Flagellar switch inverted repeats impact heterogeneity in flagellar gene expression and thus C. difficile RT027/MLST1 virulence

SUMMARY

Clostridioides difficile (C. difficile) RT027 strains cause infections that vary in severity from asymptomatic to lethal, but the molecular basis for this variability is poorly understood. Through comparative analyses of RT027 clinical isolates, we determine that isolates that exhibit greater heterogeneity in their flagellar gene expression exhibit greater virulence in vivo. C. difficile flagellar genes are phase-variably expressed due to the site-specific inversion of the flgB 5^′ UTR region, which reversibly generates ON vs. OFF orientations for the flagellar switch. We find that longer inverted repeat (IR) sequences in this switch region correlate with greater disease severity, with RT027 strains carrying 6A/6T IR sequences exhibiting greater phenotypic heterogeneity in flagellar gene expression (60%–75% ON) and causing more severe disease than those with shorter IRs (>99% ON or OFF). Our results reveal that phenotypic heterogeneity in flagellar gene expression may contribute to the variable disease severity observed in C. difficile patients.

In brief

Nguyen et al. demonstrate that variation in inverted repeats within C. difficile’s flagellar switch impacts the phase-variable expression of flagellar genes, which contributes to the variable disease outcomes observed within a single RT027 strain type. They find that greater heterogeneity in flagellar gene expression increases C. difficile’s virulence.

Graphical Abstract

INTRODUCTION

Clostridioides difficile (C. difficile) is the leading cause of hospital-acquired infection in the United States, with an estimated incidence of approximately 224,000 cases per year.¹ The incidence of community-acquired C. difficile infection (CDI) is also increasing, with up to 51.2 cases per 100,000 population.² The major risk factors for developing CDI include broad-spectrum antibiotic usage, as antibiotics deplete the commensal bacterial species that provide colonization resistance against CDI. However, CDI disease severity in humans ranges from asymptomatic colonization to diarrhea to severe pseudomembranous colitis and even death. Recent studies have revealed that the gut microbiota and host immunity impact CDI pathogenicity,³ but genetic features of C. difficile likely also regulate disease severity in humans.

C. difficile virulence requires the production of at least one of its glucosylating toxins, TcdA and/or TcdB.⁴ These toxins are internalized by epithelial cells via endocytosis^5–7 and are released into the cytosol, where they glucosylate and inactivate intracellular guanosine triphosphatases.^8–10 These activities disrupt cell signaling pathways and cytoskeletal structure, leading to cell death.¹¹ The impact of TcdA and TcdB on disease progression and severity, however, can vary, depending on the toxin titer, the strain of C. difficile, and the host’s innate and adaptive immune defenses.^12–14

In addition to toxin expression, flagellar-mediated motility has been suggested to modulate C. difficile virulence. Flagellar motility contributes to the virulence of Salmonella enterica and Escherichia coli by enhancing host cell invasion, adhesion to epithelial cells, and systemic inflammation.^15–17 However, the impact of flagellar motility on C. difficile infection in vivo remains controversial and is likely strain and host dependent.^18–20 Deletion of fliC, which encodes flagellin, did not impact the virulence of C. difficile R20291 in a germ-free mouse model, yet it reduced its virulence in antibiotic-treated SPF mice.^18,19 In contrast, the deletion of fliC in C. difficile strain CD630 increased its virulence in a hamster model of CDI.²⁰ Notably, flagellar genes are heterogeneously expressed in C. difficile due to the inversion of a flagellar switch sequence by the tyrosine recombinase RecV.²¹ Inversion of this region in the 5^′ UTR of flgB to the ON orientation allows for the expression of a flagellar gene operon that includes the gene encoding the sigma factor, SigD, which directly induces the expression of flagellar gene operons as well as tcdR, which encodes another sigma factor that activates toxin gene expression.²² Thus, flagellar gene expression is coupled with toxin gene expression.^21,23 Despite these insights, the impact of heterogeneity in flagellar gene expression on CDI virulence remains unclear.

In this study, we investigated the relationship between flagellar gene expression and virulence in multi-locus sequence type 1 (ST1) C. difficile strains isolated from patients with CDI. ST1 strains, also known as the NAP1/B1/ribotype 027, are highly transmissible, have increased multidrug resistance,^24,25 and were initially reported to be hypervirulent.^26,27 However, heterogeneity in the virulence^28,29 and toxin production of ST1 strains^30,31 has been reported, revealing that there is considerable variation between individual ST1 strains. By examining 22 ST1 clinical isolates in a mouse model of infection, we observed marked differences in virulence. We found that disease severity correlated with sequence differences in the flagellar switch region between ST1 strains, specifically differences in the inverted repeat (IR) region. Our data reveal that these sequence differences alter the invertibility of the flagellar switch, with longer IR sequences causing greater heterogeneity in flagellar gene expression within the population (i.e., generating a mixture of flagellar gene ON and OFF), and shortened IR sequences being associated with more fixed populations of either flagellar gene ON or OFF. Furthermore, exchanging a longer IR sequence (6A/6T) with a shorter IR (5A/5T) sequence in C. difficile R20291 significantly reduced its virulence, particularly when combined with the OFF flagellar switch orientation. Our results argue that flagellar switch IR sequences alter the invertibility of the flagellar switch, which in turn contributes to the variable disease outcomes observed between C. difficile ST1 strains.

RESULTS

Flagellar gene expression correlates with the virulence of clinical C. difficile RT027 isolates

We previously showed that the in vivo virulence of C. difficile ST1 isolates in C57BL/6 mice varies widely, ranging from mortality to the absence of any weight loss or diarrhea.³¹ While we showed that the avirulence of two ST1 strains was due to a small internal deletion in the cdtR gene, the remaining ST1 strains did not exhibit any genetic variability in their pathogenicity and binary toxin loci, which encode the TcdA and TcdB glucosylating toxins and the binary toxin CDT, respectively³¹ (Figure S1A). In accordance, the virulence differences (% weight loss) observed between these strains did not correlate with the fecal toxin levels measured using a cell-based toxicity assay (Figures S1B and S1D). In addition, no correlation in the colonization (colony-forming units [CFUs]) levels on day 1 post-infection and virulence of the strains was observed (Figures S1B and S1C). To identify additional mechanisms for variable virulence among ST1 strains, we selected two isolates that induced >10% weight loss (ST1–12 and ST1–53) and two isolates that resulted in <10% weight loss (ST1–6 and ST1–27) for further study (Figure S1). Here, we refer to them as high-virulence and low-virulence isolates.

We first compared the transcriptional profile of the four isolates during infection of antibiotic-treated mice by harvesting cecal contents 1 day post-infection and conducting RNA sequencing (RNA-seq) profiling of the isolates (Figure 1A). The transcriptomic profile revealed that flagella-related genes are overexpressed in high-virulence isolates relative to low-virulence isolates (Figure 1B). These differences in flagellar gene expression were also observed when RT-qPCR was used to compare the expression of two flagellar genes, fliE and fliS1, between the strains using the same cecal content samples (Figure S2A). High-virulence isolates pre-grown in broth culture also exhibited greater flagellation and spread further on swim plates 24 h (Figures 1C and 1D). In contrast, the low-virulence isolates were aflagellate and exhibited delayed spreading on the swim plates (Figures 1C and 1D). Notably, the RNA-seq analyses confirmed that toxin genes, including tcd genes and cdt genes, were expressed at similar levels between the four strains during murine infection (Figure 1B), consistent with the similar levels of toxins detected during infection in mice (Figures S2B and S2C). While ST1–27 appeared to colonize mice at higher levels compared to the other three isolates (Figure S2C), the growth of these four isolates did not differ from one another in broth culture (Figure S2D). Interestingly, RT-qPCR analyses performed with the same cecal content samples revealed that the high-virulence strain ST1–12 expressed tcdA and tcdB to higher levels than the other three strains, including the other high-virulence strain ST1–53. While these data did not fully reproduce the RNA-seq observations, they support the finding that toxin production differences do not fully explain the variation in virulence observed among ST1 isolates (Figure S2A).

Figure 1. High virulence isolates make more flagella.

(A) The schematic of the mouse infection.

(B) Heatmap of RNA sequencing results from the cecal contents of mice infected with the four ST1 isolates shown (n = 5 mice per isolate). The heatmap is generated based on the number of transcripts per sample after normalization.

(C) Transmission electron micrographs of the four isolates. Scale bar: 1 μm

(D) Swim plate results of 4 isolates after 24 h incubation (n = 3 replicates per isolate). Data are represented as mean ± SD. High-virulence isolates are pink. Low-virulence isolates are black. Statistical significance was calculated by one-way ANOVA; *p < 0.05.

See also Figures S1 and S2.

Variation in the flagellar switch IRs of RT027/ST1 clinical isolates

The difference in flagellar gene expression between high-virulence and low-virulence isolates led us to compare their flagellar genomic regions. While the flagellar genes were identical between the isolates, we identified sequence differences in the flagellar switch region, which plays an important role in regulating the expression of the flagellar operons.²¹ The flagellar switch region comprises a central region flanked by a left-IR (LIR) and a right-IR (RIR) (Figure 2A). The IRs are recognized by the tyrosine recombinase RecV, which inverts the central region. This leads to the switch region exhibiting either an ON or OFF orientation, which allows or inhibits flagellar gene expression, respectively.²¹ When we expanded the analysis of the flagellar switch region to all 68 ST1 isolates in our collection, we found that 65 of the isolates had a flagellar switch central sequence that is identical to the reference RT027/ST1 strain, R20291, and thus carried for further analyses. Most notably, we found that the sequence of the LIRs and RIRs flanking the central flagellar switch region varies between strains. Specifically, while most isolates have IRs identical to the reference strain C. difficile R20291, 40% of isolates have at least one less A or T in either the LIRs or RIRs (Figure 2A; Table S1). In total, four types of IR flanking regions were identified in our strain collection; we named them Common1 (C1–59.38% 6A/6T), Common2 (C2–32.81%, 6A/5T-ON or 5A/6T-OFF), Rare1 (R1–6.25%, 5A/6T-ON or 6A/5T-OFF), and Rare2 (R2–1.56%, 5A/5T). The C1 switch region, the most common sequence, is identical to that observed in C. difficile R20291, with 6 As on the LIR and 6 Ts on the RIR. When C1 inverts between the OFF vs. ON orientation, the number of As and Ts in both IRs remains the same. In contrast, the composition of the C2 region differs depending on the orientation of the switch region: in the OFF orientation, the LIR consists of 5 As and the RIR consists of 6 Ts (C2-OFF); in the ON orientation, the LIR consists of 6 As and the RIR consists of 5 Ts (C2-ON). The R1 region also has an asymmetric distribution depending on the switch orientation, except that in the OFF orientation, the LIR consists of 6 As and the RIR consists of 5 Ts (R1-OFF). The R2 region IRs remain the same regardless of the switch orientation, with the LIR consisting of 5 As and the RIR consisting of 5 Ts. However, the R2 is only observed in one strain, ST1–67, where it is in the OFF orientation. Notably, Sanger sequencing of the flagellar switch region confirmed the HiSeq whole-genome sequencing analyses (Table S1).

Figure 2. Variations in flagellar inverted repeat types observed among ST1 isolates.

(A) Schematic of the flagellar switch and the alignment of the flagellar switch regions from representative isolates. The full list is in Table S1.

(B) Inverted repeat (IR) types of isolates collected from around the world.

(C) IR types of isolates collected in the United States and by the CDC.

To assess the conservation of the IR types observed, we screened 1,359 RT027/ST1 isolates whose whole-genome sequences were available from National Center for Biotechnology Information (NCBI) BioSamples. These isolates derive from the Centers for Disease Control and Prevention (CDC) healthcare-associated infection sequencing (HAI-seq) C. difficile collection³² and Texas and Michigan medical centers.³³ While isolates from more recent publications were also included,^34–37 all C. difficile isolates were collected from 1988 to 2020. Most isolates (1,318/1,359) derive from continental Europe, the United Kingdom, Ireland, and the United States, with a small number (20/1,359) of isolates from east Asia and Australia. Most C. difficile ST1 isolates carry the flagellar switch IR type C1, representing 96.5% (1,311/1,359) of the isolates analyzed. Type C2 was observed in only 2.4% (33/1,359) of isolates, most (27/33) of which were obtained in the United States between 2007 and 2020. The other IR types were detected at even lower frequencies, representing only 0.66% of isolates (Figure 2B). We additionally analyzed 15 ST1 isolates from Latin American regions, as they were reported to have distinct characteristics compared to other areas of the world.^38,39 C1 remains the dominant IR type among C. difficile isolates from Latin American countries (Figure 2B).

When more recent CDC HAI-seq collection strains were analyzed, which were predominantly submitted between 2009 and 2020, the IR type C1 remained dominant, but the C2 represented 30% (19/63) isolates, similar to our collection (32.8%). Interestingly, the proportion of C2 increased in isolates collected in 2020, accounting for 53% (9/17) of isolates collected in 2020 (Figure 2C). Thus, our data suggest that variability in C. difficile flagellar IR regions has increased over time, although their impact on C. difficile virulence remains unclear.

Flagellar invertibility has a strong association with the virulence of RT027/ST1

To analyze the impact of IR types on RT027/ST1 virulence, we correlated the weight loss in mice caused by C. difficile infection with different IR types. We found a strong correlation between IR type C1 and more severe weight loss in infected mice (p = 3.2e–5) (Figures 2A and 3A). Since a small deletion in the RIR and surrounding region greatly reduces the invertibility of C. difficile flagellar switch, essentially locking it in one orientation,²³ we tested whether the different IR types impact the invertibility of the flagellar switch region. Using orientation-specific qPCR analyses to quantify the fraction of ON and OFF cells in C. difficile populations from liquid culture (Figure 3B),²¹ we found a weak correlation between greater weight loss and strains with a higher proportion of ON cells in a population by linear regression (R² = 0.39). However, the most virulent isolates had ~70% of cells in the ON orientation and consisted of the C1 IR type (Figure 3C, green circles). Furthermore, isolates with the other IR types exhibited lower virulence than the C1 isolates and were either nearly 100% ON or 100% OFF (Figure 3C). Thus, our results suggest that IR type governs the flexibility of the flagellar switch region and that IR types that lead to more heterogeneous flagellar gene expression (i.e., the C1 IR type) enhance the virulence of C. difficile strains.

Figure 3. High-virulence isolates exhibit more heterogeneous flagellar gene expression.

(A) Percentage maximum weight loss graphed by IR types. Boxplots show all data points, with the center line representing the median.

(B) Schematic of the qPCR strategy to measure the proportion of Flagella-ON vs. -OFF cells in the population.

(C) Correlation between percentage of maximum weight loss and percentage of Flagella-ON cells. DNA was extracted from bacterial cultures during logarithmic growth.

The IR type correlates with the invertibility of the flagellar switch

To further investigate whether different IR types impact the flagellar switch invertibility, we analyzed the flagellar orientation of our 64 ST1 isolates across 3–4 biological replicates per isolate. We grew individual isolates to exponential phase in broth culture and prepared the genomic DNA for orientation-specific qPCR analyses. When calculating the proportion of C1 cultures in the ON orientation, we found that the C1 IR type is slightly biased toward the ON orientation, with approximately 73% (55%–99%) in the ON orientation (Figure 4A). In contrast, the biased-OFF isolates, including C2-OFF, R1-OFF, and R2-OFF, have 95.6% (88.7%–100.0%) of their population in the OFF orientation, whereas the biased-ON isolates, including C2-ON and R2-ON, have 96.7% (92.6%–100.0%) of their population in the ON orientation (Figure 4A). Such a strong correlation between IR types and biased-flagellar orientations indicates that mutations in the IR regions greatly impact the switching invertibility.

Figure 4.

Flagellar IR types are associated with flagellar switch invertibility in culture and in mice

(A) Proportion of the population with the flagellar switch region in the ON orientation during logarithmic growth in broth culture and (B) during infection over time as determined by qPCR. The results reflect samples obtained from the feces (days 1–7) and cecum (day 8) of C. difficile-infected mice. Boxplots show all data points, with the center line representing the median. Statistical significance was calculated by one-way ANOVA; **p < 0.01; ***p < 0.001; ****p < 0.0001.

See also Figure S3.

To test whether the biased IR types also restrict the invertibility of C. difficile isolates in vivo, we infected mice with C. difficile R20291 and clinical isolates with different IR types and monitored the ON and OFF cell fractions in fecal or cecal contents over the course of 8 days. On day 1 post-infection, we found that C. difficile isolates with the C1 IR type had about ~75% of their population with the ON orientation (Figure 4B), similar to our broth culture analyses (Figure 4A). Interestingly, the proportion of C1 IR type strains in the ON vs. OFF cells exhibited greater variability over the course of the infection compared to the isolates with IR types C2, R1, and R2 (Figures 4B and S3A). These three IR type strains maintained their flagellar switch region in a biased orientation over the 8-day infection course (Figure 4B). No clear correlation was observed between the proportion of Flagella-ON vs. Flagella-OFF cells in the population and the colonization levels (Figure S3B). We also measured fecal toxin levels from fecal pellets on days 1 and 7 and cecal toxin levels on day 8; again, the proportion of Flagella-ON and Flagella-OFF cells does not appear to impact the total toxin levels (Figure S3C). Overall, these data further highlight the strong correlation between IR type and flagellar switch invertibility. Specifically, the C1 IR type (6A/6T) exhibits greater invertibility in its flagellar switch region, while the other IR types appear to “lock” the flagellar switch in either the ON or OFF orientation.

C. difficile flagellar IR type impacts RecV-mediated DNA inversion

Since IR types correlate with the invertibility of ST1 isolates (Figure 4), we wondered whether the IR types specifically affect RecV-mediated DNA inversion. To test this possibility, we assessed the invertibility of IR types in a heterologous host using an E. coli-based colorimetric assay.⁴⁰ E. coli were co-transformed with plasmids encoding C. difficile recV and a second plasmid containing a promoter flanked by either the C1, C2, or R2 IR type. Depending on the initial orientation of the promoter within the switch region flanked by the IR variants, the switch will drive the expression of either rfp (left panel) or gfp (right panel). RecV-mediated DNA inversion of the reporter plasmid results in a switch from red to green (or green to red) (Figure 5A). When the reporter plasmid carried the C1 IR type, a high percentage of dual plasmid-transformed colonies exhibited a color switch (Figure 5B). In contrast, when the reporter plasmid carried the C2 IR type, only one to two colonies exhibited a color switch, whereas no colonies were observed to have color switched when the reporter plasmid carried the R2 IR type (Figure 5B). These results align with our qPCR analyses of the switch region, strongly suggesting that the IR sequence controls the invertibility of the flagellar switch region.

Figure 5.

Flagellar IR types determine the flagellar switch invertibility

(A) Schematic of the color switch assay in E. coli.

(B) Color-switching assay when different types of flagellar IRs flank the invertible region. White arrows point to colonies that underwent a promoter inversion (i.e., switched colors). Two dilutions of the cultures were plated (left, lower dilution).

IR determines the flexibility of flagellar switch in C. difficile and impacts in vivo virulence

To directly test the ability of the flagellar IR type to alter the virulence of C. difficile in mice, we determined the impact of mutating the IR switch region of the reference RT027/ST1 strain, R20291, on its virulence using CRISPR.⁴¹ To this end, we deleted ~200 bp of the region upstream of the R20291 flgB operon first and then knocked in the wild-type C1 IR (6A/6T), as well as the variants including C2 variants 6A/5T-ON (C2-ON) and 5A/6T-OFF (C2-OFF) and R2 variants 5A/5T-ON (R2-ON) and 5A/5T-OFF (R2-OFF). Consistent with our prior data, we observed that both the R20291_C2-ON (6A/5T-ON) and R20291_R2-ON (5A/5T-ON) variants exhibited levels of flagellar motility on plates similar to those of the parental R20291 strain in swim plates 24-h after inoculation, while the R20291_C2-OFF (5A/6T-OFF) and R20291_R2-OFF (5A/5T-OFF) were largely non-motile (Figure S4A). Interestingly, a proportion of R20291_C2-OFF (5A/6T-OFF) cells became Flagella-ON, as they spread more on the swim plates 48 h after inoculation compared to R20291_R2-OFF (5A/5T-OFF), suggesting C2 is a more flexible IR type than R2 (Figure S4B).

To explore whether these R20291 flagellar IR variants have differential virulence in a mouse model, we infected antibiotic-treated mice with the mutant strains and monitored their weight loss (Figure 6A). All mutant strains colonized mice to levels similar to those of the parental R20291 strain (Figure 6B). Interestingly, fecal CFU levels were below the limit of detection for 4 out of 10 mice infected with R20291_R2-OFF (5A/5T-OFF) and 3 out of 20 mice infected with R20291_R2-ON (5A/5T-ON), implying that the ability to invert the flagellar switch region enhances the initial colonization of C. difficile (Figure 6B). To focus on the impact of the IR variants on the infection dynamics of C. difficile, we excluded these 7 mice from the downstream analyses in mice.

Figure 6.

The flagellar repeat variants and flagellar switch orientation impact the virulence of C. difficile R20291

(A) Schematic of the experimental procedure (n = 6–21 mice per strain).

(B and C) Fecal colony-forming units were measured by plating on selective agar on 1 day (B) and 14 days (C) post-infection. Data are represented as mean ± SD.

(D) Percentage weight loss relative to the baseline of mice infected with the indicated strains up to 7 days post-infection. Data are represented as mean ± SD.

(E) Percentage maximum weight loss relative to the baseline of mice infected with indicated strains up to 7 days post-infection. Boxplots show all data points with the center line representing the median.

(F) Percentage Flagellar-ON vs. -OFF cells in spore inoculum and in fecal pellets by qPCR. Data are represented as minimum to maximum, with a center line indicating the mean. Statistical significance was calculated by one-way ANOVA; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

See also Figure S4.

All colonized strains maintained high levels of C. difficile colonization and shedding up to 14 days post-infection (Figures 6C and S4C). Mice infected with the parental R20291 strain exhibited severe weight loss on day 2 post-infection, while the clinical isolate ST1–67, which carries the R2-OFF variant (5A/5T-OFF), did not induce weight loss, consistent with our previous observations (Figure 6D).³¹ Mice infected with the “restored” C1 variant, R20291_C1 (6A/6T) phenocopied the weight loss of the parental R20291 strain (Figures 6D and 6E), while the R20291 variant carrying R2-OFF, R20291_R2-OFF (5A/5T-OFF), caused minimal weight loss, essentially phenocopying the avirulence of the ST1–67 strain (Figures 6D and 6E). However, the R20291_C2-ON (6A/5T-ON) mutant virulence was equal to if not greater than that of the parental R20291_C1 (6A/6T) strain (Figures 6D and 6E). It did not completely phenocopy the clinical C2 isolate behavior in our prior work, but this is not necessarily surprising given that the flagellar switch of the C2 mutants inverts relatively frequently (Figures 6F and S4B). Additionally, we were able to observe a subpopulation of C. difficile cells flipping to the ON orientation 6 days post-infection in the R20291_C2-OFF (5A/6T-OFF) infection (Figure 6F). This flexibility likely contributes to the increased virulence in some infected mice (Figure 6E).

Since the orientation-specific qPCR revealed that C. difficile strains carrying the R2 (5A/5T) IRs remained skewed to either ~100% ON or ~100% OFF throughout the infection time course (Figure 6F), these data confirm that the R2 IR type largely prevents the flagellar switch from inverting (Figure S4B). Notably, since the R20291_R2-OFF variant caused only mild disease, and the only change between the R20291_R2-OFF variant and the parental R20291 strain is a single nucleotide in each of its flagellar switch IR sequences, these data demonstrate that the loss of flagellar gene expression at the population level reduces the virulence of C. difficile. This is consistent with the observation that the R20291_C2-ON strain is more virulent than the R20291_C2-OFF strain and that the R20291_R2-ON strain is more virulent than the R20291_R2-OFF strain (Figures 6D and 6E). While these data strongly suggest that flagellar gene expression during infection enhances the virulence of C. difficile during murine infection, we also note that the majority of mice (11/16, 69%) infected with R20291_R2-ON showed a low-virulence phenotype, causing a less than 10% weight loss (which we defined as “low virulence” for the clinical isolates; Figure S1), whereas only 6/17 (35%) of mice infected with the parental R20291 or R20291_C1 variant and none of mice infected with the R20291_C2-ON exhibited a less than 10% weight loss. These data imply that the increased invertibility of the flagellar switch, although not essential for virulence, enhances the virulence of C. difficile during murine infection. Thus, our data demonstrate that the IR types impact the invertibility of C. difficile flagellar switch during murine infection, contributing to heterogeneity in flagellar gene expression at the population level. This heterogeneity appears to promote more severe disease because greatly reducing the invertibility of the flagellar switch region by introducing the R2 IR type variant reduces disease severity in the R20291 strain background, particularly when the R2 IR variant is fixed in the OFF orientation.

Proposed model of flagellar switch inversion

RecV, a tyrosine recombinase, has been reported as necessary for the inversion of several switches in C. difficile.⁴² However, the preferred DNA sequence for RecV binding has not been identified, nor was it known how RecV avoids recombining incorrect pairs of sites (e.g., the ends of two different invertible regions). Since the general mechanism for tyrosine recombinase-mediated recombination has been well studied,^43,44 we used these prior analyses to inform an in silico model of RecV action. We first used AlphaFold2 to predict the structure of a RecV tetramer (Figure 7A). The predicted structure is highly similar to those of the catalytic domains of other tyrosine recombinases such as Cre (which recombines loxP DNA sites), including the C-terminal helix that mediates protein-protein contacts.⁴⁵ Tyrosine recombinase catalytic domains generally bind DNA in a sequence-specific manner. However, many family members include an additional N-terminal DNA-binding domain that RecV lacks. Second, we compared the IRs of all six switches whose recombination is known to be catalyzed at least in part by RecV in C. difficile R20291, taking into consideration the approximate DNA cleavage sites previously detected.⁴² We identified a conserved ~13-bp potential RecV-binding motif that is found in the IR sequence and flanks a 6-bp spacer, as expected for canonical tyrosine recombinase binding sites^43,44 (Figure 7B). In further support of this motif, previous work has reported that inversion in C. difficile R20291 is locked when the conserved nucleotides in the RIR are deleted,²³ suggesting that the activity of RecV was impaired. Searching the entire C. difficile R20291 genome for a pattern based on this, AWAGTWNCCNTTWNNNNNN-WAANGGNWACTWT (W = A or T; N = any base; 1 mismatch allowed per 10 bp) produced 17 hits, including both ends of C. difficile inversion sites (cdis) 2–5 and one end each of cdis 1 and 6 (the other ends contain similar sequences but too many mismatches) and 7 additional sites that may be false positives. In the future, more sophisticated position-weighted searches should yield more accurate results.

Figure 7.

The proposed model for RecV binding to the flagellar switch

(A) The predicted tetramer structure of RecV when binding to the IRs of the flagellar switch.

(B) The conserved sequence in the IR of six switches mediated by RecV in C. difficile R20291.

(C) The proposed spacer sequence in the flagellar switch IRs (underlined by arrows). We showing here the sequence of the upper strand. Boxed portions of the CDI4 sequences correspond to the IRs in Figure 2A.

(D) Recombination mediated by a RecV tetramer created heteroduplex DNA within the spacer (explicitly drawn sequence) and inverts the switch.

Based on these observations, we propose that the IRs that flank each invertible switch sequence each contain degenerate IRs consisting of two copies of a RecV-binding motif flanking a central 6-bp spacer region. Recombination would occur within a tetramer of RecV that pairs two such sites, with cleavage and strand exchange occurring at the outer edges of the spacers. Notably, recombinase binding is expected to be independent of the spacer sequence. However, because the spacer segments become heteroduplex in the product (one strand deriving from each partner site), recombination of the two binding sites only proceeds to completion if the spacers have identical sequences.

Indeed, further examination of the ends of the six switches suggests how the system avoids aberrant recombination between incorrectly paired repeats (Figure 7C). For each individual switch, the spacer sequences within the RecV dimer-binding sites at each end are identical (and arranged in IR). However, each switch has its own “spacer code”—that is, its own unique spacer sequence. Therefore, even if a RecV tetramer were to pair two ends of different switch regions, it would be difficult to complete recombination between them. A similar requirement has been well documented for Cre recombinase.^46,47 The length of the RecV spacer region is similar to the length of the spacer in other tyrosine recombinase systems, and we propose that variations in its length alter the efficiency of recombination (i.e., switch inversion). Efficient RecV-catalyzed recombination likely requires not only matching spacer sequences but also a particular spacer length (Figure 7D). In our collection, type C1 has the spacer sequence LIR-tacaaa(6)/RIR-tttgta(6). The other types are missing a base pair on either the left or right of the spacer region. For example, the spacer sequences of C2 and R1 are LIR-tacaa (5)/RIR-tttgta(6) (“_” implies a missing base pair) or LIR-tacaaa(6)/RIR- ttgta(5) and R2 as LIR-tacaa (5)/RIR- ttgta(5). Early studies of Cre showed that recombination was inefficient when the spacer length was changed.⁴⁸ With our model, C1 has the most efficient spacer sequence, allowing it to invert more frequently and leading to greater heterogeneity in flagellar gene expression. Because the other IR types have unequal or fewer base pairs in the spacer sequence, which may result in failure to complete recombination and a reversion to the initial substrate configuration, further research focusing on this spacer should be done to validate the model.

DISCUSSION

C. difficile infection induces a wide range of disease severity in patients, yet the molecular basis for the variation is poorly understood. Here, we applied a mouse model to study how genetic variation between C. difficile isolates impacts their virulence. Using a collection of more than 60 ST1/RT027 clinical isolates, we discovered that variation in the IR regions flanking the flagellar switch region of C. difficile regulates the invertibility of this switch by the RecV recombinase. Our data indicate that the most common variant found in ST1 strains, C1 (6A/6T), results in greater heterogeneity in flagellar gene expression within a population, with ~70% of the population carrying the flagellar switch in the ON orientation (Figures 4 and 6). We further showed that the C1 variant leads to greater disease severity relative to strains with shorter IR variants (Figure 3). The shorter IR variant types, C2, R1, and R2, restrict the RecV-mediated invertibility of the flagellar switch, effectively “locking” the flagellar switch region into either the ON or OFF orientation in C. difficile and E. coli (Figures 4 and 5). Notably, the decreased invertibility of the flagellar switch region was associated with decreased virulence in both ST1 clinical isolates (Figure 3) and CRISPR-engineered R20291 strains (Figure 6), especially when the engineered R2 switch was in the OFF orientation. Since the IR sequences are sufficient to regulate the inversion of promoter regions in E. coli (Figure 5) and C. difficile R20291 (Figure 6), our data reveal that flagellar switch region IR types control the inversion frequency of this locus (i.e., phase variation), which impacts the virulence of C. difficile ST1 strains. Our study reveals another mechanism by which RT027 strains can cause disease of varying severity, and this may help explain the wide range of disease outcomes observed in CDI patients.

The invertibility of the flagellar switch likely confers an evolutionary advantage, as more than 95% of isolates worldwide harbor C1 IR type. Interestingly, in both our strain collection and in the CDC HAI-seq C. difficile strain collection, we found that the C2-OFF IR variant accounts for approximately 30% of isolates. This could be due to rigorous testing and C. difficile cultivation in the United States,^49,50 as C2-OFF isolates are mostly non-flagellated and are associated with reduced virulence. Since flagellar synthesis and motility is costly,⁵¹ and flagella are targeted by mucosal innate immune response,^19,52 reducing flagellar motility could be energetically favorable and allow C. difficile to evade the immune system as a strategy to increase the persistence of C. difficile. However, the ability to switch between flagellar ON and OFF (and vice versa) could enhance the fitness of C. difficile during infection by allowing sub-populations of C. difficile-expressing flagellar genes to inhabit different locations within the gut or function at different stages during the infection process. For example, flagellar motility may help bring C. difficile cells closer to the host epithelium to improve toxin binding to its target cells. This could explain why C1 isolates, which have the most flexible flagellar switch, exhibit the highest virulence among the ST1 strains tested. Future investigations could focus on examining the relative spatial locations of flagella-ON vs. flagella-OFF cells for C. difficile with various IRs in situ in the colon to gain mechanistic insight into the relationship between flagellar heterogeneity and in vivo virulence.

Our R20291_C2 mutants did not fully recapitulate the clinical C2 isolates, as R20291_C2 still harbors a relatively flexible flagellar switch and demonstrates high virulence in infected mice, as compared to the clinical C2 isolates. This discrepancy suggests that there may be additional regulation or variation in clinical isolates that were selected for reduced flagellar flexibility and/or virulence. Further study on potential selective pressure resulting in C2 isolates being less heterogeneous and less virulent will help us to understand how C. difficile may leverage flagellar heterogeneity for its persistence and fitness.

Bacteria apply phase variation to reach population phenotypical heterogeneity to promote adaptation to various environmental conditions. Such heterogeneity may increase motility, antibiotic resistance, and pathogenic potentials.^53,54 Notably in C. difficile, RecV regulates multiple phase-variable regions encoding flagella as well as a cell wall protein, CwpV.^55,56 CwpV deposited on a proteinaceous layer on the cell surface of C. difficile promotes bacterial aggregation in vitro and potentially promotes intestinal colonization.⁵⁷ Moreover, C. difficile regulates its surface motility via regulation by the CmrRST system, which induces cell-chaining phenotype on surfaces and contributes to disease development in hamsters.^58,59 Here, we observed the heterogeneity of C. difficile flagella upon infection in mice, and the flagellar heterogeneity is important for C. difficile virulence. We and other researchers underscore the importance of phenotypical heterogeneity in bacterial survival and virulence and encourage more studies on such phase-variable traits of otherwise isogenic strains.

C. difficile flagellar expression has been also linked to toxin production. Turning on the expression of flagellar operon increases the expression of sigD, which encodes a sigma factor that not only further regulates flagellar synthesis and motility but also positively regulates tcdR for toxin production.²² Thus, whether flagellar gene expression contributes to virulence beyond regulating toxin levels is unclear. Here, we demonstrated that the engineered R20291 mutant with a biased ON flagellar switch orientation is significantly more virulent than its counterpart with a biased OFF flagellar switch orientation, supporting the role of flagella in C. difficile virulence. However, the increase in virulence observed when the flagellar switch is in the ON orientation did not surpass the virulence of C. difficile strains with more heterogeneous flagellar switch orientations, as was the case for both clinical isolates and CRISPR-engineered R20291 mutants. Moreover, we did not find an association between toxin levels in fecal samples and the degree of virulence in a panel of 22 isolates. These findings suggest that the coupling of flagellar gene expression to toxin gene expression contributes to virulence and is likely strain dependent, while the invertibility of the flagellar switch may provide major fitness and impact on C. difficile virulence (i.e., via impacting the locations of toxin production).

A key question that remains to be addressed is how flagellar phase variation is regulated by DNA recombinase RecV and how the environment affects the orientation of the switch during infection.^21,42 Indeed, while RecV reversibly inverts the flagellar switch between ON and OFF orientations, the proportion of ON vs. OFF cells in the resulting population is likely selected for their respective environments.⁶⁰ Since a flexible switch region generates a more heterogeneous population, with different proportions of ON vs. OFF compositions being observed in different growth conditions (e.g., vegetative broth culture, prepared spores, during mouse infection), the less flexible switches result in their more biased orientations, regardless of the conditions. Here, we proposed a RecV working model with potential recombination sites as it targets IRs for inversion. Further experiments should validate the cleavage sites and dissect out at what steps the IR variants impact RecV, such as sequence-specific DNA binding, DNA cleavage, strand exchange, or religation.

In summary, our study identifies the flagellar switch IR as a determinant of the heterogeneity C. difficile flagellar phase variation, which also introduces variable virulence outcomes within a single RT027 strain type. Since the invertibility of the flagellar switch is highly associated with the virulence of clinical isolates, we highlight the potential of using flagellar switch IRs as an easily accessible genetic trait to predict pathogen virulence. Further research on the flagellar switch regions of clinical non-ST1 strains may provide additional insights into the wide range of disease severities in patients infected with C. difficile.

Limitations of the study

We acknowledge the limitations of using mouse models of CDI as they do not fully reproduce human infections. Future analyses on large clinical datasets will help validate whether the flagellar IRs correlate with variable disease severity in infected patients.

Our survey on the flagellar IR types on C. difficile genomes from around the globe was based on the next-generation sequencing data. Since A- and T-track sequences are more prone to errors during such sequencing methods, the percentages of various IR types will likely require further validation. This limitation is likely exacerbated in analyses of geographic regions with smaller sample sizes, including Australia, east Asia, and Latin America.

We have found that increased heterogeneity promotes virulence in vivo using isogenic strains with different flagellar switch sequences and thus different frequencies of Flagella-ON. However, the mechanisms leading to this heightened virulence remain inconclusive.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Qiwen Dong (qiwendong0721@gmail.com).

Materials availability

All unique reagents and plasmids generated in this study are available from the lead contact, Qiwen Dong (qiwendong0721@gmail.com), with a completed materials transfer agreement.

Data and code availability

• RNA-seq raw data were uploaded to the NCBI Sequence Read Archive (SRA) under BioProject: PRJNA1232116.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Clinical C. difficile isolates collection

Clinical C. difficile isolates were collected from 2013 to 2017 at Memorial Sloan Kettering Cancer Center (MSKCC) from patients receiving bone marrow transplants and cancer chemotherapy.⁴⁹ The isolates were sequenced whole genome in an Illumina Hiseq platform³³ and circularized at Duchossois Family Institute by MinION Nanopore sequencing (Oxford Nanopore Technologies). Whole genome sequences are available at National Center for Biotechnology Information, BioProject PRJNA595724.

Bacterial strains and growth conditions

C. difficile isolates were grown on brain heart infusion (BHI) agar plates supplemented with yeast extract and L-cysteine (BHIS) or in BHIS broth at 37°C in an anaerobic chamber (Coylabs). Antibiotics may be supplemented as described in the detailed methods.

Mice

Wild-type female C57BL/6J mice, aged between 6 and 8 weeks, were purchased from the Jackson Laboratories. Mice housed in the BSL2 animal room are fed irradiated feed (Envigo 2918) and provided with acidified water. All mouse experiments were performed in compliance with the University of Chicago’s or Tufts University School’s institutional guidelines and were approved by its Institutional Animal Care and Use Committee.

Mammalian cell culture

Chinese hamster ovary cells (CHO/dhFr-, ATCC#CRL-9096) were grown in complete media (alpha-modified MEM supplemented with 10% FBS, 2% HEPES, 0.4% L-Glutamine, 0.007% β-mercaptoethanol, 1% Penicillin/Streptomycin/Gentamycin) at 37°C with 5% CO₂ and split every 2–3 days for subculture. HT-29 cells, a human colorectal adenocarcinoma cell line (ATCC HTB-38), were grown in DMEM supplemented with 10% FBS and 1% Pen/Strep at 37°C with 5% CO₂. Cells were split when reaching ~80% confluence. Cell lines have been tested for mycoplasma contamination.

METHOD DETAILS

Bacterial growth and spore collection

The frozen stock of C. difficile was struck onto BHIS agar plus 0.1% (w/v) sodium taurocholate hydrate (BHIS-TA) and grew overnight. Colonies were picked and sub-cultured in BHIS medium. To prepare spores for the mouse infections, C. difficile was either incubated in BHIS broth for approximately two months or on 70:30 agar for 4–5 days to encourage sporulation.⁸¹ Next, spores were separated from cell debris by gradient centrifugation using a 20%/50% (wt/vol) HistoDenz (D2158, Sigma-Aldrich) or 50% (wt/vol) sucrose gradient, then washed five times in sterile water at 14,000 × g for 5 min.⁸¹ The spore solution was further incubated at 65°C for 20 min to kill vegetative cells. Spore purity was confirmed by the absence of vegetative cell growth on BHIS plates or by microscopy.

Mouse experiment

C57BL/6 female mice from six-to eight-week-old were purchased from Jackson laboratory and housed in the specific-pathogen-free (SPF) facility at University of Chicago or the Tufts University School of Medicine. Mice were randomized and administered ad libitum with a cocktail of metronidazole (0.25 g/L), neomycin (0.25 g/L), and vancomycin (0.25 g/L) for three days. Two days after antibiotics removal, mice were intraperitoneally injected with 200μg/mouse clindamycin. After 24 h, mice were oral gavage with approximately 200 spores of C. difficile. All mice infected with C. difficile were single-housed then monitored for weight loss and clinical sign of disease for seven days. CFU/g feces counting, and toxin titer measurement was done from fecal pellet collected one day after infection as described elsewhere.³¹ Cecal toxin levels were measured by a real-time cellular analysis assay.⁸²

RNA sequencing

Mice were infected with spores from the four selected isolates (ST1–6, ST1–12, ST1–27, and ST1–53). After 24 h, mice were sacrificed to collect cecal content for RNA extraction (Qiagen RNeasy PowerMicrobiome kit). Total RNA was sent to Genewiz for rRNA removal and library preparation before sequencing. Adapters were trimmed off from the raw reads, and their quality was assessed and controlled using Trimmomatic (v.0.39),⁶⁷ then human genome was identified and removed by kneaddata (v0.7.10, https://github.com/biobakery/kneaddata), while ribosomal RNA was removed by aligned the clean non-host reads to silva database (138.1 SSURef).^68–70 The remaining reads from each sample were mapped to their corresponding circularized genome using bowtie2,⁷¹ and reads counts of each gene were obtained by running featureCounts from Subread (v2.0.1),⁷² and the core gene counts were normalized by DESeq2.⁸⁰

Sequencing the inverted-repeat sequences

64 isolates with the flagellar switch sequence resembling that of C. difficile R20291 were submitted for Sanger sequencing. Firstly, the ON and OFF sequence of each isolate was amplified using the primer pairs as designed.²¹ The PCR product was purified using QIAquick PCR purification kit (Cat. 28104, Qiagen). Purified products were sent to University of Chicago DNA Sequencing facility and sequenced using the same primer pairs for PCR reaction.

Swim plates

C. difficile was grown in BHIS broth to late-log phase then diluted to OD_600nm–0.5. To ensure a small number of C. difficile cells was added for testing, we submerged a 10 μL pipette tip into diluted C. difficile culture without drawing any liquid. The pipette tip carrying C. difficile cells on it was then stabbed into 0.3% agarose buffered with BHIS, pH 7.0. The swim plates were then incubated at 37°C. After 24 h, the swim zone size was measured and compared between isolates.

Growth curve

To ensure that only live cells were used for growth curve, we prepared a fresh culture from overnight culture at the ratio 1:100. When the back-diluted culture reached log phase, we subculture it into fresh BHI broth at a 1:100 dilution. The culture was then loaded into a 96-well plate for OD₆₀₀ measurements. The incubation time is 24 h with OD measurements every 10 min at 37°C with shaking.

Transmission electron microscope (TEM)

C. difficile was inoculated from frozen stocks onto a BHIS agar plate and incubated overnight. To collect cells for TEM, 10μL of distilled water was dropped onto a colony. After 2 min, a copper grid (CF400-Cu, EMS) was placed on top of the soaked colony so that vegetative cells were passively transferred to the grid. To stain the cells, one drop of uranyl acetate 1% was added to the grid and incubate for 30 s. TEM pictures of the stained cells were then taken at 2900X1.4 or 5900X1.4 using FEI Tecnai F30 microscope at University of Chicago Advanced Electron Microscopy Facility.

Quantitative PCR of the flagellar switch

C. difficile was grown in BHIS broth to late log phase (OD_600nm–0.5) with three to four replicates per isolate. C. difficile DNA was extracted by Qiamp PowerFecal Pro DNA kit (Cat. 51804). Fecal samples were harvested from infected mice and fecal DNA was extracted either using QIAamp PowerFecal Pro DNA Kit or as previously described.²³ Quantitative real-time PCR was done using 20–30 ng template and flagellar switch primers.²¹ C. difficile R20291 mutated strains with flagellar switch locked ON and OFF were used as the control.²³ Adenosine kinase (adk) gene is used as a reference gene. qPCR was set up using PowerUp SYBR Green Master Mix (Cat. A25742) and run by QuantStudio Real-Time PCR Systems or NEB Luna qPCR Master Mix and run by Applied Biosystems StepOnePlus system. The switch direction percentage was calculated using the ΔΔCt method as previously reported,²³ while integrating the primers’ amplification efficiencies.

Generation of C. difficile mutants using CRISPR

CRISPR editing on C. difficile strains R20291 was performed as described previously.⁴¹ Briefly, donor regions for homology were generated by separately amplifying regions ~500 bp upstream and ~500 bp downstream of the target of interest. The resulting regions were cloned into pCE677 between NotI and XhoI sites by Gibson Assembly. Geneious Prime (v11) was used to design sgRNAs targeting each deleted target. sgRNA fragments were then amplified by PCR from pCE677, using an upstream primer that introduces the altered guide and inserted at the MscI and MluI sites of the pCE677-derivative with the appropriate homology region. The regions of plasmids constructed using PCR were verified by Sanger sequencing. Plasmids were then passaged through NEBturbo E. coli strain before transformation into Bacillus subtilis strain BS49. The CRISPR-Cas9 deletion plasmids which harbor the oriT (Tn916) origin of transfer, were then introduced into C. difficile strains by conjugation.⁸³ C. difficile colonies were then screened for proper mutations in the genomes by PCR and Sanger sequencing. To generate C. difficile IR repeats mutants, two rounds of CRISPR editing were conducted. The first round was to delete ~200 bp region containing the flagellar switch and IRs while introducing an RFP landing pad (GGCGCCCAGACCGCTAAACTGAAAGTT) into the place. The second round gRNA targeted the RFP landing pad, with the mutant IR variant flagellar switch region (either in the ON or OFF orientation) template supplemented for repair. Primers used for CRISPR editing were included in Table S2.

Flagella switch alignment and IR type survey in public databases

131 ST1 C. difficile isolates from BioProject: PRJNA595724, PRJNA561087, and PRJNA594943, and 1198 ST1 isolates from four different studies^34–37 were downloaded from NCBI and assembled into contigs using SPAdes (V. 3. 13. 0).⁶³ Eleven of those isolates did not pass the assembling process. In addition, a collection of 66 ST1 C. difficile genomes from BioProject: PRJNA629351 and 15 genomes from BioProject: PRJNA551724 (N = 8) and BioProject: PRJEB31271 (N = 7) were also downloaded. MLST was determined on those contigs by mlst (Seemann T, mlst Github https://github.com/tseemann/mlst).⁸⁴ Flagellar switch region and 50 bp upstream and downstream of 68 isolates in our collection of both the ON and OFF sequences were used as query to BLAST⁶⁵ against the assembled contigs, and hits with at least 85% identity and 85% coverage of the query are considered a valid match.

E. zoli colorimetric assay

Plasmids for the E. coli experiments (see Table S3) were ordered from Twist Bioscience and were based on backbone vectors kindly onboarded with Twist by Dr. Femi Olorunniji (Liverpool John Moores University). Plasmids were checked by full-plasmid sequencing (Plasmidsaurus). pQD1 is a pBAD derivative for tightly controlled arabinose-inducible expression of RecV. Test plasmids pQD2–7 are based on pϕC31-invPB described previously,⁴⁰ with IRs from the flagellar switch replacing the att site for ϕC31 integrase.

Assays were performed as described previously with minor variations.⁸⁵ E. coli DS941⁶² was co-transformed with pQD1 plus one of the test plasmids, then after recovery grown overnight at 37C in 10mL LB supplemented with 0.2% glucose (to enhance repression of RecV), kanamycin (50 μg/mL; to maintain the test plasmid) and chloramphenicol (30 μg/mL; to maintain pQD1). In the morning, the OD₆₀₀ was ~2. 200 mL of that culture was diluted into 10mL LB plus 0.2% glucose, kanamycin (50 μg/mL) and chloramphenicol (30 μg/mL) and grown at 37C until the OD₆₀₀ reached ~0.5. They were then switched to arabinose to induce RecV expression by pelleting the cells, removing the supernatant, and resuspending in 10mL LB plus 0.2% arabinose, kanamycin (50 μg/mL) and chloramphenicol (30 μg/mL), and growing at 37C for 4 h. When cultures were plated immediately after the RecV induction period the colonies for experiments with pQD1 plus test plasmids containing the 6A/6T IRs (pQD2 or pQD3) were mostly yellow due to mixed populations of substrate and product plasmid within each founder cell (the test plasmid replicates to high copy number within each host cell).

To be separated from one another the test plasmids needed to be recovered then retransformed. After the 4 h of RecV expression in arabinose, the cultures were switched back to glucose for overnight growth: 1 mL was removed, pelleted, and resuspended in 1 mL LB supplemented with 0.2% glucose and kanamycin (50 μg/mL), then 50ul of that was used to inoculate 5 mL of LB supplemented with 0.2% glucose and kanamycin (50 μg/mL), which was grown overnight. Plasmids were recovered by miniprep. 1ul of each plasmid was used to transform competent DS941 E. coli, then two different volumes were plated on LB plus kanamycin. Colony color was visualized using a ChemiDoc imager (BioRad). The red and green channel images are overlaid in Figure 5.

Full-plasmid sequencing (Plasmidsaurus) was used to verify the recombination products. Plasmids recovered from a green colony resulting from the pQD1 + pQD2 (6A/6T – OFF/red) experiment were identical to pQD3 (6A/6T – ON/green), confirming the expected inversion. To isolate products from the experiments using pQD4 (5A/6T – OFF/red) and pQD5 (5A/6T – ON/green), the 1 or 2 product-color colonies seen in Figure 5 were picked, then restreaked to ensure separation from their substrate-containing neighbors. Sequencing confirmed that product plasmids from the pQD4 experiment matched the sequence of pQD5, and vice versa.

RNA extraction, reverse transcription and RT-qPCR

Cecal RNA was extracted using Rneasy PowerMicrobiome Kit (Qiagen) according to the manufacturer’s instructions. Complementary DNA was generated using the QuantiTect reverse transcriptase kit (Qiagen) according to the manufacturer’s instructions. Quantitative PCR was performed on complementary DNA using primers with PowerTrack SYBR Green Master Mix (Thermo Fisher) with primers listed in Table S2. Reactions were run on a QuantStudio 6 pro (Thermo Fisher). Relative abundance was normalized by ΔΔCt.

Structure prediction and conserved sequence analysis

The RecV tetramer structure was predicted by Alphafold-multimer using a web-based Colab implementation, and default parameters.^74,86,87 Sequences of the inverted repeats flanking the switch regions were located in the Clostridium difficile R20291 genome (GenBank: CP029423) using Snapgene (https://www.snapgene.com/). After manual alignment, the sequence logo shown in Figure 7B was created using https://weblogo.berkeley.edu/logo.cgi⁷⁵ An approximation of the proposed consensus sequence for binding of a RecV dimer was based on that logo (AWAGTWNCCNTTWNNNNNNWAANGGNWACTWT) and was tested by searching the full genome in Snapgene for related sequences with 1 mismatch allowed per 10bp.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analyses were performed using Prism GraphPad software v10.2.2 or R. Kruskal-Wallis, T-tests, and One-way ANOVA tests were performed to test the difference in maximum weight loss, swim plate, and RT-qPCR results. Statistical significance was determined by using a p value of <0.05. Linear regression analysis was used to estimate the correlation between IR type and weight loss. The size of each study and other statistical details can be found in figure legends.

Supplementary Material

SUPPLEMENTAL INFORMATION

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2025.115830.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
B. subtilis BS49	Wilson and Bott61	NA
C. difficile R20291	Human isolate	NCBI:NC_013316 NCBI:CP029423
C. difficile R20291_C1	This study	N/A
C. difficile R20291_C2-ON	This study	N/A
C. difficile R20291_C2-OFF	This study	N/A
C. difficile R20291_R2-ON	This study	N/A
C. difficile R20291_R2-OFF	This study	N/A
64 C difficile ST1 isolates (from Table S1)	Human isolate	BioProject: PRJNA595724
E.coli NEB5-alpha	NEB	NEB #C2987
E.coli NEBturbo	NEB	NEB #C2984
E. coli DS941	Summers and Sherratt62	N/A
Chemicals, peptides, and recombinant proteins
Metronidazole	Sigma-Aldrich	Cat# M3761
Neomycin Sulfate	Fisher Bioreagents	Cat# BP2669
Neomycin Sulfate hydrate	Thermo scientific	Cat# J61499.14
Vancomycin Hydrochloride	Hospira	UoS NDC # 00409-1319-01
Vancomycin Hydrochloride	Slate Run Pharmaceuticals	UoS NDC # 70436-021-82
Clindamycin hydrochloride	Sigma-Aldrich	Cat# C5296
Kanamycin Sulfate	Fisher Bioreagents	Cat# BP9065
Histodenz	Sigma-Aldrich	Cat# D2158
Chloramphenicol	Sigma-Aldrich	Cat# C1919
Thiamphenicol	Sigma-Aldrich	Cat# T0261
Taurocholic acid sodium salt hydrate	Sigma-Aldrich	Cat# T4009
Xylose	Sigma-Aldrich	Cat# X3877
Cefoxitin sodium salt	Sigma-Aldrich	Cat# C4786
D-Cycloserine	Sigma-Aldrich	Cat# C6880
L-Cysteine	Sigma-Aldrich	Cat# C7352
Minimum Essential Medium Eagle	Sigma-Aldrich	Cat# M8042
HEPES	Gibco	Cat# 845-1344
PEN/STREP	Gibco	Cat# 15140-122
L-Glutamine	Gibco	Cat# 810-1051
2-Mercaptoethanol	Applied Biosystems	Cat# AB1340
Gentamycin Sulfate	Gemini	Cat# 400-108
DNAseI	NEB	Cat# M0303L
Critical commercial assays
QuantiTect Reverse Transcription Kit	Qiagen	Cat# 205311
RNeasy PowerMicrobiome Kit	Qiagen	Cat# 26000
QiAamp PowerFecal pro DNA Kit	Qiagen	Cat# 51804
DNeasy PowerSoil Pro Kit	Qiagen	Cat# 47016
Gibson Assembly® Cloning Kit	NEB	Cat# E5510S
QIAquick PCR Purification Kit	Qiagen	Cat# 28104
C. DIFFICILE TOXIN/ANTITOXIN KIT	TechLab	Cat# T5000
PowerTrack SYBR Green Master Mix	Thermo Fisher	Cat# A46109
Deposited data
RNA sequence raw data	This study	BioProject:PRJNA1232116
Experimental models: Cell lines
Hamster: CHO/dhFr-	ATCC	ATCC #CRL-9096
Human: HT-29	ATCC	ATCC # HTB-38
Experimental models: Organisms/strains
Mouse: C57BL/6J	The Jackson Laboratory	JAX: 000664; RRID: IMSR_JAX:000664
Oligonucleotides
See Table S2	IDT/GENEWIZ	N/A
Recombinant DNA
Plasmid: pCE677	Kaus et al.41	N/A
Plasmid: pCE677-flgKO-CR1	This study	N/A
Plasmid: pCE677-R20291-LDCR2	This study	N/A
Plasmid: pCE677-ST1-67-LDCR2	This study	N/A
Plasmid: pCE677-ST1-67E-LDCR2	This study	N/A
Plasmid: pCE677-ST1-19-LDCR2	This study	N/A
Plasmid: pCE677-ST1-6-LDCR2	This study	N/A
Plasmid: pQD1-pQD7 (See Table S3)	This study	N/A
Software and algorithms
SPAdes	Prjibelski et al.63	N/A
GraphPad Prism v.10	GraphPad Software	N/A
Geneious Prime v.11	Geneious by Dotmatics	N/A
PATRIC web resources	Wattam et al.64	N/A
BLAST	Camacho et al.65	N/A
mlst	Jolley and Maiden66	N/A
Trimmomatic v.0.39	Bolger et al.67	N/A
KneadData v.0.7.10	https://github.com/biobakery/kneaddata	N/A
Silva database (138.1 SSURef)	Quast et al.68; Glöckner et al.69; Yilmaz et al.70	N/A
Bowtie 2	Langmead and Salzberg71	N/A
Subread v2.0.1	Liao et al.72	N/A
seqtk	https://github.com/lh3/seqtk 73	N/A
AlphaFold-Multimer	Evans et al.74	N/A
SnapGene software	www.snapgene.com	N/A
WebLogo 3	Crooks et al.75; Schneider and Stephens76	N/A
R	R Core Team (2022). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.	N/A
R package: tidyverse	Wickham et al.77	N/A
R package: ComplexHeatmap	Gu; Gu et al.78,79	N/A
R package: ggplot2	Wickham et al.77	N/A
R package: DESeq2	Love et al.80	N/A

Highlights.

• C. difficile RT027 strains exhibit SNPs in their flagellar switch region

• Shorter inverted repeats in the switch region reduce its invertibility by RecV

• Reduced heterogeneity in flagellar gene expression reduces C. difficile virulence

• SNPs in C. difficile’s flagellar switch region lead to variable disease severity