Gut bacterial taxonomic abundances vary with cognition, personality, and mood in the Wisconsin Longitudinal Study

Abstract

Animal studies have shown that the gut microbiome can influence memory, social behavior, and anxiety-like behavior. Several human studies show similar results where variation in the gut microbiome is associated with dementia, depression, and personality traits, though most of these studies are limited by small sample size and other biases. Here, we analyzed fecal samples from 313 participants in the Wisconsin Longitudinal Study, a randomly selected population-based cohort of older adults, with measured psycho-cognitive dimensions (cognition, mood, and personality) and key confounders. 16s V4 sequencing showed that Megamonas is associated with all measured psycho-cognitive traits, Fusobacterium is associated with cognitive and personality traits, Pseudoramibacter_Eubacterium is associated with mood and personality traits, Butyvibrio is associated with cognitive traits, and Cloacibacillus is associated with mood traits. These findings are robust to sensitivity analyses and provide novel evidence of shared relationships between the gut microbiome and multiple psycho-cognitive traits in older adults, confirming some of the animal literature, while also providing new insights. While we addressed some of the weaknesses in prior studies, further studies are necessary to elucidate temporal and causal relationships between the gut microbiome and multiple psycho-cognitive traits in well-phenotyped, randomly-selected population-based samples.

Keywords: Cohort; Depression; Five-factor; Gut bacteria; Gut-brain; Memory.

Fig. 1

[1 column]. Pairwise Pearson’s correlation coefficient estimates between outcomes.

Fig. 2

[2column]. Phylogenetic distribution of P-values testing that linear model coefficients for each psycho-cognitive dimension equal zero. Tests are based on regressing taxonomic counts collapsed at the (A) genera-, (B) species-, and (C) family-level, on each psycho-cognitive dimension, adjusted for all confounders. Taxa are labelled if 3 or more traits had p ​< ​1e-4.

Fig. 3

[2-column]. Volcano plots highlighting the strongest associations with each phenotype. Log-fold-change (logFC) estimates are displayed for taxonomic counts collapsed at the family, genus, and species level, regressed on (A) cognition, (B) mood, and (C) personality variables, adjusted for all confounders. Taxa are labelled if the logFC had absolute value greater than 2.

Fig. S1

Phylogenetic distribution of P-values testing that linear model coefficients for each psycho-cognitive dimension equal zero, in the entire sample and adjusted for all confounders except self-reported diagnoses of hypertension, cardiovascular disease, and diabetes. Taxa are labelled if 3 or more traits had p ​< ​1e-4.

Fig. S2

Phylogenetic distribution of P-values testing that linear model coefficients for each psycho-cognitive dimension equal zero, in the sample restricted to graduates (not siblings) and adjusted for all confounders. Taxa are labelled if 3 or more traits had p ​< ​1e-4.