Gluten-free diet affects fecal small non-coding RNA profiles and microbiome composition in celiac disease supporting a host-gut microbiota crosstalk

Figure 1.. a) Workflow of the study. b) Violin plots reporting the estimated daily intake of nutrients significantly different among the investigated groups. The nutrient intake was normalized on the daily intake of kilocalories estimated for each subject. Wilcoxon Rank-Sum test: *p < .05; **p < .01; ***p < .001; tCD, treated Celiac Disease; TG, transglutaminase 2; GFD, gluten-free diet; sncRNAs, small non-coding RNAs.

Figure 2.. a) Upset plot reporting the number of differentially expressed miRNAs (DEmiRNAs) in each comparison performed among the study groups. b) Scatterplot relating the DEmiRNA log2 Fold Changes (log2FCs) from the comparison between the tCD groups and age- and sex-matched healthy controls (x-axis) or an independent group of healthy controls (y-axis) from.16 The color code represents the comparison in which each DEmiRNA was detected. c) Heat map of log2FCs of the DEmiRNAs in the comparison groups described in the study. c1 to c4 labels depict four clusters from the hierarchical clustering analysis. d) Correlation analyses between DEmiRNA levels and years of gluten-free diet (GFD) in tCD-TG- (left) and all tCD (middle), or subjects’ age (right). *p < .05; **p < .01. e) Dot plots showing the five most representative Gene Ontology Biological Processes enriched in the functional analysis of the DEmiRNAs belonging to each of the clusters reported in panel c. The dot size is proportional to the significance, while the color code refers to the RBiomirGS coefficient. Negative and positive coefficients are related to processes predicted to be, respectively down- and up-regulated based on the…

Figure 3.. A-B) Plots showing the differentially abundant bacteria by comparing tCD-TG- (a) or tCD-TG+ (b) with healthy controls. For each study group, on the left are reported the microbial relative abundances and on the right the prevalence shift. Wilcoxon Rank-Sum test: *adj. p < .05; **adj. p < .01; ***adj. p < .001. C) Heat map representing the log2FCs of microbial metabolic pathways associated with a significantly different prevalence among the comparisons performed. Wilcoxon Rank-Sum test: *p < .05; **p < .01; ***p < 0.001 . On the right are reported the correlation coefficients computed between the pathway prevalence and the years of GFD or tCD subjects’ age: *p < .05.

Figure 4.. a) Sample plots from DIABLO analysis showing the contribution of stool DEmiRNAs, microbial species, or nutrients in distinguishing both CD categories and controls. The samples are represented as dots plotted according to their projection in the subspace spanned by the latent variables defined by integrating the three data types. b) Correlation between variate 1 (top) and variate 2 (bottom) defined for each data type. c) Network representation of the significant correlations (adj. p < .05) between DEmiRNAs and bacteria identified as discriminating features in the DIABLO analysis. The node size is proportional to their degree, while edge color and width are related to the correlation coefficients.