Bacterial clade-specific analysis identifies distinct epithelial responses in inflammatory bowel disease

Abstract

Abnormal immune responses to the resident gut microbiome can drive inflammatory bowel disease (IBD). Here, we combine high-resolution, culture-based shotgun metagenomic sequencing and analysis with matched host transcriptomics across three intestinal sites (terminal ileum, cecum, rectum) from pediatric IBD (PIBD) patients (n = 58) and matched controls (n = 42) to investigate this relationship. Combining our site-specific approach with bacterial culturing, we establish a cohort-specific bacterial culture collection, comprising 6,620 isolates (170 distinct species, 32 putative novel), cultured from 286 mucosal biopsies. Phylogeny-based, clade-specific metagenomic analysis identifies key, functionally distinct Enterococcus clades associated with either IBD or health. Strain-specific in vitro validation demonstrates differences in cell cytotoxicity and inflammatory signaling in intestinal epithelial cells, consistent with the colonic mucosa-specific response measured in patients with IBD. This demonstrates the importance of strain-specific phenotypes and consideration of anatomical sites in exploring the dysregulated host-bacterial interactions in IBD.

Keywords: Enterococcus; cell death; epithelial cells; gastrointestinal tract; host-microbe interactions; inflammatory bowel disease; metagenomic; microbiome; pediatric IBD; ulcerative colitis.

Graphical abstract

Figure 1

Cohort-specific bacterial culture collection containing 6,620 isolates, derived from intestinal mucosal biopsies Phylogenetic tree of 16S rRNA sequences (average sequence length: 1,104 base pairs) from 6,620 isolates (green: Bacteroidetes, 3,294; blue: Firmicutes, 1,553; yellow: Actinobacteria, 382; red: Proteobacteria, 1,369; orange: Fusobacteria, 18; and purple: Verrucomicrobia: 4), representing 204 distinct species. Isolates were cultured directly from 286 mucosal samples, obtained from 100 patients (PIBD: 58 patients, control: 42 patients) across three intestinal regions (terminal ileum: 93 samples; cecum: 96 samples; rectum: 97 samples) and includes 745 isolates that have been whole-genome sequenced (black bars surrounding the phylogenetic tree). Tree scale represents substitutions per site.

Figure 2

Integration of metagenomic sequencing and transcriptional profiling, from matched mucosal samples, allows for identification of bacterial clades associated with upregulation of inflammatory signatures in patient-derived mucosal samples (A) Relative Hallmark gene set scores grouped by state of sample inflammation, defined by k-means clustering using the Hallmark inflammatory response gene set. The values represented are relative to the median value for each gene set. N, molecularly non-inflamed; I, molecularly inflamed (∗∗∗p < 0.0001; Welch’s two-sample t test, ns, non-significant; n = 231 mucosal samples). (B) Correlations between the Hallmark gene set scores and the microbial species-level clade abundance selected via a sparse partial least squares (sPLS) approach from 231 mucosal samples. The color gradient indicates the spectrum between positive (red) and negative (blue) correlations.

Figure 3

In vitro phenotypic validation of candidate bacteria demonstrates a clade-specific difference in cell cytotoxicity (A) Cell cytotoxicity at 24 h post-stimulation, assessed by the LDH assay, in Caco2 intestinal epithelial cells stimulated with one of twelve isolates representing the IBD-associated clade (6 isolates; purple) or the control-associated clade (6 isolates; orange) at a multiplicity of infection (MOI) of 10:1 (n = 3 biological replicates, mean ± SEM). Relationship between bacterial clades is represented by the whole-genome phylogenetic tree below the graph. Statistics represent a comparison between IBD-associated and the control-associated clade (p = 0.0006; ratio paired t test by experiment). Scanning electron microscopy images of each isolate allow for morphological visualization of each bacteria isolate. Black bar indicates 500 nm. (B) Heatmap showing genes significantly differentially expressed between isolate CC00064 or CC00149 and the non-stimulated control (NS) at 4, 8, or 24 h post-stimulation. Color scale represents log₂ fold changes, truncated to ±4. Each isolate at each time point represents the average of n = 3 biological replicates.

Figure 4

In vitro clade-specific gene signatures present in inflamed patient-derived biopsies (A) Genes for which there was significant differential expression in the in vitro Caco2 experiment between isolates CC00149 and CC00064 at 8 h were compared with the differential expression results between the inflamed and non-inflamed mucosal samples from each intestinal region. Vertical lines represent genes in the set that showed upregulation in strain CC00149 compared with CC00064 (weighted by log₂ fold change). Genes are positioned according to the moderated t statistic in the clinical group of interest, with highly upregulated genes positioned on the far right, while downregulated genes appear on the far left. Overall gene set enrichment is represented by the enrichment line shown above. (B) Heatmap showing all genes that were significantly differentially expressed between CC00149 and CC00064 at 8 h post-stimulation that were also significantly changed in the same direction, between the molecularly inflamed compared with non-inflamed samples, within any of the three clinical sample sites. Significance (p < 0.05) is represented by the asterisk (∗) in the box with n = 3 biological replicates in each group. Values displayed are log₂ fold changes, truncated to ±3.