Depleted gut microbiome α-diversity is associated with several human diseases, but the extent to which this is reflected in the host molecular phenotype is poorly understood. We attempted to predict gut microbiome α-diversity from ~1,000 blood analytes (laboratory tests, proteomics and metabolomics) in a cohort enrolled in a consumer wellness program (N = 399). Although 77 standard clinical laboratory tests and 263 plasma proteins could not accurately predict gut α-diversity, we found that 45% of the variance in α-diversity was explained by a subset of 40 plasma metabolites (13 of the 40 of microbial origin). The prediction capacity of these 40 metabolites was confirmed in a separate validation cohort (N = 540) and across disease states, showing that our findings are robust. Several of the metabolite biomarkers that are reported here are linked with cardiovascular disease, diabetes and kidney function. Associations between host metabolites and gut microbiome α-diversity were modified in those with extreme obesity (body mass index ≥ 35), suggesting metabolic perturbation. The ability of the blood metabolome to predict gut microbiome α-diversity could pave the way to the development of clinical tests for monitoring gut microbial health.
Access optionsAccess options
Subscribe to Journal
Get full journal access for 1 year
only $20.83 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
The model summary statistics for all metabolites, proteins and clinical laboratory results analyzed are available to download in Supplementary Tables 1–3. Qualified researchers can access the full deidentified dataset for research purposes. Requests should be sent to firstname.lastname@example.org.
The packages and code used in this study are available at https://github.com/PriceLab/ShannonMets.
Duvallet, C., Gibbons, S. M., Gurry, T., Irizarry, R. A. & Alm, E. J. Meta-analysis of gut microbiome studies identifies disease-specific and shared responses. Nat. Commun. 8, 1784 (2017).
Shreiner, A. B., Kao, J. Y. & Young, V. B. The gut microbiome in health and in disease. Curr. Opin. Gastroenterol. 31, 69–75 (2015).
Bäckhed, F. et al. The gut microbiota as an environmental factor that regulates fat storage. Proc. Natl Acad. Sci. USA 101, 15718–15723 (2004).
Markle, J. G. et al. Sex differences in the gut microbiome drive hormone-dependent regulation of autoimmunity. Science 339, 1084–1088 (2013).
Belkaid, Y. & Hand, T. W. Role of the microbiota in immunity and inflammation. Cell 157, 121–141 (2014).
Zierer, J. et al. The fecal metabolome as a functional readout of the gut microbiome. Nat. Genet. 50, 790–795 (2018).
Wikoff, W. R. et al. Metabolomics analysis reveals large effects of gut microflora on mammalian blood metabolites. Proc. Natl Acad. Sci. USA 106, 3698–3703 (2009).
Pallister, T. et al. Hippurate as a metabolomic marker of gut microbiome diversity: modulation by diet and relationship to metabolic syndrome. Sci. Rep. 7, 13670 (2017).
Parker, B. L. et al. An integrative systems genetic analysis of mammalian lipid metabolism. Nature 567, 187–193 (2019).
Pascal, V. et al. A microbial signature for Crohn’s disease. Gut 66, 813–822 (2017).
Chang, J. Y. et al. Decreased diversity of the fecal microbiome in recurrent Clostridium difficile-associated diarrhea. J. Infect. Dis. 197, 435–438 (2008).
Schnorr, S. L. et al. Gut microbiome of the Hadza hunter–gatherers. Nat. Commun. 5, 3654 (2014).
Clemente, J. C. et al. The microbiome of uncontacted Amerindians. Sci. Adv. 1, e1500183 (2015).
Falony, G. et al. Population-level analysis of gut microbiome variation. Science 352, 560–564 (2016).
Sze, M. A. & Schloss, P. D. Looking for a signal in the noise: revisiting obesity and the microbiome. MBio 7, e01018-16 (2016).
Guh, D. P. et al. The incidence of co-morbidities related to obesity and overweight: a systematic review and meta-analysis. BMC Public Health 9, 88 (2009).
Caleyachetty, R. et al. Metabolically healthy obese and incident cardiovascular disease vents among 3.5 million men and women. J. Am. Coll. Cardiol. 70, 1429–1437 (2017).
Kasselman, L. J., Vernice, N. A., DeLeon, J. & Reiss, A. B. The gut microbiome and elevated cardiovascular risk in obesity and autoimmunity. Atherosclerosis 271, 203–213 (2018).
Price, N. D. et al. A wellness study of 108 individuals using personal, dense, dynamic data clouds. Nat. Biotechno.l 35, 747–756 (2017).
Tibshirani, R. Regresion shrinkage and selection via the lasso. J. R. Stat. Soc. Series B Stat. Methodol. 73, 9 (1996).
Kowalska, K., Socha, E. & Milnerowicz, H. Review: the role of paraoxonase in cardiovascular diseases. Ann. Clin. Lab. Sci. 45, 226–233 (2015).
Myers, M. G., Leibel, R. L., Seeley, R. J. & Schwartz, M. W. Obesity and leptin resistance: distinguishing cause from effect. Trends Endocrinol. Metab. 21, 643–651 (2010).
Zaura, E. et al. Same exposure but two radically different responses to antibiotics: resilience of the salivary microbiome versus long-term microbial shifts in feces. MBio 6, e01693-15 (2015).
Dethlefsen, L. & Relman, D. A. Incomplete recovery and individualized responses of the human distal gut microbiota to repeated antibiotic perturbation. Proc. Natl Acad. Sci. USA 108, 4554–4561 (2011).
Antunes, L. C. et al. Effect of antibiotic treatment on the intestinal metabolome. Antimicrob. Agents Chemother. 55, 1494–1503 (2011).
Zarrinpar, A. et al. Antibiotic-induced microbiome depletion alters metabolic homeostasis by affecting gut signaling and colonic metabolism. Nat. Commun. 9, 2872 (2018).
Han, T. S. & Lean, M. E. A clinical perspective of obesity, metabolic syndrome and cardiovascular disease. JRSM Cardiovasc. Dis. 5, 2048004016633371 (2016).
Cirulli, E. T. et al. Profound perturbation of the metabolome in obesity is associated with health risk. Cell Metab. 29, 488–500 (2018).
Kivimäki, M. et al. Overweight, obesity, and risk of cardiometabolic multimorbidity: pooled analysis of individual-level data for 120,813 adults from 16 cohort studies from the USA and Europe. Lancet Public Health 2, e277–e285 (2017).
Ganz, M. L. et al. The association of body mass index with the risk of type 2 diabetes: a case-control study nested in an electronic health records system in the United States. Diabetol. Metab. Syndr. 6, 50 (2014).
Le Chatelier, E. et al. Richness of human gut microbiome correlates with metabolic markers. Nature 500, 541–546 (2013).
Larsen, O. F. A. & Claassen, E. The mechanistic link between health and gut microbiota diversity. Sci. Rep. 8, 2183 (2018).
Segata, N. Gut microbiome: westernization and the disappearance of intestinal diversity. Curr. Biol. 25, R611–R613 (2015).
van Duynhoven, J. et al. Metabolic fate of polyphenols in the human superorganism. Proc. Natl Acad. Sci. USA 108, 4531–4538 (2011).
Rechner, A. R. et al. Colonic metabolism of dietary polyphenols: influence of structure on microbial fermentation products. Free Radic. Biol. Med. 36, 212–225 (2004).
Parkar, S. G., Trower, T. M. & Stevenson, D. E. Fecal microbial metabolism of polyphenols and its effects on human gut microbiota. Anaerobe 23, 12–19 (2013).
Zamora-Ros, R. et al. Differences in dietary intakes, food sources and determinants of total flavonoids between Mediterranean and non-Mediterranean countries participating in the European Prospective Investigation into Cancer and Nutrition (EPIC) study. Br. J. Nutr. 109, 1498–1507 (2013).
Koh, A. et al. Microbially produced imidazole propionate impairs insulin signaling through mTORC1. Cell 175, 947–961 (2018).
Gryp, T., Vanholder, R., Vaneechoutte, M. & Glorieux, G. p-Cresyl sulfate. Toxins 9, E52 (2017).
Manor, O. et al. A multi-omic association study of trimethylamine N-oxide. Cell Rep. 24, 935–946 (2018).
Wang, Z. et al. Gut flora metabolism of phosphatidylcholine promotes cardiovascular disease. Nature 472, 57–63 (2011).
Yao, C. K., Muir, J. G. & Gibson, P. R. Review article: insights into colonic protein fermentation, its modulation and potential health implications. Aliment. Pharmacol. Ther. 43, 181–196 (2016).
Fui, M. N., Dupuis, P. & Grossmann, M. Lowered testosterone in male obesity: mechanisms, morbidity and management. Asian J. Androl. 16, 223–231 (2014).
Vanholder, R., Schepers, E., Pletinck, A., Nagler, E. V. & Glorieux, G. The uremic toxicity of indoxyl sulfate and p-cresyl sulfate: a systematic review. J. Am. Soc. Nephrol. 25, 1897–1907 (2014).
Bogiatzi, C. et al. Metabolic products of the intestinal microbiome and extremes of atherosclerosis. Atherosclerosis 273, 91–97 (2018).
Salgado, R., López-Doval, S., Pereiro, N. & Lafuente, A. Perfluorooctane sulfonate (PFOS) exposure could modify the dopaminergic system in several limbic brain regions. Toxicol. Lett. 240, 226–235 (2016).
Onishchenko, N. et al. Prenatal exposure to PFOS or PFOA alters motor function in mice in a sex-related manner. Neurotox. Res. 19, 452–461 (2011).
Lai, K. P. et al. Dietary exposure to the environmental chemical, PFOS on the diversity of gut microbiota, associated with the development of metabolic syndrome. Front. Microbiol. 9, 2552 (2018).
Wahlström, A., Sayin, S. I., Marschall, H. U. & Bäckhed, F. Intestinal crosstalk between bile acids and microbiota and its impact on host metabolism. Cell Metab. 24, 41–50 (2016).
Hibberd, M. C. et al. The effects of micronutrient deficiencies on bacterial species from the human gut microbiota. Sci. Transl. Med. 9, eaal4069 (2017).
Watson, H. et al. A randomised trial of the effect of omega-3 polyunsaturated fatty acid supplements on the human intestinal microbiota. Gut 67, 1974–1983 (2018).
Li, M. et al. Symbiotic gut microbes modulate human metabolic phenotypes. Proc. Natl Acad. Sci. USA 105, 2117–2122 (2008).
Poesen, R. et al. Microbiota-derived phenylacetylglutamine associates with overall mortality and cardiovascular disease in patients with CKD. J. Am. Soc. Nephrol. 27, 3479–3487 (2016).
Moss, C. W., Lambert, M. A. & Goldsmith, D. J. Production of hydrocinnamic acid by clostridia. Appl. Microbiol. 19, 375–378 (1970).
Nealon, N. J., Worcester, C. R. & Ryan, E. P. Lactobacillus paracasei metabolism of rice bran reveals metabolome associated with Salmonella typhimurium growth reduction. J. Appl. Microbiol. 122, 1639–1656 (2017).
Beuers, U., Fischer, S., Spengler, U. & Paumgartner, G. Formation of iso-ursodeoxycholic acid during administration of ursodeoxycholic acid in man. J. Hepatol. 13, 97–103 (1991).
Caporaso, J. G. et al. QIIME allows analysis of high-throughput community sequencing data. Nat. Methods 7, 335–336 (2010).
DeSantis, T. Z. et al. Greengenes, a chimera-checked 16S rRNA gene database and workbench compatible with ARB. Appl. Environ. Microbiol. 72, 5069–5072 (2006).
Weiss, S. et al. Normalization and microbial differential abundance strategies depend upon data characteristics. Microbiome 5, 27 (2017).
Hughes, J. B., Hellmann, J. J., Ricketts, T. H. & Bohannan, B. J. Counting the uncountable: statistical approaches to estimating microbial diversity. Appl. Environ. Microbiol. 67, 4399–4406 (2001).
Faith, D. P. & Baker, A. M. Phylogenetic diversity (PD) and biodiversity conservation: some bioinformatics challenges. Evol. Bioinform. Online 2, 121–128 (2007).
Anderson, M. J. A new method for non-parametric multivariate analysis of variance. Austral Ecol. 26, 32–46 (2001).
Anderson, M. J. Distance-based tests for homogeneity of multivariate dispersions. Biometrics 62, 245–253 (2006).
Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. Series B Stat. Methodol. 57, 289–300 (1995).
We thank C. Diener, A. Zimmer and M. Robinson for helpful discussions. This research was supported by the M.J. Murdock Charitable Trust (L.H. and N.D.P.), Arivale and a generous gift from C. Ellison. S.M.G. was supported by a Washington Research Foundation Distinguished Investigator Award and by start-up funds from the Institute for Systems Biology.
At the time this research was conducted, L.H. and N.D.P. were co-founders of Arivale (where these data come from) and held stock in the company; N.D.P. was on the Arivale board of directors; L.H. was chair of, and G.S.O. a member of, Arivale’s scientific advisory board; A.T.M., O.M. and J.L. were employees of Arivale and had stock options in the company, as did G.S.O. and J.C.E. Arivale’s program is now closed, so there are no longer any competing interests. The authors declare no other current competing interests.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Integrated supplementary information
Supplementary Figure 1 Investigating collinearity among the identified predictors of Shannon diversity and β-diversity in the discovery cohort.
A) Heatmap showing the strength of correlation between each metabolite-metabolite pair. B) Histogram of all calculated Pearson r values for the 1560 metabolite-metabolite comparisons in the discovery cohort (N=399). Only six comparisons yielded a Pearson r value |r|>0.80. C) Metabolites showing significant association with inter-individual gut microbial variations (Benjamini-Hochberg FDR<0.05). Highlighted in red are the 4 metabolites (out of 11) identified to be the top predictors of Shannon diversity in our analysis. PERMANOVA was performed on the OTU-level Bray-Curtis dissimilarity matrix, including Shannon diversity as a covariate. D) β-dispersion across obesity subtypes. Dissimilarity between individuals’ microbiomes is significantly different across obesity (ANOVA P-Value=0.005). Post-hoc comparisons between BMI groups demonstrated that Obese II/III, but not Obese I individuals, were significantly more dispersed relative to normal weight individuals (Tukey HSD test P<0.05, two-sided). Sample sizes for individual BMI groups are as follows: normal weight n=143, overweight n=134, obese I n=66, obese II/III n=57. Boxplots represent the interquartile range (25th to 75th percentile, IQR), with the middle line demarking the median. Whiskers span 1.5*IQR. Points beyond this range are individually shown. E) Blood metabolome capacity to predict Shannon diversity across increasing time lag between stool and blood samples. Participants in the discovery cohort were stratified into tertiles based on the number of days between when the stool and blood samples were collected. LASSO using the 40 metabolites identified in MEP_L_fig1Fig. 1 was used to evaluate out-of-sample performance in predicting Shannon diversity across 5-fold CV. Tertiles 1 and 2 demonstrated similar performance to results reported on the whole cohort, while tertile 3 (with the largest time lag) demonstrated a decrease in performance. Importantly, limiting the allowed time gap between when samples were collected (Tertiles 1 & 2) did not result in improved performance relative to the model fitted on the entire cohort. Comparison between the whole cohort and Tertile 3 performance was performed using a two-sided nonparametric Mann-Whitney U-Test. Values are shown as mean +/- the Standard Error of the Mean.
Supplementary Figure 2 Variable relationships of sex steroids and bile acids with Shannon diversity.
A) Shannon diversity is not significantly different across sex (male n=111, female n=288), assessed using an OLS model adjusted for age and BMI. B) 5α-androstan-3β-17α blood concentration is higher in men than women, modelled using OLS regression and adjusting for age and BMI (P-Value=3.79e-13). Boxplots represent the interquartile range (25th to 75th percentile, IQR), with the middle line demarking the median. Whiskers span 1.5*IQR. Points beyond this range are individually shown. C) 5α-androstan-3β-17α is positively associated with Shannon diversity in both males and females. D&E) Secondary bile acids retained by LASSO in the prediction model show opposite association with Shannon diversity. Lines shown are y~x regression lines for the whole cohort (black) or for males (blue) and females (red) separately. The shaded region corresponds to the 95% confidence intervals for the slope of the line. Abbreviations: 5α-androstan-3β-17α: 5α-androstan-3β-17α-diol disulfate.
Spearman correlations of each of the 11 metabolites retained by all 10 LASSO models (rows) with microbiome genera (columns), in the discovery cohort (N=399), correcting for multiple hypothesis testing (Benjamini-Hochberg FDR<0.05). Only genera with at least one significant correlation value are displayed. Top color row labels the phylum for each genus. Left color column labels the sign of the mean β-coefficient for that metabolite across the 10 LASSO models generated to predict Shannon diversity (blue - negative, red - positive). The top bar graph represents the fractional abundance of each genus, with bars colored by phylum.
A) Comparison of Precision-Recall curves and B) Receiver Operator Characteristic (ROC) curves for clinical laboratory tests and 11 blood metabolites classifying participants in the bottom quartile of Shannon diversity using 10-fold CV implementation of Random forests. C) Out-of-sample R2 scores from penalized regression models predicting Shannon diversity using each omics platform individually, or in different combinations. Values are presented as mean R2 across 10-fold CV +/- standard error of the mean. Performance of metabolomics alone was compared to performance of metabolomics in combination with clinical labs, proteomics, and clinical labs and proteomics using a two-sided nonparametric Mann-Whitney U test. The P-Values for these comparisons are shown. No multiple hypothesis correction was implemented in this analysis. There was no significant improvement or decrease in performance when all analytes were combined relative to metabolomics data alone.
Supplementary Figure 5 Evaluating the blood metabolome-gut microbiome relationship under different disease conditions.
A&B) Relationship of observed and metabolome predicted Chao1 (A) and PD whole tree (B) diversity with self-reported measures of gastrointestinal (GI) health, as well as antibiotics use, using an OLS regression model adjusted for sex, age, and BMI. Conditions colored in red are significantly associated with both the observed and metabolome predicted metric (P-Value<0.05). C) The relationship of various GI measures with Shannon and metabolome predicted (mShannon) diversity, modelled using OLS regression adjusted for the same covariates as in (A) and (B) as well as antibiotics use (N=311). D) Comparison of mean out-of-sample performance of LASSO in predicting Shannon diversity in the whole cohort (black) and in participants who have not taken antibiotics in the last three months (gray - all 659 metabolites, red - only the 40 metabolites identified in the original analysis). Values are presented as mean +/- the Standard Error of the Mean. E) A scatter plot of metabolome predicted (mShannon) diversity and observed Shannon diversity in participants who have not taken antibiotics in the last three months, generated using only the 40 metabolites originally identified in this study. The mean R2 across the 10 cross validations, Pearson r of observed versus mShannon values, and corresponding P-Value are shown.
Supplementary Figure 6 Relationship of the plasma metabolome and Shannon diversity changes across BMI classes.
A) β-coefficients for each of the 11 metabolites retained by all 10 LASSO models from an OLS regression model with Shannon diversity as the dependent variable and sex and age included as covariates in the discovery cohort. The cohort was stratified based on defined BMI cutoffs and models were fitted independently for each BMI class. B) Scatter plot of PFOS and Shannon diversity for participants whose BMI is less than 25 (normal weight), and equal to or greater than 35 (Obese II/III) in the discovery cohort. C) Comparison of strengths of correlations for 5α-androstan-3β-17α and PFOS with Shannon diversity across obesity in the discovery and validation cohorts. D) Scatter plot of 5α-androstan-3β-17α and Shannon diversity for participants whose BMI is less than 35, and greater than or equal to 35 (Obese II/III) in the validation cohort. Pearson r and P-Values are shown. Lines shown are y~x regression lines, while the shaded region corresponds to the 95% confidence intervals for the slope of the line. Abbreviations: 5α-androstan-3β-17α: 5α-androstan-3β-17α-diol disulfate; PFOS: perfluorooctanosulfic acid.
A) Spearman correlation of each of the 11 strongest metabolites identified in the discovery set (rows) with microbiome genera (columns) in the validation cohort, correcting for multiple hypothesis testing (Benjamini-Hochberg FDR<0.05). Only genus-metabolite correlations that were significant in the discovery cohort were considered. The top bar graph represents the fractional abundance of each genus in the validation cohort, with bars colored by phylum. B) The number of significant Spearman correlations of each of the 11 metabolites retained by all 10 LASSO models (rows) with microbiome genera in the discovery and validation cohorts, correcting for multiple hypothesis testing (Benjamini-Hochberg FDR<0.05).
Supplementary Figs. 1–7 and Supplementary Table 4
Metabolites with non-zero β-coefficients in the LASSO models predicting Shannon diversity.
Association of clinical laboratory results with Shannon diversity in the discovery cohort.
Association of blood proteomics with Shannon diversity in the discovery cohort.