Redefining the Clostridioides difficile σB Regulon: σB Activates Genes Involved in Detoxifying Radicals That Can Result from the Exposure to Antimicrobials and Hydrogen Peroxide

Sigma B is the alternative sigma factor governing stress response in many Gram-positive bacteria. In C. difficile, a sigB mutant shows pleiotropic transcriptional effects. Here, we determine genes that are likely direct targets of σB by evaluating the transcriptional effects of σB overproduction, provide biochemical evidence of direct transcriptional activation by σB, and show that σB-dependent genes can be activated by antimicrobials. Together, our data suggest that σB is a key player in dealing with toxic radicals.

by the host, the resident microbiota, or given externally during medical therapy (5). The physiological response of C. difficile to these insults and the inflammatory responses triggered by CDI can result in the production of reactive oxygen species (ROS), reactive nitrogen species (RNS), and nitric oxide (NO) (2,6).
Bacteria need to adapt to changing environmental conditions, including stresses, by adapting their physiology in a timely manner. This is achieved by fast transcriptional reprogramming, followed by briefly delayed changes at the translational level (7). The alternative sigma factor sigma B ( B , encoded by the sigB gene), which regulates the general stress responses in a variety of Gram-positive organisms, is central to the maintenance of cellular homeostasis during stress adaptation (8,9).
Sigma factor B activity in Firmicutes species is regulated at the protein level by a partner-switching mechanism in which the anti-sigma factor RsbW binds and inhibits B association with the RNA polymerase under nonstressed conditions. When a Bactivating stress is sensed, the dephosphorylated anti-anti-sigma factor RsbV sequesters RsbW, allowing for the association of free B with the RNA polymerase core enzyme (8,10). In C. difficile, the phosphatase RsbZ is responsible for RsbV dephosphorylation (11). The tight regulation of B activity by a partner-switching mechanism is necessary, as the energy burden associated with B activity was found to be disadvantageous in several different organisms (12,13).
Despite the burden associated with its expression, B is essential for survival for several pathogenic bacterial species in response to host-dependent stressors or antimicrobials. For example, in Listeria monocytogenes, B is involved in counteracting the effects of the acidic pH encountered in the stomach and upon invasion of intestinal epithelial cells in the lysosome (14,15). In Staphylococcus aureus, B overproduction leads to thickening of the cell wall and increased resistance to beta-lactam antimicrobials (16). The sigB homologue sigF of Mycobacterium tuberculosis is induced by small amounts of rifamycin (17). Analogously, Bacillus subtilis B is involved in resolving a rifampin-induced growth arrest (18). There is also evidence for the involvement of B in C. difficile in the response to antimicrobial substances. Mutants of sigB show increased susceptibility to rifampin and mitomycin C and are also more sensitive to hydrogen peroxide, nitroprusside, and di᎑ethylamine NONOate (19). However, the underlying molecular mechanisms remain unknown. Finally, indirect activation of B -dependent genes as the result of a gene dosage shift has been demonstrated for C. difficile exposed to DNA polymerase inhibitors such as the phase II drug ibezapolstat/ ACX-362E (20).
In this study, we demonstrate that B overexpression is detectable and is tolerated for short periods of time. This allowed for the experimental identification of a set of genes that is most likely directly regulated by B by performing transcriptome analyses under conditions of acute B overexpression. The results obtained show that genes involved in the oxidative and nitrosative stress response form the core of the regulon. Additionally, we show that various antimicrobials and hydrogen peroxide induce the expression of B -regulated genes in a B -dependent manner, suggesting a link between the lethal exposure to antimicrobials and oxidative and nitrosative stresses in C. difficile.

RESULTS
C. difficile B is measurably overproduced upon induction of the sigB gene. Previous investigations of B in C. difficile have used a sigB mutant and characterized its gene expression in the stationary growth phase in comparison with that of a wild-type strain (19). Although informative, this method is likely to result in indirect effects of B due to stationary-phase heterogeneity, prolonged incubation, and possible positive or negative feedback in the B regulatory circuit. To circumvent these issues and identify genes likely to be regulated by B directly, we set out to uncouple sigB expression from its native regulatory circuit by expressing it from an inducible promoter.
First, in order to confirm overproduction of B , we measured cellular B levels using immunoblotting. For this purpose, we heterologously overproduced and purified B containing a C-terminal His tag (Fig. 1A) and used this protein to raise a polyclonal antiserum. Corresponding polyclonal antibodies were affinity purified to prevent unspecific immune reactions.
Next, we set out to validate the overproduction of B in transconjugant C. difficile cells harboring plasmids containing sigB under the control of the anhydrotetracycline (ATc)-dependent promoter P tet (21). For this purpose, B was produced in a sigB mutant background (strain IB58; sigB::CT P tet -sigB). As a control, we introduced a nonrelated expression construct in the same background (IB61; sigB::CT P tet -sluc opt ) such that this control strain carries a plasmid with the same replicon, resistance marker, and inducible promoter.
We expected a signal at approximately 30 kDa in Western blot experiments for cells grown in the presence of the inducer ATc for strain IB58, but not for the uninduced cultures of IB58 or the control strain IB61. Additionally, by growing cultures in the presence or absence of thiamphenicol, we investigated whether overproduction of B required selection for the P tet -sigB expression plasmid.
When strains were grown in brain heart infusion (BHI) broth supplemented with 0.5% (wt/vol) yeast-extract (BHIY) supplemented with 20 g/ml lincomycin and induced for 1 h with or without 100 ng/ml ATc in the presence or absence of 20 g/ml thiamphenicol, we did not detect any signal at the molecular weight expected for B in the ATc-induced control samples (sigB::CT P tet -sluc opt ) or in any of the uninduced samples (Fig. 1B). In contrast, after 1 h of induction, a clear band of the expected molecular weight of B (Ϸ30 kDa) was observed only in the IB58 (sigB::CT P tet -sigB) samples (Fig. 1B). Plasmid selection by inclusion of thiamphenicol in the growth medium did not influence B overproduction in this time frame, which might have occurred as a tradeoff between B overexpression and cellular toxicity (see further below).
We conclude that the affinity-purified rabbit anti-B antibody is specific for B and can be used for its detection in lysates of C. difficile. Furthermore, it is possible to uncouple sigB expression from its tight regulatory network by ATc-inducible overexpression for 1 h in trans.
Prolonged overexpression of B is lethal and leads to a loss of plasmids harboring P tet -sigB. Above, we showed that it is possible to overproduce B in C. difficile and that this is tolerated by the bacterium for 1 h. This observation is somewhat 6ϫHis was used to generate a Clostridiodes difficile B -specific antibody for intracellular detection. (A) Coomassie blue-stained 12.5% SDS-PAGE gel of purified recombinant B 6ϫHis . (B) Western blot using affinity purified B antibody (1:500) on strains IB58 (sigB::CT P tet -sigB) and IB61 (sigB::CT P tet -sluc opt ). Cells were grown in lincomycin (20 g/ml) and in the presence or absence of thiamphenicol (20 g/ml) until an optical density at 600 nm (OD 600 ) of Ϸ0.3, after which the indicated samples were induced with 100 ng/ml anhydrotetracycline (ATc). Samples were collected directly after the addition of ATc (or at the time ATc would have been added in the uninduced controls) at T ϭ 0 h and after 1 h of induction (T ϭ 1).
at odds with the previously reported toxic nature of overproduced B (8,11). To reconcile these two observations, the effect of long-term overexpression of sigB and the stability of the plasmids used for B overproduction under such conditions were investigated. First, overnight cultures of 630Δerm (wild-type), AP34 (P tet -sluc opt ), and JC096 (P tet -sigB) strains were adjusted for their optical density at 600 nm (OD 600 ) values and 10-fold serially diluted. Subsequently, 2-l spots per dilution were made on selective (20 g/ml thiamphenicol) and nonselective BHIY agar plates, some of which contained 200 ng/ml ATc to induce P tet -dependent gene expression. All plates were then incubated anaerobically for 24 h. On plates without thiamphenicol, regardless of the presence of the inducer ATc, comparable growth was observed for all three strains ( Fig. 2A). As expected, when selecting for the plasmid using thiamphenicol, no growth was observed for the susceptible 630Δerm strain (which lacks the catP gene contained on the expression vector). In the absence of the inducer, no difference in growth was observed for the vector control strain (AP34; P tet -sluc opt ) compared to the strain carrying the P tet -sigB plasmid (JC096). However, upon induction of sigB expression on selective plates, a 3-to 4-log growth defect was observed for the strain carrying P tet -sigB compared to the vector control strain. We conclude that prolonged induction of sigB expression is toxic when cells are cultured in the presence of thiamphenicol. Our results thus corroborate the finding that B overproduction is toxic to C. difficile cells in liquid culture (11).
The lethality associated with B overproduction was not seen when cells were grown without thiamphenicol in our experiment ( Fig. 2A). We considered two possible explanations for this observation. As thiamphenicol is used for ensuring plasmid maintenance, its absence might result in plasmid loss due to segregation or negative FIG 2 Overexpression of B is toxic for C. difficile and leads to plasmid loss. (A) Tenfold serial dilutions on brain heart infusion broth supplemented with 0.5% (wt/vol) yeast-extract (BHIY) agar plates of the 630Δerm (wild-type), AP34 (P tet -sluc opt ), and JC096 (P tet -sigB) strains. Similar results were obtained for strains IB58 and IB61 (data not shown). (B) Percentages of cells retaining the plasmid in AP34 (P tet -sluc opt ) and JC096 (P tet -sigB). (C) Percentages of cells retaining the plasmid in strains IB61 (sigB::CT P tet -sluc opt ) and IB58 (sigB::CT P tet -sigB). Percentages were calculated based on the ratio of CFU/ml of the paired selective (with thiamphenicol) and nonselective (without thiamphenicol) plates. *, P Ͻ 0.05; ****, P Յ 0.0001, as determined by an unpaired Student's t test (n ϭ 3). selection pressure when a toxic protein such as B is overproduced. The remaining cells that no longer express B would consequently be susceptible to thiamphenicol (due to the loss of catP) but might outgrow those carrying the plasmid. Alternatively, the combination of B and thiamphenicol might be toxic to the bacteria. To test whether plasmid loss was the cause of the observed lethality of bacteria overproducing B in the presence of thiamphenicol, cells from the plates without thiamphenicol (with and without ATc) were resuspended in phosphate-buffered saline (PBS) at 1.0 McFarland turbidity and 10-fold serially diluted in brain heart infusion (BHI) medium. Spots (10 l) of these dilutions were plated on plasmid-selective (thiamphenicol) and nonselective (no thiamphenicol) plates. Based on the ratio of CFU/ml of the selective and nonselective plates, the percentage of cells which lost their plasmid was calculated. If B overproduction led to the loss of the plasmid under conditions that do not select for its maintenance (no thiamphenicol), we expected significantly reduced growth on plates containing thiamphenicol. Although some plasmid loss was observed under uninduced conditions, as well as for the negative-control strain AP34 (P tet -sluc opt ), all cells originally containing the P tet -sigB plasmid (strain JC096) completely lost this plasmid upon induction of B overproduction with ATc (Fig. 2B). Similar results were obtained for sigB mutant strains IB58 (sigB::CT P tet -sigB) and IB61 (sigB::CT P tet -sluc opt ), indicating that the observed effects were solely due to in trans B overproduction and did not result from an interference of the native sigB regulatory network (Fig. 2C). Together, these results are consistent with a model in which the vector with the low-copy-number pCD6 replicon is rapidly eradicated upon expression of a gene (here sigB) that causes lethal defects (22,23). B primarily activates genes relating to oxidative/nitrosative stress responses. Above, we have shown that long-term overproduction B is detrimental and that this leads to loss of the expression plasmid in the absence of thiamphenicol (Fig. 2), but that B overproduction nevertheless could clearly be demonstrated when induction is limited to 1 h (Fig. 1C). Therefore, we used the time-limited induction to refine the previously proposed regulon (19) in both the presence and absence of thiamphenicol to strike a balance between potential secondary effects due to toxicity associated with B overproduction (with thiamphenicol), and loss of the expression plasmid from a subpopulation of cells (without thiamphenicol) ( Table 1). We compared transcriptome data from strain IB58 (sigB::CT P tet -sigB) to that of strain IB61 (sigB::CT P tet -sluc opt ). IB61 harbors a vector for the inducible expression of a luciferase gene that does not lead to any toxicity or growth phenotype (24).
We expected no genes or a limited number of genes to be differentially expressed (log 2 fold change [log2FC] of ՅϪ1.5 or Ն1.5 and adjusted P value of Ͻ0.05) under noninducing conditions. Indeed, we found only five differentially expressed genes in the P tet -sigB strain (IB58) compared to the P tet -sluc opt control (IB61) strain (hybridizations 1 and 3) (see Data Set S1 in the supplemental material). These genes were similarly positively (CD0583 and CD0584, both GGDEF domain-containing proteins [25], and CD2214 and CD2215, both potential transcriptional regulators [26]) and negatively 3 sigB::CT P tet -sluc opt (IB61) sigB::CT P tet -sigB (IB58) Lincomycin (20 g/ml), no thiamphenicol, no ATc 5 4 1 4 sigB::CT P tet -sluc opt (IB61) sigB::CT P tet -sigB (IB58) Lincomycin (20 g/ml), no thiamphenicol, ATc (100 ng/ml) 150 (145) 136 (132) 14 (13) a Numbers in brackets correspond to the number of differentially expressed genes after subtracting the differentially expressed genes identified in hybridizations 1 and (CD1616, an EAL domain protein [25]) regulated in all hybridizations, including those where sigB expression was not induced. These results suggest that the basis for the observed differential expression of these genes was vector specific but not dependent on B induction. These genes were therefore not investigated further and are excluded from the numbers discussed below. Upon induction of sigB expression, 145 genes were differentially expressed when strains were cultured without thiamphenicol (hybridization 4), and 178 genes were differentially expressed when thiamphenicol was present during cultivation (hybridization 2) ( Fig. 3 and Table 1 and Data Set S1). The majority showed an increase in expression upon induction of sigB expression (132 in the samples without thiamphenicol and 163 in the samples with thiamphenicol), consistent with its function as a sigma factor (27), while a minority revealed a decreased expression (13 in the samples without thiamphenicol and 15 in the samples with thiamphenicol). Of note, we observed only a minor difference in the number of differentially expressed genes between the cells grown in the absence and presence of thiamphenicol (33 genes).
Together, these results demonstrate a high level of consistency in the B regulon despite potential plasmid loss (when grown in the absence of thiamphenicol) or toxic effects (when grown in the presence thiamphenicol). Our results also show that B primarily activates gene expression.
We focused our further analyses on the data obtained from hybridization 2 (with ATc and thiamphenicol), as this condition provided the broadest data set (178 differentially expressed genes) for the redefinition of the B regulon under our experimental conditions (Data Set S1).
Of the 163 genes upregulated by B , the vast majority appeared to be associated with an response to oxidative stress, since they encode various oxidoreductases, peroxidases, and thioredoxin reductases (Table 2). Notably, approximately 51% of the  98 genes previously found to be upregulated under aerobic stress (7) were also positively regulated by B ( ) were also found to be induced by B , in agreement with previous findings (19). Our findings are recapitulated in a volcano plot (28), which clearly shows that genes with lower expression upon sigB induction (in blue) cluster close to the significance threshold, whereas those with increased expression (in red) show a larger fold change (Fig. 3). We calculated the Manhattan distance for each data point (see Data Set S2 in the supplemental material), and discuss the proteins encoded by the top 10 differentially expressed genes below.
CD0051A is a small hypothetical protein of unknown function. It does not contain any recognizable domains, and a secondary structure prediction using Phyre2 does not give any clues as to its potential function (29). CD0580 (GapN) is annotated as a glyceraldehyde-3-phosphate dehydrogenase (GAPDH), a key glycolytic enzyme, and contains an aldehyde dehydrogenase domain. Interestingly, its activity has been shown to be redox controlled in other bacteria and has been implicated in the response to reactive oxygen and nitrogen species (30)(31)(32). CD1623 is a putative oxidoreductase with similarity to FAD flavoproteins and rubredoxins. CD1690 (TrxA) and CD1691 (TrxB) are likely encoded in the same operon (33) and form a thioredoxin/thioredoxindisulfide reductase couple. CD0174 (CooS; InterPro family IPR010047), CD0175, and CD0176 are likely also encoded in a single operon (33) and function as carbon monoxide dehydrogenase and two putative oxidoreductases. As mentioned above, CD0174 has been implicated in aerobic/nitrosative stress, and it is likely that CD1623, CD1690, and CD1691 also function in this pathway. Finally, CD2115A encodes another small hypothetical protein; as for CD0051A, no function could be assigned on the basis of secondary structure prediction.
As the B regulon that we define here is substantially smaller than that previously reported, the major conclusion is that at least 32% of the B regulon is involved in positively regulating oxidative/nitrosative stress responses. In the previous investigation of the B regulon they were approximately 3.2% (Ϸ32/1,000) (19). Overall, we conclude that the core functions of the B regulon lie in the regulation of the detoxification response to oxygen and nitro radicals.
In vitro runoff transcriptions demonstrate direction activation of P cd0350 , P cd2963 , P cd3412 , and P cd3605 by B . Gene expression can directly or indirectly be influenced by B , and to date no attempts have been made to discriminate these possibilities biochemically (11,19). Despite the short time of induction and the uncoupling of B from its normal regulatory network, our analyses could possibly also have picked up indirect effects. To determine if the transcription of selected genes is directly activated by B , in vitro transcription runoff reactions were performed using purified B 6ϫHis and RNA polymerase core enzyme (RNAP core ) on the upstream regions of a selection of genes. The genes cd0350 (encoding a putative hydrolase involved in oxidative stress; Table 2), cd2963 (encoding an L,D-transpeptidase), cd3412 (encoding UvrB, involved in nucleotide excision repair), cd3605 (encoding a ferredoxin), and cd3614 (encoding a hypothetical protein involved in oxidative stress; Table 2) were selected on the basis of differential expression in our transcriptome analyses (Data Set S1) and those of others (19), availability of reporter constructs that could be used to generate a template for the in vitro transcription reactions (20), and/or the presence of a putative B consensus upstream in the upstream region (11). The gene cd0872 (encoding maltose O-acetyltransferase) was not differentially expressed in our transcriptome data and was thus included as a negative control. The promoter of the toxin A gene (tcdA) in combination with purified TcdR was used as a positive control for the assay, as previously described (34).
As expected, no in vitro transcript was observed for a linear DNA fragment containing P cd0872 incubated with purified B 6ϫHis and RNAP core under our experimental conditions, whereas a specific product was obtained for the positive control P tcdA in the presence of TcdR and RNAP core (Fig. 4). An RNAP core -and B 6ϫHis -specific signal was observed for fragments containing the putative promoter regions of the genes cd0350, cd2963, cd3412, and cd3605, demonstrating that expression of these genes was directed by B . For the fragments containing the putative promoter of cd3614, we did not get a consistent product in the in vitro transcription experiments, although some smearing is visible in the lane with RNAP core and B 6ϫHis . As cd3614 demonstrates clear differential expression in the DNA array experiments and it upstream region harbors the B consensus sequence WGWTT-N 13-17 -(G/T)GGTWA (19), we consider it likely that this gene is directly regulated by B and that our failure to obtain a discrete signal is due to our experimental conditions or to the lack of an auxiliary factor in our in vitro assays.
Overall, we provide the first biochemical evidence for direct B -dependent activation of several genes identified via transcriptome analyses as part of the B regulon in C. difficile.
Antimicrobials and hydrogen peroxide activate B -directed gene transcription. The redefined B regulon points toward a substantial role for B in coordinating the oxidative and nitrosative stress response, which could result from antimicrobial treatment. In order to test for the activation of B -dependent promoters by antimicrobials, we set up a plate-based luciferase reporter assay. In this assay, cells harboring Bdependent luciferase reporter constructs were plated on BHIY agar to give confluent growth and exposed to antimicrobials either through an epsilometer test (Etest) or through a filter disc. Subsequently, luciferase activity was imaged (for details, see Materials and Methods). A strain harboring P tet -sluc opt (AP34) served as negative control, as this promoter is not expected to respond in a B -dependent manner (Fig. 5A).
First, the B -dependent response to metronidazole was investigated. Metronidazole, formerly used as a first-line treatment for CDI, is believed to cause DNA damage through the formation of nitro radicals, although its exact mode of action remains unclear (6,35). To survey a full spectrum of metronidazole concentrations, we evaluated luminescence after 24-h incubation of a metronidazole Etest. If metronidazole treatment results in B -dependent activation of gene transcription, we expect to see a luciferase signal in the wild type but not in a B knockout background. In agreement with this, activation of P cd0350 was observed at the edge of the halo resulting from the metronidazole Etest in the wild-type background but not in the B knockout strain (Fig. 5B). No signal was observed for the negative control P tet -sluc opt (Fig. 5A). The observed B -dependent activation of gene expression at the edge of the halo but not further into the plate suggests that the metronidazole-induced, B -dependent activation of P cd0350 occurred close to the MIC. Expression of the luciferase from P cd2963 was found to be strictly dependent on B , as no luciferase activity was observed in the sigB knockout strain. However, there was limited to no increase in reporter gene expression in the presence of metronidazole. Metronidazole strongly activated transcription from P cd3412 at MIC levels of metronidazole, but this appeared to be independent of B in this assay since in the sigB mutant a similar induction was observed. Finally, in a manner comparable to that of P cd0350 , the activation of P CD3614 was strongly induced by metronidazole at values close to the MIC in a B -dependent manner, but residual activity was observed in the B knockout strain independent of metronidazole levels. We noted that metronidazole-induced promoter activation appeared to occur on the inside of the Etest halo, which might be attributed to the secretion of the luciferase reporter.
The observed diverse regulatory responses at different tested promoters during the treatment of C. difficile with metronidazole (with respect to basal level, sigB dependence, and induction) pointed toward a more complex regulatory network with the participation of B but also influenced by other factors. Antimicrobial-driven (and B -dependent) activation of B target genes could be specific to metronidazole or represent a more general response to cellular (toxic) stresses. Therefore, we evaluated the effects of different antimicrobial compounds and the radical producer H 2 O 2 as a positive control (19), using the P cd0350 reporter construct, as this promoter demonstrated the clearest B -dependent activation in the presence of metronidazole (Fig. 5B). We tested the cell wall biosynthesis inhibitor vancomycin, the protein synthesis inhibitor lincomycin, and the DNA polymerase inhibitor ibezapolstat (formerly known as ACX-362E) (20). We observed clear activation of P cd0350 in the presence of all added stressors but not for a negative control containing water (Fig. 5C).
We conclude that, at least for the B -dependent promoter of cd0350, activation does not only occur upon exposure to lethal levels of metronidazole but also occurs with unrelated antimicrobials and toxic stressors such as hydrogen peroxide.

DISCUSSION
In this work, we have demonstrated by Western blotting using an affinity-purified anti-B antibody that B can be overproduced for a limited period of time, sufficient for transcriptome analyses. The induced production of B in a sigB mutant background yielded highly consistent results despite potential toxicity and plasmid loss (Fig. 2), and the results were used to redefine the surprisingly large B regulon previously proposed (19). As our approach more accurately measures changes in transcription directly related to B production, the refined regulon described here is much smaller (see Data Set S1 in the supplemental material). Its size is fully in line with that of the B regulon of other Gram-positive bacteria such as L. monocytogenes (Ϸ130 genes), B. subtilis (Ϸ150 genes), and S. aureus (Ϸ200 genes) (8). The redefined regulon underscores the importance of B in responding to oxidative/nitrosative stresses, as genes implicated in such processes are significantly enriched in the smaller regulon.
The majority of the genes in our regulon were found to be induced, rather than repressed, by B . This is in line with sigma factors acting as specificity determinants for transcription initiation (27). Similar observations have been made for the B regulon of L. monocytogenes (36,37). For the first time, direct evidence of C. difficile B -dependent gene activation is provided by the results of the in vitro runoff transcriptions (Fig. 4), which demonstrate that RNAP core and B are sufficient to generate transcripts from P cd0350 , P cd2963 , P cd3412 , and P cd3605 . Notably, these experiments pave the way for a further in vitro characterization of this sigma factor in C. difficile, including validation of the B binding sequence and the interplay with other regulators.
Although the promoters of cd3412 (uvrB) and cd3614 were reported to have a B consensus sequence and are differentially expressed upon B overexpression (19,20), our results clearly demonstrate that they can also be expressed in a B -independent manner (Fig. 5B). This is most notable for P cd3412 , which is still activated by metronidazole in the absence of B , in line with results obtained with ibezapolstat in a different study (20). Both metronidazole and ibezapolstat treatment can cause DNA damage, and DNA-damage dependent induction of cd3412 therefore likely depends on a sigBindependent pathway.
The observed B -dependent gene repression is expected to be indirect ( B induces the transcription of a repressor gene), or the result of competition ( B competes with other sigma factors for RNAP), as sigma factors by their very nature induce gene expression (27). We consider the second scenario more likely for the following reasons.
First, little overproduction of B was detected after 30 min of induction. This leaves only a limited time for indirect effects to occur in our setup. Second, the majority of genes downregulated upon overexpression of B fall into a single functional group (flagellar motility). These genes are known to be regulated by the dedicated sigma factor, D (38), supporting the model of sigma factor competition. Strikingly, in L. monocytogenes, B activity (indirectly) also results in downregulation of flagellar gene expression, but this is mediated by the repressor MogR (39). Protein BLAST analyses revealed that C. difficile does not possess a MogR homologue. Nevertheless, the conserved inverse correlation between the B -dependent general stress response and bacterial motility could represent a cost-saving strategy for bacterial cells (40). The indirect mechanism underlying the observed B -dependent downregulation in C. difficile remains to be determined.
There appears to be an intriguing link between B and the response to toxic compounds, as a sigB mutant was more susceptible to rifampin and mitomycin C (19), and exposure to antimicrobials (metronidazole, vancomycin, lincomycin, and ibezapolstat) or hydrogen peroxide leads to B -dependent promoter activation (Fig. 5C). The mechanism behind the latter is unclear. It has been suggested that antimicrobials at toxic concentrations can influence metabolism and respiration (41,42), potentially resulting in the formation of bactericidal concentrations of radical species (43)(44)(45). A strong connection between B and oxidative (and/or nitrosative) stress in C. difficile (Table 2) and other bacteria (7,18,19), as well as a recently described radical scavenging strategy that increases tolerance to antimicrobials (46), are consistent with such a model. However, additional research is necessary to determine exactly how these processes occur and are influenced by antimicrobials in anaerobic organisms under anoxic conditions.
In conclusion, we have demonstrated that B is directly involved in metabolic and oxidative stress responses and that lethal stresses may influence these processes, resulting in activation of B -targeted genes.

Construction of B expression and luciferase reporter vectors.
All oligonucleotides used in this study can be found in Table 3. Plasmids and strains are listed in Table 4. All PCR products used for sequencing or plasmid synthesis were generated with Q5 high-fidelity polymerase (NEB). The P T7 -sigB 6ϫHis expression vector pIB14 was created by restriction-ligation using the restriction enzymes NdeI and XhoI. Using primers oIB-1 and oIB-2 on C. difficile 630Δerm chromosomal DNA, the sigB coding sequence (CDS) was amplified by PCR. The resulting DNA fragment was digested and ligated into NdeI-XhoI-digested pET21b(Ϫ) vector, generating expression vector pIB14. Plasmids pIB27, pIB68, pIB69, and pIB74 have been described previously (20). The cd0872 promoter area was amplified using primers oIB-14 and oIB-15, and the P cd0872 luciferase-reporter plasmid was created by restriction-ligation using restriction enzymes KpnI and SacI in digested pAP24 backbone, generating plasmid pIB21. The P cd3605 luciferase reporter plasmid was generated by Gibson assembly as described previously (20) using primers oIB-90 and oIB-99, yielding plasmid pIB73. A plasmid containing P tet -sigB was generated by cloning the 630Δerm sigB CDS amplified with oWKS-1498 and oWKS-1499 in pMiniT (catalog no. E1202; NEB) per the manufacturer's instructions. Using restriction enzymes SacI and BamHI, this PCR fragment was cloned into pRPF185, yielding pWKS1760. All plasmids were verified by Sanger sequencing.
Bacterial strains and growth conditions. Strains of Escherichia coli were grown aerobically at 37°C in Luria-Bertani broth (Affymetrix) supplemented with ampicillin (50 g/ml), kanamycin (50 g/ml), and/or chloramphenicol (20 g/ml) when required. Plasmids were maintained in E. coli strains DH5␣ or MDS42 (Scarab Genomics) under appropriate antimicrobial selection, and cells were transformed using standard procedures (47). For plasmid conjugation into recipient wild-type C. difficile 630⌬erm and the isogenic sigB mutant strains, E. coli strain CA434 was used as a donor strain as previously described (48). C. difficile strains were cultured anaerobically at 37°C in either a Don Whitley VA-1000 or A55 workstation. Cells were cultured in brain heart infusion (BHI; Oxoid) broth supplemented with 0.5% (wt/vol) yeastextract (BHIY) and 20 g/ml thiamphenicol when appropriate. Unless additional antimicrobials/stressors were added (metronidazole Etest and sterile pads supplemented with different stressors), medium was supplemented with C. difficile selective supplement (CDSS; Oxoid).
Overproduction, purification and affinity purification of B 6؋His for synthesis of a polyclonal anti-B antibody. (i) Overproduction and purification  plasmid pIB14. These cells were cultured in Luria-Bertani (LB) broth and induced with 0.5 mM isopropyl-␤-D-thiogalactopyranoside (IPTG) for 1 h starting at an optical density of Ϸ0.6. Cells were collected by centrifugation at 4°C, and the resulting cell pellets were resuspended in lysis buffer (pH ϭ 8.0; 50 mM NaH 2 PO 4 , 300 mM NaCl, 5 mM ␤-mercaptoethanol, 0.1% NP-40, and complete protease inhibitor cocktail [CPIC; Roche Applied Science]). Through the addition of 1 mg/ml lysozyme and sonication (6 ϫ 20 s), cells were lysed. The lysate was drawn through a blunt 1.2-mm needle and was clarified by centrifugation at 13,000 ϫ g at 4°C for 25 min. Recombinant B 6ϫHis was purified from the supernatant on Talon Superflow resin (GE Healthcare) per the manufacturer's instructions. Proteins were dialyzed and stored in buffer (pH ϭ 8.0) containing 50 mM NaH 2 PO 4 , 300 mM NaCl, and 12% glycerol. Protein concentrations were determined using a Bradford assay (Bio-Rad). Two ml of B 6ϫHis protein solution containing 2 mg/ml protein was sent to BioGenes GmbH (Berlin) for generation of a polyclonal rabbit anti-B antibody.
(ii) Affinity purification of the polyclonal anti-B antibody. Affinity purification of the antibody was performed to increase specificity of B detection. Approximately 350 g of purified B 6ϫHis protein was loaded onto an SDS-PAGE gel. After electrophoresis and transfer of proteins to a polyvinylidene difluoride (PVDF) membrane using standard blotting procedures, purified B 6ϫHis protein was visualized by Ponceau S staining, and the membrane containing the protein was cut as small as possible while retaining the region with the protein. The membrane was destained and washed with Tris-buffered saline with Tween 20 (TBST) buffer (500 mM NaCl, 20 mM Tris base, and 0.05% vol/vol Tween 20 [pH ϭ 7.4]) twice for 5 min at room temperature. The membrane was then preeluted by soaking in acidic glycine solution (100 mM, pH ϭ 2.5) for 5 min prior to washing with TBST twice for 5 min at room temperature. Subsequently, the membrane was blocked in 5% nonfat milk powder solution (Campina Elk, dissolved in TBST buffer) for 1 h at room temperature after again washing twice with TBST for 5 min. Serum containing anti-B antibody was incubated on the membrane overnight at 4°C. After three 5-min washes with TBST, the membrane was washed twice for 5 min in PBS. Affinity-purified antibody was eluted from the membrane by adding acidic glycine solution and incubating for 10 min at room temperature. The pH of the eluate was adjusted to 7.0 through the addition of 1 M Tris-HCl (pH ϭ 8.0). This step was repeated twice more, and the eluates were pooled and centrifuged (1 min at maximum speed) to remove precipitated protein and membrane particles. Bovine serum albumin (BSA) and sodium azide were added to the affinity-purified anti-B antibody to end concentrations of 1 mg/ml and 5 mM, respectively, and the affinity-purified antibody was stored at Ϫ80°C.
Characterization of the B regulon. (i) B overproduction in C. difficile. Exponentially growing starter cultures of C. difficile strain IB58 and IB61 were diluted to an OD 600 of 0.05 in BHIY medium (ii) RNA extraction. Bacterial RNA was extracted and analyzed as previously described (50). Briefly, cell pellets were lysed for 30 min at room temperature in enzymatic lysis buffer consisting of 15 mg/ml lysozyme and Tris-EDTA (TE) buffer. Further disruption of cells was performed by vigorous mechanical lysis for 3 min in RLT buffer to which one spatula of glass beads was added. After samples were centrifuged (3 min at 10,000 rpm at 4°C) and 100% ethanol was added to the supernatant, RNA was purified using the Qiagen RNeasy kit protocol according to manufacturer's instructions. DNA contamination was removed by using RNase-free DNase I (Qiagen) twice prior to elution of the RNA samples in H 2 O. RNA quality and integrity numbers (RINs) were assessed with a Bioanalyzer 2100 (Agilent) and RNA 6000 Nano reagents (Agilent). Only samples with an RIN of Ն7 were used for further analysis.
(iii) DNA microarray and data analysis. A customized whole-genome DNA microarray of the 630Δerm strain was used (8 ϫ 15K format; Agilent) (50). Quadruplicate samples were analyzed for the DNA microarray. Using the ULS fluorescent labeling kit for Agilent arrays (Kreatech), 1 g of total RNA was used for labeling with either Cy3 (P tet -sigB) or Cy5 (P tet -sluc opt ). After pooling and fragmentation, 300 ng of labeled RNA per sample was hybridized according to the two-color microarray protocol from Agilent. DNA microarrays were scanned with an Agilent C scanner and analyzed as described previously (50). A gene was considered differentially expressed if the log 2 fold change (log 2 FC) was ՅϪ1.5 or Ն1.5 and the P value was Ͻ0.05. Results were visualized in VolcaNoseR (28) and are available as an interactive graph via the URL contained in Text S1 in the supplemental material.
In vitro transcription. DNA oligonucleotides oWKS-1506 and oWKS-15136 (64 and 82 bp, respectively) were end labeled with γ-32 P -ATP using T4 polynucleotide kinase (PNK; Invitrogen) and used as a size indicator for the in vitro transcription reactions. For the end labeling reaction, 1 l γ-32 P-ATP was incubated together with 200 pmol oligonucleotide and 1 l (10 U) PNK in Forward reaction buffer (70 mM Tris-HCl [pH 7.6], 10 mM MgCl 2 , 100 mM KCl, and 1 mM 2-mercaptoethanol) at 37°C for 30 min. For the in vitro runoff transcriptions, sigma factors and RNA polymerase core enzyme were preincubated with PCR-amplified promoter areas (for P CD0350 , P CD0872 , P CD2963 , P CD3412 , P CD3605 , and P CD3614 ) or XbaIlinearized pCD22 (P tcdA ) for 30 min at 37°C prior to the start of the reaction. PCR products of the promoter areas as used for the in vitro transcription reactions were loaded on and excised from agarose gels and purified using a NucleoSpin gel and PCR clean-up kit (Macherey-Nagel). In vitro transcription reactions mixtures contained 1 l (1 U) E. coli RNAP core (catalog no. M0550S; NEB), 16 pmol sigma factor, 0.5 pmol DNA, 10 mM nucleoside triphosphate (NTP) mix, and 0.3 l ␣-32 P-ATP in reaction buffer (40 mM Tris-HCl, 150 mM KCl, 10 mM MgCl 2 , 1 mM dithiothreitol [DTT], and 0.01% Triton X-100 [pH ϭ 7.5]) and were incubated for 15 min at 37°C. Transcripts and labeled oligonucleotides to be used as a size indication were purified using P-30 Bio-Gel spin columns (Bio-Rad). All reactions were stopped in gel loading buffer II (Invitrogen) containing 95% formamide, 18 mM EDTA, and 0.025% each of SDS, xylene cyanol, and bromophenol blue at 95°C for 5 min and loaded on 8% monomeric UreaGel (SequaGel; National Diagnostics). Gels were dried and exposed to phosphorimager screens overnight (approximately 17 h) and imaged with a Typhoon 9410 scanner (GE Healthcare).
Spot assay for viability upon B overproduction in C. difficile and vector stability assay. C. difficile overnight precultures were corrected for OD 600 and were subsequently 10-fold serially diluted in BHI medium. Spots (2 l) of each dilution were plated on selective (20 g/ml thiamphenicol) and unselective square (90 ϫ 90 ϫ 15 mm; VWR international) BHI plates with or without 200 ng/ml anhydrotetracycline (ATc). Growth was evaluated after 24 h, and swabs were subsequently taken from all strains grown on unselective BHI agar plates with and without 200 ng/ml ATc for the vector stability assay. These swabs used for the vector stability assay were resuspended in PBS to a McFarland turbidity of 1.0, adjusted for their OD 600 values, and 10-fold serially diluted in nonselective BHI medium. Of these serially diluted suspensions, 10-l spots of each dilution were then plated on selective (20 g/ml thiamphenicol plus CDSS) and nonselective (BHI plus CDSS) plates, and CFU/ml was counted after 24 to 48 h of growth. The percentage of cells retaining the plasmid was calculated as (CFU/ml) selective /(CFU/ ml) nonselective ϫ 100%. If no growth was detected on selective plates containing thiamphenicol, the percentage of plasmid maintained was set as 0%. To calculate statistical significance between percent plasmid maintained in strains induced or not induced by ATc, an unpaired Student's t test was used.
Plate-based luciferase assay with metronidazole Etest and disk diffusion. Strains harboring luciferase reporter plasmids were grown on prereduced, selective BHI plates for 24 h. Subsequently, bacterial suspensions corresponding to 1.0 McFarland turbidity were applied on BHI agar supplemented with 0.5% yeast extract, after which a metronidazole Etest or plain disks were applied. Disks were spotted with 10 l each of sterile H 2 O, 1 M H 2 O 2 , 3,000 g/ml lincomycin, 200 g/ml metronidazole, 400 g/ml ibezapolstat, and 200 g/ml vancomycin. After 24 h of growth, luciferase activity was visualized by spraying 1:100 reconstituted NanoGlo luciferase substrate (catalog no. N1110; Promega) on the agar plate using a disposable spray flask. One spray corresponded to approximately 250 l reconstituted NanoGlo luciferase substrate. Luminescence was recorded using a Uvitec Alliance Q9 Advanced imager (BioSPX) after a 10-s exposure time per plate. Luciferase was conjugated into a sigB knockout made by allelic coupled exchange (51), whereas a ClosTron mutant background was used for the DNA arrays. However, no differences between these backgrounds have ever been observed in our assays.