The WalRK Two-Component System Is Essential for Proper Cell Envelope Biogenesis in Clostridioides difficile

ABSTRACT

The WalR-WalK two-component regulatory system (TCS) is found in all Firmicutes, in which it regulates the expression of multiple genes required for remodeling the cell envelope during growth and division. Unlike most TCSs, WalRK is essential for viability, so it has attracted interest as a potential antibiotic target. In this study, we used overexpression of WalR and CRISPR interference to investigate the Wal system of Clostridioides difficile, a major cause of hospital-associated diarrhea in high-income countries. We confirmed that the wal operon is essential and identified morphological defects and cell lysis as the major terminal phenotypes of altered wal expression. We also used transcriptome sequencing (RNA-seq) to identify over 150 genes whose expression changes in response to WalR levels. This gene set is enriched in cell envelope genes and includes genes encoding several predicted PG hydrolases and proteins that could regulate PG hydrolase activity. A distinct feature of the C. difficile cell envelope is the presence of an S-layer, and we found that WalR affects expression of several genes which encode S-layer proteins. An unexpected finding was that some Wal-associated phenotypic defects were inverted in comparison to what has been reported for other Firmicutes. For example, downregulation of Wal signaling caused C. difficile cells to become longer rather than shorter, as in Bacillus subtilis. Likewise, downregulation of Wal rendered C. difficile more sensitive to vancomycin, whereas reduced Wal activity is linked to increased vancomycin resistance in Staphylococcus aureus.

IMPORTANCE The WalRK two-component system (TCS) is essential for coordinating synthesis and turnover of peptidoglycan in Firmicutes. We investigated the WalRK TCS in Clostridioides difficile, an important bacterial pathogen with an atypical cell envelope. We confirmed that WalRK is essential and regulates cell envelope biogenesis, although several of the phenotypic changes we observed were opposite to what has been reported for other Firmicutes. We also identified over 150 genes whose expression is controlled either directly or indirectly by WalR. Overall, our findings provide a foundation for future investigations of an important regulatory system and potential antibiotic target in C. difficile.

INTRODUCTION

To grow and divide, bacteria must thoroughly remodel their peptidoglycan (PG) sacculus, a large, covalently closed macromolecule that surrounds the cell and provides protection against rupture due to turgor pressure (1). How bacteria remodel the sacculus without making errors that lead to inadvertent lysis is a question that has fascinated microbiologists for decades. The question is also of practical significance because small molecules that undermine normal peptidoglycan biogenesis are among our most effective antibiotics.

An important advance in our understanding of cell envelope biogenesis was made over 20 years ago when Fabret and Hoch identified a Bacillus subtilis two-component system (TCS) that is essential for viability (2). This TCS is now called the Wal system and is known to coordinate expression of multiple cell envelope genes in the Firmicutes. WalRK is essential in all species studied so far, including pathogens such as Staphylococcus aureus (3–5) and Streptococcus pneumoniae (6, 7). Because of its essentiality, the Wal TCS has attracted interest as a potential antibiotic target.

The Wal TCS comprises the bifunctional kinase/phosphatase WalK together with its cognate response regulator, WalR. In addition, wal operons typically include genes for membrane proteins known or presumed to modulate WalK activity, but the accessory proteins differ between organisms (8–10). Recent evidence from B. subtilis indicates that WalK activity is regulated by PG fragments generated by PG hydrolases that open spaces to make room for elongation. These PG fragments bind to WalK’s extracellular Cache domain to modulate the phosphorylation state of WalR (11). Phosphorylated WalR, in turn, binds to various promoters to activate or repress gene expression.

The number of genes in the WalR regulon ranges from about a dozen to over 100 (5, 12–16), depending on the species, and only some of these genes are directly activated or repressed by WalR. Although WalR regulons are diverse, they invariably include multiple proteins that contribute to proper biogenesis of the cell envelope, especially PG hydrolases. Consistent with these themes, the phenotypic defects elicited by artificial up- or downregulation of WalRK signaling include abnormal cell shape, larger cells, smaller cells, ghost cells (lysis), thicker PG, abnormal division septa, and altered sensitivity to antibiotics that target PG biogenesis (8–10). In S. pneumoniae walRK is essential because it activates expression of an essential PG hydrolase gene, pcsB (6, 7). However, the essentiality of walRK in B. subtilis and S. aureus is polygenic in nature, resulting from abnormal expression of multiple genes that are not by themselves essential. In all three organisms, it is possible to bypass the walRK requirement by constitutively expressing one or more Wal regulon PG hydrolase genes, sometimes in combination with deleting genes for hydrolase inhibitors (6, 17, 18).

Here, we studied the WalRK TCS in Clostridioides difficile, an anaerobic, spore-forming member of the Firmicutes responsible for close to a quarter of a million hospitalizations and over 12,000 deaths per year in the United States (19). Unlike the envelope of most Firmicutes in which walRK has been studied previously, the C. difficile envelope has a proteinaceous S-layer whose assembly presumably must be coordinated with PG synthesis (20–22). In addition, the PG itself is somewhat unusual (23). About 90% of the N-acetylglucosamine (GlcNAc) is deacetylated (which has implications for PG hydrolase activity), and about 70% of the peptide cross-links are 3-3 rather than 4-3 (which has implications for the structure of PG fragments thought to bind to the Cache domain of WalK [11]). The C. difficile Wal system includes a unique lipoprotein gene and a noncanonical WalK that lacks an intracellular PER-ARNT-SIM (PAS) signaling domain (Fig. 1) (10). These features distinguish C. difficile’s WalRK TCS from those of previously characterized Firmicutes, although a distantly related wal-like system in Mycobacterium tuberculosis (MtrAB) includes a lipoprotein and a WalK-like histidine kinase without an intracellular PAS domain (10, 24, 25). Finally, our recent development of tools for xylose-inducible gene expression and CRISPR interference (26) helped to overcome some of the challenges inherent in phenotypic analysis of essential genes like walRK in C. difficile.

FIG 1.

The wal operon of C. difficile. Based on studies in other bacteria, WalK (red) is a bifunctional signal-transducing enzyme that can phosphorylate or dephosphorylate WalR (green). WalK is predicted to have an extracellular Cache domain, as well as intracellular HAMP, HisKA (phospho-acceptor), and HATPase domains. Unlike the B. subtilis ortholog (inset), it does not have an intracellular PAS domain. WalR~P binds DNA to activate or repress expression of Wal-regulated genes. The lipoprotein (WalA [blue]), which is unique to C. difficile, may modulate WalK activity and is predicted to have a beta-propeller lyase domain. The truA2 gene codes for a pseudouridylate synthase and is not thought to play a role in Wal signaling. The operon locus is cdr20291_1676-1679 in R20291 and cd630_17810-17840 in 630Δerm.

RESULTS

The wal operon is required for cell viability, proper rod morphology, and intrinsic resistance to some antibiotics.

According to transposon insertion sequencing (Tn-seq), the WalRK TCS is essential in C. difficile (27). To confirm and extend this finding, we leveraged the power of CRISPR interference (CRISPRi) for functional analysis of essential genes/operons (28). Our CRISPRi system comprises a xylose-inducible nuclease-defective Cas9, P_xyl::dCas9, that is targeted to a gene of interest by a single guide RNA (sgRNA) expressed constitutively from the glutamate dehydrogenase promoter, P_gdh::sgRNA (26). Because CRISPRi is polar, we decided to target the first gene in the operon, which encodes a predicted lipoprotein unique to the C. difficile wal operon (Fig. 1). We have named this gene walA. For reproducibility, we designed two guides against walA, sgRNA-walA1 and sgRNA-walA2. Subsequent transcriptome sequencing (RNA-seq) experiments described below confirmed that targeting dCas9 to walA suppressed transcription of the entire operon. As a negative control, we used an sgRNA (sgRNA-neg) that does not have a target anywhere on the C. difficile chromosome. In some experiments the CRISPRi machinery was produced from a plasmid that confers resistance to thiamphenicol (Thi), while in others it was integrated into the chromosome at the pyrE locus in a way that does not leave behind an antibiotic resistance marker (29).

Chromosomal- or plasmid-based CRISPRi knockdown of the wal operon in either 630Δerm or R20291 reduced viability ≥10⁴-fold when cells were plated on tryptone-yeast extract (TY) agar containing 1% xylose (Fig. 2A; see also Fig. S1A in the supplemental material). Additional phenotypes were characterized in TY broth using 630Δerm derivatives with the CRISPRi machinery integrated at pyrE. In the absence of xylose, sgRNA-walA1 and sgRNA-walA2 retarded growth slightly compared to the sgRNA-neg control (Fig. 2B and Fig. S1B). This finding indicates that there is leaky expression of P_xyl::dCas9, which was later confirmed by RNA-seq (see below). Under these conditions, cells were, on average, about 10% longer than the wild type and the fraction of cells with curved or irregular contours increased from ~1% in the sgRNA-neg control to ~20% for sgRNA-walA1 or ~10% for sgRNA-walA2 (Fig. 2C to E and Fig. S1E and F). Further knockdown of the wal operon by addition of xylose exacerbated the growth defect in a dose-dependent manner (Fig. 2B and Fig. S1D). In addition, morphological defects became more pronounced, with cells now averaging ~30% longer than controls and up to ~40% exhibiting curved or irregular contours (Fig. 2D and E and Fig. S1E and F).

FIG 2.

The wal operon is required for viability and normal rod morphology of C. difficile 630Δerm. (A) Viability assay. Serial dilutions of overnight cultures were spotted onto TY plates with or without 1% xylose to induce the expression of dCas9. Plates were photographed after incubation overnight. Strains carry chromosomal copies of the CRISPRi components and the following guides: sgRNA-neg (UM554), sgRNA-walA1 (UM555), and sgRNA-walA2 (UM556). (B) Growth curves in liquid TY containing 0, 0.01, 0.1, or 1% xylose as indicated. For clarity, the sgRNA-neg control strain is graphed only at 0% xylose. Samples were taken at 4 h (arrow) for phase-contrast microscopy (C), cell length measurements (D), and determination of curvature (E). Numbers at the bottom of micrographs in panel C refer to percent xylose. Note cell debris indicative of lysis at 1% xylose. Bar = 10 μm. Cell length and percent curved cells were based on about 500 cells per condition. The sgRNA-walA1 strain is longer than the sgRNA-neg control in all pairwise comparisons as determined by a t test (P < 0.0001).

Partial knockdown of the wal operon in the absence of xylose allowed us to screen for altered sensitivity to antibiotics that target cell envelope biogenesis. We found that the MIC for ampicillin and daptomycin decreased about 2-fold, while the MIC for vancomycin decreased about 4-fold (Table 1). But sensitivity to the other two cell wall-active antibiotics, imipenem and bacitracin, and the gyrase inhibitor novobiocin was not affected (Table 1).

TABLE 1.

Effect of CRISPRi-silencing wal on sensitivity to antibiotics^a

Antibiotic	MIC (μg/mL)
	sgRNA-neg	sgRNA-walA1	sgRNA-walA2
Ampicillin	6.3	3.5	3.1
Imipenem	3.1	3.1	2.6
Vancomycin	4.6	0.9	0.7
Daptomycin	11.3	4.1	5.7
Bacitracin	250	250	250
Novobiocin	13	17	16

^aResults are the averages of measurements done in duplicate on four different days.

Overexpression of WalR impairs growth, alters morphology, and slows autolysis.

We also investigated Wal function in C. difficile by overproducing the response regulator WalR to drive cells into an exacerbated Wal-ON state (5, 30). Plating C. difficile 630Δerm harboring a P_xyl::walR expression plasmid onto TY-Thi containing 1% xylose resulted in a 10⁶-fold loss of viability (Fig. 3A). This strain failed to grow when subcultured directly into TY-Thi broth containing 1% xylose, but waiting until the optical density at 600 nm (OD₆₀₀) reached ~0.2 before adding 3% xylose resulted in only a subtle growth defect (Fig. 3B). Three hours postinduction, the cells had become strikingly phase bright and resistant to autolysis when suspended in buffer containing 0.01% Triton X-100 (Fig. 3C and D). Triton X-100 is thought to reduce the inhibitory effect of lipoteichoic acids on PG hydrolases (31, 32). Similar results were obtained in the R20291 strain background (Fig. S3). It is unclear what causes cells to turn phase bright upon overexpression of walR. Differential staining with Syto 9 and propidium iodide (LIVE/DEAD staining) indicated that >99% of the cells were alive after overproduction of WalR (Fig. 3B, red arrow).

FIG 3.

Overexpression of walR from a multicopy plasmid impairs growth, alters morphology, and slows autolysis in 630Δerm. (A) Viability assay. Overnight cultures of 630Δerm harboring pCE691 (P_xyl::walR) or the empty vector control plasmid pBZ101 (P_xyl EV) were serially diluted and spotted onto TY-Thi plates with or without 1% xylose. Plates were photographed after incubation overnight. (B) Growth curve of 630Δerm/pCE691 (P_xyl::walR). Duplicate cultures were grown to an OD₆₀₀ of 0.2, at which time one was induced with 3% xylose (green arrow). Both cultures were harvested after 3 h (red arrow) for analysis by phase-contrast microscopy (C) and a lysis assay (D). Bar in panel C = 10 μm. For the lysis assay, cells were suspended in buffer containing 0.01% Triton X-100 and OD₆₀₀ was monitored over time. Error bars depict SD of 3 technical replicates. Data shown are representative of results from at least 3 independent experiments.

We repeated these experiments after integrating P_xyl::walR into the chromosome of 630Δerm. As expected, the effects of xylose induction were similar but less pronounced when P_xyl::walR was in single copy (Fig. 4A). The addition of xylose at subculture was now tolerated and reduced growth in a dose-dependent manner. At 3% xylose, cells became slightly phase bright. There were some bent or hooked cells, a defect not observed with the P_xyl::walR plasmid, perhaps because it could only be induced for shorter times (Fig. 4C). After 6 h of induction, cells were harvested and evaluated in the lysis assay. We found that overexpression of walR slowed lysis in a dose-dependent manner (Fig. 4B).

FIG 4.

Prolonged overexpression of walR from a single-copy chromosomal P_xyl::walR allele affects cell shape and slows lysis. (A) Growth curves. An overnight culture of strain UM626 (P_xyl::walR integrated at pyrE in 630Δerm) was subcultured 1:50 into TY with varied xylose concentrations as indicated, grown for 6 h, and then analyzed in a lysis assay (B) or by phase-contrast microscopy (C). For the lysis assay, cells were suspended in buffer containing 0.01% Triton X-100 and OD₆₀₀ was monitored over time. Data are graphed as the mean and SD of 3 technical replicates, but most error bars are smaller than the symbols. Cells from cultures grown in the presence of 0 to 1% xylose appeared normal and are not shown, but in cultures induced with 3% xylose, about 10% of the cells had irregular or bent morphologies, examples of which are shown in panel C. Bar = 10 μm. Data shown are representative of those from at least 3 independent experiments.

Expression profiling under Wal-ON conditions.

As a step toward understanding the basis of the various Wal-related phenotypes, we sought to identify the genes in the wal regulon. (Note that we use the term regulon for convenience to refer to genes whose expression responds to manipulation of the wal operon regardless of whether these effects are direct or indirect.) We began by performing RNA-seq on two strains that overexpress walR from a single-copy P_xyl::walR construct integrated at pyrE. One of these strains overproduced wild-type WalR; the other overproduced a WalR^D54E mutant protein expected to mimic phosphorylated WalR and thus be constitutively active (13, 33) (confusingly, D54 is annotated as D66 in the R20291 genome on the BioCyc website). The two strains behaved similarly with respect to growth and performance in the lysis assay (Fig. S4A and B). The control strain for these experiments had the gene for a red fluorescent protein (rfp) integrated at pyrE (26). Strains were grown overnight in TY and then subcultured into fresh TY at a starting OD₆₀₀ of ~0.05. When the OD₆₀₀ reached ~0.2, cultures were induced with 3% xylose and allowed to grow for 3 h (~2 mass doublings), at which time cells were harvested and processed for RNA-seq. At this time point, induction of walR had only a small effect on growth rate (Fig. S4A) and >99% were viable as judged by LIVE/DEAD staining.

Not including walR, which was induced directly from P_xyl, the RNA-seq analysis identified 77 genes whose transcript abundance changed ≥4-fold (P < 0.05) upon overexpression of either walR or walR^D54E (Table 2). Of these genes, 44 were induced and 33 were repressed. Induction ratios were very similar in the walR and walR^D54E strains with only two exceptions, cd630_03901 and cd630_10631, both of which encode hypothetical proteins. The failure of the D54E substitution to increase gene expression could mean that it does not activate in C. difficile WalR, as reported previously for some other response regulators (34). Alternatively, overexpression of walR is itself activating and might obscure the effects of the substitution. The 77 genes are predicted by BioCyc (35) to be organized into 65 operons that contain 18 additional genes which missed our 4-fold expression change cutoff. Visual inspection of the expression data revealed that these genes almost always trended in the right direction, confirming the predicted structure of most of the operons. We therefore included these 18 genes in what we refer to as the Wal-ON regulon, bringing it to a total of 95 genes in 65 operons (Table 2).

TABLE 2.

Genes affected by Wal-ON conditions

Gene, changed ≥4-fold	Operona	Gene name	Annotationb	Log2 fold change			Cell wall associationd
				WalR overexpression		CRISPRi against walc
				WalR	WalRD54E
cd630_07390	07390		Hypothetical protein	5.26	5.70	−1.38	SPI
cd630_15220	15220	pgdA	Peptidoglycan deacetylase	4.02	4.26	−1.09	SPI
cd630_05490	05490		Hypothetical protein	3.87	4.51	−1.84	SPI
	17810	walA	Lipoprotein	NCe	NC	−2.76	SPII
cd630_17820	17820	walR	TC response regulator	3.81	3.67	−2.51
	17830	walK	TC sensor histidine kinase	NC	NC	−2.33	Membrane
	17840	truA2	tRNA pseudouridine synthase A	NC	NC	−2.34
cd630_07380	07380		Hypothetical protein	3.43	4.14	−1.22	SPI
cd630_08670	08670		Hypothetical protein	3.28	3.97	NC	Membrane
cd630_07400	07400		PLP-dependent aminotransferase	3.21	2.38	NC
cd630_10880	10880		Hypothetical protein	2.88	3.14	−1.40	Membrane
cd630_01660	01660		Peptidase	2.76	2.95	NC
	01650		Amino acid transporter	NC	NC	NC	Membrane
cd630_27940	27940	cwp12	Cell wall binding protein	2.69	2.82	NC	SPI
cd630_33090	33090	ligA	DNA ligase	2.64	2.62	NC
cd630_18250	18250	metY	O-Acetylhomoserine sulfhydrylase	2.50	2.80	−0.57
cd630_18260	18260	metA	Homoserine O-succinyltransferase	2.60	2.85	−0.88
cd630_14281	14281		Hypothetical protein	2.59	3.22	NC
cd630_27860	27860	cwp5	Cell wall binding protein	2.51	2.65	NC	SPI
cd630_25380	25380		Lipoprotein	2.47	2.70	−0.70	SPII
cd630_03910	03910	asnB	Asparagine synthetase	2.47	3.30	NC	Membrane
cd630_14960	14960		Hypothetical protein	2.37	2.86	−0.74	SPI
cd630_25370	25370		5′-Nucleotidase/phosphoesterase	2.33	2.49	−0.59	SPI
cd630_27910	27910	cwp2	Cell wall binding protein	2.30	2.36	−0.59	SPI
	27900		LmbE-like deacetylase	1.56	1.46	−0.55
	27890	cwp66	Cell wall binding protein	1.41	1.42	−0.58	SPI
cd630_28620	28620		Peptidase	2.29	2.24	−2.53
cd630_34880	34880	gtaB1	UTP-G1P uridylyltransferase	2.23	2.34	−0.73
cd630_20580	20580		Hypothetical protein	2.22	2.43	−1.26	Membrane
cd630_07410	07410	glpK1	Glycerol kinase	2.21	0.63	NC
	07420	eutH	Ethanolamine utilization protein	1.46	NC	NC	Membrane
	10350	cwp16	Amidase, cell wall binding protein	1.24	1.33	NC	SPI
cd630_10360	10360	cwp17	Amidase, cell wall binding protein	2.15	2.35	NC	SPI
cd630_21170	21170	trxB2	Thioredoxin reductase	2.12	2.18	NC
cd630_00210	00210	pycA	Pyruvate carboxylase	2.10	2.16	−0.46
cd630_13890	13890		Chloromuconate cycloisomerase	2.09	2.18	−0.28
	13900		Hypothetical protein	1.30	1.64		Membrane
cd630_22490	22490		ATPase	2.07	2.20	NC
cd630_22480	22480		Hypothetical protein	1.81	2.04	NC
cd630_10890	10890 10900	rgbR rgbS	TC response regulator	2.04	2.24	−1.13
	10900	rgbS	TC sensor HK (fragment)	NC	1.16	NC	Membrane
cd630_30930	30930		Aminobutyrate hydrolase	2.15	2.35	NC
	30920		Putative amino acid transporter	1.52	1.02	NC	Membrane
cd630_26810	26810		Putative Ca-chelating protein	2.01	2.59	NC	SPI
cd630_36010	36010		d-Alanyl–d-alanine carboxypeptidase	1.99	2.35	−1.38	SPI
cd630_00220	00220		Elongation factor G	1.97	2.09	NC
cd630_14290	14290		Acyl-CoA N-acyltransferase	1.92	2.47	−1.15
cd630_22280	22280		Hypothetical protein	1.83	2.05	−0.70	Membrane
cd630_01100	01100		Hypothetical protein	1.81	2.02	NC	Membrane
	01110		Hypothetical protein	1.51	1.73	NC	SPI
	01111		Hypothetical protein	NC	1.14	NC
cd630_09940	09940		Serine-pyruvate aminotransferase	1.77	2.10	NC
cd630_19942	19942		Hypothetical protein	1.74	2.12	NC
cd630_05270	05270		Beta-lactamase-like hydrolase	1.73	2.13	−0.85
cd630_22270	22270		Radical SAM protein	1.67	2.02	−0.55	SPII
cd630_18070	18070		NADPH-dependent FMN reductase	1.49	2.07	−1.53	Membrane
cd630_03901	03901		Hypothetical protein	0.92	2.37	NC
cd630_10631	10631		Hypothetical protein	−0.07	2.18	−0.62	Membrane
cd630_30360	30360		Major facilitator superfamily transporter	−1.59	−2.13	1.11	Membrane
cd630_10540	10540	bcd2	Acyl-CoA dehydrogenase	−1.93	−2.23	1.03
cd630_10550	10550	etfB	Electron transfer flavoprotein beta	−1.78	−2.11	1.07
cd630_10560	10560	etfA	Electron transfer flavoprotein alpha	−1.95	−2.18	1.10
cd630_10570	10570	crt2	3-Hydroxybutyryl-CoA dehydratase	−1.96	−2.27	1.06
cd630_31000	31000		C4-dicarboxylate anaerobic carrier	−1.88	−2.59	NC	Membrane
cd630_30990	30990		Amidohydrolase	−1.80	−2.47	NC	Membrane
cd630_10580	10580	hbd	3-Hydroxybutyryl-CoA dehydrogenase	−1.86	−2.05	0.89	SPII
cd630_10590	10590	thlA1	Acetyl-CoA acetyltransferase	−1.96	−2.16	0.77
	16820	queK	Queuosine hydrolase monomer	−1.31	−1.60	NC
	16830		ECF transporter S-comp.	−1.39	−1.65	NC	Membrane
cd630_16840	16840	queL	Radical SAM superfamily protein	−1.90	−2.06	NC
	23900		BlaI-like	−1.47	−1.65	NC
cd630_23890	23890		BlaR-like	−1.94	−2.26	0.80	Membrane
	23880		Putative PG hydrolase	−1.50	−1.77	NC	SPI
	36040		Hypothetical protein	−0.86	−0.94	0.71
cd630_36030	36030		Phosphoesterase	−1.97	−2.01	1.09	Membrane
cd630_14040	14040		Oligopeptide transporter	−2.02	−2.03	NC	Membrane
	14041		Hypothetical protein	−2.02	−1.88	NC
	25120		PTS, subunit IIA	−1.11	−1.47	NC
	25110		PTS, antiterminator	−1.45	−1.46	NC
cd630_25100	25100		PTS, subunit IIBC	−2.04	−2.19	0.97	Membrane
cd630_25320	25320		Alanine-glyoxylate transaminase	−2.05	−2.34	1.41
cd630_25310	25310		Hypothetical protein	−2.67	−2.75	1.35	Membrane
cd630_27680	27680		Cell wall hydrolase	−2.06	−2.29	NC	SPI
cd630_29660	29660	adhE1	Acetaldehyde-CoA/alcohol dehydrogenase	−2.08	−2.37	−1.09	Membrane
cd630_18120	18120		Response regulator	−2.08	−2.28	NC
cd630_22600	22600		Major facilitator superfamily transporter	−2.13	−2.32	NC	Membrane
cd630_20160	20160		Hypothetical protein	−2.19	−2.29	4.18
cd630_25090	25090		Glycoside hydrolase	−2.22	−2.33	1.16
cd630_21070	21070		Permease family protein	−2.23	−2.32	NC	Membrane
cd630_10220	10220		YczE-like membrane protein	−2.26	−2.73	0.93	Membrane
	10230		Transcriptional regulator	−1.70	−1.94	1.63
cd630_10240	10240	potA	ABC transport ATP-binding protein	−2.27	−2.36	1.77
cd630_10250	10250	potB	ABC transport permease	−2.47	−2.82	1.72	Membrane
cd630_10260	10260	potC	ABC transport permease	−2.53	−2.39	1.60	Membrane
cd630_10270	10270	potD	ABC transport substrate binding protein	−2.29	−2.37	1.55	SPII
cd630_26640	26640	murE	UDP-N-acetylmuramoylalanyl-d-glutamate-2,6-diaminopimelate ligase	−2.85	−3.07	0.93
cd630_26670	26670	ptsG-BC	PTS, subunit IIBC	−3.17	−3.52	NC	SPI
cd630_26660	26660	ptsG-A	PTS, subunit IIA	−3.12	−3.51	NC
cd630_21510	21510		Hypothetical protein	−3.56	−3.82	0.68	Membrane

^aLists other operon members in order.

^bTC, two component; PTS, phosphoenolpyruvate-dependent sugar phosphotransferase system; CoA, coenzyme A.

^cCRISPRi against wal: fold change comparing sgRNA-neg to sgRNA-walA1 strain in the absence of xylose.

^dReports predictions by SignalP (SPI, predicted to be a substrate of signal peptidase I; SPII, predicted to be a substrate of signal peptidase II) and by BUSCA (membrane). The BUSCA prediction is given only for those genes that were not already predicted to be SPI or SPII substrates by SignalP.

^eNC, no change, i.e., the fold change was close to 1 and the P value was >0.05.

Several of the genes and trends in Table 2 are worth highlighting. Xylose induction resulted in a 14-fold increase in walR mRNA (13-fold for walR^D54E), while the levels of walK and other members of the native operon were not affected. This means that induction of P_xyl::walR integrated at pyrE does not lead to induction of the native wal operon, consistent with reports that the wal operon is not autoregulated in other bacteria (8, 16). Another striking feature of Table 2 is the abundance of genes encoding cell envelope proteins. Of the 95 genes in Table 2, 22 (23%) are predicted to encode exported proteins and 29 (31%) are predicted to encode cytoplasmic membrane proteins. For comparison, ~7% of the entire proteome is predicted to be exported and 23% predicted to be membrane proteins (see Materials and Methods). Table 2 also includes 21 hypothetical proteins (22% of the total), many of which are likely to be involved in cell envelope processes based on the presence of predicted transmembrane domains or export signals.

Peptidoglycan-associated genes, especially genes encoding PG hydrolases, are prominent members of the Wal regulon in all Firmicutes examined to date (8, 10), and C. difficile is no exception. Table 2 lists 10 peptidoglycan-associated genes, including the second most highly induced gene, pgdA, which was upregulated 16-fold. PgdA deacetylates N-acetylglucosamine in peptidoglycan and is important for C. difficile’s high lysozyme resistance (36, 37). AsnB (~8-fold induced) is annotated as an asparagine synthetase, but its primary function is to amidate diaminopimelic acid (m-DAP) in PG stem peptides (38). In B. subtilis, amidation of m-DAP inhibits PG hydrolase activity (39). MurE (7-fold repressed) catalyzes one of the cytoplasmic steps in PG synthesis, addition of diaminopimelic acid to the nucleotide-linked PG precursor (40). Downregulation of murE suggests that PG synthesis may be decreased under Wal-ON conditions. Cwp16 and Cwp17 are S-layer proteins with predicted amidase domains (41) and were upregulated about 4-fold. Cell wall amidases cleave the amide bond that links stem peptides to N-acetylglucosamine in PG glycan strands (42). CD630_36010 (4-fold induced) is a predicted d-alanyl–d-alanine carboxypeptidase that removes the terminal d-Ala moiety from peptidoglycan pentapeptide side chains. This might limit 4-3 cross-linking of stem peptides but could favor 3-3 cross-linking. Two proteins of unknown function were upregulated >10-fold, CD630_07390 and CD630_07380. We suspect that both of these proteins are involved in PG metabolism because they are predicted to be exported and also induced by lysozyme (43). Notable among the repressed genes in Wal-ON cells is cd630_27680, which is downregulated about 4-fold and is predicted to encode a cell wall hydrolase from the NlpC/P60 family, a domain common to endopeptidases (42). Interestingly, this gene is preceded by two putative WalR binding sites (Table S4), and it has also been characterized as a sortase substrate (44). Another predicted PG hydrolase, CD630_23880, is mildly repressed (~3-fold). Interestingly, the gene encoding this hydrolase is cotranscribed with an uncharacterized BlaI-BlaR regulatory system, cd630_23890 and cd630_23900. Some BlaIR regulatory systems respond to beta-lactam stress (45, 46).

One of the questions that motivated this study was whether the WalR system regulates genes for S-layer proteins in C. difficile. Indeed, the Wal-ON regulon includes six S-layer protein genes, all of which are upregulated by WalR: cwp2, -5, -12, -16, -17, and -66. Two of these genes were mentioned already, the cell wall amidase genes cwp16 and cwp17. The remaining four are of unknown function, although cwp2 and cwp12 have domains implicated in adhesion or pathogenesis and studies with cwp66 have linked it to adhesion, autolysis, and resistance to antibiotics (41, 47). WalRK has been suggested to regulate an S-layer gene in Bacillus anthracis (48).

Expression profiling under Wal-OFF conditions.

As a complementary approach to identify genes in the WalR regulon, we assessed the effect of CRISPRi knockdown of the wal operon on global gene expression. To this end, we performed RNA-seq on a 630Δerm derivative that has the CRISPRi machinery (P_xyl::dCas9 and P_gdh::sgRNA-walA1) integrated into the chromosome at pyrE. The control strain was identical except that it expressed an innocuous sgRNA that does not target anywhere in the C. difficile genome (P_gdh::sgRNA-neg). Two induction conditions were used. One of these, called “no xylose,” relied on leaky expression of dCas9. For this, overnight cultures were diluted into fresh TY to a starting OD₆₀₀ of 0.03 and harvested when they reached an OD₆₀₀ of ~0.7 (Fig. S4C). For the other condition, called “low xylose,” cultures in exponential growth were induced with 0.1% xylose at an OD₆₀₀ of 0.1 and incubated for 2.5 h (~2 mass doublings) before harvest. This induction regimen reduced cell density only modestly (Fig. S4D).

There are 49 genes, organized in 27 operons, that shift 4-fold or more under Wal-OFF conditions (Table 3). The 27 operons contain an additional 14 genes that did not meet our 4-fold cutoff, all but 3 of which nevertheless trended in the right direction. The three exceptions include a pseudogene (cd630_05260) that was not detected by our RNA-seq analysis tool and two membrane protein genes (cd630_05290 and 05280) whose expression was unchanged. Thus, for the purposes of this analysis, CRISPRi knockdown of the wal operon changed the expression of 60 genes, of which 44 were induced and 16 were repressed. Curiously, for most of these genes the fold change was larger in the absence of xylose, suggesting that many of the effects on gene expression might be indirect.

TABLE 3.

Genes affected by Wal-OFF conditions

Gene, changed ≥4-fold	Operona	Gene name	Annotation	Log2 fold change			Cell wall associationc
				CRISPRi		WalR OEb
				No Xyl	Low Xyl
cd630_05140	05140	cwpV	Hemagglutinin/adhesin	4.69	2.74	NCd	SPI
cd630_20160	20160		Hypothetical protein	4.18	1.65	−2.19
cd630_11700	11700		Hypothetical protein	3.52	1.83	NC
cd630_11710	11710	etfB4	Electron transfer flavoprotein alpha	3.30	1.40	NC
cd630_11720	11720	etfA4	Electron transfer flavoprotein alpha	2.58	1.24	NC
cd630_11730	11730		FAD-linked oxidase	2.17	1.08	NC
cd630_08320	08320	aksA	trans-Homoaconitate synthase	3.11	2.02	NC
cd630_08330	08330	acnB	Aconitate hydratase	3.06	2.05	NC
cd630_08340	08340	icd	Isocitrate dehydrogenase	3.07	2.16	NC
cd630_17290	17290		Sodium:phosphate symporter	2.76	0.45	−1.46	Membrane
cd630_13840	13840		LamB/YcsF family protein	2.17	0.88	NC
cd630_13850	13850		Hypothetical protein	2.36	1.15	NC	Membrane
cd630_13860	13860		Allophanate hydrolase subunit 1	2.52	1.28	NC
cd630_13870	13870		Allophanate hydrolase subunit 2	2.60	1.14	NC
cd630_24260	24260	buk1	Butyrate kinase	2.45	NC	−1.41
	24250	ptb2	Phosphate butyryltransferase	1.65	NC
cd630_24240	24240		Pyridoxal phosphate-dependent transferase	2.55	NC	−1.15
	24230		Putative transporter	1.17	NC		Membrane
cd630_07650	07650	srlEa	PTS, IIB N-terminal component	2.17	NC	NC
cd630_07660	07660	srlEb	PTS, IIB C-terminal component	2.29	NC	NC	Membrane
cd630_07670	07670	srlB	PTS, IIA subunit	2.27	NC	NC	Membrane
cd630_07680	07680	srlD	Sorbitol-6-phosphate dehydrogenase	2.35	NC	NC
	00350		Transcriptional regulator	1.46	0.55	−0.86
cd630_00360	00360	acoA	Acetoin dehydrogenase E1 alpha	2.20	0.94	NC
cd630_00370	00370	acoB	Acetoin dehydrogenase E1 beta	2.29	0.91	NC
cd630_00380	00380	acoC	Acetoin dehydrogenase E2	2.20	0.72	NC
cd630_00390	00390	acoL	Acetoin dehydrogenase E3	2.04	0.83	−0.57
cd630_30010	30010		Hydrolase/isomerase/hydratase	2.25	NC	NC
cd630_23920	23920		tRNA-binding protein	2.15	1.65	−1.77
cd630_30960	30960	bglA3	6-Phospho-beta-glucosidase	2.13	NC	NC
cd630_30950	30950	bglA2	6-Phospho-beta-glucosidase	2.07	NC	NC
cd630_25150	25150		Aspartate aminotransferase	2.12	2.38	−1.76
	24291		Flavo/ferredoxin oxidoreductase delta	1.19	NC	NC
cd630_24290	24290		Flavo/ferredoxin oxidoreductase alpha	2.11	NC	−0.91
	24280		Flavo/ferredoxin oxidoreductase beta	1.66	NC	NC
	24270		Flavo/ferredoxin oxidoreductase gamma	1.50	NC	NC
cd630_30040	30040	kdgT2	2-Keto-3-deoxygluconate permease	2.08	0.33	0.93	Membrane
cd630_30020	30020	uxaA	d-Galactate dehydratase/altronate hydrolase	2.06	0.25	NC
cd630_32630	32630	pstC	Phosphate ABC transporter, PstC	2.03	NC	NC	Membrane
	32620	pstA	Phosphate ABC transporter, PstA	1.03	NC	NC	Membrane
	32610	pstB	Phosphate ABC transporter, ATP binding	1.80	NC	NC
	32600	phoU	Phosphate uptake regulator	1.25	NC	NC
	30980	bglG1	PTS, antiterminator	0.95	NC	NC
cd630_30970	30970	bglF2	PTS, subunit IIABC	2.01	NC	−0.76	Membrane
cd630_28540	28540	dltD	d-Alanine transferase	−1.52	−2.41	1.03	Membrane
cd630_28530	28530	dltA	d-Ala–d-ala carrier protein ligase 1	−1.34	−2.27	0.84
cd630_28520	28520	dltB	d-Alanyl transferase	−1.34	−2.23	0.49	Membrane
	28510	dltC	d-Ala–d-ala carrier protein ligase 2	−1.27	−1.89	NC
cd630_09000	09000	opuCA	ABC transporter ATP-binding protein	−1.69	−2.36	1.46	Membrane
cd630_09010	09010	opuCC	ABC transporter permease	−1.68	−2.40	1.27
cd630_05190	05190		Hypothetical protein	−2.00	−2.59	0.57	Membrane
	05260		Aminobenzoyl-glutamate transporter	NDe	ND	ND
cd630_05250	05250		Peptidase	−2.15	−1.08	NC
cd630_17671 cd630_17680	17671 17680		Hypothetical protein Membrane protein	−2.18 −2.23	−1.99 −2.18	NC 0.56	Membrane
cd630_17810	17810	walA	Lipoprotein	−2.76	−5.14	NC	SPII
cd630_17820	17820	walR	TC response regulator	−2.51	−3.91	3.81
cd630_17830	17830	walK	TC sensor histidine kinase	−2.33	−4.03	NC	Membrane
cd630_17840	17840	truA2	tRNA pseudouridine synthase A	−2.34	−3.92	NC
cd630_05300	05300		Membrane protein	−2.47	NC	NC	Membrane
	05290		Membrane protein	NC	NC	NC	Membrane
	05280		Amidohydrolase	NC	NC	0.92
cd630_28620	28620		Peptidase	−2.53	−1.43	2.29

^aLists other operon members in order.

^bFold change comparing WalR overexpression to negative-control strain.

^cPredictions by SignalP (SPI, predicted to be a substrate of signal peptidase I; SPII, predicted to be a substrate of signal peptidase II) and by BUSCA (membrane). The BUSCA prediction is given only for those genes that were not already predicted to be SPI or SPII substrates by SignalP.

^dNC, no change, i.e., the fold change was close to 1 and the P value was >0.05.

^eND, CD630_05260 is a pseudogene that was not detected by the RNA-seq analysis tool.

The CRISPRi guide targets the first gene in the operon, walA, which was repressed 6.8-fold in the absence of xylose and 35-fold with low xylose. As expected, polarity resulted in repression of the downstream genes as well, ~5-fold with leaky CRISPRi, increasing to ~15-fold with low xylose. The observed polarity confirms the predicted operon structure, including that a seemingly unrelated (and, according to Tn-seq, nonessential) gene for a tRNA modification enzyme is indeed part of the operon.

Unexpectedly, the Wal-OFF gene set comprises mostly transporters and genes for various metabolic functions. Not counting the members of the wal operon, only 1 of the genes is predicted to be exported and 15 genes (25%) are expected to be membrane protein genes, a percentage comparable to the genome as a whole (see Materials and Methods). Remarkably, the Wal-OFF regulon does not include a single annotated PG metabolism gene. However, we observed ~5-fold repression of the dltDABC operon, which is responsible for the addition of d-alanine to teichoic acids (49). We identified a predicted WalR binding site immediately upstream of dltDABC, suggesting direct regulation by WalR. This putative WalR binding site had previously been recognized as one of two direct repeat sequences in the dlt promoter region (49) (Table S4). d-Alanylation of teichoic acids affects PG hydrolase activity (50, 51) and vancomycin resistance (52, 53) in some bacteria. In C. difficile, a dltD insertion mutant is sensitized to vancomycin and some antimicrobial peptides, but the effects are <2-fold (52). Only one gene in Table 3 codes for an S-layer protein—the hemagglutinin/adhesin gene cwpV, which was ~16-fold induced in no xylose. But this gene is subject to phase variation (54), leading us to suspect that it may not be a true member of the WalR regulon.

Comparison of the Wal-ON and the Wal-OFF gene sets reveals little overlap but opposing trends.

There are only two proteins common to the Wal-ON and the Wal-OFF gene sets, cd630_20160 (hypothetical protein) and cd630_28620 (annotated as a membrane peptidase). However, a large fraction of the 79 Wal-ON genes that met the 4-fold cutoff trended in the opposite direction when evaluated under Wal-OFF conditions, even though the magnitude of change is not as large (Fig. 5A and Table 2). A similar observation was made when genes affected 4-fold or more under Wal-OFF conditions were examined under Wal-ON conditions (Fig. 5B and Table 3).

FIG 5.

Comparison of transcript changes between Wal-ON and Wal-OFF conditions. (A) Heat map showing the 79 genes whose transcript abundance changed ≥4-fold when walR was overexpressed (Wal-ON) in comparison to when the wal operon was silenced with CRISPRi (Wal-OFF). (B) Heat map showing the 49 genes whose transcript abundance changed ≥ 4-fold when the wal operon was silenced (Wal-OFF) in comparison to when walR was overexpressed (Wal-ON). Green, induced; red, repressed.

Selected members of the WalR regulon were confirmed by plasmid-based reporter fusions.

To confirm results of the RNA-seq analysis, we constructed rfp transcriptional fusions (55) to the promoter regions for seven genes from the Wal-ON and Wal-OFF gene sets. Four genes were chosen primarily because they were highly induced when WalR was overexpressed: cd630_07390, pgdA, cd630_07380, and cd630_08670 (Table 2). Additional considerations included that pgdA has an obvious connection to PG biogenesis (36, 37), while cd630_08670 has a predicted WalR-binding site (Table S4). Two genes were chosen because they were the most strongly repressed upon CRISPRi knockdown of the wal operon and are preceded by a predicted WalR-binding site: cd630_28620 and cd630_05300 (Table 3 and Table S4). In addition, cd630_28620 is one of only two genes that made the 4-fold cutoff under both Wal-ON and Wal-OFF conditions. The final gene chosen for confirmation with an rfp reporter was dltD, which incorporates d-alanine into teichoic acids (49) and was modestly downregulated by CRISPRi knockdown of the wal operon. dltD stood out because it has a predicted WalR-binding site and is one of the few cell envelope-associated genes that was at least 4-fold repressed under Wal-OFF conditions (Table 3 and Table S4).

Reporter plasmids were constructed by PCR amplifying ~300 bp upstream of the start codon for each selected gene. These fragments contain transcriptional start sites for five of the seven genes (Clost-Base database [56]); start sites for the remaining two genes have not been mapped. The PCR products were cloned into an RFP reporter plasmid. The resulting plasmids were conjugated into the appropriate Wal-ON and Wal-OFF strain pairs: 630Δerm versus 630Δerm P_xyl::walR and 630Δerm with sgRNA-walA1 versus 630Δerm sgRNA-neg. Reporter strains were grown under the same conditions as used for the RNA-seq experiments, except that the Wal-OFF condition was limited to growth in the absence of xylose, i.e., assaying the effect of leaky dCas9 expression on the reporter gene. Harvested cells were fixed and transferred to aerobic conditions to allow RFP to mature. Fluorescence was quantified by flow cytometry.

We found that all RFP reporters responded in the expected direction when walR expression was manipulated, thus confirming the RNA-seq results by an orthogonal method (Fig. 6). However, the fold changes were uniformly about 4-fold greater by RNA-seq than by RFP fluorescence. The reasons for this difference are not known. As an aside, having RNA-seq and RFP fluorescence data for the same genes enabled us to show that these are correlated, i.e., baseline RFP fluorescence was higher for genes that returned more reads in RNA-seq experiments (Fig. S5A). This correlation was expected and further indicates that our RNA-seq data are robust.

FIG 6.

Confirmation of RNA-seq results with RFP reporter fusions to selected genes. Cells harboring plasmids with transcriptional fusions of the red fluorescent protein mCherryOpt to the indicated promoters were grown as per RNA-seq conditions, either Wal-ON (green) or Wal-OFF (red). Red fluorescence was measured by flow cytometry and is graphed as the mean log₂ fold change from two independent experiments with two technical replicates. Error bars represent the SD of all four measurements. For comparison, the mean log₂ fold change of the same genes as determined by RNA-seq is also shown. The dltD reporter was graphed only under Wal-OFF conditions because this fusion was induced by xylose even in the absence of P_xyl::walR. Genes with asterisks represent fusions that were constructed to the corresponding promoter regions from strain R20291.

Altering the expression level of selected regulon genes does not induce the wal reporter.

In B. subtilis, the WalR regulon can be induced or repressed by manipulating expression of walR-regulated genes involved in PG metabolism (11, 57). We therefore asked whether manipulating expression of wal regulon genes in C. difficile would induce or repress expression of a chromosomal P_{cd630_07390}::rfp reporter. The reporter was chosen because this promoter was highly induced under Wal-ON conditions as determined by RNA-seq (38-fold) or a plasmid-based RFP reporter fusion (7-fold). The chromosomal P_{cd630_07390}::rfp reporter was validated by confirming ~10-fold-increased red fluorescence upon overexpression of walR from a P_xyl plasmid (Fig. S5B). Next, we introduced a panel of plasmids that allowed us to directly target 13 WalR regulon genes by overexpression and/or CRISPRi. These 13 genes were chosen for various reasons, including large fold changes in expression, relevance to cell wall biogenesis, and the presence of WalR binding sites (Table S3). Of note, the gene set included genes encoding putative cell wall amidase and the predicted PG hydrolase with an NlpC/P60 domain. Unfortunately, neither overexpression nor CRISPRi knockdown of the selected wal regulon genes altered expression of the P_{cd630_07390}::rfp reporter. Further work will be needed to determine whether these negative results reflect technical difficulties or more fundamental differences in Wal signaling between B. subtilis and C. difficile.

Perturbation of selected Wal regulon genes individually does not lead to any Wal-associated phenotypic defects.

As noted above, one motivation for determining the Wal regulon was to identify genes that contribute to Wal phenotypes. Recall that induction of P_xyl::walR decreases viability, causes cells to become phase bright, and reduces autolysis in buffer containing 0.01% Triton X-100 (Fig. 3). Two of the genes most highly induced under Wal-ON conditions encode the hypothetical proteins CD630_07380 and CD630_07390. However, overproduction of these proteins either individually or from P_xyl plasmids did not result in any obvious phenotypic changes (Fig. 7). Because these genes are transcribed divergently (35), we needed to reverse the direction of one gene to express them together and decided to clone both possible arrangements: cd630_07380-cd630_07390 and cd630_07390-cd630_07380. Likewise, a CRISPRi plasmid with an sgRNA that targets cd630_27680 had no effects on growth, morphology, or lysis (Fig. 7); this gene encodes PG hydrolase with two predicted WalR-binding sites and is among the most strongly repressed genes under Wal-ON conditions.

FIG 7.

Perturbation of individual Wal-ON regulon genes does not replicate the Wal-ON phenotype. Overnight cultures of 630Δerm harboring overexpression plasmids were subcultured into TY-Thi to an OD₆₀₀ of 0.03, grown to an OD₆₀₀ of 0.2, and induced with 3% xylose for 3 h. Overnight cultures of 630Δerm harboring CRISPRi plasmids were subcultured into TY-Thi containing 1% xylose to an OD₆₀₀ of 0.03 and grown to an OD₆₀₀ of ~0.8. Each culture was then examined by microscopy (bar = 5 μm) (A) and tested in the lysis assay (B). Induction of walR is the only condition that achieved phase-bright cells and slowed lysis. The plasmids used were pBZ101 (EV), pCE691 (P_xyl::walR), pIA112 (P_xyl::cd630_07380), pIA113 (P_xyl::cd630_07390), pIA114 (P_xyl::cd630_07380-07390), pIA115 (P_xyl::cd630_07390-07380), pIA34 (sgRNA-neg), pCE744 (sgRNA-cd630_27680-1), and pCE745 (sgRNA-cd630_27680-2).

CRISPRi silencing of the wal operon is lethal, and the terminal phenotypes include lysis and loss of rod shape as reflected by an abundance of curved cells (Fig. 2). However, none of these phenotypes were observed when we used a P_xyl::cwpV plasmid to overexpress the most highly induced gene from the Wal-OFF RNA-seq data set, which we tested even though cwpV is subject to phase variation and might not be part of the WalR regulon. We also observed no effect when we used CRISPRi to knock down expression of dltD, which was among the more strongly repressed Wal-OFF genes and is preceded by a putative WalR-binding site (Fig. 8). While this study was not an exhaustive evaluation of all regulon members, our findings suggest that the phenotypic defects observed upon perturbation of the Wal system are not due to changes in expression of any single gene but are cumulative in nature, much like what has been observed in B. subtilis and S. aureus (8–10, 12).

FIG 8.

Perturbation of individual Wal-OFF regulon genes does not replicate the Wal-OFF phenotype. (A) Phase-contrast microscopy of strains harboring overexpression plasmids or CRISPRi plasmids after induction with xylose as described in the legend to Fig. 7. Bar = 5 μm. (B) Viability assay. Overnight cultures were serially diluted and spotted onto TY plates with or without 1% xylose. Plates were photographed after incubation overnight. CRISPRi knockdown of the wal operon was the only condition that resulted in curved cells, lysis, or a viability defect. The plasmids used were pBZ101 (EV), pCE791 (P_xyl::cwpV), pIA34 (sgRNA-neg), pIA50 (sgRNA-walA1), and pCE738 (sgRNA-dltD).

DISCUSSION

In C. difficile the WalRK TCS is predicted to reside in a four-gene operon: walA-walR-walK-truA2. We altered expression of the WalRK regulon by complementary approaches: depleting C. difficile of WalRK by CRISPRi knockdown of the entire operon and overproduction of WalR under P_xyl control. RNA-seq revealed that the wal operon is not autoregulated in C. difficile, as evidenced by a lack of induction when WalR (or WalR^D54E) was ectopically expressed from a P_xyl promoter. Wal operons are not autoregulated in other Firmicutes either. RNA-seq also confirmed that the genes constitute an operon, because CRISPRi targeting walA resulted in roughly equivalent knockdown of the three downstream genes. Further studies will be needed to establish the specific role of each gene in WalRK signaling. We suspect that walA is a bona fide part of the Wal system because all wal operons studied to date include accessory genes for membrane proteins that modulate WalK signaling (8, 10). But a rationale for why truA2, which encodes a nonessential tRNA modification enzyme, should have come to reside in the wal operon is not obvious. This might be the result of evolutionary happenstance.

As expected, extensive knockdown of the wal operon with CRISPRi and strong overproduction of WalR from a P_xyl plasmid were both lethal. Phenotypic defects depended on how walRK expression was manipulated, but they included elongation, loss of rod shape (curved or wavy cells), phase-bright cells, enhanced or reduced autolysis, and altered sensitivity to antibiotics that target PG synthesis. Similar phenotypes have been observed upon manipulating Wal signaling in other organisms (2, 6, 7, 9, 10, 12, 17, 18). But there are some intriguing differences. In B. subtilis, in which phosphorylated WalR promotes elongation, artificial upregulation of the Wal system causes cells to become elongated, while artificial downregulation causes cells to become short (10, 18). In C. difficile, however, overexpression of WalR from P_xyl had no obvious effect on cell length, while CRISPRi knockdown of the wal operon led to elongation.

Vancomycin resistance presents another example of an inverse phenotype. In S. aureus, downregulation of walRK expression increases resistance to vancomycin while upregulation of walRK decreases vancomycin resistance (58, 59). Vancomycin-intermediate S. aureus (VISA) strains, both laboratory derived (60) and isolated in the clinic (61), often have mutations in walR or walK that downregulate the system (62). In contrast, we found that partial CRISPRi knockdown of the wal operon in C. difficile had the opposite effect, namely, a modest increase in sensitivity to vancomycin. Vancomycin is a front-line treatment for C. difficile infections, so our findings suggest that a small molecule that interferes with Wal signaling might enhance the efficacy of vancomycin therapy.

RNA-seq identified over 150 genes whose expression responds to manipulation of Wal signaling in C. difficile. Although this gene set is diverse, there are themes: many cell envelope proteins, many hypothetical proteins, and about 10 proteins with various connections to PG metabolism. There are also 7 genes for S-layer proteins, consistent with the need to coordinate S-layer assembly and remodeling with growth and turnover of the PG sacculus. The PG hydrolase genes are of particular interest because they are prominent members of the WalR regulon in other Firmicutes (5, 12–16) and ectopic expression of one or more of the corresponding enzymes can render walRK no longer essential regulators (6, 17, 18). In C. difficile, Wal-ON conditions induced two putative cell wall amidases (cwp16 and cwp17) and repressed one potential d,l-endopeptidase (cd630_27680). Direct control of cd630_27680 by WalR is suggested by the presence of two matches to the WalR-binding site consensus sequence about 100 nucleotides upstream of the start codon (Table S4). The importance of these genes for any Wal-related phenotype is unclear, however, because overexpression and CRISPRi knockdown experiments did not result in any phenotypic changes or alter expression of an RFP reporter that is induced under Wal-ON conditions. In addition to the putative hydrolase genes, several genes in the WalR regulon may play a role in controlling PG hydrolase activity. Deacetylation of PG N-acetylglucosamine (GlcNAc) (63), m-DAP amidation (39), and d-alanylation of teichoic acids (32, 50, 51, 64) have all been shown to affect hydrolase activity in other organisms.

S. pneumoniae is so far the only organism in which a single essential gene, the PG hydrolase gene pcsB, has been identified to be responsible for the essentiality of the wal system (6, 7). In B. subtilis, essentiality of the Wal system arises from the cumulative impact of aberrant expression of multiple genes involved in turnover of PG (5, 10, 11). Wal essentiality is also polygenic in nature in S. aureus (8). This seems to be the case in C. difficile as well. Although CRISPRi silencing of the wal operon leads to cell lysis, the Wal-OFF gene set (Table 3) does not include any genes determined to be essential by transposon mutagenesis (27). In contrast, the Wal-ON regulon (Table 2) includes 3 essential genes. Two of these are induced by WalR, genes for a DNA ligase (ligA, cd630_33090) and a putative calcium-chelating protein (cd630_26810), and one is repressed, the gene for a PG precursor synthesis protein MurE (cd630_26640). It is possible that repression of murE contributes to death during overexpression of WalR but not during CRISPRi knockdown of the wal operon, because murE expression did not change under that condition. Overall, our findings suggest that essentiality of WalRK is driven by a deleterious imbalance of multiple processes. (As an aside, it is worth noting that repression of murE under Wal-ON conditions is not what one would expect if WalR promotes increased PG synthesis, as in B. subtilis [11, 57].)

Here, we have provided an initial delineation of Wal phenotypes and the WalR regulon in C. difficile. Our findings raise a number of exciting questions. Is the protein WalA involved in WalRK signaling, and if so, how? What signals are sensed by WalK? Of note, owing to differences in cross-linking (23), the PG hydrolysis products implicated in regulating WalK activity in B. subtilis (11) are not very abundant in C. difficile. Moreover, C. difficile WalK is missing an intracellular PAS domain found in most other WalK proteins, suggesting that C. difficile WalK senses different signals and/or transmits those signals differently. Another open question is which WalR regulon genes are controlled directly by WalR. It will also be important to determine the functions of the WalR-regulated genes, especially those encoding the many hypothetical proteins. Answers to these questions are likely to provide novel insights into cell wall biogenesis and might point the way toward improved therapies against this important pathogen.

MATERIALS AND METHODS

Strains, media, and growth conditions.

Bacterial strains are listed in Table 4. C. difficile strains used in this study were derived from either 630Δerm or R20291, both of which have been sequenced. C. difficile was routinely grown in tryptone-yeast extract (TY) medium, supplemented as needed with thiamphenicol at 10 μg/mL (TY-Thi). TY medium consisted of 3% tryptone, 2% yeast extract, and 2% agar (for plates). Brain heart infusion (BHI) medium was prepared per the manufacturer’s (Difco) instructions. C. difficile strains were maintained at 37°C in an anaerobic chamber (Coy Laboratory Products) in an atmosphere of 10% H₂, 5% CO₂, and 85% N₂. Escherichia coli strains were grown in LB medium at 37°C with chloramphenicol at 10 μg/mL (Cam10) or ampicillin at 100 μg/mL (Amp100) as needed. LB medium contained 1% tryptone, 0.5% yeast extract, 0.5% NaCl, and 1.5% agar (for plates). OD₆₀₀ measurements were made with the WPA Biowave CO8000 tube reader in the anaerobic chamber.

TABLE 4.

Strains used in this study

Strain	Genotype and/or description	Source or reference(s)
E. coli
OmniMAX-2 T1R	F′ [proAB+ lacIq lacZΔM15 Tn10 (Tetr) Δ(ccdAB)] mcrA Δ(mrr-hsdRMS-mcrBC) φ80(lacZ)ΔM15 Δ(lacZYA-argF) U169 endA1 recA1 supE44 thi-1 gyrA96 relA1 tonA panD	Invitrogen
HB101/pRK24	F− mcrB mrr hsdS20(rB− mB−) recA13 leuB6 ara-14 proA2 lacY1 galK2 xyl-5 mtl-1 rpsL20	66, 67
C. difficile
R20291	Wild-type C. difficile strain from UK outbreak (ribotype 027)
630Δerm	Spontaneous erythromycin-sensitive derivative of strain 630 (ribotype 012)	76
CRG1496	630Δerm ΔpyrE	29
UM275	630Δerm with Pveg::rfp downstream of pyrE	26
UM554	630Δerm with Pxyl::dCas9 Pgdh::sgRNA-neg downstream of pyrE	This study
UM555	630Δerm with Pxyl::dCas9 Pgdh::sgRNA-walA-1 downstream of pyrE	This study
UM556	630Δerm with Pxyl::dCas9 Pgdh::sgRNA-walA-2 downstream of pyrE	This study
UM626	630Δerm with Pxyl::walR downstream of pyrE	This study
UM628	630Δerm with Pxyl::walRD54E downstream of pyrE	This study
UM926	630Δerm with Pcd630_0739::mCherryOpt downstream of pyrE	This study

Plasmid and strain construction.

All plasmids are listed in Table 5; an expanded version of this table which includes additional information relevant to plasmid assembly is provided in Table S2A in the supplemental material. Plasmids were constructed by isothermal assembly (65) using reagents from New England Biolabs (Ipswich, MA). Regions of plasmids constructed using PCR were verified by DNA sequencing. The oligonucleotide primers used in this work were synthesized by Integrated DNA Technologies (Coralville, IA) and are listed in Table S2B. All plasmids were propagated using OmniMax 2-T1R as the cloning host, transformed into HB101/pRK24 (66, 67), and then introduced into C. difficile strains by conjugation. Chromosomal fusions at the pyrE locus in 630Δerm were constructed by allelic exchange (29) using C. difficile CRG1496 (630Δerm ΔpyrE) as a pyrE-deficient recipient. The allelic exchange restored a functional pyrE gene.

TABLE 5.

Plasmids used in this study

Plasmid	Relevant features	Reference or comment
pAP114	Pxyl::mCherryOpt catP	26
pBZ101	Pxyl empty vector	This study
pCE636	PdltD::mCherryOpt catP	This study
pCE738	Pxyl::dCas9-opt Pgdh::sgRNA-dltD-1 catP	37
pCE691	Pxyl::walR catP	This study
pCE738	Pxyl::dCas9-opt Pgdh::sgRNA-dltD-1 catP	37
pCE741	Pxyl::cdr2656	This study
pCE744	Pxyl::dCas9-opt Pgdh::sgRNA-cd630_27680-1 catP	This study
pCE745	Pxyl::dCas9-opt Pgdh::sgRNA-cd630_27680-2 catP	This study
pCE789	Pxyl::dCas9-opt Pgdh::cwpV-1 catP	This study
pCE791	Pxyl::cwpV	This study
pDSW1728	Ptet::mCherryOpt catP	55
pDSW2037	Pxyl in integration vector pMTL-YN1C catP	This study
pDSW2053	Pxyl::dCas9opt-Pgdh::sgRNA-neg in pMTL-YN1C catP	This study
pDSW2055	Pxyl::dCas9opt-Pgdh::sgRNA-cd630_17810-1 in pMTL-YN1C catP	This study
pDSW2057	Pxyl::dCas9opt-Pgdh::sgRNA-cd630_17810-2 in pMTL-YN1C catP	This study
pIA33	Pxyl::dCas9-opt Pgdh::sgRNA-rfp catP	26
pIA34	Pxyl::dCas9-opt Pgdh::sgRNA-neg catP	26
pIA50	Pxyl::dCas9-opt Pgdh::sgRNA-walA-1 catP	This study
pIA51	Pxyl::dCas9-opt Pgdh::sgRNA-walA-2 catP	This study
pIA75	Pxyl::walR in pMTL-YN1C catP	This study
pIA76	Pxyl::walRD54E in pMTL-YN1C catP	This study
pIA79	Pxyl::dCas9-opt Pgdh::cd630_25040-1 catP	This study
pIA80	Pxyl::dCas9-opt Pgdh::cd630_25040-2 catP	This study
pIA81	Pxyl::dCas9-opt Pgdh::cd630_36010-1 catP	This study
pIA93	Pcdr_0665::mCherryOpt catP	This study; cdr_0665 is ortholog of cd630_07380
pIA95	Pcdr_0796::mCherryOpt catP	This study; cdr_0796 is ortholog of cd630_08670
pIA97	Pcdr_2753::mCherryOpt catP	This study; cdr_2753 is ortholog of cd630_28620
pIA98	Pcdr_0455::mCherryOpt catP	This study; cdr_0455 is ortholog of cd630_53000
pIA100	Pcd630_0739::mCherryOpt catP	This study
pIA101	Pcd630_0739::mCherryOpt in pMTL-YN1C catP	This stduy
pIA102	Pxyl::dCas9-opt Pgdh::cd630_05490 catP	This study
pIA103	Pxyl::dCas9-opt Pgdh::cd630_27940 catP	This study
pIA104	Pxyl::dCas9-opt Pgdh::cd630_27860 catP	This study
pIA105	Pxyl::dCas9-opt Pgdh::cd630_27910 catP	This study
pIA106	Pxyl::dCas9-opt Pgdh::cd630_10360 catP	This study
pIA107	PpgdA::mCherryOpt catP	This study
pIA108	Pxyl::dCas9-opt Pgdh::cd630_07380-1 catP	This study
pIA109	Pxyl::dCas9-opt Pgdh::cd630_07380-2 catP	This study
pIA110	Pxyl::dCas9-opt Pgdh::cd630_07390-1 catP	This study
pIA111	Pxyl::dCas9-opt Pgdh::cd630_07390-2 catP	This study
pIA112	Pxyl::cd630_0738	This study
pIA113	Pxyl::cd630_0739	This study
pIA114	Pxyl::cd630_0738-cd630_0739	This study
pIA115	Pxyl::cd630_0739-cd630_0738	This study
pMTL-YN1C	E. coli-C. difficile shuttle vector for inserting genes into C. difficile chromosome while restoring pyrE; colE1 RP4oriT-TraJ CB102ori-repH′ catP	29

Conjugation into C. difficile.

Our experiments required conjugating plasmids into C. difficile strains R20291 and 630Δerm. In the case of R20291, we encountered problems with low conjugation efficiencies even with heat shock as described previously (68). After testing several modifications of the procedure, we found that performing conjugations on filters reliably increased efficiency more than 10-fold (Fig. S6). We now routinely use filters for all our conjugations. Briefly, the E. coli donor strain (an HB101/pRK24 derivative harboring the cargo plasmid) was grown overnight in LB Amp¹⁰⁰ Cam¹⁰. The donor strain was collected gently by centrifugation of a 0.5-mL aliquot at 5,000 × g for 1 min and then washed with 1 mL of TY and pelleted again. The washed cell pellet was moved into the anaerobic chamber. R20291 was prepared for conjugation by a heat shock step (68). For this, a 200-μL aliquot of the overnight culture of the R20291 recipient was transferred to a 1.5-mL microcentrifuge tube and incubated at 48°C for 5 min in a Fisherbrand dry bath (with water-filled wells). Then 100 μL of heat-shocked R20291 was used to take up the pellet of E. coli donor cells. The strain mixture was then pipetted onto a 25-mm-diameter, 0.45-μm-pore-size Millipore mixed-cellulose filter (HAWP02500) placed on BHI plates. After incubation for 24 h at 37°C, cells were flushed from the membrane with 500 μL of TY and then 200 to 500 μL of the resulting bacterial slurry was plated onto TY amended with thiamphenicol (10 μg/mL), kanamycin (50 μg/mL), and cefoxitin (8 μg/mL) to select for exconjugants. Conjugations into 630Δerm were done identically except that the heat shock step was omitted.

Viability assay.

The effect of CRISPRi silencing on plating efficiency was evaluated by making a 10-fold serial dilution of a culture grown overnight in TY-Thi and spotting 5 μL of each dilution onto TY-Thi agar with and without 1% xylose. Plates were photographed after overnight incubation (~18 h).

Microscopy.

Cells were immobilized using thin agarose pads (1%). Phase-contrast and fluorescent micrographs were recorded on an Olympus BX60 microscope equipped with a 100× UPlanApo objective (numerical aperture, 1.35). Micrographs were captured with a Hamamatsu Orca Flash 4.0 V2+ complementary metal oxide semiconductor (CMOS) camera. The image analysis tool MicrobeJ (69) was used to measure cell length and sinuosity, which is the length of the cell axis divided by the distance between the poles. A perfectly straight rod has a sinuosity value of 1, while a curved or wavy cell has a larger value. We classified cells as curved if their sinuosity was ≥1.03 (see Fig. S2 for examples). Viability of cells was assessed with LIVE/DEAD stain (Molecular Probes; L7012). A 1-mL culture sample was pelleted, washed with phosphate-buffered saline (PBS), and resuspended in 100 μL of PBS with 5 μM Syto 9 and 30 μM propidium iodide. Cells were incubated with the dyes for 15 min, then removed from the anaerobic chamber, and immediately imaged by microscopy. For propidium iodide red fluorescence we used filter set 41004 (Chroma Technology) with a 538- to 582-nm excitation filter, 595-nm dichroic mirror (long pass), and a 582- to 682-nm emission filter. Syto 9 green fluorescence was captured with filter set 41017 (Chroma Technology Corp.) with a 450- to 490-nm excitation filter, a 495-nm dichroic mirror (long pass), and a 500- to 550-nm emission filter.

Fixation protocol.

A 500-μL aliquot of cells in growth medium was added directly to a microcentrifuge tube containing 120 μL of a 5× fixation cocktail: 100 μL of a 16% (wt/vol) paraformaldehyde aqueous solution (Alfa Aesar, Ward Hill, MA) and 20 μL of 1 M NaPO₄ buffer (pH 7.4). The sample was mixed, incubated in the anaerobic chamber at 37°C for 30 min and then on ice 30 min, and removed from the chamber. The fixed cells were washed twice with 1 mL of PBS, resuspended in 50 μL of PBS, and left in the dark for 18 h to allow for chromophore maturation.

MIC determination.

Antibiotic sensitivity was determined in 96-well plates. A 2-fold dilution series of selected antibiotics was prepared in 50 μL TY medium. Wells were then inoculated with 50 μL of a diluted culture suspension (10⁶ CFU/mL; calculated OD₆₀₀, ~0.005). Plates were imaged and evaluated after 17 h at 37°C.

Lysis assay.

Cell cultures (1 mL) were removed from the anaerobic chamber, pelleted, and resuspended in 700 μL of 0.01% Triton X-100 in 50 mM NaPO₄ buffer (pH 7.4). Of this 700 μL, three 200-μL replicates were pipetted into wells of a clear, flat-bottom 96-well plate. The turbidity was measured at 600 nm every 15 min for 10 h in a plate reader (Tecan Infinite M200 Pro).

Flow cytometry.

Cells were analyzed at the Flow Cytometry Facility at the University of Iowa using the Becton, Dickinson LSR II instrument with a 561-nm laser, a 610/20-nm-band-pass filter, and a 600 LP dichroic filter. Data were analyzed using BD FACSDiva software.

Culture growth for RNA-seq samples.

To obtain robust, quality data, all experiments were performed on three different days for biological replicates. Cultures were grown in 100 mL of TY as follows. For the Wal-ON condition, to characterize the effect of overexpressing walR, strains were subcultured to an OD₆₀₀ of ~0.05, grown to an OD₆₀₀ of 0.2, induced with xylose to 3%, and grown for an additional 3h (Fig. S4A). For the Wal-OFF condition, to characterize the effect of partial knockdown of the wal operon under leaky CRISPRi conditions, strains were subcultured to an OD₆₀₀ of ~0.03 and grown to an OD₆₀₀ of ~0.7 before harvesting (Fig. S4C). To test the effect of stronger knockdown of the wal operon, growth conditions were identified that allowed exposing cells to moderate levels of xylose (0.1%) without a significant loss of biomass. Strains were subcultured to an OD₆₀₀ of ~0.03, grown to an OD₆₀₀ of 0.3, diluted back to an OD₆₀₀ of ~0.1, induced with xylose to a final concentration of 0.1%, and harvested after an additional 2.5 h of growth (Fig. S4D). From each condition, 25 mL of culture was fixed by rapidly mixing with 25 mL of ice-cold 1:1 ethanol-acetone that had been brought into the anaerobic chamber on dry ice. Fixed cells were stored at −80°C until further workup.

RNA isolation and RNA-seq.

Fixed culture samples were thawed and then pelleted for 10 min at 8,000 × g and 4°C, and pellets were washed once with 500 μL of 1% β-mercaptoethanol (β-ME) in water. Subsequent steps were an adaptation of the Qiagen RNeasy minikit procedure. The washed pellet was resuspended in 1 mL of the Qiagen-provided RLT buffer amended with 1% β-ME and transferred to a 2-mL screw-cap tube filled with glass beads (0.1-mm diameter) to a height of ~2 to 3 mm. Cells were lysed with the FastPrep-24 homogenizer with two cycles at 6 m/s for 45 s, resting on ice for 3 min between cycles. The homogenized material was pelleted at 14,000 × g for 15 min at 4°C. The supernatant was removed, adjusted to 900 μL with RLT buffer/β-ME, mixed with 500 μL of ethanol, and loaded onto the Qiagen RNeasy spin column. Subsequent steps followed the manufacturer’s protocol. The RNA was eluted with 50 μL of water and DNase treated twice (Ambion Turbo DNA free). Absence of DNA was evaluated by amplifying a 364-bp region of mldA with Taq DNA polymerase for 30 cycles. The RNA integrity number (RIN) was determined with an Agilent Bioanalyzer at the Iowa Institute for Human Genetics, Genomics Division. All samples had a RIN of 9.4 or higher. Typical yields from 25 mL of culture were 8 to 20 μg of RNA at 200 to 500 ng/μL.

RNA samples were submitted to the Microbial Genome Sequencing Center (MiGS) in Pittsburgh, PA. MiGS performed Illumina stranded RNA library preparation paired with RiboZero Plus (per the manufacturer’s specifications) and sequenced on a NextSeq 500 using a 75cyc high-output flow cell. Fastq files were trimmed and filtered using a combination of Trimmomatic (70) and FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Alignment, normalization, and differential expression were analyzed with SeqMan NGen version 17.2 (DNASTAR, Madison, WI). Annotation was imported from NCBI with additional information obtained from progenomes (71). All subsequent analysis was performed in Excel and GraphPad Prism 9.

Bioinformatics.

Gene sequences and operon organization were obtained from the BioCyc data collection (35, 72). SignalP 5.0 (73) was used to search for type I and type II signal peptides. BUSCA (74) was used to predict protein cellular location. When all C. difficile proteins were run through the prediction programs, SignalP predicted ~7% to be exported. BUSCA predicted ~4% to exported and ~25% to be membrane associated. Cell wall association in Tables 2 and 3 was reported as follows: (i) signal peptidase substrate if predicted by SignalP and (ii) for all remaining genes, membrane protein if predicted by BUSCA.

Putative WalR binding sites were identified with Virtual Footprint (75) by querying with (i) the B. subtilis consensus motif, TGTWAH-N5-TGTWAH (14), (ii) the same motif, but allowing one mismatch, or (iii) TGTNDH-N5-BKBWRN (8). Searches were limited to the intergenic region of the genome and generated 29, 522, and 684 hits, respectively.

Data availability.

RNA-seq data were submitted to the NCBI GEO repository and assigned accession number GSE200346.