Impaired Insulin Clearance as the Initial Regulator of Obesity-Associated Hyperinsulinemia: Novel Insight Into the Underlying Mechanism Based on Serum Bile Acid Profiles

Hyperinsulinemia has been well demonstrated as a common phenomenon associated with obesity and could occur before the development of dysglycemia (1). Previous studies have shown that hyperinsulinemia increased the risk of multiple metabolic diseases, such as type 2 diabetes mellitus (T2DM) and cardiovascular diseases (1,2). Generally, hyperinsulinemia is considered as a β-cell compensatory response to insulin resistance in obese individuals to maintain the blood glucose level in the normal range (3). Plasma insulin levels are regulated by insulin secretion, insulin clearance, and insulin sensitivity (3–5). Meanwhile, the reduction of insulin clearance is also considered as a critical factor in obesity-associated hyperinsulinemia. Several previous studies have demonstrated that insulin clearance is closely associated with obesity (5,6), insulin resistance (7), and diabetes (8,9). Although the roles of insulin secretion and clearance are still controversial and not fully understood, recent research investigations revealed that impaired insulin clearance might be the primary factor for the development of hyperinsulinemia in obese individuals (7,10,11).

Bile acids (BAs) are pleiotropic bioactive molecules. In addition to their well-established role in the digestion and absorption of dietary fats and liposoluble vitamins, it is also known that BAs regulate glucose and lipid metabolism mainly via the nuclear farnesoid X receptor and the G-protein–coupled BA receptor (TGR5) (12,13). Dysregulated BA metabolism, characterized by changes in the BA pool size and composition, has been reported to play a crucial role in the pathogenesis of obesity, diabetes, fatty liver, and atherosclerosis (13,14).

To the best of our knowledge, although multiple serum metabolites such as amino acids, BAs, and lipids have been demonstrated as closely related to the risk of obesity and T2DM (13,15,16), the potential association between these serum metabolites and insulin clearance in obese individuals has not been well established. Alterations in circulating branched-chain amino acids seemed to have no correlation with insulin clearance (17).

Therefore, to further understand the pathogenesis of obesity-associated hyperinsulinemia and to discuss the potential role of BAs in insulin clearance, we performed an oral glucose tolerance test (OGTT), intravenous glucose tolerance test (IVGTT), and hyperinsulinemic-euglycemic clamp in two cohorts and analyzed a large number of metabolic parameters in these cohorts with independent evaluation strategies. Our study is designed to: 1) evaluate the relative contributions of insulin clearance and insulin secretion to hyperinsulinemia in obese individuals; and 2) examine serum BA aberrations in obese subjects and their relationship with insulin clearance and other metabolic parameters by using ultra-performance liquid chromatography coupled to tandem mass spectrometry (UPLC-MS/MS).

The current study consisted of two separate cohorts of Chinese adults recruited from the Department of Endocrinology of the First Affiliated Hospital of Nanjing Medical University from 2014 to 2020. In this study, we used data from the baseline examinations. A detailed flow diagram describing the inclusion and exclusion criteria for participants is presented in Supplementary Fig. 1. Briefly, 460 eligible participants were enrolled in cohort 1 to receive an OGTT (0, 30, 60, and 120 min), and cohort 2 comprised 81 individuals who underwent the IVGTT and hyperinsulinemic-euglycemic clamp. A total of 56 subjects in cohort 2 underwent a 4-point OGTT (0, 30, 60, and 120 min), and the other 25 subjects underwent a simplified OGTT (0 and 120 min). Obesity is defined as BMI ≥30 kg/m² by the World Health Organization criteria. Based on the 75-g OGTT results, subjects in both cohorts were further sorted into four groups: lean subjects with normal glucose tolerance (Lean-NGT), obese subjects with normal glucose tolerance (OB-NGT), obese subjects with impaired glucose tolerance and/or impaired fasting glucose (OB-IFG/IGT), and obese subjects with T2DM (OB-DM). This study protocol was approved by the Ethics Committee of the First Affiliated Hospital of Nanjing Medical University (2014-SR-003, 2018-SR-069). Written informed consent was obtained from all participants in both cohorts.

All participants underwent a 75-g OGTT in the morning after a 12-h overnight fast. Subjects were instructed to maintain a standard diet containing at least 250 g of carbohydrate/day for at least 3 days before the test. Blood samples were collected at 0, 30, 60, and 120 min after the glucose challenge for measurement of plasma glucose, serum insulin, and C-peptide concentrations.

Each participant in cohort 2 underwent an IVGTT on a separate day after an overnight fast. Briefly, 0.3 g/kg glucose solution was given intravenously over 3 min, and blood samples for glucose, insulin, and C-peptide measurement were obtained at 0, 2, 4, 6, 8, 10, 20, 30, and 60 min following the end of glucose infusion.

Whole-body insulin sensitivity was assessed by the hyperinsulinemic-euglycemic clamp method described by DeFronzo et al. (18). Briefly, participants were admitted to the research center at 7:30 a.m. after an overnight fast. An intravenous catheter was inserted in the antecubital vein for blood sampling, and the arm was warmed throughout the study for blood arterialization. In the opposite arm, another intravenous catheter was inserted for dextrose and insulin infusion. After a 10-min priming dose to acutely increase the insulin concentration, human insulin was continuously infused at 2 mU/kg/min for 110 min. Meanwhile, 20% dextrose was periodically adjusted to maintain the plasma glucose level at a stable value of 4.5–5.5 mmol/L. Blood samples for serum insulin and C-peptide assays were collected at 0, 30, 60, 90, and 120 min during the clamp.

Blood samples collected at baseline and stored in −80°C from cohort 2 were used to quantify serum BAs based on UPLC-MS/MS (ACQUITY UPLC-Xevo TQ-S; Waters Corp., Milford, MA) as described previously (19). A total of 40 BAs were detected as listed in Supplementary Table 1. Chromatographic separation was performed with an ACQUITY BEH C18 column VanGuard precolumn (2.1 × 5 mm) and analytical column (2.1 × 100 mm). Raw data obtained from UPLC-MS/MS were analyzed and quantified using the MassLynx software (v4.1; Waters Corp.) to perform peak integration, calibration, and quantification for each BA. The self-developed platform iMAP (v1.0; Metabo-Profile Biotechnology, Shanghai, China) was used for statistical analysis, including OPLS-DA analysis and univariate test, among others.

Plasma glucose was analyzed by a blood glucose biochemical analyzer (Biosen; EKF Diagnostics, Barleben, Germany). Serum insulin and C-peptide were measured by chemiluminescent methods. Hemoglobin A_1c (HbA_1c) was analyzed by a chromatographic technique. Routine biochemical measurements, including serum lipids and liver enzymes, were measured by enzymatic methods (Chemistry Analyzer Au5800; Olympus Medical Engineering Company, Tokyo, Japan).

The total areas under the curve (AUCs) of glucose (GluAUC), insulin (InsAUC), and C-peptide (CpAUC) from the OGTT and IVGTT were calculated using the trapezoidal method. Insulin sensitivity was measured by the Matsuda index from the OGTT in cohort 1 and by the glucose infusion rate (GIR) during the last 30 min of steady-state in the hyperinsulinemic-euglycemic clamp in cohort 2. As deterioration in early phase insulin secretion is critical to dysglycemia, the main insulin secretion indices we used were the InsAUC₃₀/GluAUC₃₀ ratio during the first 30 min of the OGTT and the first-phase insulin response (FPIR) calculated by the sum of serum insulin levels in the first 4 min (2 min and 4 min) during the IVGTT (20). Basal and IVGTT insulin secretion rate (ISR) were calculated by C-peptide deconvolution based on the two-compartment model described by Van Cauter et al. (21). To evaluate β-cell function, the disposition index (DI) was calculated as InsAUC₃₀/GluAUC₃₀ × Matsuda index in cohort 1 and FPIR × GIR in cohort 2. Endogenous insulin clearance during the OGTT was assessed by CpAUC₁₂₀/InsAUC₁₂₀. The metabolic clearance rate of exogenously administered insulin during the clamp (ICR_Clamp) was measured as the insulin infusion rate divided by steady-state plasma insulin above basal level as described by DeFronzo et al. (18). Total insulin clearance during the IVGTT was calculated as ICR_IVGTT = AUC(ISR/Ins)60 − V*[ln(I60) − ln(I0)], where V is the volume of distribution of insulin estimated as 141 mL/kg (7,8). CpAUC₆₀/InsAUC₆₀ from the 1-h IVGTT was calculated as another insulin clearance marker in cohort 2 (22).

Parametric and nonparametric variables among the Lean-NGT, OB-NGT, OB-IFG/IGT, and OB-DM groups were compared by one-way ANOVA with Bonferroni post hoc test or Kruskal-Wallis test. All nonnormally distributed variables were ln-transformed before correlation analysis. The association among insulin clearance, BAs, and other metabolic parameters was analyzed by Pearson correlation. Multivariable linear regression analysis was performed to determine the relationship between the ICR_Clamp and serum BAs. Data were analyzed using SPSS (version 25.0) and GraphPad Prism (version 8.0), with significance defined as P < 0.05 (two-sided).

The demographic and metabolic parameters of the participants in the two cohorts are shown in Table 1. There was an increasing trend of the age from the NGT and IFG/IGT groups to the T2DM groups among obese subjects, but no difference in BMI within obese groups. In cohort 1, the female participants had a higher ratio in obese groups compared with the lean group, while cohort 2 was matched by sex. As we expected, obese subjects in both cohorts had a worse lipid profile and liver function compared with the lean control group. The IFG/IGT and T2DM subjects in the obese group exhibited higher total cholesterol and LDL cholesterol (LDL-C) compared with lean and obese NGT groups, and no statistic difference was found between the lean NGT and obese NGT group in cohort 1. Total triglyceride (TG), ALT, and AST were significantly elevated, and HDL cholesterol (HDL-C) was reduced in all three obese groups compared with levels in the lean NGT group. The fasting plasma glucose (FPG), 2-h plasma glucose, fasting serum insulin (FINS), fasting serum C-peptide (FCP), and HbA_1c were progressively increased from lean to obese and from NGT to IFG/IGT and T2DM in obese groups (Table 1).

There was a progressive increase in FINS from lean to obese and from NGT to IFG/IGT to T2DM in both cohorts (Fig. 1A and B). Obese participants showed remarkable hyperinsulinemia.

The insulin sensitivity was decreased from lean to obese groups, but there was no difference between the IFG/IGT and T2DM subjects within the obese group in cohort 1 (Fig. 1A). In cohort 2, insulin sensitivity (GIR), measured by hyperinsulinemic-euglycemic clamp, was significantly lower in obese subjects compared with lean individuals (P < 0.0001), but similar between obese NGT and obese IFG/IGT group, and then obese subjects in the T2DM group had the worst insulin sensitivity (vs. OB-NGT, P = 0.002) (Fig. 1B).

The early phase of insulin secretion (InsAUC₃₀/GluAUC₃₀) in cohort 1 was significantly increased in the obese NGT individuals compared with lean control group and then decreased from the obese NGT to IFG/IGT and to T2DM subjects (all P < 0.001) (Fig. 1A). In cohort 2, FPIR was lower in the obese participants with T2DM compared with the other three subgroups (Fig. 1B). Moreover, the β-cell function (DI) was gradually decreased from lean to obesity and from NGT to IFG/IGT to T2DM in cohort 1 (all P < 0.01) (Table 1). The same declining trend of the DI was found in cohort 2; however, no difference was found among lean, obese NGT, and IFG/IGT subjects (Table 1).

The endogenous insulin clearance was significantly reduced in all three obese subgroups compared with the lean control subjects in cohort 1 (P < 0.0001). There was no further reduction in the obese IFG/IGT versus the obese NGT group. Interestingly, the obese subjects with T2DM had a significant but slight increase in insulin clearance compared with the other two obese subgroups (P < 0.0001) (Fig. 1A). Similar to cohort 1, exogenous insulin clearance during the hyperinsulinemic-euglycemic clamp (ICR_Clamp) in cohort 2 was markedly reduced in the obese participants compared with lean subjects (P < 0.01), but no difference was found within obese subgroups (Fig. 1B). Endogenous insulin clearance during IVGTT (ICR_IVGTT) in cohort 2 was significantly decreased in obese NGT subjects versus lean subjects (P < 0.01) (Supplementary Fig. 2A).

To further examine the potential relationship between insulin secretion and insulin clearance during the progression of hyperinsulinemia in obesity, we divided obese NGT and IFG/IGT patients in cohort 1 into 10 quantiles based on the Matsuda index from low to high (quantile 1 presenting the mildest level of insulin resistance and quantile 10 presenting the most severe level of insulin resistance). Compared with that in the lean control group, quantile 1 to 10 presented significantly lower insulin sensitivity (P < 0.0001), and FINS displayed a stepwise increase with the exacerbation of insulin resistance as expected (Fig. 1C). Additionally, we observed that insulin clearance (CpAUC₁₂₀/InsAUC₁₂₀) decreased in quantile 1 compared with lean control subjects (P < 0.01) and continued to decrease from quantile 1 to quantile 10; meanwhile, insulin secretion (InsAUC₃₀/GluAUC₃₀) was not increased until quantile 4 (Fig. 1C).

In cohort 2, because of the limited participants, we stratified obese NGT and IFG/IGT subjects into tertiles based on the GIR (the IR-T1 group presenting the mildest level of insulin resistance and the IR-T3 group presenting the most severe level of insulin resistance). The ICR_Clamp and ICR_IVGTT of the IR-T2 to -T3 groups manifested significant reductions compared with those of lean control subjects, whereas the FPIR did not change among groups (Fig. 1D and Supplementary Fig. 2B).

Correlation analysis was performed to make comparisons among four different insulin clearance markers measured from the OGTT, IVGTT, and hyperinsulinemic-euglycemic clamp. It revealed that the OGTT-based insulin clearance measurement (CpAUC₁₂₀/InsAUC₁₂₀) was positively associated with ICR_Clamp (r = 0.579; P < 0.0001), ICR_IVGTT (r = 0.443; P < 0.0001), and CpAUC₆₀/InsAUC₆₀ from IVGTT (r = 0.654; P < 0.0001). All of these four insulin clearance measurements significantly correlated with each other (Supplementary Table 2).

We next examined the correlation of insulin clearance with the large range of metabolic indices in both cohorts. In cohort 1, the insulin clearance was positively associated with age, the Matsuda index, and HDL-C, but negatively associated with BMI, FPG, FINS, FCP, ALT, AST, TG, InsAUC, CpAUC, and InsAUC₃₀/GluAUC₃₀ (all P < 0.05) (Supplementary Table 3). In cohort 2, the ICR_Clamp was positively associated with the GIR, HDL-C, and lipoprotein (a), while negatively associated with BMI, FINS, FCP, ALT, AST, TG, LDL-C, GluAUC, InsAUC, and CpAUC (all P < 0.05).

A total of 40 BA species were identified and quantitated from 81 human serum samples in cohort 2. Based on the Pearson correlation analysis, it revealed that different serum BA components were closely correlated to the ICR_Clamp and ICR_Clamp-related metabolic parameters, as shown in Fig. 2. We found that taurodeoxycholic acid (TDCA), glycodeoxycholic acid (GDCA), tauro-α-muricholic acid (TαMCA), taurolithocholic acid (TLCA), glycohyocholic acid, taurohyocholic acid (THCA), glycochenodeoxycholic acid (GCDCA), and taurochenodeoxycholic acid (TCDCA) were inversely correlated with the ICR_Clamp (P < 0.05–0.01), β-chenodeoxycholic acid, β-cholic acid, 3-dehydrocholic acid, and unconjugated to conjugated BAs (UnconBA/ConBA) were positively correlated with ICR_Clamp (P < 0.05–0.01) (Fig. 2). TDCA, GDCA, TαMCA, glycolithocholic acid sulfate, glycolithocholic acid, glycoursodeoxycholic acid, β-ursocholic acid, and ursocholic acid were negatively associated with the GIR, and UnconBA/ConBA were positively correlated with the GIR (Fig. 2). Notably, we observed that HDL-C was strongly associated with most BAs, and it was positively correlated with the ICR_Clamp (Fig. 2 and Supplementary Table 3).

Multivariable linear regression was performed to further examine the potential BAs influencing insulin clearance by correcting confounding factors. Without covariates in the initial model, we found eight BA species were negatively correlated with the ICR_Clamp, and the other three BA species and UnconBA/ConBA were positively correlated with the ICR_Clamp (Fig. 3A and B). After adjustment with age, sex, BMI, liver enzymes, and serum lipids, most of the correlations were retained but attenuated. Additionally, the taurocholic acid (TCA) began to present a marked correlation with the ICR_Clamp (Fig. 3B). We next added glucose metabolic parameters (FPG, FINS, FCP, GluAUC, InsAUC, CpAUC, FPIR, and GIR) into the model and only found that seven BAs, including GDCA (β = −0.335; P = 0.004), TDCA (β = −0.333; P = 0.003), GCDCA (β = −0.271; P = 0.013), TCDCA (β = −0.329; P = 0.003), THCA (β = −0.223; P = 0.042), TLCA (β = −0.357; P = 0.005), and TCA (β = −0.244; P = 0.020), as well as UnconBA/ConBA (β = 0.335; P = 0.002) showed a significant association with the ICR_Clamp (Fig. 3B and Supplementary Table 4).

To further validate the correlation between the ICR_Clamp and candidate serum BAs, 81 participants in cohort 2 were stratified into 3 groups based on tertiles of BAs shown above. We found there was a continuous decrease in the ICR_Clamp from the lower (T1) BA level group to the middle (T2) and higher (T3) BA level groups. A significant difference was also observed among the tertiles of GDCA (F = 3.834; P = 0.026) and TDCA (F = 5.022; P = 0.009) and among pairwise comparisons (GDCA, T1 vs. T3, P = 0.019; TDCA, T1 vs. T3, P = 0.006) (Fig. 3C and D). However, no significant difference was found in the ICR_Clamp among the tertiles of GCDCA, TLCA, THCA, and TCA and in pairwise comparison of TCDCA (Supplementary Fig. 3).

In the current study, we outlined the variation in the features of insulin clearance and insulin secretion in the progression of hyperinsulinemia in obese individuals from two different cohorts. Moreover, we further demonstrated that decreased insulin clearance, rather than increased insulin secretion, might be the primary cause of the compensatory hyperinsulinemia in obese individuals in response to insulin resistance. Additionally, we evaluated altered serum BA profiles of subjects in cohort 2 by UPLC-MS/MS, and we have demonstrated for the first time the association between BAs and insulin clearance.

We noticed that the obese groups exhibited apparent hyperinsulinemia compared with the lean NGT group in both cohorts, which was determined by the combined action of increased insulin secretion and reduced insulin clearance and sensitivity. Interestingly, we also found that the declining trend of insulin clearance still showed in obese IFG/IGT subjects, but disappeared in obese participants with T2DM in both cohorts. This result seemed to contradict the previous studies that showed impaired insulin clearance presented in patients with T2DM, and it could predict the risk of T2DM and metabolic syndrome (9,23,24). However, the recently published study by Gastaldelli et al. (8) supported our findings. Thus, we hypothesize that it may be due to relieving the stress from hyperinsulinemia to the impaired β-cell function. Moreover, combined with impaired insulin secretion and insulin resistance, dysglycemia could be aggravated. Therefore, the insulin clearance reduction could occur at a very early stage in obese individuals without diabetes, while β-cell failure might play a more important role in the later progression to T2DM.

Although within the obese subgroups, the different insulin clearance followed different glucose tolerance, we only found a weak correlation between insulin clearance and FPG in cohort 1, and no relationship was found in cohort 2. There was no significant correlation between insulin clearance and 2-h plasma glucose in either cohort. All of these data suggest that insulin resistance is a major factor that regulates alterations in insulin clearance and secretion in addition to hyperglycemia, which is more likely to be an outcome of systemic effects and plays a less important role in the regulation of insulin clearance.

Most investigators believed that the occurrence of hyperinsulinemia is a result of increased insulin secretion in response to insulin resistance in obese individuals. Additionally, Smith et al. (25) hold the opinion that hyperinsulinemia in obese patients is due to increased insulin secretion without the defects of hepatic or extrahepatic insulin extraction. Nevertheless, a previous study conducted the insulin suppression test and graded glucose infusion test in obese individuals without diabetes and found that both changes in insulin secretion and insulin clearance compensate for insulin resistance, and insulin clearance may provide the first adaptation (7). Bergman et al. (10) put forward the hypothesis that lower insulin clearance exposes peripheral organs and β-cells to high concentrations of insulin to exacerbate insulin resistance and then ultimately results in β-cell failure and progression to T2DM. Since there were some limitations in previous studies, such as a small sample size, lack of lean normal control subjects, and evaluation of only endogenous or exogenous insulin clearance, we further evaluated the relative contributions of insulin clearance and secretion to obesity-associated hyperinsulinemia in two complementary cohorts with three different evaluation methods (OGTT, IVGTT, and hyperinsulinemic-euglycemic clamp). Obese participants without diabetes were divided into 10 quantiles in cohort 1 and into tertiles in cohort 2 based on their insulin resistance levels. In both cohorts, with the decline of insulin sensitivity, fasting insulin levels and insulin secretion gradually increased, while insulin clearance gradually decreased. However, it caught our attention that when mild insulin resistance and hyperinsulinemia just appeared, both endogenous and exogenous insulin clearance were significantly reduced, but insulin secretion was maintained at the same level as that in lean NGT subjects. Therefore, we drew the conclusion that reduced insulin clearance, rather than elevated insulin secretion, was the initial promoter of hyperinsulinemia in response to obesity and insulin resistance.

It was noticed that a recent review pointed out some limitations of C-peptide/insulin ratios for insulin clearance measurement that we used in cohort 1 (26). Cohort 2 subjects have undergone the OGTT, IVGTT, and hyperinsulinemic-euglycemic clamp, which provided us the chance to compare across different insulin clearance measurements. We found that both OGTT-based CpAUC₁₂₀/InsAUC₁₂₀ and CpAUC₆₀/InsAUC₆₀ from IVGTT significantly correlated with ICR_Clamp and ICR_IVGTT, which indicated that CpAUC/InsAUC might be a usable method to estimate insulin clearance, and the OGTT-based CpAUC₁₂₀/InsAUC₁₂₀ seems more convenient.

To date, the underlying mechanisms of impaired insulin clearance in obese individuals are still unclear. For the first time, we examined the association between BAs and insulin clearance by performing thorough target metabolomics on serum BA (40 species) changes in cohort 2. We found that a group of conjugated BAs (TDCA, GDCA, TαMCA, TLCA, glycohyocholic acid, THCA, GCDCA, and TCDCA) was negatively correlated with the ICR_Clamp, and some unconjugated BAs (β-cholic acid, β-chenodeoxycholic acid, and 3-dehydrocholic acid) and the UnconBA/ConBA were positively correlated with the ICR_Clamp. The previous study from the large-scale China Cardiometabolic Disease and Cancer Cohort (4C) study demonstrated that total taurine- and glycine-conjugated BAs were positively associated with the risk of diabetes, and perturbations in unconjugated as well as conjugated BA coregulation had already been altered even before diabetes onset (14). Moreover, several studies reported that some unconjugated BAs might contribute to metabolic improvements (27,28). Overall, the process involving conjugation and deconjugation of BAs might deserve more attention in glucose and lipid metabolic regulation.

Recent studies have demonstrated that the BA profiles were also different in subjects with different glucose and lipid metabolic statuses (13,14). Based on this, we applied the fully adjusted model and combined it with tertile analysis of the ICR_Clamp across BAs to further investigate the association between insulin clearance and BAs. We found two candidate BAs, namely GDCA and TDCA, which were likely to take part in the process of hepatic insulin clearance regulation in obesity-associated hyperinsulinemia. Both GDCA and TDCA are conjugated secondary BAs and belong to the 12α OH BAs. GDCA was reported to be positively associated with insulin resistance in 904 individuals without diabetes in an observational study, and TDCA was also shown to correlate with insulin resistance markers in obese subjects (29,30). A previous study indicated that plasma levels of 12α OH BAs were significantly correlated to insulin resistance (31), while in our study, the correlation of 12α OH BAs with GIR did not reach statistical significance. All of the above observations provide evidence that BAs are closely associated with insulin resistance and that circulating BA patterns might change before metabolic disturbance onset. Our study further revealed that conjugated BAs, especially GDCA and TDCA, might be the critical metabolites regulating insulin clearance in response to insulin resistance at the early stage of obesity-associated hyperinsulinemia.

The mechanism of hepatic insulin clearance involves several steps, including insulin receptor–mediated uptake, internalization of insulin, and degradation by the insulin-degrading enzyme and lysosomes (32). A recent study in obese mice demonstrated that the gut microbiota impaired hepatic insulin clearance by inhibiting hepatic transmembrane glycoprotein carcinoembryonic antigen–related cell adhesion molecule 1 expression, which mediates rapid insulin and insulin receptor complex endocytosis (33,34). Moreover, GDCA and TDCA levels were found to be significantly increased in subjects with nonalcoholic fatty liver disease (35). Several cross-sectional and longitudinal studies with a large sample size revealed that the intrahepatic triglyceride content is strongly related to reduced hepatic insulin clearance (36–38). Thus, we hypothesize that in addition to directly targeting the hepatic insulin clearance-related proteins carcinoembryonic antigen–related cell adhesion molecule 1 and/or insulin-degrading enzyme, TDCA and GDCA could alternatively induce hepatic steatosis to impair hepatic insulin clearance indirectly.

In contrast, a few publications have demonstrated that some conjugated BAs, including GDCA, tauroursodeoxycholic acid, and glycoursodeoxycholic acid, have beneficial roles in polycystic ovary syndrome and T2DM (39,40). Thus, another possible explanation is that the increased GDCA and TDCA levels in obese patients are to adapt to impaired insulin clearance. For the cross-sectional design of the study, the specific causality between BAs and insulin clearance remains an interesting question and awaits further mechanistic investigation.

One of the major significances and novelties of our study is that we used two different cohorts and found similar phenotypes to validate our hypothesis. We estimated the endogenous insulin clearance in cohort 1 by OGTT-based CpAUC₁₂₀/InsAUC₁₂₀. In cohort 2, we calculated both endogenous ICR via IVGTT and exogenous ICR under steady-state conditions of the hyperinsulinemic-euglycemic clamp, which is more reliable than the conventional indices to evaluate insulin clearance (e.g., fasting C-peptide/insulin). We also validated the OGTT-based insulin clearance marker against direct measurements from the IVGTT and hyperinsulinemic-euglycemic clamp in cohort 2. Furthermore, we achieved a satisfactory coverage of BA profiles in human blood samples, and 40 BA species were quantified accurately in our study. To the best of our knowledge, this article is the first to examine the association of serum BAs and insulin clearance.

Like other clinic research, there were also some limitations in our study. First, the cross-sectional design is not sufficient for clarifying the precedence relationship between the changes in insulin clearance and secretion as well as the causality between altered serum BAs and insulin clearance. Then, the complex and time-consuming hyperinsulinemic-euglycemic clamp technique limits its application in large-scale investigations. At last, our study lacked some detailed lifestyle information such as diet and exercise, which might affect the serum composition of BAs.

In conclusion, we demonstrated that hyperinsulinemia in obese individuals is not only due to excessive insulin secretion, but also that it results from markedly decreased insulin clearance. We reveal an alternative pathogenetic pathway in which obesity-associated hyperinsulinemia originates with reduced insulin clearance, followed by increased insulin secretion to response to insulin resistance. Serum-conjugated BAs, especially GDCA and TDCA, which were negatively associated with insulin clearance, might mediate the impairment of hepatic insulin clearance. Our findings in this study may provide more insights into diagnostic and novel therapeutic strategies for obesity and obesity-related metabolic diseases targeting for insulin clearance.

Acknowledgments. The authors appreciate the help and support from all of the participants who took part in the study.

Funding. This work was supported by National Key Research and Development Program of China (2018YFA0506904), National Natural Science Foundation of China (91854122, 82170882, and 81730094), and National Key Research and Development Program of China (2019YFA0802701).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Z.F., Q.W., W.G., and H.Z. were involved in the study design. Z.F., Q.W., and J.B. did the statistical analysis. Z.F., Q.W., and H.Z. drafted the manuscript. Q.W., J.Z.L., and H.Z. edited the final version of the manuscript. Z.F., J.G., X.Z., Y.G., T.Y., and H.Z. contributed to guidance of clinical data acquisition. Z.F., Q.W., X.Z., Y.G., W.J., and Q.F. contributed to data recording and collating. Z.F., Q.W., J.G., Y.G., C.L., J.Y., X.Y., W.J., M.H., B.Y., and H.Z. took part in the human hyperinsulinemic-euglycemic clamp performance. X.L. did the calculation of ISR and related insulin clearance parameters. All authors approved the final version before submission. H.Z. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.