Comprehensive Phenotyping in Inflammatory Bowel Disease: Search for Biomarker Algorithms in the Transkingdom Interactions Context

Abstract

Inflammatory bowel disease (IBD) is the most common form of intestinal inflammation associated with a dysregulated immune system response to the commensal microbiota in a genetically susceptible host. IBD includes ulcerative colitis (UC) and Crohn's disease (CD), both of which are remarkably heterogeneous in their clinical presentation and response to treatment. This translates into a notable diagnostic challenge, especially in underdeveloped countries where IBD is on the rise and access to diagnosis or treatment is not always accessible for chronic diseases. The present work characterized, for the first time in our region, epigenetic biomarkers and gut microbial profiles associated with UC and CD patients in the Buenos Aires Metropolitan area and revealed differences between non-IBD controls and IBD patients. General metabolic functions associated with the gut microbiota, as well as core microorganisms within groups, were also analyzed. Additionally, the gut microbiota analysis was integrated with relevant clinical, biochemical and epigenetic markers considered in the follow-up of patients with IBD, with the aim of generating more powerful diagnostic tools to discriminate phenotypes. Overall, our study provides new insights into data analysis algorithms to promote comprehensive phenotyping tools using quantitative and qualitative analysis in a transkingdom interactions network context.

Keywords: comprehensive-phenotyping; crohn-disease; gut-microbiota; ulcerative-colitis.

Figure 1

Evaluation of miRNAs. hs-miR-155-5p and hs-miR-223-3p expression levels in IBD patients compared to non-IBD controls (A,D). Expression of hs-miR-155-5p and hs-miR-223-3p compared to non-IBD controls in relation to medication received by UC patients or CD patients (B,C,E,F). Wilcoxon test was calculated between groups (* = p-value < 0.05, ** = p-value < 0.01). 5-ASA, 5-aminosalicylic acid (mesalamine); AZA, azathioprine; ADA, Adalimumab biological therapy.

Figure 2

Comparison of microbiota community of the IBD patients and non-IBD control groups. Shannon diversity measures for Alpha Diversity (A,D) and Wilcoxon test were calculated between groups (* = p-value < 0.05, ** = p-value < 0.01) (A). Representation of richness and evenness in a dot plot (D). PCoA plots of beta diversity (Weighted and Unweighted Unifrac distances) for UC patients and non-IBD controls (B,C) and CD patients and non-IBD controls (E,F).

Figure 3

Volcano plot. Differentially abundant genera between non-IBD controls and UC patients (A) or non-IBD controls and CD patients (B). Volcano plot of the differentially abundant metabolisms between non-IBD controls and UC patients (C) or non-IBD controls and CD patients (D). Red dash lines represent a significance threshold (p-value = 0.05). Yellow, green or blue represents significant genera or metabolisms more abundant. In gray, non-significant features are represented.

Figure 4

Heatmap core. Genera core microbiota for non-IBD controls and IBD groups. The heatmap also distinguishes active/remission within the IBD group (A). Venn diagram represents shared/exclusive core genera between groups (B). Venn core microbiota for UC patients (C) and CD patients (D) discerning active and remission patients.

Figure 5

Box plot in the biochemical model. Values of the variables present in patients with CD compared to non-IBD controls (A–C). Performance of CD biochemical model, as assessed via the area under the ROC curve (D).

Figure 6

Box plot in the biochemical model. Values of the variables present in patients with UC compared to non-IBD controls (A–C). Performance of biochemical model, as assessed via the area under the ROC curve (D).

Figure 7

Box plot in the microbiota model. Values of the variables present in patients with CD compared to non-IBD controls (A–C). Performance of microbiota model, as assessed via the area under the ROC curve (D).

Figure 8

Box plot in the microbiota model. Values present in patients with UC compared to non-IBD controls (A–C). Performance of microbiota model, as assessed via the area under the ROC curve (D).

Figure 9

Correlation CD-network. Clinical characteristics age, BMI (gray nodes), biochemical measurements (orange nodes), core microbiota and differentially abundant bacterial taxa (yellow nodes), statistically significant differentially abundant metabolic pathways (p < 0.05; green nodes), and miRNAs from BCF (blue nodes) and FCF (red nodes) of the CD patients. Only correlations with Spearman’s correlation coefficients over 0.7 are shown. Green edges represent positive correlations. Red edges represent negative correlations. Variables from the logistic regression model of CD patients are indicated in red font.

Figure 10

Correlation UC-network. Clinical characteristics, age, BMI (gray nodes), biochemical measurements (orange nodes), core microbiota and differentially abundant bacterial taxa (yellow nodes), statistically significant differentially abundant metabolic pathways (p < 0.05; green nodes), and miRNAs from BFC (blue nodes) and FCF (red nodes) of the UC patients. Only correlations with Spearman’s correlation coefficients over 0.7 are shown. Green edges represent positive correlations. Red edges represent negative correlations. Variables from the logistic regression model of UC patients are indicated in red font.