Nutrient agar and broth (Oxoid) were used for B. subtilis growth. All strains used in this work are listed in Table 1. DNA manipulations and transformations were carried out using standard methods. For selections of transformants, cells were plated on NA plates supplemented with 1 μg/mL erythromycin, 5 μg/mL chlroramphenicol, 60 μg/mL spectinomycin, 1 μg/mL phleomycin, 2.5 μg/mL kanamycin and/or 10 μg/mL tetracycline. IPTG, xylose or Mg²⁺ was supplemented, as appropriate. ΔmreB and Δmbl mutants were selected on NA containing added 10 mM MgSO₄. ΔrodA mutant was selected on isotonic NA plates, composed of 2x magnesium-sucrose-maleic acid (MSM) pH7 (40 mM magnesium chloride, 1 M sucrose, and 40 mM maleic acid) mixed 1:1 with 2x NA. The concentration of kanamycin or tetracycline was increased to 10 μg/mL or 30 μg/mL in the presence of added Mg²⁺. Since transformation in Δ4 mutant cells was less effective, an temperature sensitive unstable pLOSS-P_spac-ponA plasmid was used to improve the transformation efficiency by providing ponA expression with IPTG in Δ4 mutant backgrounds (Claessen et al., 2008; Emami et al., 2017). Antibiotics were added to media for growth of strains carrying Ωerm-∆cwlO, Ωerm-P_spac-ispA, pLOSS-P_spac-ponA, Ωkan-P_spac-rodA or Ωkan-P_spac-ftsW at the following concentrations: 1 μg/mL erythromycin, 60 μg/mL spectinomycin or 2.5 or 10 μg/mL kanamycin.

L-form growth was induced on osmoprotective plates (NA containing MSM) in the presence of 200 μg/mL D-cycloserine (DCS) or genetic inhibition of PG precursor synthesis at 30°C. 1 μg/mL FtsZ inhibitor, 8 J (Adams et al., 2011), was used in L-form culture to prevent the growth of walled cells when required. IPTG or xylose was supplemented, as appropriate. For L-form escape experiments in liquid media, walled cells were pre-cultured in NB at 37°C. The exponentially growing cells (OD₆₀₀ 0.2–0.3) were diluted in the fresh osmoprotective medium (NB containing MSM) with 200 μg/mL D-cycloserine (DCS) or genetic inhibition of PG precursor synthesis, and incubated without shaking at 30°C.

For snapshot live cell imaging, walled cells were mounted on microscope slides covered with a thin film of 1.2% agarose in water, essentially as described previously (Glaser et al., 1997). L-form cultures were mounted on microscope slides covered with MSM. All microscopy experiments were conducted using a Nikon Ti microsope equipped with a Nikon CFI Plan Apo DM Lambda x100 oil objective and a Photometrics Prime camera, using MetaMorph software (version 7.7, Molecular Devices). Images were analysed and processed using FIJI.¹

We previously found that L-form escape following the inhibition of de novo PG precursor synthesis requires aPBPs, but not RodA (Figure 1A; Kawai et al., 2018). Figure 1B shows growth and morphology of otherwise wild type B. subtilis in the presence of an ispA* mutation (xseB::Tn-kan) that prevents IspA synthesis (Mercier et al., 2013). We used ispA mutations in all of the L-form experiments in this work so as to enable stable L-form growth by reducing the generation of toxic ROS (Kawai et al., 2015). The cells were cultured on isotonic nutrient agar (NA) plates containing osmoprotective sucrose and extra Mg²⁺ with (lower panels) or without (upper panels) D-cycloserine (DCS), which inhibits the PG precursor pathway (Figure 1A). Under these conditions, deletion of all four aPBPs (“Δ4 mutant”) prevented the switch to the L-form state (Figure 1C). In contrast, L-form growth occurred significantly in the absence of RodA (Figure 1D, lower panels), whereas it resulted in a spherical mode of growth in the walled state (upper panels) (RodA is normally essential in the walled state but growth can be rescued by the addition of osmolytes) (e.g., sucrose and Mg²⁺ to standard culture media) (Kawai et al., 2011).

We tested for the effects of repression of other morphogenic proteins required for cell elongation or division (brought about by repression of these genes under IPTG or xylose control). Repression of the division specific PG synthase FtsW did not affect escape (Figure 2, FtsW lower left). L-form escape also occurred efficiently in the absence of the division complex organiser FtsZ, or of the elongation organiser MreB, whether it was eliminated along with its “partners” (MreC and MreD) or its isoforms (Mbl and MreBH) (Figure 2, lower panels). In contrast, the combined absence of LytE and CwlO inhibited L-form growth (Figure 2, lower right), as previously shown (Kawai et al., 2018). All of these mutations abolished growth under selection for the walled state (without added sucrose and Mg²⁺) (Figure 2, OFF, middle panels), as expected. Thus, L-form escape seems to depend specifically on the aPBP pathway and the autolytic activity of either LytE or CwlO.

We have previously shown that L-form escape induced by DCS occurs via a gradual bulging of the walled cell before the L-form emerges (Figure 3A; Kawai et al., 2018). Our previous work has also shown that the aPBP system and the autolytic activity of either LytE or CwlO are needed for the bulging and L-form escape in the presence of DCS (Figure 3B). The bulging can also be induced by repressing genes acting in the PG precursor pathway, for example, the murE operon (Figure 3B, MurE OFF) (Kawai et al., 2018). GlmM protein catalyses a key step in the synthesis of uridine 5′-diphospho-N-acetylglucosamine (UDP-GlcNAc), which is an essential intermediate in the lipid II pathway (Figure 1A). The repression of GlmM again induced bulging (Figure 3B, GlmM OFF). Crucially, under each of these conditions, bulging of the walled cells was blocked by deletion of the aPBP genes (Figure 3B, lower panels). Thus, it seems that under conditions of limiting PG precursors, the aPBP pathway takes precedence over the Rod pathway, leading to dispersed PG synthesis and bulging.

Previous reports showed that upregulation of rodA can ameliorate growth deficiencies in walled cells caused by the inactivation of aPBPs under certain conditions (e.g., moenomycin treatment, which blocks aPBP GTase activity) (Meeske et al., 2016; Emami et al., 2017). Consistent with this, overproduction of RodA, using a second copy of the rodA gene under xylose control at an ectopic locus (amyE::P_xyl-rodA), improved the slow growth phenotype that is typical of Δ4 (aPBP) mutants (Figure 4A). We wondered if rodA upregulation might also rescue the defect in L-form escape of a Δ4 mutant. To test this we built a derivative of the Δ4 (ispA repressible) strain containing the extra copy of rodA (P_xyl-rodA). We grew this strain with and without xylose and then streaked the cells on L-form selective plates containing DCS. Cells grown with overproduced RodA gave rise to an abundance of L-form colonies (Figure 4B, middle), whereas no growth occurred for the uninduced cells (Figure 4B, left). Thus, RodA overproduction works not only to compensate for the loss of aPBPs in growth of walled cells but also for L-form escape.

PG synthesis by the RodA GTase depends on the TPase activity of its cognate bPBPs, PbpA and/or PbpH (Wei et al., 2003; Dion et al., 2019; Sjodt et al., 2020). The TPase activity of bPBPs, like that of aPBPs, is targeted by β-lactam antibiotics (Figure 1A; Lovering et al., 2012). If the rescue of L-form growth in the absence of aPBPs by RodA overproduction worked by compensating for the loss of PG synthesis, the effect should be blocked by the broad targeting β-lactam penicillin G. This was indeed the case (Figure 4B, right). Thus, escape by this pathway requires not only increased levels of the RodA GTase but also at least one of the cognate bPBPs, and probably other components of the Rod elongation system.

We introduced mutations of the elongation organiser MreB and its isoforms (ΔmreB, Δmbl and ΔmreBH) separately into a Δ4 strain carrying an ectopic copy of P_xyl-rodA. No strong effect of those deletions on normal walled growth was observed on isotonic L-form plates, that contain excess Mg²⁺ (Figure 4C, upper panels) (The growth deficiency of mreB and mbl mutants under standard conditions [e.g. NA] can be rescued by the addition of high concentrations of Mg²⁺ (Formstone and Errington, 2005; Leaver and Errington, 2005; Schirner and Errington, 2009) In the presence of DCS, although no significant effect on L-form growth was seen in the ΔmreBH strain, deletion of either of the “major” MreB encoding genes, mreB or mbl, impaired the emergence of L-form colonies (Figure 4C, lower panels), supporting the idea that the escape in a Δ4 RodA overproducing background requires function of the Rod elongation complex.

To confirm if RodA overproduction could work to stimulate L-form escape without the need for aPBPs, we cultured Δ4 mutant cells carrying an ectopic P_xyl-rodA construct in liquid nutrient broth (NB) with or without xylose, and then diluted those cells in isotonic NB containing DCS. In the absence of xylose, cells retained their cylindrical shape for at least 60 min but also began to die, as indicated by both phase pale appearance (lysis, arrowheads) and extreme nucleoid condensation (Figure 5, −Xylose). This is consistent with no appearance of L-form colonies of Δ4 mutant cells on L-form selective plates (Figures 1C, 4B). When RodA was overproduced, most cells again retained their cylindrical shape but in this case, neither lysis nor nucleoid condensation occurred, at least during the first hour (Figure 5, +Xylose/60 min). However, after further incubation (> 120 min), L-forms, began to emerge from the still cylindrical parental cells (Figure 5 i), consistent with the L-form growth on plates shown in Figure 4B. Figure 5 ii shows an example of a cell in which the nucleoid appears to have been passing out of the parental shell in association with L-form emergence. These results suggest that two distinct pathways of L-form emergence exist, depending on whether PG synthesis is being supported by the aPBP or RodA PG synthases. Presumably, in both cases the crucial event could be the formation of cell wall lesions large enough to accommodate the exit of an L-form containing an intact nucleoid. This also suggests that RodA overproduction can enable the elongation PG system to function at the reduced PG precursor availability exerted by DCS treatment, and that this can facilitate the autolytic activity needed for L-form escape.

Given that L-form escape is dependent on the aPBPs PG synthases, as well as the autolytic enzymes, as described above (Figures 1C, 2), autolytic enzymes could be also crucial in L-form escape by the Rod system in the absence of aPBPs. Since LytE and CwlO were obvious candidates, we introduced ΔlytE and ΔcwlO mutations separately into a Δ4 strain that carried an IPTG inducible copy of the major aPBP gene, ponA. Figure 6A shows that the single deletion of cwlO had little if any effect on cell growth, with or without expression of ponA (left panels). In sharp contrast, however, repression of ponA in the strain with a deletion of lytE severely impaired growth (Figure 6A, right panels), suggesting that LytE has a more important role than CwlO during growth in cells lacking aPBPs. This would be consistent with the suggestion that LytE function is associated with the Rod system (Carballido-Lopez et al., 2006; Patel et al., 2020).

We then tested whether L-form escape also required the LytE. (Note that, during the course of experiments, we found that the growth deficiency of a ΔlytE Δ4 mutant was suppressed in the presence of extra Mg²⁺ (Figure 6B) as seen for various elongation mutants (Formstone and Errington, 2005; Leaver and Errington, 2005; Schirner and Errington, 2009; Kawai et al., 2011), and taking advantage of this we generated a ΔlytE Δ4 strain carrying an ectopic copy of P_xyl-rodA) Figure 6C shows that the deletion of lytE largely prevented the emergence of L-form colonies. In contrast, no strong effect of ΔcwlO on L-form escape was observed. The escape defect in the ΔlytE Δ4 mutant was rescued by the addition of an exogenous PG hydrolase, lysozyme, leading to the emergence of L-form colonies (Figure 6C, right), confirming a crucial role for LytE activity for the escape via the Rod pathway. Thus, just as for growth of walled cells, activity of the Rod system in supporting L-form escape was largely dependent on LytE, rather than the CwlO autolysin when the aPBP system did not operate.