Bacterial associations in the healthy human gut microbiome across populations

Abstract

In a microbial community, associations between constituent members play an important role in determining the overall structure and function of the community. The human gut microbiome is believed to play an integral role in host health and disease. To understand the nature of bacterial associations at the species level in healthy human gut microbiomes, we analyzed previously published collections of whole-genome shotgun sequence data, totaling over 1.6 Tbp, generated from 606 fecal samples obtained from four different healthy human populations. Using a Random Forest Classifier, we identified 202 signature bacterial species that were prevalent in these populations and whose relative abundances could be used to accurately distinguish between the populations. Bacterial association networks were constructed with these signature species using an approach based on the graphical lasso. Network analysis revealed conserved bacterial associations across populations and a dominance of positive associations over negative associations, with this dominance being driven by associations between species that are closely related either taxonomically or functionally. Bacterial species that form network modules, and species that constitute hubs and bottlenecks, were also identified. Functional analysis using protein families suggests that much of the taxonomic variation across human populations does not foment substantial functional or structural differences.

Figure 1

‘Abundant cores’ and Signature Species. (a) All cohorts exhibit a bimodal distribution for species prevalence. Species that are prevalent in 90% or more samples within a cohort is considered a member of that cohort’s ‘abundant core’. (b) The proportion of total sample relative abundance each cohort’s ‘abundant core’ species and the union of all ‘abundant cores’ species (i.e., Signature Species/Sig). The ‘abundant core’ microbiota is shown to account for the bulk of reads mapped within each sample. Each dot represents a sample from that cohort. (c) PCA demonstrating the lack of distinct clustering of samples from different cohorts based on the CLR-transformed relative abundance data of the signature species. Samples from the Indian and American cohorts appear to separate from the rest of the cohorts however, samples from the other two cohorts demonstrate little separation. The diamonds indicate cluster centroids.

Figure 2

Species-level bacterial association networks. Network modeling of associations between (173/202) signature species within each network. A total of 29 species were not shown as they had zero edges in all networks. Node color designates the phylum each species belongs to, node size is reflective of node degree, and edge color represents if the association is positive (blue) or negative (orange). Nodes are ordered counterclockwise around the circle by the alphabetical order of the concatenated string of all taxonomic levels. Nodes that are numbered correspond to species with the highest degree centrality within modules, designated as “hubs”. Brackets around [Bacteroides] pectinophilus indicate that it is misclassified (i.e., placed incorrectly in a higher taxonomic rank and awaiting to be formally renamed). We utilized Blast to designate [Bacteroides] pectinophilus as belonging to the phylum Firmicutes. For a full list of species shown and not shown within network models see Supplemental.

Figure 3

Cohort network association analysis. (a) The distribution of bacterial association weights within each cohort’s network, dots (black and yellow) and (n) represent total associations. Yellow dots represent species associations that were found shared across all networks. (b) The proportion of associations within each cohort’s network that are unique (red) or shared (blue) with at least one other network. (c) Sub-graph displaying only the 20 conserved nodes (species) and 14 edges (associations) retained across all cohorts.

Figure 4

Taxonomic and functional relationships between species. (a) Proportion of associations within each cohort’s network that are either positive or negative at the lowest level of taxonomic relation (n = total associations). Most positive associations appear between taxonomically similar species. (b) Association weight vs Bray–Curtis distance of genome functional profiles between network partners. Positive associations between functionally similar species are both common and greater in strength than negative associations. There appears to be a minimal distance between genome functional profiles before a negative association is demonstrated. (c) An asynchronous LPA was used to analyze the modules composing the networks of each cohort. Each dot represents the aggregated TIGRFAM profiles of an individual module found by aLPA and the diamonds represent the cohort centroids. Four distinct clusters were found, and each cohort was represented within each cluster. The American cohort appears to be biased towards Cluster IV, however the other cohorts do not appear overtly biased to any one cluster.

Figure 5

Pie plots of the cluster taxonomy. Pie plots demonstrating genus-level taxonomic compositions within each of the module clusters. Clusters were determined using PCA of module functional profiles for each module. (a) Cluster I is dominated by members of the Streptococcus and Bifidobacterium genera and no genus represents less than 3% relative abundance. (b) Members of the Bacteroides genus are also the most abundant in the Cluster II, however the Prevotella and Allistipes genera are also abundant and account for > 70% of abundance when combined with Bacteroides. There are 6 genera with relative abundances below 3%. (c) Members of the Bacteroides genus are the most abundant in the Cluster III and there are 49 genera with relative abundances below 3%. (d) There are only 5 genera above 3% relative abundance and 44 genera below 3% with no one genus showing greater than 15% relative abundance. Genera with < 3% relative abundance were placed in the ‘Others’ category.