Society: AGA

Background:
We and others have demonstrated a tight coupling between dietary intake, gut microbial ecology, and near-term alterations in cardiometabolic biomarkers in the ZOE PREDICT-1 study (Zeevi 2015; Asnicar & Berry 2021). However, it is unknown whether adherence to a de novo data-driven microbially-informed diet derived from the PREDICT-1 is associated with long-term weight change and its complications.
Methods:
To evaluate the impact of a “Metabolic Microbiome Score” (MMS) on long-term weight change, we pooled data from 3 ongoing U.S.-based prospective cohorts with lifestyle and dietary data collected through biennial questionnaires and validated food frequency questionnaires. MMS was calculated by summing reported food intake multiplied by food quality scores derived from the PREDICT-1, while normalizing for the servings of foods consumed. To validate prior findings, we assessed the association between MMS, food groups, and gut microbial features using partial Spearman correlations adjusted for age, sex, and body mass index. We used a multivariable generalized linear regression model to examine the association between 4-year changes in MMS and 4-year weight change after adjusting for relevant clinical and lifestyle confounders. We fitted Cox proportional hazards models to examine the associations between MMS and incident diabetes, cardiovascular disease (CVD), and mortality.
Results:
Among 186,661 participants with more than 20 years of follow-up, we observed that compared to the lowest quartile, those in the highest quartile of MMS were associated with 1.36 kg less weight gain (95% CI, -1.40 to -1.32; Table). A 1-SD increase in MMS was associated with 0.54 kg less weight gain over 4-year study intervals (95% CI, -0.55 to -0.52). There was a modest correlation between MMS and other established dietary patterns, such as the Western (Spearman rho = -0.43) and prudent diets (rho = 0.54). We also confirmed previously observed links between MMS, higher quality foods, and the relative enrichment of health-associated gut microbial features, such as Eubacterium eligens (Fig. A). The link between MMS and weight change appeared to differ according to age, sex, and body mass index, suggesting that those younger, female, or with obesity would derive greater benefit from adhering to diet high in MMS (P for interactions <0.05). In addition, adherence to MMS diet was associated with statistically-significant reductions in obesity-related complications including type 2 diabetes, CVD, and mortality (Fig. B).
Conclusion:
A data-driven dietary score corresponding with better near-term cardiometabolic markers was associated with less long-term weight gain and reduced risk of obesity-related complications in 3 independent U.S. cohorts, suggesting a role for population-wide public health initiatives geared towards personalized improvement of diet quality.

<b>Table. Weight change over 4-year periods according to per 1-SD increase and quartiles of change in Metabolic Microbiome Score </b><br /> Abbreviations: SD, standard deviation; MV, multivariable<br /> <sup>a</sup>Additionally adjusted for sex, questionnaire cycle (4-year intervals), White (yes/no), height (continuous), body mass index (continuous), smoking pack-years (continuous), physical activity (continuous), change in physical activity (continuous), and post-menopausal hormone use (yes/no [for women])<br /> <sup>b</sup>Additionally adjusted for total energy intake (continuous)

Table. Weight change over 4-year periods according to per 1-SD increase and quartiles of change in Metabolic Microbiome Score
Abbreviations: SD, standard deviation; MV, multivariable
^aAdditionally adjusted for sex, questionnaire cycle (4-year intervals), White (yes/no), height (continuous), body mass index (continuous), smoking pack-years (continuous), physical activity (continuous), change in physical activity (continuous), and post-menopausal hormone use (yes/no [for women])
^bAdditionally adjusted for total energy intake (continuous)

<b>Figure A. Metabolic Microbiome Score (MMS) validation. </b>Partial Spearman correlation heatmap for food groups and microbiome adjusted for sex, age, and body mass index. Red color represents positive correlation while blue color represents negative correlation. Asterisks represent significance. ***: p<sub>FDR </sub>≤ 0.001; **: 0.001 < p<sub>FDR </sub>≤ 0.01; *: 0.01 < p<sub>FDR </sub>≤ 0.05.<br /> <b>Figure B. Metabolic Microbiome Score and risk of obesity-related complications.</b> Adjusted for age, sex, White (yes/no), body mass index (continuous), smoking pack-years (continuous), physical activity (continuous), alcohol intake (continuous), post-menopausal hormone use (yes/no [for women]), and disease-specific confounders.

Figure A. Metabolic Microbiome Score (MMS) validation. Partial Spearman correlation heatmap for food groups and microbiome adjusted for sex, age, and body mass index. Red color represents positive correlation while blue color represents negative correlation. Asterisks represent significance. ***: p_FDR≤ 0.001; **: 0.001 < p_FDR≤ 0.01; *: 0.01 < p_FDR≤ 0.05.
Figure B. Metabolic Microbiome Score and risk of obesity-related complications. Adjusted for age, sex, White (yes/no), body mass index (continuous), smoking pack-years (continuous), physical activity (continuous), alcohol intake (continuous), post-menopausal hormone use (yes/no [for women]), and disease-specific confounders.

Background: Coffee intake has been associated with lower risks of many chronic diseases and with features of the gut microbiome. However, no study has systematically examined the interplay between coffee intake and the gut microbiome in relation to host metabolome.
Methods: We collected detailed dietary information every 4 years since 1986 from 51,529 men in the Health Professionals Follow-up Study (HPFS). In 2012-2013, a subcohort of 307 healthy HPFS men provided up to 2 pairs of stool and 2 blood samples. Intakes of overall coffee and caffeinated, decaffeinated, filtered, instant, and espresso coffee were used as exposures. We profiled 236 plasma metabolomes as compared with 925 stool metagenomes, for which microbial species, functional pathways (MetaCyc), enzymes (EC), and gene family (UniRef90) profiles were generated. We validated our findings in a subcohort of 220 healthy women from the Nurses’ Health Study II.
Results: Higher coffee intake was strongly associated with increased abundances of several microbial species, particularly Clostridium phoceensis (FDR-adjusted p=5*10^-11) as previously observed (prev. Lawsonibacter asaccharolyticus). These associations were similar for caffeinated and decaffeinated coffee. A plasma metabolomic signature of coffee intake developed using elastic net regression was also strongly positively correlated with the abundance of C. phoceensis (p<0.001) (Fig.1). Although a variety of small molecular metabolites are jointly correlated both with coffee intake and with C. phoceensis abundance – primarily those known to be derived from coffee itself – the strongest individual correlation was between C. phoceensis and quinic acid (FDR-adjusted p=1*10^-6), a metabolite of chlorogenic acid, one of the most abundant polyphenols in coffee. Stratified analyses of pathways, enzymes, and gene families contributed by C. phoceensis also revealed a particularly strong positive association with plasma levels of quinic acid (FDR-adjusted p<0.005). In contrast, none of the metabolic pathways or enzymes of caffeine metabolism were significant in the paired multi-omic analysis (concordant with the microbial correlation with decaffeinated coffee). Also, coffee intake was associated with a reduced plasma level of high-sensitivity C-reactive protein (p=0.008). Notably, this inverse association was only observed in individuals with a higher C. phoceensis level (p = 0.004), but not in those with a lower C. phoceensis level (p=0.79) (p-interaction=0.009) (Fig. 2). We validated the strong association of coffee intake with abundance of C. phoceensis in the women’s cohort.
Conclusions: We reproduced a strong association of coffee intake with a specific gut microbial species, C. phoceensis, and showed that C. phoceensis might be involved in the metabolism of coffee polyphenols and modify the anti-inflammatory effect of coffee intake in the host.

<b>Figure 1. Associations of coffee intake with gut microbial species and plasma metabolites abundances.</b> <b>a,</b> Significant associations of coffee intake measures with microbial species (FDR-adjusted <i>p</i> < 0.25) by MaAsLin analysis. <b>b</b>-<b>d</b>, Associations of total coffee intake with the species of <i>Clostridium phoceensis</i>,<i> SGB14966</i>, and<i> SGB4777</i>. <b>e</b>-<b>f</b>, Taxonomy and spearman correlation between <i>C. phoceensis</i>,<i> SGB14966</i>, and<i> SGB4777</i>. <b>g</b>, Significant spearman partial correlations of total coffee intake with all known plasma metabolites (<i>n</i>=373). <b>h</b>, Metabolomic signature of total coffee by elastic net regression. <b>i</b>, Association of metabolomic signature of coffee intake with <i>C. phoceensis</i>.

Figure 1. Associations of coffee intake with gut microbial species and plasma metabolites abundances. a, Significant associations of coffee intake measures with microbial species (FDR-adjusted p < 0.25) by MaAsLin analysis. b-d, Associations of total coffee intake with the species of Clostridium phoceensis, SGB14966, and SGB4777. e-f, Taxonomy and spearman correlation between C. phoceensis, SGB14966, and SGB4777. g, Significant spearman partial correlations of total coffee intake with all known plasma metabolites (n=373). h, Metabolomic signature of total coffee by elastic net regression. i, Association of metabolomic signature of coffee intake with C. phoceensis.

<b>Figure 2. Interplay between coffee intake and the microbial species <i>Clostridium phoceensis</i> in relation to plasma metabolome. </b><i>Clostridium phoceensis</i> may be involved in the metabolism of polyphenols in coffee. <b>a, </b>HAllA (high-sensitivity pattern discovery in large, paired multi-omic datasets) heatmap showing significant clusters between all microbial species (<i>n</i> = 333) and all known metabolites (<i>n</i> = 375). <b>b</b>, The metabolism of the significant <i>Clostridium phoceensis</i>-correlated clusters of coffee component metabolites. <b>c</b>, HAllA heatmap between <i>Clostridium phoceensis</i>-stratified EC enzymes (<i>n</i> = 277) and known metabolites (<i>n</i> = 375). <b>d</b>, the modification effect of <i>Clostridium phoceensis</i> on the association between coffee intake and a reduced plasma level of high-sensitivity C-reactive protein (CRP).

Figure 2. Interplay between coffee intake and the microbial species Clostridium phoceensis in relation to plasma metabolome. Clostridium phoceensis may be involved in the metabolism of polyphenols in coffee. a, HAllA (high-sensitivity pattern discovery in large, paired multi-omic datasets) heatmap showing significant clusters between all microbial species (n = 333) and all known metabolites (n = 375). b, The metabolism of the significant Clostridium phoceensis-correlated clusters of coffee component metabolites. c, HAllA heatmap between Clostridium phoceensis-stratified EC enzymes (n = 277) and known metabolites (n = 375). d, the modification effect of Clostridium phoceensis on the association between coffee intake and a reduced plasma level of high-sensitivity C-reactive protein (CRP).

Kai Wang

Appearances