Skip to main content
NCBI home page
Diabetes logoLink to Diabetes
. 2024 May 3;73(8):1215–1228. doi: 10.2337/db23-0676

Comprehensive Clinical and Genetic Analyses of Circulating Bile Acids and Their Associations With Diabetes and Its Indices

Ibrahim Choucair 1,2,*, Deepthi P Mallela 1,2,*, James R Hilser 3,4, Jaana A Hartiala 3, Ina Nemet 1,2, Valentin Gogonea 1,2,5, Lin Li 1,2, Aldons J Lusis 6,7, Michael A Fischbach 8, WH Wilson Tang 1,2,9, Hooman Allayee 3,4,✉, Stanley L Hazen 1,2,9,✉
PMCID: PMC11262044  PMID: 38701355

Abstract

Bile acids (BAs) are cholesterol-derived compounds that regulate glucose, lipid, and energy metabolism. Despite their significance in glucose homeostasis, the association between specific BA molecular species and their synthetic pathways with diabetes is unclear. Here, we used a recently validated, stable-isotope dilution, high-performance liquid chromatography with tandem mass spectrometry method to quantify a panel of BAs in fasting plasma from 2,145 study participants and explored structural and genetic determinants of BAs linked to diabetes, insulin resistance, and obesity. Multiple 12α-hydroxylated BAs were associated with diabetes (adjusted odds ratio [aOR] range, 1.3–1.9; P < 0.05 for all) and insulin resistance (aOR range, 1.3–2.2; P < 0.05 for all). Conversely, multiple 6α-hydroxylated BAs and isolithocholic acid (iso-LCA) were inversely associated with diabetes and obesity (aOR range, 0.3–0.9; P < 0.05 for all). Genome-wide association studies revealed multiple genome-wide significant loci linked with 9 of the 14 diabetes-associated BAs, including a locus for iso-LCA (rs11866815). Mendelian randomization analyses showed genetically elevated deoxycholic acid levels were causally associated with higher BMI, and iso-LCA levels were causally associated with reduced BMI and diabetes risk. In conclusion, comprehensive, large-scale, quantitative mass spectrometry and genetics analyses show circulating levels of multiple structurally specific BAs, especially DCA and iso-LCA, are clinically associated with and genetically linked to obesity and diabetes.

Article Highlights

  • The association between specific circulating bile acid (BA) molecular species and diabetes in humans is not extensively studied.

  • In the present study, we found that multiple 12α-hydroxylated BAs are positively associated with diabetes and insulin resistance, whereas multiple 6α-hydroxylated BAs are inversely associated with diabetes and obesity.

  • Mendelian randomization analyses show isolithocholic acid levels are causally associated with reduced risk for diabetes and obesity, and deoxycholic acid levels are causally associated with obesity risk.

  • The findings of this study provide new avenues for investigating the role of BAs in metabolic diseases and may have substantial therapeutic implications.

Introduction

Diabetes is a growing public health concern (1), imposing direct health care costs of >$300 billion/year in the U.S. (2) and necessitating more efficient, long-term solutions. Progression of diabetes is associated with increased risk of cardiovascular disease, stroke, renal disease, and death (3,4). In recent years, bile acids (BAs) have gained importance as a novel, potential therapeutic target for diabetes treatment (5,6). In addition to their well-recognized roles in digestion, lipid absorption, and cholesterol metabolism (7), BAs have also been causally linked to several metabolic processes (7,8). For example, BAs can act as signaling molecules in glucose homeostasis by regulating glucose, lipid, and energy metabolism in the liver and gut via receptor-dependent and independent pathways (9,10). Thus, BAs may play a crucial role in the pathogenesis of diabetes (11).

Structurally, BAs are cholesterol-derived molecules with a steroid nucleus and are broadly classified into primary and secondary BAs (9,10). The primary BAs, synthesized by the liver, in humans include cholic acid (CA), chenodeoxycholic acid (CDCA), and hyocholic acid (HCA) (9,10,12). In the classical pathway, CA is produced by 12α-hydroxylation of the steroid ring by sterol 12α-hydroxylase (CYP8B1), whereas CDCA is synthesized when 12α-hydroxylation does not occur (9,10,12). Through 6α-hydroxylation of CDCA, the liver enzyme CYP3A4 produces HCA, a primary BA synthesized in lower concentrations in humans (12). The metabolism of primary BAs by gut microbes yields many secondary BAs, including multiple distinct 12α-hydroxylated (12α-OH) and 6α-hydroxylated (6α-OH) BAs (9,10). Recent studies have suggested that 12α-OH BAs are associated with insulin resistance (IR) (13–15). Furthermore, studies in diabetic mouse models suggest 6α-OH BAs can improve glucose homeostasis through glucagon-like peptide-1 (GLP-1) activation via both farnesoid X receptor and Takeda G protein-coupled receptor 5 (TGR5) signaling (16,17). A deeper understanding of BA synthesis pathways, the relevant structural modifications of BAs in humans, and their associations with metabolic phenotypes could be beneficial in developing therapies for the treatment of diabetes and obesity (13).

Despite the growing appreciation of both the importance of BAs to metabolic regulation and their expanding structural diversity, there is limited understanding of the distinct molecular species of BAs and their synthesis pathways in the context of metabolic disorders. BAs are a large and diverse group of structurally similar molecules whose quantification can pose significant challenges. Because of the complexity involved in quantification of structurally similar BAs, we developed a stable-isotope dilution, liquid chromatography with online tandem mass spectrometry (LC/MS/MS) method to simultaneously quantify a large panel of individual BA molecular species in biological matrices, including human plasma (18).

In this study, we conducted comprehensive clinical and genetic analyses of a large panel of multiple structurally specific BA molecular species and explored their associations with diabetes, indices of IR, and obesity. Given their apparent quantitative associations with diabetes, we further investigated the association between diabetes and both 12α- and 6α-OH BA molecular species by superimposing them on the BA metabolic pathways. We also performed genome-wide association studies (GWAS) to identify the loci associated with variations in circulating levels of BAs clinically linked with metabolic disease. Finally, we conducted Mendelian randomization (MR) analyses to investigate any potential causal relationships between the significantly associated BA levels and metabolic disorders.

Research Design and Methods

Study Population

Sequential participants (N = 2,145) from the Cleveland Clinic GeneBank Study participated in this study. GeneBank is a large, single-center, prospective observational repository with a connecting clinical database consisting of stable participants undergoing elective diagnostic cardiac evaluations at the Cleveland Clinic (Genebank; ClinicalTrials.gov identifier NCT00590200). All study protocols were approved by the institutional review board of the Cleveland Clinic, and all participants gave written informed consent. This study adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.

Clinical Outcomes

The primary outcome was adjudicated diagnosis of diabetes, and secondary outcomes were HOMA-IR and obesity. Diabetes was defined as a history of type 1 or type 2 diabetes, or by laboratory tests (performed on the day of sample acquisition) showing hemoglobin A1c (HbA1c) ≥6.5% or fasting blood glucose ≥126 mg/dL or history of an oral glucose tolerance test with a 2-h glucose level of ≥200 mg/dL, or antidiabetic medication use. HOMA-IR was calculated using the following formula: fasting glucose (mg/dL) × fasting insulin (μU/mL)/405 (19) and was used to define individuals as IR, using a cutoff of 1.97 (20). Participants were defined as obese if their BMI was ≥30.0 kg/m2.

Measurement of BAs

Recently described and validated, stable-isotope dilution LC-MS/MS methodology (18) was used to quantify 60 structurally specific molecular species of BAs found in humans and rodents. Samples examined were overnight (≥12 h) fasting plasma samples from clinical patients. Only the subset of BAs whose circulating levels met predetermined benchmark quality control parameters (n = 36; see Supplementary Material) were carried forward for analysis. Analyte limit of detection (LOD) used a minimal threshold signal-to-noise ratio of 3:1, and a minimal threshold signal-to-noise ratio of 10:1 was used for limit of quantification. Only BAs with concentrations higher than the LOD in plasma samples of a third or more of the study participants were included in our analyses, leading to inclusion of 36 BAs in the present analyses. For statistical analysis, participants whose BA levels were not detectable (below LOD) were assigned a value of one-half of the lowest concentration of the analyte (minimum value) quantified.

Statistical Analysis

Statistical analysis was restricted to individuals (N = 2,145; n = 1,372 men, n = 773 women) with no missing data on the primary outcome (diabetes) and exposure (36 BAs). The normality of variable distribution was determined using the Shapiro-Wilk test and nonparametric tests were used for the nonparametric data comparisons. For baseline characteristics (Table 1), the Wilcoxon rank-sum test for continuous data and Pearson χ2 test for categorical variables were performed using R (R, version 4.1.3, for Windows 64 bit) with scripts developed in-house. Continuous variables were summarized as the median (25th and 75th percentiles), and categorical variables were reported as percentages. In Fig. 1, P values were calculated using Wilcoxon rank-sum tests to evaluate the statistical differences in the median BA concentrations between participants with diabetes (n = 475) and those without diabetes (n = 1,670). Stata statistical software (StataCorp, release 17, 2021) was used to calculate odds ratios (ORs) and 95% CIs as well as the Spearman (Supplementary Fig. 3) and partial Spearman (Supplementary Fig. 4) correlations between BA concentrations and diabetes- and obesity-relevant clinical phenotypes.

Table 1.

Baseline characteristics of study participants (N = 2,145)

Characteristic* Participants without diabetes (n = 1,670) Participants with diabetes (n = 475) P value†
Age, years 62.4 (54.5–71.3) 64.4 (57.5–72.1) 0.001
Male sex, n (%) 1,120 (67.1) 252 (53.1) <0.001
Systolic blood pressure, mmHg 132.0 (119.0–145.0) 132.0 (119.0–149.0) 0.810
Diastolic blood pressure, mmHg 75.0 (68.0–82.0) 71.0 (64.0–80.0) <0.001
Current smoking, n (%) 225 (13.5) 49 (10.3) 0.069
BMI, kg/m2 28.0 (25.2–31.4) 30.4 (27.0–35.7) <0.001
LDLc, mg/dL 96.0 (78.0–117.0) 95.0 (74.0–117) 0.238
HDLc, mg/dL 34.5 (28.4–42.5) 33.5 (27.2–40.4) 0.002
TG, mg/dL 110.0 (82.0–154.0) 135.0 (93.0–190.0) <0.001
Glucose, mg/dL 97.3 (89.9–105.0) 135.0 (113.8–168.8) <0.001
Insulin, µU/mL 7.0 (4.5–10.5) 12.3 (7.7–19.4) <0.001
HbA1c, % 5.6 (5.2–5.9) 6.8 (6.2–7.6) <0.001
HbA1c, mmol/mol 38.0 (33.0–41.0) 51.0 (44.0–60.0) <0.001
Statin use, n (%) 954 (57.1) 312 (65.7) 0.001
Aspirin use, n (%) 1,197 (71.7) 354 (74.5) 0.221
Antidiabetic medication, n (%) 0 (0) 260 (54.7) NA
Open in a new tab

Continuous data are reported as median (IQR) and categorical variables are presented as n (%). IQR, interquartile range.

*

Baseline characteristics of the participants studied are included, with participants stratified by the presence of diabetes. Of note, of the total number of participants with diabetes in our cohort (n = 475), only 9 had type 1 diabetes and 466 had type 2 diabetes.

†

P values were calculated using Wilcoxon rank-sum test for continuous data and Pearson χ2 test for categorical variables.

Figure 1.

Figure 1

Open in a new tab

Plasma BA concentrations showing significant differences between participants with and without diabetes. Box and whisker plots of BA metabolites in human plasma in participants with (n = 475) and without (n = 1,670) diabetes. The lower and upper lines of the box indicate the 25th and 75th percentiles; the line in the middle is the median; and the upper and lower whiskers are 10th and 90th percentiles. BAs in red showed significantly higher concentrations in participants with diabetes, and BAs in green showed significantly lower concentrations in participants with diabetes compared with participants without diabetes. P value was calculated using Wilcoxon rank-sum test. P < 0.05 is significant. **FDR-corrected P < 0.05, ***FDR-corrected P < 0.001.

Univariable and multivariable logistic regression analyses were performed to calculate the ORs and 95% CIs using diabetes, HOMA-IR, and obesity as the dependent variables and the plasma concentrations of various BAs as the independent variables. Through forest plots, we reported the ORs and 95% CIs for all 36 BAs for diabetes, HOMA-IR, and obesity (Figs. 2 and 3 and Supplementary Figs. 1 and 2) in univariable models (unadjusted) and then in multivariable models, after adjusting for traditional risk factors, including age, sex, smoking, systolic blood pressure, and levels of HDL cholesterol (HDLc), LDL cholesterol (LDLc), triglycerides (TGs), and C-reactive protein (CRP). Similar analyses were also performed for grouped plasma BAs (namely, total, free, conjugated, total glycine and taurine conjugated, total primary, total secondary, total 12α-OH, total 6α-OH, and ratios of grouped plasma BAs [12α-OH to total; 12α-OH to non–12α-OH; 6α-OH to total; and 6α-OH to non–6α-OH) (Fig. 4). We also performed the same multilogistic regression analyses on the subcohort of participants who were not taking diabetes medications (n = 1,885) (Supplementary Fig. 1), as well as with additional adjustments including HOMA-IR and BMI in the fully loaded multivariable models (after excluding all participants taking exogenous insulin) (Supplementary Fig. 2). Partial Spearman correlation analyses were performed after adjusting for confounders (traditional risk factors including BMI, HOMA-IR, and/or diabetic status [Supplementary Fig. 4]). ORs presented for diabetes, HOMA-IR, and obesity for all participants, as well for the subset of participants not taking any antidiabetes medications, indicate the risk of fourth quartile (Q4) versus first quartile (Q1) of each analyte. We performed Hosmer-Lemeshaw diagnostic tests to assess the model fits for the logistic regression models. P < 0.05 was considered statistically significant.

Figure 2.

Figure 2

Open in a new tab

CA and its derivatives in metabolic diseases. A: Metabolic pathways of CA metabolism by the liver (solid arrows) and gut microbial enzymes (dashed arrows). Red chemical structures represent BAs positively associated with diabetes; green structures represent BAs negatively associated with diabetes. The purple chemical structure represents the primary BA, CA. B: Associations of CA-derived BAs with diabetes and its indices. Forest plots of ORs for diabetes, HOMA-IR (defined as ≥1.97), and obesity (defined as BMI ≥30 kg/m2) of Q4 vs. Q1; bars represent 95% CIs. Closed circles represent unadjusted ORs, and the open circles represent adjusted ORs for age, sex, smoking, systolic blood pressure, and levels of HDLc, LDLc, TGs, and CRP. When the association was statistically significant (P < 0.05), the OR (circle) and 95% CI (line) are red (positive association) or green (inverse association). If both the unadjusted and adjusted associations of BA with diabetes were significant, the label names for BAs are in red (positive association) or green (inverse association), and in black if just the unadjusted or adjusted association was significant or if both were not significant.

Figure 3.

Figure 3

Open in a new tab

CDCA and its derivatives in metabolic diseases. A: Metabolic pathways of CDCA metabolism by liver (solid arrows) and gut microbial enzymes (dashed arrows). Red chemical structures represent BAs positively associated with diabetes, and green structures represent BAs negatively associated with diabetes. The purple chemical structure represents the primary BA, CDCA. HCA is also another primary BA colored in green (indicating a significant inverse association with diabetes). HDCA (indicated by *) generated from LCA, HCA, α-muricholic acid (3,6,7-trihydroxy acid), and ω-muricholic acid (3,6,7-trihydroxy acid) in mice, as reported (19,20). B: Associations of CDCA-derived BAs with diabetes and its indices. Forest plot of ORs for diabetes, HOMA-IR (defined ≥1.97), and obesity (defined as BMI ≥30 kg/m2) of Q4 vs. Q1; bars represent 95% CIs. Closed circles represent unadjusted ORs, and open circles represent adjusted ORs for age, sex, smoking, systolic blood pressure, and levels of HDLc, LDLc, TGs, and CRP. When the association was statistically significant (P < 0.05), the OR (circle) and 95% CI (line) are red (positive association) or green (inverse association). If both the unadjusted and adjusted associations of BA with diabetes were significant, the label names for BAs are in red (positive association) or green (inverse association), and in black if just the unadjusted or adjusted association was significant or if both were not significant.

Figure 4.

Figure 4

Open in a new tab

Associations of grouped plasma BA concentrations with diabetes, HOMA-IR, and obesity. Forest plots of ORs for diabetes, HOMA-IR, and obesity of Q4 vs. Q1 for the cohort including participants taking antidiabetic medication (n = 2,145) vs. the cohort excluding participants taking antidiabetic medication (no diabetes medications; n = 1,885). Bars represent 95% CIs. Closed circles represent unadjusted ORs, and open circles represent ORs adjusted for age, sex, smoking, systolic blood pressure, and levels of HDLc, LDLc, TGs, and CRP. When the association was statistically significant (P < 0.05), the OR (circle) and 95% CI (line) are red (positive association) or green (inverse association). If both the unadjusted and adjusted associations of BA with diabetes (N = 2,145) were significant, the label names for BAs are in red (positive association) or green (inverse association), and in black if just the unadjusted or adjusted association was significant or if both were not significant. Conj., conjugated; 12α-OH/Total, ratio of 12α-OH BAs to total BAs; 12α-OH/non–12α-OH, ratio of 12α-OH BAs to non–12α-OH BAs; 6α-OH/Total, ratio of 6α-OH BAs to total BAs; 6α-OH/non–6α-OH, ratio of 6α-OH BAs to non–6α-OH BAs.

An expanded Methods section with a detailed summary of the mass spectrometry quantification method and the clinical, genetic, and MR analyses is provided in the Supplementary Material.

Data and Resource Availability

Source data for figures and tables, summary statistics for individual BAs and BA groupings, and custom code used for both clinical analyses and GWAS studies are all available at https://doi.org/10.5281/zenodo.10557984. There are restrictions to the availability of some of the clinical data generated in this study for the GeneBank cohort, because the informed consent from research patients in this study does not permit data sharing outside our institution without authorization. Where permissible, the data sets created and/or analyzed during this study are available upon request from the corresponding author.

Results

Clinical Characteristics of Study Population

In this study, we examined the association of circulating levels of a large panel of structurally specific BAs with prevalent diabetes and several of its quantitative indices of glycemic control and IR. Characteristics of patients (N = 2,145) and laboratory results for overnight fasting blood samples were obtained at the time of participant enrollment; results are listed in Table 1. Participants with diabetes (compared with participants without diabetes) were older, had higher BMI and TG levels, lower HDLc levels, and were more likely to be taking statins (Table 1).

Association of Plasma BA Levels With Diabetes, IR, and Obesity

We first quantified fasting plasma levels of a panel (n = 60) of structurally specific BAs by stable-isotope dilution LC-MS/MS (Supplementary Methods). Only the subset of BAs whose circulating levels met predetermined benchmark quality control parameters (Supplementary Methods) were included in this study (n = 36). Supplementary Table 1 provides the common names, systematic names, and chemical structures of these 36 BAs, and Supplementary Table 2 lists the subgroup classification of all measured BAs. Notably, the fasting plasma concentrations of BAs in humans span four orders of magnitude and often are not normally distributed, yet numerous specific BA species showed statistically significant enrichment (e.g., n = 12 BAs) and others depletion (n = 8 BAs), in participants with diabetes (Fig. 1 and Supplementary Table 3). The statistical differences between the participants without diabetes and those with diabetes, for most BAs, remained significant even when looking only among the subgroup of participants who were not taking antidiabetic medication (n = 1,885) and after false discovery rate (FDR) correction for multiple hypothesis testing (Supplementary Table 3).

To assist in interpreting whether patterns in the significant differences observed had underlying potential biochemical meaning, we superimposed results of analyses on BA metabolic pathways, with the odds of fasting plasma levels (Q4 vs. Q1) of a given BA being significantly associated with diabetes, HOMA-IR (using a cutoff of ≥1.97) (21,22), and association with obesity (defined by BMI ≥30 kg/m2) (23) (Figs. 2 and 3).

Participants with diabetes had significantly higher levels of several 12α-OH BAs derived from CA, including lithocholenic acid and deoxycholic acid (DCA), and numerous of its derivative metabolites (namely, 12-ketolithocholic acid [12-keto-LCA], 23-nordeoxycholic acid [23-nor-DCA], glycodeoxycholic acid [GDCA], and taurodeoxycholic acid [TDCA]; P < 0.05 for all [Fig. 2]). Furthermore, several 12α-OH BAs exhibited similar patterns in terms of being significantly associated (or trending) with HOMA-IR and obesity with taurocholic acid (TCA) and isodeoxycholic acid being positively associated with both, whereas DCA and glycocholic acid were associated with HOMA-IR and not with obesity. Ursocholic acid was inversely associated with diabetes after risk-factor adjustment but did not approach statistical significance with HOMA-IR or obesity.

Plasma BA Associations With Diabetes, IR, and Obesity Among Participants Not Taking Antidiabetic Medication

Because diabetes medications are reported to potentially affect BA composition (24–26), in additional studies we restricted analyses to only the subset of participants not taking diabetes medications at the time of study enrollment and blood sample collection (n = 1,885). The observed pattern of BAs whose circulating levels were associated with diabetes, HOMA-IR, and obesity were highly similar when comparing all participants versus the subset not taking diabetes medications (Supplementary Fig. 1). Specifically, some of the same 12α-OH BAs continued to show strong positive associations (or trends) with either diabetes or HOMA-IR (e.g., TDCA, and tendency with GDCA and TCA for HOMA-IR), or the combination of both HOMA-IR and obesity (e.g., TCA, and tendency with isodeoxycholic acid). Similarly, among the subset of participants not taking diabetes medications, circulating levels of many 6α-OH BAs continued to show strong negative associations with both diabetes and obesity (e.g., glycohyocholic acid [GHCA], HCA, taurohyocholic acid, and hyodeoxycholic acid [HDCA] [Fig. 3 and Supplementary Fig. 1]).

Interestingly, despite the associations with diabetes and obesity, none of the 6α-OH BAs were significantly inversely associated with HOMA-IR (except for a trend with HDCA, which was no longer associated with HOMA-IR when the analysis was carried out in the subset of participants who did not take diabetes medications) (Supplementary Fig. 1). By comparison, higher levels of the CDCA-derived metabolite isolithocholic acid (iso-LCA) remained strongly and consistently inversely associated with both diabetes and obesity regardless of participants taking or not taking diabetes medications (Supplementary Fig. 1). Similarly, 7-ketolithocholic acid remained inversely associated with diabetes within the subset of participants not taking any diabetes medications (Supplementary Fig. 1). In further analyses, the associations of individual 12α-OH BAs and 6α-OH BAs with diabetes and HOMA-IR remained essentially unchanged when further adjusted for BMI and HOMA-IR (in the case of diabetes as outcome) in the fully loaded model. The only exceptions noted were for TDCA, GDCA, and GHCA, which lost significance in associations with diabetes after further adjustments, and also when examined in the subcohort of participants without diabetes medication (Supplementary Fig. 2).

Association of Broad Categories of Plasma BA Concentrations With Diabetes and Metabolic Indices

In further analyses, rather than examining the BAs as individual molecular species, we looked at various groupings of the BAs based on sites of sterol ring hydroxylation (Fig. 4). Notably, in keeping with published studies (13,27), total BAs (i.e., the sum of all individual molecular species concentrations), various groupings of total BAs, free and conjugated forms of BAs, and total primary BAs, or total secondary BAs, all were positively associated with HOMA-IR, even after adjustments for traditional diabetes risk factors. In contrast, however, none of these groupings was significantly associated with either diabetes or obesity (Fig. 4).

When examining the 12α-OH BAs, the positive association noted with diabetes for the sum total of 12α-OH BAs was lost when adjusted for traditional risk factors and the positive association noted with diabetes for ratio relative to total BAs (i.e., 12α-OH BAs to total BAs) or to non–12α-OH BAs, like other BA groups, attenuated when limiting analyses to among only the subset of participants not taking diabetes medications. The 12α-OH BAs, along with many of the BA groups, were significantly associated with HOMA-IR (including in the subset of participants not taking any diabetes medications), but not with obesity (Fig. 4). Notably, 6α-OH BAs, whether analyzed as sum total levels or as a ratio relative to total BAs or to non–6α-OH BAs, were consistently inversely associated with prevalent diabetes, HOMA-IR, and obesity, including after adjustments for traditional risk factors or whether examined in the subset of participants not taking diabetes medications (Fig. 4).

Spearman Correlations Between BAs and Diabetes- and Obesity-Relevant Phenotypes

To further examine the relationships of individual BAs with various relevant diabetes- and obesity-related phenotypes (namely, fasting glucose, insulin, glucose to insulin ratio, HbA1c, HOMA-IR, HOMA for β-cell function, weight and BMI), we analyzed their Spearman correlations (Supplementary Fig. 3). Plasma levels of multiple 12α-OH BAs were significantly correlated with insulin, glucose to insulin ratio, HOMA-IR, and HOMA for β-cell function (red boxes in Supplementary Fig. 3). Most of these associations remained significant even after adjusting for confounders (red boxes in Supplementary Fig. 4) on a partial Spearman correlation. Moreover, multiple 6α-OH BAs, and iso-LCA, showed strong inverse correlations with diabetes and obesity-related phenotypes (blue boxes in Supplementary Fig. 3), and iso-LCA concentrations remained consistently significantly inversely correlated with BMI even after risk-factor adjustment (Supplementary Fig. 4).

Genetic Determinants of Plasma BA Levels and Their Relationships to Metabolic Disease

All participants examined with quantitative LC-MS/MS analyses for the BAs had genotype data available. Therefore, we next sought to perform GWAS to identify host genetic factors associated with variations in circulating levels of BAs and determine their relationship with metabolic disease. At the genome-wide threshold (P = 5.0 × 10−8), GWAS analyses revealed 12 significant associations (Table 2, Supplementary Fig. 5A–I, and Supplementary Fig. 6A–K) for 9 of the 14 BAs clinically associated with diabetes (Figs. 2 and 3). All minor alleles of the lead single nucleotide polymorphisms (SNPs) at these loci were associated with increased BA levels, and nearly half of the variants had relatively rare frequencies (∼5% or less) (Table 2). Although most of the genetic associations were specific to only one BA, some variants, such as those for DCA, TDCA, and 12-keto-LCA, exhibited pleiotropic associations with other BAs, as well (Supplementary Table 4). By comparison, no genome-wide significant loci were identified for levels of 7-ketolithocholic acid, HCA, taurohyocholic acid, GHCA, or HDCA, although several regions did yield suggestive (P = 5.0 × 10−6) association signals (Supplementary Fig. 7A–E).

Table 2.

Loci identified for BAs associated with diabetes in GeneBank

Bile acid SNP Chr:Pos Nearest gene EA/OA EAF β (SE*) P value†
DCA rs773141 9:77664595 NMRK1 A/G 0.5 0.19 (0.03) 1.7 × 10−8
Lithocholic acid rs73187945 21:27726757 CYYR1 G/A 0.02 2.33 (0.41) 1.8 × 10−8
TDCA rs73127925 3:81284712 GBE1 C/T 0.02 0.76 (0.13) 1.3 × 10−8
TDCA rs111835596 21:41178635 IGSF5 A/G 0.03 0.53 (0.10) 4.4 × 10−8
12-Keto-LCA rs67981690 12:21343886 SLCO1B1 G/A 0.14 0.69 (0.10) 2.2 × 10−11
23-Nor-DCA rs181813865 7:146068542 CNTNAP2 A/G 0.02 1.66 (0.30) 4.7 × 10−8
GDCA rs73127925 3:81284712 GBE1 C/T 0.02 0.75 (0.13) 1.5 × 10−8
6-Keto-LCA rs3816076 4:73175530 ADAMTS3 C/T 0.03 2.31 (0.42) 3.4 × 10−8
6-Keto-LCA rs76270623 8:75156017 JPH1 G/T 0.01 3.35 (0.53) 2.9 × 10−10
6-Keto-LCA rs535408158 16:48666314 N4BP1 CA/C 0.06 2.17 (0.39) 3.6 × 10−8
Iso-LCA rs11866815 16:387867 AXIN1 T/C 0.25 0.49 (0.09) 4.7 × 10−8
Taurohyodeoxycholic acid rs67981690 12:21343886 SLCO1B1 G/A 0.14 0.66 (0.12) 4.2 × 10−8
Open in a new tab

Chr:Pos, chromosome to base pair position (hg19); EA, effect allele; EAF, effect allele frequency; OA, other allele.

*

β values refer to the effect of EA on plasma BA levels.

†

P values were obtained after adjustment for age, sex, and genotyping array.

To prioritize candidate causal genes at the significantly associated loci, we used publicly available expression quantitative trait locus (eQTL) data from the GTEx Project (28), STARNET (Stockholm-Tartu Atherosclerosis Reverse Network Engineering Task) cohort (29), and eQTLGen (30) databases. Of the 10 regions identified for BA levels (Supplementary Table 4), data were only available for rs773141 and rs11866815, which revealed multi-tissue eQTLs for several genes localizing to these two variants for DCA and iso-LCA, respectively (Supplementary Table 5).

Association of BA-Associated Loci With Diabetes and BMI and MR Analyses

We next evaluated the significant BA-associated loci for associations with BMI and diabetes, using the results of recently published, large-scale GWAS analyses (31,32). Notably, association of the BA-increasing alleles rs773141 (DCA) and rs76270623 (6-keto-LCA) with increased BMI and risk of diabetes, respectively, and association of the iso-LCA-raising allele rs11866815 with lower BMI and decreased diabetes risk (Supplementary Table 6) exceeded the Bonferroni-corrected significance threshold of P = 5.0 × 10−3 for testing 10 variants (P = 0.05/10 = 5.0 × 10−3). Furthermore, the direction and specificity of the genetic associations of rs773141 (DCA), rs76270623 (6-keto-LCA), and rs11866815 (iso-LCA) with obesity and diabetes were consistent with the clinical associations observed between these BAs (Fig. 3 and Supplementary Table 6).

Based on the observed associations between genome-wide significant variants for DCA (rs773141), 6-keto-LCA (rs76270623), and iso-LCA (rs11866815) with obesity and diabetes, we performed further studies using two-sample MR analyses to evaluate the effects of DCA, 6-keto-LCA, and iso-LCA on metabolic outcomes. The results showed that genetically increased DCA levels were causally associated with higher BMI (β = 0.051, SE = 0.010; P = 9.3 × 10−7) (Table 3). Similar evidence was observed for a causal association between 6-keto-LCA levels and increased diabetes risk (OR 1.02 [95% CI 1.01–1.03]; P = 9.3 × 10−6), based on only one of the variants tested. By comparison, MR analysis showed that genetically elevated iso-LCA levels were causally associated with reduced BMI (β = −0.031, SE = 0.005; P = 1.6 × 10−11) and reduced diabetes (OR 0.94 [95% CI 0.92–0.97]; P = 4.8 × 10−7) (Table 3). Furthermore, Stieger tests (Supplementary Methods) to evaluate directionality of our instrument variables yielded significant P values and provided consistent evidence for DCA, 6-keto-LCA, and iso-LCA being causal drivers of BMI and diabetes (data not shown).

Table 3.

Two-sample MR analyses of BA-associated variants for diabetes and obesity

Bile acid Variant Diabetes BMI
OR (95% CI) P value β (SE) P value
DCA rs773141 1.02 (0.97–1.07) 0.48 0.051 (0.010) 9.3 × 10−7
6-Keto-LCA rs76270623 1.02 (1.01–1.03) 9.3 × 10−6 0.003 (0.003) 0.19
6-Keto-LCA rs3816076 1.00 (0.99–1.01) 0.88 −0.004 (0.003) 0.14
Iso-LCA rs11866815 0.94 (0.92–0.97) 4.8 × 10−7 −0.031 (0.005) 1.6 × 10−11
Open in a new tab

Identified variants from the GWAS analyses in GeneBank were used as instrument variables to test indicated BAs for causal association with diabetes and BMI, using Wald ratio tests with published effect sizes (29).

Finally, in additional studies, we used our GWAS results to examine 338 previously reported diabetes loci (32) for association with levels of the 14 diabetes-associated BAs. However, only nominally significant (P < 0.05) associations were observed, none of which exceeded the Bonferroni-corrected significance threshold of P = 1.4 × 10−4 for multiple testing (P = 0.05/338 = 1.4 × 10−4) (Supplementary Table 7). On the basis of our MR analyses suggesting that iso-LCA and 6-keto-LCA levels were causally associated with diabetes, we also carried out reverse MR analyses for levels of these BAs with diabetes-associated variants (n = 308 and 293 SNPs for iso-LCA and 6-keto-LCA levels, respectively). However, these analyses did not provide evidence that genetically increased diabetes risk was causally associated with either iso-LCA or 6-keto-LCA (data not shown).

Discussion

We used a recently reported, stable-isotope dilution LC-MS/MS method to quantify a comprehensive panel of circulating BAs (18) and report here both clinical associations and genetic links between circulating levels of individual BA molecular species and metabolic disease. The present examination of BA molecular species associations with diabetes- and obesity-relevant phenotypes demonstrates a link between specific 12α-OH BAs and IR, and multiple specific 6α-OH BAs showing a pronounced inverse association with diabetes and obesity. Moreover, plasma iso-LCA levels showed a striking inverse association with diabetes and obesity. Most importantly, analyses using both GWAS and MR approaches revealed host genetic determinants of circulating levels of specific BAs with causal links to metabolic diseases. Specifically, following our GWAS analyses, MR studies showed genetically increased DCA levels were causally associated with higher BMI, and genetically elevated iso-LCA levels were causally associated with both reduced diabetes and reduced obesity. However, a causal link between genetically increased 6-keto-LCA levels and increased diabetes risk was less consistent (Table 3). Thus, by performing comprehensive biochemical and genetic studies of circulating levels of multiple BA molecular species, our studies have identified new, potential BA-related drivers of altered susceptibility to both diabetes and obesity.

As potent regulators of hepatic cholesterol and TG metabolism, BAs affect insulin sensitivity and glucose regulation (14–17). To our knowledge, this study is the first to identify a clinical and genetic association between individual circulating plasma 12α-OH and 6α-OH BA molecular species with diabetes. The positive associations observed between the plasma levels of some 12α-OH BAs (namely, TDCA, GDCA, 23-Nor-DCA, and DCA [Fig. 2]) and diabetes in the present human clinical observational studies are consistent with results recently reported from animal studies (15,18,33,34) and human studies. Furthermore, in diet-induced obesity animal models, high-fat-fed and high-sucrose-fed rodents had increased levels of 12α-OH BAs (CA, DCA, ursocholic acid, TCA, TDCA, glycocholic acid, and GDCA) in the enterohepatic circulation, implying a direct association between 12α-OH BAs and glucose metabolism in rats (15). Furthermore, in high-fat-fed mice, increased 12α-OH BAs were linked to enhanced mRNA expression of 7α-hydroxylase (Cyp7a1), Cyp8b1, and increased synthesis of both hepatic TG and total cholesterol in the liver (33). DCA was the most abundant circulating 12α-OH BA observed in this study. In prior studies, direct supplementation of DCA in mice was reported to impair glucose regulation (34). CYP8B1 is the 12α-hydroxylase that determines the production of CA (a 12α-OH BA) versus CDCA (a non–12α-OH BA) (35). Recent research has shown that CYP8B1 plays a role in several syndromes and metabolic diseases, including diabetes, obesity, metabolic dysfunction-associated steatotic liver disease, and metabolic dysfunction-associated steatohepatitis (36). Specifically, the absence of functional CYP8B1 both eliminates 12α-OH BAs and results in a reduction in Western diet–induced weight gain, improved insulin sensitivity, and resistance to hepatic steatosis (35).

The results of our MR analyses provide direct genetic causal support for a role for variations in DCA levels in contributing to obesity, because the DCA-raising allele (rs773141) was significantly associated with increased BMI. It is worth noting that large-scale clinical analyses with ∼10,000 subjects from the EPIC-Norfolk study also revealed DCA levels to be significantly and positively associated with BMI (β = 0.067; P = 8.2 × 10−12), obesity (β = 0.103; P = 2.1 × 10−4), and diabetes (β = 0.148; P = 4.8 × 10−3) (37). Thus, the directionality of the clinical and genetic associations we observed between several structurally specific 12α-OH BAs and diabetes, IR (Fig. 2 and Supplementary Figs. 1 and 2), and other diabetes-relevant phenotypes (Supplementary Fig. 3) are consistent with connections between some 12α-OH BAs and diabetes-relevant phenotypes reported in other human cohorts and in animal models of disease (15,33,34).

One of the more consistent findings observed in this study involves 6α-OH BAs and their inverse association with diabetes and obesity. 6α-OH BAs, such as HCA, HDCA, and their conjugates, are dominant BAs in pigs and rodents (which, incidentally, are recognized for being resistant to developing diabetes (17,38)), whereas CA and CDCA and their derivatives are prominent in humans (38). We note that most of the downstream derivatives of HCA in the present study were significantly inversely associated with both diabetes and obesity (Fig. 3). The inverse associations we noted between 6α-OH BA levels and diabetes phenotypes are consistent with some recently reported animal studies (16–18,40). For example, oligofructose supplementation in murine models fed a Western-style diet had increased levels of 6α-OH BAs and improved glucose metabolism. The dietary fiber effects on 6α-OH BAs were microbiota dependent and required functional TGR5 signaling to reduce body weight gain and improve glucose metabolism in the mice (16). In another murine study, HDCA-stimulated TGR5 signaling increased GLP-1 secretion and improved host glucose metabolism in both in vitro and in vivo models (16). HCA and its derivatives have been shown to improve glucose homeostasis by increasing GLP-1 production in both diabetic mouse models and obesogenic swine models by activating TGR5 and inhibiting farnesoid X receptor (17). The effects of 6α-OH BAs on glucose metabolism in humans has not been extensively studied; however, our present studies establish important inverse associations between circulating plasma levels of 6α-OH BAs and both diabetes and obesity.

The origins of 6-keto-LCA, one of the BA molecular species that showed genetic linkage to diabetes in the present study, is worth further discussion. Classically, the origins of 6-keto-LCA are reported to occur from LCA as a precursor via mammalian transformation pathways (41,42). For example, 6-keto-LCA is produced when LCA is incubated with rodent liver microsomes (41), human hepatic microsomes, and human recombinant cytochrome P450 enzymes (42). However, we speculate that 6-keto-LCA may potentially be synthesized via gut microbes, such as from HDCA as precursor, though this has not been corroborated in humans. We suspect a role of gut microbes as a primary participant in the synthesis of plasma 6-keto-LCA based on our recent reported mouse studies in which an oral cocktail of poorly absorbed antibiotics was given to suppress gut microbiota, using established methods and were found to reduce circulating 6-keto-LCA concentrations by 96% (mean ± SD, before antibiotics: 17.2 ± 49.6 nmol/L; after ≥5 days of oral antibiotics: 0.7 ± 0.8 nmol/L; n > 10 mice; P < 0.0001] (18). These results suggest a significant contribution of gut microbes to plasma levels of 6-keto-LCA (at least in mice).

One BA that had both clinical and genetic associations with both diabetes and obesity is iso-LCA. In contrast to either 12α-OH and 6α-OH BAs, iso-LCA is a 3β-hydroxy epimer of LCA, generated by microbial metabolism via the intermediate 3-keto-LCA (43). 5α-Reductase and 3β-hydroxysteroid dehydrogenase are microbial enzymes that have been identified in the production of iso-LCA from LCA in human gut microbiota (44). Higher iso-LCA levels tracked with decreased metabolic risk, and both the direction and specificity of the genetic associations observed with the lead variant (rs11866815) for iso-LCA were consistent with the directionality of clinical associations observed between (higher) iso-LCA and (reduced) diabetes and (reduced) obesity risks. In this regard, numerous antidiabetic therapies are emerging in diabetes management, and BA research has garnered increased scientific interest in recent years. The role of BAs in diabetes management has been demonstrated by the use of BA chelating agents in patients with type 2 diabetes, as well as the significance of bariatric surgery as an effective therapeutic option for obesity and diabetes, with effects thought to be linked, in part, to modifying BA metabolism (5,45). In this context, given the MR analyses demonstrating causality for both diabetes and obesity, we speculate that provision of iso-LCA (either directly, or via a prebiotic or probiotic that fosters enhanced iso-LCA formation) may have potential to be a promising diabetes or obesity pharmacotherapy.

Our GWAS analyses revealed 12 loci associated with several BAs that were clinically associated with metabolic outcomes. For those loci for which the minor allele frequencies of the lead variants were relatively rare, replication in independent data sets will be required to consider them as true association signals. By comparison, the frequency of the BA-promoting allele of the lead variant (rs11866815) at the chromosome 16 locus for iso-LCA levels was common among participants of European ancestry (0.25). Although this locus did not appear to harbor any positional candidate gene with an obvious role in BA metabolism, publicly available databases (28–30) revealed that rs11866815 yielded eQTLs for MRPL28 and ARHGDIG in the esophagus, stomach, pancreas, and/or colon. In addition, other genes at this locus (Supplementary Fig. 6J) have been implicated in diabetes and metabolic traits, including DECR2 (46), PDIA2 (47), and AXIN1 (48). Additional studies will be needed to functionally validate one or more of these genes as the underlying molecular basis for the association between the chromosome 16 locus and iso-LCA levels. AXIN1 is a component of the β-catenin destruction complex regulating the Wnt-signaling pathway (49) and linked to mTOR pathway (50). Although dysregulated mTOR signaling has been linked to various metabolic diseases, including type 2 diabetes and obesity (50), a direct link among AXIN1, BA metabolism, and diabetes has yet to be identified. In addition, it should be noted that the association of rs11866815 with diabetes was attenuated after adjustment for BMI, suggesting that at least a portion of the causal association we identified between iso-LCA levels and diabetes could be mediated through BMI.

Another notable finding from our genetic analyses was the identification of a locus on chromosome 12 harboring SLCO1B1. This locus has been linked to numerous other traits in prior GWAS, including the pharmacogenetic response to statins and other drugs, levels of various clinical biomarkers, vitamins, aberrant lipid profiles, both recognized and undiscovered metabolites, and elevated pre-diabetes risk (51,52). Nearly all these pleiotropic association signals have implicated rs4149056 or other tightly linked (r2 > 0.7) variants. In this regard, rs4149056 leads to a computationally predicted (53) and functionally validated (54) deleterious Val174Ala (GTG > GCG) substitution in SLCO1B1, which encodes an organic anion transporter that mediates hepatic uptake of various compounds (55). Hence, rs4149056 is likely the major causal variant at this locus where carriers of the minor C (Ala) allele have higher circulating levels of the various substrates transported by SLCO1B1, due to diminished hepatocyte influx activity. Consistent with this biological model, the minor allele of our lead SNP on chromosome 12 (rs67981690), which is in high linkage disequilibrium with rs4149056 (r2 = 0.86), was not only significantly associated with increased plasma 12-keto-LCA and taurohyodeoxycholic acid levels but also suggestively associated with increased circulating levels of several other BAs. Taken together, our study, and the observations of other groups (56), add several BA species to the list of compounds likely transported by SLCO1B1, and further illustrate the broad structural substrate recognition of this transporter.

This study has several limitations worth noting. The majority of the participants with diabetes included in our cohort had type 2 diabetes (n = 466)—only nine participants with diabetes had type 1 diabetes; therefore, all analyses presented are for total diabetes and are not stratified on the basis of diabetes subtype. In addition, analyses based on individual antidiabetic medication types could not be performed other than answering yes or no to either insulin use or to taking any antidiabetes medications, because diabetes medication subtype information was not available. Another potential limitation is that results shown represent overnight fasting blood levels of BAs at only one time point; whether serial measures would provide enhanced association with diabetes or obesity remains unknown. In addition, our cohort was recruited at a quaternary referral center and shows a high prevalence of comorbidities, including numerous cardiometabolic disease risk factors; thus, the generalizability of our findings to a community-based cohort remains to be established. Another limitation of the study is its cross-sectional design. We recognize that clinical observational studies, by design, only show association and not causation; thus, there always exists the possibility of unmodelled confounding that may have affected our results through factors not included in our models. To address the issue of causation, we also performed accompanying genetic MR analyses, which can provide statistical evidence for causal relationships. Inspection of data from the GWAS catalog did not reveal other traits associated with the loci identified for the clinically associated BAs, suggesting that conclusions drawn with respect to the causal associations between DCA and iso-LCA levels with BMI and diabetes were likely not confounded by pleiotropic associations of the variants. However, as with any MR analysis, not being able to fully rule out pleiotropy is still a limitation, and follow-up studies are needed to more definitively conclude that DCA and iso-LCA are causal drivers (positive or negative) of metabolic traits.

Conclusion

In conclusion, our study represents a comprehensive clinical and genetic analyses examining the relationship of multiple structurally specific BAs with diabetes, IR, and obesity. Our findings indicate that fasting levels of specific BAs circulating in human plasma are differentially associated with diabetes and obesity, and that a subset of these associations may represent causal biological relationships. These data provide new avenues for exploring the role of BAs in metabolic disease and may have important therapeutic implications.

This article contains supplementary material online at https://doi.org/10.2337/figshare.25731687.

Article Information

Acknowledgments. The authors thank Taylor Weeks, Cleveland Clinic, for assistance with the preparation of the manuscript, its figures, and its submission.

Funding. This work was supported by the National Institutes of Health and the Office of Dietary Supplements (grants P01-HL147823, R01-HL167831, R01-HL103866, R01-HL133169, R01-HL148110, and DP1-DK113598) and the Foundation Leducq (grant 17CVD01).

Duality of Interest. S.L.H. reports being named as co-inventor on pending and issued patents held by the Cleveland Clinic relating to cardiovascular diagnostics and therapeutics; being a paid consultant for and having received research funds from Zehna Therapeutics; and being eligible to receive royalty payments for inventions or discoveries related to cardiovascular diagnostics or therapeutics from Cleveland HeartLab, Quest Diagnostics, and Procter & Gamble. M.A.F. is a cofounder and director of Federation Bio and Viralogic; a cofounder of Revolution Medicines; and reports consultancy work for NGM Bio; ownership interest in Kelonia and NGM Bio; patents or royalties involving Federation Bio; advisory or leadership roles with Federation Bio and Kelonia; is on the board of NGM Biopharmaceuticals; and is an innovation partner at The Column Group. W.H.W.T. reports being a consultant for Sequana Medical A.G., Owkin Inc., Relypsa Inc., and PreCardiac Inc.; having received honorarium from Springer Nature for authorship or editorship and American Board of Internal Medicine for exam writing committee participation, all of which are unrelated to the subject and contents of this article. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. I.C. and D.P.M. contributed equally to data collection, formal analysis, and writing and editing of the manuscript. J.R.H., J.A.H., and I.N. contributed to data acquisition and formal analysis, and review and editing of the manuscript. L.L. contributed to initial statistical analyses. V.G., A.J.L., M.A.F., and W.H.W.T. contributed to formal analysis; writing, review, and editing of the manuscript; and funding acquisition. H.A. and S.L.H. contributed equally to study conceptualization; analysis; writing, reviewing, and editing the manuscript; project supervision and administration; and funding acquisition. H.A. and S.L.H. are the guarantors of this work and, as such, had full access to all the data in the study and take full responsibility for the integrity of data and the accuracy of the data analysis.

Funding Statement

This work was supported by the National Institutes of Health and the Office of Dietary Supplements (grants P01-HL147823, R01-HL167831, R01-HL103866, R01-HL133169, R01-HL148110, and DP1-DK113598) and the Foundation Leducq (grant 17CVD01).

Footnotes

I.C. is currently affiliated with the Department of Pathology and Laboratory Medicine, University of Kentucky College of Medicine, Lexington, KY.

This article is featured in a podcast available at diabetesjournals.org/diabetes/pages/diabetesbio.

References

  • 1. Khan MAB, Hashim MJ, King JK, Govender RD, Mustafa H, Al Kaabi J. Epidemiology of type 2 diabetes - global burden of disease and forecasted trends. J Epidemiol Glob Health 2020;10:107–111 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. American Diabetes Association . Economic costs of diabetes in the U.S. in 2017. Diabetes Care 2018;41:917–928 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Fonseca VA. Defining and characterizing the progression of type 2 diabetes. Diabetes Care 2009;32(Suppl. 2):S151–156 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Raghavan S, Vassy JL, Ho YL, et al. Diabetes mellitus-related all-cause and cardiovascular mortality in a national cohort of adults. J Am Heart Assoc 2019;8:e011295. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Handelsman Y. Role of bile acid sequestrants in the treatment of type 2 diabetes. Diabetes Care 2011;34(Suppl. 2):S244–250 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Mudaliar S, Henry RR, Sanyal AJ, et al. Efficacy and safety of the farnesoid X receptor agonist obeticholic acid in patients with type 2 diabetes and nonalcoholic fatty liver disease. Gastroenterology 2013;145:574–82.e1 [DOI] [PubMed] [Google Scholar]
  • 7. Staels B, Fonseca VA. Bile acids and metabolic regulation: mechanisms and clinical responses to bile acid sequestration. Diabetes Care 2009;32(Suppl. 2):S237–245 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Watanabe M, Houten SM, Mataki C, et al. Bile acids induce energy expenditure by promoting intracellular thyroid hormone activation. Nature 2006;439:484–489 [DOI] [PubMed] [Google Scholar]
  • 9. Li T, Chiang JY. Bile acid signaling in metabolic disease and drug therapy. Pharmacol Rev 2014;66:948–983 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. Fiorucci S, Distrutti E. The pharmacology of bile acids and their receptors. Handb Exp Pharmacol 2019;256:3–18 [DOI] [PubMed] [Google Scholar]
  • 11. Alberto González-Regueiro J, Moreno-Castañeda L, Uribe M, Carlos Chávez-Tapia N. The role of bile acids in glucose metabolism and their relation with diabetes. Ann Hepatol 2017;16(Suppl. 1):S15–S20 [DOI] [PubMed] [Google Scholar]
  • 12. Elliott WH, Hyde PM. Metabolic pathways of bile acid synthesis. Am J Med 1971;51:568–579 [DOI] [PubMed] [Google Scholar]
  • 13. Haeusler RA, Astiarraga B, Camastra S, Accili D, Ferrannini E. Human insulin resistance is associated with increased plasma levels of 12α-hydroxylated bile acids. Diabetes 2013;62:4184–4191 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Haeusler RA, Pratt-Hyatt M, Welch CL, Klaassen CD, Accili D. Impaired generation of 12-hydroxylated bile acids links hepatic insulin signaling with dyslipidemia. Cell Metab 2012;15:65–74 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Yoshitsugu R, Kikuchi K, Hori S, et al. Correlation between 12α-hydroxylated bile acids and insulin secretion during glucose tolerance tests in rats fed a high-fat and high-sucrose diet. Lipids Health Dis 2020;19:9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Makki K, Brolin H, Petersen N, et al. 6α-hydroxylated bile acids mediate TGR5 signalling to improve glucose metabolism upon dietary fiber supplementation in mice. Gut 2023;72:314–324 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Zheng X, Chen T, Jiang R, et al. Hyocholic acid species improve glucose homeostasis through a distinct TGR5 and FXR signaling mechanism. Cell Metab 2021;33:791–803.e7 [DOI] [PubMed] [Google Scholar]
  • 18. Choucair I, Nemet I, Li L, et al. Quantification of bile acids: a mass spectrometry platform for studying gut microbe connection to metabolic diseases. J Lipid Res 2020;61:159–177 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Einarsson K. On the formation of hyodeoxycholic acid in the rat. Bile acids and steroids 154. J Biol Chem 1966;241:534–539 [PubMed] [Google Scholar]
  • 20. Eyssen HJ, De Pauw G, Van Eldere J. Formation of hyodeoxycholic acid from muricholic acid and hyocholic acid by an unidentified gram-positive rod termed HDCA-1 isolated from rat intestinal microflora. Appl Environ Microbiol 1999;65:3158–3163 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Matsuda M. Measuring and estimating insulin resistance in clinical and research settings. Nutr Metab Cardiovasc Dis 2010;20:79–86 [DOI] [PubMed] [Google Scholar]
  • 22. Lee CH, Shih AZ, Woo YC, et al. Optimal cut-offs of homeostasis model assessment of insulin resistance (HOMA-IR) to identify dysglycemia and type 2 diabetes mellitus: a 15-year prospective study in Chinese. PLoS One 2016;11:e0163424. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. American Diabetes Association Professional Practice Committee . 8. Obesity and Weight Management for the Prevention and Treatment of Type 2 Diabetes: Standards of Medical Care in Diabetes—2022. Diabetes Care 2022;45(Suppl. 1):S113–S124 [DOI] [PubMed] [Google Scholar]
  • 24. Wu H, Esteve E, Tremaroli V, et al. Metformin alters the gut microbiome of individuals with treatment-naive type 2 diabetes, contributing to the therapeutic effects of the drug. Nat Med 2017;23:850–858 [DOI] [PubMed] [Google Scholar]
  • 25. Scarpello JH, Hodgson E, Howlett HC. Effect of metformin on bile salt circulation and intestinal motility in type 2 diabetes mellitus. Diabet Med 1998;15:651–656 [DOI] [PubMed] [Google Scholar]
  • 26. Caspary WF, Zavada I, Reimold W, Deuticke U, Emrich D, Willms B. Alteration of bile acid metabolism and vitamin-B12-absorption in diabetics on biguanides. Diabetologia 1977;13:187–193 [DOI] [PubMed] [Google Scholar]
  • 27. Wewalka M, Patti ME, Barbato C, Houten SM, Goldfine AB. Fasting serum taurine-conjugated bile acids are elevated in type 2 diabetes and do not change with intensification of insulin. J Clin Endocrinol Metab 2014;99:1442–1451 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. GTEx Consortium . The Genotype-Tissue Expression (GTEx) project. Nat Genet 2013;45:580–585 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Franzén O, Ermel R, Cohain A, et al. Cardiometabolic risk loci share downstream cis- and trans-gene regulation across tissues and diseases. Science 2016;353:827–830 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Võsa U, Claringbould A, Westra HJ, et al.; BIOS Consortium ; i2QTL Consortium. Large-scale cis- and trans-eQTL analyses identify thousands of genetic loci and polygenic scores that regulate blood gene expression. Nat Genet 2021;53:1300–1310 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. Loh PR, Kichaev G, Gazal S, Schoech AP, Price AL. Mixed-model association for biobank-scale datasets. Nat Genet 2018;50:906–908 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32. Mahajan A, Spracklen CN, Zhang W, et al.; FinnGen ; eMERGE Consortium. Multi-ancestry genetic study of type 2 diabetes highlights the power of diverse populations for discovery and translation. Nat Genet 2022;54:560–572 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33. Hori S, Abe T, Lee DG, et al. Association between 12α-hydroxylated bile acids and hepatic steatosis in rats fed a high-fat diet. J Nutr Biochem 2020;83:108412. [DOI] [PubMed] [Google Scholar]
  • 34. Zaborska KE, Lee SA, Garribay D, Cha E, Cummings BP. Deoxycholic acid supplementation impairs glucose homeostasis in mice. PLoS One 2018;13:e0200908. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. Bertaggia E, Jensen KK, Castro-Perez J, et al. Cyp8b1 ablation prevents Western diet-induced weight gain and hepatic steatosis because of impaired fat absorption. Am J Physiol Endocrinol Metab 2017;313:E121–E133 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Liu J, Carlson HA, Scott EE. The structure and characterization of human cytochrome P450 8B1 supports future drug design for nonalcoholic fatty liver disease and diabetes. J Biol Chem 2022;298:102344. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Pietzner M, Stewart ID, Raffler J, et al. Plasma metabolites to profile pathways in noncommunicable disease multimorbidity. Nat Med 2021;27:471–479 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Zheng X, Chen T, Zhao A, et al. Hyocholic acid species as novel biomarkers for metabolic disorders. Nat Commun 2021;12:1487. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Petersen AØ, Julienne H, Hyötyläinen T, et al. Conjugated C-6 hydroxylated bile acids in serum relate to human metabolic health and gut Clostridia species. Sci Rep 2021;11:13252. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40. Chávez-Talavera O, Wargny M, Pichelin M, et al. Bile acids associate with glucose metabolism, but do not predict conversion from impaired fasting glucose to diabetes. Metabolism 2020;103:154042. [DOI] [PubMed] [Google Scholar]
  • 41. Deo AK, Bandiera SM. Biotransformation of lithocholic acid by rat hepatic microsomes: metabolite analysis by liquid chromatography/mass spectrometry. Drug Metab Dispos 2008;36:442–451 [DOI] [PubMed] [Google Scholar]
  • 42. Deo AK, Bandiera SM. 3-Ketocholanoic acid is the major in vitro human hepatic microsomal metabolite of lithocholic acid. Drug Metab Dispos 2009;37:1938–1947 [DOI] [PubMed] [Google Scholar]
  • 43. Kelsey MI, Sexton SA. The biosynthesis of ethyl esters of lithocholic acid and isolithocholic acid by rat intestinal microflora. J Steroid Biochem 1976;7:641–647 [DOI] [PubMed] [Google Scholar]
  • 44. Sato Y, Atarashi K, Plichta DR, et al. Novel bile acid biosynthetic pathways are enriched in the microbiome of centenarians. Nature 2021;599:458–464 [DOI] [PubMed] [Google Scholar]
  • 45. Koliaki C, Liatis S, le Roux CW, Kokkinos A. The role of bariatric surgery to treat diabetes: current challenges and perspectives. BMC Endocr Disord 2017;17:50. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46. GeneCards . DECR2 Gene - 2,4-dienoyl-CoA reductase 2. Accessed 16 January 2023. Available from: https://www.genecards.org/cgi-bin/carddisp.pl?gene=DECR2.
  • 47. Nieto A, Mira E, Castaño JG. Transcriptional regulation of rat liver protein disulphide-isomerase gene by insulin and in diabetes. Biochem J 1990;267:317–323 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48. Wen X, Zhang B, Wu B, et al. Signaling pathways in obesity: mechanisms and therapeutic interventions. Signal Transduct Target Ther 2022;7:298. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49. Song X, Wang S, Li L. New insights into the regulation of Axin function in canonical Wnt signaling pathway. Protein Cell 2014;5:186–193 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50. Ardestani A, Lupse B, Kido Y, Leibowitz G, Maedler K. mTORC1 signaling: a double-edged sword in diabetic β cells. Cell Metab 2018;27:314–331 [DOI] [PubMed] [Google Scholar]
  • 51. Buniello A, MacArthur JAL, Cerezo M, et al. The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019. Nucleic Acids Res 2019;47(D1):D1005–D1012 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52. Saber-Ayad M, Manzoor S, El-Serafi A, Mahmoud I, Abusnana S, Sulaiman N. Statin-induced myopathy SLCO1B1 521T>C is associated with prediabetes, high body mass index and normal lipid profile in Emirati population. Diabetes Res Clin Pract 2018;139:272–277 [DOI] [PubMed] [Google Scholar]
  • 53. Adzhubei IA, Schmidt S, Peshkin L, et al. A method and server for predicting damaging missense mutations. Nat Methods 2010;7:248–249 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54. Tirona RG, Leake BF, Merino G, Kim RB. Polymorphisms in OATP-C: identification of multiple allelic variants associated with altered transport activity among European- and African-Americans. J Biol Chem 2001;276:35669–35675 [DOI] [PubMed] [Google Scholar]
  • 55. König J, Cui Y, Nies AT, Keppler D. A novel human organic anion transporting polypeptide localized to the basolateral hepatocyte membrane. Am J Physiol Gastrointest Liver Physiol 2000;278:G156–G164 [DOI] [PubMed] [Google Scholar]
  • 56. Neuvonen M, Hirvensalo P, Tornio A, et al. Identification of glycochenodeoxycholate 3-O-glucuronide and glycodeoxycholate 3-O-glucuronide as highly sensitive and specific OATP1B1 biomarkers. Clin Pharmacol Ther 2021;109:646–657 [DOI] [PMC free article] [PubMed] [Google Scholar]

Articles from Diabetes are provided here courtesy of American Diabetes Association

RESOURCES