A regulatory pathway that selectively up-regulates elongasome function in the absence of class A PBPs

Bacteria surround themselves with peptidoglycan, an adaptable enclosure that contributes to cell shape and stability. Peptidoglycan assembly relies on penicillin-binding proteins (PBPs) acting in concert with SEDS-family transglycosylases RodA and FtsW, which support cell elongation and division respectively. In Bacillus subtilis, cells lacking all four PBPs with transglycosylase activity (aPBPs) are viable. Here, we show that the alternative sigma factor σI is essential in the absence of aPBPs. Defects in aPBP-dependent wall synthesis are compensated by σI-dependent upregulation of an MreB homolog, MreBH, which localizes the LytE autolysin to the RodA-containing elongasome complex. Suppressor analysis reveals that cells unable to activate this σI stress response acquire gain-of-function mutations in the essential histidine kinase WalK, which also elevates expression of sigI, mreBH and lytE. These results reveal compensatory mechanisms that balance the directional peptidoglycan synthesis arising from the elongasome complex with the more diffusive action of aPBPs.


Introduction
Nearly all bacterial cells are surrounded by a peptidoglycan (PG) cell wall that provides a protective barrier, helps resist cell swelling and lysis under hypoosmotic conditions, and contributes to cell shape determination (Egan et al., 2020;Zhao et al., 2017). PG functions as a large, covalently linked macromolecular enclosure and is actively remodeled to allow cell growth and division. The basic processes of PG synthesis are broadly conserved, and the detailed pathways are well documented. PG synthesis initiates with the diversion of sugars from central metabolism to form the two amino-sugars, N-acetylglucosamine (NAG) and N-acetylmuramic acid (NAM), and the incorporation of amino acids to form the stem peptide (Barreteau et al., 2008). The ultimate product of these cytosolic reactions is lipid II, a disaccharide pentapeptide precursor unit linked to an undecaprenyl pyrophosphate carrier lipid (van Heijenoort, 2007). Lipid II is flipped across the membrane (Sham et al., 2014;Meeske et al., 2015) where it interacts with two key enzymatic activities to assemble the PG layer: a transglycosylase (TG) function joins the disaccharide unit to form long, linear chains of alternating NAG-NAM residues, and a transpeptidase (TP) activity crosslinks a subset of the pentapeptide side chains to link the glycan strands together. Crucially, insertion of new glycan strands requires endopeptidases that can cleave existing crosslinks to facilitate cell wall expansion (Singh et al., 2012;Hashimoto et al., 2012;Do et al., 2020).
Most bacteria require PG for survival, except under very specific conditions (Claessen and Errington, 2019). This, combined with the absence of PG in eukaryotes, makes PG synthesis and stability an excellent target for antibiotics. One class of PG-targeting antibiotics, the beta-lactams, account for more than 60% of the global market (Klein et al., 2018). Beta-lactam antibiotics interfere with PG synthesis by covalently modifying penicillin-binding proteins (PBPs), named for their affinity for the first widely used member of this drug family. All PBPs have TP activity, and beta-lactams mimic the substrate of the transpeptidation reaction (Tipper and Strominger, 1965). Many PBPs also have TG activity, and these bifunctional PBPs are designated class A PBPs, or aPBPs . Other PBPs, designated bPBPs, only have TP activity, and must work in coordination with enzymes that provide TG activity (Wei et al., 2003;Taguchi et al., 2019;Rohs et al., 2018;Ö zbaykal et al., 2020).
While the basic outline of PG assembly has been understood for decades, the last few years have seen major strides in our understanding of how PG synthesis is coordinated in time and space (Zhao et al., 2017;Egan et al., 2020). Moreover, PG synthesis can be regulated as a function of cell growth, division, nutritional status, and in response to externally imposed stresses such as the action of antibiotics (Delhaye et al., 2019;Typas et al., 2012;Helmann, 2016). B. subtilis has been a leading model system for understanding PG synthesis in rod-shaped, Gram-positive bacteria. Seminal work in this system established, for example, that the sites of PG synthesis during cell elongation seem to be correlated with cytoskeletal filaments assembled from MreB and its paralogs, MreBH and Mbl (Kawai et al., 2009). This synthesis occurs in arcs that are perpendicular to the long access of the cell and is driven by a putative complex known as the elongasome (Garner et al., 2011). Cell division, in contrast, occurs at mid-cell during vegetative growth and is directed by a different cytoskeletal filament, FtsZ, in a complex called the divisome (Mahone and Goley, 2020). In early models, it was suggested that the major aPBP, PBP1 (encoded by the ponA gene), shuttled between the elongasome and divisome to provide the needed TG and TP activities (Claessen et al., 2008). However, bPBPs clearly also play important roles in synthesis (Wei et al., 2003). The composition and dynamic nature of these complementary systems has been subject of intensive study.
A key finding that challenged our understanding of PG synthesis in B. subtilis was the observation that a strain lacking all four known aPBPs was viable and still synthesized an apparently normal PG layer . This implied that there must be another protein with TG activity and, unlike aPBP-associated TG activity, this activity was insensitive to inhibition by moenomycin (MOE). MOE, like many PG synthesis inhibitors, activates the s M stress response (Mascher et al., 2007). Moreover, sigM null mutants are highly MOE sensitive (Mascher et al., 2007), which suggested that the missing TG might be part of the s M regulon. Indeed, the elongasome-associated TG has been identified as the SEDS family protein RodA (Meeske et al., 2016;Emami et al., 2017), a known member of the s M regulon (Eiamphungporn and Helmann, 2008;Helmann, 2016). A RodA paralog, FtsW, provides TG activity in the context of the divisome (Taguchi et al., 2019;Liu et al., 2018).
Our current understanding of PG synthesis during cell elongation in B. subtilis suggests that the bulk of synthesis is provided by the elongasome, with RodA serving as TG and PBP2a and PbpH, and perhaps also aPBPs, serving as TP (Emami et al., 2017;Meeske et al., 2016). This action is directional, largely oriented perpendicular to the long cell axis, and is balanced by a more diffusive activity of aPBPs (Dion et al., 2019;Vigouroux et al., 2020). Cells that rely exclusively on the elongasome for growth are longer and thinner, whereas those that rely predominantly on aPBPs tend to be wider and shorter (Dion et al., 2019). Many PG synthesis inhibitors activate the s M regulon, and this leads to elevated expression of many key PG biosynthetic enzymes (MurB, Amj, BcrC), elongasome components (MreB, RodA, MreCD), and the major aPBP (PBP1) (Eiamphungporn and Helmann, 2008;Helmann, 2016). However, some antibiotics may act selectively on the aPBPs or the elongasome, and it is less clear how cells might act to balance these two biosynthetic activities.
Here, we sought to define pathways important for fitness in cells that rely exclusively on the elongasome for cell elongation. We demonstrate that cells lacking aPBPs, or even just PBP1 (ponA), require a regulatory pathway that selectively increases expression of elongasome-associated proteins. Specifically, DponA mutant cells are unable to grow in the absence of s I , which induces transcription of genes encoding MreBH and an associated autolysin, LytE. Factors that facilitate s I activity, including the RasP intramembrane peptidase and its regulator EcsAB, are therefore also essential under these conditions. Further support for the importance of MreBH and LytE derives from analysis of a suppressor mutation that activates the WalKR two-component system, and thereby also restores viability to a DrasPDponA double mutant by up-regulating these same elongasome components. These results suggest that the s I stress response acting in concert with the WalKR system helps to maintain balanced activity of the elongasome and the aPBPs during cell elongation.

Results
The EcsAB-RasP pathway is essential in the absence of class A PBPs Bacteria often use overlapping or redundant systems to sustain essential pathways such as PG synthesis. To identify genes with significant roles in elongasome activity in B. subtilis, we constructed a strain (designated D4) lacking all four class A PBPs (aPBPs), and which therefore relies solely on the elongasome for PG synthesis during cell elongation . A Tn-Seq approach was employed to identify genes essential in the D4 strain but not in the wild-type (WT) background. We identified the ecsAB operon as having numerous mariner transposon insertions in WT, but very few in the D4 strain ( Figure 1-figure supplement 1). We verified conditional essentiality of ecsA by determining the plating efficiency of a clean, unmarked deletion mutant (DecsA) in a ponA depletion background in the presence and absence of the genes encoding the other 3 aPBPs (pbpD, pbpF, pbpG). Interestingly, ecsA was not only essential in the D4 background but also with depletion of ponA alone ( Figure 1A). Mutations that impair PG synthesis can often be rescued by growth on plates amended with 20 mM MgSO 4 , which leads to decreased activity of autolysins and thereby helps restore balance between PG synthesis and degradation pathways (Formstone and Errington, 2005). Indeed, an DecsADponA mutant was viable when streaked on high Mg plates, and growth was Mg-dependent ( Figure 1B).
EcsA has been designated as part of an ABC-type transporter involved in the expression and secretion of proteins (Leskelä et al., 1999). Deletion of ecsA has a profound effect on the intramembrane protease RasP, with similar phenotypes noted for the ecsA and rasP deletion mutants (Heinrich et al., 2008). Consequently, we tested whether the essential role of EcsA in the DponA strain was due to RasP. Indeed, viability of DrasPDponA, like DecsADponA, depended on high Mg concentrations ( Figure 1B). The above data highlight the importance of the EcsAB-RasP pathway in maintaining viability in the absence of aPBPs.
Mutants defective in the EcsAB-RasP pathway are sensitive to antibiotics that inhibit aPBPs Upregulation of elongasome activity is known to alleviate aPBP defects (Meeske et al., 2016). Based on the observed conditional essentiality, we hypothesized that the EcsAB-RasP pathway might functionally compensate for the absence of aPBPs. As a first test of this hypothesis, we measured sensitivity to moenomycin (MOE), a specific inhibitor of aPBP-associated TG activity (Van Heijenoort et al., 1978;Chen et al., 2019). Indeed, ecsA and rasP mutants were MOE sensitive with a four-fold decrease in minimum inhibitory concentration (MIC) relative to WT (Table 1). This was not due to a general growth defect: ecsA and rasP single mutants grew as well as WT in the absence of MOE, albeit with some lysis in stationary phase (Figure 2A), consistent with previous observations (Heinrich et al., 2008). This antibiotic sensitivity could be complemented by ectopic expression of ecsAB or rasP, respectively (Figure 2-figure supplement 1). Moreover, DecsADrasP had a similar MOE sensitivity as DrasP (Figure 2A), suggesting that the synthetic lethality of ecsA with ponA is mediated through its known downstream effect on the activity of RasP (Heinrich et al., 2008). In contrast to MOE, the DrasP and DponA mutants had a similar sensitivity as WT when tested for sensitivity to antibiotics that act on substrates common to both the elongasome and aPBP-dependent pathways of PG synthesis. For example, both nisin (Wiedemann et al., 2001) and vancomycin (Watanakunakorn, 1984) bind the common lipid II intermediate ( Figure 2-figure supplement 1). Together, these results suggest that the EcsAB-RasP pathway is critical when aPBPs are compromised, but not as a general response to inhibition of PG synthesis.
We next sought to test antibiotics that, unlike MOE, inhibit aPBPs at their TP active site. We reasoned that a stress response important for elongasome activity should also provide resistance to antibiotics that inhibit aPBPs, assuming they do not also interfere with the bPBPs essential for the elongasome. We tested 4 b-lactams (cefuroxime, oxacillin, ampicillin and penicillin G) for their inhibition profiles against DrasP and DponA strains. Oxacillin and cefuroxime (CEF) were previously suggested to preferentially inhibit aPBPs (Sassine et al., 2017;Sharifzadeh et al., 2020), whereas penicillin G preferentially inhibits bPBPs (Sassine et al., 2017). Consistently, oxacillin and CEF had highest activity against DrasP, whereas penicillin G and ampicillin had the highest activity against DponA, which encodes the major aPBP, PBP1 ( Figure 2B). These results support the idea that the EcsAB-RasP pathway functionally compensates either for the absence of aPBPs or for their chemical inhibition at either the TG (MOE) or TP (CEF) active sites.
Interestingly, the DponA mutant was actually more CEF resistant than WT. Thus, PBP1 inactivated by CEF may be deleterious to the cell. This is suggestive of futile cycling, a process in which inactivation of the TP active site leads to an ongoing generation and degradation of uncrosslinked PG strands driven by the aPBP-associated TG (Cho et al., 2014;Waxman et al., 1980). To explore this idea further, we treated WT cells with sub-inhibitory concentrations of two drugs simultaneously, MOE and CEF, that inhibit the two different active sites of the aPBP proteins. If CEF results in futile  cycling, we reasoned that MOE might antagonize this effect. In contrast, MOE and CEF together resulted in synergistic inhibition ( Figure 2-figure supplement 2). This is consistent with the same target drug synergy model, as previously described for E. coli protein synthesis inhibitors (Yilancioglu, 2019) and drugs used to treat human diseases (Jia et al., 2009), but does not support the hypothesis of CEF-dependent futile cycling.
EcsAB-RasP functions through s I to sustain cell wall synthesis in the absence of aPBPs RasP functions as an intramembrane protease for the activation of multiple stress response pathways, and our results suggest it may be important for PG synthesis when aPBPs are missing or inhibited. RasP proteolytically inactivates the anti-sigma factors RsiW (regulator of s W ) (Schö bel et al., 2004), RsiV (regulator of s V ) (Hastie et al., 2013) and RsgI (regulator of s I ) (Liu et al., 2017). In the absence of RasP, these s factors can not be activated. RasP also cleaves FtsL, a cell division protein (Bramkamp et al., 2006). To determine which of these RasP targets may contribute to elongasome activity, we took advantage of the fact that MOE and CEF selectively inactivate aPBPs. Therefore, MOE and CEF resistance provides a readout of elongasome function. We tested mutants lacking each of the three RasP-dependent sigma factors or containing D25FtsL, coding for a functional, but truncated FtsL (deleted in amino acids 2-26) variant that is not subject to cleavage by RasP (Bramkamp et al., 2006). The DecsA and DrasP mutants were 4-fold more sensitive to MOE than   WT (0.4 vs. 1.6 mg/mL), whereas for DponA the (MIC) was >1.6 mg/mL (Table 1; Figure 3-figure supplement 1). The MIC was unaffected by deletion of sigW or sigV or by the non-cleavable FtsL (1.6 mg/mL). However, the DsigI mutant was significantly more sensitive to MOE with the MIC being 0.4 mg/mL, similar to DrasP. This suggests that s I is required for optimal function of the MOE-insensitive elongasome. Similar results were observed when CEF sensitivity was monitored ( Figure 3A). Of the known RasP targets, s I contributes the most to CEF resistance. Moreover, the DsigWDsigI mutant phenocopies the DrasP mutant, suggesting that activation of s I and s W largely accounts for the role of RasP in CEF resistance. In addition, the sensitivity of the DecsA and DrasP mutants was not further increased by mutation of sigW or sigI (Figure 3-figure supplement 2), indicative of them being in the same pathway. Finally, deletion of rsgI, encoding the s I anti-sigma factor, led to a significant decrease in CEF sensitivity of the DecsA and DrasP mutants. DrsgI was more sensitive to CEF compared to WT, which may be due to increased activity of s I and its associated autolysins. In contrast, deletion of rsiW, encoding the s W anti-sigma factor, led to a much less pronounced effect (Figure 3-figure supplement 2). Thus, s I plays a dominant role in intrinsic CEF resistance, and as expected this activity relies on the RasP-dependent degradation of the RsgI anti-sigma factor.
The importance of s I in the absence of aPBPs was confirmed by determining the plating efficiency of DsigIDponA double mutant ( Figure 3B). The double mutant could survive with high Mg 2+ , but was unable to grow on LB. This synthetic lethality of the DsigIDponA and DrasPDponA strains was suppressed by ectopically expressing the sigI gene from the leaky promoter P spac(hy) . Thus, decreased s I activity can fully explain the DrasP antibiotic sensitivity phenotypes, and we therefore conclude that one or more members of the s I regulon must facilitate growth under conditions of impaired aPBP activity.

s I supports elongasome function by regulating MreBH and LytE
Next, we sought to identify the s I -dependent genes important for survival in the absence of aPBPs. Of the genes directly regulated by s I (Ramaniuk et al., 2018), five (mreBH, lytE, gsiB, fabI and bcrC) have known or likely roles related to cell envelope functions. GsiB is a general stress response protein (Michna et al., 2016) and FabI is involved in fatty acid synthesis (Heath et al., 2000). BcrC   functions in undecaprenylpyrophosphate recycling (Bernard et al., 2005;Zhao et al., 2016;Radeck et al., 2017b), and MreBH and LytE are both elongasome-associated proteins. MreBH, one of three MreB-family proteins that associate with the elongasome, sequesters and directs the LytE endopeptidase to the sites of insertion of new peptidoglycan (Carballido-Ló pez et al., 2006). To further define the role of s I in sustaining viability during aPBP inhibition, we conducted CEF/MOE sensitivity assays using single mutants of s I -controlled genes. The mreBH, lytE and bcrC single mutants exhibited slightly higher sensitivity for both CEF and MOE (Figure 4-figure supplement 1), however, they did not entirely phenocopy the sigI phenotype. The DmreBHDlytE double mutant exhibited the same level of CEF and MOE sensitivity as both the rasP and sigI mutants ( Figure 4A-  B). Thus, these results suggest that the EcsAB-RasP-s I pathway primarily acts through MreBH and LytE to control elongasome function.
To further validate the importance of MreBH and LytE, we created deletion mutants in the DponA background ( Figure 4C). A DmreBHDponA double mutant could be constructed only when the cells were initially plated on LB supplemented with high Mg 2+ . Once constructed, however, this mutant and the DlytEDponA double mutant did not exhibit a plating defect on LB. In contrast, the triple mutant of DmreBHDlytEDponA was synthetic lethal and could not be plated on LB agar without Mg 2+ supplementation. These data suggest an additive role for MreBH and LytE in the effective functioning of the elongasome, likely due to the ability of LytE to retain some function in the absence of MreBH, and MreBH having functional roles beyond localization of LytE. B. subtilis has two partially redundant D,L-endopeptidases, LytE and CwlO, which are collectively essential for cell viability (Hashimoto et al., 2012). The involvement of s I in the expression of lytE has already been established since both DlytEDcwlO and DsigIDcwlO are synthetic lethal (Salzberg et al., 2013). Consistently, DrasPDcwlO was also synthetic lethal (Figure 4-figure supplement 1). To confirm that LytE activity was reduced in the rasP and sigI mutants we evaluated the autolytic potential of the cells. Cells were treated with sodium azide, which disrupts membrane potential and activates autolysins (Jolliffe et al., 1981;Wang et al., 2014). By monitoring the time taken for a 50% reduction in optical density, we found that the DlytE mutant had a lower rate of autolysis ( Figure 4D). Similar to DlytE, we observed that DrasP, DsigI and DmreBH also had lower autolytic potential, consistent with a role in affecting LytE expression or activity.
Next, we evaluated the expression levels of mreBH and lytE in DrasP, DsigI and DponA mutants ( Figure 4E). In the DponA mutant, mreBH was significantly upregulated, whereas lytE was unchanged. In DsigI, both mreBH and lytE expression was significantly lower. This suggests that DponA cells require higher levels of MreBH to direct the autolytic activity of LytE to support optimal elongasome function, and that activation of s I mediates increased mreBH expression. As a result, the reduced expression of mreBH in DrasP and DsigI strains likely contributes to the synthetic lethality with DponA.

Balance in the MreBH-LytE activity is essential for optimal elongasome function
We complemented the conditional essentiality of mreBH and lytE by ectopically expressing each of these genes individually as well as in combination in different mutant backgrounds. These strains were used to evaluate the relative importance of each gene upon inhibition of PBP1 by monitoring their CEF resistance. Although ectopic expression of mreBH complements the CEF sensitivity of DmreBH, it is unable to restore CEF resistance to the DmreBHDlytE double mutant ( Figure 5A). However, when both mreBH and lytE were ectopically expressed, the strain was significantly more CEF resistant than WT ( Figure 5A). Similarly, induction of mreBH modestly increased CEF resistance of DrasP ( Figure 5B), but not a DrasPDlytE double mutant. Similar results were obtained in cells where pbpD, pbpF and pbpG were deleted (data not shown) indicating no indirect effect of MreBH on these aPBPs. In DsigI, however, mreBH expression alone had no significant impact on CEF resistance, perhaps due to reduced availability of LytE. Thus, increasing MreBH levels likely functions to increase elongasome activity by facilitating the localized action of LytE. Conversely, the P spac(hy) lytE overexpression construct could not be introduced into the DrasP and DsigI mutants. We speculate that high LytE, in cells that have reduced expression of mreBH, leads to delocalized and unregulated autolysin activity. Collectively, these results further support a model in which a major role of MreBH is in directing LytE to sites of ongoing, elongasome-dependent PG synthesis.
The elongasome is critical for the maintenance of rod-shape, as judged by the spherical morphology of conditional mutants that are depleted for either the RodA transglycosylase or the two class B PBPs that provide transpeptidase activity (Boylan and Mendelson, 1969;Wei et al., 2003). The maintenance of rod shape is also affected by the balance between the directional motion of the elongasome and the random diffusive motion of PBP1 (Dion et al., 2019). Any imbalance in the activities of the two systems can lead to change in cell morphology. Overexpression of MreB or other elongasome proteins leads to cells that are longer and thinner, whereas overexpression of PBP1 leads to shorter and wider cells (Dion et al., 2019). Thus, we hypothesized that the effects of the s I regulatory system (acting through mreBH and lytE) on elongasome function would be revealed by monitoring cell morphology. We imaged WT, DrasP, DsigI, DmreBH, DlytE, DmreBHDlytE and DponA cells and quantified the cell length and width using MicrobeJ (Ducret et al., 2016). Indeed, DrasP, DsigI and DmreBHDlytE mutants were significantly shorter ( Figure 5C) and wider ( Figure 5D) compared to the WT, which indicates that these cells were primarily utilizing PBP1 for PG synthesis.
DmreBH and DlytE mutants individually also had lower elongasome activity. In contrast, the DponA mutant formed significantly thinner cells due to PG synthesis being contributed mainly by the elongasome. These data support the conclusion that the rasP, sigI and mreBH-lytE genes all support elongasome function. DlytE, and DmreBHDlytE strains with and complementation by ectopic expression of genes from the leaky promoter, P spac(hy) , or (for the DmreBHDlytE strain) expression of mreBH from a xylose inducible promoter (P xyl ) and lytE from the P spac(hy) . P-value cut-off of <0.0001 was used. (B) CEF sensitivity (as for panel A) for DrasP and DsigI mutants with ectopic expression of mreBH from P spac(hy) in the presence and absence of lytE. P-value cut-off of <0.0001 was used. Cell length (C) and width (D) of WT, DponA, DrasP, DsigI, DmreBHDlytE, and DmreBH and DlytE strains was determined using at least 100 cells for each strain. P-value cut-off of <0.005 was used. The online version of this article includes the following source data for figure 5: Source data 1. Data of zone of inhibition and cell size measurements.
Suppressor analysis confirms the importance of mreBH and lytE in cells dependent on elongasome Next, we took advantage of the DrasPDponA synthetic lethality to isolate suppressors that grow on LB agar plates. Using whole-genome resequencing, we identified three strains with point mutations in walK (Ala241Asp, Ser385Leu, Asp274Ala). WalK is the sensor kinase of the essential two-component system WalKR, which regulates cell wall metabolism (Takada and Yoshikawa, 2018). WalR has binding sites upstream of sigI, mreBH and lytE and activates expression of these genes under heat stress (Huang et al., 2013). In addition to their regulation by s I , sigI and lytE also have s A -dependent promoters. WalR may function in conjunction with the s A holoenzyme, which together with s I controls lytE expression (Tseng et al., 2011). Taking into account the importance of WalKR in the expression of sigI, mreBH and lytE, we characterized one of the suppressor mutants of WalK, wherein aspartate 274 is changed to alanine (D274A).
Residue 274 lies in the cytoplasmic Per-Arnt-Sim (PAS) domain of WalK ( Figure 6A). PAS domains have been linked to signal sensing (Taylor and Zhulin, 1999) and may be involved in protein dimerization (Huang et al., 1993). Recently, the cytoplasmic PAS domain of S. aureus WalK was found to bind zinc at a site including D274. Moreover, mutation in this binding site, which is highly conserved in WalK orthologs (Monk et al., 2019), led to increased kinase activity. We therefore hypothesized that the WalK D274A suppressor (denoted as WalK*) led to increased activity of the WalKR two-component system. We used CRISPR mutagenesis to introduce the walK* allele into WT cells and then confirmed that this allele suppressed the synthetic lethality of DrasPDponA ( Figure 6B).
We next aimed to test the effect of WalK * on gene expression and cell wall homeostasis. The sigI and lytE genes can be expressed through their s A promoter after activation by WalR (Salzberg et al., 2013;Tseng et al., 2011). However, mreBH lacks an annotated s A promoter, implying that the expression of mreBH may rely on WalR activation of the s I holoenzyme. To test this hypothesis, we measured CEF sensitivity of walK*DrasP and walK*DsigI strains ( Figure 6C). Although walK* increased CEF resistance of the DrasP strain, it could not rescue the DsigI strain. This supports the idea that WalR may act in conjunction with s I to activate transcription of mreBH, and thereby augment elongasome activity. Increased activation of WalK* can lead to increased expression of not only lytE, but also cwlO (Takada and Yoshikawa, 2018). This could lead to elevated autolysin levels that might account for the higher CEF sensitivity of walK* alone compared to WT.
We further quantified the mRNA levels of mreBH and lytE in the walK* strain and in the walK*DrasP and walK*DsigI strains ( Figure 6D). The walK* allele led to increased expression of both mreBH and lytE. Moreover, these levels were similar to that observed in the DponA background, suggesting that deletion of ponA leads to a compensatory increase in mreBH and lytE mediated by the WalKR. However, they were lower for the walK*DsigI strain. These data suggest that walK* leads to increased activation of WalR, which then leads to increased transcription of sigI and thereby of mreBH and lytE. This ultimately leads to the survival of the DrasPDponA strain. These data also validate the importance of RasP and s I in the regulation of MreBH and LytE and their significant impact on elongasome activity, especially in the DponA background.
Additive role of s I and s M in regulating the elongasome activity While our results suggest a critical role for s I in aPBP-elongasome homeostasis through its regulation of MreBH and LytE, previous studies have indicated that the extracytoplasmic (ECF) sigma factor s M also plays a significant role in B. subtilis cell wall homeostasis. s M regulates the expression of rodA, mreB, mreC and mreD (core components of the elongasome), as well as ponA and other genes involved in PG synthesis (Eiamphungporn and Helmann, 2008;Luo and Helmann, 2012). To determine the relative contribution of s M to cell survival during aPBP inhibition, we used P M * mutations that selectively inactivate s M -dependent promoters of genes encoding elongasome components. We constructed the P M *rodA and P M *ponA strains that are unable to upregulate rodA and ponA, respectively, and a P M *maf strain that cannot upregulate the mreBCD genes located downstream of the intragenic P M inside maf (Eiamphungporn and Helmann, 2008). We also constructed the double mutant P M *rodA P M *maf strain. The CEF sensitivity of P M *rodA and P M *rodA-P M *maf was similar to that of the sigM mutant ( Figure 7A). Neither P M *maf nor P M *ponA were CEF sensitive. Thus, under conditions where CEF has inhibited PBP1, s M helps restore peptidoglycan synthesis primarily by increasing the expression of rodA to increase elongasome activity. In contrast, the double mutants of DecsADsigM, DrasPDsigM and DsigIDsigM revealed an additive effect with respect to CEF sensitivity ( Figure 7B). Thus, the role of the elongasome in PG synthesis can be regulated through two-independent pathways: the EcsAB-RasP-s I pathway acts by regulating MreBH and LytE, and the s M pathway acts through RodA.

Discussion
Peptidoglycan (PG) is a defining feature of bacteria. This cellular enclosure must provide stability, yet at the same time be highly dynamic and adaptable. During growth, PG is continuously remodeled, which involves the action of autolysins, hydrolytic enzymes that cleave links within and between the glycan strands (Vollmer et al., 2008;Egan et al., 2020). These hydrolases are essential for the insertion of new glycan strands into the existing structure (Hashimoto et al., 2012;Singh et al., 2012). Cell shape maintenance requires that the sites of new PG synthesis be spatially regulated, often in response to the activity of cytoskeletal filaments such as the MreB (Domínguez-Escobar et al., 2011) and FtsZ proteins (Mahone and Goley, 2020).
B. subtilis, a genetically tractable model organism, has provided an important system for investigating the pathways of PG synthesis in rod-shaped, Gram positive bacteria. During cell elongation, a multiprotein complex designated the elongasome is the primary biosynthetic machine for inserting new glycan strands. In B. subtilis, there are three MreB paralogs (MreB, Mbl and MreBH), which colocalize to form elongasome-associated cytoskeletal filaments along the cell periphery (Carballido-Ló pez et al., 2006;Garner et al., 2011). Cells lacking all three paralogs lose their rod shape and become spheres which ultimately lyse (Kawai et al., 2009). Whereas MreB and Mbl are critical for the circumferential motion of the elongasome, the role of MreBH is less clear, and seems related to its ability to recruit LytE (Carballido-Ló pez et al., 2006). PG synthesis by the elongasome relies on the activity of RodA as TG, with bPBPs providing TP activity ( Figure 8A). A separate complex, the divisome, builds the cross-walls prior to cell separation (Mahone and Goley, 2020).
Because of its unique chemical composition, PG synthesis requires numerous highly conserved enzymes, which thereby present attractive targets for antibiotics (Bugg et al., 2011). Inhibitors of PG synthesis may result in spheroplast formation, cell lysis, or morphological defects, depending on the antibiotic target and the organism (Cross et al., 2019;Emami et al., 2017). Many of our most familiar antibiotics are natural products of soil bacteria, including Bacillus spp. (Kaspar et al., 2019;Stein, 2005) and many actinobacteria (Mahajan, 2012). Like other soil bacteria, B. subtilis has Figure 8. s I co-ordinates with WalKR to regulate elongasome function, and complements the s M dependent stress response. (A) PG synthesis potential is dictated by the activity of the elongasome in collaboration with aPBPs. Cell wall stress activates s M (left), which up-regulates both pathways. In the absence of aPBPs, cells upregulate elongasome activity through s I , which increases expression of genes (mreBH and lytE) important for elongasome function. Synthetic lethal relationships are shown here between deletion of ponA and genes in the s I pathway (black circles). Bypass of synthetic lethality can be compensated by a gain of function mutation in walK (star). (B) The promoter regions of sigI, mreBH and lytE are shown, depicting the binding sites of WalR and s I as annotated before (Huang et al., 2013). s I and WalR act as activators for the expression of sigI and lytE from the s A promoter. The downstream WalR binding site is important for expression of sigI and lytE at 37˚C whereas the upstream binding site is crucial for the heat induction of these genes at 51˚C. substantial intrinsic resistance to many antibiotics (Kingston et al., 2013;Radeck et al., 2017a;Helmann, 2016). We have explored these intrinsic resistance mechanisms by analysis of cell envelope stress responses, including those controlled by alternative sigma factors (Helmann, 2016). For example, s V is induced by and provides resistance to lysozyme by covalently modifying PG (Guariglia-Oropeza and Helmann, 2011), whereas s W is induced by and provides resistance to membraneactive bacteriocins (Butcher and Helmann, 2006;Kingston et al., 2011).
The s M response is selectively induced by stresses during PG synthesis and contributes to resistance to a wide-variety of PG synthesis inhibitors, including MOE, CEF, and bacitracin (Helmann, 2016;Mascher et al., 2007). The s M regulon serves to both upregulate PG synthetic capacity, and to compensate for stresses resulting from PG inhibition. The former includes the upregulation of elongasome components ( Figure 8A) and PG biosynthetic enzymes (PBP1, Ddl, MurB, MurF, BcrC, Amj) (Eiamphungporn and Helmann, 2008). The latter includes the large regulon controlled by the Spx transcription factor that protects cells against antibiotic-associated oxidative stress (Rojas-Tapias and Helmann, 2018). Finally, it has recently been shown that induction of a s M -regulated ppGpp synthase, YwaC, increases the number of persister cells following antibiotic exposure (Fung et al., 2020).
Here, we identify a major role for another alternative sigma factor, s I , in conferring intrinsic resistance to important cell wall antibiotics, MOE and CEF. Induction of s I , which requires the EcsAB-RasP regulatory pathway (Liu et al., 2017), selectively elevates elongasome function by increasing the expression of the MreB paralog, MreBH, and the associated autolytic endopeptidase LytE (Carballido-Ló pez et al., 2006). This stress response is critical in cells lacking PBP1, as judged by the synthetic lethality of DsigI DponA mutants ( Figure 3B). This stress response functions in coordination with both the s M stress response ( Figure 7A), which increases elongasome function by upregulation of the RodA TG (Meeske et al., 2016;Emami et al., 2017), and the essential WalKR two-component system (Figures 6 and 8). Although s I was previously linked to heat-stress (Zuber et al., 2001), virulence in B. anthracis (Kim and Wilson, 2016), and control of autolysin synthesis (Salzberg et al., 2013), our results reveal new insights into its role in cell envelope stress.
This study also highlights the complex regulation of the mreBH and lytE genes. WalR, s I and s A binding sites have been previously annotated in the promoters of sigI, mreBH and lytE ( Figure 8B). The WalK (D274A) gain of function mutant suppresses the lethal phenotype of DrasPDponA by induction of mreBH and lytE ( Figure 6). However, induction was not significant in the s I mutant. We conclude that co-activation by WalR and s I is required for induction of these two genes. The signals sensed by WalK are unclear, but it was recently suggested that peptidoglycan cleavage products generated by LytE and CwlO can be sensed by WalK to balance the activity of these proteins (Dobihal et al., 2019). Moreover, it was previously observed that sigI activation enhances the growth of mbl mutants (Schirner and Errington, 2009), which we suggest was likely due to increasing elongasome activity through mreBH and lytE.
Collectively, our results reveal that WalKR and s I act in coordination to maintain optimal elongasome activity, and these pathways complement the general PG stress response activated by s M (Figure 8). One general theme that has emerged is that PG synthesis involves multiple, functionally overlapping systems, often with one being inducible by antibiotic inhibition of the other. For example, the inducible UPP phosphatase BcrC complements the activity of UppP (Radeck et al., 2017b;Zhao et al., 2016), and the s M -regulated Amj functions as a second lipid II flippase that is critical when MurJ is inhibited (Chamakura et al., 2017;Meeske et al., 2015). Similarly, inhibition of aPBPs by MOE leads to an essential, compensatory induction of RodA (Meeske et al., 2016;Emami et al., 2017). Here, it is shown that this single s M -regulated target gene can largely account for the CEF sensitivity of sigM mutants (Figure 7). This increase in RodA, together with the induction of MreBH and LytE, serves to boost the biosynthetic potential of the elongasome. These results reveal mechanisms that allow diverse PG biosynthetic complexes to coordinate their activities, in both time and space. The highly orchestrated processes that direct and coordinate PG synthesis are important both for intrinsic antibiotic resistance, as explored here and are ultimately responsible for the enormous diversity of bacterial morphologies (Caccamo and Brun, 2018).  Table. For all deletion mutants, primary strains were ordered from the BKK/BKE collection available at the Bacillus Genetic Stock Centre (BGSC) (Koo et al., 2017). These gene deletions with the antibiotic cassette (kanamycin or erythromycin) were then transformed into our WT 168 strain using natural competence induced in modified competence (MC) medium. rasP, ecsA and ponA deletion strains had very low natural competence. Thus, other mutations were introduced using SPP1 phage transduction as described (Kearns et al., 2005). The null mutants were constructed using pDR244, which removes the resistance cassette leading to clean in-frame deletions (Koo et al., 2017). The resulting gene deletions (designated D) were confirmed with check primers listed in Supplementary file 1.

Materials and methods
Genes were ectopically expressed at amyE under promoter P spac(hy) using pPL82 plasmid (Quisel et al., 2001). MreBH was also expressed at the lacA locus under xylose inducible promoter P xyl using plasmid pECE741 (Popp et al., 2017). The respective genes were amplified from genomic DNA using primers listed in Supplementary file 1. The required restriction enzyme sites (and if required a ribosome binding site (RBS)) were incorporated in the primers used for gene amplification. CRISPR-Cas9 mutagenesis was carried out using pJOE8999 plasmid as described before (Altenbuchner, 2016). The primers used to construct the repair fragment and guide RNAs are in Supplementary file 1. The whole sequence of the genes was confirmed by Sanger sequencing (Biotechnology Resources core facility at Cornell University).

Transposon mutagenesis
The transposon-sequencing (Tn-Seq) was performed using modified pMarA (Le Breton et al., 2006). pMarA is a plasmid harboring the mariner-Himar1 transposase gene and a temperature-sensitive replicon to select for transposition events. Two MmeI sites were introduced flanking the BstXI and PstI sites to generate plasmid pMarA1 (HE8334). The plasmid was transformed into WT Bacillus subtilis and DpbpDFG ponA::erm mutant at 28˚C selecting for Kan R on LB plates supplemented with 10 mM MgSO 4 (final concentration) to generate strain HB20725 and HB20738, respectively. Liquid cultures of HB20725 and HB20738 harboring plasmid-borne transposons were grown at 28˚C in liquid LB medium with 10 mM MgSO 4 to mid-exponential phase (OD 600~0 .4), diluted and spread on LB plates containing kanamycin and 10 mM MgSO 4 . Plates were incubated overnight at 48˚C to select for transposition events, and the ones with distinct single colonies (not too crowded, and about 500 colonies per plate) were pooled together. Two hundred and forty plates with a total of >100,000 independent colonies were pooled together for each strain, and their genomic DNA was isolated. For each strain, 10 mg of genomic DNA was digested using MmeI, purified and ligated with sequencing adaptors. Illumina sequencing was performed and DNA adjacent to the transposon insertion sites were matched to Bacillus subtilis reference genome NC_000964.3 using CLC workbench version 8.5.1. Matching results were visualized using CLC workbench, and quantified using Tn-seq Explorer software (Solaimanpour et al., 2015). For visualization of transposon insertions, IGV genome browser was used (Robinson et al., 2011).

Plating efficiency
For plating efficiency (spot dilution) assays, the cultures were grown in LB medium with 20 mM MgSO 4 to~0.4 OD 600 . 1 mL of culture was centrifuged at 5000 rpm for 5 min and resuspended in LB medium (without MgSO 4 ). 10-fold serial dilutions were done in LB medium and 10 mL was plated/ spotted on LB agar plates, allowed to air-dry for 10-15 min, and incubated overnight at 37˚C.

Growth kinetics and MIC determinations
Cultures were grown in LB medium to~0.4 OD 600 . 1 mL of this culture was inoculated in each well containing 200 mL of LB media with the required drug concentration. Honeycomb 100-well plates were used for the assay. The increase in the OD 600 of the culture was monitored real-time using Bioscreen C growth curve analyzer (Growth curves USA). Readings were taken at every 15 min interval up to 24 hr under constant shaking conditions at 37˚C. For MIC determination, two-fold increase in the drug concentration was screened ranging from (0.2 to 1.6 mg/mL). The minimum concentration which inhibited the growth (less than 0.2 OD 600 ) up to at least 10 hr of incubation was considered as the MIC for the strain.

Disc diffusion assays
Antibiotic sensitivity was screened by determining the zone of inhibition using a disc diffusion assay. Cultures were allowed the grow up to~0.4 OD 600 . 100 mL of this culture was added to 4 mL of top agar (0.75% agar) kept at 50˚C to prevent it from solidifying. This was poured on to 15 mL LB agar plates (1.5% agar). The top agar was allowed to air-dry for 30 min. A Whatmann paper filter disc of 6 mm was then put on the top agar. The required amount of drug was added on the disc immediately. The plates were incubated overnight at 37˚C and the diameter of the clear zone of inhibition was measured. For all histograms, the zone of inhibition (Y-axis) starts from 6 mm which is the disc diameter. For strains having the inducible promoter P xyl , both the top agar and LB agar plates were made with 0.1% xylose.
Autolytic potential 200 mL of cells (~0.4 OD 600 ) were added in each well of a 100-well honeycomb plate. To this, 0.05 M of sodium azide (from 5 M stock) was added. Immediately, the real-time monitoring of the decrease in OD 600 was started with Bioscreen C. Readings were taken every 15 min for up to 24 hr. The time at which 50% of the cells had lysed was noted for each mutant. The time taken (in hours) was plotted as lysis time for each strain. Sodium azide stock was prepared fresh before every experiment.

Real-time PCR
Gene expression for mreBH and lytE was determined by real-time PCR using primers in Table S2. RNA was purified from 1.5 mL of~0.4 OD 600 cells using the RNeasy Kit from Qiagen as per the manufacturer's instructions. 2 mg of RNA was used to prepare 20 mL of cDNA to achieve a final concentration of 100 ng/mL using High capacity cDNA reverse transcription kit from Applied Biosystems. The gene expression levels were measured using 100 ng of cDNA using 0.5 mM of gene specific primers and 1X SYBR green (Bio-Rad) in CFX connect real-time system from Bio-Rad. gyrA was used a house-keeping gene. Gene expression values (2 -Dct ) were plotted after normalization with gyrA.

Cell size measurements
A very thin agar pad was prepared on slides from 0.8% agarose. 10 mL of cells (~0.4 OD 600 ) were spotted and allowed to air dry for 10 min before putting on a cover slip. Cells were imaged using Olympus BX61. Images were captured using Cooke Sensicam camera system under 100X magnification with immersion oil. The images were then analyzed for their length and width using MicrobeJ (Ducret et al., 2016), a plugin for imageJ (Schneider et al., 2012).

Suppressor analysis
Spontaneous suppressors were picked from LB agar plates for DecsADponA and DrasPDponA. 12 suppressors were selected from each background and their chromosomal DNA extracted using Qiagen DNA extraction kit. DNA was sequenced using the Illumina platform at the Biotechnology Resources core facility at Cornell University. The results were trimmed, mapped and aligned with the DecsADponA and DrasPDponA backgrounds using CLC genomics workbench.

Statistical analysis
All the experiments were performed with a minimum of 3 biological replicates. For microscopy images, at least 100 cells per strain were quantified for their cell length and width. One-way ANOVA was used to calculate the statistical significance. Tukey's comparison test was used to determine significance between all the strains. P-value cut-offs have been mentioned in the figure legends. Different letters represent data which are significantly different. Same letter represents mean values which are not statistically different. Significance between two strains was determined using student's t-test.