Vitamin A carotenoids, but not retinoids, mediate the impact of a healthy diet on gut microbial diversity

Abstract

Background: Vitamin A is essential for physiological processes like vision and immunity. Vitamin A's effect on gut microbiome composition, which affects absorption and metabolism of other vitamins, is still unknown. Here we examined the relationship between gut metagenome composition and six vitamin A-related metabolites (two retinoid: -retinol, 4 oxoretinoic acid (oxoRA) and four carotenoid metabolites, including beta-cryptoxanthin and three carotene diols).

Methods: We included 1053 individuals from the TwinsUK cohort with vitamin A-related metabolites measured in serum and faeces, diet history, and gut microbiome composition assessed by shotgun metagenome sequencing. Results were replicated in 327 women from the ZOE PREDICT-1 study.

Results: Five vitamin A-related serum metabolites were positively correlated with microbiome alpha diversity (r = 0.15 to r = 0.20, p < 4 × 10^-6). Carotenoid compounds were positively correlated with the short-chain fatty-acid-producing bacteria Faecalibacterium prausnitzii and Coprococcus eutactus. Retinol was not associated with any microbial species. We found that gut microbiome composition could predict circulating levels of carotenoids and oxoretinoic acid with AUCs ranging from 0.66 to 0.74 using random forest models, but not retinol (AUC = 0.52). The healthy eating index (HEI) was strongly associated with gut microbiome diversity and with all carotenoid compounds, but not retinoids. We investigated the mediating role of carotenoid compounds on the effect of a healthy diet (HEI) on gut microbiome diversity, finding that carotenoids significantly mediated between 18 and 25% of the effect of HEI on gut microbiome alpha diversity.

Conclusions: Our results show strong links between circulating carotene compounds and gut microbiome composition and potential links to a healthy diet pattern.

Keywords: Carotenoids; Gut microbiome composition; Retinoids; Vitamin A metabolites.

Fig. 1

Schematic representation of the study. The figure describes the links between dietary intake and the studied vitamin A-related metabolites, and highlights the research questions we are addressing, the data, and the analytical workflow used

Fig. 2

Associations between vitamin A metabolites from serum and stool and the Shannon diversity index (A and B) and the healthy eating index (C and D). Data shown. Circles show effect sizes in the TwinsUK and ZOE PREDICT 1 cohorts, diamonds are summary effect sizes derived from fixed effects meta-analysis (MA) combining both cohorts. Whiskers represent the 95% confidence intervals. Statistical significance is indicated by asterisks as follows: * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001

Fig. 3

A Heatmap representing correlation of vitamin A-related metabolites from serum and gut bacterial species. Each cell of the matrix contains the regression coefficient between one vitamin A-related metabolite and a bacterial species. The table is colour-coded by correlation according to the table legend (red for positive and blue for negative correlations). * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001. B Prediction of the gut microbiota in vitamin A-related metabolites estimated by random forestregressors (using Spearman’s correlations) and classifiers (using AUC values) in TwinsUKand ZOE PREDICT-1 participants. Boxplots represent the mean AUC and the 95% confidence intervals across fivefold for TwinsUK and ZOE PREDICT-1. Dark blue and light blue circles indicate the mean of Spearman’s correlations between the real value of each vitamin A-related metabolite and the value predicted by the models across fivefold in TwinsUK and ZOE PREDICT-1

Fig. 4

Mediation analysis diagram. A healthy diet, represented by the healthy eating index, was modelled as the exposure. Alpha diversity of the gut microbiome, measured by the Shannon diversity index, was modelled as the outcome. Each vitamin A metabolite was modelled as a mediator. Summary statistics from the mediation analysis are depicted as beta (95% confidence interval). Abbreviations: HEI, healthy eating index; CI, confidence interval