Introduction

The rescue of fitness-compromised mutants by selection is a useful strategy to obtain new understanding into the molecular determinants of complex phenotypes (Remigi et al. 2019; LaBar et al. 2020). The approach has shed light on factors such as the flexibility of regulatory networks (Lind et al. 2015), the diversity of structural components affecting ecological success (Lind et al. 2017), the predictability of evolution (Lind et al. 2019), the origins of new genes (Näsvall et al. 2012), and the causes of changes in interactions among bacteria and their hosts (Remigi et al. 2014). Here we turn attention to the evolutionary rescue of a Pseudomonas fluorescens mutant lacking MreB, a key component of cell shape (Errington 2003) that is conserved in a broad range of rod-shaped bacteria (Shi et al. 2018).

For several model organisms, notably, Escherichia coli and Bacillus subtilis, the molecular mechanisms underpinning cell shape have been the subject of intensive investigation (Shi et al. 2018; van Teeffelen and Renner 2018; Egan et al. 2020). The primary determinant of bacterial cell shape in both Gram positive and negative cells is peptidoglycan. Its synthesis and assembly is directed by two main multiprotein complexes that each comprise a combination of enzymes and regulatory elements involved in cell wall dynamics, DNA segregation and cell division (Daniel and Errington 2003; Young 2003; Osborn and Rothfield 2007; Young 2007; Huang et al. 2008; Young 2010). The first of these, the elongasome (or rod complex), inserts newly synthesized peptidoglycan strands into the lateral wall of the growing cell, while the divisome participates in septum positioning and cell wall constriction (van Teeffelen and Renner 2018). Both elongasome and divisome rely on penicillin-binding proteins (PBPs) to polymerize (transglycosylate) and / or to cross-link (transpeptidate) newly synthesized peptidoglycan strands.

The elongasome consists of several catalytically interacting components including the transglycosylase (TGase) RodA (Cho et al. 2016), the transpeptidase (TPase) Pbp2 (Ishino et al. 1986), that function in concert with the semi-autonomous class A PBP (aPBP), Pbp1A, which has both TGase and TPase activities (Randich and Brun 2015). Activity of Pbp1A, while not strictly a member of the elongasome, depends on the outer membrane-anchored protein LpoA (Sardis et al. 2021) with Pbp1A interacting with Pbp2 (Banzhaf et al. 2012; Egan et al. 2020; Kang and Boll 2022). Central to elongasome function – and essential for construction of rod-shaped cells – is the actin-like protein MreB. MreB forms helical structures tethered to the inner membrane and coordinates the pattern of new wall growth along the lateral part of the cell cylinder through interactions with other components of the elongasome (Typas et al. 2012; Shi et al. 2018).

Most bacteria contain a second aPBP, Pbp1B, which like Pbp1A has both TGase and TPase activity. The two aPBPs are partially redundant (Yousif et al. 1985), but based on localisation of Pbp1B to the mid-cell during division (Bertsche et al. 2006), the standard view is that Pbp1B contributes to divisome function (particularly through interaction with PBP3), whereas Pbp1A contributes to cell elongation (Egan et al. 2020). However, it is apparent that Pbp1B also contributes to elongation (Typas et al. 2012; Cho et al. 2016) with evidence from E. coli that its role may be more important than Pbp1A (Vigouroux et al. 2020). In particular it was recently shown that Pbp1B responds to cell wall damage by repairing defects (Vigouroux et al. 2020) and that the bound (enzymatically active) form of Pbp1B is elevated in the absence of Pbp1A (Vigouroux et al. 2020).

In both E. coli and B. subtilis (and most other rod-shape bacteria), MreB is required for maintenance of rod-shape (Shi et al. 2018). While deletion of the gene is lethal under normal growth conditions, inhibition of MreB function by depolymerisation via the drug A22 causes abnormal cell wall growth leading to spherical cells that swell and eventually lyse (Iwai et al. 2002). An earlier study in B. subtilis – where cells devoid of MreB can be maintained in a viable state by culture in the presence of magnesium – found that mutations in Pbp1A could rescue cell viability in the absence of magnesium, but could not return cells to rod-shape (Kawai et al. 2009).

Here, with interest in evolution of the machinery underpinning cell shape, we took advantage of prior knowledge that inactivation of mreB in the rod-shaped bacterium P. fluorescens SBW25, while reducing fitness, is not lethal (Spiers et al. 2002), and asked whether selection could compensate for this cost. Of particular interest was: 1, the capacity for compensatory evolution to restore fitness; 2, the number, timing and impact of fitness-restoring mutations and; 3, the connection between fitness-restoring mutations, growth dynamics and cell morphology.

Materials and Methods

Bacterial strains and culture conditions

Escherichia coli and Pseudomonas aeruginosa were grown at 37°C; and P. fluorescens SBW25 at 28°C. Antibiotics were used at the following concentrations for E. coli and / or P. fluorescens SBW25: 12 μg ml^-1 tetracycline; 30 μg ml^-1 kanamycin; 100 μg ml^-1 ampicillin. Bacteria were propagated in Lysogeny Broth (LB: 10 g tryptone, 5 g yeast extract, 10 g NaCl per litre).

Strain construction

The ΔmreB and Δpbp1A mutants were constructed using SOE-PCR (splicing by overlapping extension using the polymerase chain reaction), followed by a two-step allelic exchange protocol¹. All media for construction of mreB mutants were supplemented with 10 mM MgCl₂. Genome sequencing confirmed the absence of suppressor mutations. The same procedure was used to reconstruct the mutations from the evolved lines (pbp1A G1450A, pbp1A (A1084C), Δpflu4921-4925 (5399934-5403214del)) and ectopic Mini-Tn7T-mreB into ancestral SBW25 and ΔmreB backgrounds. DNA fragments flanking the mutation of interest were amplified using two primer pairs. The internal primers were designed to have overlapping complementary sequences that allowed the resulting fragments to be joined together in a subsequent PCR reaction. The resulting DNA product was TA-cloned into pCR8/GW/TOPO (Invitrogen). This was then subcloned into the pUIC3 vector, which was mobilized via conjugation into SBW25 with the help of pRK2013. Transconjugants were selected on LB plates supplemented with nitrofurantoin, tetracycline and X-gal. Allelic exchange mutants identified as white colonies were obtained, where possible, by cycloserine enrichment to select against tetracycline resistant cells, and tetracycline sensitive clones were examined for the deletion or mutations using PCR and DNA sequencing. Complementation of mreB in the ΔmreB background was achieved by cloning the PCR-amplified DNA fragment into pUC18-mini-Tn7T-LAC, which was subsequently integrated into the P. fluorescens SBW25 genome at a unique site located downstream of the glmS gene (Liu et al. 2014).

Growth dynamics

Strains were initially grown in 5 mL LB broth at 28°C overnight, except for ΔmreB which was grown for 2 days. Growth curves were started by adding 2 µl aliquots of overnight cultures to LB to a final volume of 200 µl (1:100 dilution) in 96-well microplates (3 replicates). Plates were incubated at 28°C (with shaking) for 24 to 48 hours in a BioTek Synergy 2 plate reader. Growth was measured at an absorbance of 600 nm and recorded at 5-minute intervals using the Gen5 software (Bio-Tek). Growth curves were processed as semi-log plots to identify the lag and exponential phases. Doubling times were calculated as: t = ln (2)/k.

Evolution experiment

Ten replicate populations of the ΔmreB strain were grown in 5 mL aliquots of LB broth at 28°C with shaking at 180 rpm. Every 24 h, 5 μL was transferred to fresh media. Every 5 days, samples of each population were collected and stored at −80°C in 15% (v/v) glycerol. The number of generations per transfer changed over the course of the experiment but is roughly ten generations per 24 h period and ∼1,000 generations (100 transfers) were performed.

Competitive fitness assay

Competitive fitness was determined relative to ancestral SBW25 marked with GFP. This strain was constructed using the mini-Tn7 transposon system, expressing GFP and a gentamicin resistance marker (mini-Tn7 (Gm)P_rrnBP1 gfp-a). Strains were grown to exponential phase in shaken LB at 28°C before beginning the competition. Competing strains were mixed with SBW25-GFP at a 1:1 ratio by adding 150 uL of each strain to 5 mL LB and grown for 3 hours. Initial ratios were determined by counting 100,000 cells using flow cytometry (BD FACS Diva II). Suitable dilutions of the initial population were plated on LB agar plates to determine viable counts. The mixed culture was diluted 1,000-fold in LB, then incubated at 28°C for 24 hours. Final viable counts and ratios were determined as described above. The number of generations over 24 hours of growth were determined using the formula ln(final population/initial population)/ln (2). Selection coefficients were calculated using the regression model s = [ln (R (t)/R (0))]/[t], where R is the ratio of the competing strain to SBW25-GFP, and t is time. Control experiments were conducted to determine the fitness cost of the GFP marker in SBW25. For each strain, the competition assay was performed with a minimum of three biological replicates. Ancestral SBW25 had a relative fitness of 1.0 when compared to the marked strain, indicating that the GFP insert is neutral, and that the SBW25-GFP strain was a suitable reference strain for this assay.

Mutation detection

In order to detect mutations in the experimental lines, total DNA was extracted from the ΔmreB strain and from each of the 10 populations at 500 and 1,000 generations. Samples were submitted for 100 bp single-end Illumina DNA sequencing. An average of 6,802,957 single reads were obtained per sample and these were aligned to the reference genome sequence (NC_012660.1) to give approximately 100-fold coverage using Geneious.

To detect the frequency of the pbp1A mutations in experimental lines at generation 50, a colony sequencing approach was used. Colony sequencing was used to prevent the confounding effects of altered cell size, and potentially genomic content per cell, on estimates of the frequency of mutants. To ensure sequenced colonies were isogenic, the frozen glycerol stock from generation 50 was streaked on LB agar plates and resulting individual colonies were each re-streaked on LB agar. Resulting colonies were then used as a template for PCR, using primers that allowed amplification of all the mutations identified by Illumina Sequencing at 500 and 1000 generations. Sanger sequencing was used to identify mutations, with mutations only called if present in sanger sequencing from both forward and reverse primers, with sequences mapped to the pbp1A sequence from the NC_012660 genomic reference using Geneious 11.1.5 (https://www.geneious.com).

Microscopy

Phase contrast and fluorescence images were captured using the BX61 upright microscope (Olympus) at 100x using a Hamamatsu Orca Flash 4.0 camera. The microscope is equipped with an HXP lamp with appropriate multi-band pass filters for different channels. Cell^P/ Cellsens software (Olympus) were used to control the microscope. Cells were grown in LB, and harvested at log phase (OD₆₀₀ 0.4). Viability assays were conducted using the LIVE/DEAD BacLight Bacterial Viability Kit (Thermo Fisher). Viability was measured as the proportion of live cells in the total population (live/ (live +dead)). Nucleoid staining was done using the DAPI nucleic acid stain (Thermo Fisher) following the manufacturer’s protocols.

Time-lapse on agarose pads

Strains were inoculated in LB from glycerol stocks and shaken overnight at 28°C. Cultures were then diluted 100-fold in fresh LB and seeded on a gel pad (1% agarose in LB). The preparation was sealed on a glass coverslip with double-sided tape (Gene Frame, Fischer Scientific). A duct was cut through the center of the pad to allow for oxygen diffusion into the gel. Temperature was maintained at 30°C using a custom-made temperature controller. Bacteria were imaged on a custom-built microscope using a 100X/NA 1.4 objective lens (Apo-ph3, Olympus) and an Orca-Flash4.0 CMOS camera (Hamamatsu). Image acquisition and microscope control were actuated with a LabView interface (National Instruments) (Julou et al. 2013). Typically, we monitored 10 different locations; images were taken every 5 min in correlation mode. Segmentation and cell lineage were computed using a MatLab code implemented from Schnitzcell (Locke and Elowitz 2009). Bacteria were tracked for 3 generations. In order to limit the number of generations, we monitored bacteria harvested from a liquid stationary phase. Since, bacteria shorten after entering stationary phase, bacteria lengthened when resuming growth under the microscope, so that the elongation rate is larger than the septation rate for the first generations (Fig. 3B). To get the control line in our experimental configuration, we measured these rates in three species (ancestral P. fluorescens SBW25, E. coli MG1655 and P. aeruginosa PAO1). Points below the line indicate a reduction of cell volume beyond the fact that cells simply resumed growth.

Scanning Electron Microscopy (SEM)

Cells were grown in LB, and harvested in log phase. Cells were fixed in modified Karnovsky’s fixative then placed between two membrane filters (0.4μm, Isopore, Merck Millipore LTD) in an aluminum clamp. Following three washes of phosphate buffer, the cells were dehydrated in a graded-ethanol series, placed in liquid CO₂, then dried in a critical-point drying chamber. The samples were mounted onto aluminum stubs and sputter coated with gold (BAL-TEC SCD 005 sputter coater) and viewed in a FEI Quanta 200 scanning electron microscope at an accelerating voltage of 20kV.

Image analysis

Compactness and estimated volume measurements of cells from liquid culture. The main measure of cell shape, compactness or C, was computed by the CMEIAS (Liu et al. 2001) software as: √ (4Area/π)/major axis length (Liu, J. 2001). Estimated volume or V_e was estimated with different formula, according to cell compactness, for spherical cells that have a compactness ≥ 0.7, V_e was computed using the general formula for ellipsoids: V=4/3π (L/2) (W/2)², where L=length and W=width. V_e of rod-shaped cells, defined as having a compactness value ≤ 0.7, were computed using the combined formulas for cylinders and spheres: V_e = (π (W/2)² (L-W))+ (4/3π (W/2)³).

Cell volume and division axis of cells on agarose pads. Cell volume was computed from the mask retrieved after image segmentation. Masks are fitted to an ellipse. The elongation axis is given by the direction of the major axis. The division axis relative to the mother was computed by comparing the elongation axis between mother and sister cell, through the following formula: |sinθ|, where θ is the angle between mother and sister cell. |sinθ|=0 means that mother and daughter cells are aligned, |sinθ|=1 means that they are perpendicular.

Probability p to pass to the next generation: For the second generation on the pad, we computed the probability to pass to the next generation as the capability of cells to progress through the cell cycle and divide. Bacteria that did not grow or stop elongating before dividing were classified as non-proliferating. For all non-proliferating bacteria, we confirmed that no division occurred during the next 5 hours.

Elongation rate, septation rate, growth asymmetry and the precision of septum positioning: The elongation rate, r, is measured as the rate at which cell volume increases: V (t)=V₀e^rt, where t is the relative time in the cell cycle. The septation rate is given by , where τ is the average division time, i.e., the time between birth and septation. For all sister cell pairs, we computed relative difference in growth between sister cells by measuring the relative elongation rate (i.e., volume extension rate) of sister cells at generation , where r_1,2 is the elongation rate for each cell of sister pairs. The population average C =<abs (r -r)/(r₁+r₂)>_cells. was computed on the proliferating sub-population in order to avoid trivial bias due to cell proliferation arrest of one of the two sister cells. We did the same for the relative difference in division time and measured C_T=<abs (T₁-T₂)/ (T₁+T₂)>_cells., where T_1,2 are the division time of sister cells. For all cells at generation F1, we computed the precision of septum positioning by measuring the relative cell volume of the two daughter cells: , where V_1,2 are the volumes of each daughter cell. The population average C =<abs (V -V)/(V₁+V₂)>_cells was computed for every F1 cell having offspring.

Laplace argument

For a given geometry, cell volume is determined, at steady state, by an analog of the Young-Laplace law for thin elastic membranes, which relates the pressure discontinuity ΔP at the cell wall to membrane tension and radius of curvature R_s. The pressure jump is powered by turgor pressure in the cytoplasm. For spheres, having two identical curvature radii R_s, the law is . For cylinders, having only one finite radius of curvature R_c, the law is . Assuming that turgor pressure and membrane tension are the same, and ΔP_S = ΔP_C, we obtain R_s = 2R_c. Hence, the ratio of volumes between the two geometries can be expressed as a function of the original ancestral SBW25 length L and width 2R_c: , where V_s and V_c are the volume of the sphere and the cylinder, respectively. Hence a change from cylindrical to spherical shape should result in a volume increase by a factor 1.78 for P. fluorescens, with a diameter of 1µm and a length of 3µm (Fig. 3A).

Protein sequence alignment and modeling

Protein sequences were obtained from NCBI BLAST (http://blast.ncbi.nlm.nih.gov) and The Pseudomonas Genome Database (Winsor et al. 2016), and aligned using MEGA7 (Kumar et al. 2016). The sequence alignment was visualised using ESPript ‘Easy Sequencing in PostScript’ (Robert and Gouet 2014). Protein visualisation was performed using Visual Molecular Dynamics (VMD) (Humphrey et al. 1996) based on the crystal structure of Acinetobacter baumannii Pbp1a in complex with Aztreonam as the base model (Pbp1A from A. baumannii shares a 73% amino acid sequence identity (E value = 0.0) to the Pbp1A of P. fluorescens SBW25). Sequences were aligned, and locations of the mutations in the evolved lines were mapped in the corresponding regions. The PDB file was downloaded from the RCSB Protein Data Bank (www.rcsb.org) using PDB ID 3UE0.

Membrane preparation and detection of penicillin-binding proteins

For the membrane preparation of the different genotypes of P. fluorescens SBW25, 150 mL cell cultures were grown to an OD₆₀₀ of 0.6 and harvested by centrifugation at 15,000 x g for 15 minutes at 4 °C (Beckman JLA-16.250 rotor, Beckman Avanti J-25, Beckman Coulter). Pellets were washed with 20 mM potassium phosphate (pH 7.5) and 140 mM sodium chloride buffer. Following another 30 minutes of centrifugation at 360 x g at 4 °C, cells were disrupted by passing twice through a French press (‘Pressure Cell’ Homogenizer FPG12800, Stansted Fluid Power Ltd). Cell lysates were subjected to centrifugation at 360 x g for 10 minutes at 4 °C. Supernatant fractions were collected and centrifuged at 75,600 x g for 90 minutes at 4 °C (Beckman JA25.50 rotor, Beckman Avanti J-25, Beckman Coulter). The resulting pellets were washed once again and resuspended in 100 µL potassium phosphate buffer. The concentration of the membrane preparations was measured using Bradford reagent (BioRad) with bovine serum albumin as a standard (Sigma).

For detection of PBPs, membranes were incubated for 30 minutes at 37 °C with a fluorescent labeling agent, Bocillin FL Penicillin (Invitrogen, Thermo Fisher Scientific) (final concentration: 0.05 mM). Finally, the samples were denatured with 5x SDS loading buffer with ß-mercaptoethanol as a reducing agent at 100 °C for 5 minutes, followed by centrifugation to remove membrane debris. 20 µL of reaction mixture was subjected to SDS-PAGE analysis on an 8% polyacrylamide gel, protected from the light. Bocillin FL-labeled PBPs were visualized with a laser scanner, Typhoon FLA 9500 (GE Healthcare Life Sciences) at 473 nm with a 530DF20 emission filter.

Peptidoglycan isolation and analysis

For peptidoglycan profiling and analysis, P. fluorescens genotypes were grown in LB medium to an OD₆₀₀ of 0.2 and harvested. Purification of the peptidoglycan was carried out as previously described (Desmarais et al. 2013) with some minor changes. Briefly, the cell pellets were boiled in an equal volume of 5% (w/v) SDS for approximately an hour and left stirring overnight at room temperature. The sacculi were washed repeatedly with MilliQ water by ultracentrifugation (150 000 × g, 15 min, 20 °C TLA100.3 Beckman rotor; OptimaTM Max ultracentrifuge Beckman, Beckman Coulter) until SDS was eliminated from the samples. The clean sacculi were digested with muramidase (Cellosyl 100 μg.mL^-1) overnight at 37 °C. The enzyme reaction was stopped by heat-inactivation at 100 °C. Coagulated proteins were removed by centrifugation (21,000 x g for 15 min, Heraeus^TM Pico^TM 21 Microcentrifuge, Thermo Fisher Scientific). For sample reduction, supernatants were adjusted to pH 8.5 - 9.0 with borate buffer, followed by incubation for 30 minutes at room temperature with freshly prepared NaBH₄ solution (final concentration of 10 mg.mL^-1). Finally, the pH was adjusted to 3.5 by the addition of phosphoric acid. Muropeptides were separated by UPLC on a Waters UPLC system (Waters) equipped with a Kinetex C18 UPLC Column, 130 Å, 1.7 μm, 2.1 mm × 150 mm (Phenomenex) and a dual wavelength absorbance detector. Elution of muropeptides was detected at 204 nm. Separation of the muropeptides was carried out at 45 °C using a linear gradient from Buffer A (formic acid 0.1% (v/v) to Buffer B (formic acid 0.1% (v/v), acetonitrile 40% (v/v) in an 18 min run with 0.250 mL.min^-1 flow.

Fluorescent D amino Acid labeling

Short-pulse staining of FDAA incorporation into growing P. fluorescens cell walls was accomplished using FDAA BODIPY-FL 3-amino-D-alanine (BADA) (Hsu et al. 2017) (green; emission 502 nm, excitation 518 nm). Briefly, BADA was diluted in dimethyl sulfoxide to a concentration of 100 mM. For each test strain, one mL of exponential culture (OD 600 of 0.4) was pelleted by centrifugation (10,000g: 2,000g for ΔmreB) and resuspended in 100 uL. BADA was added to achieve a final concentration of 1mM BADA. Incubation at 28°C for 20% of strain generation time. Following this, 230 µL of ice-cold 100% ethanol was added to culture. An additional 1 mL of ice cold 70% ethanol was added to remove excess dye. Cells were fixed on ice for 15 minutes before being washed with 1 mL of PBS three times to remove excess BADA. Finally, cells were resuspended in 20-100 µL of PBS and subjected to standard fluorescent microscopy on agarose pads. Fluorescent microscopy was performed as described above.

Results

Deletion of mreB from P. fluorescens SBW25 generates viable spherical cells

Previous observations of viable mreB mutants of SBW25 came from transposon screens of adaptive (wrinkly spreader) mutants (Spiers et al. 2002). To establish that cells can indeed proliferate despite the absence of MreB, mreB was deleted from ancestral P. fluorescens SBW25 (hereafter SBW25). In LB broth, SBW25 ΔmreB cells were viable, spherical and variable in cell volume (Fig. 1A).

Figure 1.

Cell shape, width and length were used to estimate volume (V_e). The volume of ΔmreB mutant cells was larger (20.65 ± 16.17 µm³) than those of ancestral SBW25 (3.27 ± 0.94 µm³) cells. A negative correlation between V_e and compactness (the ratio of spherical to circle shape with a compactness of 1.0 being a circle) as a result of cell elongation was evident for SBW25 (R² = 0.63) (Fig. 1B). In contrast, only a weak correlation was observed for the ΔmreB mutant (R² = 0.13) (Fig. 1B). As cell volume increases, DNA content is expected to increase provided that DNA replication continues independently of septation. Increased DNA content and spherical cell shape are both predicted to further perturb cell division (Jun and Mulder 2006; Jun and Wright 2010). Ancestral SBW25 and ΔmreB cells were stained with a nucleic acid stain (DAPI) and analysed using flow cytometry. In both genotypes DNA content scaled with cell volume. The largest ΔmreB cells had many times the DNA content of ancestral cells, scaling approximately with volume (Supp Fig 1, Supp Fig 2).

Figure 2.

In comparison with SBW25, the ΔmreB mutant had reduced fitness (Fig. 1C) with a longer effective generation time, extended lag phase, and lower maximum yield (Supp Fig 3). Complementation of ΔmreB with the ancestral gene restored rod-shaped morphology and fitness (Supp Fig 4).

Figure 3.

Figure 4.

To understand the causes of reduced fitness, the division of single cells was monitored by time-lapse video microscopy on agar pads. Although SBW25 ΔmreB cells appear spherical, measurement of cell aspect ratio revealed existence of an elongation axis that remained constant throughout the cell cycle (Supp Fig 5). As in rod-shaped bacteria and ovococci, the septum formed perpendicularly to the elongation axis. After cell division, the elongation axis rotated 90° relative to the mother cell (Supp Fig 6, Supp Movie 1), which is reminiscent of ovococci¹⁷. But more surprisingly, measurement of the rate of volume extension (hereafter termed elongation rate) showed that during exponential phase it was indistinguishable from the cell elongation rate of ancestral SBW25 (Fig 1D).

Since elongation rate was unchanged upon deletion of mreB, the reduced fitness of the mutant was likely due to production of non-viable cells. We thus measured cell viability in LB broth with a “live / dead” assay and showed that 95% of ancestral SBW25 cells were viable, while the fraction of viable ΔmreB mutant cells was reduced to 81%. Thus, the decreased fitness of the mutant is primarily a consequence of the production of non-viable cells (Fig. 1E). To confirm this result, we performed time lapse microscopy of cells grown on agar pads to measure the probability p to pass to the next generation. For ancestral SBW25, the probability p of producing offspring equals 1, whereas for ΔmreB cells, p = 0.85. In instances where the probability to pass to the next generation p is not equal to one, the equation for growth dynamics is represented by dN=p*r*N*dt, where N is the number of cells at time t and r is the cell elongation rate. Hence, the effective growth rate of the population is given by p*r. Because ΔmreB and ancestral SBW25 cells elongate at the same rate, the fitness of ΔmreB cells is directly given by p. Consistently, the probability to pass to the next generation p = 0.85 as measured in single cell experiments is approximately equal to the measured fitness of 0.77 for ΔmreB in bulk experiments (Fig 1C).

We next examined the causes for decreased proliferation capacity by measuring, at the single cell level, the correlation between the probability p to pass to the next generation and the fidelity of cell division. To this end, the precision of septum positioning, C_s, and correlation between the relative difference in elongation rate between sister cells at generation F1, C_g, and probability p of producing offspring at the end of generation F1 (Fig 1F) were calculated. As a proxy to determine the precision of septum positioning, we measured the relative difference C_s in cell volume V_e, between two daughter cells immediately after division at generation F2. In ancestral SBW25, septum position was accurately placed, with only 5 % difference in the volume of sister cells. In ΔmreB mutant cells, the accuracy of septum position was significantly altered, with 19.4 % difference in the volume of sister cells. To determine if sister cells had different growth dynamics, we measured the relative difference C_g in elongation rate r between two sister cells at generation F1. In ancestral SBW25, the elongation rate in pairs of sister cells differed by 5 %, whereas in ΔmreB cells the difference in rate was on average 55 %. Combined together, error in septum positioning and growth asymmetry between sister cells explains why the ΔmreB mutant produces cells of variable volume, which then have different proliferative capacity. To examine how cell volume affects proliferation ability, the probability p to pass to the next generation was plotted as a function of cell volume. As evident in Fig. 1G the volume distribution has two extreme tails, with both large and small cells being unlikely to produce offspring cells.

Selection rescues fitness of a compromised mreB deletion mutant

To determine whether selection could discover mutation (s) that might compensate for reduced fitness of ΔmreB, SBW25 ΔmreB was propagated for 1,000 generations in shaken broth culture with daily transfer. The competitive fitness of derived types was measured relative to ancestral SBW25 using neutral markers to distinguish competitors. Fitness of all replicate lines (relative to ancestral SBW25) significantly improved (Fig. 2A) to the point where fitness was comparable to ancestral SBW25 (Supp Fig 7). Subsequent measurement of fitness at generations 50, 250 and 500 showed that compensation of fitness was rapid, with a highly similar trajectory for each of the 10 replicate populations (Supp Fig 8). Such rapid restoration of fitness is indicative of a small number of readily achievable compensatory mutations.

Examination of derived genotypes at generation 1,000 by microscopy showed that all lines retained the coccoid shape (Fig. 2B, Supp Fig 9). Cell volume decreased compared to that of the ancestral cell type (Fig. 2B) and did not overlap with the ΔmreB population (Fig. 1B and 2B) showing that the derived cells define a new phenotype and not a subset of the spherical ΔmreB ancestor. Analysis of the pattern of cell division in the derived lines showed evidence of an elongation axis, which rotated at cell division, as observed for ΔmreB mutant cells (Supp Fig 6).

Identifying compensatory mutations

In order to identify the genetic basis of adaptation, populations were sequenced at generations 500 and 1,000 with the ensuing reads mapped to the genome of SBW25 (Table 1). A single gene, pbp1A (PFLU0406) had mutations in seven independent lines. In four of these, Line 1 (D484N), Line 3 (D721A), Line 4 (T362P) and Line 6 (N698K), the mutation has risen to high frequency, but only in Line 6 was the mutation fixed. Some lines (4, 9 and 10) have more than a single pbp1A mutation and in some instances the mutation detected at generation 500 differs from that evident at generation 1,000. Clonal interface between competing mutations thus seems likely. In the three lines with no evidence of pbp1A mutations at generations 500, or 1,000, two exhibited mutations in genes involved in septum formation (ftsA for Line 2 and ftsZ for Line 5). Finally, Line 7, contained a five gene deletion spanning PFLU4921 to PLFU4925 which included oprD, but also has a non-synonymous mutation in ftsE that was not detected at generation 500, but composed 65% of sequencing reads at generation 1,000. The prominence of pbp1A mutations led us to further study two lines showing pbp1A mutations (Line 1 and Line 4) and to compare them with the five gene deletion (PFLU4921 to PLFU4925 Δ5399934-5403214) evident in Line 7.

The pbp1A gene is a class A penicillin-binding protein with both transpeptidase (TP) and transglycosylase (TG) activities. It is semi-autonomous and interacts with the rod complex and specifically with PBP2 (Banzhaf et al. 2012; Geisinger et al. 2020). Pbp1A contains three known domains (Fig. 2C). Structure mapping of the mutations shows that the mutation in Line 1 occurred in the transpeptidase (TP) domain, which is proximal to the active site (Supp Fig 10). Similar mutations in Streptococcus pneumoniae cause a loss-of-function (Job et al. 2008) and rescue viability of both mreC and mreD mutants (Land and Winkler 2011). The mutation in Line 4 occurred in the oligonucleotide/oligosaccharide binding or Outer membrane Pbp1A Docking Domain (OB/ODD) domain⁶. Each mutation was reconstructed by allelic exchange in SBW25 and SBW25 ΔmreB backgrounds.

Given the very rapid fitness increase (Supp Fig 8) we looked for evidence of pbp1A mutations at generation 50. The gene was amplified by PCR from 16 colonies from each replicate population and the product Sanger sequenced. All lines had at least one mutation (out of the 16 colonies) demonstrating the relative ease at which pbp1A mutations arise. Results are shown in Supp Table 2.

Analysis of pbp1A mutations

Both ΔmreB pbp1A reconstructed genotypes (hereafter ΔmreB pbp1A*) produced spherical to ovoid cells with near ancestral SBW25 volume (ΔmreB pbp1A (D484N) M = 3.86 ± 0.89 µm³; ΔmreB pbp1A (T362P) M = 5.51 ± 0.87 µm³) and DNA content (Fig. 2B, Supp Fig 2). Analysis of the pattern of cell division in the reconstructed lines showed evidence of an elongation axis that rotated at cell division as observed for SBW25 ΔmreB (Supp Fig 6). ΔmreB pbp1A* genotypes showed growth dynamics and fitness effects similar to ancestral SBW25 (Supp Fig 11) (Fig. 2A), indicating that these pbp1A mutations are each sufficient to restore ancestral fitness and to recapitulate the major phenotypes of derived Lines 1 and 4. Elongation rate was identical to SBW25 for SBW25 ΔmreB carrying the reconstructed mutations, (Supp Fig 12). Fitness of a genetically engineered ΔmreB Δpbp1A mutant (relative to ΔmreB) was 1.35 (SD + 0.04, n = 6) was indistinguishable from fitness of the Line 4 mutation reconstructed in the same ΔmreB background (1.38; SD + 0.11, n = 6) demonstrating that loss-of-function of pbpB1A is sufficient to restored fitness (Supp Fig 13). Loss-of-function is also consistent with the finding that at generation 50 all lines harboured mutations in pbp1A (Supp Table 2).

When reconstructed in ancestral SBW25, pbp1A mutations had little effect on population dynamics (Supp Fig 11A) or shape (Fig. 2B), however elongation rates were reduced (Supp Fig 12), suggesting that the pbp1A mutations result in a reduction – or more likely even complete loss-of-function. In addition, cells were narrower than WT cells: 0.89 µm (SD ± 0.07) and 0.94 µm (SD ± 0.05) for Pbp1A TP and OB/ODD mutations, respectively, compared to 1.00 µm (SD ± 0.06) for ancestral SBW25. This resulted in smaller cell volumes (Fig. 2B, Supp Fig 14 & 15). The reduced widths / volumes -- which also correlate with reduced DNA content (Supp Fig 2) – might be indicative of reduced enzymatic function leading to reduced lateral cell wall synthesis as the case for loss-of-function mutations in pbp1A reported in both B. subtilis and E. coli (Murray et al. 1998; Claessen et al. 2008; Kawai et al. 2009; Banzhaf et al. 2012).

Analysis of Δpflu4921-4925

The five-gene deletion (Δ5399934-5403214 (Δpflu4921-4925)) in derived Line 7 (Supp Fig 16, Supp. Table 1) includes three hypothetical proteins, a cold shock protein (PFLU4922, encoding CspC), and an outer membrane porin, PFLU4925, which encodes OprD. The latter is responsible for the influx of basic amino acids and some antibiotics into the bacterial cell (Skurnik et al. 2013). The deletion was also reconstructed and characterised in the ΔmreB and ancestral SBW25 backgrounds. The relative fitness of the ΔmreB Δpflu4921-4925 genotype was similar to the ancestral SBW25 type (Fig. 2A), but with an extended lag time (Supp Fig 11). Cells were spherical with an average Ve of 5.26 µm³ (SD ± 3.13) (Fig. 2B, Supp Fig 14 & 15) and harboured less DNA compared to ΔmreB SBW25, but more than the Pbp1A* reconstructions in the same background (Supp Fig 2). In addition, a fraction of the cell division events were defective resulting in clumps of spherical cells with imperfectly formed septa (Fig 2B, Supp Fig 14-17). Spherical cells with notably incomplete septa were not observed in the Evolved Line 7 population and appeared to be a consequence of the ΔmreB and Δpflu4921-4925 in the absence of other compensatory mutations (Supp. Table 1).

When genes pflu4921-4925 (Δ5399934-5403214) were deleted from ancestral SBW25 the cells remained rod-shaped and displayed aberrant growth characteristics relative to the ancestral strain (Supp Fig 14). The elongation rate was slower (Supp Fig 12), cells were significantly thinner (width = 0.74µm SD ± 0.06; p < 0.001, two sample t-test), and had a smaller average Ve (2.47 µm³, SD ± 1.18) (Fig. 2B and Supp Fig 15). As in the ΔmreB background, a fraction of cells failed to completely separate at cell division and DNA was observed to be dispersed between incomplete septa (Supp Fig 14-17). The average DNA content of the SBW25 Δpflu4921-4925 cells was also larger than the pbp1A* reconstructions in the same background (Supp Fig 2).

Cell volume homeostasis

The large average cell volume observed in stationary phase SBW25 ΔmreB cells (Fig. 2B) suggests that some of the physical parameters governing cell volume may have been altered. Cell volume at steady state is determined by the equilibrium between turgor pressure (Rojas and Huang 2018) and membrane tension through the Laplace law for thin elastic shells (Verge-Serandour and Turlier 2021) (see Materials and Methods for elaboration). Since it is difficult to imagine a scenario where the cell-wall would strengthen in the absence of mreB, mreB loss is expected to result in a significant decrease of internal pressure compared to ancestral SBW25 cells, synonymous with the influx of water through pores formed in a cell-wall.

During exponential growth, the cell volume of daughter cells is limited by the amount of new material that can be incorporated by the mother cell. According to the “adder” model (Campos et al. 2014; Taheri-Araghi et al. 2017), the average cell volume in the population converges within a few generations towards the average added volume. We thus quantified the added volume for all strains (Fig. 3A) and confirmed that cell volume rapidly reduces for exponentially growing SBW25 ΔmreB cells (Supp Movie 2). Since, the volume of ΔmreB cells in exponential phase is much lower than in stationary phase, it suggests that cells swell during stationary phase, probably due to defects in cell wall integrity with deleterious consequences for cell viability. Hence, without restoring cell wall integrity, cell volume homeostasis cannot be maintained. Remarkably, the added-volumes of spherical cells in each of the derived mutant lines (Fig. 3A) were close to the volumes corresponding to the passage from a cylindrical to spherical geometry with constant turgor pressure and membrane tension (see Materials and Methods). This suggests that adaptive mutations restored the mechanical integrity of the cell wall.

To maintain cell volume homeostasis, elongation and septation rates must be balanced so that cells maintain a steady volume throughout successive generations. Indeed, if septation is faster than elongation, cell volume of the progeny decreases, while if elongation is faster than the rate of septation rate, cells increase in volume. In both instances the fate of populations is extinction. The rate of septation was significantly higher than the elongation rate in ΔmreB cells showing that the distribution of cell volume is unstable and confirms that the average volume is expected to decrease in subsequent generations (Fig. 3B). In contrast to SBW25 ΔmreB cells, the rates of elongation and septation were balanced in the derived lines (Fig. 3B), indicating that cell volume homeostasis has been reached. Reconstruction of pbp1A mutations in SBW25 ΔmreB also restored this balance, suggesting that the mutations contribute to maintenance of volume homeostasis in the absence of mreB. In contrast, the Line 7 deletion reconstructed in SBW25 ΔmreB did not restore this balance (Fig. 3B,).

Another important feature for cell volume homeostasis is the fidelity of cell division: two daughter cells with equal volume must themselves generate four daughters of equal volume. Therefore, we measured the relative difference in elongation rates (C_g) and division time (C_T) between sister cells at generation F1 on agar pads, as well as the relative difference in precision of septum positioning (C_s) for all derived and reconstructed lines. All evolved lines regained symmetry both in elongation rate and in precision of septum positioning, but was most notable for the symmetry of elongation rate (Fig. 3C,D). These improvements correlate with an increase in proliferative capacity, while no significant correlation was observed in the relative difference of division time C_T between sister cells (Supp Fig 18). Reconstructions in ancestral SBW25, showed a slight alteration of symmetry, which is more pronounced in the case of Line 7 (five gene deletion). Reconstructions in evolved lines Line 1 and Line 4 showed improvement in growth asymmetry. Interestingly, recovery was only partial for Line 7, raising the possibility of additional effects due to a secondary mutation whose effects depend on the mechanics of the environment (liquid vs solid culture). Indeed, as indicated above and shown in Supp Table 1, Line 7, at generation 1,000 also contains a non-synonymous mutation in ftsE (T188A) that has reached high frequency. Given that FtsE promotes elevated peptidoglycan synthesis at cell septa the possibility of synergistically beneficial effects with the OprD-inclusive deletion is not improbable (Mallik et al. 2023).

The mutations observed in Line 1 or Line 4 showed a significant improvement in the precision of septum position, suggesting that asymmetry in elongation rate between daughter cells has greater impacts on cell survival than septum positioning. This can also be seen from the shape of the curves (Fig. 3C,D), where p decreases linearly with growth asymmetry up to 50%, while precision of septum positioning saturates at 20% regardless of p, probably because nuclear occlusion is able to achieve a sufficiently good level of septum positioning.

Investigating the molecular consequences of mutations

To obtain insight into the molecular mechanisms by which mutations in pbp1A compensate for loss of mreB, we performed a range of biochemical analyses on SBW25 and SBW25 ΔmreB engineered with the fitness-rescuing mutations in pbp1A: SBW25 / SBW25 ΔmreB Pbp1A (D484N) (mutation in the TPase domain); SBW25 / SBW25 ΔmreB Pbp1A (T362P) (mutation in the OB/ODD domain). SBW25 ΔmreB Δpbp1A was used as a negative control. In ancestral SBW25 36.34% (+ 2.26) of muropeptides are cross-linked (Fig. 4A). The ΔmreB mutant showed a marked increase in cross-linking (40.47% +/- 1.50). The degree of cross-linking in the pbp1A mutants in either SBW25 or the ΔmreB genetic background were lower (31.28% +/- 0.98 to 33.66% +/- 0.52).

Lower levels of cross-linking observed in the ΔmreB and SBW25 reconstructed lines suggests that mutations in Line 1 and 4 either decreased the amount of transpeptidation or increased the amount of transglycosylation. To discriminate between the two hypotheses, we utilized Bocillin-FL labeling, a fluorescent penicillin derivative that binds to the active site of PBPs, as a proxy to measure PBP1a TPase activity. We included SBW25 Δpbp1A as a control to specifically identify the band corresponding to Pbp1A in the SDS-PAGE gel (Fig. 4B). In agreement with higher levels of cross-linking activity measured in SBW25 ΔmreB (Fig. 4A), the intensity of the Pbp1A band was slightly higher in ΔmreB cells compared to the ancestral genotype. In comparison to the ancestor of the evolved lines (ΔmreB SBW25), this signal was reduced in the Line 1 mutant (TP domain) (Fig. 4B) and completely vanished in the Line 4 mutant (OB/ODD domain). The activity of this domain is not well characterized, but is thought to be an important regulatory domain that licenses TPase domain function through interaction with LpoA (Sardis et al. 2021). The relative peptidoglycan density and UPLC traces are shown (Supp Fig 19).

In addition, we studied how the ΔmreB and pbp1A OB/ODD and TP domain mutations affect the location of active cell wall synthesis. This was achieved by incorporation of fluorescently labelled D-amino acids (FDAAs) via brief pulses of BADA into exponentially growing cells followed by microscopy to visualize locations of active cell wall synthesis (Fig. 4C). In SBW25, BADA incorporation into the cell wall was homogeneous across the lateral surface during growth, but enriched at the mid-cell during septation. In SBW25 ΔmreB cells, BADA labelling was heterogeneous across the entire cell wall. This is contrary to expectation – at odds with observations of diffuse and homogeneous wall synthesis in MreB-depleted E. coli (Ursell et al. 2014) – and indicates that even in the absence of MreB, cell wall synthesis is active. SBW25 ΔmreB containing pbp1A* mutations resulted in small spherical cells in which the prominence of cell wall synthesis in the septal region was restored as the primary location of peptidoglycan production, particularly in longer cells (Fig. 4C). Thus pbp1A mutations appear to restore cell homogeneity by eliminating the disorganized pattern of cell wall synthesis (Fig 4C).

Discussion

Ability to generate a viable mreB deletion mutant of P. fluorescens SBW25 – albeit with significantly compromised fitness – has provided opportunity to investigate the phenotypic consequences of loss of MreB function without use of MreB-inhibiting drugs, and to explore mutational routes leading to fitness restoration.

After 250 generations of selection, fitness of all derived lines was restored to wild type levels (Fig 2A). Measurement of fitness at earlier time points showed that compensation occurred early in the selection regime and in single steps (Fig 2A, Supp Fig 8), consistent with compensation occurring via one or a small number of mutations. Sequencing and genetic reconstruction experiments confirmed, particularly for the class of mutant with defects in pbp1A, that single point mutations reducing or eliminating gene function were sufficient. In all derived lines, cell shape remained spherical. This sits in accord with evidence suggesting that there are no genetic routes or growth conditions able to restore rod-like shape in the absence of MreB (Shi et al. 2018). This stated, we recognise that rod-shape bacteria belonging to the Actinobacteria and Rhizobiales lack MreB (Margolin 2009; Zhao et al. 2021) suggesting that in principle it is possible to restore rod-shape in the absence of MreB.

In experiments where selection is used to find mutations that compensate for deleterious effects arising from deletion of a gene that compromises fitness (Laan et al. 2015; Lind et al. 2015; Rainey et al. 2017), a central issue concerns the focus of selection: the defect that selection might stand to correct. In this work, where deletion of mreB provided the compromised starting genotype, the most obvious difference between SBW25 ΔmreB and each of the ten derived lines is cell volume. It is therefore tempting to suggest that volume is a target of selection, and especially so given known adaptive changes in cell volume in the long-term Lenski experiment (Lenski and Travisano 1994; Monds et al. 2014). However, a closer examination of data, particularly from time-resolved microscopy, patterns of peptidoglycan synthesis and biophysical considerations, suggests a more complex set of effects and adaptive responses.

At first glance, the most notable distinction between cells of ancestral and derived types is shape and volume: SBW25 cells are rod shaped (thus with low compactness) with relatively little variance in the volume of daughter cells, whereas the shape of ΔmreB cells tends toward spherical with large variance in volume. While the rate of elongation is the same in SBW25 and ΔmreB cells, substantial variance is apparent in the rate of elongation among daughter (ΔmreB) cells (Fig 1D). Also evident are significant differences in cell viability, with viability of daughter cells being high in SBW25, but much reduced in the ΔmreB mutant. With focus on ΔmreB cells, viability of daughter cells depends on cell volume and declines significantly in both very large and very small cells (Fig 1G). This is understandable in terms of effects on DNA segregation, with small cells likely nonviable through lack of DNA, and large cells being similarly non-viable as a consequence of multiple copies of intertwined chromosomes and imbalance of osmotic pressure.

A key issue concerns the causes of variability in cell volume, toward which both septation asymmetry (C_s) and differential elongation rate (C_g) between daughter cells are the primary contributory factors. As evident by comparing C_s and C_g for ancestral SBW25 (black open square in Figs 3C and 3D) and ΔmreB (blue open circle in Figs 3C and 3D) cells, respectively, the greatest difference is the relative difference in elongation rate among sister cells (C_g). This indicates that this is the primary cause of variability in cell volume in the mreB mutant. Moreover, if correct, it leads to the prediction that fitness restoration will depend on selection finding mutational routes to restore population-level homogeneity in elongation rate among offspring.

What then, underlies differences in elongation rate? Here, data on peptidoglycan cross-linking, TPase activity and patterns of cell wall synthesis in SBW25 versus ΔmreB cells are suggestive. Relative to SBW25, cells of the mreB mutant showed increased peptidoglycan cross-linking (Fig 4A) likely resulting from enhanced TPase activity of Pbp1A (Fig 4B) and heterogeneous (uneven) patterns of cell wall synthesis (Fig 4C). These data indicate that despite the absence of MreB, cell wall synthesis continues, but is dysregulated in space and time. This is in marked contrast to cell wall synthesis in MreB-impaired (by treatment with A22) E. coli where cell wall synthesis is diffuse and homogeneous (Ursell et al. 2014).

Just what contributes to cell wall synthesis on removal of MreB is unclear. The standard view is that on deletion (or depolymerisation) of MreB, the elongasome is registered non-functional (Park and Uehara 2008), but compelling evidence appears to be lacking (Cho et al. 2016). In E. coli, PBP2 can trigger elongasome initiation independently of MreB by promoting interaction with a hypothetical protein that in turn interacts with PBP2 and the cell wall (Özbaykal et al. 2020). Additionally, in A. baumannii high Pbp1A activity inhibits elongasome function (Simpson et al. 2021) with ΔponA (pbp1A) mutants having enhanced elongasome function (Kang et al. 2021). Accordingly, in the following discussion we present two hypotheses to account for our findings: the first assumes that the elongasome is non-functional, whereas the second considers the possibility that the elongasome remains functional, but is stalled (or trapped).

Irrespective of whether the elongasome is rendered non-functional on removal of MreB, it is necessary to explain the observed heterogeneity of cell wall synthesis. Assuming, firstly, a non-functional elongasome, then obvious candidates for cell wall synthesis are the aPBPs, Pbp1A and Pbp1B, that can function independently of the elongasome (Cho et al. 2016) (although depend on activation by LpoA and LpoB respectively), with Pbp1B being particularly important in repair of cell wall lesions (Vigouroux et al. 2020). Perhaps heterogeneity of cell wall synthesis in ΔmreB arises from attempts by Pbp1B to repair large scale lesions, but there being insufficient bound (active) Pbp1B to fully accomplish this.

Assuming that the peptidoglycan-synthesising activity of the elongasome remains active on deletion of mreB, then heterogeneity of cell wall synthesis indicates that the multi-protein complex has become either stalled or trapped. This is reminiscent of observations in B. subtilis where inactivation of MreB causes failure in the dynamic redistribution of PBP1 (Kawai et al. 2009). Stalling – and concomitant heterogeneity of synthesis – could arise from elevated TPase activity that could in turn stem from altered interactions between Pbp2 and Pbp1A, which are known to mutually affect the catalytic activity of the other (Banzhaf et al. 2012; Egan et al. 2020). In support of this hypothesis, the TPase activity of Pbp1A in SBW25 ΔmreB is elevated compared to the same protein in ancestral SBW25 (Fig 4B).

In the populations derived from the 1,000 generation selection experiment, solutions to variability in cell volume stemmed largely from reduction (or complete abolition) of Pbp1A function. As predicted above and shown in Fig 3D, derived lines showed greatest improvement in the degree to which daughter cells differ in elongation rate (C_g). Such improvement correlates directly with significant increases in the probability that daughter cells are viable. By way of comparison changes in the relative positioning of the septum among daughter cells in derived lines (C_s) were relatively minor. Selection thus appears to have restored fitness by changes that reduced variation in the rate of elongation among daughter cells. Assays of TPase and peptidoglycan cross-linking indicate that this is linked to reduction of TPase function of Pbp1A and a reduction in peptidoglycan cross-linking, with the two together likely favouring a disorganized cell wall architecture suited to the topology of spherical cells (Fig 4D)

That selection compensated for the maladapted effects of mreB deletion by mutations in pbp1A can now be viewed in light of the two hypotheses for heterogeneity in cell wall synthesis. In the case of heterogeneity stemming from absence of elongasome function and localized activity of Pbp1B – and drawing upon the findings that inactivation of Pbp1A in E. coli results in a four-fold elevation of the amount of bound (and thus active) Pbp1B (Vigouroux et al. 2020) – inactivation of Pbp1A is expected to significantly increase Pbp1B activity perhaps to a level that would allow for wall repair uniformly throughout the cell (Vigouroux et al. 2020). Should heterogeneity in cell wall synthesis arise from a stalled, but otherwise functional elongasome, then elimination of Pbp1A might serve to correct the balance between TPase and GTase activity and thus allow the elongasome to move more diffusively along the inner membrane (Simpson et al. 2021). Future work is planned to test these competing hypotheses aided by the fact that deletion of mreB in SBW25 is not lethal.

The reason that compromised, albeit viable ΔmreB cells, can be generated in SBW25 under standard laboratory conditions is unknown, but answers may shed new light on the complexities of elongasome function and particularly the connection between elongasome function and MreB. Deletion of mreB in many bacteria is lethal including the closely related P. aeruginosa (Robertson et al. 2007), although in E. coli the defect can be rescued by slow growth or by increased activity of the cell division protein FtsZ (Kruse et al. 2005; Bendezú and de Boer 2008). In B. subtilis lethality can be ameliorated by growth in the presence of magnesium (Formstone and Errington 2005). This suggests that factors contributing to lethality involve subtleties in interactions among elongasome components (including associated factors such as pbp1A and pbp1B), which are disregulated in SBW25 ΔmreB, but are insufficiently out of kilter to result in a lethal phenotype.

In this context there exists possible explanations for the repeated and independent transitions from ancestral rod-shape cells to cocci (Siefert’t and Fox 1998). Recent work on the rod to coccus transition in nasopharyngeal pathogens of the Neisseriaceae family, where deletion of mreB and other determinants of the Rod complex, while not lethal, result in a significant fitness decrease (Veyrier et al. 2015). Such a fitness reduction was not apparent when coccoid cells arose from deletion of the FtsZ ring assembly-promoting protein YacF (ZapD), suggesting the requirement for a series of events beginning with loss of yacF followed by evolutionary compensation with eventual loss of the elongasome machinery (Veyrier et al. 2015). Our work here shows that the evolutionary route from rod to sphere is readily achievable under laboratory conditions via loss of mreB followed by compensatory mutation in pbp1A and reorganization of peptidoglycan structure. Whether such a route is ever realizable in nature will depend on the initial fitness effects of the mreB mutation, which in turn will depend on environmental context, but it is not implausible that inactivation of mreB may confer a fitness advantage under some conditions. This argument is further elaborated by Yulo & Hendrickson (Yulo and Hendrickson 2019).

While fitness in the majority of experimental lines was restored by mutations in pbp1A, one line, Line 5, contained a mutation in ftsZ, but oddly this mutation, while apparent at generation 500, was undetected by population sequencing at generation 1,000 (Supp Table 1). Interestingly, the FtsZ D176A mutation occurred at precisely the same position as was found in a separate selection experiment in which compensatory mutations that ameliorate harmful fitness effects associated with deletion of rodA were sought (D. W. Rogers, E. Franceschini & P. B. Rainey, unpublished). The prevalence of pbp1A mutations in the work reported here likely reflects the fact that loss-of-function mutations in pbp1A are more readily achieved compared to gain-of-function mutations in ftsZ. This is strongly evident in screening of mutations in pbp1A at generation 50, where all lines carried at least one pbp1A mutation (Supp Table 2). It will be of interest to determine whether ftsZ mutations that restore fitness to ΔrodA cells also compensate for defects due to elimination of mreB.

Derived Line 7, which carries a five gene deletion spanning the outer membrane porin OprD – and with the possibility of additional effects via mutation in ftsE – provides a contrasting route for compensatory evolution. Although oprD has not been specifically deleted, it seems reasonable to predict that in the absence of the porin, influx of water is likely limited thus damping changes in turgor pressure and reducing cell swelling. Curiously, deletion of the focal five genes from SBW25 leads to incomplete septation (Fig 4C and Supp Fig 15F), with an indication from time-resolved movies of ΔmreB ΔPFLU4921-4925 (Supp Movie 3) that there is also fault with septation in this background. This suggests that while ΔPFLU4921-4925 contributed to rescue of ΔmreB, it came at a cost to septum formation, which was subsequently rescued by mutation in the transmembrane divisome protein FtsE.

In conclusion, we return to the possibilities for new understanding of complex cellular processes arising from use of natural selection to rescue genetically perturbed cells (Lind et al. 2019; Remigi et al. 2019; LaBar et al. 2020). Here, a fitness-compromised, but viable, mreB deletion mutant of P. fluorescens SBW25 provided opportunity to explore consequences of both the initial genetic lesion and the bases of fitness compensation. In part attributable to the fact that ΔmreB SBW25 is viable, it was possible to glimpse functional effects arising from deletion of MreB, namely, uneven rates of growth among daughter cells caused by heterogeneity in patterns of cell wall synthesis. Restoration of fitness was readily achieved by mutations that reduced (or eliminated) Pbp1A function or increased septation. The primary effect was to reduce variance in growth rate among daughter cells, which is connected to resumption in homogeneity of cell wall synthesis. Disentangling two explanatory hypotheses – one invoking a role for Pbp1B and the other that elongasome function in SBW25 depends on a correct balance of among cell wall synthesizing enzymes – awaits future work.

Acknowledgements

We thank Olin Silander for helpful discussions and for his assistance to PRY with principal components analysis of cell shape, Sebastian Schmeier and Saumya Agrawal for bioinformatics analysis, Dave Rogers for valuable discussion and both Dave Rogers and Tim Cooper for comments on the manuscript. PBR is especially grateful to Malavika Venu, Norma Rivera and Dave Rogers who made the ΔmreB Δpbp1A mutant and Ellen McConnell who introduced fluorescent markers and measured fitness. Electron Microscopy was provided by Massey University and was performed by Niki Murray, Manawatu Microscopy and Imaging Centre, Massey University, Palmerston North, NZ. PBR thanks the Royal Society of New Zealand (James Cook Fellowship) and acknowledges generous core support from the Max Planck Society. Research in the Cava lab was supported by the Swedish Research Council (2018-02823 and 2018-05882), Umeå University, the Knut and Alice Wallenberg Foundation and the Kempe Foundation (SMK2062). Open access funding is provided by Max Planck Society. PBR dedicates the paper to the memory of Andrew Spiers, colleague and ardent wrinkliologist, whose 2002 manuscript first reported the fact that a transposon insertion in mreB abolishes wrinkly morphology, but not viability.

Author contributions

PRY: Genetics, construction of SBW25 GFP, phase contrast and fluorescence (BADA) microscopy, flow cytometry, growth assays, data collection / analysis, protein modeling, and manuscript review. ND: microscopy, data collection / analysis, manuscript preparation. MG: Construction of SBW25 ΔmreB, selection experiment and fitness measures. BRR: Peptidoglycan analysis, Bocillin-FL PBP labeling assays, data analysis and manuscript review. ADF: Sequencing and data analysis, manuscript review. XXZ & YH: Genetics and mutation construction and manuscript review. MM: Mutation construction, aztreonam analysis. FC: Supervision, funding acquisition, data analysis and manuscript review. PBR: Experimental conception and design, funding acquisition, microscopy, data analysis, manuscript preparation. HH: DNA extraction and sequencing, genomic analysis, supervision of PRY, and manuscript preparation.

Supplementary Material

Supplementary Table 1.

Supplementary Table 2.

Supplementary Figure 1.

Supplementary Figure 2.

Supplementary Figure 3.

Supplementary Figure 4.

Supplementary Figure 5.

Supplementary Figure 6.

Supplementary Figure 7.

Supplementary Figure 8.

Supplementary Figure 9.

Supplementary Figure 10.

Supplementary Figure 11.

Supplementary Figure 12.

Supplementary Figure 13.

Supplementary Figure 14.

Supplementary Figure 15.

Supplementary Figure 16.

Supplementary Figure 17.

Supplementary Figure 18.

Supplementary Figure 19.