Sex-Specific Associations between Gut Prevotellaceae and Host Genetics on Adiposity

Abstract

The gut microbiome has been recognized as a tool for understanding adiposity accumulation and for providing personalized nutrition advice for the management of obesity and accompanying metabolic complications. The genetic background is also involved in human energy homeostasis. In order to increase the value of nutrigenetic dietary advice, the interplay between genetics and microbiota must be investigated. The purpose of the present study was to evaluate interactive associations between gut microbiota composition and 95 obesity-related single nucleotide polymorphisms (SNPs) searched in the literature. Oral mucosa and fecal samples from 360 normal weight, overweight and obese subjects were collected. Next generation genotyping of these 95 SNPs and fecal 16S rRNA sequencing were performed. A genetic risk score (GRS) was constructed with 10 SNPs statistically or marginally associated with body mass index (BMI). Several microbiome statistical analyses at family taxonomic level were applied (LEfSe, Canonical Correspondence Analysis, MetagenomeSeq and Random Forest), and Prevotellaceae family was found in all of them as one of the most important bacterial families associated with BMI and GRS. Thus, in this family it was further analyzed the interactive association between BMI and GRS with linear regression models. Interestingly, women with higher abundance of Prevotellaceae and higher GRS were more obese, compared to women with higher GRS and lower abundance of Prevotellaceae. These findings suggest relevant interrelationships between Prevotellaceae and the genetic background that may determine interindividual BMI differences in women, which opens the way to new precision nutrition-based treatments for obesity.

Keywords: genetic risk score; gut microbiome; metagenomics; nutrigenetics; obesity.

Figure 1

Linear discriminant analysis effect size between normal weight and overweight+obese subjects. Red, bacterial taxa statistically overrepresented in normal weight participants; green, bacterial taxa overrepresented in overweight and obese volunteers.

Figure 2

Linear discriminant analysis effect size between subjects with high Genetic Risk Score and subjects with low Genetic Risk Score Red, bacterial taxa statistically overrepresented in high Genetic Risk Score participants; green, bacterial taxa overrepresented in low Genetic Risk Score participants.

Figure 3

Biplot of the analysis of multivariate ordination with canonical correspondence analysis (CCA) applied to taxonomic abundance comparison at family level (including families obtained by LEfSe). CCA was performed to assess the variance in microbiota profiles at the family level in lean and overweight+obese, high GRS and low GRS and men and women. Vectors represent the environmental variables (in red) and black points represent abundance of families.

Figure 4

(A) Box plot of Prevotellaceae abundance in metagenomeSeq analysis between normal weight and overweight+obese subjects. (B) Box plot of Prevotellaceae abundance in metagenomeSeq analysis between low-GRS and high-GRS subjects. FDR: false discovery rate adjusted p value.

Figure 5

Ranking of mean decrease in accuracy (MDA) values in random forest analysis, a statistical classification that indicates the importance of each variable. Random forest calculates feature importance by removing each feature from the model and measuring the decrease in accuracy (for presence) or the increase in the mean-square error (for abundance). According to these importance scores, features were ranked in increasing order across models. The plot shows each variable on the y-axis, and their importance on the x-axis. Thus, the most important variables are at the top and an estimate of their importance is given by the position of the dot on the x-axis. (A) The figure shows the hierarchical rank of 30 families listed as responsible for the differences between groups of BMI. (B) The graph shows the hierarchical rank of importance of 30 bacterial families implicated in the differences between groups of GRS.

Figure 6

Predicted values of BMI in all population studied (A), men (B) and women (C) according to the Genetic Risk Score (GRS) calculated with 5 (dash line and square dots) and 15 (solid line and circle dots) relative abundance of Prevotellaceae, using linear regression models adjusted for age, sex, physical activity and energy intake, showing association between relative abundance of Prevotellaceae family and GRS.