Variety of Fruit and Vegetables and Alcohol Intake are Associated with Gut Microbial Species and Gene Abundance in Colorectal Cancer Survivors

Abstract

Background: Adherence to the American Cancer Society (ACS) guidelines of avoiding obesity, maintaining physical activity, and consuming a diet rich in fruits, vegetables, and whole grains is associated with longer survival in colorectal cancer (CRC) survivors. Dietary components of the ACS guidelines may act in part by changing the microbiome, which is implicated in CRC outcomes.

Objectives: We conducted a pilot cross-sectional study to explore associations between ACS guidelines and the gut microbiome.

Methods: Stool samples and questionnaires were collected from 28 CRC survivors at the University of California, San Francisco from 2019 to 2020. ACS scores were calculated based on validated questionnaires. Gut microbial community structure from 16S amplicons and gene/pathway abundances from metagenomics were tested for associations with the ACS score and its components using ANOVA and general linear models.

Results: The overall ACS score was not significantly associated with variations in the fecal microbiota. However, fruit and vegetable intake and alcohol intake accounted for 19% (P = 0.005) and 13% (P = 0.01) of variation in the microbiota, respectively. Fruit/vegetable consumption was associated with increased microbial diversity, increased Firmicutes, decreased Bacteroidota, and changes to multiple genes and metabolic pathways, including enriched pathways for amino acid and short-chain fatty acid biosynthesis and plant-associated sugar degradation. In contrast, alcohol consumption was positively associated with overall microbial diversity, negatively associated with Bacteroidota abundance, and associated with changes to multiple genes and metabolic pathways. The other components of the ACS score were not statistically significantly associated with the fecal microbiota in our sample.

Conclusions: These results guide future studies examining the impact of changes in the intake of fruits, vegetables, and alcoholic drinks on the gut microbiome of CRC survivors.

Keywords: alcohol; colorectal cancer; fruits and vegetables; human gut microbiome; nutrition; survivorship.

FIGURE 1

Components of American Cancer Society (ACS) scores significantly account for variations in the microbiota of colorectal cancer survivors. (A) Derivation of our study population (Table 1). (B) A schematic for calculating the ACS score. Fruit/veg points include the sum of points for amount (5+ servings/d = 1 point; <5 servings/d = 0 points) and tertiles of variety (0, 1, 2 points). When examining the ACS score with alcohol, the alcohol points (0–2) are added to the ACS score for a total range of 0–8. (C) Permutational multivariate analysis of variance (PERMANOVA) testing of ACS score and component points, clinical history, demographics, other health behavior variables, and continuous variables used to calculate ACS scores. Nominal P values were reported from the PERMANOVA test (ADONIS function in the Vegan package in R) using the weighted-UniFrac beta diversity metric of bacterial community composition, and false discovery rates q were calculated with Benjamini-Hochberg multiple-testing corrections (Supplementary Table 2). A PERMANOVA test with a different beta diversity metric (Bray-Curtis) is shown in Supplementary Fig. 2a. N = 27 participants. BMI, body mass index.

FIGURE 2

Fruit and vegetable intake is associated with gut microbial diversity and composition. (A) The Fruits and Vegetable score is calculated as 1 point for consuming 5 or more servings/d and 0–2 points for tertiles of the number of unique fruits and vegetables usually consumed in a year. (B) Principal coordinate (PC) analysis plots of weighted-UniFrac distance matrices of bacterial community composition colored based on fruits/veg points, servings, and the number of unique fruit/veg consumed. R² and P values were extracted from permutational multivariate analysis of variance (PERMANOVA) tests. (C) Alpha diversity analysis using Shannon diversity metric. (D,E) Proportions of phyla that were significantly (q < 0.2) correlated with (D) Fruit/veg points and (E) The number of unique fruits and vegetables consumed. (F, G) Centered-log-ratio (CLR) transformed abundances of genera significantly associated with (F) Fruit/veg points and (G) the number of unique fruits and vegetables consumed. Adjusted P values (q) were calculated using Benjamini-Hochberg multiple-testing correction. The line and gray ribbons represent 95% confidence intervals. N = 27 participants.

FIGURE 3

Fruit and vegetable intake is associated with microbial gene and pathway abundances. (A, B) Nonmetric multidimensional scaling (MDS) of variations in (A) gene family-level abundances and (B) pathway abundances colored by fruit/vegetable points. R² and P values were extracted from permutational multivariate analysis of variance (PERMANOVA) tests. (C,D) Significant (false-discovery-rate q < 0.2; ANOVA test with Benjamini-Hochberg correction) (C) Gene families (Supplementary Table 3) and (D) pathways associated with fruit/veg points (Supplementary Table 4). Bolded pathways are significant by general linear model (GLM). Z-score was calculated from average normalized abundances. The identifier in the parentheses refers to gene family entries in the UniRef90 database. N = 27 participants.

FIGURE 4

Alcohol consumption is associated with the gut microbiome of cancer survivors. (A) Principal coordinate (PC) analysis of weighted-UniFrac distance matrices of bacterial community composition from 16S ribosomal RNA gene sequencing colored based on alcohol intake. Light drinking—defined as >0 and ≤1 drink/d for women, >0 and ≤2 drinks/d for men; amounts above these cut-offs are labeled as heavy drinking. R² and P values were extracted from permutational multivariate analysis of variance (PERMANOVA) tests. (B) Alpha diversity analysis using Shannon diversity index of community composition for alcohol intake. (C) The proportion of phylum-level abundances associated with alcohol intake. (D) Centered-log-ratio (CLR)-transformed abundance of a species significantly associated with alcohol intake. (E) A volcano plot of gene families tested for association with alcohol intake using 2 statistical models - a general linear model (GLM) and ANOVA (Supplementary Table 5). GLM tests for the linear associations between normalized gene family-level abundances and alcohol intake. (F) Significant pathways associated with alcohol intake (Supplementary Table 6). Bolded pathways have a significant relationship based on GLM. Significance is defined as false-discovery-rate q < 0.2 using Benjamini-Hochberg multiple-testing correction. N = 27 participants.