Advertisement

Red Wine Consumption Associated With Increased Gut Microbiota α-Diversity in 3 Independent Cohorts

      Keywords

      Abbreviations used in this paper:

      BMI (body mass index), FGFP (Flemish Gut Flora Project), GM (gut microbiota)
      See editorial on page 48.
      Alcohol consumption leads to adverse health outcomes.
      • Wood A.M.
      • et al.
      However, moderate red wine intake has been shown to exert beneficial effects on metabolic health.
      • Artero A.
      • et al.
      This is mostly attributed to red wine’s rich and varied polyphenol content.
      • Chiva-Blanch G.
      • et al.
      Polyphenols have antimicrobial properties that can beneficially affect the gut microbiota (GM), which may have a knock-on effect on host health, as proven by animal studies and clinical trials.
      • Cardona F.
      • et al.
      Surprisingly, the consequences of red wine consumption compared to other type of alcoholic beverages on the GM remain poorly explored in epidemiologic studies. We aimed to investigate and compare the effect of various alcoholic drinks on the GM and subsequent health outcomes in large population-based cohorts.

      Methods

      We explored the effect of beer and cider, red wine, white wine, spirits, and sum of all alcohols on the α-diversity of the GM (profiled via 16S ribosomal RNA sequencing) in a discovery cohort of 916 UK female individuals (TwinsUK), using a linear mixed-effect model adjusted for age, body mass index (BMI), Healthy Eating Index, education, and family structure (Supplementary Figure 1). Alcohol consumption was derived from food frequency questionnaires, where the average number of glasses (ie, 2 units of alcohol) consumed monthly were reported. Alcohol patterns associated with α-diversity (P < .002) were evaluated on GM β-diversity using permutational multivariate analysis of variance on Bray-Curtis dissimilarity matrix, and their association with 85 genera present in at least 10% of the population. To evaluate the relevance of significant findings in relation to host’s health, we performed mediation analysis where α-diversity was considered as a mediator of the association between alcohol consumption and BMI or blood fasting glucose, insulin, total cholesterol, chylomicron, low-density lipoprotein and high-density lipoprotein levels, including all covariates described previously.
      We aimed to replicate significant results (Bonferroni P < .002 for α-diversity and P < .0006 for genera) in 2 independent cohorts: the Flemish Gut Flora Project (FGFP; n = 1104) and the American Gut Project (n = 904, alcohol consumption reported as a binary variable), as well as in discordant twin analysis (50 pairs not included in the discovery cohort). Cohort descriptions as well as statistical analysis and data processing are detailed in the Supplementary Material.

      Results

      Red wine consumption was positively associated in a frequency-dependent manner with α-diversity, but even rare consumption showed an effect (Table 1). White wine also displayed a lesser but suggestive positive association with α-diversity, while we saw no associations with other alcohol categories.
      Table 1Results of the Association Between Gut Microbiota and Alcoholic Drinks Consumption
      CohortGM TraitAlcoholnEffect sizeSEP valueOther
      “Other” indicates other α-diversity metrics tested in this study that were significantly associated with the drinking category of interest (P < .002), such as Sh, Si, ISi, or Fisher. For the FGFP cohort only, Sh and Fisher were available for replication and for the AGP cohort, Sh only.
      CohortGM trait
      Discovery
       TwinsUKRichnessAll alcohol (F)9162.91.710.09
      Beer (F)9162.32.880.42
      Spirits (F)9161.32.320.59
      White wine (F)9164.91.910.01
      Red wine (F)9167.91.725.09E-06
      Significant results in the discovery cohort (P value below Bonferroni threshold of 0.002 for α-diversity and 0.0006 for genera).
      Sh, Fisher
      Red wine (B)91621.34.821.11E-05
      Significant results in the discovery cohort (P value below Bonferroni threshold of 0.002 for α-diversity and 0.0006 for genera).
      Fisher
      Red wine (RN)5092.12.734.00E-03
      Results nominally significant and not included in the multiple testing correction.
      Fisher
      Red wine (RD)1976.53.20.047
      Results nominally significant and not included in the multiple testing correction.
      Fisher
      PhascolarctobacteriumRed wine (F)9160.03.0087.32E-05
      Significant results in the discovery cohort (P value below Bonferroni threshold of 0.002 for α-diversity and 0.0006 for genera).
      NA
      BarnesiellaRed wine (F)9160.03.0072.44E-04
      Significant results in the discovery cohort (P value below Bonferroni threshold of 0.002 for α-diversity and 0.0006 for genera).
      NA
      Prevotellaceae_NK3B31Red wine (F)9160.01.0043.67E-04
      Significant results in the discovery cohort (P value below Bonferroni threshold of 0.002 for α-diversity and 0.0006 for genera).
      NA
      Replication
       DiscordantRichnessRed wine (F)10027.918.543.91E-03
      Results nominally significant and same direction of association in the replication analysis.
      Fisher
       TwinsPhascolarctobacteriumRed wine (F)1000.013.0250.59
      Results not significant but with the same direction of association in the replication analysis.
      NA
      BarnesiellaRed wine (F)1000.047.050.35
      Results not significant but with the same direction of association in the replication analysis.
      NA
      Prevotellaceae_NK3B31Red wine (F)1000.049.0240.05
      Results not significant but with the same direction of association in the replication analysis.
      NA
       FGFPRichnessRed wine (F)11041.1.450.02
      Results nominally significant and same direction of association in the replication analysis.
      Sh
      Red wine (B)11041.9.830.02
      Results nominally significant and same direction of association in the replication analysis.
      Sh
      PhascolarctobacteriumRed wine (F)11040.11.070.14NA
      BarnesiellaRed wine (F)11040.26.071.66E-04
      Results nominally significant and same direction of association in the replication analysis.
      NA
      Prevotellaceae_NK3B31Red wine (F)1104NANANANA
       AGPRichnessRed wine (B)9040.70.310.03
      Results nominally significant and same direction of association in the replication analysis.
      Sh
      NOTE. The table reports the results (effect size, SE, and P value) of the linear regression between alcohol consumption (the predictor) and several GM traits (the response).
      ↑, positive association between the GM trait and drinking pattern; ↓, negative association between the GM trait and drinking pattern; B, binary; when alcohol consumption was used as binary variable; F, frequency; when alcohol consumption was used as frequency variable; Isi, inverted Simpson; NA, not applicable; RD, linear regression including only volunteers drinking rarely or daily; RN, linear regression including only volunteers drinking rarely or never; Sh, Shannon; Si, Simpson.
      a “Other” indicates other α-diversity metrics tested in this study that were significantly associated with the drinking category of interest (P < .002), such as Sh, Si, ISi, or Fisher. For the FGFP cohort only, Sh and Fisher were available for replication and for the AGP cohort, Sh only.
      b Significant results in the discovery cohort (P value below Bonferroni threshold of 0.002 for α-diversity and 0.0006 for genera).
      c Results nominally significant and not included in the multiple testing correction.
      d Results nominally significant and same direction of association in the replication analysis.
      e Results not significant but with the same direction of association in the replication analysis.
      Twins allow control for host genetics, early-life environmental exposure, and socioeconomic status, which are important confounders of both red wine consumption and GM composition. Twins drinking red wine at a frequency at least 2 categories above their co-twins had significantly higher α-diversity (Table 1). Moreover, associations between red wine and 2 α-diversity metrics were independently replicated in FGFP. Finally, red wine drinkers had significantly greater α-diversity than non–red wine drinkers in the TwinsUK, FGFP, and American Gut Project cohorts (Table 1).
      Next, we observed that frequency of red wine drinking only accounted for a modest proportion of the total GM β-diversity (ie, between individuals) variance (0.5%; P = .001) and significantly associated with 3 genera (P < .0006; Table 1) of which 1, Barnesiella, was replicated in the FGFP cohort. The same direction of association was also observed in discordant twin analysis (Table 1).
      Finally, α-diversity acted as a potential partial mediator in the negative association between red wine consumption and BMI or chylomicron levels (Supplementary Table 1; 18.6% and 23% of mediation, respectively; P < .001). We also observed a direct association between red wine and blood insulin and high-density lipoprotein. BMI was the only health biomarker common to all 3 cohorts. We replicated the potential mediation of α-diversity on the association between BMI and red wine in the American Gut Project (22% of mediation), but not in FGFP.

      Discussion

      Here, we demonstrated that red wine consumption is positively associated with GM α-diversity, a marker of gut health in 3 independent cohorts and discordant twin analysis. Furthermore, our results suggest that even rare consumption may be sufficient to increase α-diversity. We also showed that this may contribute to some but not all of the highly debated health benefits conferred by moderate red wine consumption, such as improvement of cholesterol metabolism or reduced adiposity.
      • Artero A.
      • et al.
      ,
      • Chiva-Blanch G.
      • et al.
      This may be related to the high content of polyphenols in red wine, such as anthocyanin, resveratrol, and gallic acid.
      • Corder R.
      • et al.
      This was suggested by the weaker association detected with white wine that is similar to red wine in terms of alcohol percentage (and potential social confounders), but is on average 6–7 times less concentrated in polyphenols.
      • Sánchez-Moreno C.
      • et al.
      Notably, Barnesiella level was higher in red wine consumers, consistent with previous findings reporting that Barnesiella doubled in the gut of rats fed black raspberry diets.
      • Gu J.
      • et al.
      Black raspberry is 4 times more concentrated in polyphenols than red wine; importantly, it also contains resveratrol, an antimicrobial found to contribute to red wine’s potential beneficial effects on cardiovascular functions.
      • Pastor R.
      • et al.
      Despite consistency in the results among the 3 cohorts, some limitations must be considered. Self-reported alcohol consumption is often underestimated and was captured differently in all cohorts. Complete homogenization of all 3 datasets was not possible, so results were not adjusted for identical covariates. Additionally, BMI was the only common health surrogate available in all cohorts, limiting the generalization of our other findings. Finally, the cross-sectional and observational nature of this study does not allow us to determine causal relationships between red wine drinking and GM composition and health, where ideally randomized controlled trials are needed (but highly unlikely).
      In conclusion, red wine consumption was associated with an increase in gut microbial α-diversity, potentially mediating host BMI reduction in 2 cohorts. This was not observed in response to any of the other alcohols studied. These results could be due to the high polyphenol content of red wine, contributing to the global debate about its potential health benefits and furthers our understanding of gut microbiota mechanisms.

      Acknowledgments

      The authors wish to express their appreciation to all study participants of the TwinsUK cohort. We thank Julia K. Goodrich, Ruth E. Ley, and the Cornell technical team for generating the microbial data. The authors thank Rob Knight and Daniel McDonald and other members of the American Gut/British Gut project for the valuable open source data. The authors acknowledge Ruth Bowyer for providing the Healthy Eating Index and socioeconomic data. Author contribution: TDS and CILR planed the study. CILR conducted the statistical analysis. PMW and generated the ASVs. JS and JR carried out the analysis performed in the Flemish Gut Flora Project cohort. CILR drafted the manuscript. JTB and TDS had leadership in collecting the data. All authors approved the final version of the manuscript.

      Supplementary Methods

      Gut Microbiota Profiling

      We used 16s ribosomal RNA microbiome data acquired, as described previously, available on 1421 volunteers from the TwinsUK cohort.
      • Goodrich J.K.
      • et al.
      Ethical approval was provided by the National Research Ethics Service Committee London–Westminster (Research Ethics Committee reference no. 12/LO/0227). All data are available for request.
      Sequences were demultiplexed, and separate forward and reverse read files generated for each sample. Amplicon sequence variants (ASVs) were then produced using the DADA2
      • Callahan B.J.
      • et al.
      package in R. Sequences were assessed for quality and trimmed at the ends to remove deteriorating quality reads. Error was estimated within the forward and reverse reads for each sample, and the ASV algorithm applied (DADAfs, DADArs) to produce ASVs. Forward and reverse ASVs were then joined. Removal of chimeras was followed by taxonomic assignment with SILVA1.3.2. Following recommendation of McMurdie et al,
      • McMurdie P.J.
      • et al.
      α-diversity metrics were calculated on the full untrimmed ASV count table using the Phyloseq package
      • McMurdie P.J.
      • et al.
      in R. ASVs were aggregated at the genus level before being transformed to zero inflated and log 10 transformed relative abundance.

      Evaluation of Alcohol Consumption From TwinsUK

      Alcohol consumption was measured by self-administered food frequency questionnaire data, collected following Epic-Norfolk guidelines. Twins reported the numbers of alcoholic drinks consumed per month for different alcoholic drinks, including red wine, white wine, beer and cider, and spirits. Volunteers overall drinking frequency was classified under the following categories: never (individuals who reported to never drink alcohol), rarely (less that once a month), occasionally (less than once a week but more than once a month), regularly (more than once a week but less than daily), and daily (at least once a day). Individual reporting to drink more than 2 glass per day of each category were removed from the analysis. Overall drinking frequency was calculated by summing up all reported alcohol consumption.

      Covariates

      In all analysis, we considered the effects of age, BMI (kg/m2), and socioeconomic status
      • Bowyer R.C.
      • et al.
      due to their previous associations with GM composition. We used education assessed using questionnaires as a surrogate for socioeconomic status, as it was available in the TwinsUK and American Gut Project (AGP) cohort. Lastly, alcohol consumption is often related/confounded by specific dietary patterns.
      • Hansel B.
      • et al.
      Thus, we aimed to correct our associations for dietary patterns using the Healthy Eating Index 2010,
      • Guenther P.M.
      • et al.
      calculated from food frequency questionnaires.
      • Bowyer R.C.
      • et al.
      Any missing data were labeled as NA and the participant was removed from the analysis. The following technical covariates were included: sequencing depth, sequence run, person who extracted the DNA, person who loaded the DNA, and the sample collection method. After inclusion of all covariates, the initial population was reduced to 916. Due to the nature of this article, the effect of individual covariates on the results are not presented.

      Statistical Analysis

      To evaluate the effect of alcohol consumption on GM α-diversity measurements (richness, Shannon, Simpson, inversed Simpson, and Fisher indexes), we performed linear mixed-effect regression (lme4 package in R) where α-diversity measurements were a response to alcohol consumption, including family structure as random effects and age, gender, BMI, diet, and level of education as covariates. We applied a Bonferroni cutoff to correct for multiple testing (P < .002).
      We then evaluated GM β-diversity variance explained by alcohol consumption patterns (associated with α-diversity) using permutational multivariate analysis of variance (vegan package in R) on Bray-Curtis dissimilarity. To identify specific taxa associated with drinking patterns, we performed linear regression as described for α-diversity on the 85 genera present in at least 10% of the population (as a response). Results displaying a P value <.0006 (Bonferroni cutoff) were considered as significant.
      α-Diversity measurements of alcohol discordant twins were compared using a paired Wilcoxon test.
      To evaluate the clinical significance of our findings, we selected 7 health biomarkers: BMI, fasting blood levels of glucose, insulin, total cholesterol, chylomicron, high-density lipoprotein, and very-low-density lipoprotein to conduct mediation analysis (mediation package in R). In case of significant association between alcohol consumption and health biomarker, we posit α-diversity as the mediating variable for the causal effect of alcohol consumption on biomarker. Results are presented as average causal mediation effect of α-diversity in the association between alcohol and biomarkers, the average direct effect of association between alcohol and biomarkers and, finally, the percentage of mediation is also indicated (Supplementary Table 1). Significant results (for BMI only, which was the only health surrogate available in other cohorts) were then replicated in the FGFP and AGP cohorts as described for TwinsUK.

      Replication

      Replication of significant results was pursued using data from the AGP
      • McDonald D.
      • et al.
      and FGFP
      • Valles-Colomer M.
      • et al.
      cohorts processed, as described previously. Alcohol consumption was measured through a web portal (http://www.microbio.me/americangut) for AGP and through an online questionnaire for FGFP. For AGP, alcohol consumption was described as categorical variables (yes/no) only for all alcohol categories. For FGFP, alcohol consumption was reported based on intake from the previous week and frequency was categorized as never, sometimes, often, and always. For the AGP cohort, the same covariates as the ones described for TwinsUK were used plus gender, center project name, extraction robot, DNA extraction kit lot, mastermix lot, processing robot, and country of origin for AGP as technical covariates. After covariates inclusion, the initial population was reduced to 905. For FGFP, age, gender, and BMI were used as covariates. Transit time, which can be captured by the Bristol Stool Scale and is highly associated with GM composition,
      • Vandeputte D.
      • et al.
      was also included, although this was not recorded for the 2 other cohorts. Results from the discovery analysis were considered as replicated when nominal significance was observed in both replication cohorts.
      Figure thumbnail fx1
      Supplementary Figure 1Organogram of the overall analysis. Dark boxes linked by dark arrows represent the succession of the analysis conducted in TwinsUK (discovery cohort) based on significance of the results (indicated in a circle P < .002) in the first set of analyses (association between drinking frequency and α-diversity). The dark gray boxes are linked to light boxes when replication was conducted in discordant twins, FGFP or AGP based on results significance (indicated in circles). For instance, associations from the TwinsUK binary analysis presenting a P value <.002 were replicated in FGFP and AGP. All boxes for the discovery analysis contain the type of analysis conducted, the covariates included and the R package used in italic. For the replication analysis, only the covariates used are indicated.
      Supplementary Table 1Summary Statistics of the Mediation Analysis for the TwinsUK, American Gut Project, and Flemish Gut Flora Project Cohort for Available α-Diversity Metrics
      CohortTraitEffectnACMEPADEP%
      Discovery
       TwinsUKBMIPartial916-0.06<2e-16-0.230.0418%
      GlucoseNull4650.010.600.070.48NA
      InsulineDirect412-0.050.96-3.00<2e-16NA
      Total cholesterolDirect4240.001.000.060.56NA
      ChylomicronPartial4240.23<2e-16-0.010.0223%
      LDLNull4240.000.960.070.42NA
      HDLDirect4240.000.760.100.04NA
      Replication
       FGFPBMIDirect1104-0.040.08-0.640.03NA
       AGPBMIFull904-0.13<2e-16-0.240.5422%
      NOTE. Models were built to assess the percentage of the association between red wine consumption and health biomarkers mediated by red wine consumption. All covariates described in the Methods section were included except for BMI in the BMI–red wine association. The effect observed are described in the “effect” column where: Null, corresponds to no association observed between red wine consumption ant the trait of interest; Partial, the association between the biomarker and red wine consumption is partially mediated by richness; Direct, the association between the biomarker and red wine consumption is not mediated by richness; Full, the association between the biomarker and red wine consumption is fully mediated by richness.

      Supplementary Material

      References

        • Wood A.M.
        • et al.
        Lancet. 2018; 391: 1513-1523
        • Artero A.
        • et al.
        Maturitas. 2015; 80: 3-13
        • Chiva-Blanch G.
        • et al.
        Alcohol Alcohol. 2013; 48: 270-277
        • Cardona F.
        • et al.
        J Nutr Biochem. 2013; 24: 1415-1422
        • Corder R.
        • et al.
        Nature. 2006; 444: 566
        • Sánchez-Moreno C.
        • et al.
        J Sci Food Agric. 1999; 79: 1301-1304
        • Gu J.
        • et al.
        Mol Nutr Food Res. 2019; 63e1800636
        • Pastor R.
        • et al.
        Crit Rev Food Sci. 2019; 59: 1237-1255

      Linked Article

      • Untangling the 2-Way Relationship Between Red Wine Polyphenols and Gut Microbiota
        GastroenterologyVol. 158Issue 1
        • Preview
          The production of red wine, a product of the Vitis vinifera L grape and one of the most widely consumed beverages around the world, dates back ≥8000 years, principally originating from the Caucasus region between Europe and Asia. Recent evidence from observational studies demonstrates a role for red wine in the promotion of beneficial gut microbiota.1 In this issue of Gastroenterology, Le Roy et al1 report an association between red wine polyphenols and gut microbiota α-diversity, with only weak associations with white wine and no association with beer, cider, spirits, or all alcohol.
        • Full-Text
        • PDF