How the Anaerobic Enteropathogen Clostridioides difficile Tolerates Low O2 Tensions

Although the gastrointestinal tract is regarded as mainly anoxic, low O2 tension is present in the gut and tends to increase following antibiotic-induced disruption of the host microbiota. Two decreasing O2 gradients are observed, a longitudinal one from the small to the large intestine and a second one from the intestinal epithelium toward the colon lumen. Thus, O2 concentration fluctuations within the gastrointestinal tract are a challenge for anaerobic bacteria such as C. difficile. This enteropathogen has developed efficient strategies to detoxify O2. In this work, we identified reverse rubrerythrins and flavodiiron proteins as key actors for O2 tolerance in C. difficile. These enzymes are responsible for the reduction of O2 protecting C. difficile vegetative cells from associated damages. Original and complex detoxification pathways involving O2-reductases are crucial in the ability of C. difficile to tolerate O2 and survive to O2 concentrations encountered in the gastrointestinal tract.

and in the mammalian gastrointestinal tract (1). Nowadays, this pathogen is considered the major cause of antibiotic-associated diarrhea and pseudomembranous colitis (2). C. difficile infection (CDI) is transmitted via the oral-fecal route (3) through ingestion of spores, the form for the transmission, resistance, and persistence of this bacterium (4). Colonization of the gastrointestinal tract usually occurs following antibiotic-induced dysbiosis of the host microbiota, which leads to substantial changes in the metabolic pool, particularly in bile acids (5). These metabolic modifications enable germination of the spores in the small intestine and thereafter colonization of the intestinal tract by the vegetative cells (6). Then, C. difficile cells produce virulence factors, including two toxins, TcdA and TcdB, that cause the pathology associated with CDI. These toxins are responsible for alteration of the actin cytoskeleton of intestinal epithelial cells, which induces intestinal cell lysis and triggers an important inflammation (7,8). The inflammation process results in the production by the host immune system of reactive oxygen species (ROS), nitric oxide (NO), and reactive nitrogen species (RNS) at bactericidal concentrations (9,10).
Apart from inflammation-induced oxidative and nitrosative stresses, oxygen (O 2 ) is also a major stress that C. difficile faces during gut colonization. Indeed, although a healthy gut, with a diverse microbiota, is regarded as mainly anoxic, a longitudinal decreasing O 2 gradient is observed along the gastrointestinal tract (11). From 4% in the small intestine lumen, which is the location of spore germination, the O 2 tension decreases to 0.1 to 0.4% in the large intestine lumen, where vegetative cells multiply (11). A second increasing O 2 gradient from the colon lumen toward the intestinal epithelium also occurs (from 0.1 to 0.4% to 5%) (12,13). In addition, antibiotic-induced disruption of the host microbiota leads to an increased O 2 level within the gut (12,14). Thus, O 2 concentration fluctuations within the gastrointestinal tract present a challenge to anaerobic bacteria such as C. difficile. While strictly anaerobic, C. difficile is able to grow in nonstrict anoxic conditions (1 to 3% O 2 ) (15), indicating that this bacterium encodes an arsenal of proteins involved in O 2 detoxification and protection against oxidative stress. Recently, the deletion of the iscS2 gene encoding a cysteine desulfurase likely involved in Fe-S cluster biogenesis was shown to cause a severe growth defect in the presence of 2% O 2 (16). Another essential actor in the C. difficile ability to tolerate low O 2 concentrations is the alternative sigma factor involved in the general stress response, B . Indeed, a sigB mutant is unable to grow in the presence of 0.1% O 2 , a tension lower than that physiologically found in the large intestine (17). Of note, both sigB and iscS2 mutants present a colonization defect in axenic mice, and a delay of colonization is observed for the iscS2 mutant in conventional mice (17). These results suggest that O 2 tolerance might be an important mechanism during the C. difficile colonization process. However, little is known about the proteins involved in the ability of C. difficile to tolerate O 2 .
In other organisms, both flavodiiron proteins (FDPs) and rubrerythrins (Rbrs) play an important role in protecting the cells from O 2 , oxidative, or nitrosative stresses (18)(19)(20)(21). FDPs are enzymes composed of a minimal core containing a diiron catalytic center in a metallo-␤-lactamase-like domain and a flavin-mononucleotide (FMN)-containing flavodoxin domain, located at the N-and C-terminal ends, respectively, of the protein (22)(23)(24). FDPs reduce oxygen to water or NO to nitrous oxide. Although some of these enzymes show substrate specificity toward O 2 or NO, others have a dual activity.
Canonical Rbrs are composed by two types of iron sites: an N-terminal four-helix bundle domain harboring a non-sulfur diiron center and a rubredoxin (Rd)-like [Fe(S-Cys) 4 ] domain in the C-terminal part (25,26). In reverse Rbrs (revRbrs), the position of these two domains are reversed. Rbrs act as peroxidases through NAD(P)H oxidation in partnership with another redox partner, and O 2 -reductase activity has also been reported for a revRbr from Clostridium acetobutylicum (20,25,27,28). Both FDPs and Rbrs are broadly distributed among anaerobic and microaerophilic bacteria and archaea, but they are also present in aerobes and in eukaryotes. Most FDPs, as well as Rbrs and revRbrs, require electron transfer proteins as physiological partners to couple the oxidation of NAD(P)H to the reduction of their substrates, such as rubredoxins (Rds), NAD(P)H/FAD dependent oxidoreductases or the F 420 coenzyme in methanogens (20,(29)(30)(31)(32).
The genome of C. difficile harbors genes encoding two FDPs (CD1157 and CD1623, here named FdpA and FdpF, respectively, according to their domain composition [24]), two Rbrs (CD0825 and CD2845) and two revRbrs (CD1474 and CD1524, here named revRbr1 and revRbr2, respectively). While CD2845 is expressed under the control of G (33) and its product is present in the spore (34), the expression of CD0825, revRbr1, revRbr2, fdpA, and fdpF is positively controlled by B (17). Among the six aforementioned proteins, only FdpF is functionally characterized (35). FdpF is a large class F FDP (24) that harbors, in addition to the minimal core universally present in FDPs, two other domains: a Rd domain of the short type, which distinguishes from canonical Rds by its smaller number of amino acids in between the two pairs of iron-binding cysteines (ϳ12 amino acids versus ϳ30 amino acids) (36) and a NAD(P)H:Rd oxidoreductase-like domain (NROR) (see Fig. S1 in the supplemental material). This protein acts mainly as an NADH:O 2 oxidoreductase reducing O 2 to H 2 O but also exhibits a detectable H 2 O 2 reductase activity (35). Due to the presence in the same polypeptide chain of the extra domains, FdpF receives electrons directly from NADH and transfers them to O 2 without the participation of additional protein partners (35). In the present study, we demonstrate a key role for both FDPs and both revRbrs, whose genes are controlled by B , in the ability of C. difficile to tolerate physiological O 2 tensions encountered within the gut.

RESULTS
Regulation of revRbr1, revRbr2, fdpF, and fdpA expression. Genome-wide transcriptional start site (TSS) mapping allowed us to identify promoters upstream of CD0825, revRbr1 (CD1474), revRbr2 (CD1524), fdpA (CD1157), and fdpF (CD1623) (37). A B consensus sequence [WGWTT-N 13-17 -(G/T)GGTAWA] was identified upstream of the TSSs mapped in the promoter region of revRbr1, revRbr2, fdpF, and fdpA genes ( Fig. 1A to C) (17). In contrast, a canonical Ϫ10 box (TATACT) and a Ϫ35 box (TTGACA) of a A -dependent promoter were identified upstream of the CD0825 mapped TSS (Fig. 1D), suggesting indirect control by B , in agreement with the weaker transcriptional control by B observed for this gene compared to the others (17). Interestingly, upstream of the B -dependent promoter of the fdpA gene, a second TSS was mapped corresponding to a promoter with a canonical extended Ϫ10 box (TGNTATATT) of a A -dependent promoter and a poorly conserved Ϫ35 box, as often observed with a Ϫ10 extended box (Fig. 1C). We next focused on the genes harboring a B consensus sequence in their promoter region, and we wanted to determine whether their expression was strictly dependent on B . Using transcriptional fusions between the promoter regions of the revRbr2 and fdpF genes and the gene encoding the fluorescent reporter SNAP Cd (38), we detected expression of the P revRbr2 -and P fdpF -SNAP Cd fusions in the wild-type (WT) strain in 77 and 61% of the cells, respectively, after 16 h of growth. In contrast, the expression of both fusions was completely abolished in the sigB mutant as observed for the revRbr1 gene ( Fig. 1E and F) (39). To study the more complex expression of the fdpA gene from its two promoters, we first constructed a transcriptional fusion between SNAP Cd and the two promoters. Expression of the P A/B(fdpA) -SNAP Cd fusion was detected in 12% of the cells in the WT strain (Fig. 1G). Contrary to the P revRbr2 -and P fdpF -SNAP Cd fusions, expression of the P A/B(fdpA) -SNAP Cd fusion in the sigB mutant was still detected in a small fraction of the cells in agreement with the existence of the A -dependent promoter (Fig. 1G). When we monitored the expression of a P B(fdpA) -SNAP Cd fusion containing only the B -dependent promoter, we detected fluorescent cells in the WT strain but not in the sigB mutant, confirming that this promoter is recognized by B (Fig. 1H). Thus, the fdpA expression is dependent on both A and B , while the expression of revRbr2 and fdpF is only B dependent in C. difficile, as observed for revRbr1 (39). Fig. S1A in the supplemental material) harboring only the minimal core common to all FDPs (24). The sequence alignment of both FdpA and FdpF cores (diiron and FMN domains) against the structural alignment of all FDPs with available structures (see Fig. S2A) reveals that the sequence identity between the core domains of FdpA and FdpF is only 29%. Despite this relatively low identity percentage, the structural alignment shows that the most conserved regions of characterized FDPs are also present in both FdpA and FdpF. The predicted amino acids involved in the coordination of the diiron center in FdpA are histidines (His 80 , His 85 , His 146 , and His 224 ), aspartates (Asp 84 and Asp 165 ), and one glutamate (Glu 82 ) (see Fig. S2A) (35). The sequence alignment also reveals a significant conservation of other features, such as the putative chain of aromatic amino acids (22) formed by Tyr 191 , Tyr 192 , Tyr 239 , and Trp 242 .

FDPs and revRbrs of C. difficile. FdpA (397 amino acids) is a class A FDP (see
RevRbr1 and revRbr2 share 96% amino acid identity among themselves (Fig. S1B). However, only ca. 25% identity is observed with other rubrerythrins, when a sequence alignment of the four-helix bundle domain of these two revRbrs is performed against a structural alignment of the four-helix bundle domain of three canonical rubrerythrins with known structures (see Fig. S2B). The major predicted difference occurs in the loop connecting the two pairs of helices, which is shorter in revRbrs. Contrary to what was observed for the C. difficile FDPs, there is no significative conservation of large protein regions apart from the diiron center ligands (see Fig. S2B), which are predicted to be glutamates (Glu 67 and Glu 130 ) and histidines (His 70 and His 133 , revRbr1 numbering) (26). The Rd domain contains two CysXXCys motifs (Cys 6 /Cys 9 and Cys 22 /Cys 25 ) separated by 12 residues (see Fig. S1B), which is typical of short-spaced Rd domains found in most Rbrs, as well as in the Rd domains of several classes of FDPs, including FdpF (24,35,36).
Proteins characterization and redox properties. To characterize their properties and functions, the FdpA, revRbr1, and revRbr2 proteins were overexpressed in Escherichia coli and purified through several chromatographic steps in a process analogous to those previously used for similar proteins (32,(40)(41)(42)(43)(44)(45)(46). The molecular masses determined by SDS-PAGE were ϳ45 kDa for FdpA and ϳ22 kDa for the two revRbrs, in agreement with those calculated from their respective amino acid sequences, 44.8 and 20 kDa, respectively (see Fig. S3A). Size exclusion chromatography of the purified proteins revealed that FdpA is a homodimer in solution, with a molecular mass of ϳ82 kDa, while both revRbrs are isolated as tetramers with molecular masses of ϳ90 kDa (see Fig. S3B). The presence of homodimers and homotetramers in solution is typical of other FDPs and Rbrs (23,47,48). For FDPs, the homodimer organization is essential for efficient electron transfer between the electron accepting site in one monomer, the FMN, and the diiron center of the other monomer.
For FdpA, quantification of iron and flavin yielded iron/FMN/protein monomer ratios of 1/0.3/1. These values are lower than expected (i.e., 2/1/1), indicating an incomplete incorporation of both iron and FMN. Even after incubation with FMN following purification, no significant improvement was observed. For revRbr1 and revRbr2, the iron/ protein monomer ratios were 2.1/1 and 2.9/1, respectively, instead of the expected 3/1 ratio. Thus, while revRbr2 is almost fully loaded with iron, revRbr1 is only partially loaded.
By reversed-phase high-pressure liquid chromatography, FMN was identified as the FdpA flavin cofactor, as observed so far for all FDPs (22,24). FdpA exhibited a UV-visible absorbance spectrum typical of a flavin-containing protein, with maxima at 380 and 450 nm (see Fig. S4A). The UV-visible spectra of revRbr1 and revRbr2 were characteristic of proteins containing a Rd center, with maxima at 380, 490, and 580 nm (see Fig. S4B and C). For these three proteins, the spectral contribution of the diiron sites was not detected due to their low molar absorptivities (49). Nevertheless, the diiron site of FdpA was detected by electron paramagnetic resonance (EPR) upon partial reduction with menadiol. As already observed for the E. coli FDP (50), the EPR spectrum revealed a mixture of species, with g values of 1.69, 1.71, 1.98, 1.83, 1.84, and 1.93 (Fig. 2), typical of S ϭ 1/2 anti-ferro-magnetically coupled diiron sites in the mixed valence form [Fe(III)-Fe(II)]. For the revRbrs, we could only observe the typical resonances at g Ϸ 4.3 and a minor resonance at g Ϸ 9.3, attributable, respectively, to the middle (|Ϯ3/2Ͼ) and lower (|Ϯ1/2Ͼ) doublets of a high-spin (S ϭ 5/2) ferric center, with a rhombicity (E/D) close to 0.33, characteristic of the ferric [Fe(SCys) 4 ] sites (51). Overall, the spectroscopic properties of FdpA and revRbrs are identical to those of analogs (42)(43)(44).
The reduction potential of the Rd center for both revRbrs was obtained by redox titrations monitored by visible spectroscopy in an anaerobic chamber at pH 7.5, following the decrease in absorbance of the ferric Rd site (see Fig. S5A and B). We obtained values of ϩ10 mV Ϯ 5 mV and of Ϫ10 mV Ϯ 5 mV for revRbr1 and revRbr2, respectively, which are similar to those of previously isolated Rds (generally in the range of Ϫ50 to ϩ50 mV [reviewed in reference 51]) but significantly lower than those thus far reported for the Rd centers of the few studied canonical rubrerythrins: ϩ213 and ϩ281 mV for the Desulfovibrio vulgaris nigerythrin and rubrerythrin, respectively, and ϩ185 mV for the Campylobacter jejuni desulforubrerythrin (43, 47, 52). Interestingly, for Oxygen Detoxification in Clostridioides difficile the enzyme from C. jejuni, the desulforedoxin-like center has a high positive potential (ϩ240 mV) (47), compared to other known desulforedoxin-like centers (ϳ0 mV). This reinforces the idea that there is a very large variability in the reduction potentials of Rd and Rd-like centers that remain to be assigned to specific features of the polypeptide chains close to the metal center, in spite of its identical structure. The flavin center of FdpA was titrated similarly, and the reduction potentials obtained were Ϫ60 Ϯ 15 and Ϫ175 Ϯ 15 mV for the FMN ox /FMN semiquinone and FMN semiquinone /FMN red redox transitions (see Fig. S5C), values within the range of those available for a few other FDPs (24,53).

revRbrs and FDP H 2 O 2 -and O 2 -reductase activities.
Having established the presence of the FMN and iron cofactors and determined the redox properties of FdpA, revRbr1, and revRbr2, we addressed the catalytic activities of these proteins. The O 2and H 2 O 2 -reductase activities of the three enzymes were determined using NADH as the primary electron donor. As expected, FdpA, revRbr1, and revRbr2 alone did not show any NADH oxidase activity (Fig. 3, before the addition of substrate). Since the C.  difficile physiological electron donors of these proteins are still unknown and its genome does not encode any Rd that could act as a redox partner, as happens in a few organisms, we used a heterologous system to perform the assays. We found that the truncated Rd domain (Rd-D) of E. coli FDP (flavoRd) and its reductase, homologous to NADH:Rd oxidoreductases (NROR), were able to reduce FdpA and the two revRbrs in the presence of NADH (see Fig. S4, insets).
To test the H 2 O 2 -reductase activity, we spectrophotometrically measured the NADH consumption upon addition of H 2 O 2 to a premix containing NADH, NROR, Rd-D, and either FdpA, revRbr1, or revRbr2. A clear H 2 O 2 -reductase activity was observed for revRbr1 and revRbr2 with rates of 2.1 Ϯ 0.1 s Ϫ1 and 7.7 Ϯ 0.7 s Ϫ1 , respectively ( Fig. 3A and B; Table 1). The reduced activity detected for revRbr1 might be related to its lower content of iron compared to revRbr2. In contrast, we detected only a negligible H 2 O 2 -reductase activity with a rate of 0.27 Ϯ 0.03 s Ϫ1 for FdpA ( Fig. 3C and Table 1). It should be mentioned that thus far such activity was reported only for the FdpF from C. difficile (35).
The O 2 -reductase activity was measured by following O 2 consumption using an O 2 Clark-type selective electrode. FdpA, revRbr1, or revRbr2 were added to an air-saturated buffer (ϳ260 M O 2 ) containing both NROR and Rd-D from E. coli and an excess of NADH (5 mM). We observed an O 2 -reductase activity for FdpA, revRbr1, and revRbr2 with rates of 1.4 Ϯ 0.3 s Ϫ1 , 1.2 Ϯ 0.1 s Ϫ1 , and 2.0 Ϯ 0.4 s Ϫ1 , respectively ( Fig. 4 and Table 1). These activities may be underestimated since the physiological partners in C. difficile were not used in the assays as they remain to be identified. However, we can conclude from these results that FdpA acts as an NADH-linked O 2 -reductase and that the two revRbrs act both

Involvement of revRbrs and FDPs in H 2 O 2 stress management.
To study the physiological role of the two revRbrs and the two FDPs in C. difficile, we deleted the genes encoding revRbr1, revRbr2, and FdpF using the allelic chromosomic exchange (ACE). We generated ΔrevRbr1, ΔrevRbr2, and ΔfdpF single mutants and a ΔrevRbr1/2 double mutant, as well as a ΔrevRbr1/2-ΔfdpF triple mutant. We also inactivated fdpA by insertion of an intron into this gene using the clostron system to generate a fdpA single mutant, a fdpA-ΔfdpF double mutant, a ΔrevRbr1/2-fdpA triple mutant, and a ΔrevRbr1/ 2-ΔfdpF-fdpA quadruple mutant. Since we showed that revRbr1 and revRbr2, as well as FdpF (35), have significant in vitro H 2 O 2 -reductase activities (Table 1), we first tested the possible involvement of these proteins in C. difficile H 2 O 2 resistance. Using disk diffusion assays, we did not observe any difference in the growth inhibition area for the ⌬revRbr1/2, the ⌬revRbr1/2-⌬fdpF, and the ΔrevRbr1/2-ΔfdpF-fdpA mutants compared to the WT strain in the presence of H 2 O 2 (see Fig. S6B). These results suggest either that these H 2 O 2 -reductase activities have no physiological role in the detoxification of H 2 O 2 in C. difficile or that the existence of other enzymes scavenging H 2 O 2 prevents to observe a phenotype in our conditions. Involvement of revRbrs and FDPs in C. difficile low O 2 tolerance. Since both revRbrs and both FDPs act as NADH dependent O 2 -reductases (Table 1), we assessed the ability of the different mutants to tolerate low O 2 tensions. For this purpose, we incubated tryptone-yeast extract (TY) plates spotted with serial dilutions of each strain in anaerobiosis or in the presence of different low O 2 tensions (0.1 or 0.4%). No difference in growth between the WT strain and the mutants was observed in anaerobiosis (Fig. 5). Similarly, we did not detect any growth defect for the ΔrevRbr1, ΔrevRbr2, and ΔfdpF single mutants compared to the WT strain when exposed to O 2 tensions up to 0.4% (Fig. 5A and B). In contrast, there was a growth defect for the fdpA mutant under these conditions, whereas no differences were observed in the presence of lower O 2 tensions (i.e., 0.1 and 0.2%) ( Fig. 5B; see also Fig. S7A). We were not able to complement this reduced O 2 tolerance by expressing fdpA on a plasmid either from its own promoter (both A and B dependent promoters in pMTL84121 plasmid) or from an anhydrotetracycline (aTc)-inducible promoter (P tet ) (pDIA6103 plasmid) (see Fig. S7B). Of note, overexpression of fdpA in the presence of 100 ng ml Ϫ1 of aTc resulted in a growth defect even in anaerobiosis, suggesting that the overexpression of fdpA is toxic for C. difficile. On the other hand, the growth defect observed for the fdpA mutant in the presence of 0.4% O 2 was restored when the fdpA disrupted by the clostron insertion was replaced by a wild-type copy of fdpA using the ACE technique (Fig. 5B). In addition, we observed the same growth defect for a ΔfdpA mutant generated by ACE in the presence of 0.4% O 2 (see Fig. S7B). Altogether, these results indicate that FdpA plays a role in low O 2 tolerance in C. difficile. Strikingly, deletion of both revRbrencoding genes strongly affected the growth compared to each single mutant and the WT strain in the presence of 0.1 or 0.4% O 2 (Fig. 5A). The growth of this double mutant was fully restored when we expressed either the revRbr1 or the revRbr2 gene under the control of their own promoter, indicating a functional redundancy between both revRbrs in their ability to reduce O 2 and confirming a crucial role of revRbrs in O 2 tolerance in C. difficile. In contrast to revRbrs, the inactivation of both FDP-encoding genes only resulted in a slight but reproducible decreased O 2 tolerance compared to the fdpA single mutant at 0.4% ( Fig. 5B; see also Fig. S7C). In addition, the introduction of pMTL84121-P B -fdpF did not increase the growth of the fdpA-ΔfdpF mutant, indicating that FdpF cannot replace FdpA to restore O 2 tolerance at 0.4% (Fig. 5B).
To determine whether the O 2 -reductase activities of both FDPs and revRbrs could be additive, we exposed the ΔrevRbr1/2-ΔfdpF and the ΔrevRbr1/2-fdpA triple mutants, as well as the quadruple mutant, to low O 2 tensions. Inactivation of fdpF did not decrease the O 2 tolerance of the ΔrevRbr1/2 double mutant whatever the O 2 tension tested ( Fig. 5C; see also Fig. S8). Interestingly, inactivation of fdpA increased the O 2 tolerance of the ΔrevRbr1/2 double mutant when exposed to 0.1% but not 0.4% O 2 . This effect was FdpF dependent since the quadruple mutant (ΔrevRbr1/2-⌬fdpF-fdpA::erm) was less tolerant than the ΔrevRbr1/2 double mutant and both triple mutants to O 2 from concentrations as low as 0.1% ( Fig. 5C; see also Fig. S8). Remarkably, we noticed that the introduction of pMTL84121-P B -fdpF restored the O 2 tolerance observed in the ΔrevRbr1/2-fdpA::erm triple mutant in the presence of 0.1% O 2 , confirming the role of FdpF in O 2 tolerance at extremely low O 2 tension (Fig. 5C). In addition, at 0.4% O 2 , no colony was observed for the quadruple mutant while both triple mutants were able to grow when they were non-diluted ( Fig. 5C; see also Fig. S8). From these results, we conclude that O 2 -reductase activities of both FDPs and revRbrs are additive, and we propose that FdpF plays a minor but visible role in O 2 tolerance under these conditions. Altogether, these results demonstrate the role of both revRbrs and FDPs in O 2 tolerance mechanisms in C. difficile.

DISCUSSION
In the present study, we identified two revRbrs and two FDPs as the O 2 -reductases in C. difficile. We showed that these enzymes are important for O 2 tolerance of this bacterium and that their activities are additive. Indeed, a quadruple mutant inactivated for all genes encoding these proteins has a strong growth defect at 0.1% O 2 and is unable to grow in the presence of 0.4% O 2 . This demonstrates for the first time the physiological role of these two classes of enzymes in the ability of a strict anaerobe to tolerate low physiological O 2 tensions (Ͻ0.5%). It is worth noting that O 2 tolerance varies among the different C. difficile ribotypes. Indeed, whereas the survival of R20291 (ribotype 027) or 5235 (ribotype 078) vegetative cells sharply decreases after 3 h of air exposure, vegetative cells of 630Δerm strain (ribotype 012) are still detectable after 24 h (54). Moreover, the R20291 strain has a decreased O 2 tolerance compared to the 630Δerm strain in the presence of 0.1 and 0.4% O 2 (17). However, in both 630Δerm and Oxygen Detoxification in Clostridioides difficile ® R20291 backgrounds, this low O 2 tolerance process depends on the alternative sigma factor involved in general stress response, B , which is involved in transcription of fdpA, fdpF, revRbr1, and revRbr2 (17; this study). Accordingly, both revRbr-and FDP-encoding genes harbor a B -consensus sequence in their promoter, and the phenotype of the quadruple mutant at low O 2 tension phenocopies that of the sigB mutant. However, while the transcription of fdpF, revRbr1, or revRbr2 appears to be strictly B dependent (Fig. 1) (39), fdpA is transcribed from two promoters depending on A and B . Of note, the expression of fdpA, fdpF, revRbr1, or revRbr2 is not induced by amoxicillin and clindamycin exposure, two antibiotics known to promote CDI through the disruption of a healthy microbiota that results in an increased O 2 level in the gut (see below) (55). On the other hand, the expression of these genes is upregulated upon brief aeration of C. difficile, whereas no induction is observed after a short exposure to 5% O 2 or when grown in microaerophilic conditions (2% O 2 ) (15,55,56). These discrepancies could be linked to the differences in stress exposure conditions. Further studies are required to elucidate the exact role of air or low O 2 exposure on the expression of genes encoding revRbrs and FDPs and whether this effect is mediated by B . Interestingly, homologs of revRbr1/2 (Rbr3A/B) and of two class A FDPs (FprA1/2) are present in C. acetobutylicum, and the expression of the corresponding genes is induced upon 5% O 2 flushing (21,57,58). In this bacterium lacking B , the expression of rbr3A/B and fprA1/2 is negatively controlled by the PerR repressor (59,60), and a perR mutant has increased O 2 consumption, aerotolerance, and ROS resistance (60). In C. difficile, a PerR-encoding gene cluster is present. This cluster contains genes encoding a canonical rubrerythrin (CD0825) and a desulfoferrodoxin (CD0827) (a superoxide reductase) that are homologs to PerR targets in C. acetobutylicum (59). Using the consensus defined in C. acetobutylicum, we failed to identify an identical PerR box in the promoter region of fdp and revRbr genes but also upstream the CD0825 operon. The role of PerR in the physiology of C. difficile and in the control of fdp and revRbr genes remains to be studied.
We also demonstrated that both purified revRbrs reduce O 2 and H 2 O 2 into H 2 O, although H 2 O 2 is the preferred substrate, as observed for revRbrs in C. acetobutylicum (20). On the other hand, FdpA only acts as NADH-linked O 2 -reductase in vitro. Although some FDPs have been shown to scavenge both NO and O 2 (32,45), FdpA and FdpF lack this dual activity (see Fig. S8) (35). In C. acetobutylicum, O 2 -reductase activity of class A FDPs FprA1/2 are 100 times greater than those of Rbr3A/B (20,21). In our conditions, O 2 -reductase activities of both revRbrs and FdpA are rather similar and 10-fold lower than the one described for FdpF (between 1.2 and 2 s Ϫ1 compared to 16 s Ϫ1 ) (35). However, in FdpF, all the domains involved in the electron transfer from NADH to O 2 are present, while revRbrs and FdpA need intermediate proteins to perform such a transfer. Since the natural electron transfer proteins associated with revRbrs and FdpA remain unknown in C. difficile, our assays were performed with Rd-D and NROR from E. coli that most probably result in a lower O 2 -reductase activity. Therefore, the activities determined in these experimental conditions may be considerably underestimated. C. difficile 630 has no genes encoding Rds, as also observed in many organisms having FDPs and in which the electron donors of these enzymes remain also to be determined (24).
Even if the enzymatic properties of revRbrs and FDPs from several microorganisms have been characterized, less is known about their physiological role, especially in the ability of anaerobes to deal with O 2 (24,61). In C. difficile, the relative importance of both revRbrs and FDPs in O 2 tolerance seems to be variable. Indeed, only the fdpA single mutant showed a reduced O 2 tolerance, while we need to inactivate both revRbrs-encoding genes to observe a growth defect in the presence of low O 2 tensions. This clearly indicates a functional redundancy of revRbrs sharing both a high level of similarity and a common control of their production by B . On the other hand, the double mutant ΔrevRbr1/2 showed a strongly reduced tolerance to 0.1% O 2 , highlighting the crucial role of these proteins in the growth of C. difficile at very low O 2 tensions. Surprisingly, fdpA inactivation in the ΔrevRbr1/2 mutant increased the tolerance at 0.1% O 2 . Although the underlying mechanism responsible for this higher tolerance is not characterized, this effect is dependent on FdpF, since its inactivation completely abolished the growth whatever the O 2 concentration tested (Fig. 5). Further investigations are needed concerning the expression or the activity of FdpF in different genetic backgrounds and oxygenation conditions. While the two revRbrs and FdpF also harbor significant H 2 O 2reductase activities in vitro (35), inactivation of these genes does not result in increased H 2 O 2 sensitivity, strongly suggesting that other enzymes efficiently scavenging H 2 O 2 are present in C. difficile. Three catalases and a peroxidase (CotE) are present in C. difficile but only produced during sporulation (33,62). However, a secreted glutamate dehydrogenase has been shown to participate in resistance to H 2 O 2 (63). In addition, other proteins such as a canonical Rbr (CD0825) and a bacterioferritin comigratory protein (CD1822), that are homologs to proteins with peroxidase activity in other organisms (61) might contribute to protect C. difficile vegetative cells from H 2 O 2 stress. Further work is needed to characterize all the pathways involved in H 2 O 2 protection and to demonstrate the possible physiological role of the revRbrs and FdpF in H 2 O 2 detoxification.
Previous studies have shown that NO-specific FDPs play a role in the protection of bacteria from NO-related killing within macrophages (64,65). However, no studies have linked O 2 -specific FDPs and revRbrs with virulence or colonization. Several recent studies demonstrate that antibiotic administration, a CDI risk factor, increased the O 2 level within the gut by modifying the microbiota. Indeed, antibiotherapy notably depletes clostridia that produce butyrate, a compound directly consumed by enterocytes through O 2 conversion to CO 2 (14,66). Moreover, some vegetative cells have been found associated with the mucus (67-69), a place in which the O 2 level is higher than in the lumen (11,70). Thus, C. difficile vegetative cells cope with O 2 tensions during the colonization process and, in such nonstrict anaerobic environments, the ability to tolerate small amounts of O 2 appears to be essential (55). Accordingly, both sigB and iscS2 mutants, affected in their ability to tolerate low O 2 tensions, have been shown to be impaired in colonization in axenic mice (16,17). Since both revRbrs and FDPs play a crucial role in C. difficile low O 2 tolerance, further studies are needed to show that these proteins are important during the colonization process.

MATERIALS AND METHODS
Bacterial strains and growth conditions. The C. difficile strains and plasmids used in this study are listed in Table S1 in the supplemental material. C. difficile strains were grown anaerobically (5% H 2 , 5% CO 2 , 90% N 2 ) in TY (Bacto tryptone, 30 g liter Ϫ1 , yeast extract, 20 g liter Ϫ1 [pH 7.4]), in Pep-M (71), or in C. difficile minimal medium (CDMM) (72). For solid media, agar was added to a final concentration of 17 g liter Ϫ1 . When necessary, cefoxitin (Cfx; 25 g ml Ϫ1 ), thiamphenicol (Tm; 15 g ml Ϫ1 ), or erythromycin (Erm; 2.5 g ml Ϫ1 ) were added to C. difficile cultures. E. coli strains were grown in LB broth or in M9 minimal medium supplemented with FeSO 4 to a final concentration of 200 M. When indicated, ampicillin (100 g ml Ϫ1 ), chloramphenicol (15 g ml Ϫ1 ), or kanamycin (50 g ml Ϫ1 ) was added to the culture medium.
Construction of C. difficile strains. The clostron gene knockout system (73,74) was used to inactivate the fdpA (CD1157) gene, yielding the insertional mutant strain 630Δerm fdpA::erm (CDIP588). We designed primers to retarget the group II intron of pMTL007-CE5 to insert it into the fdpA gene in sense orientation after nucleotide 225 in the coding sequence (see Table S1). The PCR product generated by overlap extension was cloned between the HindIII and BsrG1 sites of pMTL007-CE5 to obtain pDIA6374. C. difficile transconjugants obtained with E. coli HB101(RP4) containing pDIA6374 were selected on brain heart infusion (BHI) agar containing Tm and Cfx and then plated on BHI agar containing Erm. PCRs on transconjugants chromosomal DNA were performed to verify the integration of the intron into the fdpA gene and the splicing of the group I intron from the group II intron after integration.
The ΔrevRbr1 (ΔCD1474), ΔrevRbr2 (ΔCD1524), ΔfdpA (ΔCD1157), and ΔfdpF (ΔCD1623) knockout mutants were obtained using the codA-mediated allele exchange method (75,76). Then, 1.5-kb fragments located up-and downstream of these genes were PCR amplified from 630Δerm genomic DNA using the pairs IMV917/NK70 and NK71/IMV918, IMV919/NK64 and NK65/IMV920, CF80/CF81 and NK180/CF83, and IMV914/NK58 and NK59/IMV915 (Table S1B) for ΔrevRbr1, ΔrevRbr2, ΔfdpA, and ΔfdpF, respectively. Purified PCR fragments were then introduced into the pMTLSC7315 plasmid using a Gibson Assembly master mix (Biolabs). The sequences of the resulting plasmids were verified by sequencing. The plasmids obtained introduced in HB101(RP4) E. coli strain were transferred by conjugation into the C. difficile 630Δerm strain. Transconjugants were selected on BHI plates supplemented with Tm and C. difficile selective supplement (SR0096; Oxoid). Isolation of faster growing single-crossover integrants was performed by serial restreaking on BHI plates containing Cfx and Tm. Single-crossover integrants were then restreaked on CDMM plates supplemented with fluorocytosine (50 g ml Ϫ1 ), allowing the isolation of double-crossover events. After confirmation of plasmid loss (Tm-sensitive clones), the presence of the expected deletion in clones was checked by PCR. Steps were repeated in each different mutant in order to generate C. difficile multi mutants (see Table S1A).
Oxygen Detoxification in Clostridioides difficile ® September/October 2020 Volume 11 Issue 5 e01559-20 mbio.asm.org 11 Complementation of the different mutants. To complement revRbr1, revRbr2, and fdpF, the promoter region and the open reading frame of the corresponding gene were amplified by PCR using oligonucleotide pairs CF115/CF116, CF113/CF114, and IMV699/IMV700, respectively. The PCR products were cloned into pMTL84121 to produce pDIA6538, pDIA6537, and pDIA6388 (see Table S1A). For fdpA, the region encompassing both A -and B -dependent promoters and the open reading frame was amplified by NK259/NK260 and inserted by Gibson Assembly into pMTL84121 amplified by inverse PCR with NK261/NK262 to yield pDIA6870. To express fdpA under the control of the inducible P tet promoter, the open reading frame was amplified by PCR using oligonucleotides IMV978 and IMV979. The PCR product was cloned into pDIA6103 (37) to produce pDIA6806. E. coli HB101(RP4) strain containing each plasmid was mated with single or double mutants (see Table S1). To complement the fdpA mutant at the same locus, the clostron was removed by the chromosomal introduction of a fdpA wild-type copy using ACE. First, a PCR fragment encompassing fdpA and 1.5-kb fragments located upstream and downstream of fdpA gene was amplified from 630Δerm genomic DNA using primers CF80/CF83. This fragment was then introduced into pMTLSC7315 by Gibson Assembly, yielding pDIA6956. The resulting plasmid was transferred from E. coli HB101(RP4) to the fdpA::erm mutant by mating. Selection of transconjugants, isolation of faster growing single-crossover integrants, and isolation of double-crossover events were performed as described above. After confirmation of plasmid loss, the presence of the fdpA gene in lieu of the intron was checked by PCR.
Oxygen and H 2 O 2 stress tolerance assays. For O 2 tolerance assays, 5 l of different serial dilutions (from 10 0 to 10 Ϫ5 ) of C. difficile strains grown for 7 h in TY were spotted on TY plates containing 0.05% taurocholate. Plates were then incubated in the presence of different O 2 tensions or in anaerobiosis in a microaerophilic workstation from Baker Ruskinn. When needed, aTc was added to the plate to induce fdpA expression. The last dilution allowing growth was recorded after incubation at 37°C for 64 h. Disk diffusion assays were conducted as follows. Cultures of C. difficile strains grown in Pep-M medium were plated on calibrated Pep-M agar. A sterile 6-mm paper disk was placed on the agar surface, and 4 l of 1 M hydrogen peroxide (H 2 O 2 ) was added to the disk. The diameter of the growth inhibition area was measured after 24 h of incubation at 37°C.
Transcriptional SNAP Cd fusions. To construct transcriptional fusions, we used pFT47, a plasmid containing the SNAP Cd gene (38). A fragment containing either the complete promoter region of fdpA (positions Ϫ308 to ϩ114 from the start codon) or only the B -dependent promoter region (Ϫ104 to ϩ114) was amplified using genomic DNA and the primer pairs IMV949/IMV950 and IMV976/IMV950, respectively. The DNA fragments were inserted between the EcoRI and XhoI sites of pFT47 to obtain pDIA6517 and PDIA6669, respectively. To construct the revRbr2-or fdpF-transcriptional SNAP Cd fusion, the promoter region (Ϫ190 to ϩ30 or Ϫ202 to ϩ14 from the TSS) was amplified using genomic DNA and primer pairs CF23/CF24 or CF21/CF22, respectively. The DNA fragments were inserted into pFT47 between the EcoRI and XhoI sites to obtain pDIA6459 and pDIA6458. The plasmids introduced into E. coli HB101(RP4) were then transferred by conjugation into C. difficile 630Δerm strain and the sigB mutant (see Table S1).
Fluorescence microscopy and image analysis. To monitor the expression of the different transcriptional fusions, the strains were grown for 16 h in BHI. SNAP labeling and fluorescence microscopy were performed as previously described (38). The images were taken with exposure times of 200 ms for autofluorescence and 500 ms for SNAP. Cells were observed on a Nikon Eclipse TI-E microscope 60ϫ objective and captured with a CoolSNAP HQ2 Camera. For quantification of the SNAP-TMR Star signal resulting from transcriptional fusions, the pixel intensity was measured and corrected by subtracting the average pixel intensity of the background. Images were analyzed using ImageJ.
Protein production, purification, and quaternary structure determination. The coding region of revRbr1 and revRbr2 were PCR amplified using genomic DNA and primer pairs IMV970/IMV971 or IMV968/IMV969 to produce a 540-bp PCR product that was cloned into pET20 (Novagen), yielding pDIA6635 or pDIA6671. The amino acid sequence of FdpA was used to synthesize its encoding gene (GenScript, Inc., USA) using codon optimized for expression in E. coli, which was cloned into a pET24a plasmid. After verification by sequencing, the plasmids were introduced in E. coli BL21(DE3)Gold. Strains overexpressing revRbr1, revRbr2, or fdpA were grown aerobically at 37°C at 150 rpm in M9 minimal medium, with 20 mM glucose, 0.1 mM FeSO 4 , and ampicillin (for revRbr1 and revRbr2) or kanamycin (for fdpA). When the optical density at 600 nm reached 0.4, 0.1 mM FeSO 4 was added, and gene expression was induced by 1 mM IPTG (isopropyl-␤-D-thiogalactopyranoside) for both revRbrs and 0.1 mM IPTG for fdpA. After 6 h of growth at 30°C, the cells were harvested by centrifugation, resuspended in a buffer containing 50 mM Tris-HCl (pH 7.5), and stored at -20°C. Cells were disrupted by at least three cycles in a French press apparatus at 16,000 lb/in 2 (Thermo) in the presence of DNase (Applichem). The crude extract was cleared by low-speed centrifugation at 25,000 ϫ g for 25 min and then at 138,000 ϫ g for 90 min at 4°C to remove cell debris and the membrane fraction, respectively. The soluble extract was dialyzed overnight at 4°C against buffer A (20 mM Tris-HCl [pH 7.5], 18% glycerol) and subsequently loaded onto a Q-Sepharose Fast Flow column (65 ml; GE Healthcare) previously equilibrated with buffer A. Proteins were eluted with a linear gradient from buffer A to buffer B (buffer A containing 500 mM NaCl). The eluted fractions were monitored throughout the purification process by SDS-PAGE and UV-visible spectroscopy. Fractions containing the desired protein were pooled and concentrated. For both revRbrs, the concentrated fraction was then loaded onto a size exclusion Superdex S75 column (330 ml; GE Healthcare) equilibrated with buffer A containing 150 mM NaCl. The fractions containing the protein were pooled and concentrated. For revRbr1, the last fraction was loaded onto a Fractogel column (20 ml; Merck Millipore) previously equilibrated with buffer A and eluted with a linear gradient from buffer A to buffer B. After purification, FdpA was incubated overnight at 4°C in the presence of 1 mM FMN and subsequently applied to a PD-10 desalting column to eliminate unbounded FMN. Fractions containing purified proteins were verified by SDS-PAGE (see Fig. S3A).
Protein, metal, and flavin quantification. Purified protein samples were quantified using a bicinchoninic acid kit (Thermo) and bovine serum albumin as the standard. The iron content was determined by the phenanthroline colorimetric method. Protein samples were incubated for 15 min with 1 M HCl and for further 30 min with 10% trichloroacetic acid at room temperature, centrifuged at 5,000 ϫ g for 20 min, and neutralized by the addition of 15% ammonium acetate. Samples were then incubated with 10% hydroxylamine and 0.3% 1-10-phenantrhroline. The absorbance spectrum was measured, and the iron content was quantified by using 510 ϭ 11.2 mM Ϫ1 cm Ϫ1 . The flavin type and content were determined as previously described (35).
Spectroscopic methods. UV-visible spectra were obtained in a Perkin-Elmer Lambda 35 spectrophotometer. Electron paramagnetic resonance (EPR) spectroscopy was performed using a Bruker EMX spectrometer equipped with an Oxford Instruments ESR-900 continuous-flow helium cryostat and a high-sensitivity perpendicular mode rectangular cavity. Protein samples were prepared aerobically to final concentrations of 1 mM (FdpA and revRbr2) or 600 M (revRbr1). Partially reduced samples were also prepared anaerobically by incubation with 1 equivalent of menadiol. . The reduction potentials are reported in relation to the standard hydrogen electrode. The reduction potentials determined for revRbrs and FdpA were calculated by adjusting a Nernst equation to the experimental data obtained by UV-visible spectroscopy. In all cases, the reduction potential of the diiron sites could not be determined since they have very low molar absorptivities. For the revRbrs, the absorption values obtained at 490 nm (typical of the oxidized Rd-like site) were normalized in relation to the full oxidized protein, and a theoretical curve of the oxidized population was fitted using a SciLab routine and applying a Nernst equation for a one-electron transfer process. The populations were calculated as follows: where Q is calculated as follows: Therefore, where R ϭ 8.314 J K Ϫ1 mol Ϫ1 , T ϭ 298.15 K, n ϭ 1 electron, and F ϭ 96485 C mol Ϫ1 . The theoretical model and curve adjustment were performed using Scilab 6.0.2. The E 0 for the rubredoxin center of the two revRbrs was determined from the fit. For FdpA, the FMN moiety is the only contributor for the visible spectrum and undergoes two consecutive redox reactions. The reduction potentials were determined as for the revRbrs, using the appropriate Nernst equation for two sequential monoelectronic processes and taking into consideration the molar absorptions of the oxidized and semireduced forms of the flavin. and where R ϭ 8.314 JK Ϫ1 mol Ϫ1 , T ϭ 298.15 K, n ϭ 1 electron, and F ϭ 96485 C mol Ϫ1 . As before, the theoretical model and curve adjustment were performed using Scilab 6.0.2. The E 01 and E 02 were determined from the fit.
Spectrophotometric measurement of the H 2 O 2 -reductase activity. Because the physiological electron donor to the C. difficile revRbrs and FdpA is unknown, we used an artificial electron-donating system: a mixture of the Rd domain (Rd-D) of the E. coli FDP (flavoRd) and of the flavoRd reductase (NROR, gene ID 947088, E. coli strain K-12). The E. coli proteins were overexpressed and purified as previously described (50). The enzymatic activity for H 2 O 2 was determined by UV-visible spectroscopy, inside an anaerobic chamber (Coy Lab Products). The assays were performed in buffer C. The reaction was monitored at 340 nm, determining the NADH consumption ( 340 ϭ 6,220 mM Ϫ1 cm Ϫ1 ). A mixture of buffer containing 200 M NADH, NROR (2.5 or 5 M for the revRbrs or FdpA assays, respectively), and Rd-D (2.5 or 5 M for revRbrs or FdpA assays, respectively) was used. Different amounts of NROR and Rd-D were tested and optimized in combination with each enzyme in order to ensure that the reaction rates were maximized. The reaction was initiated by the addition of H 2 O 2 . Concentrations of revRbr1, revRbr2, or FdpA were 0.1, 0.03, and 1 M, respectively. Different amounts of H 2 O 2 were used for each of the three enzymes to evaluate the dependence of the rates with the amount of substrate. No significant differences were verified within the range of concentrations used (10 to 150 M). This indicates that, if these enzymes follow a Michalis-Menten behavior, the K m value is low and therefore not covered by the range used. This also indicates that all the assays are in Vmax conditions. The calculated rates (s Ϫ1 ) presented in Table 1 were calculated subtracting the experimental slope (M s Ϫ1 ) before and after the addition of H 2 O 2 , divided by the protein concentration (M).
Measurement of O 2 -reductase activity. The O 2 -reductase activity of the proteins was measured amperometrically with a Clark-type electrode selective for O 2 (Oxygraph-2K; Oroboros Instruments, Innsbruck, Austria). The assays were performed in buffer C. The O 2 -reductase activity was evaluated at 25°C in air equilibrated buffer (ϳ260 M O 2 ) in the presence of 5 mM NADH, NROR (2.5 M), and Rd-D (2.5 M). Different amounts of NROR and Rd-D were tested and optimized in combination with each enzyme in order to ensure that the reaction rates were maximized. The reaction was initiated by addition of revRbr2 (0.1 M), revRbr1 (0.1 M), or FdpA (1 M). Assays were performed in the presence of catalase (640 nM). The calculated rates (s -1 ) presented in Table 1 were calculated subtracting the experimental slope (M s Ϫ1 ) before and after the addition of each enzyme, divided by the protein concentration (M).
Measurement of NO-reductase activity. The NO-reductase activity of FdpA was measured amperometrically with a Clark-type electrode selective for NO (ISO-NOP; World Precision Instruments, Sarasota, FL). The assays were performed in buffer C. The NO-reductase activity was evaluated at 25°C in degassed buffer in the presence of an O 2 -scavenging system (10 mM glucose, 375 nM glucose oxidase, and 750 nM catalase), NROR (2.5 M), and Rd-D (2.5 M). Sequential additions of NO (up to 12 M) were followed by the addition of 5 mM NADH, and the reaction was initiated by the addition of FdpA (0.1 M). Stock solutions of 1.91 mM NO were prepared by saturating a degassed buffer C in a rubber seal-capped flask with pure NO gas (Air Liquide) at 1 atm on ice: gaseous NO was flushed through a 5 mM NaOH trap to remove higher N-oxides, and a second trap with deionized water to remove aerosols. After this, the solution was allowed to equilibrate at room temperature.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.