High-resolution analysis of the treated coeliac disease microbiome reveals strain-level variation

Abstract

Background: Coeliac disease (CeD) is an immune-mediated disorder primarily affecting the small intestine, characterized by an inflammatory immune reaction to dietary gluten. CeD onset results from a multifaceted interplay of genetic and environmental factors. While recent data show that alterations in gut microbiome composition could play an important role, many current studies are constrained by small sample sizes and limited resolution.

Methods: To address these limitations, we analyzed fecal gut microbiota from two Dutch cohorts, CeDNN (128 treated CeD patients (tCeD), 106 controls) and the Lifelines Dutch Microbiome Project (24 self-reported tCeD, 654 controls), using shotgun metagenomic sequencing. Self-reported IBS (570 cases, 1710 controls) and IBD (93 cases, 465 controls) were used as comparative conditions of the gastrointestinal tract. Interindividual variation within the case and control groups was calculated at whole microbiome and strain level. Finally, species-specific gene repertoires were analyzed in tCeD patients and controls.

Results: Within-individual microbiome diversity was decreased in patients with self-reported IBS and IBD but not in tCeD patients. Each condition displayed a unique microbial pattern and, in addition to confirming previously reported microbiome associations, we identify an increase in the levels of Clostridium sp. CAG:253, Roseburia hominis, and Eggerthella lenta, amongst others. We further show that the observed changes can partially be explained by gluten-free diet adherence. We also observe increased interindividual variation of gut microbiome composition among tCeD patients and a higher bacterial mutation frequency in tCeD that contributes to higher interindividual variation at strain level. In addition, the immotile European subspecies of Eubacterium rectale, which has a distinct carbohydrate metabolism potential, was nearly absent in tCeD patients.

Conclusion: Our study sheds light on the complex interplay between the gut microbiome and CeD, revealing increased interindividual variation and strain-level variation in tCeD patients. These findings expand our understanding of the microbiome's role in intestinal health and disease.

Keywords: Coeliac disease; gluten; gluten-free diet; gut microbiome; metagenomic sequencing.

Figure 1.

Overlap in microbiome changes between treated coeliac disease (tCeD), inflammatory bowel disease (IBD), and irritable bowel syndrome (IBS). Bacterial species with (A) higher abundance or prevalence in disease and (B) lower abundance or prevalence in disease. Vertical bars indicate the number of species affected in the disease groups, as indicated by the black dots below each respective bar. Horizontal black bars at bottom left indicate the total number of species affected in each disease group.

Figure 2.

Spearman correlation coefficients between GFD duration and species abundance for all the species significantly enriched in tCeD patients. Nominal (p < 0.05) and corrected (p_adj < 0.05) significance are indicated in light and dark blue, respectively.

Figure 3.

Beta diversity in tCeD vs. controls using Bray–Curtis dissimilarity as a distance metric. a. Principal coordinate plot of tCeD samples and matched controls. b. Within-group dissimilarity histograms for tCeD (top), IBD (middle), and IBS (bottom). Blue: pairs of unrelated controls. Orange: pairs of unrelated patients. Values at the left are the number of different permutations of two samples that fall into each group.

Figure 4.

Distribution of pairwise genetic distances between strains of (a) Bacteroides ovatus, (b) Bacteroides uniformis, (c) Bacteroides xylanisolvens, and (d) Eubacterium rectale. Control vs. control (top) and tCeD vs. tCeD (bottom). Pairs of relatives were excluded from both comparisons. The significance of increased distance among tCeD patients is indicated by p-values from permutation tests (1000 iterations).

Figure 5.

Strain-level differences of Eubacterium rectale between tCeD patients and controls. a. Abundance and prevalence of E. rectale in patients and controls. Violin plots represent the distribution of abundance (log-scale) in the two groups. Samples in which E. rectale was absent are excluded from the plot. Shaded bars in the background represent prevalence among all samples. b. Non-metric multidimensional scaling plots of pairwise genetic distances between E. rectale strains. Symbol shapes and colours indicate their assigned cluster (PAM clustering). Ellipses include 95% of the samples in the indicated group. c. Heatmap of the presence/absence of E. rectale genes in all the E. rectale strains analysed. Each row represents an E. rectale strain from one participant. CeD patients are indicated by the orange lines at the left of the heatmap. Each column represents a UniRef90 cluster (gene) that is differentially prevalent in tCeD patients (nominal p < 0.05). Coloured bars below indicate whether the genes are predicted to encode glycosyltransferases (green) or flagellum-related proteins (pink). Black bars at bottom (FDR-significant) indicate the genes with p_adj < 0.05.

Figure 6.

Correlation between prevalence in a specific sample of genes encoding the RadC-like JAB domain (‘# genes’) and the median mutational distance (StrainPhlan) from other strains within the same sample group for (a) Bacteroides ovatus, (b) Bacteroides uniformis, (c) Bacteroides vulgatus, and (d) Bacteroides xylanisolvens. Significance within sample groups is indicated by the p-values inside the plot. The p-values in the plot title report the association in both groups combined (linear mixed-effects model).