Many of the gut microbiome’s effects on health parameters are the result of microbes metabolising diverse compounds from our diet [1
]. Dietary (poly)phenols, which include flavonoids (such as flavan-3-ols, flavonols, anthocyanins, flavanones, flavones and isoflavones) and non-flavonoids (such as stilbenes, ellagitannins, lignans and phenolic acids), are actively studied for their potential health benefits in chronic diseases, such as cardiovascular diseases (CVD) and age-related cognitive decline [2
]. Common (poly)phenol-rich foods include fruits, berries, nuts, tea, cocoa products, coffee, vegetables, olive oil, soy products and red wine [3
Studies have shown that diets based on a high intake of plant (poly)phenols and fermentable fibers alter microbial activities within the gut [4
]. Indeed, the gut microbiome converts (poly)phenols and fibers into biologically active compounds, thereby regulating energy and lipids metabolism [5
] with potential impact on metabolic diseases [6
]. (Poly)phenols, for instance, exert a protective effect on vascular endothelial cells by preventing the oxidation of low-density lipoprotein [7
]. In animal studies, supplementation with high concentrations of polyphenols inhibits the growth of detrimental species, such as Clostridia
, and increases the abundance of Lactic acid bacteria and Akkermansia
Small interventional studies have shown that (poly)phenol consumption (e.g., from wine, cocoa, tea, blueberries or aronia berries) results in increases in specific beneficial bacterial communities in the gut, particularly Bifidobacterium
], while the relative abundance of Bacteroides
increases in subjects consuming red wine or aronia berry (poly)phenols [10
]. A recent study from our group found that red wine consumption was associated with an increase in gut microbial alpha diversity, potentially mediating host body mass index (BMI) reduction in two cohorts [15
]. (Poly)phenols have also been implicated in the reduced abundance of pathogenic bacterial species in the gut microbiome. A study examining the antibacterial activity of fruit (poly)phenols found a high sensitivity to these compounds in the enteropathogens Staphylococcus aureus
and Salmonella typhimurium
], while reductions in pathogenic Clostridium
and C. histolyticum
) have been reported following consumption of fruit, seed, wine and tea polyphenols [17
(Poly)phenols are well known antioxidants [21
], and the pivotal role of oxidative stress as a key pathway in weight gain has been highlighted before [22
]. It has therefore been hypothesized that both flavonoids, such as curcumin, and stilbenes such as resveratrol, have beneficial effects on adipose tissue by alleviating intracellular oxidative stress, reducing chronic low-grade inflammation, inhibiting adipogenesis and lipogenesis, and suppressing the differentiation of preadipocytes to mature adipocytes [23
However, it is not clear whether polyphenols and fermentable fibers act in tandem (or synergistically) through the gut microbiota to improve different physiological processes linked to metabolic disease risk, or whether they have independent effects [4
Here we assess the association between (poly)phenol intake, as measured by food frequency questionnaires (FFQs), and gut microbial diversity and abundances, independently of fiber intake (measured as non-starch polysaccharides) in a large-population cohort. We then investigate whether the gut microbiome mediates part of the benefits of (poly)phenols intake on obesity. This can help prioritize interventions.
The demographic characteristics of the study population are presented in Table 1
Correlations between dietary (poly)phenol intake and gut microbiome diversity and composition.
In 1810 female twins from the TwinsUK cohort, with microbiome data available and (poly)phenol intake from FFQs measured within 1.59 years (SD = 1.08), we find that both Shannon diversity and the number of observed OTUs were significantly associated with total (poly)phenol intake (Shannon
: Beta (SE) = 0.19 (0.05), p
= 3.77 × 10−4
, Observed number of OTUs
: 0.21 (0.05), p
= 2.99 × 10−5
), stilbene (Shannon
: 0.06 (0.02), p
= 5.42 × 10−5
, Observed number of OTUs
: 0.07 (0.01), p
= 6.51 × 10−3
) and flavonoid intake (Shannon
: 0.09 (0.03), p
= 9.67 × 10−3
, Observed number of OTUs
: 0.14 (0.03), p
= 2.31 × 10−5
) after adjusting for age, BMI, fiber intake, energy intake and family relatedness (Figure 1
). We validated the association between Shannon diversity and total (poly)phenol intake, adjusting for fiber intake, age and BMI, in a small independent cohort of 64 men from the Aronia study [13
] (Beta (SE) = 0.25 (0.11), p
Analyses were adjusted by age, BMI, dietary fiber intake, total energy intake and family relatedness. CI indicates confidence interval.
We then examined, in the TwinsUK cohort, the association between stilbene and flavonoid intakes and bacterial abundances (genus with abundance > 0.001). We identified five genera significantly associated with stilbene intake and three genera associated with flavonoid intake, after adjusting for age, BMI, fiber intake, energy intake, family relatedness and multiple testing using FDR correction (FDR < 0.05) (Figure 2
). These include (i) the positive correlation of flavonoid intake and Veillonella
, a performance-enhancing microbe that functions via its metabolic conversion of exercise-induced lactate into the short-chain fatty acid propionate [34
], and (ii) the positive correlation of stilbene intake with some known butyrate-producing bacteria, such as Lachnospira
In addition, we found that Veillonella abundances were positively correlated with fecal abundance of lactate (Beta (SE) = 0.27 (0.04), p = 1.45 × 10−10) in TwinsUK.
Correlations between dietary stilbene and flavonoid intakes, obesity and gut microbiome diversity.
We next investigated, in the TwinsUK cohort, the associations between stilbene and flavonoid intake and obesity, adjusting for covariates, fiber intake and family structure. We found that both stilbene and flavonoid intake (in the log scale) were also associated with a significantly lower prevalence of obesity. (Stilbenes
: OR(95%CI) = 0.80 (0.73,0.87), p
= 4.90 × 10−7
: OR(95%CI) = 0.77 (0.65,0.91), p
= 0.002) (Table 2
Analyses are adjusted for age, fiber intake, total daily energy intake and family relatedness.
We therefore conducted a formal mediation analysis with TwinsUK to determine the indirect effect of the gut microbiome (Shannon index) on the relationship between (poly)phenol intake (stilbene- and flavonoid-high versus -low) and obesity. The analysis found that the Shannon diversity acted as a potential partial mediator in the negative association between stilbene intake and obesity (VAF = 11.16% (8.13%, 16.66%) p < 0.001), and in the negative association between flavonoids intake and obesity (11.11% (4.71%, 93.80%)).
In the largest study to date, we report that (poly)phenol intake, and particularly stilbenes and flavonoids, are associated with higher gut microbiota diversity and lower prevalence of obesity, independent of fiber intake. Furthermore, gut microbial diversity mediated the association between stilbene and flavonoid intakes and obesity. In addition, we identified a positive correlation between flavonoid intake and Veillonella, the energy-enhancing bacteria, and a positive association between stilbene intake and Faecalibacterium, a key marker of a healthy gut.
Stilbenes are non-flavonoid (poly)phenols, characterized by the presence of a 1,2-diphenylethylene nucleus in their structure, and include resveratrol and its derivatives [35
]. They are present in different foods and plants at high levels, such as red grapes, red wine, some kinds of tea, berries and peanuts [36
], though their levels are very low in foods overall. Because of this, the association between gut microbiome diversity and stilbene intake may be a marker of other factors, such as consumption of red wine and berries, which are also very rich in other (poly)phenols, such as anthocyanins and flavan-3-ols. However, when we repeated the analyses adjusting for anthocyanins and flavan-3-ols, the results remained the same, suggesting that the effect of stilbenes is independent.
Several small studies have found that polyphenols might be effective in preventing small increases in weight during periods of overfeeding, although there is no evidence for them reducing weight [37
]. Our finding of a 20–23% lower prevalence of obesity with higher intakes of total flavonoids and stilbenes, of which up to 11% is mediated by microbial diversity, highlights the public health importance of these results. By increasing the dietary intake of polyphenols, it may be possible to prevent some of the weight gain which is constantly increasing worldwide [38
The intakes associated with these findings could be readily incorporated into the diet by encouraging a Mediterranean diet, which facilitates a higher polyphenol intake than the traditional western diet [39
], or by including plants such as Hibiscus [21
] as part of the diet.
The flavonoid class embraces the largest and most widely studied group of (poly)phenols, with a structure characterized by two phenyl rings and a heterocyclic ring. The main food sources of flavonoids in the TwinsUK cohort are tea, dark chocolate, cocoa and fruit (Table S1
). Previous studies have suggested that (poly)phenols and fiber may provide complementary health benefits [40
]. Only a small percentage (<10%) of the (poly)phenols ingested is absorbed in the small intestine. The rest accumulates in the lumen, and is subjected to the catabolic reactions of the gut microbial community, which breaks them down into a series of low molecular weight phenolic metabolites that, being absorbable, may be responsible for the health benefits of (poly)phenols [41
According to current UK dietary guidelines, the recommended average fiber intake for adults is 30g per day [42
]. However, though the dietary guidelines suggest that adults should consume five portions of fruits and vegetables per day [43
], and though the newly published Cardiovascular disease risk Reduction (CRESSIDA) study shows that adherence to UK dietary guidelines is associated with higher dietary intake of total and specific (poly)phenols [44
], there is no indication as to the amount of total and specific (poly)phenols that should be ingested. We found that (poly)phenols, and in particular stilbenes and flavonoids, are associated with microbiome diversity independently of fiber intake. Our results in both the TwinsUK and Aronia cohorts indicate that even moderate consumption of (poly)phenols in the diet may have a positive impact on gut microbiome diversity. As the range of (poly)phenols in fruits and vegetables is highly variable [45
], our results could inform new guidelines for increasing awareness of the beneficial effects of (poly)phenol consumption. More importantly, our study highlights that different types of phenolic compounds have different links to gut microbiome composition, which may in turn have different effects on human health. Future studies are needed to look at more specific sub-classes of polyphenols, and the food sources of these polyphenols, as they will have structurally diverse polyphenol profiles, which is likely to lead to differing interactions with the gut microbiome.
In our exploratory study, we also report a link between flavonoid dietary intake and the abundance of Veillonella
. A recent study by Schelman and coworkers [34
] found that not only is the abundance of this bacterium linked to improved exercise performance, but that this is due to a transformation of lactate into propionate. Moreover, in our data, we find a significant positive association between Veillonella
and the fecal abundance of lactate. The association between Veillonella
and flavonoid intake then suggests one of the mechanisms whereby flavonoids may improve metabolic health, as up to certain doses, propionate is expected to improve energy metabolism. It also suggests that there may be scope for investigating the role of flavonoid supplementation or dietary interventions in improving athletic performance.
Another interesting aspect to come out of our study is the higher abundance of Faecalibacterium
in individuals with higher stilbene intakes. Faecalibacterium prausnitzii
is known to be a major butyrate producer [46
], and it has been widely claimed that among the main benefits of prebiotic fibers, such as inulin, is the increase in butyrate production [47
], a powerful anti-inflammatory and histone deacetylase regulator [48
] shown to be protective against colon cancer [49
]. This suggests that some of the health benefits commonly associated with stilbenes, such as resveratrol, may be due to their link to butyrate production, and this is an aspect that deserves further investigation.
We note some study limitations. First, we have used FFQs rather than other methods for assessing nutrient intake. FFQs have become a well-accepted method for the quantitative assessment of usual nutrient intake [50
], but being recalled, the data are subject to some bias. However, the value of FFQs for assessing dietary composition has been documented objectively, by correlations with biochemical indicators and the prediction of outcomes in prospective studies [51
]. Second, the discovery cohort was based on middle-aged white female twins. However, we replicated our main results in a small independent study of males, suggesting that the results are generalizable to both sexes. Thirdly, the cross-sectional and observational nature of this study does not allow us to determine causal relationships between (poly)phenol intake, gut microbiome composition and health, where ideally randomized control trials are needed. It does not allow us to understand some associations, which go in the opposite direction for flavonoids and stilbenes, such as the one with Sutterellla
. Moreover, we here investigated the distinct impact of polyphenols on the microbiome using only bio-informatics tools. Additional investigations, involving diets with omitted polyphenols or dietary fibers, are recommended to validate the current findings. Further, an important future avenue for research will be investigating whether any of these interactions between (poly)phenols and gut microbes are different in men vs women, or young vs old, and if the type and amount of polyphenol intake can be optimized based on demographic considerations. On the other hand, we note several strengths, including the sample size of the study and the detailed clinical and molecular phenotyping of the study subjects, which has allowed us to test the relative contributions of (poly)phenol intake to metabolic phenotypes.