Homeostatic control of cell wall hydrolysis by the WalRK two-component signaling pathway in Bacillus subtilis

Bacterial cells are encased in a peptidoglycan (PG) exoskeleton that protects them from osmotic lysis and specifies their distinct shapes. Cell wall hydrolases are required to enlarge this covalently closed macromolecule during growth, but how these autolytic enzymes are regulated remains poorly understood. Bacillus subtilis encodes two functionally redundant D,L-endopeptidases (CwlO and LytE) that cleave peptide crosslinks to allow expansion of the PG meshwork during growth. Here, we provide evidence that the essential and broadly conserved WalR-WalK two component regulatory system continuously monitors changes in the activity of these hydrolases by sensing the cleavage products generated by these enzymes and modulating their levels and activity in response. The WalR-WalK pathway is conserved among many Gram-positive pathogens where it controls transcription of distinct sets of PG hydrolases. Cell wall remodeling in these bacteria may be subject to homeostatic control mechanisms similar to the one reported here.


Introduction
The cell wall peptidoglycan (PG) is composed of long glycan strands cross-linked together by short peptides. This three-dimensional exoskeleton specifies shape and protects the cell from osmotic rupture. For cells to grow they must enlarge this covalently closed macromolecule and this requires both the synthesis of new material and hydrolysis of the PG meshwork to allow for its expansion. How the cell maintains the appropriate levels of these potentially autolytic enzymes remains unclear. Here, we report that the WalR-WalK (WalRK) two-component signaling pathway functions in the homeostatic control of the cell wall hydrolases required for expansion of the PG during growth.
The WalRK two-component system (TCS) was discovered over two decades ago and is among the most broadly conserved TCS in Firmicutes (Fabret and Hoch, 1998;Dubrac et al., 2008a;Takada et al., 2018). WalK (also referred to as YycG, VicK, or MicA) is a membrane-anchored sensor kinase and WalR (also called YycF, VicR, or MicB) is a DNA binding response regulator of the OmpR family (Fabret and Hoch, 1998;Dubrac et al., 2008a;Dubrac and Msadek, 2008b;Okajima et al., 2008). In most bacteria that encode this TCS, two additional genes, walH (yycH) and walI (yycI), reside in an operon with them. In B. subtilis, WalH and WalI are negative regulators of the WalK sensor kinase and the three integral membrane proteins assemble into a multimeric complex (Szurmant et al., 2005;Szurmant et al., 2007;Szurmant et al., 2008). The WalRK system is essential in most Firmicutes, making it an attractive antibiotic target (Barrett and Hoch, 1998;Gotoh et al., 2010). Accordingly, the WalRK pathway has been extensively studied in B. subtilis as well as several important Gram-positive pathogens. In all cases where it has been examined, the WalR regulon contains genes encoding cell wall hydrolases (Bisicchia et al., 2007;Howell et al., 2003;Ahn and Burne, 2007;Ng et al., 2005;Liu, 2006;Dubrac et al., 2007). Furthermore, cells engineered to constitutively express a subset of these enzymes can bypass the essentiality of the signaling pathway (Ng et al., 2003;Delaune et al., 2011;Takada et al., 2018). These findings have led to the view that the essential role of WalRK is to coordinate cell wall metabolism with growth. However, despite two decades of research, what the WalK sensor kinases senses and how this pathway functions in cell wall homeostasis have remained mysterious.
In B. subtilis, phosphorylated WalR (WalR~P) controls the synthesis of several cell wall hydrolases; among them are two enzymes (CwlO and LytE) that are critical for cell wall elongation. Cells lacking either PG hydrolase are viable but depletion of one in the absence of the other causes a cessation of growth followed by lysis (Bisicchia et al., 2007;Hashimoto et al., 2012). Both enzymes are D,Lendopeptidases and cleave the peptide bond between the second (g-D-Glu) and third (mDAP) amino acid in the stem peptide of PG (Ishikawa et al., 1998;Yamaguchi et al., 2004). CwlO is controlled by a membrane complex composed of the non-canonical ABC transporter FtsEX and two integral membrane proteins SweC and SweD (Meisner et al., 2013;Domínguez-Cuevas et al., 2012;Brunet et al., 2019). LytE is a secreted enzyme with LysM domains that direct it to the lateral cell wall (Margot et al., 1998;Ishikawa et al., 1998;Buist et al., 2008). How the level and activity of these essential elongation hydrolases are regulated remains incompletely understood. Here, we report that the WalK sensor kinase monitors the activity of CwlO and LytE by sensing the cleavage products generated by them. In response, WalK controls WalR-dependent changes in the expression and activity of these enzymes. Thus, this essential two component system functions in the homeostatic control of PG hydrolysis required for growth. This represents the first homeostatic pathway for cell wall hydrolysis in bacteria and we propose that cell wall remodeling in related Gram-positive pathogens is subject to similar regulatory control.

LytE levels increase in the absence of CwlO maintaining cell envelope integrity
In the course of characterizing LytE protein levels in various mutant backgrounds, we discovered that LytE levels increase approximately 2-fold in the absence of CwlO ( Figure 1A). To determine whether this increase was due to changes in lytE transcription, we fused the lytE promoter to lacZ and compared ß-galactosidase activity in wild-type and cells lacking CwlO. As can be seen in Figure 1B, transcription from the P lytE promoter increased~2 fold in the DcwlO mutant. A similar increase in P lytE transcription was observed in cells lacking the FtsEX complex, which is required for CwlO activity ( Figure 1B) (Meisner et al., 2013). Furthermore, a point mutation in the Walker A motif in FtsE, predicted to impair ATP binding (Yang et al., 2011;Meisner et al., 2013) but not CwlO association with FtsX (Brunet et al., 2019) also resulted in increased lytE transcription (Figure 1-figure supplement 1). From these experiments we conclude that cells lacking CwlO activity increase expression of lytE. We also observed a modest but reproducible increase in lytE transcription in cells lacking LytE ( Figure 1B), suggesting that B. subtilis increases lytE expression in response to reduction in D,L-endopeptidase activity in general.
To investigate whether the~2 fold change in LytE levels in the DcwlO mutant has any physiological consequences, we used a strain lacking both cwlO and lytE that contained an IPTG-regulated allele of lytE. First, we determined the inducer concentration that resulted in LytE levels equivalent to wild-type ( Figure 1A) and then examined the cells by fluorescence microscopy ( Figure 1C, Figure 1-figure supplement 2). As reported previously, cells lacking cwlO were shorter and fatter than wild-type and these morphological phenotypes were largely homogenous throughout the population (Hashimoto et al., 2012;Meisner et al., 2013;Brunet et al., 2019). Furthermore, based on cytoplasmic mCherry fluorescence ( Figure 1C) and propidium iodide staining ( Figure 1D), the DcwlO mutant cells had intact membranes. By contrast, cells lacking cwlO in which LytE was artificially maintained at wild-type levels had heterogeneous morphologies with >20% lysis or membrane permeability defects ( Figure 1C and D, Figure 1-figure supplement 2). These results indicate that the increase in LytE levels allows the DcwlO mutant to maintain membrane integrity and a homogenous morphology. Thus, these data suggest that cells lacking CwlO compensate for the reduction in D,L-endopeptidase activity by increasing expression of a second D,L-endopeptidase, LytE.
B. subtilis modulates lytE transcription in response to changes in D,Lendopeptidase activity The experiments described above indicate that B. subtilis increases lytE expression in the absence of CwlO activity. We next investigated whether cells decrease lytE transcription in the presence of too much D,L-endopeptidase activity. Because CwlO is regulated post-translationally by FtsEX (Domínguez-Cuevas et al., 2012;Meisner et al., 2013), we used LytE to generate high D,L-endopeptidase activity. We introduced a strong IPTG-inducible promoter fusion to lytE and a catalytic mutant (C247S) into a strain harboring our P lytE -lacZ reporter and monitored ß-galactosidase activity after induction. Strikingly, cells with increased levels of wild-type LytE had reduced lytE transcription, while cells over-expressing the catalytic mutant had P lytE promoter activity similar to wild-type (Figure 2A   (A) Immunoblot analysis of LytE produced under the control of its native promoter or under IPTG control. The indicated strains (DlytE, wild-type (wt), DcwlO, and DcwlO DlytE P (IPTG) -lytE (LytE = wt)) were grown in CH medium with or without 500 mM IPTG and harvested at an OD600~0.4. SigA protein levels were analyzed to control for loading. (B) Bar graph showing b-galactosidase activity from a lytE promoter (P lytE ) fusion to lacZ in wild-type (wt), DcwlO, DftsEX, and DlytE strains. Activity was assayed in exponentially growing cultures in LB. Error bars represent standard deviation from three biological replicates. Asterisks indicate p-values calculated using Welch's unequal variances t-test compared to wildtype (*<0.02, **<0.005). (C) Representative fluorescent images of strains from (A) harboring cytoplasmic mCherry grown under identical conditions as in (A). Membranes were visualized with TMA-DPH (top), and merged images of cytoplasmic mCherry and phase-contrast are shown (bottom). Scale bar indicates 2 mm. (D) Bar graph showing the percentage of cells with envelope integrity defects in wild-type (wt), DcwlO, and a DcwlO mutant in which LytE levels are held at levels equivalent to wild-type (LytE = wt). Cells without cytoplasmic fluorescence and/or that stained with propidium iodide were scored as lysed or PI positive. >500 cells were scored per strain. The images and immunoblots in this figure were representatives from three independent experiments. The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1.     Immunoblot analysis and lacZ reporters are population-based assays. To address whether the changes in lytE transcription in response to high or low D,L-endopeptidase activity were homogenous throughout the population, we built a lytE promoter fusion to the gene encoding the yellow fluorescent protein variant Venus (P lytE -venus). The fluorescent reporter was introduced into strains with high or low D,L-endopeptidase activity and then monitored during exponential growth by fluorescence microscopy. As can be seen in Figure 2C, all cells lacking cwlO had increased Venus fluorescence while all cells over-expressing lytE had reduced fluorescence. Collectively, these results indicate that B. subtilis modulates lytE expression in response to both an increase and decrease in D,L-endopeptidase activity. -galactosidase activity from the P lytE -lacZ reporter in wild-type (wt), and strains harboring the lytE gene or a catalytic mutant (C247S) under the control of the strong IPTG-inducible promoter P hyperspank (P hy ) and with an optimized ribosome binding site. Activity was assayed in exponentially growing cultures 60 min after the addition of 50 mM IPTG. Error bars represent standard deviation from three biological replicates. (B) Immunoblot analysis of LytE in the same strains as in (A), 60 min after induction with 50 mM IPTG. SigA protein levels were analyzed to control for loading. (C) Representative images of cytoplasmic Venus fluorescence from the P lytE -venus reporter in the indicated strains visualized 30 min after addition of 50 mM IPTG. (D) Bar graph showing bgalactosidase activity from P lytE -lacZ in the indicated strains. Cells were grown in LB medium or LB medium supplemented with the indicated concentrations of IPTG. b-galactosidase activity was assayed 60 min after induction of lytE. Error bars represent standard deviation from three biological replicates. (E) Immunoblot analysis of LytE and SigA in the strains used in (D). All representative images and immunoblots in this figure are from one of three independent experiments. The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1. Figure 2A B-galactosidase assay Miller Units. Source data 2. Figure 2D B-galactosidase assay Miller Units. The experiments presented thus far indicate that B. subtilis can increase or decrease lytE transcription in cells lacking cwlO or over-expressing lytE. However, it seemed unlikely that B. subtilis evolved a mechanism to compensate for gene deletion and over-expression. To investigate whether B. subtilis modulates lytE transcription in response to more physiological changes in D,L-endopeptidase activity, we used an IPTG-regulated allele of lytE and grew cells at different inducer concentrations to produce a range of LytE levels that were both above and below wild-type levels ( Figure 2E). As can been seen in Figure 2D, we found that P lytE transcription inversely correlated with the amount of D,L-endopeptidase produced. Importantly, using the same IPTG-regulated promoter fused to gfp, we found that at IPTG concentrations similar to those used to express lytE all cells in the population had equivalent GFP fluorescence (Figure 2-figure supplement 1). Thus, the graded response to the changes in LytE levels observed in the ensemble assays reflect similar changes in D, L-endopeptidase levels in all cells in the population.

Modulation of lytE expression requires WalR and not SigI
The lytE gene is has two promoters that influence each other (Tseng et al., 2011;Salzberg et al., 2013) ( Figure 3A). One is recognized by the alternative sigma factor Sigma I (SigI) and the other is Schematic diagram of the lytE gene with its two promoters recognized by SigI and SigA. Phosphorylated WalR binds to two sites (box 1 and 2) flanking the SigA promoter. (B) Bar graph showing b-galactosidase activity from the P lytE -lacZ reporter in the indicated strains. Activity was assayed in exponentially growing cultures in LB medium. Error bars represent standard deviation from three biological replicates. (C) Immunoblot analysis of LytE protein levels in the strains used in (B). SigA levels were analyzed to control for loading. Immunoblots from one of two biological replicates are shown. (D) Spot dilutions of the indicated strains in the presence and absence of cwlO expression. Strains were grown in the presence of IPTG (500 mM) to OD 600~2 .0. The cultures were washed twice without inducer, normalized to an OD 600 = 1.5, 10-fold serially diluted, and spotted (5 mL) onto LB agar plates with or without 500 mM IPTG. Representative plates from one of two biological replicates are shown. (E) Bar graph showing b-galactosidase activity from the P lytE -lacZ reporter in indicated strains. Strains harboring the strong IPTG-inducible promoter fusion to lytE (P hy -lytE) were induced for 60 min with 50 mM IPTG. Error bars represent standard deviation from three biological replicates. The online version of this article includes the following source data for figure 3: Source data 1. Figure   b-galactosidase activity from P yocH -lacZ and P iseA -lacZ reporters in the indicated strains. Activity was assayed in exponentially growing cultures in LB medium. Strains harboring strong IPTG-inducible promoter fusion to lytE (P hy -lytE) and the LytE catalytic mutant (C247S) (P hy -lytE*) were induced for 60 min with 50 mM IPTG. Error bars represent standard deviation from three biological replicates. (D) Representative fluorescent images of the indicated strains harboring P yocH -venus. Cells were grown to OD 600~0 .3 in LB medium supplemented with 1 mM IPTG, washed in medium lacking inducer and imaged before and 30 min after growth in LB medium lacking inducer. (E) Representative fluorescent images of the indicated strains harboring P iseA -venus. Strains were grown to OD 600~0 .3 in LB medium and imaged before and 30 min after addition of IPTG (50 mM). Representative images are from one of three independent experiments. (F) Quantification of the average fluorescence intensity, normalized to cell area, of strains shown in (D) and (E).>1000 cells were analyzed from three independent experiments. Dark gray bars (0 m) are from cells prior to IPTG removal for the CwlO depletion experiment monitoring P yocH -venus and prior to IPTG addition for the lytE over-expression experiment monitoring P iseA -venus. Light gray bars (30 m) are from cells 30 min after IPTG removal (left graph) and 30 after IPTG addition (right graph). (G) Representative fluorescent images of P iseA -venus at the indicated times after addition of IPTG (50 mM). P iseA -venus is de-repressed within 15 min after induction of lytE.. The online version of this article includes the following source data and figure supplement(s) for figure 4: Source data 1. Figure 4B B-galactosidase assay Miller Units. Source data 2. Figure 4C B-galactosidase assay Miller Units. Source data 3. Figure 4F cwlO depletion, cellular fluorescence intensity. Source data 4. Figure 4F lytE over-expression, cellular fluorescence intensity. Figure supplement 1. Validation of P iseA -venus and P yocH -venus reporters. Figure 4 continued on next page controlled by Sigma A (SigA) but requires the phosphorylated form of the response regulator WalR (Salzberg et al., 2013). Cells lacking SigI have reduced P lytE -lacZ expression and reduced LytE protein levels, while cells lacking the anti-SigI factor RsgI have increased lytE transcription and increased LytE protein levels ( Figure 3B and C). Similarly, a point mutation in one of the two WalR binding sites in the lytE promoter (box2*) (Salzberg et al., 2013) abolishes lytE transcription and LytE protein levels ( Figure 3B and C). Furthermore, cells lacking WalH, a negative regulator of the WalRK two-component system (Szurmant et al., 2005), results in increased lytE transcription and LytE protein levels ( Figure 3B and C). Consistent with these observations and the synthetic lethal relationship between lytE and cwlO, cells lacking SigI or harboring a point mutation in the WalR binding site in the lytE promoter are inviable when CwlO is depleted (Salzberg et al., 2013) ( Figure 3D). To investigate whether the RsgI-SigI or the WalRK signaling pathway is involved in the response to changes in D,L-endopeptidase activity, we monitored the P lytE -lacZ response to increased LytE levels in DrsgI and DwalH mutants in which SigI and WalR were constitutively active. As can be seen in Figure 3E, over-expression of lytE in the DrsgI mutant resulted in reduced lytE transcription, while in the DwalH mutant lytE transcription was unchanged. These data suggest that the WalRK signaling pathway is responsible for mediating the observed response to changes in D,L-endopeptidase activity.
The WalRK signaling pathway responds to changes in D,Lendopeptidase activity To further test whether the WalRK pathway responds to changes in D,L-endopeptidase activity, we generated transcriptional reporters for two well-characterized genes (yocH and iseA) that are specifically regulated by WalR and not SigI (Bisicchia et al., 2007). yocH is positively regulated by phosphorylated WalR (WalR~P) while iseA is negatively regulated by WalR~P ( Figure 4A). We fused both promoters to lacZ and separately to venus. To validate these reporters, we monitored their activity in strains lacking WalH or WalI, two negative regulators of WalK (Szurmant et al., 2005;Szurmant et al., 2007). In the absence of either, WalR activity is high and P yocH -venus transcription increased, while P iseA -venus transcription decreased ( Next, we used these WalR-specific reporters to investigate whether the WalRK pathway responds to changes in D,L-endopeptidase activity. When monitored by ß-galactosidase assay, P yocH transcription increased in the absence of CwlO and decreased when lytE, but not lytE(C247S), was overexpressed ( Figure 4B and C). Reciprocally, P iseA transcription was reduced in the absence of CwlO and was strongly de-repressed when lytE was over-expressed ( Figure 4B and C). Similar results were obtained with our fluorescent reporters. Within 30 min after shutting off cwlO transcription, P yocHdirected transcription of venus increased ( Figure 4D). Furthermore, 15 min after inducing lytE transcription, we could detect de-repression of the P iseA -venus reporter ( Figure 4E and G), and after 30 min down-regulation of the P yocH -venus was apparent ( Figure 4-figure supplement 2). Quantification of Venus fluorescence ( Figure 4F) indicates that our fluorescent and lacZ reporters respond similarly to changes in D,L-endopeptidase levels, although for unknown reasons the magnitude of P iseA de-repression was not as great with the fluorescent reporter. Importantly, these WalR-specific promoters were not affected by a deletion of sigI ( Figure 4B and C and   Figure 3E, these results suggest that the WalRK signaling pathway monitors the extent of D,L endopeptidase activity and modulates lytE and cwlO transcription in response.
Recent studies indicate that the serine/threonine kinase PrkC also controls WalR activity (Libby et al., 2015). PrkC principally regulates WalR in stationary phase and therefore was unlikely to mediate the response to changes in D,L-endopeptidase activity during exponential growth observed here. However, to directly test this we analyzed P iseA and P yocH transcription in wild-type and the DprkC mutant before and after inducing lytE (Figure 4-figure supplement 5). As Representative fluorescence and phase-contrast images of cells harboring P iseA -venus before and 30 min after induction of lytE with 50 mM IPTG. Images on the right are from a strain that also contained the FtsZ inhibitor MciZ under xylose control (P (xyl) -mciZ) and were grown in the presence of 10 mM xylose for 60 min prior to IPTG addition. (B) Representative fluorescence and phase-contrast images of cells harboring P yocH -venus before and 30 min after removal of IPTG to shut off cwlO transcription. Images on the right are from a strain that also contained P (xyl) -mciZ and was grown in the presence of 10 mM xylose for 40 min, prior to IPTG removal. A subset of the division-inhibited cells did not de-repress P iseA -venus (A) or induce P yocH -venus (B); this could be due to loss of viability (see   anticipated, the transcriptional responses were similar in the presence and absence of PrkC, ruling out a role for this signaling kinase in responding to D,L-endopeptidase activity. The WalK sensor kinase responds to changes in D,L-endopeptidase activity in the absence of cell division Previous work on the WalRK pathway indicates that WalK, but not WalH or WalI, localizes to the divisome and does so in an FtsZ-dependent manner (Fukushima et al., 2008). Furthermore, depletion of FtsZ or other divisome components resulted in a decrease in yocH mRNA levels and an increase in pdaC transcript levels, consistent with a reduction in WalR activity (Fukushima et al., 2008;Fukushima et al., 2011). These studies formed the basis of the prevailing model for WalRK function in which WalK signaling and WalR-dependent expression of cell wall hydrolases are linked to growth Figure 6. The extracellular domains of WalH and WalI are not required for WalK to respond to changes in D,Lendopeptidase activity. (A) Representative fluorescence images of the indicated strains harboring P iseA -venus and a strong IPTG-inducible promoter fused to lytE (P hy -lytE). Strains were grown to OD 600~0 .3 in LB medium and imaged before and 30 min after addition of 50 mM IPTG. The medium was supplemented with 10 mM xylose for the strains harboring xylose-regulated alleles of walH or walI with deletions of their extracellular domains (DECD). Schematic diagram of the putative WalK/WalH/WalI membrane complex is shown to the right. (B) Representative images of P iseA -venus expression before and 30 min after lytE over-expression in strain lacking both walH and walI and complemented by xylose-induced full-length genes or DECD deletion variants. Schematic diagram of the putative membrane complex with WalH and WalI lacking their ECDs is shown on the right. Representative images are from one of three independent experiments. Scale bar indicates 5 mm. The online version of this article includes the following figure supplement(s) for figure 6: via the divisome (Fukushima et al., 2008;Dubrac et al., 2008a;Fukushima et al., 2011). Specifically, it was proposed that in non-dividing cells WalK is held inactive along the lateral membranes by its negative regulators WalH and WalI, while in actively growing and dividing cells, WalK localizes to the septum without its inhibitors where it is competent to activate WalR. Our findings that WalK responds to changes in PG hydrolases that act along the lateral cell wall appear inconsistent with a model in which WalK is only active at the septum. To address this discrepancy, we investigated whether the WalRK pathway could respond to changes in D,L-endopeptidase activity in cells lacking a divisome.
To prevent divisome assembly, we took advantage of the FtsZ inhibitor MciZ (Handler et al., 2008). Cells harboring a xylose-regulated allele of mciZ were grown for 60 min in the presence of inducer to block FtsZ-ring assembly ( Figure 5-figure supplement 1). The resulting filaments were then induced to express lytE, and WalRK activity was monitored 30 min later using our P iseA -venus reporter. As can be seen in Figure 5A, lytE over-expression resulted in similar de-repression of P iseAvenus in wild-type and cells lacking a divisome. Reciprocally, depletion of CwlO in filamenting cells for 30 min triggered induction of P yocH -venus to the same extent as CwlO depletion in wild-type cells ( Figure 5B). We note that a subset of the filaments in both experiments did not respond to changes in D,L-endopeptidase activity. We suspect this is due to a loss in viability, which is largely consistent with propidium iodide staining (Figure 5-figure supplements 2 and 3). Altogether, our data indicate that the WalRK signaling pathway is capable of responding to changes in PG hydrolase activity in the absence of division, and therefore argue that WalK can function along the lateral membranes.
WalK responds to changes in D,L-endopeptidase activity in the absence of the extracellular domains of the WalH and WalI regulators The two WalK inhibitors, WalH and WalI, are single-pass integral membrane proteins with large extracellular domains (ECDs) (Szurmant et al., 2005) (Figure 6A -schematic). Both regulators reside in a membrane complex with WalK (Szurmant et al., 2007) and could therefore function in signal recognition. Previous work from Szurmant and Hoch showed that basal WalRK activity was unaffected by deletions of the ECDs of WalH and WalI . To investigate whether either of the extracellular domains were required for WalRK signaling in response to changes in D,Lendopeptidase activity, we sought to test strains in which these domains were deleted. We generated xylose-regulated alleles of full-length walH and walI and deletion variants (walH(D61-455) and walI(D36-280)) that were identical to those used previously . We introduced these alleles at an ectopic genomic locus in strains lacking walH or walI that harbored the P iseA -venus reporter. As expected, in the absence of xylose, the DwalH and DwalI strains had virtually undetectable P iseA -venus expression, indicative of high WalRK activity and strong repression of the P iseA promoter ( Figure 6A, Figure 6-figure supplement 1). Similar to what was reported previously, expression of both the full-length and truncated alleles of walH and walI restored basal P iseA -venus transcription ( Figure 6A, Figure 6-figure supplement 1). Importantly, cells harboring the fulllength and truncated alleles of walH and walI responded similarly to wild-type when lytE was overexpressed ( Figure 6A, Figure 6-figure supplement 1). In all cases, P iseA -venus transcription was de-repressed, indicative of a reduction in WalRK signaling. These data indicate that the extracellular domains of WalH and WalI are not individually required for WalRK signaling in response to changes in D,L-endopeptidase activity. To test whether cells lacking the extracellular domains of both WalH and WalI can respond to lytE over-expression, we built stains harboring xylose-regulated alleles of both walH and walI or their deletion variants. As can be seen in Figure 6B, in both strains P iseA -venus transcription was de-repressed upon over-expression of lytE, although the response was more modest in the strain harboring the deletion variants. Collectively, these data argue that the extracellular domains of WalH and WalI are not necessary for WalRK signaling in response to changes in D,Lendopeptidase activity and suggest that the extracellular domain of WalK functions as the sensor.
The WalK extracellular domain is required to respond to changes in D,Lendopeptidase activity WalK contains a 148 amino acid extracellular loop that is homologous to Per-Arnt-Sim (PAS)-like domains, also known as an sCache domain (Upadhyay et al., 2016) (Figure 6A -schematic). These domains are a common feature of sensor kinases and in some cases have been shown to bind signaling ligands (Chang et al., 2010;Henry and Crosson, 2011;Upadhyay et al., 2016). Previous studies indicate that deletion of the extracellular sCache domain of WalK (WalKDsCache) results in constitutive signaling and high WalR activity (Fukushima et al., 2011) and we confirmed this using both P yocH -venus and P iseA -venus reporters ( Figure 7A). To determine whether this domain is required to respond to changes in D,L-endopeptidase activity, we monitored P iseA -venus transcription after over-expression of lytE. P iseA transcription was de-repressed in both wild-type cells and the DwalK mutant complemented with full-length walK, indicative of a decrease in WalRK signaling Representative fluorescence images of the indicated strains harboring P iseAvenus or P yocH -venus reporters. Cells were grown to OD 600~0 .4 in LB medium. Strains in which WalK lacks its extracellular sensory domain (DsCache) have high WalRK activity, leading to strong repression of P iseA -venus and elevated transcription of P yocH -venus. Schematic model of the signaling complex with WalK lacking its extracellular domain is shown on the right. (B) Representative fluorescence images of the indicated strains harboring P iseAvenus and P hy -lytE. Strains were grown to OD 600~0 .3 in LB medium and imaged before and 30 min after addition of 50 mM IPTG. Representative images are from one of three independent experiments. Scale bar indicates 5 mm. (C) WalK and WalKDsCache interact with WalH and WalI in the bacterial adenylate cyclase two-hybrid (BACTH) assay. The BTH101 E. coli reporter strain containing plasmids expressing the indicated protein fusions to the complementing (T18 and T25) domains of the Bordetella adenylate cyclase. The T18-WalHI plasmid contains a T18-WalH fusion and untagged WalI. Cells were grown to early stationary phase in LB at 37˚C, normalized to OD 600 = 0.2, and 3 mL of each was spotted on LB agar plates supplemented with X-Gal and IPTG. Plates were incubated overnight at 30˚C. The 'Zip' fusions, composed of the leucine zipper domain of GCN4, served as positive and negative controls. The online version of this article includes the following figure supplement(s) for figure 7: ( Figure 7B). However, the DwalK mutant harboring the walK(DsCache) variant was unresponsive to lytE over-expression ( Figure 7B). Furthermore, the walK(DsCache) variant failed to respond to depletion of CwlO (Figure 7-figure supplement 1).
These data are consistent with the idea that the sCache domain of WalK is necessary for sensing changes in D,L-endopeptidase activity. However, previous in vivo formaldehyde crosslinking experiments suggested that the interaction between WalK and WalH/WalI requires the extracellular domain of WalK (Fukushima et al., 2011). If correct, the constitutive signaling in the walK(DsCache)  mutant and the failure to respond to changes in D,L-endopeptidase activity could be due to the loss of inhibition by WalH and WalI. To more directly test the interactions among these proteins, we used the Bacterial Adenylate Cyclase Two Hybrid (BACTH) system (Ladant and Ullmann, 1999;Karimova et al., 2005). We generated fusions with complementary fragments (T18 and T25) of Bordetella pertussis adenylate cyclase to WalK, WalK(DsCache), WalH, and WalI. The fusions were cotransformed into E. coli and assayed for interaction on LB agar supplemented with X-Gal (see Materials and methods). As can be seen in Figure 6C, a strong interaction was only observed when T25-WalK and T18-WalH were co-expressed with untagged WalI, indicating that all three proteins are required for stable interaction. Importantly, we observed a similarly strong interaction using T25-WalK(DsCache) ( Figure 6C). These data argue that the WalK variant lacking its ECD stably interacts with WalH and WalI in vivo. Collectively, these results and those from Figure 5 suggest that the extracellular sCache domain of WalK functions as the sensing domain that monitors changes in D,Lendopeptidase activity.  The WalK response is specific to D,L-endopeptidase crosslink cleavage The data presented thus far indicate that the WalK sensor kinase responds to changes in cell wall hydrolysis. We envisioned two possible models for what WalK might be sensing. Since CwlO and LytE are both D,L-endopeptidases, WalK could specifically monitor D,L-endopeptidase cleavage products. Alternatively, WalK could be sensing some aspect of the cell wall that is affected by PG hydrolysis in general. For example, WalK could monitor the presence of intact crosslinks or the extent of tension on the PG meshwork. To help distinguish between these two models, we investigated whether WalK responds to PG hydrolases that cleave distinct bonds in the PG meshwork ( Figure 8B). To target these enzymes to the lateral cell wall, we fused the catalytic domains of each PG hydrolase to the N-terminal LysM domains of LytE (M3) ( Figure 8A) that direct it to the lateral wall (Buist et al., 2008;Hashimoto et al., 2012;Hashimoto et al., 2018). Each chimera was expressed under the control of the same IPTG-regulated promoter used to over-express lytE, and WalK signaling was monitored 30 min after induction using the P iseA -venus reporter. Consistent with the idea that WalK monitors D,L-endopeptidase activity, the chimeras that contained D, L-endopeptidase domains from PG hydrolases that function in cell separation (LytF and CwlS from B. subtilis) (Ohnishi et al., 1999;Yamamoto et al., 2003;Fukushima et al., 2006) de-repressed P iseA transcription ( Figure 8C, Figure 8-figure supplement 1B). By contrast, WalK signaling was not impacted by expression of chimeras containing D,D-endopeptidase domains (MepM and MepS from E. coli) (Singh et al., 2012), L,D-endopeptidase domains (LytH and CwlK from B. subtilis) (Horsburgh et al., 2003;Fukushima et al., 2007), the amidase domain from B. subtilis YrvJ (Wendrich and Marahiel, 1997), or the glucosaminidase domain from B. subtilis LytD (Margot et al., 1994) ( Figure 8C, Figure 8-figure supplement 1B). Over-expression of most of the chimeras did not cause discernable morphological defects raising the possibility that the fusions were not active enough to alter WalRK signaling, however, we note that the MepM D,D-endopeptidase chimera was able to suppress the lethality of DcwlO DlytE double mutant (Hashimoto et al., 2018) but even this fusion did not impact P iseA transcription (Figure 8-figure supplement 1A). Collectively, these data favor the model that WalK specifically monitors D,L-endopeptidase activity.
To directly test whether D,L-endopeptidase cleavage products inhibit WalRK signaling, we sought to investigate whether soluble PG cleavage products generated by a D,L-endopeptidase in vitro could de-repress the P iseA -venus reporter. To this end, we incubated purified B. subtilis sacculi at 37C for 16 hr with a commercial muramidase (mutanolysin from Streptomyces globisporus), a purified amidase (LytA from Streptococcus pneumoniae) (Flores-Kim et al., 2019), or the constitutively active D,L-endopeptidase domain from B. subtilis CwlO that lacks its regulatory coiled-coil domain (CwlODcc) ( Figure 8D). Phase-contrast microscopy was used to assess cell wall hydrolysis ( Figure 8E -top panels). Translucent sacculi containing phase-dark aggregates (Liechti et al., 2014) were readily detectable in buffer-treated sacculi, while only the liberated phase-dark aggregates were present in the sacculi incubated with the PG hydrolases. The soluble material was collected from each digestion and added to cells harboring the P iseA -venus reporter. Notably, only the cleavage products generated by D,L-endopeptidase digestion reduced WalRK activity and de-repressed P iseA -venus ( Figure 8E -bottom panels). Furthermore, the soluble material from sacculi incubated with iodoacetamide-inactivated CwlODcc failed to reduce WalRK activity (Figure 8-figure supplement 2). These results indicate that WalK specifically responds to D,L-endopeptidase cleavage products.

Discussion
Altogether, our data support a model in which the WalR-WalK two-component signaling pathway functions in homeostatic control of the cell wall elongation hydrolases LytE and CwlO, and does so by sensing and responding to PG cleavage products generated by these enzymes (Figure 9). When D,L-endopeptidase activity is low, the concentration of these cleavage products drops leading to an increase in WalK kinase activity and a concomitant increase in WalR~P. WalR~P increases transcription of lytE and cwlO, restoring homeostasis (Figure 9 -left panel). Reciprocally, when there is elevated D,L-endopeptidase activity, the concentration of cleavage products increases. High concentrations of these putative allosteric inhibitors reduce WalK activity, resulting in a decrease in lytE and cwlO transcription (Figure 9 -right panel).
In this model, transcriptional activation of lytE and cwlO rapidly boosts D,L-endopeptidase activity when the levels of these enzymes drop. However, this pathway can also rapidly reduce D,L-endopeptidase activity when it gets too high. The iseA gene, which is repressed by WalR~P, encodes a secreted inhibitor of D,L-endopeptidases (Bisicchia et al., 2007;Salzberg and Helmann, 2007;Yamamoto et al., 2008). It was given its name (inhibitor of cell separation A) because of its ability to inhibit cell separation D,L-endopeptidases when over-expressed on a multi-copy plasmid (Yamamoto et al., 2008). However, we have found that LytE is likely to be the primary target of IseA. Specifically, we found that the levels of IseA needed to inhibit LytE had no impact on cell separation (Figure 9-figure supplement 1A and B). Furthermore, we have also found that CwlO has a half-life of~7 min (Figure 9-figure supplement 2A). Accordingly, when D,L-endopeptidase activity is too high, a drop in WalRK signaling not only causes a reduction in lytE and cwlO transcription, but also inhibition of LytE activity via de-repression of iseA and a reduction in CwlO levels through degradation or shedding of the enzyme into the medium (Figure 9-figure supplement 2B). Thus, the WalRK signaling pathway is able to adjust the levels and activity of CwlO and LytE on the time-scale of a cell cycle to maintain homeostatic control of these essential enzymes during growth.
Previous work on the WalRK signaling pathway led to the model that WalK localization to the septal ring, in the absence of its negative regulators WalH and WalI, couples its activity to growth via the divisome (Fukushima et al., 2008;Dubrac et al., 2008a). Support for this model comes from experiments in which depletion of FtsZ or other divisome components results in a decrease in yocH mRNA levels and an increase in pdaC mRNA levels, suggesting that WalK-dependent phosphorylation of WalR requires an intact divisome (Fukushima et al., 2008;Fukushima et al., 2011). One caveat of these experiments is that yocH and pdaC transcript levels were monitored 3 hr after depletion of the divisome components, raising the possibility that the reduction in WalR activity could be indirect. We note that we did not detect changes in P iseA -and P yocH -directed Venus fluorescence after inhibiting FtsZ for 70-90 min ( Figure 5, Figure 5-figure supplement 2, Figure 5-figure supplement 3). However, RT-PCR is likely to be more sensitive than our fluorescence-based assay. It is therefore possible that WalRK functions in the homeostatic control of the elongation hydrolases as described here, and separately acts at the divisome to boost CwlO and LytE levels during cytokinesis in anticipation of cell wall growth following division. Time-lapse microscopy in microfluidic devices like the mother machine (Wang et al., 2010) will enable a direct test of this model in the future.
We have demonstrated that the soluble cleavage products generated by D,L-endopeptidasedigestion can inhibit WalK signaling and could therefore function as allosteric inhibitors of the kinase. Although the specific cleavage product that inhibits WalK is currently unknown, we favor the idea that the cross-linked tetra-and/or penta-peptides liberated by cleavage of peptide crosslinks ( Figures 8B and 9) are the WalK ligand because cleavage of PG crossbridges is required for expansion of the PG meshwork during growth (Bisicchia et al., 2007;Meisner et al., 2013;Hashimoto et al., 2012;Hashimoto et al., 2018). To generate these products, both g-D-Glu-mDAP bonds must be cleaved in the peptide crossbridge. Concerted cleavage of identical bonds in peptide crosslinks has been proposed previously (Wong et al., 2015), however it has not been rigorously tested. Alternatively, it is also possible that di-or tri-peptides released upon cleavage of uncross-linked stem peptides or even short glycan strands with attached di-peptides could serve as the signal. Future experiments will be directed at defining the specific cleavage product that is sensed by WalK and ultimately the ligand-binding site in the sensor domain.
As indicated above, our data are most consistent with a model in which WalK is inhibited when it binds its signal and active when un-liganded ( Figure 9). Specifically, we showed that D,L-endopeptidase cleavage products cause de-repression of P iseA , indicative of low levels of WalR~P and low WalK kinase activity. Reciprocally, we found that depletion of CwlO, and therefore low concentrations of cleavage products, leads to high P yocH transcription, indicative of high levels of WalR~P and high WalK kinase activity. Another, closely related, two-component system in B. subtilis is the PhoR-PhoP system, involved in sensing and responding to phosphate limitation (Seki et al., 1987;Botella et al., 2014). The PhoP response regulator activates genes involved in phosphate scavenging and inhibits the synthesis of the phosphate-rich surface polymers called wall teichoic acid (WTA). Interestingly, work from the Devine lab suggests that the sensor kinase PhoR is active when unliganded and inhibited when bound by intermediates in the WTA biosynthetic pathway (Botella et al., 2014). In this case, the signal is thought to be sensed by the intracellular PAS domain of PhoR. Nonetheless, it is interesting that in both cases these kinases appear to be regulated by allosteric inhibition.
Cell growth requires an intimate balance between cell wall synthesis and its hydrolysis. The homeostatic control pathway we have defined here ensures that exponentially growing cells maintain a defined amount of D,L-endopeptidase activity for cell wall elongation. However, under different growth conditions, for example entry into stationary phase, the cell is likely to require different levels of hydrolase activity. We propose that the cell could adapt by modulating the homeostatic 'set point' of the WalRK signaling pathway. This set point is likely determined by the levels of the WalK-WalH-WalI sensor kinase complex; the affinity of WalK for its putative allosteric inhibitor; and the activities of the histidine kinase and response regulator. A link between one or more of these determinants and PG synthesis could ensure coordination. Since the wal operon is not under auto-regulatory control (Fabret and Hoch, 1998), altering the set point is unlikely to occur through changes in the levels in the WalK-WalH-WalI complex. Instead, B. subtilis could modulate WalK's affinity for D,Lendopeptidase cleavage products or adjust WalK kinase activity or WalR-dependent transcription. For example, the conserved extracellular domains of WalH and WalI could function in modulating ligand affinity in response to an extracellular signal generated during PG synthesis, or in response to changes in the cell envelope itself. Similarly, the conserved intracellular PAS domain of WalK could alter kinase activity in response to a cytoplasmic signal. For example, this domain could sense intermediates in the PG precursor (lipid II) pathway, analogous to the signal sensed by the PhoR PAS domain. Finally, WalR activity is known to be modulated by PrkC-dependent phosphorylation during entry into stationary phase (Libby et al., 2015), which is thought to respond to extracellular muropeptides. By changing the homeostatic set point for hydrolase activity, extracellular signals could directly coordinate hydrolase activity with PG synthesis while the intracellular signal could link hydrolase activity to flux through the precursor synthesis pathway. Identifying these potential allosteric modulators and defining how they impact the homeostatic control of LytE and CwlO are important challenges for the future.
The WalRK two-component system is the most broadly conserved TCS in the Firmicutes and can be found in important human pathogens including Staphylococcus aureus, Listeria monocytogenes, Bacillus anthracis, Streptococcus pneumoniae, and Streptococcus mutans (Dubrac et al., 2008a). In most cases, this pathway is essential and, in the organisms in which it has been investigated, the WalR regulon contains cell wall hydrolases although not necessarily D,L-endopeptidases (Martin et al., 1999;Howell et al., 2003;Dubrac and Msadek, 2004;Ng et al., 2005;Liu, 2006;Bisicchia et al., 2007;Ahn and Burne, 2007;Delaune et al., 2011). Furthermore, as is the case with B. subtilis (Takada et al., 2018), the essentiality of this pathway can be bypassed by engineering these bacteria to express a subset of the PG hydrolytic enzymes in the WalR regulon (Ng et al., 2003;Delaune et al., 2011). Based on these similarities, we hypothesize that the WalRK pathway functions in homeostatic control of cell wall hydrolysis in these pathogens and uses distinct cleavage products to monitor PG hydrolase activity.
It is noteworthy that the WalK sensor kinases in Streptococci and Lactococci lack an extracellular sCache domain (Lange et al., 1999;Dubrac et al., 2008a;Wang et al., 2013). Accordingly, if there is homeostatic control of cell wall hydrolases in these bacteria they must use a different sensing mechanism or employ a distinct signaling pathway. Intriguingly, the StkP serine/theronine kinase in S. pneumoniae has been implicated in modulating cell wall hydrolysis and has been suggested to function in concert with WalRK pathway in this organism (Fleurie et al., 2012;Stamsås et al., 2017). Analysis of the WalK signaling in diverse Gram-positive pathogens will establish the similarities and differences in the regulatory logic we have uncovered here. Finally, the cleavage products of the PG hydrolases in the WalR regulons could potently inhibit WalK signaling and therefore offer the potential for therapeutic development.

Strains, plasmids, and routine growth conditions
All Bacillus subtilis strains were derived from the prototrophic strain PY79 (Youngman et al., 1983). Cells were grown in either Luria-Bertani (LB) or casein hydrolysate (CH) medium at 37˚C. Unless otherwise indicated, B. subtilis strains were constructed using genomic DNA and a 1-step competence method. Antibiotic concentrations were used at: 100 mg/mL spectinomycin, 5 mg/mL chloramphenicol, 10 mg/mL tetracycline, 10 mg/mL kanamycin, 1 mg/mL erythromycin and 25 mg/mL lincomycin. A list of strains and plasmids used in this study can be found in the Key Resources Table (Supplementary file 2), and oligonucleotide primers can be found in Supplementary file 1.

b-Galactosidase assays
B. subtilis strains were grown in LB medium at 37˚C to an OD 600 of~0.5. The optical density was recorded and 1 mL of culture was harvested and assayed for b-galactosidase activity as previously described (Rudner et al., 1999). Briefly, cell pellets were re-suspended in 1 mL Z buffer (40 mM NaH 2 PO 4 , 60 mM Na 2 HPO 4 , 1 mM MgSO 4 , 10 mM KCl, and 50 mM b-mercaptoethanol). 250 mL of this suspension was added to 750 mL of Z buffer supplemented with lysozyme (0.25 mg/ml), and the samples were incubated at 37˚C for 15 min. The colorimetric reaction was initiated by addition of 200 mL of 2-nitrophenyl-b-D-galactopyranoside (ONPG, 4 mg/ml) in Z buffer and stopped with 500 mL 1M Na 2 CO 3 . The absorbance at 420 nm and OD 550 of the reactions were recorded, and the bgalactosidase specific activity in Miller Units was calculated according to the formula [A 420 -1.75x (OD 550 )] / (time [min] x OD 600 ) x dilution factor x 1000 (Miller, 1972).

LytE purification and antibody production
Recombinant LytE lacking its three N-terminal LysM domains (LytEDlysMx3) was expressed in E. coli BL21 (DE3) using the P T7 -His 6 -SUMO-lytEDlysMx3 expression vector (pYB18). Cells were grown in Terrific Broth (Tartof and Hobbs, 1988) supplemented with 100 mg/ml ampicillin at 37˚C to OD 600 = 0.3. LytEDlysMx3 expression was induced for 16 hr at 22˚C by addition of 0.5 mM IPTG. After induction, cells were collected by centrifugation at 10,000 Â g for 10 min. Cell pellets were resuspended in 15 mL Buffer 1 (20 mM Tris pH 7.5, 300 mM NaCl, 5 mM imidazole, 10% glycerol, 0.1 mM Dithiothreitol) and Complete EDTA-free protease inhibitors (Roche) and lysed via passage through a French press. Cell lysates were clarified by centrifugation at 10,000 x g for 10 min at 4˚C. Clarified lysates were mixed with 0.5 mL of Ni 2+ -NTA agarose resin (Qiagen) and incubated for 2 hr at 4˚C. The mixture was loaded onto a column (BioRad) and washed with 10 mL Buffer 1.The His 6 -SUMO-LytEDlysMx3 fusion protein was eluted with Buffer 2 (20 mM Tris pH 7.5, 300 mM NaCl, 200 mM imidazole, 0.1 mM Dithiothreitol). Eluates were pooled and dialyzed into storage buffer (20 mM Tris pH 7.5, 300 mM NaCl, 10% glycerol, 0.1 mM Dithiothreitol) at 4˚C overnight. 10 mL of purified His 6 -Ulp1 (1.25 mg/ml) was added to the dialysate and was incubated overnight on ice. The reaction was then mixed with 0.5 mL Ni 2+ -NTA agarose and loaded onto a column. Flow-through fractions containing the cleaved (untagged) LytEDlysMx3 were collected and used to generate rabbit polyclonal antibodies (Covance).

Fluorescence microscopy
Exponentially growing cells were harvested and concentrated by centrifugation at 6800 x g for 1.5 min and re-suspended in 1/10th volume growth medium and then immobilized on 2% (wt/vol) agarose pads containing growth medium. Fluorescence microscopy was performed on a Nikon Ti inverted microscope equipped with a Plan Apo 100x/1.4 Oil Ph3 DM phase contrast objective, an Andor Zyla 4.2 Plus sCMOS camera, and Lumencore SpectraX LED Illumination. Images were acquired using Nikon Elements 4.3 acquisition software. The fluorescent membrane dye TMA-DPH was added to the concentrated cell suspension at 50 mM final. Propidium iodide (PI) was added at a final concentration of 5 mM. Venus and YFP were imaged using a Chroma ET filter cube for YFP (49003) with an exposure time of 800 ms; TMA-DPH was visualized using a Chroma ET filter cube for DAPI (49000) with an exposure time of 300 ms; mCherry and PI were visualized using a Chroma ET filter cube for mCherry (49008) with an exposure time of 800 ms and 500 ms, respectively. Image processing was performed using Metamorph software (version 7.7.0.0) and Oufti (Paintdakhi et al., 2016) was used for quantitative image analysis.

Bacterial two-hybrid assay
The Bacterial Adenylate Cyclase-based Two Hybrid (BACTH) system was used as previously described (Karimova et al., 1998;Bendezú et al., 2009). Briefly, pairs of proteins were fused to the complementary fragments (T18 and T25) of the Bordetella pertusis adenylate cyclase. Competent BTH101 E. coli cells were co-transformed with the two plasmids containing T18 and T25 protein fusions in one step. Transformants were selected on LB agar plates supplemented with 100 mg/mL ampicillin (Amp 100 ), 50 mg/mL kanamycin (Kan 50 ), 500 mg/mL isopropyl-b-D-thiogalactoside (IPTG 500 ), and 100 mg/mL 5-Bromo-4-chloro-3-indolyl-b-D-galactopyranoside (X-Gal 100 ). Plates were incubated at 30˚C and the homogeneity of the colony color among transformants was confirmed. Single colonies were then grown to early stationary phase in LB medium supplemented with Amp 100 and Kan 50 , normalized to OD 600 = 0.2, and spotted (3 mL) on LB agar plates containing Amp 100 , Kan 50 , IPTG 500 , X-Gal 100 . Plates were incubated at 30˚C overnight and imaged. Transformations were done in triplicate with selected images representative of three biological replicates.

PG hydrolase over-expression
Cultures of exponentially growing cells were diluted to OD 600 = 0.01 in LB and grown for~1 hr at 37˚C to an OD 600~0 .05. The cultures were then induced with IPTG (concentrations indicated in the Figure Legends). Fluorescent images were acquired before and at indicated times after induction. For experiments in which cell division was inhibited, cultures of exponentially growing cells were diluted to OD 600 = 0.04 and 10 mM xylose was added to induce mciZ. The cultures were then grown for~1 hr at 37˚C to allow cells to filament for~2 mass doublings before 50 mM IPTG was added to induce lytE expression.

PG hydrolase depletion
Cultures were grown in LB supplemented with 1 mM IPTG to mid-exponential phase, washed three times in LB lacking inducer, and diluted to OD 600 = 0.05 in LB to initiate depletion. Images were acquired before and at indicated times after depletion as described in the text. For experiments in which cell division was inhibited, cultures were grown in the presence of 1 mM IPTG (to induce cwlO) until mid-exponential phase. Cells were diluted to OD 600 = 0.04 with 1 mM IPTG and 10 mM xylose (to induce mciZ). The cultures were then grown for~1 hr at 37˚C to allow cells to filament for~2 mass doublings. The cells were then washed three times in LB lacking IPTG, and resuspended in LB containing 10 mM xylose. iseA over-expression.
Cultures of exponentially growing cells were diluted to OD 600 = 0.02 in CH medium supplemented with 10 mM xylose and grown for~1.5 hr at 37˚C to an OD 600~0 .2. The cultures were then induced with 500 mM IPTG. Fluorescent images were acquired at indicated times after induction.

In vivo protein turnover assay
Wild-type B. subtilis was grown in LB medium at 37˚C to an OD 600 of 0.5. Protein translation was blocked by the addition of both spectinomycin (200 mg/mL, final concentration) and chloramphenicol (10 mg/mL, final concentration). Samples (1 mL of culture) were collected immediately prior to antibiotic treatment and at the indicated times after. Cells were pelleted by centrifugation for 5 min and immediately flash-frozen in liquid nitrogen. The cell pellets were thawed on ice, resuspended in lysis buffer (20 mM Tris pH 7.0, 10 mM MgCl 2 , 1 mM EDTA, 1 mg/ml lysozyme, 10 mg/ml DNase I, 100 mg/ml RNase A, 1 mM PMSF, 1 mg/ml leupeptin, 1 mg/ml pepstatin), and the suspensions transferred to fresh microfuge tubes to avoid CwlO present in culture medium that non-specifically bound to the plastic tube (Brunet et al., 2019). The lysates were then analyzed by immunoblot as described above.

CwlODcc purification
Recombinant CwlO lacking its N-terminal coiled coil domain (Dcc) was expressed in E. coli BL21 (DE3) DfhuA (New England Biolabs) using the P T7 -His 6 -SUMO-cwlODcc expression vector (pJM63). Cells were grown in LB supplemented with 100 mg/mL ampicillin at 37˚C to OD 600 = 0.5. Cultures were allowed to equilibrate at room temperature for 30 min and then transferred to 30˚C. His 6 -SUMO-CwlODcc expression was induced with 0.5 mM IPTG for 3 hr. Cells were collected by centrifugation, resuspended in 50 mL Buffer A (20 mM Tris HCl pH 7.4, 500 mM NaCl, 20 mM Imidazole, and 2X complete protease inhibitor tablets (Roche), and stored at À80˚C. The cell suspension was thawed on ice and lysed by two passes through a cell disruptor (Constant Systems Ltd.) at 25,000 psi. The lysate was clarified by ultracentrifugation at 35,000 rpm for 30 min at 4˚C. The supernatant was added to 1 mL Ni 2+ -NTA beads (Qiagen) and incubated for 1 hr at 4˚C. The suspension was loaded into a 10 mL column (BioRad), washed twice with 4 mL Buffer A, and eluted with 2.5 mL Buffer B (20 mM Tris HCl pH 7.4, 500 mM NaCl, 300 mM Imidazole). 10 mL of purified His 6 -Ulp1 (1.25 mg/ml) was added to the eluate, and the mixture was dialyzed into storage buffer (20 mM Tris HCl pH 8, 100 mM NaCl, 10% glycerol) overnight at 4˚C. The next morning 10 mL more His 6 -Ulp1 was added to the dialysate and incubated for 1 hr at 30˚C. The dialysate was mixed with 1 mL of Ni 2+ -NTA beads for 1 hr at 4˚C to remove free His 6 -Ulp1 and His 6 -SUMO. The suspension was loaded onto a column and the CwlODcc-containing flow-through was collected, aliquoted, and stored at À80˚C.

B. subtilis sacculi preparation
Wild-type B. subtilis was grown in 500 mL LB at 37˚C to an OD 600 = 0.5. Cells were pelleted, re-suspended in 10 mL 0.1 M Tris HCl pH 7.5 with 2% SDS (wt/vol), and boiled for 1 hr. The sample was cooled to room temperature and incubated with Proteinase K Solution (Invitrogen) at a final concentration of 0.4 mg/mL at 50˚C for 1 hr. Sacculi were pelleted at 20,000 x g and washed five times with 10 mL ddH 2 O until free of SDS. The sacculi were then subjected to acid hydrolysis by suspension in 10 mL 1 M HCl at 37˚C for 4 hr. Sacculi were then pelleted and washed five times with 10 mL ddH 2 O. Sacculi were distributed into 10 Â 1 mL aliquots and stored at À80˚C.

Sacculi digestion
B. subtilis sacculi aliquots were resuspended in 1.4 mL cleavage buffer (25 mM MES pH 5.5) and dispersed in a water-bath sonicator for 30 min. Purified CwlODcc (final concentration 0.13 mg/mL), LytA (Flores-Kim et al., 2019), final concentration 0.17 mg/mL), and mutanolysin (Sigma Aldrich, 50 units final) were separately added to sonicated sacculi and incubated overnight at 37˚C. After the overnight incubation, insoluble material was pelleted at 20,000 g for 15 min and the soluble cleavage products were collected, lyophilized, heat-inactivated (100˚C for 20 min), and stored at À80˚C. Immediately prior to use, the lyopholized material was resuspended in 50 mL ddH 2 O. When indicated, CwlODcc was inactivated with 20 mM iodoacetamide at room temperature in darkness for 30 min prior to addition to sacculi and overnight incubation.

Assaying soluble cleavage products for inhibition of WalRK signaling
Cultures of exponentially growing cells were diluted to an OD 600 = 0.01 in LB and grown for 1 hr at 37˚C to OD 600~0 .05. 50 mL of soluble sacculi cleavage products (prepared as described above) were added to 450 mL of cells and were incubated at 37˚C with aeration. Fluorescent images were acquired before and at indicated times after addition.

Strain and plasmid construction Deletion strains
Insertion-deletion mutants were from the Bacillus knock-out (BKE) collection (Koo et al., 2017) or were generated by isothermal assembly (Gibson, 2011) of PCR products followed by direct transformation into B. subtilis. All BKE mutants were back-crossed twice into B. subtilis PY79. All deletions were confirmed by PCR. Antibiotic cassette removal was performed using a temperature-sensitive plasmid that constitutively expresses Cre recombinase (Meeske et al., 2015).
Bacterial two-hybrid plasmids pGD115 [T25-walK] pGD115 was constructed in a 2-way ligation with a PCR product containing the walK gene (amplified from B. subtilis PY79 genomic DNA using oligonucleotide primers oGD278 and oGD279) cut with XbaI and KpnI and pKT25 cut with the same enzymes. The resulting plasmid was sequencedconfirmed.

pGD118 [T18-walH]
pGD118 was constructed in a 2-way ligation with a PCR product containing the walH gene (amplified from B. subtilis PY79 genomic DNA using oligonucleotide primers oGD282 and oGD283) cut with XbaI and KpnI and pCH364 cut with the same enzymes. The resulting plasmid was sequencedconfirmed.
pGD120 [T18-walI] pGD120 was constructed in a 2-way ligation with a PCR product containing the walI gene (amplified from B. subtilis PY79 genomic DNA using oligonucleotide primers oGD284 and oGD285) cut with XbaI and KpnI and pCH364 cut with the same enzymes. The resulting plasmid was sequencedconfirmed.

pGD122 [T18-walHI]
pGD122 was constructed in a 2-way ligation with a PCR product containing the walHI genes (amplified from B. subtilis PY79 genomic DNA using oligonucleotide primers oGD282 and oGD285) cut with XbaI and KpnI and pCH364 cut with the same enzymes. The resulting plasmid was sequencedconfirmed.
pGD123 [T25-walK(D44-167)] pGD123 was constructed in a 2-way ligation with a PCR product containing the walK(D44-167) gene (amplified with genomic DNA from bGD500 using oligonucleotide primers oGD278 and oGD279) cut with XbaI and KpnI and pKT25 cut with the same enzymes. The resulting plasmid was sequenced-confirmed. pJM063 [P T7 -SUMO-6xHis-cwlODcc] pJM063 was constructed in a 2-way isothermal assembly reaction with a PCR product containing the C-terminal domain of CwlO (cwlODcc, amino acids 334-473) (amplified from B. subtilis PY79 genomic DNA using oligonucleotide primers oJM117 and oJM320) and pTB146 (Bendezú et al., 2009) cut with SacI and BamHI. The resulting plasmid was sequenced-confirmed.