Metagenomics reveals impact of geography and acute diarrheal disease on the Central Indian human gut microbiome

Abstract

Background: The Central Indian gut microbiome remains grossly understudied. Herein, we sought to investigate the burden of antimicrobial resistance and diarrheal diseases, particularly Clostridioides difficile, in rural-agricultural and urban populations in Central India, where there is widespread unregulated antibiotic use. We utilized shotgun metagenomics to comprehensively characterize the bacterial and viral fractions of the gut microbiome and their encoded functions in 105 participants.

Results: We observed distinct rural-urban differences in bacterial and viral populations, with geography exhibiting a greater influence than diarrheal status. Clostridioides difficile disease was more commonly observed in urban subjects, and their microbiomes were enriched in metabolic pathways relating to the metabolism of industrial compounds and genes encoding resistance to 3^rd generation cephalosporins and carbapenems. By linking phages present in the microbiome to their bacterial hosts through CRISPR spacers, phage variation could be directly related to shifts in bacterial populations, with the auxiliary metabolic potential of rural-associated phages enriched for carbon and amino acid energy metabolism.

Conclusions: We report distinct differences in antimicrobial resistance gene profiles, enrichment of metabolic pathways and phage composition between rural and urban populations, as well as a higher burden of Clostridioides difficile disease in the urban population. Our results reveal that geography is the key driver of variation in urban and rural Indian microbiomes, with acute diarrheal disease, including C. difficile disease exerting a lesser impact. Future studies will be required to understand the potential role of dietary, cultural, and genetic factors in contributing to microbiome differences between rural and urban populations.

Keywords: Clostridioides difficile; Central India; Gut microbiome; antibiotic resistome; diarrhea; virome.

Figure 1.

Nagpur District. Mapped locations of study participant home residences in Nagpur district.

Figure 2.

Variations in the gut microbiota by geographic location and diarrheal status. (A) Principal coordinates analysis (PCoA) of microbiota profiles based on Bray–Curtis Dissimilarity of species-level taxonomic abundance. Subject profiles vary by both geographic location and diarrheal status. (B) Comparison of microbial diversity between diarrheal and non-diarrheal control subjects from both rural and urban geographic locations. * p.corr = 0.05. (C) Summary of genus-level taxonomic profiles by subject. Subjects are grouped by geographic location and diarrheal status, with diarrheal subjects further subdivided into C. difficile toxin positive (CDT +ve) and negative (CDT – ve). Bacteroides dominant profiles are more frequent in urban subjects, while Prevotella dominant profiles are more frequent in rural subjects. (D) Differentially abundant taxa at species-level based on either geographic location (left, rural vs urban control subjects) or diarrheal status (right, non-diarrheal controls vs diarrheal). All taxa shown are significantly different between groups based on generalized linear models with FDR corrected p < .05.

Figure 3.

Analysis of antimicrobial resistance gene carriage by gut microbiota. (A) Heatmap of antimicrobial resistance (AMR) gene abundance aggregated by antibiotic class. Individual columns show subjects grouped by geography (rural – yellow vs. urban – blue), diarrheal status (non-diarrheal – green vs. diarrheal – red) and antibiotic exposure (brown). Row order represents hierarchical clustering of resistance gene count data using a Euclidean distance matrix. MLS = Macrolides, Lincosamides, and Streptogramins. (B) Heatmap of antimicrobial resistance gene cluster abundance for Beta-lactam antibiotics. Columns represent individual subjects, grouped as above. Individual gene cluster codes are shown in rows corresponding to MegaRes database entries. Beta-lactam resistance mechanisms for each gene cluster are indicated to the left of the heatmap; Ambler class A to D, Porin mutant or PBP (Penicillin Binding Protein). (C) Comparison of the Beta-lactam resistance gene counts which differed significantly between rural and urban subjects. All statistical comparisons between urban and rural subjects were made with the Mann–Whitney U test with FDR correction and results indicated in each panel. * p < .05, ** p < .01, *** p < .001.

Figure 4.

Taxonomic contributions to differentially enriched metabolic pathways. The top 10 pathways enriched in either urban or rural subjects are shown with the predicted contribution of individual taxa to the overall pathway variance (red diamonds). For each pathway, the top and bottom bars indicate urban- and rural-associated taxa, respectively, displaying the predicted contribution of each taxon to enrichment in either group; urban (positive) or rural (negative). For example, enrichment of Lipoic acid metabolism in urban subjects is associated with the positive contribution (A) of Klebsiella pneumoniae (Kp), Parabacteroides distasonis (Pd) and Bacteroides vulgatus (Bv), with only minor negative contributions from multiple other species (B). Rural-associated taxa contributing to enrichment in urban subjects (C), most likely because they encode the function sparsely, include Prevotella copri (Pc) and Eubacterium rectale (Er). Prevotella stercorea (Ps) is predicted to enrich this pathway in rural subjects (D), acting against the total observed shift.

Figure 5.

Contrasting fecal viromes by geographic location and diarrheal status. (A) Network visualization of viral clustering. Viral clusters (VCs) containing previously characterized viral sequences (viral RefSeq 85) are colored by International Committee on Taxonomy of Viruses (ICTV) family-level taxonomic assignments. While Microviridae VCs are connected to Caudovirales through shared protein clusters, these taxa are unrelated. (B) Inverse Simpson diversity comparisons of subjects by diarrheal status and geographic location. (C) Principal coordinate analysis of VC profiles based on Bray–Curtis Dissimilarity. (D) The fold change (log10) of the top 25 most abundant rural and urban VCs, with superimposition of the same VC’s association with either health or diarrheal status. (E) The fold change (log10) of all VCs relative abundance that is targeted by CRISPR spacers from identifiable bacterial genera. Each point represents a VC, with size representing the aggregate number of CRISPR spacers targeting individual viruses within a cluster.

Figure 6.

Examination of the auxiliary metabolic potential of human fecal viruses. (A) Shared proteins encoded by Viral Clusters (VCs) shared amongst 10 or more individuals within this study. (B) The VC-encoded metabolic functions were determined per individual virome, with the similarities between subjects visualized by principal coordinate analysis using the Jaccard index. (C) Relative abundance comparisons of the protein categorical-function predictions of VCs by residence. (D & E) The observed frequency of amino acid transport and metabolism functions, and carbohydrate transport and metabolism functional predictions encoded by individual virome VCs. Only statistically significant EggNOG functional predictions are displayed (Mann–Whitney U test with Bonferroni correction, p adj = 0.05).