Clostridioides difficile concentration-dependant alterations in gut microbiota of asymptomatic infants

The stool samples used in this study were collected as part of a previously published study [23]. This was a randomized, double-blind, placebo-controlled trial conducted at three Slovenian neonatal units between November 2016 and March 2019. The trial was registered at ClinicalTrials.gov (Identifier: NCT02865564) and received approval from the National Medical Ethics Committee (No. 89/03/15). Detailed methodologies are provided in [23]. Briefly, stool samples used in this study were obtained from 76 infants along with clinical, anthropometric, demographic and feeding data. The inclusion criteria into original study were antibiotic administration for suspected early or late neonatal sepsis for at least five days (ampicillin/flucloxacillin and gentamicin), at the age of less than 21 days, and gestational age of at least 37 weeks. In present study, we included only stool samples collected from these infants at follow-up visit, one year after the completion of antibiotic treatment. None of the infants included in this study were diagnosed with or actively treated for C. difficile infection during the study period.

Stool samples were collected in sealed sterile containers and promptly refrigerated for less than 24 h before storage at − 80 °C for subsequent analyses. Initially, suspensions of stool samples weighing between 0.1 and 0.5 g were prepared in Ringer’s solution, diluted at a ratio of 1:100. Following centrifugation at 3600 g for 10 min at 10 °C, the resultant pellet underwent lysis and sonication according to the methodology outlined by Matijašić et al. 2014 [24]. DNA extraction was then carried out using the automated Maxwell 16 System protocol (Promega, USA).

Number of copies of C. difficile 16S rRNA gene and toxin tcdB gene were determined using digital PCR (dPCR) technology on the QX200 system (Bio-Rad, USA), employing ddPCR EvaGreen Supermix (Bio-Rad, USA) for fluorescence detection with C. difficile specific primers Fcdiff (5‘-TTGAGCGATTTACTTCGGTAAAGA-3’) and Rcdiff (5’-CCATCCTGTACTGGCTCACCT-3’) [25] and tcdB gene specific primers TcdB_398F (5’- GAAAGTCCAAGTTTACGCTCAAT-3’) and TcdB_399R (5’- GCTGCACCTAAACTTACACCA-3’) [26]. The dPCR analysis was performed to quantify the absolute number of copies of C. difficile per ng of DNA. To achieve this, droplet PCR events were normalized to the input DNA concentration, which was assessed using the Quant-iT™ PicoGreen™ dsDNA Assay (ThermoFisher, USA).

Bacterial community structure was acquired through targeted amplicon sequencing of the V3-V4 variable region of the 16S rRNA gene utilizing the primer pair Bakt_341F (5’-CCTACGGGNGGCWGCAG-3’)-Bakt_805R (5’-GACTACHVGGGTATCTAATCC-3’) [17]. Sequencing was conducted on the MiSeq Illumina platform (2 × 300 bp, Illumina, USA). Sequence data processing involved quality filtering of reads and construction of Zero-radius Operational Taxonomic Unit (ZOTUs) using the UNOISE3 pipeline implemented in USEARCH v.11.0.667 [27, 28] with default settings except the addition of parameter -fastq-minlen 400 (fastq_filter command). Taxonomy was assigned using the RDP trainset (version 18). The sequencing data generated in this study are available in the NCBI Sequence Read Archive (SRA) under the accession number PRJNA954698.

Statistical analysis and visualization of the microbiome data was performed in R (v4.2.2, R Core Team, 2022) implementing packages ‘vegan’, ‘picante’ and ‘ggplot’. Alpha diversity was presented as Faith’s phylogenetic diversity calculated based on neighbour-joining tree constructed from ZOTU representative sequences and Shannon diversity index. Beta diversity analysis was conducted using Bray-Curtis dissimilarities. Differences in community composition between groups were evaluated with permutation-based test ANOSIM (R package ‘vegan’).

C. difficile colonization and concentration was analysed in relation to an extensive set of subject metadata. Risk for C. difficile colonization did not significantly correlate with metadata collected, including factors typically associated with CDI in adults such as antibiotic use and hospitalization (Table 1).

C. difficile colonization was in our cohort not associated with specific microbial community structure characteristics. No significant differences were found in either alpha diversity (Shannon diversity, p = 0.235; Faith’s phylogenetic diversity, p = 0.245) or beta diversity (ANOSIM, p = 0.369) (Fig. 1A and B). Furthermore, among C. difficile colonized, no differences between tcdB positive and negative samples were found in either alpha (Shannon diversity, p = 0.980; Faith’s phylogenetic diversity, p = 0.823) or beta diversity (ANOSIM, p = 0.220) (Fig. 1B). Finally, population level analysis did not yield any differentially abundant taxa between C. difficile colonized and non-colonized infants.

On the other hand, we found strong associations between microbiome composition and C. difficile concentration. Higher concentration of C. difficile was significantly positively correlated with the mean Bray-Curtis dissimilarity from the cohort average (Spearman’s rho = 0.595, p = 0.001; Fig. 1C), indicating alterations in community composition, found to be predominantly associated with decreased alpha diversity. While concentration-associated decreases in both richness and diversity were observed (observed ZOTUs, p = 0.016; Shannon diversity, p = 0.017; Shannon evenness, p = 0.044; Supplementary Fig. 1), the correlation was most significant with Faith’s phylogenetic diversity (Spearman’s rho = -0.514, p = 0.005, Fig. 1D). The lowest alpha diversity was observed in samples with highest concentrations of C. difficile, but group minimum values were comparable to those observed in the group of C. difficile non-colonized infants, therefore not indicating any exacerbating effect of C. difficile colonization on dysbiosis-like features in gut microbiota. Among detected ZOTUs, the one linked to C. difficile (ZOTU171) exhibited strongest positive correlation with increasing mean Bray-Curtis dissimilarity (Pearson’s r = 0.427, p < 0.001). Along with negative correlation observed with Fusicatenibacter (ZOTU13) (Pearson’s r = -0.470, p < 0.001), these two represented the only taxa significantly correlated with mean Bray-Curtis dissimilarity beyond false discovery rate (FDR) < 0.05 (Supplementary Fig. 2). C. difficile concentration also positively correlated with the relative abundance of Escherichia, with the correlation most evident at the highest concentrations of C. difficile (Spearman’s rho = 0.585, p = 0.001) (Fig. 1E).