The C. difficile strains used in this study were of clinical or non-clinical background (personal communication), with four pairs of clinical and non-clinical strains corresponding to ST and one additional non-clinical strain (Table 1). Strains were routinely cultivated under anaerobic conditions in supplemented Brain Heart Infusion Broth (BHIS; supplemented with 0.5% yeast extract, 0.05% L-cysteine, 0.0001% Na-resazurin, purged with nitrogen) at 37°C. Putative prophage regions of the strains were predicted with PHASTEST (Wishart et al., 2023).

The stress effect of DCA on the various strains was assessed via a minimum inhibitory concentration (MIC) assay and relative growth determinations at different concentrations. Cultivation was performed in cell culture plates (24 well for suspension cells, Sartorius AG, Göttingen, Germany) in an anaerobic tent (Coy Laboratory, Grass Lake, United States). Overnight cultures of C. difficile strains were cultivated as described above and used to inoculate 2 mL medium to a final OD₆₀₀ of 0.05, with either BHIS medium alone or supplemented with DCA (sodium deoxycholate, Sigma-Aldrich Chemie Gmbh, Taufkirchen, Germany) in concentrations ranging from 0.255 mM to 1.2 mM, which cover the physiological concentration range in humans (Hamilton et al., 2007). Culture plates were incubated at 37°C for 22 h and kept anaerobic during the OD₆₀₀ measurement in a Synergy 2 Plate Reader (Biotek Agilent Technologies Deutschland GmbH, Böblingen, Germany). Relative growth in the presence of DCA was calculated in relation to the untreated cultures. Each experiment was performed in triplicate.

Relative growth was visualized with ggplot2 (v3.4.2; Wickham, 2016) in RStudio (v2022.06.0; RStudio Team, 2020), and significance was determined with the Tukey’s ‘honest significant difference’ method implemented in the stats package (v4.2.0; Team, R Core, 2013).

Phage particles were isolated from untreated and DCA-induced cultures. Two pre-warmed flasks of 55 mL BHIS for each isolate were inoculated 1:100 from an overnight culture and incubated at 37°C. Growth was monitored via OD₆₀₀ measurements until an OD of ~0.6 (0.5–0.7). One culture for each isolate was induced with 0.255 mM (0.01%) DCA final concentration. The physiological concentration of DCA varies between individuals (Hamilton et al., 2007). We, therefore, used this concentration that is within the physiological range and was also used in various C. difficile studies regarding growth behavior or spore germination (Theriot et al., 2016; Lewis et al., 2017). The DCA solution was freshly prepared under anaerobic conditions, with DCA suspended in distilled water so that 100 μL inducing solution was required per 10 mL culture. The solution was sterilized by filtration (Filtropur S 0.2 μm, Sarstedt AG & Co. KG, Nümbrecht, Germany) and anaerobically added to the induction cultures. The second culture was not induced for analysis of spontaneous phage activity. Induced and non-induced cultures were further incubated until 22-h total incubation. The final OD₆₀₀ of each culture was determined before isolating phage particles.

For phage particle isolation, the cells were pelleted via centrifugation at 4°C and 3,000 x g for 15 min. The remaining cells were removed by filtration of the supernatant with a 0.45-μm Filtropur S filter (Sarstedt). Phage particles were pelleted via centrifugation at 8°C and 20,000 x g for 1 h. The pellet was suspended in 1 mL SM buffer (50 mM Tris–HCl, 100 mM NaCl, 8 mM MgSO_4, 7 H₂O, pH 7.4) supplemented with 0.5 mM CaCl₂ (supporting phage stability and upcoming DNase treatment) and let soak overnight at 4°C. Particle suspension was further supported the next day by shaking at 150 rpm (LT-V Lab-Shaker, Adolf Kühner AG, Birsfelden, Germany) at room temperature for 2 h. The suspended samples were finally transferred to 2 mL DNA Low Binding microtubes (Sarstedt) for the following treatments using cut filter tips to reduce possible shearing. The samples were stored at 4°C.

The phage DNA was isolated using the MasterPure Gram Positive DNA Purification Kit (Epicenter, Madison, WI, United States) with modifications. Prior to the phage DNA isolation, phage samples were supplemented with 2 μL of 100 mg/mL lysozyme solution (lysozyme from chicken egg white 177,000 U/mg; Serva, Heidelberg, Germany) suspended in SM buffer to remove remaining host cell debris and with 50 μg/mL final concentration RNase A (Biozym Scientific GmbH, Hess. Oldendorf, Germany) and 10 U Baseline-ZERO DNase (Biozym Scientific GmbH) to digest host nucleic acids. The samples were incubated at 37°C for 6 h with gentle inversion every 30 min. Fragments of host nucleic acids resulting from the digestion were removed by phage particle pelleting via centrifugation at 4°C and 20,000 x g for 1 h. The recovered pellet was suspended in 150 μL SM buffer. The suspension was supported by shaking at 150 rpm (LT-V Lab-Shaker) at room temperature for 10 min and slight flicking. EDTA (0.5 M, pH 8.0) was added to 10 mM final concentration for complete DNase inhibition.

Phage particles were lysed by adding 1% SDS (10% solution) and 2 μL Proteinase K (50 μg/μl; Biozym Scientific GmbH), incubation at 56°C for 1.5 h, and gentle inversion every 30 min. Subsequently, the samples were completely cooled down on ice before the addition of 130 μL MPC protein precipitation reagent (pre-cooled to −20°C). After mixing by gentle inversion, the proteins were pelleted via centrifugation at 4°C and 10,000 x g for 10 min. The DNA-containing supernatant was transferred to 1.5 mL DNA Low Binding microtubes (Sarstedt). The DNA was precipitated by addition of 0.3 M Na-acetate (3 M, pH 5.2), 10 mM MgCl₂ (2 M), and 0.8 volume isopropanol at room temperature. The samples were inverted 40 times and incubated at room temperature for 10 min before centrifugation at 4°C and 15,000 x g for 30 min. The supernatant was removed carefully, and the DNA pellet was washed twice with 150 μL 70% ethanol (pre-cooled to −20°C) and centrifugation at 4°C and 15,000 x g for 5 to 10 min. The supernatant was removed, and the sample was briefly centrifuged again to collect all residual ethanol for final removal. DNA pellets were air-dried under a sterile bench and immediately suspended in 20 μL TE buffer. Complete DNA elution was supported by brief storage at 4°C and slight flicking, before final storage at −20°C. The DNA concentration was assessed with the Qubit 3.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, United States) using the HS dsDNA assay kit.

The phage DNA was subjected to Illumina sequencing for dsDNA by paired-end library preparation with the Nextera XT DNA Library Preparation Kit (Illumina, San Diego, CA, United States) as recommended by the manufacturer. Libraries were sequenced using an Illumina MiSeq system and MiSeq Reagent Kit version 3 (2 × 300 bp, 600 cycles) as recommended by the manufacturer.

All following software was used in default mode. Sequencing raw reads were quality processed with fastp (v0.23.4; Chen et al., 2018) before removing the sequencing adapters with Trimmomatic (v0.39; Bolger et al., 2014). Processed reads were mapped onto the corresponding host genome using bowtie2 (v2.5.0), and the resulting SAM file was converted to a TDS file for bioinformatics analysis with the TraV software (Dietrich et al., 2014).

TDS files of processed reads for the same isolate were together inspected in TraV (Dietrich et al., 2014) for read coverage, and reads were normalized by calculation of nucleotide activity per kilobase of exon model per million mapped reads (NPKM). This results in a value for each CDS corresponding to its read coverage in reference to the overall read amount. NPKM values were further normalized to account for the differing growth behavior under the induction conditions by transforming values to an OD₆₀₀ of 2.0. In this way, NPKM values reflected phage abundance under the different conditions at the same cell density, which allows a better qualitative estimation of phage activity. OD normalization and visualization of NPKMs values were done in Rstudio (v2022.06.0; RStudio Team, 2020) using the packages tidyverse (v2.0.0; Wickham et al., 2019), ggforce (v0.4.1; Pedersen, 2022), and ggplot2 (v3.4.2; Wickham, 2016). NPKM values were plotted against the host genome with regard to the sequence start of the corresponding CDS, and original and normalized NPKM values were plotted together to visualize the effect of OD normalization. Phage regions predicted with PHASTEST (Wishart et al., 2023) were also implemented.

Active regions identified via sequence read mapping were extracted for a new genome annotation. Sequence ranges were thereby adopted from PHASTEST (Wishart et al., 2023) predictions or selected based on read mapping in TraV (Dietrich et al., 2014), if the predicted prophage region did not cover the entire mapped region. Annotation was customized for phage genomes using Pharokka (v1.3.2; Bouras et al., 2023) in default mode with sequence re-orientation to the large terminase subunit. If the large terminase subunit was not annotated automatically, it was determined via BLAST analysis (Johnson et al., 2008) and manually annotated. For specific genes and their encoded protein, additional analyses with protein BLAST (Johnson et al., 2008) and InterProScan (v5.63–95.0; Jones et al., 2014) were performed.

Genome-based classifications were done with different bioinformatic analysis tools. An average nucleotide identity analysis (ANI) with pyani (v0.2.12; Pritchard, 2014) and MUMmer3 alignment (Kurtz et al., 2004) (ANIm) was used in default mode to compare the phages among each other. The assessment of the DNA-packaging strategy was performed based on the study of Rashid et al. (2016). The large terminase subunit was aligned at the protein sequence level with ClustalW, and a maximum-likelihood phylogenetic tree was constructed with the Whelan and Goldman (WAG) substitution model and otherwise default parameter with the software MEGA (v11.0.13; Tamura et al., 2021). The branches were collapsed in MEGA (Tamura et al., 2021) if none of our phages clustered within, and final modifications for visualization were done in Inkscape (v0.48).¹ A nucleotide BLAST analysis (Johnson et al., 2008) with default parameters was performed to check for similarity to already known phage genomes and the prevalence in genomes of other C. difficile strains. BLAST results were ordered based on query coverage and hits with a query coverage below 10% were neglected unless relevant matches with higher coverages were not obtained. For assigning the phages to a morphological family of the order Caudovirales, phage genomes were inspected for the presence of baseplate proteins and sequence length of the tail length tape measure protein (Zinke et al., 2022). If the tail length tape measure protein was not annotated, it was identified via protein BLAST analysis (Johnson et al., 2008) and manually curated in the genome.

Before investigating the effect of the secondary bile salt DCA on prophage activity in C. difficile, individual DCA tolerance of the various strains was assessed in the form of a MIC assay with relative growth determinations (Figure 1). DCA concentrations ranged from 0.255 mM to 1.2 mM, thereby comprising the physiological human concentration of DCA (Hamilton et al., 2007). All strains already exhibited reduced growth at the lowest concentration, which further decreased with increasing concentration. At all concentrations, strains of the same ST showed no significant difference in DCA tolerance, which implied comparable stress levels. Similar stress levels in turn might imply similar cellular strategies to cope with DCA-associated cellular damage but could also hint at similar DCA-induced prophage activity. In contrast, ST-specific tolerance differences were apparent across the DCA concentration range, with ST1 exhibiting the highest tolerance, followed by ST8, ST11, ST3, and, lastly, ST340, which was the most susceptible ST. Consequently, DCA tolerance correlated with the ST but not with the clinical background. The tolerance difference between the STs was most distinct at the lowest concentration, which was also used in the prophage induction experiments. Determined MICs ranged from 1 mM (ST11 and ST340) to 1.2 mM (ST1, ST3, ST8).

Prophage prediction of all analyzed genomes exhibited various putative prophage regions, often with multiple incomplete and intact predicted regions in one genome (Supplementary Table S1). Active prophages were determined by sequencing of particle-protected DNA. Sequencing reads were mapped to the corresponding host genome. Normalized read coverage (NPKM values) represented phage abundance as a relative measure indicative of phage activity (Hertel et al., 2015). In the following, the term phage activity describes the production and resulting abundance of DNA-containing particles, and the mentioned NPKM values refer to the OD-normalized data. Sequencing of phage DNA libraries was successful for all samples except for strain DSM 29747, which was the only strain without a predicted intact prophage genomic region (Supplementary Table S1). This strain was therefore missing in further analyses.

The mapping of phage DNA sequencing reads onto respective host genomes is depicted in Figures 2–6. The overall examination of the mapping results revealed distinct activity of at least one region in all strains and under both induction conditions. This demonstrated spontaneous phage activity in all strains and simultaneous activity of multiple phages within the same host. Almost all regions matched well with the predicted and intact prophage regions. All these regions are summarized in Table 2 for name assignment to facilitate the following descriptions. As apparent in Table 2, regions with positive prophage prediction accorded in size with typical genome sizes of C. difficile phages (Heuler et al., 2021), while those without were significantly smaller.

Comparison of corresponding strains showed no apparent differences in carriage or location of active phages. Correspondingly, a correlation to the clinical background of a strain was not detected. As active prophages of corresponding strains resided at corresponding genome positions, we performed an ANIm analysis on extracted sequences of all active regions to assess their similarity among each other (Supplementary Figure S1). This confirmed identical sequences of the analogous phages TS3_3_phi/DSM28196_phi1 of ST1-strains and high similarity of both phages J2_1_phi1/SC083-01-01_phi1 and J2_1_phi2/SC083-01-01_phi2 among the ST8-strains, while the other phages exhibited only little or no similarity to the others.

All spontaneously active regions (Table 2) were inspected for their activities under DCA induction. The overall NPKM transformation to an OD of 2 revealed distinctly higher signals under DCA induction in most active regions and all strains. This verified the phage-inducing effect of DCA, which varied apparently between the different regions even within the same host and thereby implied phage dependency.

The genomes of both ST1-strains carried an identical phage (Supplementary Figure S1; TS3_3_phi/ DSM28196_phi1) as determined by ANIm analysis (Supplementary Figure S1) and similar genome position. The analogous phages showed distinct spontaneous activity with approximate magnitude, but their DCA-induced activity differed substantially. Phage DSM28196_phi1 (Figure 2B) exhibited ~3.5x higher signal increase under DCA than TS3_3_phi in the non-clinical strain (Figure 2A). Since the phages were identical, the differing reactions seemed to be host-related. The genome of strain DSM 28196 possessed another active region DSM28196_phi2, which showed minor activity under both conditions and was phage-atypical by comprising only four genes and missing a prophage prediction (Figure 2B).

The genomes of ST3-strains possessed several active regions (Figure 3), which were not similar to each other (Supplementary Figure S1). In the genome of the non-clinical strain B1_2 (Figure 3A), two of the four active regions comprised the two ECEs. B1_2_phi4 on ECE2 exhibited the highest NPKM values under DCA induction. Remarkably, B1_2_phi4 is another phage-atypical but active region without corresponding prophage prediction, as previously detected for DSM28196_phi2 (Figure 2B). B1_2_phi3 on ECE1 also showed increased activity under DCA, but both spontaneous and induced activity were not particularly high. B1_2_phi2 on the chromosome showed spontaneous activity and a substantial increase in DCA-induced activity. B1_2_phi1 on the chromosome exhibited almost no signal under spontaneous conditions, which slightly increased upon DCA induction. This might indicate true DCA induction of this phage without prior spontaneous activity. In contrast to strain B1_2, the corresponding clinical strain SC084-01-01 possessed no prominently active region on the chromosome (Figure 3B). The activity could be observed within the region SC084-01-01_phi1, but NPKM values were very low under both conditions and signals did not cover the whole phage region. The abundancy of this phage was probably insufficient to capture distinct activity by the sequencing approach. Similar activity was observed for SC084-01-01_phi2 on ECE1 of SC084-01-01. SC084-01-01_phi3 on ECE2 was the only region in SC084-01-01 with prominent activity under spontaneous conditions, and the activity further increased apparently under DCA induction.

The genomes of both ST8-strains exhibited activity for their analogous chromosomal and extrachromosomal regions (Figure 4), which were similar phages according to ANIm analysis (Supplementary Figure S1) and similar genomic location. J2_1_phi2 on the ECE of strain J2_1 showed distinct spontaneous activity and a DCA-induced activity increase (Figure 4A), whereas SC083-01-01_phi1 on the ECE of SC083-01-01 was only slightly active under spontaneous conditions, and the signal increase upon DCA induction was only little. The activity of the chromosomal phages differed apparently as well. J2_1_phi1 on the J2_1 chromosome was spontaneously active and showed increased activity under DCA induction (Figure 4A). The left part of the prophage region started with minor activity, which drastically increased at the terminase genes. Strikingly, the left part of the prophage region started with minor activity, which drastically increased at the terminase genes. Further remarkable, sequencing reads mapped beyond the predicted prophage region and spread upstream (~30 kb) and downstream (~135 kb). The downstream region adjoined the phage activity with similar NPKM values that gradually decreased over the entire section. The chromosomal phage SC083-01-01_phi1 of the corresponding clinical ST8-strain SC083-01-01 did not exhibit these peculiarities (Figure 4B). It was spontaneously active, and the activity increased substantially under DCA treatment. This increase was strikingly twice as high as observed for the analogous phage J2_1_phi1 (Figure 4A), despite their similarity (Supplementary Figure S1). Such dissimilar activity increase among analogous phages was already observed in the ST1-strains (Figure 2). In both ST8- and ST1-pairs, a distinctly stronger increase upon DCA induction was connected to the clinical background of the strains. Since DCA-stress levels did not significantly differ between corresponding clinical and non-clinical strains (Figure 1) and thereby suggested similar induction levels, the question of the underlying mechanism of these diverging activities in analogous phages arose. This prompted an undefined regulation of phage induction involved in the clinical strains.

Strain MA_1 could not be compared to its corresponding clinical strain DSM 29747, but distinct spontaneous activity was visible for MA_1_phi on the ECE, which increased under DCA treatment (Figure 5).

Strain MA_2 is, to our knowledge, the first cryptic C. difficile strain with detailed phage examination. This strain possessed four active regions (Figure 6). MA_2_phi2 on the chromosome and MA_2_phi4 on the ECE both showed prominent activity under spontaneous conditions and a distinct activity increase upon DCA induction. Interestingly, MA_2_phi4 was another active region without phage prediction, as observed previously for DSM28196_phi2 (Figure 2B) and B1_2_phi4 (Figure 3A). This was also true for MA_2_phi3 on the chromosome, where the activity was, however, very low under spontaneous conditions. The activity of this region increased upon DCA induction. Chromosomal region MA_2_phi1 exhibited the least activity in this genome even under DCA treatment. Read mapping for MA_2_phi1 and MA_2_phi2 did not cover the entire predicted prophage regions, but sections without read coverage contained only bacterial genes and were therefore evidently mispredicted.

Almost all identified active prophage regions could be induced by the secondary bile salt DCA. Mentionable, the phage activity measured after DCA treatment might be influenced by a direct effect of DCA on the phage. A study on different bacteriophages in Escherichia coli investigated the effect of bile salts on the host–phage interaction and observed varying survival rates of the phages (Scanlan et al., 2017). Consequently, DCA might not only induce but also damage phages, thereby reducing the abundance of induced phage particles and the measured phage activity, respectively. The measured activity in our experiments might therefore be lower than the actual activity after induction.

DCA is not the only secondary bile salt present in the human intestines that can stress C. difficile cells (Sorg and Sonenshein, 2008; Thanissery et al., 2017). It might therefore be assumed that the other secondary bile salts induce prophages as well. Overall, the analysis of prophage induction and activity with all secondary bile salts—individually and combined—is advisable to further understand C. difficile prophage activity in vivo. In this regard, the corresponding toxin production of the analyzed strains would be another interesting aspect for investigations. The secondary bile salts were demonstrated to negatively affect toxin production (Thanissery et al., 2017) and function (Tam et al., 2020). In contrast, prophages can affect toxin production both negatively (Govind et al., 2009) and positively (Sekulovic et al., 2011). As such, phage induction in the presence of secondary bile salts can not only promote phage-mediated HGT and drive adaptation but also influence toxin levels upon lysogenization of other toxigenic strains, thereby further affecting strain virulence and disease severity.

Moreover, almost all ECEs were detected as active phages. Although extrachromosomal prophages have already been described in C. difficile strains (Garneau et al., 2018; Ramírez-Vargas et al., 2018), only a few of these have been isolated and characterized so far (Heuler et al., 2021).