Gut microbial ecology and exposome of a healthy Pakistani cohort

Abstract

Background: Pakistan is a multi-ethnic society where there is a disparity between dietary habits, genetic composition, and environmental exposures. The microbial ecology of healthy Pakistani gut in the context of anthropometric, sociodemographic, and dietary patterns holds interest by virtue of it being one of the most populous countries, and also being a Lower Middle Income Country (LMIC).

Methods: 16S rRNA profiling of healthy gut microbiome of normo-weight healthy Pakistani individuals from different regions of residence is performed with additional meta-data collected through filled questionnaires. The current health status is then linked to dietary patterns through [Formula: see text] test of independence and Generalized Linear Latent Variable Model (GLLVM) where distribution of individual microbes is regressed against all recorded sources of variability. To identify the core microbiome signature, a dynamic approach is used that considers into account species occupancy as well as consistency across assumed grouping of samples including organization by gender and province of residence. Fitting neutral modeling then revealed core microbiome that is selected by the environment.

Results: A strong determinant of disparity is by province of residence. It is also established that the male microbiome is better adapted to the local niche than the female microbiome, and that there is microbial taxonomic and functional diversity in different ethnicities, dietary patterns and lifestyle habits. Some microbial genera, such as, Megamonas, Porphyromonas, Haemophilus, Klebsiella and Finegoldia showed significant associations with consumption of pickle, fresh fruits, rice, and cheese. Our analyses suggest current health status being associated with the diet, sleeping patterns, employment status, and the medical history.

Conclusions: This study provides a snapshot of the healthy core Pakistani gut microbiome by focusing on the most populous provinces and ethnic groups residing in predominantly urban areas. The study serves a reference dataset for exploring variations in disease status and designing personalized dietary and lifestyle interventions to promote gut health, particularly in LMICs settings.

Fig. 1

Core microbiome identified through species occupancy abundance diagrams. A stringent occupancy criteria A, B, C is incorporated where we clump all the samples in a single site (no site specific occupancy), and then calculate the ranking of ASVs based on their occupancy and replicate consistency within a single category. Alternatively, we calculate the occupancy and replicate consistency of these ASVs separately (site specific occupancy) for each site where for D, site represents province of residence for males; for E, site represents province of residence for female; and for F, site represents the gender. Once we have obtained the rankings depending on which criteria used, Bray–Curtis similarity is calculated for the whole dataset, and then also for only the top-ranked taxa. The contribution of the top-ranked taxa is divided by the total Bray–Curtis similarity to calculate a percent contribution of the prospective core set to beta diversity. The next-ranked taxon is added consecutively to find the point in the ranking at which adding one more taxon offers diminishing returns on explanatory value for beta diversity (G, H, I). The red line represents the stringent “Elbow approach” where the change is maximal between the left and right side of dotted red threshold in terms of first-order differences, and “Last 2% decrease” criteria where ASVs are incorporated in the core subset until there is no more than 2% decrease in beta diversity. In this study, we are only identifying core microbiome (red, green and blue points) using “Last 2% decrease” criteria. Independently, a neutral model is fitted with those ASVs that fall within the 95% confidence interval (shown in green), and those that fall outside the 95% model confidence to be inferred as deterministically assembled, i.e., non-neutral ASVs. Points above the model are selected by the (host) environment (shown in red), and points below the model are dispersal limited (shown in blue). The proportion of core ASVs belonging to different phyla are then shown with a pie chart whilst the count of neutral/non-neutral ASVs are shown with the bar plots

Fig. 2

$\mathit{\beta }$ coefficients returned from GLLVM procedure for covariates considered in this study by considering top 100 most abundant genera incorporating both continuous as well as categorical labelling of samples. Those coefficients which are positively associated with the microbial abundance of a particular species are represented in red colour whilst those that are negatively associated are represented with blue colour, respectively. Where the coefficients are non-significant, i.e., the 95% confidence interval crosses the 0 boundary, they are greyed out. Since the collation of ASVs was performed at Genus level, all those ASVs that cannot be categorized based on taxonomy are collated under “__Unknowns__” category. The acronyms are as follows: (ICT Islamabad Capital Territory, AJK Azad Jammu & Kashmir, KPK Khyber Pakhtunkhwa, LC Lower Class, LMC Lower Middle Class, MC Middle Class, UMC Upper Middle Class). Note that the GLLVM procedure additionally calculates the residual covariance matrix of the latent variables in the model which gives an additional co-occurrence relationship between microbes, and is given in Additional file 1: Fig. S47

Fig. 3

$\mathit{\beta }$ coefficients for covariates categorized under dietary items

Fig. 4

$\mathit{\beta }$ coefficients for covariates categorized under intake frequency of selected dietary items