Altered Faecal Microbiota Composition and Structure of Ghanaian Children with Acute Gastroenteritis

Abstract

Acute gastroenteritis (AGE) is a disease of global public health importance. Recent studies show that children with AGE have an altered gut microbiota relative to non-AGE controls. Yet, how the gut microbiota differs in Ghanaian children with and without AGE remains unclear. Here, we explore the 16S rRNA gene-based faecal microbiota profiles of Ghanaian children five years of age and younger, comprising 57 AGE cases and 50 healthy controls. We found that AGE cases were associated with lower microbial diversity and altered microbial sequence profiles relative to the controls. The faecal microbiota of AGE cases was enriched for disease-associated bacterial genera, including Enterococcus, Streptococcus, and Staphylococcus. In contrast, the faecal microbiota of controls was enriched for potentially beneficial genera, including Faecalibacterium, Prevotella, Ruminococcus, and Bacteroides. Lastly, distinct microbial correlation network characteristics were observed between AGE cases and controls, thereby supporting broad differences in faecal microbiota structure. Altogether, we show that the faecal microbiota of Ghanaian children with AGE differ from controls and are enriched for bacterial genera increasingly associated with diseases.

Keywords: acute gastroenteritis; bacteria; children; correlation network; disease; faecal microbiota; pathogen.

Figure 1

AGE cases are associated with lower alpha diversity and distinct beta diversity relative to healthy controls. (A) Violin plot of alpha diversity metrics. Left to right: Observed ASVs (p < 0.0001, Wilcoxon rank sum test), Shannon (p < 0.0001, Wilcoxon rank sum test), and phylogenetic diversity (PD) (p < 0.0001, Wilcoxon rank sum test). Violin plots show the kernel probability density plots of individual alpha diversity estimates. Boxplots show the median (middle line: 50th percentile), first (bottom: 25th percentile), and third quartiles (top: 75th percentile), and whiskers as 1.5 times the interquartile range. Principal coordinates analysis (PCoA) plots of (B) Bray–Curtis dissimilarity (p < 0.001, PERMANOVA R² = 0.09) and (C) Weighted (p < 0.001, PERMANOVA R² = 0.19) and (D) Unweighted UniFrac metrics (p < 0.001, PERMANOVA R² = 0.12). Percentage variation explained on axes 1 and 2 are shown. Ellipses show the 95% confidence interval for the variation within each group. Each dot represents a faecal microbiota sample from a different individual.

Figure 2

Faecal microbial sequence profiles of AGE cases and healthy controls. Relative abundance of (A) Phylum-, (B) Family-, and (C) Genus-level features. Samples are grouped according to status (AGE cases or healthy controls). Each bar represents an individual faecal microbial sequence profile. Top 10 most dominant taxa are shown, with the remaining collapsed under “Other”.

Figure 3

Core genera of AGE cases and healthy controls. (A) Heatmap of core genera across samples of the two groups. Genus names are shown. (B) Venn diagram of the number of core genera identified within AGE cases and healthy controls. Core genera were identified using a 50% prevalence cut-off and an abundance cut-off of 0.01% (0.0001).

Figure 4

Differential abundance testing identifies genera with high abundance between AGE cases and healthy controls. DESeq2 (left) and ANCOM-BC (right) were used for differential abundance testing. Genera that passed multiple test correction (Benjamini and Hochberg’s FDR < 0.05) are shown. Dots represent the estimated effect size distribution as log2 (fold-change) and W for DESeq2 and ANCOM-BC, respectively. Negative values represent genera enriched in AGE cases and positive values represent genera enriched in healthy controls. Differential abundance testing identifies genera with high abundance between AGE cases and healthy controls. DESeq2 (left) and ANCOM-BC (right) were used for differential abundance testing. Genera that passed multiple test correction (Benjamini and Hochberg’s FDR < 0.05) are shown. Dots represent the estimated effect size distribution as log2 (fold-change) and W for DESeq2 and ANCOM-BC, respectively. Negative values represent genera enriched in AGE cases and positive values represent genera enriched in healthy controls.

Figure 5

Global balance for AGE cases and healthy controls. The two groups of genera that form the global balance are shown at the top of the boxplot, which shows the distribution of balance scores. Boxplots show the median (middle line: 50th percentile), first (bottom: 25th percentile), and third quartiles (top: 75th percentile), and whiskers as 1.5 times the interquartile range. The right section of the figure shows the AUC-ROC curve value of 0.932 and the data density plot for each group.

Figure 6

Correlation network analysis of faecal microbiota. Correlation networks for (A) All groups, (B) AGE cases, and (C) Healthy controls. Module memberships identified by SCNIC are shaded with colours other than black (for instance, shared membership under module-0 is shaded light green on panel A). Nodes (with rectangular/square shapes) represent genera, and edges represent correlations with an R-value greater than 0.35. (D) Differential abundance testing with ANCOM-BC. Genera that passed multiple test correction (Benjamini and Hochberg’s FDR < 0.05) are shown. Dots with negative W values represent genera enriched in AGE cases, while positive values represent genera enriched in healthy controls.