Phenotypic analysis of various Clostridioides difficile ribotypes reveals consistency among core processes

ABSTRACT

Clostridioides difficile infections (CDI) cause almost 300,000 hospitalizations per year, of which ~15%–30% are the result of recurring infections. The prevalence and persistence of CDI in hospital settings have resulted in an extensive collection of C. difficile clinical isolates and their classification, typically by ribotype. While much of the current literature focuses on one or two prominent epidemic ribotypes (e.g., RT027), recent years have seen several other ribotypes dominate the clinical landscape (e.g., RT106 and RT078). Some ribotypes are associated with severe disease and/or increased recurrence rates, but why certain ribotypes are more prominent or harmful than others remains unknown. Because C. difficile has a large, open pan-genome, this observed relationship between ribotype and clinical outcome could be a result of the genetic diversity of C. difficile. Thus, we hypothesize that the core biological processes of C. difficile are conserved across ribotypes/clades. We tested this hypothesis by observing the growth kinetics, sporulation, germination, production of toxin A and toxin B, bile acid sensitivity, bile salt hydrolase activity, and surface motility of 15 strains belonging to various ribotypes spanning each known C. difficile clade. In viewing these phenotypes across each strain, we see that core phenotypes (growth, germination, sporulation, and resistance to bile salt toxicity) are remarkably consistent across clades/ribotypes. This suggests that variations observed in the clinical setting may be due to unidentified factors in the accessory genome or due to unknown host factors.

IMPORTANCE

Clostridioides difficile infections impact thousands of individuals every year, many of whom experience recurring infections. Clinical studies have reported an unexplained correlation between some clades/ribotypes of C. difficile and disease severity/recurrence. Here, we demonstrate that C. difficile strains across major clades/ribotypes are consistent in their core phenotypes. This suggests that such phenotypes are not responsible for variations in disease severity/recurrence and are ideal targets for the development of therapeutics meant to treat C. difficile-related infections.

INTRODUCTION

Clostridioides difficile is a gram-positive, anaerobic, endospore-forming pathogen with two major life stages: the metabolically active vegetative cell and the dormant spore (1, 2). The spore is the transmissible form and provides extreme resistance to antibiotics, environmental stresses, and disinfection techniques (2–4). When a patient experiencing antibiotic-induced dysbiosis ingests C. difficile spores, the spores traverse the gut and germinate into the vegetative form (5–7). In the absence of a healthy microbiome, vegetative cells more efficiently colonize the host and can produce toxins that induce the symptoms characteristic of C. difficile infection (CDI) (diarrhea, pseudomembranous colitis, etc.) (8). These vegetative cells will form new spores which can remain in the gut or pass into the environment contributing to the spread/recurrence of disease (6). Of approximately 300,000 CDI-related hospitalizations in the US, about 15%–30% of patients experienced disease recurrence; the chance of recurrence increases with each subsequent infection (9, 10).

The primary virulence factors in CDI are two large exotoxins: TcdA (toxin A) and TcdB (toxin B) (8). These toxins are encoded within a 19.6 kb pathogenicity locus (PaLoc) that includes a positive (tcdR) and negative (tcdC) regulator of expression and a holin-like protein (tcdE) (11). Diagnosis of CDI is typically confirmed by testing fecal samples of patients presenting with CDI symptoms for the presence of one or both C. difficile toxins (12). The presence of toxins is known to be variable between species and is often used to characterize C. difficile strains (e.g., toxinotyping) (13).

In 2006, the first fully assembled C. difficile genome was published (14). Currently, there are greater than 19,000 C. difficile genomes deposited in the NCBI database and greater than 31,000 deposited in Enterobase (15, 16). Broadly, C. difficile genomes are ~4 Mbp in size and contain ~4 k genes (14, 17–19). About 25%–50% of these genes belong to the core genome (genes shared among sequenced strains), while the pan-genome (all genes both shared and unique found in sequenced strains) is generally agreed to consist of ~6 k genes with no currently defined limit (20–22). In addition, greater than 11% of any given C. difficile genome consists of mobile elements (transposons, prophage, a prophage-like sigK intervening [skin] element, etc.) (14). Phylogenetic analyses group C. difficile strains into five main clades and three cryptic clades (23, 24). Most studies agree that clades 1 and 2 are closely related, while clade 5 exhibits the most genetic distinction, with recent studies suggesting that clade 5 is undergoing speciation (22, 25, 26). The relatively small size of the core genome, open pan-genome, large number of mobile elements, and observed evolutionary distance all indicate that C. difficile, as a species, has substantial genetic variation between its members.

In the clinical setting, C. difficile strains are typically classified using various typing methods (ribotyping, restriction endonuclease analysis, multilocus sequence typing [MLST], toxinotyping, serotyping, etc.) (13, 27–31). Of these, ribotyping remains the most popular and clinically relevant. Historically, ribotype 027 (RT027) strains have been associated with the worst CDI outbreaks and thus are generally the most well studied (18, 19, 32, 33). However, the prominent strains isolated from more recent outbreaks belong to less well-studied ribotypes (e.g., RT106 or RT078) (34–42). Because C. difficile is so genetically diverse, ribotype-specific clinical outcomes may be attributable to processes encoded by the accessory genome and not to functions encoded in the core genome. Thus, we hypothesize that processes central to C. difficile biology (e.g., vegetative growth, sporulation, germination, and resistance to bile salt toxicity) are conserved across ribotype/clade. To test this hypothesis, we collected strains from ribotypes belonging to each of the five classical clades. We then measured their growth in rich and minimal media, spore production, response to germinants, production of toxin A and toxin B, bile salt hydrolase activity (BSHA), resistance to bile acid toxicity, and surface motility. Our analyses indicate that, while the strains show variations in their ability to process taurocholate (TA)-conjugated bile salts and in their ability to metabolize various carbohydrates, they exhibit remarkably consistent core phenotypes with no strong patterns across ribotypes/clades. These results support the hypothesis that despite the evolutionary variation observed between C. difficile strains, certain processes that are central to C. difficile biology remain consistent across ribotype/clade.

MATERIALS AND METHODS

Bacterial strains and growth conditions

C. difficile R20291 was used as an experimental control for all assays performed in this study. C. difficile strains LC5624, LK3P-030, and LK3P-081 (all identified as ST-42 by in silico MLST and presumed to be RT106) were obtained from pediatric patients receiving care at the Ann and Rober H. Lurie Children’s Hospital of Chicago, Chicago, IL, USA. C. difficile strains M68 and M120 were obtained from the collection of Dr. Trevor Lawley (Sanger Institute, Hinxton, UK). The remaining strains were collected by Dr. Daniel Parades-Sabja (Texas A&M University, College Station, TX, USA) and César Rodríguez (Facultad de Microbiología and Centro de Investigación en Enfermedades Tropicales, Universidad de Costa Rica, San José, Costa Rica) from various outbreaks in Central/South America.

Strains were routinely grown anaerobically (Coy Laboratories, model B, atmospheric H₂ was kept between 3% and 4% using 10% H₂, 5% CO₂, and 85% N₂ gas mix) at 37°C. Each strain was grown on/in either brain heart infusion medium (Becton, Dickinson and Company, Franklin Lakes, NJ, USA) supplemented with yeast extract and 0.1% (wt/vol) L-cysteine (BHIS) and, where indicated, with 11 mM theophylline, 10 µg/mL thiamphenicol, 50 µg/mL kanamycin, and/or 250 µg/mL D-cycloserine. For some growth assays, strains were grown in C. difficile minimal medium (CDMM; 1% [wt/vol] casamino acids, 2 mM L-tryptophan, 4 mM L-cysteine, 35.2 mM Na₂HPO₄·7H₂O, 60 mM NaHCO₃, 7 mM KH₂PO₄, 20 mM NaCl, 45 mM glucose [or indicated carbohydrate], 400 µM [NH₄]₂SO₄, 177 µM CaCl₂·2H₂O, 96 µM MgCl₂·6H₂O, 50.6 µM MnCl₂·4H₂O, 4.2 µM CoCl₂·6H₂O, 34 µM FeSO₄·7H₂O, 4 µM D-biotin, 2 µM calcium-D-pantothenate, and 6 µM pyridoxine). Where indicated, the pH of BHIS was adjusted using acetic acid/sodium hydroxide or supplemented with 0.1% TA (GoldBio S-121-100) in a solid medium and 1 mM TA, taurodeoxycholic acid (TDCA; Millipore Sigma 580221-50GM), hyodeoxycholic acid (HDCA; MP Biomedicals 157514), cholic acid (CA; Sigma Aldrich C1129-100G), chenodeoxycholic acid (CDCA; ACROS Organics RK-88245-39), or deoxycholic acid (DCA; Sigma Aldrich D2510-100G) for liquid cultures. Except for TA, which is soluble in water, bile acids stocks were made in dimethylsulfoxide (DMSO).

Plasmid and strain construction

Theophylline-mediated allelic exchange plasmids were constructed as described previously (43). Briefly, 500 bp regions up and downstream of the indicated genes were amplified via PCR using primers 3,814–3,817 for pMAB50 and primers 4,133–4,136 for pMAB55. The PCR products were inserted via Gibson assembly into pJB94 between the NotI and XhoI restriction sites. The resulting plasmids were transformed into Escherichia coli DH5α.

The allelic exchange plasmids were then transformed into E. coli HB101 for conjugation into C. difficile. For each conjugation, 1 mL of the log-phase culture of E. coli HB101 harboring the appropriate plasmid was pelleted and mixed with 500 µL of a log-phase C. difficile culture. The mixture was spotted onto a pre-reduced BHIS plate supplemented with theophylline (Theo) and incubated overnight. The subsequent growth was suspended in 500 µL BHIS and spread across five pre-reduced, BHIS Theo plates supplemented with thiamphenicol, kanamycin, and D-cycloserine. Growth was monitored for 4 days, and colonies that appeared were tested for the presence of the plasmid using primers 173 and 174 for catP. Following confirmation of a successful conjugation, plasmid integration was selected for by passaging on BHIS supplemented with thiamphenicol and confirmed via PCR using primers 2,993 and 3,819 for pMAB50 and primers 2,980 and 3,819 for pMAB55. Colonies that harbored integrated plasmids were then plated on BHIS Theo to promote the excision of the plasmid from the genome. Mutations were screened via PCR using primers 2,980 and 2,981 for the C. difficile ΔsigG strain, which was then cured of the allelic exchange plasmid via repeated passages on BHIS (no antibiotic) medium. Using the same method, a tcdR deletion was introduced (and confirmed using primers 2,992 and 2,993) to generate the C. difficile ΔsigGΔtcdR strain.

Whole-genome sequencing

Genomic DNA was extracted for all strains except C. difficile R20291 (GenBank accession number NZ_CP029423.1), C. difficile LC5624 (GenBank accession number CP022524.1), C. difficile M68 (GenBank accession number NC_017175.1), and C. difficile M120 (GenBank accession number NC_017174.1) for which genomes are already published. Briefly, each strain was grown for 18 hours in a BHIS medium. The cells were pelleted, lysed, and subjected to phenol-chloroform extraction. This genomic DNA was sent to SeqCoast Genomics (Portsmouth, NH, USA) for long-read Oxford Nanopore sequencing. The long-read data were assembled using the Flye plugin for de novo assembly in Geneious Prime (version 2024.0.5) (44, 45). To improve genome quality, Illumina sequencing reads for each strain were mapped to the assembled genomes using BBMapper (46). The genome alignments were performed using the MAUVE plugin (47). Phylogenies were generated using the Geneious Tree Builder with a Tamura-Nei genetic distance model and the neighbor-joining method for three individual locally collinear blocks (LCBs) produced by the MAUVE alignment.

Core genome analysis

The core genome (defined here as the genes present in all strains under study) was determined using the finalized genome for all strains except C. difficile R20291 and Spine (version 0.3.1) (48). For each strain, the resulting files were concatenated in Geneious Prime and annotated against C. difficile CD630 (GenBank accession number NC_009089) in keeping with the annotation of the C. difficile core genome multilocus sequence typing scheme (20). The results of this analysis can be found in Table S1.

Protein alignment

The genes encoding each of the indicated proteins were extracted from the assembled and MAUVE-aligned genomes. Their sequences were translated and aligned using Clustal Omega in Geneious Prime (49). The amino acids were shaded according to similarity (Pam140 Score Matrix) with white indicating 100% amino acid identity across all observed strains and black indicating <60% identity.

Growth curves and doubling times

For growth curves in a BHIS medium, strains were grown in 5 mL pre-reduced BHIS liquid medium for 12 hours. Cultures were back diluted to an OD₆₀₀ = 0.05 in 5 mL and allowed to grow to an OD₆₀₀ = 0.5. This log-phase culture was then used to inoculate the experimental culture (40 mL) to an OD₆₀₀ of 0.05 at T₀. OD₆₀₀ measurements were taken every 30 minutes for 8 hours using a Biowave Cell Density Meter CO8000. When the OD₆₀₀ reached ≥0.3, the OD measurements across 2 hours of growth were used to calculate doubling time.

For growth curves in CDMM, strains were grown in 5 mL pre-reduced CDMM (made with a final concentration of either 45 mM glucose/fructose, 53 mM xylose, or 50 mM trehalose, where indicated) for 24 hours. Cultures were back diluted in 5 mL to an OD₆₀₀ = 0.05 and grown for 12 hours to an OD₆₀₀ ≈ 0.5–0.7. The culture was then used to inoculate 3 mL of the indicated medium to a starting OD₆₀₀ of 0.05 in a 12-well plate. The plate was then placed in a Stratus Microplate Reader (Cerillo, Charlottesville, VA, USA) (located in the anaerobic chamber) where OD₆₀₀ measurements were collected every 3 minutes for 22 hours. Each time point represents the average reading from three different sensors measuring a single well. Due to random blockage of sensors during the assay, the data were filtered using the z-score method. Doubling times were calculated using the most linear portion of the OD curves and the formula t₂ = ln(2)/(k), where k is the growth rate determined using the formula k = LN(OD_t2/OD_t1)/(t2 − t1).

Spore purification

Spores were purified as previously described (50–57). Briefly, the indicated strains were plated onto 10 pre-reduced BHIS plates and incubated for 5 days. The growth from each plate was scraped into 1 mL 18 MΩ dH₂O and stored at 4°C. After a minimum period of 24 hours, cells were resuspended in 1 mL 18 MΩ dH₂O and centrifuged for 1 minute at 14,000 × g. The supernatant was removed, and subsequent washes with 18 MΩ dH₂O were performed to separate the spores from vegetative cells/cell debris in distinct layers that were gradually removed, and the cell pellets were combined with each wash. Final contaminants were removed by placing the cells on 9 mL of 50% sucrose and centrifuging for 20 minutes at 4,000 × g and 4°C. The supernatant was discarded, and the pellet (containing the purified spores) was washed thrice with 18 MΩ dH₂O to remove the remaining sucrose. The final pellet was resuspended in 1 mL 18 MΩ dH₂O and stored at 4°C.

Sporulation

The sporulation assay was performed as previously described, with some slight modifications (3, 58). A 16-hour culture of the indicated strain was back diluted in 5 mL to an OD₆₀₀ = 0.05 and allowed to grow to an OD₆₀₀ = 0.5. One hundred microliter of the log-phase culture was plated on pre-reduced BHIS agar medium and incubated for 48 hours. One-quarter of the plate was scraped into 1 mL of 1× phosphate-buffered saline (PBS, pH 7.4), and 500 µL of this suspension was treated with 100% ethanol for 20 minutes. The treated cells were serially diluted in 1× PBS with 0.1% TA and plated on pre-reduced BHIS TA plates. The plates were incubated for 48 hours prior to CFU enumeration.

Germination assay

Germination was assessed using an OD-based assay as described previously (51, 52). Briefly, samples of purified spores were adjusted to an OD₆₀₀ of 0.5. The spores were heat treated at 65°C for 30 minutes. For each of the tested germinant concentrations, 5 µL of the spore sample was suspended in 95 µL of the appropriate germination buffer. For all samples, the buffer contained 0.5 M HEPES and 50 mM NaCl, pH 7.2. For co-germinant (glycine) efficiency, the buffer contained 10 mM TA and various amounts of glycine (final concentrations of 0 mM, 0.1 mM, 0.2 mM, 0.4 mM, 0.6 mM, 0.8 mM, 1 mM, 2 mM, or 5 mM). For bile-acid germinant (TA) efficiency, the germination buffer contained 100 mM glycine, 10% (vol/vol) DMSO (to control for the DMSO used to dissolve CDCA), and various amounts of TA (final concentrations of 0 mM, 0.1 mM, 0.2 mM, 0.5 mM, 1 mM, 2 mM, 5 mM, or 10 mM). For CDCA, the germination buffer was supplemented with 100 mM glycine and 1 mM CDCA and TA to a final concentration of 1 mM, 2 mM, 5 mM, 10 mM, 15 mM, 20 mM, 30 mM, or 50 mM. The OD₆₀₀ of each sample was recorded every 30 seconds over the course of 1 hour using a SpectraMax M3 spectrophotometer (Molecular Devices, San Jose, CA, USA). The plate was shaken vigorously 5 seconds before each measurement.

A germination curve was generated for each strain/germinant combination by plotting the OD₆₀₀ at a given time (T_x) divided by the OD₆₀₀ at time zero (T₀) vs time (the controls for all strains/germinants and an example for R20291 with varying concentration of glycine are shown in Fig. S1) (51, 52, 59–62). Germinant sensitivity was calculated using the maximum slope for each germination curve. The slope was plotted against (co)-germinant concentration to generate a Michaelis-Menten graph. A Lineweaver-Burke plot was generated, and from this, the Ki/EC₅₀ was calculated. As this is a multi-enzyme/protein process, this does not provide a bona fide K_m. Here, EC₅₀ is defined as the concentration of germinant that produces half the maximum germination rate. This value is used here and in many prior studies to provide a measurement for spore interactions with germinants (51, 52, 59–65). The efficiency of the competitive inhibitor was calculated as previously described using the following equation K_i = (inhibitor)/([K_CDCA/K_TA] − 1) (59–62).

ELISAs

Each strain (including C. difficile R20291 and the C. difficile R20291 ΔsigGΔtcdR negative control) was grown in 5 mL pre-reduced BHIS broth for ~4 hours until log-phase growth was reached. Each culture was then back diluted in 5 mL BHIS to an OD₆₀₀ = 0.05 and incubated for 48 hours. At 24 and 48 hours, 250 µL samples of culture were removed from the anaerobic chamber and stored at −80°C. Samples were thawed at room temperature and combined with 1 mL of the 1× fecal extraction buffer from the Fecal C. difficile Toxin A (KT-834) or Toxin B (KT-835) enzyme-linked immunosorbent assay (ELISA) kit (Epitope Diagnostics San Diego, CA, USA). The samples were centrifuged for 5 minutes at 900 × g and 100 µL of the clarified sample was added to a toxin A or toxin B antibody-coated microplate in duplicate. The plate was incubated for 1 hour, static, at room temperature. Each well was washed five times with 350 µL of 1× ELISA wash solution prior to treatment with 100 µL of the tracer antibody. The plate was then incubated for 30 minutes, static, at room temperature. The wash was repeated five more times, and 100 µL of the ELISA HRP substrate was added to each well. The plate was incubated for 20 minutes, static, at room temperature prior to the addition of 100 µL of ELISA stop solution. The absorbance of each well was read at 450 nm using a SpectraMax M3 spectrophotometer. The average absorbance value of the technical triplicates for each strain was normalized to the absorbance value of the C. difficile R20291 ΔsigGΔtcdR negative control strain.

Bile salt sensitivity

Each strain was grown in 5 mL pre-reduced BHIS broth for 16 hours, back diluted in 5 mL to an OD₆₀₀ = 0.05, and allowed to grow to an OD₆₀₀ = 0.5. Fifty microliter of these cultures was added to pre-reduced BHIS broth of the indicated pH and bile acid concentration (obtained through a series of 1:1 dilutions to a final volume of 500 µL). The samples were incubated for ~18 hours, and MICs were assessed by the presence/absence of growth as indicated by a visual inspection of culture turbidity.

Bile salt hydrolase activity

Bile salt hydrolase activity assays were performed as described previously (54). Briefly, strains were grown in 5 mL BHIS liquid medium for 16 hours. A total of 10⁸ CFU from these cultures were transferred into 5 mL BHIS supplemented with 1 mM TA or TDCA and incubated for 24 hours after which the cultures were centrifuged for 10 min at 4,000 × g. The pellet was suspended in 100% methanol. One millimolar HDCA (an internal standard) was added to the supernatant before being lyophilized and suspended in methanol. The suspended pellet and dried supernatant were combined. Each strain was run alongside three C. difficile R20291 controls: a negative control without TA, a control with 1 mM each TA/TDCA, CA/DCA, and HDCA all added to the spent supernatant, and a positive control with 1 mM TA added before the 24-hour incubation (Fig. S2).

The bile salts found in each sample were separated by reverse-phase high-performance liquid chromatography using a Shimadzu Prominence system (Shimadzu, Kyoto, Japan) (54, 66–68). For each strain, 30 µL of the sample was separated on a Synchronis C18 column (4.6 by 250 mm, 5 µm particle size, ThermoFisher, Waltham, MA, USA) using a methanol-based mobile phase (53% methanol, 24% acetonitrile, and 30 mM ammonium acetate [pH 5.6]). A Sedere Sedex model 80 low temperature-evaporative light scattering detector using 50 psi Zero Grade air at 94°C detected the bile salt peaks. Percent deconjugation was calculated using the area under the peak for CA/DCA divided by the sum of the areas under the peaks for TA/TDCA and CA/DCA.

Surface motility

Surface motility assays were performed as previously described (69). Briefly, strains were grown in 5 mL BHIS for 16 hours, diluted 1:50 in 5 mL fresh BHIS medium, and grown until the OD₆₀₀ reached ~0.5. From these cultures, 10 µL was spotted onto pre-reduced BHIS 1.5% agar medium and incubated for 5 days. Plates were then imaged using a Bio-Rad GelDoc XR+ (Bio-Rad Laboratories, Hercules, CA, USA), and the growth diameters were quantified and normalized to C. difficile R20291.

Statistical analysis

All data represent the average from three independent biological replicates with the error bars indicating the SEM. For all assays except toxin production, statistical significance was determined using the one-way analysis of variance (ANOVA) analysis function from GraphPad Prism (version 9.0.2 for Windows, GraphPad Software, San Diego, California USA). When results were compared to C. difficile R20291, Šidák’s multiple comparisons test was used, while Tukey’s multiple comparisons test was used when comparing between ribotypes/clades. For the toxin A or B ELISA results, a two-way repeated measures ANOVA analysis was performed using the analysis function in GraphPad Prism. Asterisks indicate P-values with, * ≤0.05, ** ≤0.02, *** ≤0.01, and **** ≤0.0001.

RESULTS

Strain collection and genomic analysis

To study phenotypic variation among the C. difficile clades, we collected 15 different clinical isolates of C. difficile (Table 1). C. difficile R20291 is a well-characterized member of clade 2 and was used as a control throughout (3, 43, 51–57, 66, 70–80). Each of the five main clades is represented in this study by strains belonging to six prominent ribotypes. The indicated ribotype was selected based on its clinical relevance (17, 34, 41, 81–83). Apart from those strains whose genomes have already been published (C. difficile R20291, C. difficile LC5624, C. difficile M68, and C. difficile M120), we assembled the genomes for each of the strains and deposited them in the NCBI database. Phylogenetic analysis clustered the strains into the five classical clades, consistent with previously published data (Fig. 1; Fig. S3) (22, 25, 26).

TABLE 1.

Strains used in this study

Strain	Clade	Ribotype
R20291	2	027
PUC_256	1A	014-020
HC52	1A	014-020
PUC_90	1A	014-020
LC5624	1B	106
LK3P-030	1B	106
LK3P-081	1B	106
PUC_75	3	023
S9	3	023
C103	3	023
M68	4	017
PUC_606	4	017
ICC5	4	017
M120	5	078
P8_S146	5	078
P12_S145	5	078

Fig 1.

Phylogeny of strains used in this study. The neighbor-joining phylogeny generated for the strains in this study derived from LCB 60, a 764,748 bp segment of the MAUVE alignment representing approximately 25% of any given genome in the study. The phylogeny was created using the Geneious Tree Builder application in the Geneious Prime software using the Tamura-Nei genetic distance model. Strains are grouped by their respective ribotypes/clades, with the scale bar representing the number of substitutions per 1,000 bp.

C. difficile growth is consistent in rich medium

In the laboratory setting, C. difficile is often cultured in a rich medium (BHIS). Because this was the medium in which all our assays would be performed, we first sought to determine if there were any inherent growth differences in this medium. Growth curves in a BHIS medium were obtained for each strain and indicated little variation in overall growth kinetics. We observed minimal differences between each strain in a given clade (Fig. 2A through E). All strains reached the stationary phase within 4 hours of growth and a maximum OD₆₀₀ of 2.0–3.0. Moreover, there were no significant differences in growth when the data were grouped by clade (Fig. 2F). For a more objective comparison, we determined the generation times for each strain. The generation times for individual strains, including C. difficile R20291 (clade 2), were calculated at 40–60 minutes (Fig. 2G). The average generation time for each clade was ~50 minutes (Fig. 2H). Taken together, this indicates that none of the tested strains exhibited a notable change in growth rate in the BHIS medium relative to each other.

Fig 2.

Growth of strains in rich medium. OD₆₀₀ measurements for clade 1A (A), clade 1B (B), clade 3 (C), clade 4 (D), and clade 5 (E) strains were taken every 30 minutes over the course of 8 hours. These same data are shown in panel F grouped by ribotype/clade. Data from the most linear portion of the growth curve were used to calculate the doubling time presented by strain (G) and by ribotype/clade (H). Data points represent the average from independent biological triplicates with error bars representing the SEM. For panel G, Šidák’s multiple comparisons test (comparing all strains to C. difficile R20291) was used, while Tukey’s multiple comparisons test (comparing all clades to each other) was used for panel H. No statistically significant differences between strains were found.

Clade 4 strains are limited in their ability to use various carbohydrates

Given that BHIS is a rich medium, the lack of variation in growth between strains was not entirely surprising. To determine if the strains responded differently in minimal medium, we cultured them in standard CDMM (containing glucose). Under these conditions, all strains exhibited less growth compared to growth in BHIS medium as indicated by noticeably increased generation times (Fig. 3A through C). While the C. difficile R20291 strain (clade 2) had a generation time of 100 minutes, C. difficile strains LK3P-081 (clade 1B), C103 (clade 3), M68 (clade 4), PUC_606 (clade 4), ICC5 (clade 4), M120 (clade 5), and P12 (clade 5) had slower generation times. The clade 4 strain, C. difficile M68, had the slowest generation time at ~400 minutes (Fig. 3B). When these data are grouped by clade, the clade 1A strains had a generation time of ~90 minutes, while the clade 4 strains had a generation time of ~350 minutes (Fig. 3C). Taken together, the data indicate that clade 4 strains tested demonstrate lower growth rates in standard CDMM relative to the other strains.

Fig 3.

Growth of strains in minimal medium. Strains were grown in CDMM supplemented with either glucose (A), xylose (D), fructose (G), or trehalose (J). OD₆₀₀ measurements for each strain were taken every 3 minutes for 22 hours. Data from the most linear portion of the growth curve were used to calculate generation times for growth in CDMM supplemented with glucose (B and C), xylose (E and F), fructose (H and I), or trehalose (K and L). Data points represent the average from independent biological triplicates with error bars representing the SEM. For panels B, E, H, and K, Šidák’s multiple comparisons test (comparing all strains to C. difficile R20291) was used, while Tukey’s multiple comparisons test (comparing all clades to each other) was used for panels C, F, I, and L. Asterisks indicate P-values of * ≤0.05, ** ≤0.02, *** ≤0.01, and **** ≤0.0001.

Because C. difficile can incorporate different carbohydrates into its metabolic pathways, we sought to determine if any of the strains favored one carbohydrate source over another. Previous research has identified xylose as an important carbon source for many bacteria (84, 85). Xylose forms a five-carbon, six-member ring like glucose, but it lacks a 6’ carbon and associated hydroxyl group. When testing growth in CDMM supplemented with xylose (CDMM-xyl), we observed slight variations between strains. All strains reached the stationary phase between 6 and 10 hours (Fig. 3D). Contrary to what was observed for growth in standard CDMM, generation times for all strains except C. difficile R20291 (246 minutes), C. difficile M68 (240 minutes), and C. difficile PUC_606 (220 minutes) were below 200 minutes (Fig. 3E). Additionally, while most of the strains grew better than the control in CDMM-xyl, clade 4 strains, once again, appeared to have a decreased growth rate in this medium in comparison to the other clades with an average doubling time of ~210 minutes (Fig. 3F).

Fructose, another common carbohydrate that is important for C. difficile metabolism, is the primary carbohydrate found in a medium that allows for selection/isolation of C. difficile in clinical environments (TCCFA) (86, 87). Like glucose, it contains six carbons but forms a five-member ring. When grown in CDMM supplemented with fructose, the strains reached a stationary phase between 8 and 12 hours (Fig. 3G). C. difficile strains LK3P-030 (clade 1B), PUC_75 (clade 3), and ICC5 (clade 4), as well as all clade 1A and five strains, exhibited generation times of 100–130 minutes. These times were lower than the C. difficile R20291 strain (clade 2), which had a 177-minute generation time (Fig. 3H). C. difficile strains LC5624 (clade 1B), LK3P-081 (clade 1B), S9 (clade 3), C103 (clade 3), M68 (clade 4), and PUC_606 (clade 4) all grew similarly compared to C. difficile R20291 with generation times between 130 and 175 minutes. When grouped by clade, strains from clades 1B, 3, and 4 showed a ~50-minute slower generation time relative to the remaining strains (Fig. 3I).

Finally, we tested the growth of the strains in CDMM supplemented with 50 mM trehalose (CDMM-tre), a concentration previously shown to support the growth of strains spanning multiple ribotypes (25). Trehalose is composed of two α,α−1,1-linked glucose molecules and has, arguably, been associated with an increase in CDI (25, 88, 89). Growth in CDMM-tre was variable, with clade 1B strains reaching the stationary phase around 10 hours of growth and clade 1A strains reaching the stationary phase around 20 hours of growth (Fig. 3J). All strains, except for C. difficile M120 (660 minutes), had a lower generation time than the C. difficile R20291 strain (clade 2, 620 minutes; Fig. 3K). Additionally, generation times for C. difficile strains PUC_256 (clade 1A), HC52 (clade 1A), PUC_90 (clade 1A), LC5624 (clade 1B), LK3P-030 (clade 1B), LK3P-081 (clade 1B), PUC_75 (clade 3), S9 (clade 3), C103 (clade 3), ICC5 (clade 4), P8 (clade 5), and P12 (clade 5) were similar to the generation times observed in CDMM-xyl at ~200–300 minutes. Overall, clade 4 strains had consistently slower growth in CDMM-tre relative to the other clades (Fig. 3L).

Variations in carbohydrate metabolism protein sequences are not consistent with phenotypic differences

The catabolite control protein, CcpA, is encoded by all strains in this study (Fig. S5). Because there was little variation in the CcpA protein sequence, we hypothesize that there should be no major change in the global regulation of carbohydrate metabolism. In all tested strains, the xylose utilization operon was also present. Some variants appeared in the XylA (isomerase), XylB (kinase), and XylR (transcriptional regulator) protein sequences (Fig. S6 to S8). However, none of these variations were consistent between the clades 1A, 1B, 3, and 5 strains, which had faster generation times compared to the C. difficile R20291 strain (Fig. 3D and F).

Previous reports found heterogeneity in the genes responsible for processing trehalose and corresponding differences in the ability of strains from different ribotypes to grow in media in which trehalose was the sole carbon source (25, 89). We observed the same for the strains tested here, specifically when comparing the phosphotrehalase (TreA) protein sequence (Fig. S9). The clade 3 C. difficile strains PUC_75 and S9 had nonsense mutations in TreA, and C. difficile strains C103 (clade 3), M120 (clade 5), P8 (clade 5), and P12 (clade 5) were missing TreA at this locus entirely. The trehalose operon repressor (TreR) also showed some variations and was missing entirely at the canonical locus in the clades 3 and 5 strains (Fig. S10). Both the treR and treA genes were found in an operon elsewhere in the genomes for the clade 5 strains, and these copies are included in the alignments (Fig. S9 and S10). Clade 3 strain, C. difficile C103, contained an intact copy of treA at an additional locus but did not possess a corresponding copy of treR, suggesting that treA may be regulated differently in this strain. The remaining clade 3 strains (C. difficile PUC_75 and S9) do not possess an intact copy of treA, suggesting the existence of another, non-canonical, trehalose metabolism pathway in these strains.

Spore production differences are minor between strains

Given that C. difficile is transmitted by spores, changes in sporulation levels could explain why some strains are more prevalent in the clinical setting than others; an increase in spore number may lead to increased disease spread/recurrence. We measured sporulation in each of the strains over a 48-hour period (Fig. 4A through E) and observed a small increase in spore number for the clade 1A strain C. difficile HC52 (clade 1A, ~10⁹ spores) and a decrease for the clade 1B strain C. difficile LK3P-030 (~10⁷ spores), relative to the C. difficile R20291 strain (clade 2, ~10⁸ spores). When these data were grouped by clade, the clade 1A strains produced ~10-fold more spores on average than the clade 4 strains (Fig. 4F).

Fig 4.

Spore production by strains over 48 hours. The number of spores produced on BHIS over 48 hours for clade 1A (A), clade 1B (B), clade 3 (C), clade 4 (D), and clade 5 (E) was reported on a log₁₀ scale. These same data are grouped by clade in F. Data points represent the average from independent biological triplicates with error bars representing the SEM. For A–E, Šidák’s multiple comparisons test (comparing all strains to C. difficile R20291) was used, while Tukey’s multiple comparisons test (comparing all clades to each other) was used for F. Asterisks indicate a P-value of ** ≤0.02.

C. difficile clinical isolates are more sensitive to germinants than the C. difficile R20291 lab strain

Bile salts are cholesterol derivatives found in mammalian digestive systems and contribute to the absorption of nutrients (90). These compounds are produced in the liver and circulate through the small intestine before being recycled back to the liver. A small portion of these bile acids escape enterohepatic recirculation and enter the colon (91, 92). Germination by C. difficile spores occurs in response to two signals: a bile acid germinant and an amino acid co-germinant (93). Prior work has shown that TA and glycine are the most efficient germinants in vitro, and consequently, they are used frequently in germination analyses (51). Because germination is required for the successful outgrowth of vegetative cells from the dormant spore form, we sought to determine how the strains responded to germinants. Previous work from our lab and others has demonstrated that germination efficiency could be quantified as an EC₅₀ value using Michaelis-Menten kinetics (51, 52, 59–65).

All strains except for C. difficile PUC_90 (clade 1A) had a lower EC₅₀ value for glycine (the germinant is more potent) compared to the C. difficile R20291 strain (clade 2), which had an EC_50,glycine of 0.25 mM (Fig. 5A). Clade 1A strains showed the most variation in EC_50,glycine with values ranging from 0.06 mM (C. difficile HC52) to 0.30 mM (C. difficile PUC_90). Clade 4 strains were the most consistent with EC_50,glycine values ranging from 0.05 mM to 0.08 mM. The average EC_50,glycine value for each clade fell between 0.07 mM and 0.18 mM indicating that glycine sensitivity is consistent between the strains (Fig. 5B). The EC_50,TA value for all strains ranged from 0.10 mM (C. difficile P8, clade 5) to 0.60 mM (C. difficile HC52, clade 1A), all of which were lower than the C. difficile R20291 strain (clade 2, 2.0 mM; Fig. 5C). The largest variation was observed in the clade 1A strains with EC_50,TA values ranging from 0.20 mM (C. difficile PUC_90) to 0.80 mM (C. difficile HC52). Clade 5 strains had the least variation with the EC_50,TA values ranging from 0.10 mM (C. difficile P8) to 0.15 mM (C. difficile P12). At the clade level, EC_50,TA remained consistent with only slight variations between 0.10 mM for clades 3 and 5 and 0.60 mM for clade 4 (Fig. 5D).

Fig 5.

Strain sensitivity to germinants. Germination assays for each strain were performed in the presence of various concentrations of glycine (A and B), TA (C and D), or TA + CDCA (E and F). Germinant sensitivity was calculated using the maximum slope for each condition plotted against (co)-germinant concentration. The data fitted to a linear relationship by taking the inverse of the slope vs concentration plot, and from this, Ki/EC₅₀ was calculated with EC₅₀ equaling the concentration of germinant that produces the half maximum germination rate. The efficiency of the competitive inhibitor was calculated using the following equation K_i = (inhibitor)/([K_CDCA/K_TA] − 1) (61, 62). Data points represent the average from independent biological triplicates with error bars representing the SEM. For A, C, and E, Šidák’s multiple comparisons test (comparing all strains to C. difficile R20291) was used, while Tukey’s multiple comparisons test (comparing all clades to each other) was used for panels B, D, and F. Asterisks indicate P-values of * ≤0.05, ** ≤0.02, *** ≤0.01, and **** ≤0.0001.

In addition to observing the germination of the strains in response to two of the most efficient germinants, we also determined the response to a known competitive inhibitor of TA-mediated spore germination, CDCA. K_i values for each strain ranged from 0.10 mM for C. difficile PUC_75 to 0.90 mM for C. difficile strains PUC_256 and ICC5 (Fig. 5E). The C. difficile R20291 strain had a K_i value of 0.40 mM. There were no major differences in inhibitor sensitivity compared to the control. Additionally, these data are consistent between clades with only a 0.10 mM difference between the least (0.40 mM, clade 3) and greatest (0.50 mM, clade 1A) K_i values (Fig. 5F).

The bile acid and amino acid co-germinant signals are recognized in C. difficile by the pseudoproteases CspC and CspA, respectively (52, 94). These signals are transmitted to CspB, which then activates the cortex lytic enzyme, SleC (95–97). To determine if there was any genetic variation in the proteins responsible for initiating germination, we aligned the CspBA, CspC, and SleC sequences (52, 94, 95, 97, 98). The alignments (Fig. S11 to S13) of each protein revealed some variations compared to the C. difficile R20291 strain (clade 2), but none matched any residues that were found to influence germinant sensitivity (95, 98).

Secretion of toxins A and B varies in the C. difficile clinical isolates

Here, we used TcdA- and TcdB-specific ELISAs to assess the levels of toxins secreted by the strains. At 24 hours, negligible levels of TcdA were detected in all strains except C. difficile strains HC52 (clade 1A), PUC_90 (clade 1A), and PUC_75 (clade 3), which produced an approximately fivefold greater signal than the C. difficile R20291 ΔsigGΔtcdR negative control (Fig. 6A). At 48 hours, the clades 1A, 3 (except C. difficile C103), and 5 strains produced a TcdA signal approximately two- to fivefold greater than the negative control. At both 24 and 48 hours, clade 1A strains produced the most TcdA with a signal approximately fivefold greater than the negative control strain, while absorbance values for the clade 4 strains were below the limit of detection (Fig. 6B). TcdB levels were similar at both 24 and 48 hours between the clades 1A, 1B, and 3 strains (Fig. 6C). Clade 4 strains, C. difficile M68 and PUC_606, produced approximately onefold increase in TcdB signal relative to the negative control strain at 24 hours and approximately sixfold signal increase at 48 hours. The clade 5 strains show the largest increase in TcdB signal at 48 hours with approximately 5- to 11-fold increase compared to the negative control strain. TcdB levels were similar across all clades at each tested timepoint at approximately one- to twofold greater than the negative control strain (Fig. 6D). Toxin levels for all strains/clades at both 24 and 48 hours were lower than the C. difficile R20291 strain (Fig. 6).

Fig 6.

Toxin A/B levels of strains. Exponential-phase cultures of each strain were inoculated into BHIS and grown for 48 hours. At 24 and 48 hours post-inoculation, samples were collected and frozen. Thawed samples were tested for the presence of toxins A (A and B) and B (B and D) using a commercial ELISA kit from Epitope Diagnostics. Data points represent the average from independent biological triplicates with error bars representing the SEM. The limit of detection as determined by the absorbance of the C. difficile R20291 ΔsigGΔtcdR strain is indicated by a dotted line. A repeated measure (RM) two-way ANOVA was used to compare each strain to C. difficile R20291 (A and C) or to compare all clades to each other (B and D). Asterisks indicate P-values of * ≤0.05, ** ≤0.02, *** ≤0.01, and **** ≤0.0001 with black asterisks designating comparisons between the 24-hour samples and gray asterisks designating comparisons between 48-hour samples.

Genomic analysis of the PaLoc (data not shown) indicates high variability within the tcdA gene across the clinical strains. Notably, C. difficile C103 (clade 3) contains a large (~9 kb) insert between the tcdA and tcdB genes within the PaLoc, providing a reason for the lack of TcdA signal. Additionally, C. difficile ICC5 (clade 4) is missing the vast majority of this locus, which explains the absence of both TcdA and TcdB signals in this strain.

C. difficile strains are equally resistant to bile acid toxicity

Several bile acids are known to inhibit C. difficile growth (93, 99, 100). Thus, we sought to determine if any of our strains were equally susceptible to CA (the primary bile acid used to generate TA), DCA (a secondary bile acid generated from 7α-dehydroxylation of CA), and CDCA (an analog of CA) using MIC assays. To mimic the various regions of the colon encountered by C. difficile, each assay was performed at pH 7.5, 6.8, 6.2, and 5.5 (101–103). The results indicated little variation in bile acid sensitivity between strains (Fig. 7). At the most neutral pH, the MIC for CA was 7.5 mM. As the pH became more acidic, the MIC decreased eightfold to 0.94 mM (Fig. 7A). A similar trend was seen for CDCA and DCA with an MIC of 1 mM at pH 7.5 and an eightfold decrease in MIC at pH 5.5 (Fig. 7C and E). This data, when grouped by clade, again found no difference in bile acid sensitivity between clades (Fig. 7B, D and F).

Fig 7.

Bile salt sensitivity of strains. Exponential-phase cultures of each strain were inoculated into BHIS of the indicated pH and concentration of CA (A and B), CDCA (C and D), and DCA (E and F). MICs were assessed by the presence/absence of growth after ~18 hours. Black represents the highest tested concentration of bile salt value, while white represents no inhibition. Data represent the average from independent biological triplicates.

C. difficile strains vary in their ability to modify taurine-based bile salts

Previous work from our lab demonstrated that some C. difficile strains deconjugate taurine-conjugated bile salts (54). While it is currently unknown how this activity contributes to disease outcomes, the introduction of free taurine into the host could impact disease progression. Because the examined strains had various levels of activity, we tested the BSHA of our strains using TA as a substrate. Each strain was incubated with TA, and the amount of its deconjugated product (CA) was quantified (Fig. 8). C. difficile PUC_90 (clade 1A) and C. difficile ICC5 (clade 4) processed TA at an efficiency comparable to the C. difficile control (~70%–80%). C. difficile strains HC52 (clade 1A), S9 (clade 3), and PUC_606 (clade 4) had little activity against TA, while the remaining clade 1A, 3, and 4 strains had some activity (though not to the same level as C. difficile R20291 [clade 2]). Interestingly, all strains from both clades 1B and 5 had low levels of BSHA against TA.

Fig 8.

BSHA. Each strain was grown in the presence of 1 mM TA and incubated for 24 hours. The bile salts present in each culture following incubation were identified/quantified by reverse-phase HPLC. Percent deconjugation was calculated using the following formula: percent deconjugation = CA/(TA + CA). Data points represent the average from independent biological triplicates with error bars representing the SEM.

In prior work, we found that strains that did not deconjugate TA could process TDCA (54). Thus, we tested if clades 1B and 5 strains could deconjugate TDCA (Fig. S14). No detectable amount of DCA was generated by these strains. Taken together, this suggests that BSH activity may not be a core phenotype shared among strains.

C. difficile strains show similar surface motility

Some clade 5 strains have distinct growth morphologies compared to members of other clades (69). We tested surface motility for each strain to determine if we could detect similar morphologies (Fig. S15A through F). Strains from all clades spread similarly from the point of inoculation (Fig. S15G). The size of the protrusions from the central growth ring was the most striking variation, with C. difficile LK3P-030 (clade 1B) having the largest projections, and C. difficile M68 (clade 4) producing virtually no projections. The quantification of these projections, however, indicated that these differences, while visually apparent, were not statistically significant (Fig. S15H). As previously reported (69), the clade 5 strains had some asymmetric growth of the extensions, which is especially evident in the C. difficile P12 strain (clade 5; Fig. S15Fii).

Most genes known to contribute to the observed phenotypes are found in the core genome

To determine which phenotypes are core processes, we identified the core genome for our strains using Spine (Table S2) (48). While developed for use in Pseudomonas aeruginosa, several studies have reported its utility for C. difficile (35, 104–107). The genes encoding each protein aligned in this paper were compared to the list of core genome genes produced from the Spine analysis. Most of these genes (ccpa, xylA, xylB, xylR (CD630_RS16335), cspBA, cspC, and sleC) were found in the set of core genes for our strains, confirming their classification as core processes. Unsurprisingly, the genes corresponding to most variable phenotypes (treA, treR, tcdA, and tcdB) were not included in the core genome.

DISCUSSION

C. difficile is a highly variable species with an open pan-genome (22). Phylogenetic analysis indicates significant evolutionary changes between the five main clades as indicated by widespread homologous recombination and horizontal gene transfer (23, 24). Previous research has shown that even members of the same clade/ribotype had varying phenotypes (e.g., toxin activity) (108, 109). Additionally, several studies have observed an apparent correlation between ribotype and clinical outcome, with RT106 (clade 1/2) and RT078 (clade 5) being associated with the worst outcomes/recurrence compared to the commonly studied RT027 (clade 2) (34–42, 110). We hypothesized that despite evolutionary variation, C. difficile strains would have core phenotypes that are central to C. difficile biology and consequently similar across clades/ribotypes.

All strains had similar growth rates in the rich medium tested, indicating that any differences observed between strains are not due to differences in growth while in liquid BHIS during experimental setup. When growth was assessed in a minimal medium and compared to the laboratory-adapted C. difficile R20291 strain (clade 2), most of the clinical strains (except for the clade 1A strains which had similar generation times compared to C. difficile R20291) showed decreased growth rates in glucose-containing CDMM and increased growth rates in fructose-/xylose-/trehalose-containing medium. Between the clades, there was some variation, especially in the clade 4 strains which had consistently decreased growth rates in all tested media types. While alignments of the carbohydrate metabolism protein sequences revealed some genetic variation, these variations are not consistent with the observed phenotypic differences. This suggests that variations within the protein sequences themselves are not sufficient to explain the observed phenotypic variation.

Notably, another study reported that RT023 (clade 3 strains) was unable to grow on CDMM-tre (111). Our clade 3 strains grew relatively well in CDMM-tre, which may be the result of differences in experimental setup. Specifically, the strains in this study were grown in CDMM-tre for 24 hours followed by a subculture into the same media and a further 12 hours of growth prior to inoculation of the experimental culture. In the Midani study (111), cells were grown solely in BHIS prior to dilution into the experimental culture. If trehalose metabolism occurs in these strains via an unconventional pathway, as suggested by our genomic data (these strains lack an intact version of the canonical trehalose operon), processing of trehalose may be less efficient during the lag phase of growth. The additional time spent in the trehalose-supplemented minimal medium likely allowed the strains to adapt to the nutrient-limited conditions, reducing the lag phase and resulting in growth observable over 22 hours. Additionally, variation in the makeup of CDMM (defined amino acids vs casamino acids, additional vitamins, or variable lots of each constituent) or unknown strain-specific divergences could explain the differing results between studies.

For the number of spores produced by each strain over a 48-hour period, only two strains (C. difficile HC52: clade 1A and C. difficile LK3P-081: clade 1B) demonstrated any statistically significant differences compared to C. difficile R20291; the largest difference observed between the strains was ~10-fold. Whether this difference has any biological relevance remains to be seen. Given that some studies of C. difficile infection in mice report disease development with as few as 100 spores, this difference is unlikely to impact disease formation/progression but might contribute to altered disease spread/recurrence (6, 112).

We observed a decreased EC₅₀ value for both TA and glycine for most of our strains compared to the control. This indicates increased sensitivity to these germinants. The EC_50,glycine values are less than or similar to the physiologically relevant concentration of glycine (~0.1–0.3 mM), suggesting that these strains have become better adapted to sensing co-germinants in the mammalian gut (113–115). Additionally, the EC₅₀,_TA values are similar to the estimated concentration of TA in the colon following antibiotic treatment, suggesting that TA-mediated germination in all tested strains occurs efficiently in the gut (116, 117). Previously, the amounts of the Csp and SleC germination proteins present in C. difficile spores have been quantified (52). While not tested in this study, variation in the relative amounts of each protein found in each strain or changes in the transcriptional regulation of the proteins could contribute to the observed differences. Moreover, differences in spore coat composition could influence germinant permeability as seen in Bacillus anthracis (118). Additionally, we observed no differences in germination in the presence of CDCA, a competitive inhibitor of TA-mediated spore germination. CDCA was an effective inhibitor of TA-mediated germination in all tested strains. These results suggest that anti-germination-based therapies could be broadly applicable in the treatment of CDI. In support of this, recent work on the bile salt analogs, CamSA and CaPA, shows protection against CDI in mouse models of infection (119, 120).

Investigations of C. difficile strains isolated during outbreaks have assessed genotypic diversity of ribotypes/sequence types and reported upon virulence (i.e., toxin A/B) related phenotypes (121–123). These studies identified variation within the C. difficile PaLoc, which is often correlated to variation in disease severity. This is consistent with our phenotypic observations in vitro, especially regarding the variable levels of TcdA secreted by each clade. Notably, the clade 4 strains do not secrete significant amounts of TcdA consistent with their designation as an A⁻B⁺ toxinotype (124). Interestingly, the clinical strains all present with lower toxin levels compared to the C. difficile R20291 strain (clade 2). This could be the result of strain specificity in the antibodies from the ELISA kit used to assess toxin levels. Alternatively, differences in toxin secretion could contribute to the variable phenotypes, a possibility that would require further analysis of PaLoc gene expression in each of these strains.

Each of the strains was assessed for its ability to resist the toxic effects of certain bile acids. This assay was performed at various pHs to mimic conditions experienced by vegetative cells in the colon (101–103). We observed little variation in MIC when compared to C. difficile R20291 or between strains. As expected, the MIC of each bile salt increased with pH, reflecting the deprotonation of each bile acid (more negative charges), which limits its ability to interact with and disrupt the negatively charged bacterial membrane. At higher pH values, the MIC values for these strains are well above the physiological concentrations of CA, CDCA, and DCA in the colon (~0.12–0.23 mM) (125). This suggests that bile acid inhibition could be primarily effective in lower pH environments. The lack of variation suggests that bile acid toxicity is a core phenotype for C. difficile. Moreover, our results suggest that these bile acids may have different effects on C. difficile growth depending on the location within the gut.

Recent work from our lab found that C. difficile is capable of processing taurine-conjugated bile salts by removing the taurine group (54). When this assay was performed on the strains in this study, they showed varying abilities to deconjugate TA, with clades 1B and 5 members showing low processing levels. These same strains could also not process TDCA, indicating that the proteins responsible for this behavior are either missing, are not expressed, or have different substrates than the two tested here. Interestingly, this phenotype is unique to the clades 1B and 5 strains, which we previously noted have been associated with more severe/recurrent CDI. Because a bile salt hydrolase has not yet been identified in C. difficile, it remains unclear if BSHA, or the lack thereof, is relevant to the clinical outcome of CDI. Regardless, because not all the tested strains demonstrated BSHA, this may not be a core phenotype.

Most of the variation seen within each assay is observed in comparison to the control strain, C. difficile R20291 (clade 2). This strain was isolated during the Stoke-Manderville outbreak in the early 2000s, has been passed between laboratories, and may have become a laboratory-adapted strain (though still virulent in animal models) (126). Observed differences correspond to an increased fitness of the clinical isolates compared to the C. difficile R20291 strain (clade 2) as indicated by increased growth rates in fructose-/xylose-/trehalose-containing medium and increased sensitivity to germinants. This could indicate either a loss of some functions within the C. difficile R20291 strain or that C. difficile, as a species, has evolved to become more fit in the gut.

Spine analysis of our strains indicated that most of the genes known to encode proteins involved in the various tested phenotypes are members of the core genome. These proteins have functions related to global metabolism regulation, xylose usage, and germination. Additionally, proteins responsible for regulating the C. difficile sporulation program (spo0A, sigE, sigF, sigG, etc.) are also found in the core genome. This supports our designation of these phenotypes as core processes. Consistent with previous reports on the variation in trehalose metabolism and the non-typical arrangement of the trehalose metabolism genes in our clades 3 and 5 strains (Fig. S9 and S10), the treR and treA genes were not present in the core genome (25, 88, 111). This suggests, consistent with previous findings, that trehalose metabolism is highly variable between species and likely contributes to the variation in disease severity/recurrence observed between strains (25, 88, 89, 111).

When considering the strains independently of C. difficile R20291 (clade 2), we observed remarkable phenotypic similarity between strains and no major patterns corresponding to clade/ribotype. This is true of most tested phenotypes except for carbohydrate use, toxin production, and BSHA. This is consistent with previous research that demonstrates that C. difficile metabolism and toxin production are highly variable between strains (121, 122, 127, 128). Only identifying the factor responsible for BSHA will test the hypothesis that it also is not a core process. Because CDI severity has been shown to be highly variable within individual outbreaks, we hypothesize that those phenotypes classified here as non-core processes contribute more to variation than do the core processes. This is further supported by an in vivo study in which a set of accessory genes could be correlated with high virulence in mice infected with various clinical isolates of C. difficile that shared similar core genomes (129). Taken together, these data suggest that the previously observed relationships between ribotype and CDI severity may not be due to changes in these core phenotypes but rather to other influences including currently unidentified host factors.

This study focused only on a small portion of the known C. difficile strains and a limited number of phenotypes. Examples of other relevant phenotypes not tested in this paper include cytotoxicity, adherence, serotype, O₂ tolerance, and outgrowth of a cell post-spore germination (130–132). Variation in these phenotypes between strains would likely indicate a contribution to disease severity. Further analysis of phenotypic variation between strains of all ribotypes/clades both in vivo and in vitro will expand upon what we have learned here and provide valuable insight into how C. difficile might manifest itself in a clinical setting.

DATA AVAILABILITY

Long-read and Illumina data for all strains have been uploaded to the NCBI BioProject database under accession PRJNA1252054.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/aem.00964-25.

Supplemental figures, part I. aem.00964-25-s0001.pdf.

Figures S1 to S8.

Supplemental figures, part II. aem.00964-25-s0002.pdf.

Figures S9 to S15.

Table S1 and legends. aem.00964-25-s0003.docx.

Legends for Fig. S1 to S15; Table S1, strains, primers, and plasmids used in ELISA.

Table S2. aem.00964-25-s0004.xlsx.

Core genome list of genes.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.