leading to a need for more FtsZ in order to form a functional Z-ring. The viability of our round mutants without the overexpression of ftsZAQ indicates these mutants have retained the viability function of MreB but lost the rod shape regulation function, and that the change to a sphere does not require more FtsZ for the cells to grow. Taken together, this suggests that the two functions of MreB, cell shape regulation and viability, can be separated.

Intragenic mreB mutations suppress the round shape-phenotype

Previously, elongasome components were shown to restore rod shape in a rodZ mutant [18,28]. While it has been suggested that MreB is necessary for rod shape, similar suppressor screens have been unfeasible until now. To determine if MreB can be bypassed in a similar manner to RodZ we attempted to find suppressors of the cell shape defect in six of the spherical mutants (Table 3). We chose mutants that have a variety of different phenotypes. P108A, A118L and V181A all lead to cells with bright foci, while G167A has cytoplasmic localization. M90A has both weak foci and cytoplasmic localization of MreB while I15A has a higher MIC to cephalexin and cefsulodin than the other spherical strains. While most of the mutations we found are reversions back to WT, we did identify two suppressor mutations in the P108A parent that restore rod shape. Both suppressors are intragenic to mreB (Fig 6). G168A is a better suppressor of the round shape than F84V (Fig 6) and is in the ATP binding pocket (Fig 6A). It may influence ATP binding/hydrolysis, possibly leading to more stable polymers [106]. In support of this, the G168A mutation results in A22 resistance (Fig 4B). The suppressor strain with the P108A G168A mutation is much more sensitive to mecillinam than the P108A parent, suggesting it may have restored the elongasome activity that was lost in the P108A parent. Overall, the results indicate the shape maintaining function of MreB cannot be compensated by other elongasome components making MreB essential for rod-shape maintenance.

PG synthesis may be independent of elongasome activity in spherical strains with mreB mutations

The activity of PBP2, a transpeptidase (TP) that crosslinks muropeptide strands is crucial for elongasome activity. The loss of rod shape in our round mutants suggests that the activity of the elongasome is perturbed. To test this, we measured the susceptibility of 18 round mutants to mecillinam, an amidinopenicillin that specifically inhibits the TP activity of PBP2. All the mutants tested were resistant (>30 folds than WT), suggesting that PBP2 activity is no longer required and therefore that the elongasome is not functioning in these mutants (Table 2). In addition, we assayed grow in a PBP2 and RodA temperature-sensitive background and found that many mutants maintain growth at the non-permissive temperature (S9-S10 Figs).

PBP1A and PBP1B have been shown to synthesize glycan strands independent of elongasome activity [90,92]. We measured the MIC of these mutants against cefsulodin (targets: PBP1A and PBP1B) and cephalexin (targets PBP3) (Table 2). All mutants but one (I15A) were more sensitive to them indicating other major PBPs may have a greater role in PG synthesis in these strains, suggesting that these enzymes are compensating for the loss of MreB. However, we are able to delete either PBP1B or PBP1A from the spherical strains, indicating these enzymes are not responsible for cell wall synthesis in the spherical mutants. In fact, many of the strains grow better in the deletion background, suggesting that the activity of PBP1A or 1B is actually detrimental in the spherical cells.

PBP3 is part of the divisome, suggesting that there may be increased divisome activity in these mutants and FtsZ has been shown to direct PG incorporation into the lateral cell wall when MreB is perturbed [107]. This suggests that the suppressive nature of overexpression of FtsZAQ is not due to making a larger Z-ring as previously suggested, but rather to increasing general cell wall synthesis, allowing the cells to grow without a functioning elongasome.

Bacterial growth

Bacteria were grown using standard laboratory conditions. Unless stated otherwise, cultures were grown overnight in LB medium (10 g/L NaCl, 10 g/L tryptone, 5 g/L yeast extract) and subcultured in the morning 1:1000 at 37^oC in a shaking incubator.

Library construction

An ordered array of alanine or lysine substitutions was made by Twist Bioscience consisting of a linear DNA molecule including DNA upstream of mreB, a kanamycin resistance cassette, and the mreBCD operon (S1A Fig). Each construct has a gfp-sandwich fusion between a non-conserved loop of mreB [46]. The constructs were amplified using PCR and transformed into an E. coli MG1655 strain with the PKD46 plasmid that facilitates λ-red recombination [108]. The recombinants were selected on LB plates supplemented with kanamycin (30 μg/mL) at 30°C and screened for GFP fluorescence. Suspension of desired colonies were incubated at 42°C to cure them of the PKD46 plasmid. Recombinants with mutations near the C-terminal region of mreB were sequence verified to ensure proper recombination occurred.

Library fixation

Strains were inoculated into 96-well plates from freezer stocks in LB and grown overnight at 37°C. Overnight cultures were diluted 1:1000 in fresh medium and grown for three hours. Cultures were diluted 1:1000 into fresh medium and grown for two more hours and then diluted 1:1000 a second time to keep the in exponential phase and grown for two hours. Cells were fixed with paraformaldehyde for 15 minutes, washed with PBS three times and resuspended in PBS before being stained with 5 µg/ml FM-464 and 2 µg/ml DAPI. Cells were stored at 20^oC and imaged.

Microscopy

All imaging was done on 1% M63-glucose agarose pads at room temperature. Phase contrast and fluorescent images were collected on a Nikon Ni-E epifluorescent microscope equipped with a 100X/1.45 NA objective (Nikon), Zyla 4.2 plus cooled sCMOS camera (Andor), and NIS Elements software (Nikon). GFP-fluorescence frames were captured with 200ms excitation unless mentioned.

For initial imaging of the library see Library Fixation for cell prep. For all live cell imaging, cultures were diluted to equal numbers (0.1 OD₆₀₀) and 2 μL of suspension was inoculated into 2 mL of LB and incubated at 37°C to early exponential phase (0.2 to 0.25 OD₆₀₀). All analyses from live cells include data pooled from three independent experiments.

Cell shape statistics were calculated using the MATLAB software Morphometrics [56], and custom software as described previously [26]. Only non-diving cells that could be computationally separated from neighboring cells were used for analysis unless stated otherwise. Custom MATLAB software was used for quality control to remove any “cells” that where not contoured correctly. For a list of data from the fixed cell imaging see S1 Data. Pearson correlation coefficient was used to analyze relationships between sets, Student’s t test, Dunnett’s and Tukey’s multiple comparison tests were used for statistical analyses as indicated. All data from the fixed cell imaging can be found in the S1 Data.

Growth rate analyses

Overnight cultures were grown in LB at 37°C and diluted in LB in equal numbers (0.1 OD) as described previously. 1 μL of the suspension was inoculated into 100 μL of LB in 96-well plates for each strain with at least three biological replicates on each plate and repeated three times. Plates were incubated at 37°C shaking for 16hr in a Biotek plate reader. Doubling times are reported as the max doubling time as determined by the maximal instantaneous growth rate.

Suppressor screen analysis

Overnight cultures were grown in LB at 37°C and diluted in LB in equal numbers (0.1 OD) as described previously and grown to 0.3 OD. 100 μLs of cells were plated onto LB plates supplemented with differing concentrations of cephalexin (3.0, 4.0, 5.0, 6.0 & 7.0 μg/mL) in triplicate and incubated overnight at 37°C. Isolated colonies were suspended in 10 μL LB and imaged to analyze the cell shape.

Sequencing

For whole-genome sequencing, genomic DNA was extracted from strains from overnight cultures using Qiagen DNeasy Blood and tissue kit. Illumina sequencing was performed by MIGS (now SeqCenter) with over 1.3 million reads per strain. Variant analysis was performed using Breseq [109] by mapping SNPs in the mutant DNA to a reference MG1655 strain.

For single construct sequencing, the entire mreB cloning construct or spoT was PCR amplified and sequenced using nanopore technology (Plasmidsaurus).

Mapping mutations

Pymol was used to map the mutations on the AlphaFold model of EcMreB.

Antibiotic sensitivity profiling

Overnight cultures were diluted into equal numbers (0.1 OD) and 1 μL was inoculated into 100 μL of LB supplemented with a gradient of antibiotic concentrations of two-fold dilutions starting with: Mecillinam (512 μg/mL), Cephalexin (16 μg/mL), Cefsulodin (16 μg/mL) and Ampicillin (16 μg/mL). Plates were incubated at 37°C for 16h with shaking. The MIC was defined as the concentration of drug that resulted in an OD less than 0.1 times the OD of the control well without any drug. For six-hour growth ratio assay, overnight cultures were inoculated in equal numbers into 2mL of LB with and without with different antibiotic concentrations and incubated at 37°C for 6h with shaking. It was repeated for three separate days.

Additional mutant constructs

P1 transduction was used to move the mreB variants into different spoT mutation backgrounds or MC4100. The transductants were selected on LB plates supplemented with 30 μg/mL of kanamycin incubated at 37°C overnight.

The temperature-sensitive derivatives for PBP2 and RodA were similarly constructed via transduction of the mreB variants into the respective backgrounds. The transductants were selected from LB plates supplemented with 30 μg/mL of kanamycin and incubated at 30°C overnight. MC4100 was used as the strain to create respective native PBP2 and RodA counterparts in a similar way as described above.

Kanamycin resistance was eliminated from all strains using the recombinase plasmid pCP20 as described previously [108].

Spot assays

Overnight cultures of WT and mutant cells were diluted to equal numbers (0.1 OD₆₀₀) and serially ten-fold diluted. A 4 μL spot of each dilution was plated on an LB and incubated at 37°C overnight.

Immunoblotting

Cells were collected from cultures grown in LB at 37°C (OD₆₀₀ 0.2-0.25) and pellets were resuspended in 100 μL of 2X Laemmli buffer (Bio-Rad Laboratories, Inc.). The suspension was boiled for 20 minutes and ran in stain-free Mini-PROTEAN-TGX gels (Bio-Rad Laboratories, Inc.). Prestained Precision Plus Protein standard (Bio-Rad Laboratories Inc.) was used as a protein standard. After electrophoresis, the resolved proteins were transferred onto a PVDF membrane (Trans-Blot Turbo Transfer System, Bio-Rad Laboratories, Inc.) and blocked with 5% skimmed milk in TBS buffer (Bio-Rad Laboratories, Inc.). Anti-RpoB (EPR18704, Abcam Limited) and FtsZ antibodies (AS10715, Agrisera AB) were diluted 1:10000 in TBST buffer and added to the membrane for 1h at room temperature. The membrane was washed 3X with TBST. A secondary horseradish peroxidase-conjugated antibody (1:3000 dilution, Blotting Grade Goat Anti-Rabbit IgG (H + L), Bio-Rad Laboratories, Inc.) in TBST buffer was added for 1hr at room temperature and then the membrane was washed 3X. Clarity Western ECL Substrate (Bio-Rad Laboratories Inc.) was used to detect the chemiluminescent protein bands, which were visualized using a ChemiDoc MP Imaging System (Bio-Rad Laboratories Inc.). RpoB was used as the internal loading control for normalization. The intensity of the bands were quantified using ImageJ.