The secondary bile acid isoursodeoxycholate correlates with post-prandial lipemia, inflammation, and appetite and changes post-bariatric surgery

Abstract

Primary and secondary bile acids (BAs) influence metabolism and inflammation, and the gut microbiome modulates levels of BAs. We systematically explore the host genetic, gut microbial, and habitual dietary contribution to a panel of 19 serum and 15 stool BAs in two population-based cohorts (TwinsUK, n = 2,382; ZOE PREDICT-1, n = 327) and assess changes post-bariatric surgery and after nutritional interventions. We report that BAs have a moderately heritable genetic component, and the gut microbiome accurately predicts their levels in serum and stool. The secondary BA isoursodeoxycholate (isoUDCA) can be explained mostly by gut microbes (area under the receiver operating characteristic curve [AUC] = ∼80%) and associates with post-prandial lipemia and inflammation (GlycA). Furthermore, circulating isoUDCA decreases significantly 1 year after bariatric surgery (β = -0.72, p = 1 × 10^-5) and in response to fiber supplementation (β = -0.37, p < 0.03) but not omega-3 supplementation. In healthy individuals, isoUDCA fasting levels correlate with pre-meal appetite (p < 1 × 10^-4). Our findings indicate an important role for isoUDCA in lipid metabolism, appetite, and, potentially, cardiometabolic risk.

Keywords: bariatric surgery; bile acids; liver function; post-prandial; triglycerides.

Graphical abstract

Figure 1

Schematic representation of the study In this schematic representation of the study, we highlight for each step the research question we want to answer, the analysis workflow, and the data used. We first aimed to estimate how much of the variance in serum and fecal bile acids is explained by the host genome, the gut microbiome, and diet in the TwinsUK and ZOE PREDICT-1 cohorts (step 1). We then investigated which primary and secondary bile acids are correlated with post-prandial lipemia, inflammation, and liver function in ZOE PREDICT-1 and replicated the top hit in the BARIA study (step 2). Finally, we further identified the gut microbiome and dietary features correlating with the BA (isoUDCA) associated with post-prandial lipemia, inflammation, and liver function in TwinsUK and ZOE PREDICT-1 (step 3) and determined how much dietary intervention (Omega-3 and Fiber Intervention to Improve Metabolic Health study) and bariatric surgery (BARIA study) affect its circulating levels.

Figure 2

Host genetics, gut microbiome, and dietary contributions to serum and stool BA concentrations in the TwinsUK and ZOE PREDICT-1 studies (A) Heritability estimates of BAs calculated in TwinsUK and ZOE PREDICT-1 participants. The A, C, and E labels indicate the percentage of variance attributed to the additive genetic factors, common/shared environmental factors, and unique environmental factors, respectively. Labels in bold show the BAs detected in both samples. (B) Prediction of the gut microbiota in BAs levels estimated by random forest regressors (using Spearman’s correlations) and classifiers (using AUC values) in 845 TwinsUK and 327 ZOE PREDICT-1 participants. Dark and light blue bars represent the mean AUC and the 95% confidence intervals across 5-fold for TwinsUK and ZOE PREDICT-1, respectively. Gray and orange bars indicate the mean and the 95% confidence intervals of the Spearman’s correlations between the real value of each BA and the value predicted by the models across 5-fold in TwinsUK and ZOE PREDICT-1, respectively. (C) Random effects inverse variance meta-analyzed coefficients from linear models in TwinsUK and ZOE PREDICT-1 between serum and stool BAs and dietary quality measures (oPDI, original plant diversity index; uPDI, unhealthy plant diversity index; hPDI, healthy plant diversity index; aMED, alternative Mediterranean diet score; HEI, Healthy Eating Index). Significant correlation is denoted by ∗p < 0.05.

Figure 3

Relationship between isoursodeoxycholate, metabolic health, and the gut microbiome Association between fasting serum and stool isoUDCA levels and (A) fasting triglycerides, triglyceride peak, and post-prandial triglyceride change in the ZOE PREDICT-1 cohort and in pre-surgery BARIA study participants (serum isoUDCA only) (for δ triglycerides in BARIA, we investigated the correlation between isoUDCA and δ first and third tertiles); (B) liver function parameters, basal and post-prandial metabolic factors (inflammation, glucose, insulin), and appetite in ZOE PREDICT-1; and (C) single gut microbial species in TwinsUK and ZOE PREDICT-1. Partial Spearman’s correlations calculated adjusting for age, BMI, and sex assigned at birth in 845 TwinsUK participants. Only the characterized species with a prevalence >20% that had significant correlations (FDR < 0.05) and presented the same directional effects in TwinsUK and ZOE PREDICT-1 are shown. Species are hierarchically clustered (complete linkage, Euclidean distance).

Figure 4

Interventions that modify isoUDCA (A) Effect of bariatric surgery on fasting levels of isoUDCA in the BARIA cohort (94% laparoscopic Roux-en-Y gastric bypass). isoUDCA levels were measured pre-surgery and 12 months after surgery. (B) Effect on fasting isoUDCA levels following a 6-week dietary supplementation intervention with 20 g/day inulin (C) or with 500 mg omega-3. (D) Correlation between fasting isoUDCA levels and self-reported fasting appetite in the ZOE PREDICT-1 study.