Our study enrolled 117 patients hospitalized at the First Affiliated Hospital, Zhejiang University School of Medicine (China), after informed consent. Considering patients were hospitalized for different causes and subsequently underwent different treatments, no exclusion criteria except minimum age of 18 was used in this study. Stool samples collected from these 117 patients were divided into four groups associated with diarrhea and C. difficile colonization statuses: (1) Control group (CN, n = 30): patients without diarrhea symptoms and a negative C. difficile test by glutamate dehydrogenase (GDH), enzyme immunoassay (EIA) toxin A/B and quantitative PCR for toxin B gene. (2) Non-CDI diarrhea (NCD, n = 30): patients with diarrhea (a passage of more than 3 times of unformed stools) and a negative C. difficile test same as the control group. (3) CDI group (CDI, n = 30): patients diagnosed with new-onset diarrhea and a positive C. difficile test. (4) Asymptomatic colonization group (AC, n = 27): patients without diarrhea and a positive C. difficile test. Diarrhea was defined as ≥3 loose stools in 24 h (Vázquez-Cuesta et al., 2023). Patients who had CDI episodes or CDI treatment within the past 3 months were excluded.

Stool samples were collected in sterile containers and immediately transported to the laboratory for CDI diagnosis, and then stored at −80 °C until DNA extraction.

Stool DNA was extracted using the DNeasy PowerSoil Pro Kit (Qiagen, Germany) according to the manufacturer’s instructions. DNA concentration and purity were measured using NanoDrop 2000 (Thermo Fisher Scientific, MA, United States).

Primer sequences used for toxigenic C. difficile quantification were F: 5’-AGC AGTTGAATATAGTGGTTTAGTTAGAGTTG-3’and R: 5’-CATGCTTTTTTAGTTT CTGGATTGAA-3′ (Maestri et al., 2022). Quantitative PCR (qPCR) was performed on a C1000-Touch Real-time PCR instrument (Bio-Rad, United States). The reaction system and conditions of qPCR were established by following the protocol of ChamQ SYBR qPCR Master Mix (Vazyme, China). All samples were performed in duplicate, and samples were considered positive when both replicates were positive.

The hypervariable V3V4 region of bacterial 16S rRNA gene was amplified using 343F (5’-TACGGRAGGCAGCAG-3′) and 798R (5’-AGGGTATCTAATCCT-3′). Amplicons were tested by 2% agarose gel electrophoresis and purified with Agencourt AMPure XP beads (Beckman Coulter Co., United States). The purified DNA samples were sent for sequencing on Illumina NovaSeq6000 platform (Shanghai OE Biotech, China). Raw sequencing data were deposited in the China National Center for Bioinformation (CNCB) under accession number PRJCA050088.

The raw sequencing data were trimmed and filtered to obtain high-quality data (clean data). Then amplicon sequence variants (ASV) and the ASV abundance table were generated using DADA2 with the default parameters of QIIME2 (Callahan et al., 2016). All ASVs were annotated and blasted against SILVA 138 database to obtain taxonomic classification.

The core microbiota of each group based on ASV level was identified here. The richness of gut bacterial communities in CN, NCD, CDI and AC groups were estimated by Chao1 index. Principal coordinates analysis (PCoA) was generated to assess bacterial community structure among different groups based on unweighted unifrac distances. Co-occurrence network analyses were performed using the Molecular Ecological Network Analyses (MENA, http://ieg4.rccc.ou.edu/MENA/), and the network was visualized using Gephi 0.10.1 (Deng et al., 2012). We use linear discriminant analysis (LDA) and LDA effective size analysis (LEfSe) to illustrate species difference at the level of genus between AC and CDI groups. Spearman’s correlation analysis was applied to analyze the correlation between differential genus and disease severity. CDI severity score index was calculated in accordance with previous studies (Peretz et al., 2016; Mulki et al., 2017). For each patient the score incorporated nine parameters, each variable added one point: serum albumin <3 g/dL, a WBC count that is markedly high (>50,000 cells/mL) or low (<2000 cells/mL), serum creatinine >130 μmol/L, serum lactate levels >2.2 mmol/L, temperature >38.5 °C, systolic blood pressure <100 mmHg, altered mental status, severe abdominal tenderness or distention, intestinal symptoms such as ileus, obstruction, perforation, toxic mega-colon.

C. difficile strain (CD) used for co-culture was clinical isolate (sequencing type 54, tcdA+/tcdB+) that was preserved in our laboratory. O. splanchnicus strain ATCC 29572 (OS) was purchased from BeNa Culture Collection, Henan, China. CD and OS were grown at 37 °C in brain-heart-infusion broth (Oxoid) supplemented with 0.5% yeast extract (Oxoid) and 0.1% L-cysteine (Sigma-Aldrich) (BHIS), under the anaerobic environment (80% nitrogen, 10% hydrogen, 10% carbon dioxide) (Smith et al., 2022).

To construct co-culture systems, single colonies of CD and OS were separately inoculated into BHIS and grown anaerobically at 37 °C overnight. Each strain was sub-cultured into fresh media and grown to mid-exponential phase (OD_600nm = 0.3 to 0.7). Then, the co-cultures of the two strains were prepared at different inoculation ratios (CD: OS = 9:1, CD: OS = 1:1, CD: OS = 1:9), and the final inoculum concentration was 1 × 10⁷ CFU/mL.

At each timepoint (6 h,12 h,18 h,24 h and 30 h), 1 mL samples of co-cultures were taken, centrifugated at 12000 g for 1 min and the supernatant discarded. Cell pellets were frozen at −80 °C until gDNA extractions were performed. The gDNA was extracted using a DNeasy PowerSoil Pro Kit (Qiagen) as per the manufacturer’s instructions. Primers specific to OS were designed based on the sequence of the hypervariable regions of 16S rRNA gene (F: 5′- GTGGGTAGCGAACAGGATTAG-3′; R: 5′- CCCAGGTGGCTCACTTAATAC-3′). Primers specific to CD and the qPCR reaction was performed as mentioned above. For absolute quantitation of copy numbers of CD and OS, a standard curve method was employed. Briefly, a fragment containing the target qPCR amplicon for each bacterium was amplified from genomic DNA using specific primers and cloned into a standard plasmid vector using a TA cloning kit (Promega, United States). The resulting recombinant plasmids were verified by sequencing. Plasmid DNA was extracted and quantified using a spectrophotometer. The copy number of each plasmid per microliter was calculated and a ten-fold serial dilution series of each plasmid was prepared to generate the standard curve. The amplification efficiency was 105.9%, R² = 0.9995 for OS primers and 107.8%, R² = 0.9979 for CD primers (Supplementary Figure S1). The copy numbers of CD and OS in each experimental sample were then interpolated from their respective standard curves.

Toxin production was monitored over time by collecting culture supernatants at 12-, 24-, 48-, and 72-h post-inoculation. Toxins (both TcdA and TcdB) concentrations in co-cultures were determined using ELISA kit (tgcBiomics, Germany). The blank media used for culturing was set as negative control, and supernatant of CD monocultures was set as positive control (Sulaiman et al., 2024).

For gene expression analysis, 1 mL aliquots of the co-culture were harvested at 12, 24, 48, and 72 h. Bacterial cells were pelleted by centrifugation at 12,000 g for 1 min. Total RNA was then extracted from the pellet using the RNeasy mini kit (Qiagen, Germany) with on-column DNase I digestion to eliminate genomic DNA contamination. The isolated total RNA was reverse transcribed by PrimeScript™RT Master Mix (Takara, Japan) according to manufacturer’s instructions. A 7500 real-time PCR system (Bio-Rad, United States) was used to perform quantitative PCR reactions. The system and conditions of RT-qPCR were established by following the protocol of the SYBR^® Premix Ex Taq Kit (Takara, Japan). Primer sequences used for tcdA quantification were tcdA-F: 5’-GCGGAAATGGTAGAAATG-3′ and tcdA-R: 5’-ATCAGGTGCTATCAATACTT-3′ (Zhu et al., 2021). Primer sequences used for tcdB quantification were F: 5’-AGC AGTTGAATATAGTGGTTTAGTTAGAGTTG-3’and R: 5’-CATGCTTTTTTAGTTT CTGGATTGAA-3′ (Maestri et al., 2022). 16 s rRNA of C. difficile used as an internal control, primer sequences were CD16-F: 5’-TTGAGCGATTTACTTCGGTAAAGA-3′ and CD16-R: 5’-CCATCCTGTACTGGCTCACCT-3′. Relative gene transcript levels was calculated by the 2^−ΔΔCT method (Livak and Schmittgen, 2001). All reactions were carried out in triplicate.

Cytotoxicity assays were done on Vero cells (Trzilova et al., 2022). For the assay, Vero cells were passaged in a 96-multiwell plate and incubated for 24 h prior to use. Cells were seeded in a six-well plate (3 × 10⁵ cells/well), in Dulbecco’s Modified Eagles Medium (DMEM), supplemented with 10% fetal bovine serum. Cells were grown for 24 h at 37 °C, 5% CO₂ in air atmosphere. Supernatants collected at various time points were centrifuged and filtered through a 0.22 μm filter. 50 μL of the final content was added to each well plate and plates were incubated at 37 °C, in 5% CO₂, for 18 h. Empty BHIS medium served as negative control. Cell rounding was assessed using a 10x objective on a light microscope.

Ethanol resistance was used to determine the sporulation efficiency of monoculture and co-culture systems based on previously described procedures (Edwards and McBride, 2014; Donnelly et al., 2022). 0.5 mL aliquot of culture was mixed with 0.5 mL 95% ethanol, subjected to vortex mixing, incubated at room temperature for 15 min, serially diluted in 1 × PBS, and plated on BHIS with 0.1% taurocholate to enumerate spores.

Our monoculture and co-culture growth contain 3 biological replicates, data are expressed as mean ± SEM. Statistical analysis among multiple groups was performed using SPSS (24.0) for one-way ANOVA analysis, and statistical significance was set at p < 0.05. These results were visualized Origin 2024 software.

The 117 patients included in our study were divided into the control group (CN, 30 subjects), non-CDI diarrhea group (NCD, 30 subjects), CDI group (CDI, 30 subjects) and asymptomatic carrier group (AC, 27 subjects). The patients’ characteristics are presented in Table 1. The median age of all group patients was around of 60 years, and the majority of AC and CDI groups were long stay inpatients (over 30 days). There were no significant differences in the rate of use of antibiotics among groups (p = 0.348).

In total, 2,879 ASVs were obtained from the 117 fecal samples, and the number of ASVs of each sample ranged from 25 to 258. According to previous studies, the ASVs that presented in all groups were defined as core taxa, the ASVs that only presented in one group were defined as unique taxa, and the ASVs shared by two or three groups were defined as other taxa (Shade and Handelsman, 2012). As shown in Figure 1, the numbers of total ASVs were similar between CN (1,250 ASVs) and AC (1,206 ASVs) groups, while the total ASV numbers of NCD (1,077 ASVs) and CDI (1,071 ASVs) groups were very close. The number of ASVs that were shared by all groups was 264, and the four groups had large proportion of unique ASVs, and the number of unique ASVs was highest in AC (511 ASVs), followed by CN (493 ASVs), CDI (466 ASVs) and NCD (460 ASVs). The shared ASVs in CN&AC were highest (126 ASVs), followed by CN&AC&NCD (97 ASVs). The α-diversity analysis suggested that the microbial diversity was significantly decreased in CDI and AC groups, compared to CN group (Figure 2A). Furthermore, β-diversity analysis indicated that the composition of bacterial communities in CDI group significantly differed from other groups (PERMANOVA, p = 0.001), while CN, NCD and AC groups had similar microbial community structure (Figure 2B). Altogether, these findings indicated patients with true infection had unique gut microbial community, and the gut microbiome may be a potential biomarker source.

To further understand the bacterial composition differences among CN, NCD, CDI and AC groups, we analyzed the relative abundance of fecal microbes in each group at phylum and genus level, respectively. Generally, the phylum Firmicutes displayed the highest relative abundance in all groups (CN, 61.8%; NCD, 51.3%; CDI, 56.5%; AC, 54.7%) (Figure 2A). The phylum Proteobacteria was increased in CDI (27.0%), AC (27.0%), and especially in NCD (38.8%) compared to CN (22.2%). In addition, CDI and AC group had increased Bacteroidota and decreased Actinobacteriota at phylum level compared to CN and NCD groups. Moreover, the differences among groups at the genus level (top 10 in abundance) were further investigated. As shown in Figure 2B, the relative abundance of Lactobacillus and Bifidobacterium decreased in NCD, CDI and AC groups, while the relative abundance of Escherichia, Klebsiella and Streptococcus increased in comparison with CN group.

Furthermore, microbial co-occurrence network was constructed on genus level (>0.1% relative abundance) to show the interactions among individual species. The network analysis results highlighted the differences of interaction properties between the four groups (Figure 3). In comparison with CN (119 nodes, 398 edges), NCD (99 nodes, 292 edges) and CDI (88 nodes, 170 edges) had lower number of nodes and edges. Moreover, NCD and CDI also showed the decreasing average degree and lower proportion of negative interactions, indicating weaker bacterial interactions in these two groups. AC (117 nodes, 426 edges) had more edges and higher average degree, however, the proportion of negative interactions decreased in AC in comparison with CN. CDI had distinct network structure compared to the other groups, shown with lower number of nodes and edges, decreased average degree and higher modularity (Table 2). These results suggested that the bacterial interactions were much weaker in CDI group.

Bacterial community structure and network analysis had suggested the significant gut microbiota alteration in CDI patients. To explore the biomarkers that could distinguish the true infection from asymptomatic colonization, we explored the most discriminative bacteria between AC and CDI groups using LefSe analysis. The cladogram indicated 76 discriminative features between AC and CDI groups (LDA score ≥ 2, Figure 4). Some known and potential probiotics including Bifidobacterium, Faecalibaculum and Odoribacter were significantly decreased in CDI group, while Dialister, Pasteurella, Pseudomonas and Veillonellaceae were enriched in CDI group.

Next, all CDI patients were evaluated for the severity of the disease according to the criteria described above (Supplementary Figure S2), and genera associated with CDI disease severity were analyzed. In total, 46 genera were significantly associated with severity scores, of which 14 were positively correlated with severity scores and 32 were negatively correlated with severity scores (Figure 5A). Among them, Odoribacter was the most negatively correlated with CDI severity scores, suggesting it’s abundance may correlate with CDI disease progression. Subsequently, the abundance of Odoribacter among four groups was analyzed. Concordantly, we found that the abundance of Odoribacter was significantly lower in CDI group compared with CN, NCD and AC group (Figure 5B).

To determine whether Odoribacter influences key pathogenesis phenotypes of C. difficile (CD) including bacterial growth, sporulation, and toxin production, we selected the type strain O. splanchnicus ATCC 29572 (OS) and established CD-OS co-culture systems with different bacterial concentration ratio and quantified the proportions of the two strains. We first determined the growth kinetics of CD and OS in monoculture. The growth curve of OS was characterized by a prolonged lag phase compared with CD (Figure 6A). And we noticed the optical density of CD cultures began a sustained decrease after 12 h, which might be indicative of the initiation of sporulation. The results of co-culture systems showed that OS bacterial load finally stabilized at 5 × 10⁸ CFU/mL across all co-culture systems. However, the final bacterial load of C. difficile exhibited a significant reduction specifically in the 1:9 (CD: OS) ratio system (Figure 6B). In contrast, the CD load remained stable at around 7 × 10⁷ CFU/mL in the 1:1 and 9:1 ratio systems (Figures 6C,D). The shifting compositional ratio of the two species over time was also evaluated. The data show that the proportion of CD increased progressively until 18 h, after which it underwent a consistent decrease. This reversal ultimately resulted in the dominance of OS in all systems at 30 h (Figures 6B–D).

We next investigated the impact of OS on toxin production by CD. A time-course analysis of toxin levels in culture supernatants was performed by ELISA. At the 12-h time point, toxin accumulation was not detectable in any culture group, with levels falling below the assay’s detection limit (data not shown). By 24 h, toxins were detected in the CD monoculture and the CD: OS = 9:1 co-culture, while it remained undetectable in the CD: OS = 1:1 and 1:9 groups (Figure 7A). A significant difference emerged at 48 h, where the toxin titer in the CD monoculture reached approximately 8 μg/L, which was significantly higher than in all co-culture conditions (Figure 7B). At 72 h, only the CD: OS = 1:9 co-culture maintained a statistically significant reduction in toxin levels compared to the CD monoculture. The toxin levels in the CD: OS = 9:1 and 1:1 groups, although numerically lower, were not significantly different from the monoculture control (Figure 7C).

To elucidate the molecular mechanism underlying this suppression, we quantified the expression of the toxin genes tcdA and tcdB via qPCR at corresponding time points. At the early 12-h time point, the expression of both tcdA and tcdB was significantly lower in all co-culture groups compared to the CD monoculture (Figures 7D,E). At subsequent time points, tcdA expression in co-cultures showed no significant difference from the monoculture. In contrast, a sustained transcriptional repression of tcdB was observed specifically in the CD: OS = 1:9 and 1:1 co-cultures at 24 h, where its expression remained significantly suppressed compared to the CD-alone control.

To validate that functional toxin was being released, a Vero cell cytotoxicity assay was performed using culture supernatants harvested at 12, 24, 48, and 72 h. Morphological analysis revealed a clear time-dependent and co-culture-dependent cytopathic effect (Figure 8). Supernatants from the 12-h time point, including that from the CD monoculture, induced no apparent morphological changes in Vero cells compared to the negative control. Exposure to 24-h supernatants resulted in initial cell rounding, which was markedly more severe in cells treated with the CD monoculture supernatant than in those treated with co-culture supernatants (CD: OS = 9:1, 1:1, and 1:9), where a greater proportion of cells remained adherent. Treatment with 48-h and, particularly, 72-h supernatants led to extensive cell rounding and detachment. Notably, cells exposed to the 72-h supernatant from the CD monoculture showed substantial death. Across these later time points, the cytopathic effect was consistently mildest in cells treated with the supernatant from the CD: OS = 1:9 co-culture.

We further assessed the dynamics of CD spore formation under different co-culture conditions. Utilizing the ethanol tolerance of spores, we collected cultures at various time points, and quantified viable spores by plate counting. The results indicated that no significant spore formation was detected in any group at the early 12-h time point (Figures 9A,E). Although trace spores were observed in some replicates of the CD: OS = 1:9 group, the high variability between biological replicates precluded statistical significance. By 24 h, the spore counts in all co-culture groups (CD: OS = 9:1, 1:1, and 1:9) were significantly higher than in the CD monoculture (Figures 9B,F). At the subsequent 48-h and 72-h time points, total spore numbers increased across all groups, with the CD: OS = 1:9 co-culture consistently yielding the highest spore count (Figures 9C,D,G,H).