Duodenal transcriptomics demonstrates signatures of tissue inflammation and immune cell infiltration in children with environmental enteric dysfunction across global centers

Abstract

Background: Environmental enteric dysfunction (EED) is an inflammatory condition of the small intestine that is prevalent in children residing in low- and middle-income countries. EED is accompanied by profound histopathologic changes in the small bowel, loss of absorptive capacity, increased intestinal permeability, increased microbial translocation, and nutrient loss.

Objectives: We sought to identify dysregulated genes and pathways that might underlie pediatric EED.

Methods: RNA-sequencing libraries were generated from endoscopically obtained duodenal tissue from undernourished children with EED from 3 prospective cohorts of children with EED. The EED transcriptome was defined in comparison to North American children without EED. Weighted gene coexpression network analysis (WGCNA) was tested for gene modules associated with EED and its histologic features.

Results: The 1784 upregulated genes in EED were highly enriched for immune and inflammatory processes, including IL-17 and JAK-STAT signaling, and cytokine-cytokine receptor interactions. The 1388 downregulated genes included genes corresponding to xenobiotic metabolism, detoxification, and antioxidant capacities. A gene coexpression module enriched for antimicrobial responses and chemokine activity was significantly associated with villous blunting, goblet cell depletion, and overall histologic severity of EED.

Conclusions: The transcriptome signatures of EED include specific innate and adaptive immune responses that are consistently elevated across study centers, coupled with reduced detoxification and antioxidant capacities. These data may have implications for targeted interventions to improve EED outcomes.

Keywords: RNA-sequencing; WGCNA; biopsy; environmental enteric dysfunction; environmental enteropathy; small intestine.

FIGURE 1

The EED transcriptome. (A) Volcano plot of differentially expressed genes in EED. Downregulated, upregulated, and nonsignificant genes are shown in green, red, and gray plots, respectively. Genes are plotted according to the LFC (X-axis) and the -log10 (P_adj) (Y-axis). (B) Bland–Altman plot showing overall expression (X-axis) relative to LFC (Y-axis) in EED. (C) The number of differentially expressed genes corresponding to specific biotypes in EED. Upregulated genes in EED are shown in red, downregulated genes are shown in green. TR, T-cell receptor; IG, immunoglobulin. (D) Gene set enrichment of up and downregulated genesets in EED. Enriched categories (Y-axis) colored by source and normalized enrichment scores (X-axis) are shown.

FIGURE 2

Gene coexpression modules in EED. (A) Heatmap of gene modules associated with the EED histologic index score and component histologic scores. Colors correspond to the Pearson correlation coefficient. P values adjusted for the number of comparisons using a Bonferroni correction are shown in parentheses. Correlations with an absolute value <0.2 were omitted for simplicity (complete results are in Supplemental Data set 3). (B) Expression of hub genes in EED (gray n = 141) and non-EED (black n = 39) cohorts. (C–L) Relationship between module hub gene expression and corresponding histologic scores. Coefficients were estimated using mixed-effects linear regression, adjusting for site, and RNA counts are log transformed. Only subjects with histologic scores were included (TSP-5 = 127, villus architecture = 110, goblet cell score = 141, Paneth cell score = 107, and IEL = 141).

FIGURE 3

TheLCN2module contains coregulated antimicrobial genes. (A) Heatmap of differentially expressed genes in the LCN2 module. (B) ToppGene enrichment analysis of the LCN2 module. Dot size corresponds to the number of genes in the enriched category. Dot color indicates the database source. Enrichment was calculated as [# of genes in query geneset/# of genes in genome]. (C) Network plot of LCN2 gene module protein–protein interactions. Node color corresponds to LFC values in EED.

FIGURE 4

Expression of genes related to the histologic features of EED. The relationships between the expression of selected gene markers (Y-axis) and their respective EED histopathologic feature (X-axis) were evaluated in the EED and comparison cohorts by a linear mixed-effects model. Samples without histologic scores were omitted. RNA counts were log transformed and exponentiated coefficients and 95% CIs are shown on each plot. (A) IEL infiltration score and CD3E expression (n =180). (B) Goblet cell depletion score and MUC2 expression (n = 180). (C) Paneth cell depletion score and LYZ expression (n = 180). (D) Brunner’s gland score and MUC6 expression (n = 178). (E) Villus architecture score and ALPI expression (n = 146).

FIGURE 5

Relationship between quantitative immunohistochemical markers of EED and small intestinal gene expression. The relationship between the transcript abundances (X-axis) and quantitative immunohistochemical measure of protein (Y-axis) in paired biopsies was evaluated using a linear mixed-effects model. Symbols correspond to the study group. Transcript abundances and IHC measures are log transformed and coefficients and 95% CIs are shown on each plot. Only samples with IHC values are included. (A) CD19 (n = 132); (B) CD3 (n = 132); (C) CD45 (n = 108); (D) CXCL10 (n = 131); (E) GZMB (n = 118); (F) LCN2 (n = 119); (G) DEFA5 (n = 130); (H) MUC2 (n = 132); (I) REG1B (n = 131); (J) SLC15A1 (n = 111); (K) SI (n = 132); and (L) DUOX2 (n = 118).