d-Proline Reductase Underlies Proline-Dependent Growth of Clostridioides difficile

ABSTRACT

Clostridioides difficile is a nosocomial pathogen that colonizes the gut and causes diarrhea, colitis, and severe inflammation. Recently, C. difficile has been shown to use toxin-mediated inflammation to promote host collagen degradation, which releases several amino acids into the environment. Amino acids act as electron donors and acceptors in Stickland metabolism, an anaerobic process involving redox reactions between pairs of amino acids. Proline, glycine, and hydroxyproline are the three main constituents of collagen and are assumed to act as electron acceptors, but their exact effects on the growth and physiology of C. difficile are still unclear. Using three standard culture media (supplemented brain heart infusion [BHIS], tryptone-yeast [TY], and C. difficile minimal medium [CDMM]) supplemented with proline, glycine, or hydroxyproline, we grew C. difficile strains R20291, JIR8094, and a panel of mutants unable to express the Stickland selenoenzymes d-proline reductase and glycine reductase. In the wild-type strains, growth yields in rich media (BHIS and TY) were higher with proline and hydroxyproline but not glycine; moreover, proline-stimulated growth yields required the activity of d-proline reductase, whereas hydroxyproline-stimulated growth yields were independent of its activity. While assumed to be a proline auxotroph, C. difficile could surprisingly grow in a defined medium (CDMM) without proline but only if d-proline reductase was absent. We believe the mere presence of this enzyme ultimately determines the organism’s strict dependence on proline and likely defines the bioenergetic priorities for thriving in the host. Finally, we demonstrated that addition of proline and hydroxyproline to the culture medium could reduce toxin production but not in cells lacking selenoproteins.

IMPORTANCE Stickland metabolism is a core facet of C. difficile physiology that likely plays a major role in host colonization. Here, we carefully delineate the effects of each amino acid on the growth of C. difficile with respect to the selenoenzymes d-proline reductase and glycine reductase. Moreover, we report that d-proline reductase forces C. difficile to strictly depend on proline for growth. Finally, we provide evidence that proline and hydroxyproline suppress toxin production and that selenoproteins are involved in this mechanism. Our findings highlight the significance of selenium-dependent Stickland reactions and may provide insight on what occurs during host infection, especially as it relates to the decision to colonize based on proline as a nutrient.

INTRODUCTION

Clostridioides difficile (formerly Clostridium difficile) is a Gram-positive, rod-shaped, spore-forming anaerobe and the leading cause of antibiotic-associated diarrhea, often representing 15–25% of all cases (1, 2). Depending on the severity, symptomatic C. difficile infections (CDI) manifest in various disease states ranging from mild diarrhea to life-threatening fulminant colitis (3). C. difficile primarily thrives in the dysbiotic gut environment, which is typically induced by treatment with broad-spectrum antibiotics (3). Following the disruption of the intestinal flora, ingested C. difficile spores travel through the small intestine and germinate into vegetative cells after encountering certain primary bile acids and amino acids (4, 5). These vegetative cells proliferate and eventually release two glucosyltransferases, toxin A (TcdA) and toxin B (TcdB), which are the causative agents of disease; both toxins glucosylate the host’s small Rho-family GTPases in intestinal epithelial cells, ultimately resulting in massive inflammation and colonic injury (6). Though, despite the necessary role that toxins play in establishing CDI, the rationale behind creating incredibly intense gut inflammation and the metabolic strategies that C. difficile employs to thrive in this environment are both poorly understood.

Inflammation is a key element of the host’s immune system, especially in the context of bacterial infections; the complex events that underlie this response have been thoroughly defined (7, 8). However, despite the fact that the immune response is expected to eliminate invading microbes, an unintended consequence of inflammation is the generation of various host-derived metabolites that instead drive the selective expansion of many enteric pathogens in the gut lumen (9–14). Likewise, this concept has recently been demonstrated in the context of CDI, as toxigenic C. difficile was shown to utilize sorbitol generated by aldose reductase, a host enzyme in immune cells that was upregulated during toxin-dependent inflammation (15). Another study using in vivo transcriptomics reported that toxin-mediated inflammation remodels the nutritional environment of the gut in a way that favors the metabolic preferences of C. difficile, particularly amino acid catabolism (16). Specifically, Fletcher et al. (16) suggest a mechanism where toxin-dependent inflammation releases nutrients for C. difficile in the form of amino acids via degradation of host collagen. Indeed, in infected mice, genes encoding collagen-degrading enzymes (e.g., matrix metalloproteinases) are upregulated while certain members of the microbiota that potentially compete for collagen-derived proline and hydroxyproline (e.g., Bacteroidaceae) are suppressed (16). Additionally, the nutrients liberated by this mechanism could help satisfy the strict amino acid requirements for C. difficile growth (17, 18). In support of this idea, C. difficile uniquely excels at performing the Stickland reaction, a biochemical pathway that directly harvests energy from amino acids (19–21). Therefore, if Stickland metabolism is a relevant strategy in vivo, collagen may likely act as an energy depot for invading C. difficile, thereby providing a reasonable argument for why this pathogen easily blooms in the inflamed colon.

In many proteolytic clostridia, Stickland metabolism is a core bioenergetic scheme defined by the coupled oxidation of one amino acid (Stickland donor) and the reduction of another (Stickland acceptor) (22, 23). Briefly, Stickland donors are either oxidatively deaminated or decarboxylated to yield reducing equivalents (NADH) and ATP through substrate-level phosphorylation; Stickland acceptors are reduced or reductively deaminated in an NADH-dependent manner, ultimately regenerating NAD⁺ for further oxidations (20, 22, 23). While Stickland donors are typically aliphatic amino acids (e.g., alanine, leucine, isoleucine, and valine) (22–24), a variety of other nutrients and metabolic pathways may also functionally act as Stickland donors, since reducing power (e.g., NADH) can be derived from sugar oxidation or possibly even gaseous H₂ through hydrogenases (25, 26). In contrast, Stickland acceptors are comparatively less diverse, including only proline, glycine, and leucine; however, others have shown that phenylalanine, tyrosine, and methionine can be reduced at a lower efficiency (22, 24, 27–33). Therefore, Stickland metabolism is theoretically limited by its narrow selection of electron acceptors, which potentially forces C. difficile to occupy a niche where these amino acids are prioritized for rapid disposal of reducing equivalents. This niche, however, may be attainable through the degradation of host collagen, given that most of the amino acids found in high abundance are proline, glycine, and hydroxyproline (34). Although not a true Stickland acceptor, hydroxyproline can be converted to proline in C. difficile by 4-hydroxyproline dehydratase (HypD) and Δ¹-pyrroline-5-carboxylate reductase (P5CR) (35, 36). Thus, the metabolic rationale behind toxin-mediated inflammation might be to scavenge proline, glycine, and hydroxyproline from collagen, which C. difficile could use to overcome the bioenergetic hurdles imposed by Stickland metabolism.

When acting as electron acceptors, proline and glycine undergo reduction by the C. difficile enzymes d-proline reductase (Prd) and glycine reductase (Grd), respectively (19, 20). The true substrate of Prd is generated from the conversion of l-proline to d-proline via proline racemase in C. difficile (19). Subsequent reduction of d-proline by Prd results in cleavage of the ring, ultimately regenerating NAD⁺ and forming 5-aminovalerate as a by-product; glycine is reductively deaminated by Grd to produce NAD⁺ and acetyl phosphate, which yields acetate and ATP through substrate-level phosphorylation (19, 20). In C. difficile, the Prd and Grd complexes share a unique molecular trait in that some of their respective subunits (PrdB, GrdA, and GrdB) contain selenocysteine residues, effectively classifying them as selenoenzymes (19). To synthesize selenoproteins, UGA-encoded selenocysteine must be co-translationally inserted into polypeptides; however, the synthesis of this particular amino acid requires selenophosphate, a selenium donor with high group transfer potential (37, 38). In bacterial selenium metabolism, selenophosphate is generated from selenide in an ATP-dependent manner via the activity of selenophosphate synthetase (SelD) (39). Because of the importance of proline and glycine in C. difficile bioenergetics, the presence of selenium in both Prd and Grd implies a critical role in physiology. We have previously demonstrated that selenium is required for optimal growth on Stickland pairs containing either proline or glycine (19). However, despite the apparent role that SelD plays during important physiological events such as vegetative cell growth, sporulation, and outgrowth of germinated spores (19, 20, 40, 41), selenoproteins have been shown to be nonessential in C. difficile (20, 40), calling into question the actual significance of selenium-dependent Stickland reactions. Overall, our current understanding of the role of selenoproteins in C. difficile physiology is unclear, and the effects of proline, glycine, and hydroxyproline have yet to be fully elucidated in the context of cells lacking Prd and Grd.

In this study, we measured the protein levels of Prd and Grd in C. difficile strains R20291 and JIR8094 grown in several routine culture media to better understand the prevalence of selenium-dependent Stickland reactions. Using these same media supplemented with proline, glycine, and hydroxyproline, we then evaluated the effects of these amino acids on the growth of both wild-type strains and mutants deficient in Prd, Grd, and selenoproteins. Finally, we determined the effects of these amino acids on the ability of R20291 to produce toxins. In this work, we report various growth phenotypes of C. difficile in rich and defined media supplemented with proline, glycine, and hydroxyproline; a SelD-dependent regulatory effect on toxin production by proline and hydroxyproline; and an unexpected discovery that proline-dependent growth of C. difficile is largely due to a unique dependence on Prd.

RESULTS

d-Proline reductase is highly expressed throughout the majority of batch culture.

Proline, glycine, and hydroxyproline are among the most abundant amino acids in collagen; therefore, if C. difficile were to promote collagen degradation, these nutrients would be released into the colonic environment and potentially serve as a rich source of Stickland acceptors for the pathogen. In this scenario, Prd and Grd would likely play a key role in scavenging these nutrients for electrons since the former is induced by proline and hydroxyproline while the latter is induced by glycine (19). However, the abundance of these selenoenzymes has not been critically studied in the context of a classical in vitro growth study, which can reveal significant and relevant information about the organism in a carefully controlled manner. Here, we measured the protein levels of Prd and Grd in C. difficile by labeling R20291 and JIR8094 cells with ⁷⁵Se and growing them in three different routine culture media—BHIS (supplemented brain heart infusion) (42), TY (tryptone-yeast) (43), and CDMM (C. difficile minimal medium) (18)—at 37°C for 72 h (Fig. 1).

FIG 1.

d-Proline reductase persists throughout the entirety of in vitro culture. In the presence of 2 μCi ⁷⁵Se per mL of culture and 50 nM sodium selenite, C. difficile strains R20291 (A, C, E) and JIR8094 (B, D, F) were grown in BHIS (A, B), TY (C, D), and CDMM (E, F) at 37°C for 72 h and harvested at the indicated time points. After harvest and lysis of cells, approximately 7.5 μg (BHIS), 2.5 μg (TY), and 1.0 μg (CDMM) of soluble cellular protein were resolved on 15% acrylamide gels by SDS-PAGE. The known selenoproteins GrdB, PrdB, and GrdA are indicated with arrows based on previous literature (19, 20, 40). The asterisk (*) denotes a low-molecular-weight selenium species of unknown identity.

The selenoprotein subunits of Prd (PrdB) and Grd (GrdA and GrdB) were effectively revealed with this radiolabeling method and were used as proxies for the overall abundance of both complexes. We first discovered in both strains that PrdB was present at 6 h and persisted throughout the 72 h in rich media (BHIS, TY) (Fig. 1A–C, and 1D). We found this surprising considering these growth media were not augmented with excess proline to induce expression of the prd operon (19, 20). There were similar observations in minimal medium (CDMM), although the levels of PrdB differed between strains (Fig. 1E and F). The constant presence of PrdB in three different growth media suggests a strong dependence on the Prd complex regardless of the nutritional environment. Next, we observed varying levels of GrdA and GrdB. GrdA appeared during the early stages of growth in BHIS and TY and slowly decreased in intensity over the course of the entire study (Fig. 1A-D). In CDMM, GrdA presented with less intensity in R20291 compared to JIR8094 where it reached a maximum at 24 h (Fig. 1E and F). On the other hand, the ⁷⁵Se profile of GrdB exhibited the most variability and did not correlate with GrdA in BHIS (Fig. 1A and B), but both selenoproteins were found in similar levels in TY (Fig. 1C and D). Compared to GrdA in CDMM, GrdB was almost nonexistent (Fig. 1E and F). We also curiously spotted an intense low-molecular-weight species (*) of unknown identity that migrated far below 10 kDa (Fig. 1E and F). The differential expression of GrdA and GrdB was surprising and prompted us to evaluate whether depleting glycine from the medium would yield any differences in the ⁷⁵Se profile of each subunit. When we omitted glycine from CDMM, both Grd subunits disappeared from the gel, leaving only PrdB and the low-molecular-weight selenium species (*) (Fig. S1).

Next, in order to carefully determine the role of Prd and Grd in C. difficile physiology, we employed a panel of mutants that harbored deficiencies in either enzyme or the ability to synthesize selenoproteins (Table 1). We first verified the expected phenotypes of these mutants by analyzing their selenoprotein profiles with ⁷⁵Se labeling. The selenoprotein profiles of the JIR8094 mutant strains used in this study have already been verified in previous reports (20, 40). In similar fashion, we labeled the R20291 mutant strains KNM6 (ΔselD) and KNM9 (ΔselD::selD⁺) with ⁷⁵Se and simultaneously confirmed the absence of radiolabeled proteins in the ΔselD mutant and the restoration of PrdB, GrdA, and GrdB in the ΔselD::selD⁺ mutant (Fig. S2). This information allowed us to interpret the resulting phenotypes from future experiments with high confidence. Overall, these findings painted a clearer picture of cellular selenoprotein levels in routine culture media which we could then reference when conducting controlled growth studies with proline, glycine, and hydroxyproline.

TABLE 1.

Bacterial strains used in this study

Bacterial strains	Description (relevant genotype)	Reference/source
R20291	Wild type, ribotype 027	(40)
KNM6	R20291 (ΔselD) CRISPR-Cas9 mutant	(40)
KNM9	KNM6 (ΔselD::selD+) CRISPR-Cas9 mutant	(41)
JIR8094	Wild type, ribotype 012, ErmS derivative of strain 630	(20)
LB-CD4	JIR8094 (prdB::ermB) TargeTron mutant	(20)
LB-CD7	JIR8094 (selD::ermB) TargeTron mutant	(40)
LB-CD8	JIR8094 (prdR::ermB) TargeTron mutant	(20)
LB-CD12	JIR8094 (grdA::ermB) TargeTron mutant	(20)

Proline and hydroxyproline enhance growth yields in rich media, but the latter does not require proline reductase.

To determine the physiological effects of the three Stickland acceptors, we grew a panel of C. difficile strains (Table 1) in rich media (BHIS and TY) augmented with either 30 mM l-proline, glycine, or L-4-hydroxyproline at 37°C and monitored the growth of each culture by measuring the optical density at 600 nm (OD₆₀₀) over a 48-h period. We then performed analyses of the resulting growth plots to identify significant changes in growth rates and growth yields, as these parameters typically define the organism’s ability to compete within the host gut.

In both rich media, the parent strains R20291 and JIR8094 grew to substantially higher yields with added proline and hydroxyproline (Fig. 2A, 2D, 3A, and 3D). Interestingly, compared to base media without supplemented amino acids, the growth yield of the ΔselD mutant did not change with added proline (Fig. 2B and 3B). In the ΔselD::selD⁺ strain, the proline-dependent growth yield was similar to that of R20291 (Fig. 2C and 3C), demonstrating that this effect was due to the presence of SelD. In JIR8094, we observed a similar SelD-dependent phenotype since the yield of LB-CD7 (selD::ermB) was likewise unaffected by the addition of proline (Fig. 2D, 2E, 3D, and 3E), confirming that this phenomenon was not strain dependent. Based on these results, we hypothesized that these growth yield enhancements were due to proline reduction by Prd.

FIG 2.

Proline and hydroxyproline increase growth yield in BHIS, but the latter does not rely on d-proline reductase. The following C. difficile strains were grown in BHIS at 37°C for 48 h: (A) R20291, (B) KNM6, (C) KNM9, (D) JIR8094, (E) LB-CD7, (F) LB-CD4, (G) LB-CD8, and (H) LB-CD12. When indicated, BHIS was supplemented with proline (Pro), glycine (Gly), and hydroxyproline (Hyp) at 30 mM. The experiment was repeated twice. Data points represent the means of triplicate cultures while error bars represent standard deviations. The first 24 h are shown to better visualize relevant phenotypes.

FIG 3.

Growth yield stimulation from proline and hydroxyproline is enhanced in TY. The following C. difficile strains were grown in TY at 37°C for 48 h: (A) R20291, (B) KNM6, (C) KNM9, (D) JIR8094, (E) LB-CD7, (F) LB-CD4, (G) LB-CD8, and (H) LB-CD12. When indicated, TY was supplemented with proline (Pro), glycine (Gly), and hydroxyproline (Hyp) at 30 mM. The experiment was repeated twice. Data points represent the means of triplicate cultures while error bars represent standard deviations. The first 24 h are shown to better visualize relevant phenotypes.

We have previously demonstrated that the addition of proline to a growth medium increases the levels of Prd (19), and it is now known that this phenomenon requires the transcriptional regulator PrdR (20). We found that Prd was indeed responsible for proline-dependent growth stimulation, as either the disruption of the enzyme complex in LB-CD4 (prdB::ermB) (Fig. 2F and 3F) or PrdR in LB-CD8 (prdR::ermB) (Fig. 2G and 3G) was sufficient to remove the organism’s ability to respond to added proline. On the other hand, glycine apparently had no beneficial effect on growth yield. We found both observations to be consistent with another report (20). However, we unexpectedly noticed that glycine occasionally lowered the growth yields of C. difficile. Since Grd levels are modulated by the presence of glycine (19, 20), we assumed that Grd activity might have been somehow responsible for the reduction in growth yield; however, this was not the case as the apparent biomass of LB-CD12 (grdA::ermB) still dropped in the presence of added glycine (Fig. 2H and 3H). Finally, we found that hydroxyproline consistently gave better growth yields than the other two Stickland acceptors. In surprising contrast to proline, we found that these growth enhancements were not due to either SelD or Prd (Fig. 2 and 3), suggesting that the role of hydroxyproline in C. difficile physiology is not simply limited to its ability to act as an additional source of proline for Prd.

Overall, our analyses indicated that the maximum growth yield of each 48-h culture (interpreted as the highest OD₆₀₀ of each curve) was significantly enhanced with proline and hydroxyproline supplementation in a Prd-dependent and Prd-independent manner, respectively (Fig. S3). In comparison, the doubling times of each strain (calculated as the inverse of growth rate) did not substantially change when proline, glycine, and hydroxyproline were added in excess (Table S1 and S2). Finally, we noted that the growth effects due to proline and hydroxyproline were greatly exacerbated in TY compared to BHIS, suggesting that nutrient status of the culture medium influences the organism’s ability to use both amino acids as growth substrates. This line of reasoning is supported by the fact that C. difficile produces substantially less 5-aminovalerate in Eggerth-Gagnon medium compared to BHIS, demonstrating that the rate of proline reduction changes in response to nutrient composition (44).

In the absence of proline reductase, proline is no longer required for growth in a defined minimal medium.

While our observations in rich media are potentially interesting, both of these peptide-based media unfortunately contain substantial amounts of the amino acids we were studying. Instead, in order to clearly define the effects of each Stickland acceptor, we found it advantageous to use a defined minimal medium (i.e., CDMM) as it allowed us to control the amount of each variable accurately and reproducibly. When we first grew R20291 and JIR8094 in CDMM at 37°C, we found that they exhibited very distinct growth profiles compared to rich media (Fig. 4A and 4D). Specifically, after both strains reached maximum culture density, the growth behavior was characterized by a sharp decrease in turbidity, a phenomenon that has been reported in other physiological studies of C. difficile (33, 45).

FIG 4.

Proline-dependent growth of C. difficile requires the presence of d-proline reductase. The following C. difficile strains were grown in CDMM at 37°C for 48 h: (A) R20291, (B) KNM6, (C) KNM9, (D) JIR8094, (E) LB-CD7, (F) LB-CD4, (G) LB-CD8, and (H) LB-CD12. Proline (Pro), glycine (Gly), and hydroxyproline (Hyp) were omitted (^-) from or added (⁺) to CDMM as indicated. Refer to Materials and Methods for amino acid concentrations. The experiment was repeated twice. Data points represent the means of triplicate cultures while error bars represent standard deviations.

To cleanly determine the effects of the three Stickland acceptors, we then grew R20291 and JIR8094 in several versions of CDMM lacking either proline, glycine, or both amino acids. For studies with hydroxyproline, we used CDMM deficient in both amino acids and substituted proline with equimolar hydroxyproline. When proline was removed from CDMM, R20291 and JIR8094 behaved as proline auxotrophs (Fig. 4A and D), which we expected since it has been previously established that proline is required for growth in minimal media (17, 18). However, based on these results, we believed that the experiments to determine the effects of proline were therefore confounded by the fact that it apparently must be present in every preparation of the medium to guarantee growth of C. difficile. Despite this assumption, we surprisingly found that the ΔselD mutant grew very well in the absence of proline (Fig. 4B), whereas the proline-dependent phenotype from R20291 was restored in the ΔselD::selD⁺ mutant (Fig. 4C). Moreover, this behavior was not specific to cells with the R20291 genetic background since the selD::ermB strain (with the JIR8094 genetic background) also grew in the absence of proline (Fig. 4E). Since C. difficile cells lacking SelD gained the ability to grow without added proline, we hypothesized that selenoproteins were somehow playing a major role in establishing this absolute growth requirement. Considering that the effect was related to proline, we suspected that Prd was the selenoenzyme responsible for this phenotype. Indeed, additional growth experiments revealed that the prdB::ermB and prdR::ermB strains also grew in the absence of proline (Fig. 4F and G). On the other hand, the grdA::ermB strain failed to grow unless proline was present (Fig. 4H), verifying that Prd (not Grd) is the selenoenzyme responsible for establishing the growth requirement. Overall, these data suggest that Prd activity forces the organism to strictly depend on environmental proline, which likely explains why wild-type strains behave as proline auxotrophs.

In contrast, we found that the removal of glycine had little to no effect on growth, possibly substituted by threonine present in the medium (18). The grdA::ermB strain, however, gave a heavily distorted growth profile compared to JIR8094 (Fig. 4H), implying that Grd plays an important role in physiology that is not inherently obvious from studies using rich media. Finally, all C. difficile strains were still able to grow even when proline was replaced with hydroxyproline, likely explained by conversion of hydroxyproline to proline via HypD and P5CR (35, 36). However, while strains with the R20291 background grew similarly as in regular proline-containing CDMM (Fig. 4A–C), most strains with the JIR8094 background comparatively exhibited reduced growth yields and highly unusual growth curves that varied drastically with each mutation (Fig. 4D–F and 4H). Curiously, while hydroxyproline distorted the growth curve of the prdB::ermB mutant, it did not appear to have any negative effect on the prdR::ermB mutant as its growth curve remained similar to the proline-containing CDMM control (Fig. 4G). These data suggest that PrdR plays a Prd-independent regulatory role in the organism’s response to hydroxyproline.

Proline and hydroxyproline addition leads to a decrease in toxin production that is diminished in a ΔselD mutant strain.

Toxin production in C. difficile can be modulated by various environmental factors, including nutrient status of the growth medium (46, 47). Since amino acids are known to have varying effects on toxin synthesis (47–53), it is likely that the three Stickland acceptors play a role in modulating toxin levels in C. difficile, either directly through regulation or indirectly through impacts on growth and bioenergetics. Indeed, proline negatively affects tcdA expression on a transcriptional level via PrdR (20). Thus, we sought to determine the effects of proline, glycine, and hydroxyproline on toxin production. We first assessed the ability of R20291 and JIR8094 to produce toxin in BHIS, TY, and CDMM at 37°C for 72 h (Fig. 5A). We then estimated overall toxin production by measuring the levels of TcdA in spent media by immunoblot with a TcdA-specific antibody. We found that R20291 made a substantial amount of TcdA at 48 h in BHIS, whereas in TY, toxin saturation began much earlier at 24 h. In contrast, TcdA produced by JIR8094 was not detectable in rich media until 48 h. Toxin production in CDMM was incredibly low for both strains and could not be reliably analyzed.

FIG 5.

Proline and hydroxyproline suppress toxin production in a SelD-dependent manner in BHIS. Culture supernatants were harvested, resolved on 7.5% acrylamide gels via SDS-PAGE, and probed with a TcdA-specific monoclonal antibody. (A) Toxin profiles from 72-h time courses of R20291 and JIR8094 in BHIS, TY, and CDMM, as estimated by representative immunoblot against TcdA. (B) Modulation of toxin production by proline, glycine, and hydroxyproline in 48-h BHIS cultures. Proline (Pro), glycine (Gly), and hydroxyproline (Hyp) were added at 30 mM as indicated. The experiment was repeated twice. Representative TcdA immunoblots are shown. (C) TcdA band intensity as quantified by densitometry and with respect to the amount of total protein in each culture. Data points represent the means of normalized band volumes derived from triplicate cultures (including Fig. 5B) while error bars represent standard deviations. Statistical analysis was performed in GraphPad Prism 8 using two-way ANOVA with Dunnett’s multiple-comparison test in which all comparisons were made to BHIS. ns, not significant; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001.

Based on these results, we similarly probed for TcdA using the spent media from the previous BHIS growth studies (Fig. 2) and evaluated toxin production in response to excess proline, glycine, and hydroxyproline at 48 h. We found that proline and hydroxyproline substantially reduced toxin production in R20291, though only the hydroxyproline-dependent effect was deemed to be significant in our analysis. Comparatively, glycine did not have any noticeable effect on toxin levels (Fig. 5B and C). Interestingly, the toxin profile in the ΔselD mutant was strikingly different, as TcdA levels did not significantly change in response to added proline and hydroxyproline. In the ΔselD::selD⁺ mutant, the toxin-reducing effects from supplemented proline and hydroxyproline were restored, and both were found to be significant compared to the BHIS control. These results indicate that the activity of selenoproteins, likely Prd, leads to conditions in the cell that result in lower toxin production through an unknown mechanism. Overall, our findings suggest some involvement of these Stickland acceptors in controlling toxin production, but further experiments are required in order to fully map out the regulatory pathways involved.

DISCUSSION

The strict physiological dependence on Prd and evidence for a hierarchy of electron acceptors.

Our results demonstrated that proline enhances the growth of C. difficile in a manner dependent on Prd (Fig. 2 and 3, and Fig. S3), which is consistent with a previous report (20), though we did not observe substantial changes in the growth rates with our method of analysis (Table S1 and S2). The Prd-dependent biomass increase in vitro likely provides an advantage in vivo, especially since the prdB::ermB mutant has been shown to exhibit a colonization defect in mice transplanted with a dysbiotic human gut microbiota (54). According to a recent report suggesting that proline reduction plays a larger role in establishing colonization resistance compared to bile acid metabolism (55), current therapies may benefit from a greater understanding of the role of proline reduction in C. difficile.

The historical incapability of C. difficile to grow in the absence of proline has long been interpreted as proline auxotrophy (17, 18, 54, 56), but this term is a misnomer since it generally describes a genetic inability to synthesize the nutrient of interest. Recently, C. difficile strain 630 was shown to grow in a glucose-containing medium lacking proline and leucine—another amino acid assumed to be “essential” for the organism (17, 18)—only after repeated subcultures in glucose media containing decreasing concentrations of these Stickland acceptors (45). Based on that report and the fact that several mutants in our study grew in the absence of proline (Fig. 4B, E, F, and G), C. difficile therefore must have the ability to biosynthesize its own proline for anabolic processes (e.g., protein synthesis). Indeed, C. difficile can generate proline from ornithine using ornithine cyclodeaminase; moreover, ornithine can be produced from intermediates derived from other nonessential amino acids like arginine and glutamate via ornithine carbamoyltransferase and acetylornithine aminotransferase, respectively (57). Though, if this biosynthetic pathway is active, it is clearly unable to support the growth of wild-type cells. Therefore, we propose that this unusual behavior stems from a unique “addiction” to using proline as an electron acceptor based on several observations. First, PrdB was abundant throughout almost all stages of growth in several different culture media (Fig. 1), suggesting that vegetative cells highly depend on the Prd complex. Second, Prd- cells can grow in the absence of proline (Fig. 4B, E, F, and G), implying that the presence of this enzyme significantly raises the minimum amount of proline needed for growth. Third, proline is one of the first amino acids rapidly consumed before entering stationary phase, accompanied by the simultaneous production of 5-aminovalerate (31, 33, 45). Lastly, proline reduction lowers the NADH/NAD⁺ ratio, thereby stimulating Rex, a redox-sensing transcriptional regulator, to repress alternative NAD⁺-regenerating pathways (58). We also find it very intriguing that strict proline dependence can be eliminated either by genetically disrupting Prd or by biochemically “weaning” cells off proline (45). However, in the latter strategy, Gencic and Grahame (45) demonstrated via enzyme assays of 630 extracts that Prd activity was unexpectedly higher in the proline-deficient glucose-containing medium compared to BHIS; the authors reasoned that glucose may have induced expression of the prd operon via activation by CcpA, a transcriptional regulator involved in carbon catabolite repression (59). This observation conflicts with our idea of “proline addiction” as this is a clear example of Prd+ cells growing in the absence of proline, suggesting that the presence of Prd may not always force the organism to depend on proline; however, it must be noted that their glucose-only medium was also deprived of leucine, another preferred Stickland acceptor that was present in our preparations of CDMM. Therefore, we believe that the adaptation to conditions deprived of two preferred Stickland acceptors required for growth (proline and leucine) ultimately induces a metabolic reprogramming that is radically different from Prd- cells growing in a culture medium lacking only proline (45). An investigation into the molecular mechanisms behind this adaptation is certainly warranted.

The ability for C. difficile to maintain intracellular redox poise is potentially hampered due to the limitation of electron acceptors in Stickland metabolism. Proline reduction is likely the organism’s most preferred method to dispose of reducing equivalents, but since Prd- cells can grow in the absence of proline, an alternative NAD⁺-regenerating pathway must be compensating for the loss of proline reduction. While the identity of the predominant electron-accepting pathway in Prd- strains is unknown, a possible candidate may be the Wood-Ljungdahl pathway (WLP), a metabolic scheme used by acetogens to reduce two molecules of CO₂ to form acetate (45, 60). Though minimally active in rich media containing substantial amounts of Stickland acceptors (e.g., BHIS), the WLP was found to be highly upregulated in 630 after it had successfully adapted to glucose-containing media lacking proline and leucine (45). In these conditions, the WLP generates acetate in order to fuel the production of butyrate (termed “acetobutyrogenesis”), thereby simultaneously removing electrons and generating ATP for C. difficile. Gencic and Grahame (45) proposed that the coupling of the WLP to butyrate fermentation creates an efficient electron sink even with low flux through the pathways, as the former disposes of reducing equivalents (e.g., reduced ferredoxin) generated by the latter. In glucose-only media, it seems rational for this coupled metabolism to dominate as it would make use of the CO₂ and electrons generated from the oxidation of glucose-derived pyruvate. Since our growth experiments were conducted in glucose-supplemented CDMM under an atmosphere containing 5% CO₂ (Fig. 4), it is possible that the Prd- cells shifted to “acetobutyrogenesis” to compensate for the loss of proline reduction. Furthermore, this metabolic shift is likely controlled by Rex, as it has been shown that genes involved in butyrate fermentation are derepressed in the prdR::ermB and rex::ermB mutants (58). In support of this idea, Gencic and Grahame (45) even noted several Rex-binding sites in the WLP promoter. Based on these studies, evaluating whether the rex::ermB mutant could grow in proline-deficient conditions may help outline the unknown hierarchy of electron acceptors in C. difficile.

Our proline experiments provide further evidence that C. difficile harbors distinct preferences for different electron acceptors (Fig. 6). While reducing power can be derived indiscriminately from multiple sources such as classical Stickland donors (i.e., aliphatic amino acids), sugars and sugar alcohols, or even H₂, the availability of certain electron acceptors mainly determines the process of NAD⁺ regeneration. While this concept has been studied heavily in the model organism Escherichia coli (61), little is known about the electron acceptor hierarchy in C. difficile other than the fact that proline and leucine are the most preferred acceptors and that the former is consumed first before the latter (31, 33, 45). In the event that both preferred acceptors are absent, it is likely that one or more Rex-mediated alternative pathways will attempt to compensate for the decrease in NAD⁺ regeneration (58), though the preferential order of these pathways is unknown. An interesting example of a potential alternative electron sink involves H₂ metabolism, as C. difficile possesses various hydrogenases that mediate the consumption or production of this molecule (62). It must be noted, however, that H₂ is almost always present within the anaerobic chamber due to its requirement for palladium-catalyzed O₂ decontamination. Physiological experiments are unfortunately confounded by the presence of H₂ in the environment, especially if the levels of H₂ are inconsistent between experiments. Indeed, McAllister et al. (41) observed that plasmid complementation of selD only occurred under low H₂ (~1.7%), not high H₂ (4%). Because of this, we deliberately performed experiments at an atmosphere containing a low level of H₂ (~1.0%) that was still sufficient to remove O₂ from our chamber. Further investigations of the electron acceptor hierarchy should be conducted at a low and consistent level of H₂ to lessen the impact of this confounding variable and to allow for direct comparisons between experiments.

FIG 6.

An overview of electron donors and acceptors in C. difficile. Reducing power (NADH) is generated from various electron donors (red). To regenerate NAD⁺ for continued oxidation of fuel sources, certain enzymes dispose of electrons using specific molecules as electron acceptors in a set order (blue). C. difficile harbors preferences for electron acceptors based on a poorly understood hierarchy. The most preferred substrate is proline. Leucine is also a preferred substrate but is probably lower on the hierarchy. Prd elimination, which cures “proline addiction” as demonstrated in this study, may inadvertently activate alternative NAD⁺-regenerating pathways in an unknown order (indicated by blue circles with dashed outlines). Electron acceptors that are the focus of this study are in bold. Hydroxyproline (*) is not a true Stickland acceptor but serves as an additional source of proline for Prd. NADH was chosen to represent electron flow for simplicity, but it must be noted that additional carriers besides NADH exist (e.g., ferredoxin). The donors and acceptors listed here are representative and by no means comprehensive. NADH, NAD (oxidized/reduced); SCFA, short-chain fatty acids; WLP, Wood-Ljungdahl pathway.

The complex role of glycine and the effects of Grd.

Our growth experiments revealed that glycine offered no obvious benefit to vegetative C. difficile. Furthermore, in rich media supplemented with 30 mM glycine, the growth yields of several strains decreased in a Grd-independent manner (Fig. 2 and 3, and Fig. S3). Although the idea of glycine-mediated toxicity seemed puzzling considering the amino acid’s role in Stickland bioenergetics, it is well established that glycine exhibits an inhibitory effect as a result of its misincorporation into the peptidoglycan of several bacteria (63). Even in C. difficile, glycine likely functions as a cell wall-weakening agent when in excess, as Bhattacharjee and Sorg (64) demonstrated its requirement for the generation of competent cells; indeed, the authors observed that R20291 and 630 cells curiously adopted a curved morphology after growth in BHIS supplemented with 1% glycine (~133 mM) (64). On the basis of these studies, the glycine-dependent decrease in growth yield could possibly be explained as cell lysis due to the weakening of cell walls resulting from glycine misincorporation. Previous studies, however, did not always describe growth inhibition at 30 mM glycine (19, 20), so our interpretation is that this concentration approaches, but does not necessarily equate to, a reproducibly toxic amount of glycine for C. difficile.

Despite the fact that glycine reduction supports the generation of ATP via substrate-level phosphorylation, we observed no change in the growth of the grdA::ermB mutant compared to JIR8094 in rich media (Fig. 2H and 3H), an observation consistent with another report (20). While appearing to have no obvious phenotype in those media, this strain exhibited a highly distorted growth profile in CDMM compared to wild type (Fig. 4H), in that there was no rapid loss in turbidity immediately after achieving maximum culture density (which decreased substantially compared to wild type). It is interesting that the loss of Grd caused a growth defect compared to Prd- strains, which still grew as well as wild type (Fig. 4E, G, and H), indicating that Grd plays an important physiological role in this medium that is not readily compensated by another process in C. difficile. Although, while the significance behind its role as a substrate for Grd is still unknown, glycine has been well established to be a known cogerminant for C. difficile spores (4). Recently, Leslie et al. (65) demonstrated that gnotobiotic mice precolonized with nonlethal strain 630 were protected by lethal strain VPI 10463 in that the former apparently prevented germination of the latter by depleting glycine, though the study did not examine whether this was due to a Grd-dependent mechanism. Given the fact that a recent report suggests the importance of glycine reduction in establishing colonization resistance (55), an interesting link between Grd and spore biology may likely exist in CDI. McAllister et al. (41) reported that, while the presence of SelD apparently plays no role in spore germination, it does specifically affect sporulation and outgrowth from germinated spores. If Grd is one of the selenoenzymes responsible for those phenotypes, the ATP and NAD⁺ yield from glycine reduction would likely be more relevant during spore outgrowth, as this process relates to vegetative cell development. Indeed, since Grd- cells exhibited a growth defect in CDMM (Fig. 4H), it would be interesting to evaluate the outgrowth kinetics of the grdA::ermB mutant to determine if glycine, in addition to acting as a cogerminant, plays a larger role in spore biology than originally assumed. Investigating the potential relationship between Grd and spore-related processes may reveal critical mechanisms that could be exploited by therapeutics.

Our growth experiments using ⁷⁵Se revealed vast differences in the levels of labeled GrdA and GrdB (Fig. 1). We found this puzzling considering that grdA and grdB are adjacent to each other in the grd operon and are likely under the control of the same unknown promoter. This discrepancy could be explained by a difference in the molar ratio of GrdA to GrdB in the Grd complex; however, as there are no reports that describe the biochemical characterization or modeling of Grd, the true stoichiometry of the native complex is unknown. Unfortunately, interpretation of the ⁷⁵Se profile of GrdA is additionally confounded by what appears to be a similarly-sized band that migrated slightly higher than GrdA; this band is most obvious within the range of 6 h to 12 h in BHIS (Fig. 1A and B). We are currently unaware if GrdA undergoes a chemical modification that slightly changes its molecular weight or if a different selenoprotein of similar size to GrdA (~16.8 kDa) appears only during early growth.

The varying physiological effects of hydroxyproline in different strains.

The addition of hydroxyproline to rich media increased the growth yield of C. difficile even in the absence of Prd (Fig. 2 and 3, and S3). This result was surprising as we initially assumed that hydroxyproline would primarily fuel Prd through full conversion to proline via HypD and P5CR; furthermore, it has recently been demonstrated that hydroxyproline-dependent growth yield enhancement requires HypD (66). Given that hydroxyproline is not a true substrate for Prd (19), its conversion to proline would mainly fuel non-Stickland processes in Prd- strains; however, it is interesting to note that P5CR also regenerates NAD⁺ in the process of generating proline (36), allowing Prd- strains to fulfill the bioenergetic requirement with hydroxyproline using a Stickland-independent mechanism. Thus, in an infection, C. difficile potentially benefits from scavenging hydroxyproline from collagen (in which it is a major constituent), by providing itself with two unique opportunities to dispose of electrons: (i) conversion to proline in a Stickland-independent manner using P5CR, and (ii) Stickland-dependent reduction of converted proline by Prd.

Our experiments in CDMM curiously revealed that the effect of hydroxyproline varied greatly between strains. In the case of R20291 and related mutants, the growth yields in regular proline-containing CDMM and proline-deficient CDMM supplemented with hydroxyproline were approximately similar (Fig. 4A, B, and C). On the other hand, JIR8094 and several of its mutants grew very strangely and at substantially lower yields in comparison (Fig. 4D–F and 4H), implying that hydroxyproline poorly compensates for the absence of proline in this strain. A recent report on the role of HypD in CDI likely sheds some light on this phenomenon as it was demonstrated that a ΔhypD mutant exhibited a significant growth defect in CDMM when proline was replaced with hydroxyproline (compared to its parent strain 630Δerm which grew better in the same medium) (66). Moreover, Reed et al. (66) found that hypD expression varied in several strains of C. difficile; for example, as determined by qRT-PCR, the expression of hypD (reported as relative copy number) increased in R20291 and VPI 10463 in the presence of hydroxyproline (between 1 and 10 copies), whereas 630 did not exhibit any change in its minimally-expressed hypD (between 0.1 and 1 copy). Based on these observations, there may be a rational explanation for why R20291 and JIR8094 respond differently to hydroxyproline. On one hand, R20291 probably expresses enough HypD to facilitate a 1:1 conversion of hydroxyproline to proline, allowing for optimal growth. On the other hand, one might assume that JIR8094 barely expresses HypD even in the presence of hydroxyproline (similar to its parent strain 630), likely resulting in inefficient conversion and, ultimately, substandard growth. In the context of an infection, these results suggest that not all strains have the same capacity to benefit from hydroxyproline released from degraded collagen.

Another element to consider is the fact that the prdR::ermB mutant grew better in hydroxyproline-containing CDMM compared to JIR8094; in fact, the mutant growth curves in this medium and the proline-containing control were almost identical, as if disrupting prdR somehow alleviated the growth defect (Fig. 4G). Since this effect was not observed in the prdB::ermB mutant, it may be that PrdR performs another function unrelated to the expression of prd. The binding capability of PrdR has not been characterized, but it is assumed that it binds proline; therefore, if PrdR also had an affinity for hydroxyproline based on its structural similarity to proline, one could assume that the transcriptional regulator functions to “sense” the levels of hydroxyproline in the cell for an unknown process. Whether or not PrdR directly regulates other genes besides the prd operon is unknown, but an investigation of the potential effects of PrdR on the hydroxyproline utilization genes (e.g., hypD) may yield interesting results.

Toxin production and the impact of selenoproteins on its regulation.

We found that the addition of proline and hydroxyproline to BHIS caused a reduction in the amount of TcdA, but this effect was abolished in a ΔselD mutant (Fig. 5B and C). These results suggest that toxin production is not affected by the presence of these amino acids but, rather, by their involvement in pathways regulated by selenoproteins. Since Prd is a selenoenzyme, it is tempting to think that proline reduction creates a condition that results in the downregulation of toxins. In support of this, tcdA expression was shown to decrease in JIR8094 when grown in proline-supplemented TY; moreover, the prdR::ermB mutant expressed higher levels of tcdA regardless of whether proline was present (20). Additionally, this same mutant was more virulent than JIR8094 in a hamster model (58). On the basis of these observations, Bouillaut et al. (58) suggested that PrdR indirectly influences Rex to downregulate toxin production; specifically, Prd would lower the NADH/NAD⁺ ratio, thereby causing Rex to repress the pathway involved in the synthesis of butyrate, a toxin-inducing molecule (53). Indeed, a rex::ermB mutant had a higher abundance of tcdA transcript compared to JIR8094 (58). While these results indicate a role for Prd in toxin suppression, the prdB::ermB mutant paradoxically harbored lower levels of tcdA transcript in TY regardless of proline addition; moreover, the authors found no significant difference in the relative abundance of tcdA mRNA from this mutant cultured in TY and JIR8094 cultured in proline-supplemented TY (20). Furthermore, in mice, the prdB::ermB strain made considerably less TcdB compared to JIR8094 (54), while the rex::ermB strain was actually less virulent in hamsters than JIR8094 (58). To help make sense of these issues, the authors proposed that PrdR likely regulates toxin production via a mechanism independent of Prd and Rex (58). Given the fact that PrdR is an enhancer-binding protein for the alternative sigma factor σ⁵⁴, an additional layer of regulation may be influencing toxin production. In a recent investigation of SigL (σ⁵⁴) in C. difficile, Clark et al. (67) reported pleiotropic effects on toxin production and other processes in two sigL::erm mutants from strains of two different ribotypes. Since a SigL-dependent promoter does exist within the prd operon, it would be interesting to understand how PrdR and SigL specifically influence this operon in different strains, which may give us insight into the mechanism behind proline-mediated suppression of toxin production.

Our experiments to determine the effects of amino acids on toxin production were performed mainly with R20291 and its related mutants (Fig. 5B and C). Evaluating the amount of TcdA secreted from the prdB::ermB and prdR::ermB strains in the presence and absence of proline and hydroxyproline may help determine the nature of the toxin phenotype in the ΔselD mutant and clarify any strain-dependent discrepancies and issues mentioned in the previous section; however, these strains did not produce enough TcdA to be detected even up to 72 h, regardless of the medium. This is likely due to the fact that JIR8094 has been reported as a poor producer of toxins (68, 69). Moreover, since our experiments with R20291 were performed in BHIS, it is possible that the toxin profile differs based on the medium. Unfortunately, we were unable to obtain reproducible results in similar experiments using TY. Finally, because the prdR::ermB and rex::ermB strains were reported to express higher levels of tcdA transcript in CDMM compared to JIR8094 (58), we attempted to evaluate TcdA levels in CDMM with our method. However, in this medium, none of our strains produced enough TcdA to be detected even at 72 h.

Due to the recent discovery that C. difficile can degrade collagen for nutrients in an infection (16), our findings potentially describe an intriguing mechanism for nutrient scavenging. According to our results, the release of proline and hydroxyproline from collagen would simultaneously boost the growth of C. difficile and suppress toxin production. While this downregulation is puzzling considering that toxin-mediated inflammation is required for collagen degradation, uncontrolled toxin expression would likely cause the death of the host, essentially depriving the pathogen of a vital source of amino acids. We propose that the metabolism of collagen-derived proline and hydroxyproline signals a “fed state” in C. difficile, which responds by downregulating toxin production in order to avoid killing the host, thereby maintaining its niche. Overall, our findings describe a CDI model in which the organism makes key decisions in several important scenarios depending on the presence or absence of specific nutrients (Fig. 7). As this model is based on the limited information regarding the pathogen’s electron acceptor preferences, investigations focused on uncovering the specific order of these electron acceptors in the hierarchy will prove beneficial. Understanding the nature of this hierarchy will likely paint a clearer picture of C. difficile pathogenesis and the organism’s ability to thrive in infected patients.

FIG 7.

A model depicting C. difficile infection as a series of decisions based on nutrient environment. Newly ingested C. difficile spores travel to the small intestine and will germinate into vegetative cells if taurocholate and glycine are present; depletion of either nutrient would result in no spore germination. Early colonization of vegetative C. difficile requires proline, likely due to a forced “addiction” imposed by the presence of Prd. If proline is absent, the pathogen may rely on alternative NAD⁺-regenerating pathways mediated by Rex. In this scenario, C. difficile scans the environment for other electron acceptors. If leucine or glycine is present, reductive Stickland metabolism occurs. If Stickland acceptors are absent, C. difficile may resort to utilizing non-Stickland bioenergetic schemes such as the Wood-Ljungdahl pathway coupled with butyrate fermentation. On the other hand, the presence of proline likely promotes effective colonization and growth of vegetative cells. In order to maintain the niche and thrive in the host, the pathogen will produce toxins to obtain host-derived nutrients (e.g., Stickland acceptors from collagen). Although not fully understood, the lifestyle of C. difficile is ultimately defined by a delicate balance of controlled growth, toxin production, and sporulation. Box colors indicate the following: optimal infection scenarios (green), suboptimal infection scenarios (yellow), and a failure to colonize (red).

MATERIALS AND METHODS

Bacterial strains and culture media.

All C. difficile strains used in this study are listed in Table 1. Experiments were routinely performed under an atmosphere of ~1% H₂, 5% CO₂, and >90% N₂ using a Coy anaerobic chamber. H₂ was maintained within a difference of 0.2%, as measured by a Coy anaerobic monitor (CAM-12). For physiological studies in rich media, strains were grown in either BHIS (37 g/L brain heart infusion [Oxoid], 5 g/L yeast extract, 0.1% l-cysteine) or TY (30 g/L tryptone, 20 g/L yeast extract, 0.1% mercaptoacetic acid) (42, 43). For physiological studies in a defined minimal medium, strains were grown in CDMM (18). The amino acids (mg/L) in CDMM were as follows: l-tryptophan (100), l-valine (100), l-isoleucine (100), l-leucine (1,000), l-cysteine hydrochloride monohydrate (500), l-proline (800), l-arginine hydrochloride (100), l-histidine monohydrochloride monohydrate (100), l-methionine (100), glycine (100), and l-threonine (100). When indicated, l-proline and glycine were omitted while l-4-hydroxyproline was substituted for l-proline at equimolar concentration. The vitamins (mg/L) in CDMM were as follows: calcium-d-pantothenate (1), pyridoxine (0.1), and biotin (0.01). Glucose was present in CDMM at 0.2%. The following salts and metals (mg/L) in CDMM were as follows: potassium phosphate monobasic (300), sodium phosphate dibasic (1,500), sodium chloride (900), calcium chloride dihydrate (26), magnesium chloride hexahydrate (20), manganese chloride tetrahydrate (10), ammonium sulfate (40), ferrous sulfate heptahydrate (4), cobalt chloride hexahydrate (1), and sodium bicarbonate (5,000). Additionally, the recipe was modified to include nickel chloride hexahydrate (1 mg/L), sodium selenite (1 μM), sodium molybdate dihydrate (1 μM), and sodium tungstate dihydrate (1 μM).

Growth studies and analysis.

Briefly, single colonies of each strain were used to inoculate 5-mL broths (BHIS, TY, or CDMM) and were expanded overnight at 37°C. Overnight cultures were then diluted 100-fold the following day in their respective media and transferred to 96-well plates in 200-μL volumes in triplicate. Diluted cultures were grown at 37°C in a BioTek Epoch 2 Microplate Spectrophotometer, and the OD₆₀₀ of each culture was automatically recorded every 30 min over a 48-h period. Before every read, cultures were resuspended for 5 sec using the double orbital function on the fast speed.

Radiolabeling studies with ⁷⁵Se.

Overnight cultures were diluted 100-fold in their respective media, each containing 2 μCi ⁷⁵Se per mL medium along with “cold” sodium selenite at 50 nM. Cultures were then grown at 37°C for 72 h. At several time points, culture aliquots (2 mL) were harvested at 12,000 × g (BHIS, TY) or 16,000 × g (CDMM) for 5 min. After discarding supernatants, cell pellets were frozen at −80°C until ready for use. Cell pellets were resuspended in 50 μL cold lysis buffer (50 mM Tris-HCl, 0.5 mM EDTA, 0.1 mM benzamidine, [pH 8.0]) and homogenized with a sonic dismembrator (Fisher Scientific, Model 1000). Lysates were clarified at 16,100 × g for 5 min and subsequently loaded onto Tris-glycine gels (15% acrylamide) to achieve either 7.5 μg (BHIS), 2.5 μg (TY), or 1.0 μg (CDMM) soluble cellular protein each lane, as estimated by Bradford assays calibrated with bovine serum albumin (BSA) (70). Proteins were resolved at 200 V for 50 min and stained with GelCode Blue for 1–2 h. After overnight destaining with dH₂O, gels were soaked with a drying solution (30% methanol, 5% glycerol) for 15 min and then dried overnight. Dry gels were exposed to a phosphor screen for at least 48 h and subsequently analyzed using a Personal Molecular Imager (Bio-Rad).

Analysis of TcdA production.

Toxin production of 48-h cultures derived from each growth study were analyzed by TcdA immunoblot. Briefly, total protein was estimated with the Bradford assay as described above. Triplicate cultures were harvested at 2,500 × g and 4°C for 20 min (Hermle Z400K, Labnet). Cell-free supernatants were collected and frozen at −20°C until day of use. Thawed supernatants were mixed with equal volumes of 2× Laemmli buffer (1:1) and subsequently incubated in a 100°C sand bath for 5 min. Denatured samples (20 μL) were loaded onto Tris-glycine gels (7.5% acrylamide) for SDS-PAGE and resolved for 1 h at 200 V. Transfer of electrophoresed samples to polyvinylidene difluoride membranes was performed at 4°C overnight at 30 V. Membranes were incubated in a blocking buffer (20 mM Tris-HCl, 150 mM NaCl, 0.01 mM EDTA, 0.1% Tween 20, 1% BSA, [pH 7.5]) for 1 h at room temperature. Detection of TcdA was performed using a monoclonal mouse anti-TcdA antibody (PCG4.1, Novus Biologicals) and a rabbit anti-mouse IgG antibody conjugated with alkaline phosphatase. Blots were visualized with a ChemiDoc XRS+ imaging system (Bio-Rad).

Semiquantitative analysis of TcdA was performed using Image Lab 6.0 software (Bio-Rad). Briefly, after subtracting background noise, peaks corresponding to the target band were selected with the software’s Lane Profile tool, in which the peak density (arbitrary units) was normalized to the amount of total protein measured from each sample. Normalized TcdA values derived from identical triplicate cultures were averaged together and subsequently converted to percentages using the Normalize tool in GraphPad Prism 8. The mean TcdA value from R20291 in BHIS was set as 100% while 0 was set as 0%.