Intestinal bile acid sequestration improves glucose control by stimulating hepatic miR-182-5p in type 2 diabetes

Abstract

Colesevelam is a bile acid sequestrant approved to treat both hyperlipidemia and type 2 diabetes, but the mechanism for its glucose-lowering effects is not fully understood. The aim of this study was to investigate the role of hepatic microRNAs (miRNAs) as regulators of metabolic disease and to investigate the link between the cholesterol and glucose-lowering effects of colesevelam. To quantify the impact of colesevelam treatment in rodent models of diabetes, metabolic studies were performed in Zucker diabetic fatty (ZDF) rats and db/db mice. Colesevelam treatments significantly decreased plasma glucose levels and increased glycolysis in the absence of changes to insulin levels in ZDF rats and db/db mice. High-throughput sequencing and real-time PCR were used to quantify hepatic miRNA and mRNA changes, and the cholesterol-sensitive miR-96/182/183 cluster was found to be significantly increased in livers from ZDF rats treated with colesevelam compared with vehicle controls. Inhibition of miR-182 in vivo attenuated colesevelam-mediated improvements to glycemic control in db/db mice. Hepatic expression of mediator complex subunit 1 (MED1), a nuclear receptor coactivator, was significantly decreased with colesevelam treatments in db/db mice, and MED1 was experimentally validated to be a direct target of miR-96/182/183 in humans and mice. In summary, these results support that colesevelam likely improves glycemic control through hepatic miR-182–5p, a mechanism that directly links cholesterol and glucose metabolism.

NEW & NOTEWORTHY Colesevelam lowers systemic glucose levels in Zucker diabetic fatty rats and db/db mice and increases hepatic levels of the sterol response element binding protein 2-responsive microRNA cluster miR-96/182/183. Inhibition of miR-182 in vivo reverses the glucose-lowering effects of colesevelam in db/db mice. Mediator complex subunit 1 (MED1) is a novel, direct target of the miR-96/182/183 cluster in mice and humans.

INTRODUCTION

Type 2 diabetes (T2D) is a major health problem worldwide and will continue to expand due to rising obesity levels (29). The pathophysiology of T2D is characterized by decreased glycemic control and hyperglycemia, which arises from loss of pancreatic β-cell integrity, insulin resistance, and metabolic dysfunction (29). Key dysfunctions include insufficient insulin secretion from pancreatic β-cells, impaired suppression of hepatic gluconeogenesis in response to insulin, and defects in glucose uptake by skeletal muscle and adipocytes. Each of these metabolic processes has been targeted by drug therapy to improve glycemic control. Nonetheless, some drugs that are indicated to treat other diseases have been repositioned to treat T2D despite a lack of clear mechanism for their glucose-lowering properties. Colesevelam, a bile acid sequestrant (BAS), has emerged from clinical data to lower blood glucose levels; however, the mechanisms for its glycemic effects have yet to be determined (40).

BAS is a class of drugs that binds to bile acids (BA) in the intestine and prevent their reabsorption. In the liver, cholesterol 7-α-hydroxylase (CYP7A1) converts cholesterol to BAs, which are secreted into the bile to facilitate absorption of lipids and lipid-soluble vitamins in the intestine (10). Approximately 95% of BAs are reabsorbed by the intestine and transported back to the liver via enterohepatic circulation (10). Absorbed bile-acids activate farnesoid X receptor (FXR), which promotes biliary secretion and represses BA synthesis to prevent BA overload and toxicity (47). Colesevelam binds and sequesters BAs in the gut and prevents their absorption. In turn, decreased BA absorption limits the activation of FXR and promotes increased BA synthesis through activation of CYP7A1. Increased hepatic sterol flux promotes activation of the sterol response element binding protein (SREBP2) activity and transcription of its target genes, 3-hydroxy-3-methylglutaryl-coenzyme A reductase (HMGCR), squalene monooxygenase (SQLE), and low-density lipoprotein receptor (LDLR). This transcriptional program results in increase cholesterol biosynthesis, increased low-density lipoprotein cholesterol (LDL-C) uptake, and increased cholesterol conversion to BAs (24, 44). As a result of promoting LDL-C uptake for cholesterol conversion to bile, colesevelam reduces plasma LDL-C levels, which lowers the risk for cardiovascular disease. In clinical trials, colesevelam decreased LDL-C levels by 15% and total plasma cholesterol levels by 10% (11). In 2000, colesevelam was Food and Drug Administration approved as a monotherapy with diet and exercise to treat primary hyperlipidemia and is prescribed with statins to manage hypercholesterolemia and lower LDL-C levels. Subsequently in 2008, based on the results of three pivotal clinical studies showing that colesevelam improves glycemia (5, 19, 22), colesevelam was approved as adjunct therapy for T2D and is the only BAS currently approved to treat T2D. Although colesevelam is not first line therapy for either hypercholesterolemia or T2D, colesevelam is the only single ingredient drug approved to treat both diseases. Nevertheless, the molecular mechanism(s) of colesevelam-mediated glycemic control are unknown.

Work in the last decade has demonstrated that hepatic sterol metabolism is tightly regulated by microRNAs (miRNA) (17, 26). miRNAs are small, noncoding RNAs that regulate gene expression by binding to and destabilizing mRNA targets (4, 16). In the liver, miRNAs have emerged as critical regulators of both glucose and cholesterol metabolism (15, 39). For example, miRNAs have been found to regulate insulin signaling, glucose uptake, glycogen synthesis, and gluconeogenesis in the liver (30, 39, 51). Moreover, many of the nuclear receptors that respond to changes in sterol metabolism also regulate miRNA expression (14, 45, 53). Therefore, we hypothesized that investigating hepatic miRNAs may provide a link among the intestinal, liver, and systemic effects of colesevelam on sterol and glucose regulation. We found that the polycistronic miRNA cluster miR-96/182/183, composed of miR-96-5p, miR-182-5p, and miR-183-5p (26), is elevated in the liver of diabetic rats treated with colesevelam. The promoter for this cluster contains a sterol response element (SRE) and is a transcriptional target of SREBP2 (26). Moreover, these miRNAs have been reported to be suppressed in numerous tissues in rodent models and humans with T2D (56). In liver and islets, dysregulation of miR-182-5p has been proposed to contribute to T2D, in part, through altered regulation of forkhead box protein O1 (Foxo1), a key transcription factor regulating gluconeogenesis in the liver and insulin secretion in islets (23, 43, 46). Furthermore, miR-182 knockout mice have impaired glucose tolerance, a phenotype that was attributed to decreased glucose oxidation and glycolytic capacity in muscle (55). Thus far the influence of colesevalam on hepatic miRNA expression has not been explored, and as such, we designed a study to assess hepatic miRNA changes in colesevelam-treated rodents and their impact on glucose metabolism.

In this study, improvements in glucose control with colesevelam were associated with a significant increase in hepatic expression of miR-96/182/183. Most importantly, inhibition of miR-182–5p partially inhibited colesevelam’s glucose-lowering effects in db/db mice. Moreover, we found that miRNAs in the miR-96/182/183 cluster directly target mediator complex subunit 1 (Med1), a RNA polymerase II cofactor that links nuclear receptor activity to transcription (8). Therefore, colesevelam-induced activation of SREBP2 may drive miR-96/182/183 regulation of glucose metabolism through Med1 and other target genes. This pathway likely contributes to colesevalam’s effects on glucose and supports a paradigm in which miRNAs link cholesterol and glucose regulation.

MATERIALS AND METHODS

Animals.

All animal studies were submitted to, approved by, and completed under protocols by KineMed or Vanderbilt Institutional Animal Care and Use Committee. Rats and mice were maintained in a 12-h:12-h light-dark cycle with unrestricted access to food and water. Eight-week-old male Zucker diabetic rats (ZDF fa/fa; Charles River, Wilmington, MA) were maintained (ad libitum) on Purina 5001 chow diet (vehicle) (n = 24) or chow diet supplemented with 2% colesevelam (n = 24) for 4 wk. Rats and food were weighed weekly to monitor body weight and food intake, respectively. Morning blood glucose concentrations (OneTouch Ultra; Lifescan, Milapitas, CA) and HbA1c (DCA 2000 analyzer; Bayer Healthcare, Elkhart, IN) were measured weekly in fed rats. Portal blood was collected from all rats at death after a 12-h fast to quantify blood glucose, plasma insulin and total gastric inhibitory peptide (GIP) concentrations (Rat Metabolic Hormone MAG Bead Kit 2 Plex; Millipore Bioscience Division, St. Charles, MO) and total glucagon like peptide-1 (GLP-1) and active GLP-1 (ECLIA; Pacific Biomarkers, Seattle, WA). For glucose tolerance tests, a subset of rats (n = 12) were fasted for 4 h and given a bolus of [6,6-²H₂] glucose (2 g/kg body wt) by oral gavage. Blood glucose, plasma insulin, and plasma total GIP concentrations were measured at 0, 10, 20, and 60 min, and plasma active GLP-1 concentrations were quantified at 10 min postglucose injection. Whole body glycolytic disposal of the administered glucose was determined by measuring the production of deuterated water (²H₂O) from the administered [6,6-²H₂]glucose during the glucose tolerance tests. Islet cell proliferation was measured by heavy water (²H₂O) labeling as described by Chen et al. (7). Briefly, rats received ²H₂O in the drinking water for 4 wk. At the end of the heavy water-labeling period, rats were euthanized, pancreatic islet cells and bone marrow cells were isolated, and the incorporation of ²H from ²H₂O into the deoxyribose moiety of purine deoxyribonucleotides in genomic DNA was measured by gas chromatography-mass spectrometry. The fraction of newly divided islet cells was then calculated from the ratio of ²H incorporation into DNA from islets to ²H incorporation into bone marrow cells. Hepatic free cholesterol synthesis was measured by heavy water (²H₂O) labeling as described by Lee et al. (33). Briefly, rats received ²H₂O in the drinking water for 4 wk. At the end of the heavy water-labeling period, rats were euthanized, free cholesterol was extracted from the liver and the incorporation of ²H in free cholesterol was compared with the ²H incorporation of body water to determine the fraction of free cholesterol that was newly synthesized in the liver during the labeling period. Adult male BKS.Cg-Dock7^m +/+ Lepr^db/J (db/db) mice were obtained from Jackson Laboratories and maintained (ad libitum) on NIH31 chow diet alone or supplemented with 2% colesevelam (Envigo) for the 9 wk. Mice were injected (intravenously) at retro-orbital site with 10 mg/kg in vivo grade locked-nucleic acid (LNA) miRNA inhibitor against mmu-miR-182-5p (Exiqon). Glucose tolerance tests were performed at 4 and 8 wk; fasted (5 h) mice were injected intraperitoneally with glucose (1.5 g/kg body wt), and blood glucose levels were quantified at 0, 15, 30, 60, and 120 min postinjection. Insulin tolerance tests were completed at 5 wk; fasted (5 h) mice were injected (ip) with 1.5 U/kg body wt insulin (Humalin; Eli Lilly), and blood glucose levels were quantified at 0, 15, 30, 45, 60, 90, and 120 min postinjection. To quantify plasma incretin levels, fasted mice were assayed at 6 wk before and after 1.5g/kg body wt glucose injection intraperitoneally, and insulin and GIP levels were quantified by multiplex immunoassays (Luminex). GLP-1 levels were undetectable. At the termination of the study, blood was collected from retro-orbital site and tissues were harvested and flash frozen in liquid nitrogen and stored at −80°C. Blood glucose concentrations were assayed within two freeze-thaw cycles using the glucose colorimetric assay kit (Cayman Chemical).

Cell culture.

Huh7 and HEK293 cells were maintained in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin. Huh7 cells were transfected with 20 nM miRNA mimics (Dharmacon) or LNAs (Qiagen) using Dhamafect1 (Dharmacon). HEK cells were transfected with 50 nM miRNA mimics (Dharmacon) and 1 µg/ml gene reporter (luciferase) plasmids using Dhamafect Duo (Dhamacon). Mouse primary hepatocytes were isolated from chow-fed male C57B/6J mice (Jackson Laboratory) by perfusion and digestion (137 mM NaCl, 7 mM KCl, 0.7 mM Na₂HPO₄-12H₂O, 10 mM HEPES, 5.1 mM CaCl₂, and 0.4 mg/ml collagenase from Clostridium histolyticum type IV), as previously described (20). Cells were seeded in 12- or 6-well BioCoat Collagen I plates (BD) and incubated at 37°C with 5% CO₂ in William’s E media + hepatocyte supplements (Invitrogen). Five hours after plating, cells were transiently transfected with 50 nM microRNA mimics (Dharmacon), 50 nM LNAs (Qiagen), 50 nM ON-TARGET Plus Med1 siRNAs (set of 4), or scramble control siRNAs (Dharmacon) using Lipofectamine RNAiMAX Transfection Reagent (ThermoFisher) for 48 h.

Transcriptomics.

Total RNA was isolated from cells or 30 mg of tissue using the miRNeasy Mini kits (Qiagen). cDNAs were generated using the High Capacity cDNA Reverse Transcription kits (ThermoFisher) for mRNAs or TaqMan MicroRNA Reverse Transcription kits (ThermoFisher) for miRNAs. Real-time PCR was completed with Universal PCR Master Mix and TaqMan mRNA and miRNA assays (ThermoFisher). U6 and Ppia were used as internal controls for miRNA and mRNA reactions, respectively, and relative quantitative values were calculated by the ΔCt method (relative quantitative value = 2^−dCt) and reported as fold changes.

Small RNA sequencing.

Total RNA from rat liver (n = 6) was prepared for high-throughput small RNA sequencing (sRNA-seq) using TruSeq small RNA library kits (Illumina). Library preparation was performed according to manufacturer’s protocol; 1 μg of input total RNA and 11 amplification cycles were used. Before sequencing, sRNA libraries were size-selected (135–200 nts in length) by Pippin-Prep (Sage Science). Individual libraries were purified and concentrated using the DNA Clean and Concentrator 5 kit (Zymo), tested for quality (High-Sensitivity DNA chips, 2100 Bioanalyzer; Agilent), and quantified (High-Sensitivity DNA assays, Qubit; Life Technologies). Equal molar concentrations of libraries were pooled for multiplexing and the pool was concentrated using the DNA Clean and Concentrator 5 kit (Zymo). Single-end sequencing (SR50) of multiplexed libraries was performed on the Illumina HiSeq2500 platform at the Vanderbilt Technologies for Advanced Genomics (VANTAGE) DNA Sequencing Core. sRNA-seq data were analyzed using the TIGER sRNA-seq data analysis pipeline (2). All reads ≥16 nts in length were aligned to the rat genome (Rnor5.0) by Bowtie1 (v1.1.2) (31) allowing one mismatch. DEseq2 (v1.18.1) (35) was used to identify significant (P < 0.05) differentially (absolute fold change >1.5) expressed liver miRNAs. Raw and processed data from this sequencing experiment have been deposited to Gene Expression Omnibus: GSE116826.

Proteomics.

Cell lysates were collected using protein extraction buffer (50 mM HEPES, 150 mM NaCl, 0.2 mM EDTA, 2 mM MgCl2, 1% Triton X-100, and 1× protease inhibitors). Protein extracts from livers were collected in RIPA buffer (150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS 50 mM Tris·HCl, 1× protease inhibitors, pH 8.0). Protein lysates were incubated for 1 h at 4°C, and cell debris was pelleted through centrifugation; 40–100 µg of protein lysates were separated on 3–8% Tris-acetate gels (ThermoFisher). Rabbit anti-mouse antibodies against MED1 (lot 2G073735; Abnova PAB8661; polyclonal, 1:1,000 dilution in Odyssey blocking buffer) and β-actin (lot 14; Cell Signaling 13E5; monoclonal, 1:5,000 dilution in Odyssey blocking buffer) were used for Western blotting and detected using the LICOR system. The Med1 antibody was validated in cell culture using a Med1 siRNA.

Gene reporter (Luciferase) assays.

Putative miR-96/182/183 target sites (30- to 40-bp fragments) within the human MED1 3′-untranslated region (3′-UTR) were cloned down-stream of Firefly luciferase (pEZX-MT01 vector, GeneCopoeia) and confirmed by Sanger sequencing (GeneWiz). Luciferase assays were performed in HEK293 cells dual transfected with gene reporter (luciferase) plasmids and miRNA mimics (miR-96/182/183) using the Luc-Pair Duo-Luciferase Assay Kit 2.0 (GeneCopoeia). Firefly luciferase activity was normalized to Renilla (transfection control) luciferase activity and expressed as fold changes.

Statistics.

Data are shown as means ± SE. Differences between two groups were analyzed by Student’s t-test (n < 4) or Mann-Whitney nonparametric test (n ≥ 4) correction, and differences between more than two groups were analyzed by one-way ANOVA with Dunn’s posttest. Area under the curve for glucose tolerance tests were calculated using GraphPad Prism.

RESULTS

Colesevelam reduces plasma glucose levels and increases glycolysis in ZDF rats.

To assess the impact of colesevelam on glucose metabolism, ZDF rats were treated with colesevelam or vehicle (control) for 4 wk. Colesevelam-treated rats maintained glycemic control over the course of study, whereas vehicle-treated ZDF rats had significantly elevated morning blood glucose levels beginning at 1 wk (Fig. 1A). Fasting blood glucose concentrations were also decreased in colesevelam, compared with vehicle-treated ZDF rats after a 12-h fast (Fig. 1B). Moreover, for vehicle-treated ZDF rats, hemoglobin A1c (A1C) levels increased over 4 wk, whereas colesevelam-treated rats had decreased A1C levels for the duration of the study (Fig. 1C). There were no differences in food intake (Fig. 1D) or body weight (body wt) gain (Fig. 1E) between the two groups. To further investigate colesevelam’s impact on glycemic control in ZDF rats, glucose tolerance tests were performed, and colesevelam-treated ZDF rats were found to have improved glucose tolerance (Fig. 1, F and G) and increased glycolytic disposal of glucose compared with vehicle-treated rats (Fig. 1H) during the tests. The changes in glucose tolerance and glycolysis were despite no changes in fasting or postglucose plasma insulin levels (Fig. 1, I and J).

Fig. 1.

Colesevelam (Col) treatment improves glucose tolerance in Zucker diabetic fatty (ZDF) rats. A: morning glucose levels in ZDF rats treated with vehicle or 2% colesevelam-supplemented diet; n = 24. B: fasting glucose in portal blood after 12-h fast; n = 22–23. C: hemoglobin A1c levels; n = 24. D: food intake of ZDF rats treated with vehicle or 2% colesevelam supplemented diet; n = 12. E: body weights; n = 24. F: plasma glucose levels after intraperitoneal glucose challenge; n = 12. G: area under the curve (AUC) of glucose tolerance tests; n = 12. H: total glucose disposal via glycolytic pathway following glucose disposal tests; n = 12. I: insulin levels during glucose tolerance test; n = 12. J: 12-h fasting insulin levels; n = 24. For comparisons between 2 groups, Mann-Whitney nonparametric tests were used, and for repeated measures across time, two-way ANOVAs with Bonferonni’s posttests were used.

Previous studies have suggested that colesevelam’s glucose-lowering effects are linked to intestinal production of incretins, GLP-1 and GIP (41, 42). To quantify the impact of colesevelam on incretin levels in ZDF rats, GLP-1 and GIP were quantified in the fasting state and during a glucose tolerance test (Fig. 2, A–C) We failed to detect any significant differences in levels of either fasting GLP-1 or GIP or postglucose active GLP-1 concentrations with colesevelam treatments (Fig. 2, A and B). Total GIP levels were, however, increased 20 min after an oral glucose bolus in colesevelam-treated rats compared with vehicle-treated rats (Fig. 2C). Pharmacological agents that increase circulating incretins, i.e., dipeptidyl peptidase-4 inhibitors and GLP-1 receptor agonists, have been shown to stimulate β-cell proliferation (48). However, no significant difference in islet cell proliferation was found after 4 wk of colesevelam or vehicle treatment in ZDF rats (Fig. 2D). Similarly, homeostatic model assessment (HOMA)-IR (primarily reflecting hepatic insulin sensitivity) and β-cell function (HOMA-β) were not significantly different between the two groups (Fig. 2, E and F). In addition, ZDF rats treated with colesevelam were found to have significantly increased hepatic de novo cholesterol biosynthesis compared with vehicle-treated rats, as well as a slight, albeit not significant (P = 0.06), increase in hepatic de novo lipogenesis (Fig. 2, G and H). Together these data suggest that colesevelam treatments significantly alter hepatic lipid metabolism and improve glucose control by increasing whole body glycolysis independent of any changes in β-cell functionality and only modest changes in incretin concentrations. Therefore, it is likely that other mechanisms contribute to these early metabolic effects of this drug.

Fig. 2.

Colesevelam (Col) alters hepatic lipid metabolism but has modest effects on incretins. A: levels of glucagon like peptide-1 (GLP-1) in portal blood from overnight-fasted animals; n = 24. B: levels of gastric inhibitory peptide (GIP) in portal blood from overnight-fasted animals; n = 24. C: plasma GIP levels after intraperitena glucose challenge; n = 12. D: percentage of new islets by 5-bromo-2'-deoxyuridine labeling; n = 12. E: homeostatic model assessment (HOMA) method for assessing insulin resistance from basal (fasting) glucose and insulin concentrations; n = 22–24 F: homeostatic model assessment (HOMA) method for assessing insulin β-cell function from basal (fasting) glucose and C-peptide concentrations; n = 22–23. G: percent de novo cholesterol synthesis over 4 wk; n = 6. H: de novo lipogenesis, percent new triglyceride-palmitate after 4 wk of labeling; n = 6. For comparisons between 2 groups, Mann-Whitney nonparametric tests were used, and for repeated measures across time, two-way ANOVAs with Bonferonni’s posttests were used.

Colesevelam induces hepatic expression of the miR-96/182/183 cluster.

Due to the hepatic effects of colesevelam in vivo, hepatic miRNA changes were investigated. To quantify the impact of colesevelam on hepatic miRNAs, sRNA-seq was performed on livers from ZDF rats (n = 6) after colesevelam or vehicle (control) treatments for 4 wk. Strikingly, colesevelam treatments significantly altered 29 hepatic miRNAs (4 up, 25 down) compared with vehicle-treated rats (Fig. 3A). Two of the most upregulated hepatic miRNAs were rno-miR-182 and rno-miR-183-5p, which are part of a poly-cistronic miRNA cluster on chromosome 4 with rno-miR-96-5p in rats (12). rno-miR-182 is expressed at levels 10 times higher than rno-miR-183-5p, and rno-miR-96-5p was not detectable by sequencing in livers of ZDF rats (Fig. 3A). Real-time PCR was used to validate sequencing results, and rno-miR-182 and rno-miR-183-5p were confirmed to be significantly increased in livers from colesevelam-treated compared with vehicle-treated ZDF rats (Fig. 3, B and C). Hepatic rno-miR-96-5p levels were also found to be significantly increased with colesevelam treatments, as quantified by real-time PCR (Fig. 3D). Similar to sequencing, results from real-time PCR assays suggest that rno-miR-182-5p was the most abundantly expressed miRNA from the cluster (Fig. 3, B–D). The miR-96/182/183 cluster has previously been reported to harbor an SRE in its promoter and be directly activated by the nuclear receptor SREBP2 (26). Most interestingly, colesevelam’s lipid-lowering effects in humans are attributable to changes in sterol metabolism that increase activity of SREBP2. To link changes in miRNA expression with SREBP2 activity, we measured the expression of Srebf2 and downstream SREBP2 target genes, Sqle, Ldlr, and Hmgcr, and we found that Sqle and Hmgcr mRNA levels were significantly increased in the livers of colesevelam-treated ZDF rats (Fig. 3E). These results suggest that colesevelam stimulates expression of hepatic miR-96/182/183 in ZDF rats, likely through SREBP2.

Fig. 3.

The microRNA (miR)-96/182/183 cluster is upregulated in the livers of colesevelam-treated Zucker diabetic fatty (ZDF) rats. A: volcano plot depicting significantly altered miRNAs between colesevelam (Col) and vehicle-treated ZDF rat livers (left) and expression (reads per million total reads) of the miR-96/182/183 cluster (left); n = 6. B–D: liver expression of the miR-96/182/183 cluster by real-time PCR; n = 9–12. E: liver gene (mRNA) expression changes of Srebf2 and its target genes Hmgcr, Ldlr, and Sqle; n = 8–12. For sequencing data, unpaired t-tests were used. For comparisons between 2 groups, Mann-Whitney nonparametric tests were used; n.d., not detectable.

Inhibition of miR-182 blocks colesevelam’s glucose-lowering effects.

miR-182 is the most abundant hepatic miRNA in the cluster. Therefore, to next determine if miR-182-5p contributes to the glucose-lowering effects of colesevelam, leptin receptor-deficient (db/db) mice were cotreated with colesevelam and miR-182-5p inhibitors. Adult male db/db mice were fed control diet (Chow) or diet supplemented with 2% colesevelam (Col) for 9 wk. Mice were intravenously injected biweekly with 10 mg/kg body wt LNA miRNA inhibitors against mmu-miR-182-5p (LNA-182) or PBS (control) (Fig. 4A). Three groups were studied, db/db mice treated with vehicle (Chow + PBS), colesevelam (Col + PBS), or colesevelam + LNA-182 (Col + LNA). Similar to observations in ZDF rats, hepatic levels of all three miRNAs were significantly upregulated in the livers of Col + PBS db/db mice (Fig. 4B). Real-time PCR was also used to confirm that mmu-miR-182-5p levels were significantly decreased in mice that received Col + LNA treatment, compared with Col + PBS (Fig. 4B). Surprisingly, LNAs against mmu-miR-182-5p also decreased mmu-miR-183-5p levels in db/db mouse livers; however, mmu-miR-96-5p levels were not significantly affected (Fig. 4B). Furthermore, the expected changes in the SREBP2 target genes Hmgcr, Ldlr, and Sqle were observed, and LNA-182 treatment did not interfere with these effects (Fig. 4C). Although SREBP2 likely promotes its own transcription, the Srebf2 mRNA levels were not altered in livers from db/db mice treated with colesevelam (Col +PBS) or LNA-182 (Col + LNA-182) as compared with control-treated mice (Chow + PBS) (Fig. 4C). To determine if miRNAs from the miR-96/182/183 cluster were also upregulated in other metabolic tissues in response to colesevelam treatments, miR-96-5p, miR-182-5p, and miR-183-5p were measured in white adipose tissue (WAT) from treated db/db mice. Colesevelam treatments failed to increase the expression of miRNAs in the miR-182/183/96 cluster in WAT; however, miR-182-5p levels were decreased by 50% in Col + LNA compared with Col + PBS db/db mice (Fig. 4D). These results suggest that the induction of the miR-96/182/183 cluster is specific to liver, likely due to increased hepatic SREBP2 activity.

Fig. 4.

Colesevelam (Col) stimulates the microRNA miR-96/182/183 cluster in livers of db/db mice and is inhibited with locked-nucleic acid (LNA)-182 treatment. A: schematic of animal study. Twelve-week-old db/db mice treated with vehicle or 2% colesevelam-supplemented diet for 9 wk. GTT, glucose tolerance test; ITT, insulin tolerance test. B: liver expression of miR-96/182/183 cluster miRNAs; n = 8–9. C: liver gene (mRNA) expression changes for SREBP2 target genes Hmgcr, Ldlr, Sqle, and Srebf2; n = 8–9. Actb, β-actin. D: white adipose tissue (WAT) miRNA expression of miR-182/183/96 cluster; n = 8–9. For comparisons among 3 groups, Kruskal-Wallis one-way ANOVA with Dunn’s posttest was used; α = 0.05.

To determine if colesevelam’s glucose-lowering properties are mediated by miR-182-5p in mice, we performed metabolic studies in these db/db mice. (Fig. 4A). Before injections, there was no difference in plasma glucose or body weights between groups (Fig. 5, A and B). In Chow + PBS db/db mice, glucose levels steadily increased over the 9-wk study, and these mice lost a significant amount of weight beginning at 6 wk (Fig. 5, A and B). Conversely, Col + PBS mice displayed variable glucose levels over 9 wk; as glucose levels were decreased at 1, 2, 6, and 7 wk of treatments but not at 3, 4, 5, and 8 wk (Fig. 5A). Most importantly, LNA-182 treatments blocked colesvelam’s glucose-lowering effects early in the study at 1 and 2 wk but failed to alter colesevelam’s effects on fasting glucose levels later in the study (Fig. 5A). The high variability in fasting glucose may be due to repeated testing on these severely diabetic mice, including glucose tolerance test (weeks 4 and 8), insulin tolerance test (week 5), and glucose bolus for incretin measurements (week 6). Nonetheless, colesevelam treatments prevented the observed weight loss (likely a result of diabetic wasting) in the Col + PBS db/db mice, and LNA-182 treatments did not affect this parameter (Fig. 5B). To determine if colesevelam improves glucose metabolism in db/db mice, glucose tolerance tests were performed at the 4- and 8-wk time points within the study. At 4 wk, Col + PBS mice, but not Col + LNA mice, showed improved glucose tolerance compared with Chow + PBS mice (Fig. 5, C and D). LNA treatments were found to partially reverse the beneficial effects of colesevelam (P = 0.0778), as we did not find a significant difference between Chow + PBS and Col + LNA-treated mice (Fig. 5D). These results suggest that miR-182-5p is at least partially required for improved glucose tolerance associated with colesevelam treatments in mice (Fig. 5, C and D).

Fig. 5.

Inhibition of microRNA (miR)-182 in vivo abrogates the early improvements in glucose tolerance conferred by colesevelam in db/db mice. A: 4-h fasted glucose levels in Chow + PBS-, Col + PBS-, and Col + locked-nucleic acid (LNA)-treated db/db mice; n = 7–9. B: body weight of db/db mice; n = 8–9. C: plasma glucose levels after intraperitoneal glucose challenge at wk 4 of study; n = 9. D: area under the curve (AUC) of glucose tolerance test at wk 4; n = 9. E: plasma glucose levels after intraperitoneal glucose challenge at wk 4 of study; n = 9. F: area under the curve of glucose tolerance test at wk 4; n = 9. G: plasma insulin levels after 4-h fast and 15 min postglucose bolus at wk 5 of study; n = 9. H: plasma gastric inhibitory peptide (GIP) levels after 4-h fast and 15 min postglucose bolus at wk 5 of study. n = 6–7. For comparisons among 3 groups, Kruskal-Wallis one-way ANOVA with Dunn’s posttest was used (α = 0.05); or two-way ANOVA with Bonferonni’s posttest for repeated measures across time was used. *P < 0.05, Chow + PBS vs. Col + PBS; #P < 0.05, Chow + PBS vs. Col + LNA; $P < 0.05, Col + PBS vs. Col + LNA.

At 8 wk, glucose tolerance levels were significantly improved in Col + PBS-treated mice compared with Chow + PBS-treated mice; however, glucose tolerance levels in Col + LNA-treated mice were not significantly different compared with either Chow + PBS or Col + PBS (Fig. 5, E and F). To determine if colesevelam and/or LNA-182 treatments altered fasting or postglucose incretin or insulin levels in db/db mice, insulin, GLP-1, and GIP levels were measured in fasted (4 h) mice before and after (15 min) glucose injections at 6 wk of treatment (Fig. 5, G and H). Col + PBS db/db mice, but not Col + LNA mice, were found to have elevated fasting and postglucose insulin levels, compared with Chow + PBS mice (Fig. 5G). GLP-1 levels were undetectable in either fasting or postglucose samples, and no significant differences in GIP were detected between Col + PBS and Chow + PBS mice in fasted and postglucose states (Fig. 5H). Together these results suggest that miR-182-5p partially mediates the effects of colesevelam on glucose metabolism.

The miR-96/182/183 cluster directly regulates Med1.

miRNAs have emerged as critical regulators of metabolic gene expression and represent a new class of drug targets to treat metabolic dysfunction (36). To demonstrate that colesevelam-induced miR-182-5p likely mediates hepatic gene expression changes, we sought to experimentally validate miR-182-5p regulation of a novel predicted target gene in vitro that is suppressed by colesevelam in vivo. Therefore, in silico miRNA prediction studies were used with real-time PCR to identify potential gene targets that are expressed in the liver. Using this approach, we identified Med1 as a potential target of miR-182-5p, as well as miR-183-5p, and miR-96-5p, the two other members of the miRNA cluster. In humans, the MED1 3′-UTR harbors multiple putative binding sites for each of the miRNAs in the miR-96/182/183 cluster; two for hsa-miR-182-5p, one for hsa-miR-183-5p, and two for hsa-miR-96-5p (TargetScanSv7.2). In humans and rodents, miR-182-5p and miR-96-5p share the same seed sequence (bases 2-7 on the 5′-end of the mature miRNA), a critical region used to recognize mRNA targets, and thus are predicted to regulate the same genes at the same putative target sites. We validated the Med1 antibody using siRNA knockdown of Med1 in primary hepatocytes (Fig. 6, A and B). To test whether the miR-96/182/183 cluster regulates hepatic Med1 in mice, real-time PCR was used to quantify gene expression in the livers of db/db mice treated with Chow + PBS, Col + PBS, and Col + LNA. In Col + PBS-treated mice, miR-96/182/183 expression levels were increased (Fig. 4B), and hepatic Med1 mRNA levels were found to be significantly decreased compared with Chow + PBS mice (Fig. 6C). A modest, but not statistically significant restoration, was observed in Col + LNA-treated mice (Fig. 6C). Furthermore, Med1 protein levels were significantly reduced in Col + PBS mice compared with Chow + PBS-treated mice, as determined by Western blotting and densitometry (Fig. 6, D and E). Most interestingly, LNA-182 treatments were found to partially reverse the observed colesevelam (Col)-mediated decrease in Med1 protein levels, as we found no significant differences between Col + LNA-182-treated mice and Chow + PBS-treated mice; however, we also failed to find a significant difference between Col + LNA-182- and Col + PBS-treated mice (Fig. 6, D and E).

Fig. 6.

Mediator complex subunit 1 (Med1) is regulated by microRNA (miR)-96/182/183 in mice. A: expression of Med1 by real-time PCR in mouse primary hepatocytes transfected with 50 nM of Med1 or scramble siRNA; n = 3 B: Western blot of Med1 and β-actin (loading control) from protein samples of primary hepatocytes transfected with Med1 or scramble siRNA. Ratio of Med1/β-actin densitometry is indicated between the blots and graphed on the right; n = 3. A.U., arbitrary units. C: hepatic Med1 mRNA levels in db/db mice treated with vehicle or 2% colesevelam (Col)-supplemented diet and PBS or locked-nucleic acid (LNA)-182-5p; real-time PCR; n = 8–9. D: hepatic Med1 and β-actin (loading control) protein levels from 3 representative liver protein samples for Chow + PBS, Col + PBS, and Col + LNA-182. Western blotting. Ratio of Med1/β-actin densitometry is indicated between the blots. E: densitometry quantification of Med1 protein levels normalized to β-actin; n = 8–9. F: expression of miR-96-5p, miR-182-5p, and miR-183-5p in mouse primary hepatocytes transfected with 50 nM miR-96-5p, miR-182-5p, miR-183-5p mimics individually or in combination; real-time PCR; n = 3. G: Med1 mRNA levels in mouse primary hepatocytes transfected with 50 nM miR-96-5p, miR-182-5p, and miR-183-5p mimics individually or in combination; real-time PCR; n = 3. H: protein levels of Med1 and β-actin (loading control) from protein samples of primary hepatocytes transfected with mock or 50 nM miR-96-5p, miR-182-5p, and miR-183-5p mimics individually or in combination. Ratio of Med1/β-actin densitometry is indicated between the blots and graphed on the right; n = 3. For comparisons between 2 groups, Students’ t-tests were used, and for more than 2 groups, Kruskal-Wallis one-way ANOVA with Dunn’s posttest was used (α = 0.05) or one-way ANOVA with Bonferonni’s posttest for comparison to mock group only was used (α = 0.05).

To experimentally test whether miR-96/182/183 directly target Med1 in mice, in vitro studies were performed in mouse primary hepatocytes. Primary hepatocytes were isolated from male c57BL/6J wild-type mice and transiently transfected with 50 nM miRNA mimics of mmu-miR-182-5p, mmu-miR-183-5p, and mmu-miR-96-5p individually or in combination for 48h (Fig. 6F). We validated successful transfection of miRNA mimics and observed significantly decreased Med1 mRNA levels (Fig. 6, F and G). Furthermore, overexpression of all three miRNAs resulted in a significant reduction in Med1 protein levels in mouse primary hepatocytes (Figs. 6H). These results support that miRNAs in the miR-96/182/183 cluster regulate Med1 expression in mouse liver.

To determine if miR-96/182/183 directly regulates MED1 in humans, putative MED1 3′-UTR targets sites were cloned downstream of luciferase in gene reporter plasmids. In humans, MED1 3′-UTR is predicted to harbor three distinct regions that contain predicted target sites for the miR-96/182/183 cluster miRNAs (Fig. 7A). Gene reporter (luciferase) assays were performed in HEK293 cells dually transfected with the luciferase reporters and 50 nM miRNA mimics. Site 1 contains the putative binding sites for hsa-miR-182-5p and hsa-miR-96-5p (Fig. 7B), and we found that only overexpression of miR-96-3p, but not miR-182-5p, suppressed luciferase activity (Fig. 7C). Site 2 is a putative binding site only for hsa-miR-183-5p, but we found that overexpression of all three miRNAs suppressed normalized luciferase activity (Fig. 7, B and C). Similar to site 1, site 3 contains predicted binding sites for hsa-miR-182-5p and hsa-miR-96-5p, but not hsa-miR-183-5p; however, unlike site 1 where only miR-96-3p suppressed luciferase activity, both hsa-miR-182-5p and hsa-miR-96-5p suppressed normalized luciferase activity in gene reporter assays (Fig. 7, B and C). To demonstrate that the miRNAs suppress reporter luciferase activity through these putative target sites, two bases of the predicted seed sites were mutated for each of the loci (sites 1–3) (Fig. 7B). Most importantly, miRNA overexpression failed to suppress luciferase activity when the predicted target sites were mutated, indicating that this regulation is likely direct (Fig. 7C). Together these data suggest that all three miRNAs in the miR-96/182/183 cluster directly target and suppress MED1 at multiple target sites within the 3′-UTR.

Fig. 7.

Mediator complex subunit 1 (MED1) is a direct target of microRNA (miR)-182-5p, miR-183-5p, and miR-96-5p in humans. A: schematic of human MED1 mRNA and the 3 putative miR-182/183/96 target sites in the 3′-untranslated region (3′-UTR). B: predicted MED1 3′-UTR target sites. Bases highlighted in dark gray represent mutations for gene reporter (luciferase) assays. C: normalized luciferase activity in HEK293 cells after dual transfection of miR-96/182/183 mimics and gene (luciferase) reporters harboring putative target sites for miR-96-5p, miR-182-5p, and/or miR-183-5p; n = 4. D: miR-96-5p, miR-182-5p, and miR-183-5p levels in Huh7 cells transfected with 50 nM miR-96-5p, miR-182-5p, and miR-183-5p mimics individually or in combination; real-time PCR; n = 6. E: MED1 mRNA levels in Huh7 cells transfected with 50 nM miR-96-5p, miR-182-5p, and miR-183-5p mimics individually or in combination; real-time PCR; n = 6. F: miR-96-5p, miR-182-5p, and miR-183-5p levels in Huh7 cells transfected with mock or 50 nM locked-nucleic acids (LNAs) against miR-182-5p and/or miR-183-5p; real-time PCR; n = 6. G: MED1 mRNA levels in Huh7 cells transfected mock or 50 nM LNAs against miR-182-5p and/or miR-183-5p; real-time PCR; n = 6. H: protein levels of Med1 and β-actin (loading control) from protein lysates of Huh7 cells transfected with mock or 50 nM miR-96, miR-182, and miR-183 LNAs. Ratio of Med1/β-actin densitometry is indicated between the blots and graphed on the right; n = 6. A.U. arbitrary units. For comparisons between more than 2 groups to mock only, one-way ANOVA with Bonferonni’s posttest was used, α = 0.05. For comparisons between 2 groups, Mann-Whitney nonparametric tests were used.

To further demonstrate that the miR-96/182/183 cluster regulates MED1 in human hepatocytes, miRNA mimics were transiently transfected into Huh7 hepatoma cells (Fig. 7D). In Huh7 cells, overexpression of both hsa-miR-182-5p and hsa-miR-183-5p was required to significantly decrease MED1 mRNA levels (Fig. 7E). Conversely, LNAs against hsa-miR-182-5p and hsa-miR-183-5p were transiently transfected in Huh7 cells, individually and in combination. Markedly, inhibition of hsa-miR-182-5p (LNA-182) alone or in combination with LNA-183 significantly increased MED1 mRNA levels (Figs. 7, F and G). Most importantly, LNA inhibition of miR-96, miR-182, and miR-183 (in combination) resulted in a significant increase in MED1 protein levels in Huh7 cells (Fig. 7H). Together these data support that MED1 is a novel target of the miR-96/182/183 cluster in both mice and humans.

DISCUSSION

Colesevelam hydrochloride is a BAS approved for the treatment of T2D (18, 38, 41, 42). Nevertheless, the molecular mechanisms by which colesevelam improves glucose metabolism remain unclear. Both human studies and studies in diabetic rodent models suggest that colesevelam likely mediates its metabolic effects, in part, through the liver (6, 32, 38, 41, 52). Thus the goal of this study was to investigate whether a hepatic mechanism contributes to colesevelam’s glucose-lowering capacity. Strikingly, we discovered a mechanism that bridges both cholesterol and glucose metabolic pathways in the liver. We found that colesevelam induces the expression of the miR-96/182/183 cluster of miRNAs in the liver, likely through increased SREBP2 activity, known to be associated with colesevelam. We found that colesevelam-induced changes to glucose metabolism in db/db mice were partially blocked by LNA inhibition of miR-182 in vivo. Next, we demonstrated that miR-96/182/183 directly targets and suppresses MED1 in humans and mice, a gene previously linked to hepatic lipid metabolism (3). Results from this study support that colesevelam improves glycemic control through cholesterol-linked posttranscriptional gene regulation in the liver.

Although a variety of hypotheses have been proposed (6, 18, 40), the specific mechanism(s) by which colesevelam improves glycemic control remains unknown. Colesevelam is not absorbed in the intestine; therefore, it has been postulated that colesevelam’s effects are mediated through the intestine, and previous studies have implicated gut incretins. We and others have previously reported that colesevelam treatment increases plasma levels of gut-derived incretins that promote insulin secretion, e.g., GLP-1 and GIP in humans (6) and animal models (42). However, the observed increase in circulating incretins are modest, especially compared with dipeptidyl peptidase-4 inhibitors, and while they may contribute to the improved glycemic control with colesevelam treatment, signals beyond incretins are likely involved in the metabolic benefits of colesevelam (6). In human studies, we previously suggested an effect on gluconeogenesis and glycogenolysis (6), and these effects were confirmed in rodents (41). Similarly, Watanabe et al. (50) and Yamakawa et al. (52) recently reported that colestimide, another BAS, improved glycemic control and suggested this effect is driven by hepatic changes. In this study, hepatic SREBP2 (Srebf2) mRNA levels were found to be increased with colestimide treatments by PCR; however, in our study, we failed to find a significant increase in Srebf2 mRNA levels in mice treated with colesevelam, despite evidence that other SREBP2 transcriptional target genes (e.g., Hmgcr) were significantly increased with BAS (colesevelam) treatments (50, 52). Here, we present evidence that colesevelam lowered plasma glucose levels and increased glycolysis in ZDF rats. This could occur through either improvements in hepatic or peripheral insulin sensitivity or improvements in β-cell function. However, it is unlikely that colesevelam affects peripheral insulin sensitivity, as previous studies showed no difference in clamp studies in humans. Furthermore, we observed only modest effects on gut incretin levels, no difference in islet proliferation rate, HOMA-IR, or HOMA-β, suggesting that, at least at 4-wk, β-cell function and insulin sensitivity are not substantially improved. However, longer treatments periods may be required to observe changes in these parameters. Therefore, taken together, these studies suggest that colesevelam improves glucose metabolism through incretin-dependent and incretin-independent mechanisms. As discussed above, the effects on incretins have been inconsistent in the literature and other factors likely also contribute to the glucose-lowering effects of colesevelam. Therefore, 4 wk of colesevelam treatments in ZDF rats provide a window to study new incretin-independent mechanisms, particularly in the liver, that contribute to colesevelam’s metabolic effects.

High-throughput sRNA-seq and real-time PCR studies identified a hepatic miRNA cluster that contributes to the glucose-lowering effects of colesevelam. We found that miR-182, miR-183-5p, and miR-96-5p, which form the poly-cistronic miRNA cluster miR-96/182/183, are upregulated in the livers of colesevelam-treated diabetic rodents, ZDF rats, and db/db mice. Furthermore, inhibition of miR-182-5p in vivo was found to partially reverse the beneficial effects of colesevelam on glucose control. miR-182-5p is expressed at levels 10- to 100-fold greater than miR-183-5p and miR-96-5p in the liver, and therefore, it is likely that the physiological and hepatic effects observed in this study are driven by changes in miR-182-5p and its accompanying effect’s on gene expression. However, due to the observed decrease in miR-183-5p in LNA-182-treated mice and the sequence similarity between the miRNAs, we cannot rule that miR-183-5p may contribute to some of the effects on glucose metabolism observed in vivo. Nevertheless, whether the effect is specific to miR-182-5p or whether miR-183-5p contributes to the effects observed in vivo, our findings suggest that the glucose-lowering properties of colesevelam are likely mediated through its effects on cholesterol metabolism as the miR-96/182/183 miRNA cluster is directly regulated in the liver by the master transcriptional regulator of cholesterol metabolism, SREBP2 (26). Colesevelam has previously been demonstrated to indirectly promote SREBP2 activity in the liver (50). Our results suggest that SREBP2-mediated expression of the miRNA cluster in response to colesevelam improves glycemic control through suppression of target gene(s) that promote metabolic dysfunction (55, 56). Paradoxically, statins, which classically induce SREBP2 activity and have been shown to induce miR-96/182/183 expression, have been associated with increased risk of T2D (1, 26). Jeon et. al. (26) first demonstrated that SREBP2 transcriptionally regulates the miR-96/182/183 cluster in the liver in response to statins. They proposed that these miRNAs coordinate hepatic lipid metabolism by directly targeting negative regulators of SREBP2, insulin-induced gene 2 (Insig2) and F-box/WD repeat-containing protein (Fbxw7), as a feed-forward network (26). However, we failed to find changes in Insig2 or Fbxw7 mRNA expression in the livers of colesevelam-treated mice (data not shown). These discrepancies are likely due to differences in cellular metabolic conditions that may affect miRNA-mRNA interactions. Moreover, statins inhibit, while colesevelam promotes, hepatic cholesterol synthesis, which may contribute to the differential effects on glycemia. Colesevelam and statins also differ in their effects on BAs (13), signaling molecules that have been implicated in glucose control and have also been shown to regulate miRNAs-gene networks (14). Colesevelam may also mediate the effects on the miR-96/182/183 miRNA cluster through changes in the FXR transcriptional network, as FXR signaling is reduced with bile acid sequestration. The expression of the bile acid enzyme CYP7A1 is increased with reduced FXR activation, and recently, Cyp7a1-overexpressing mice were shown to have increased hepatic SREBP2 (mRNA) expression compared with control mice (34). Moreover, miR-33, an intronic miRNA harbored by and cotranscribed with SREBF2, was also found to be increased with Cyp7a1 overexpression (34), suggesting that Cyp7a1 may also contribute to the observed induction of SREBP2 activity and the increase in miR-96/182/183 cluster in our study. As such, further work will be required to fully elucidate differences in metabolic physiology between colesevelam and statins and to understand whether the changes in BAs contribute to the regulation of this miRNA cluster and its target genes.

Previous studies have linked individual miRNAs of the miR-96/182/183 cluster to glucose metabolism. For example, miR-182 knockout mice have impairments in glucose tolerance (55). Furthermore, miR-182-5p has been reported to be downregulated in numerous tissues (blood, adipose tissue, pancreas, muscle, and liver) in diet-induced obesity (DIO) mouse models. Likewise, miR-182-5p levels have been found to be decreased in blood from DIO nonhuman primates and T2D patients (56). Recently, Zhang et al. (55) reported the miR-182 knockout mice have impaired glucose tolerance, which they linked to decreased glucose utilization by skeletal muscle. Similarly, we observed an increase in glycolysis in colesevelam-treated ZDF rats, which may also be due to improvements in whole body insulin sensitivity. In addition, miR-182 targets a member of the forkhead box family of transcription factors, FoxO3, which has been implicated in muscle atrophy (25). Therefore, it is possible that miR-96/182/183 in nonhepatic tissues may contribute to glucose control. However, the amount of cholesterol flux in other peripheral tissues (i.e., muscle and WAT) in response to colesevelam and T2D is unlikely to be elevated. Nevertheless, miR-96/182/183 expression in these tissues may be independent of SREBP2 and influence glucose metabolism through unique mechanisms.

Results from this study identified MED1 as a novel target of all three miRNAs in the miR-96/182/183 cluster in both rodents and humans. MED1 is a large multisubunit protein, originally identified as a peroxisome proliferator-activated receptor (PPAR) binding protein (57). MED1 has been found to bind to a number of nuclear receptors, including PPARα, PPARγ, estrogen receptor-α (ER-α), glucocorticoid receptor (GR), constitutive androstane receptor (CAR), and thyroid hormone receptor-α, among others, and its function has been proposed as integrating ligand-dependent activity of nuclear receptors with transcription (8, 27, 54, 57). MED1 is also a cofactor of the RNA polymerase II; therefore, it is likely that through MED1, miR-96/182/183 regulates numerous, unexplored genes, and biological pathways. Nonetheless, mice with mutations in the nuclear receptor binding motif of Med1 are viable, and it is postulated that these mice are protected from insulin resistance and glucose intolerance that occur in wild-type mice fed a high-fat diet (8, 28). These observations are in line with our findings that diabetic mice treated with colesevelam have increased miR-182-5p expression, decreased hepatic Med1 expression, and improvements in glucose tolerance. Nevertheless, further work will be required to understand whether hepatic Med1 regulates glucose metabolism. Med1-null mice are embryonic lethal (58); however, tissue-specific Med1 knockout mice have been generated. For example, skeletal muscle-specific Med1 knockout mice showed improved glucose tolerance, insulin sensitivity, and resistance to high-fat DIO (9). Liver-specific Med1 knockout mice have been used to demonstrate a role for Med1 in liver regeneration, hepatocellular proliferation, induction of PPARα responsive genes in the liver and development of hepatic steatosis; however no glucose phenotypes were reported for these mice (3, 27, 37). Of note, liver specific Med1 knockout mice were found to phenocopy PPARα knockout mice (27, 58) and were reported to have decreased expression of genes involved in fatty acid oxidation, fatty acyl-CoA oxidase (ACOX), and PDK4 (27). In addition, liver specific Med1 knockout mice are protected from developing hepatic steatosis (3). Although we did not find alterations in fatty acid oxidation or lipogenic genes in this study (data not shown), it is possible that miR-96/182/183-dependent downregulation of Med1 alters PPARα or PPARγ gene regulation under conditions distinct from those assessed in this study, e.g., stress, PPARα or PPARγ activation, or an earlier diabetic phenotype. Furthermore, Med1 has also been shown to be posttranslationally regulated by phosphorylation and cellular localization (49), and it is possible that while a 50% reduction in Med1 mRNA and protein levels were observed in response to miR-96/182/183 overexpression, these posttranslational regulatory mechanisms maintain the regulation of PPARα or PPARγ gene expression intact.

One key limitation to this study is that we only investigated the effects of colesevelam on hepatic gene expression. While it is unlikely that SREBP2 is altered in other tissues in response to colesevelam, further studies will be required to elucidate whether bile acid changes and signaling through the G protein-coupled bile acid receptor 1 (TGR5) alter miR-96/182/183 expression in other metabolic tissues, e.g., islet, adipose, and muscle. Similarly, the regulation of Med1 by miR-96/182/183 in these tissues also requires further investigation, as Med1 has also been found to affect metabolic phenotypes outside the liver, e.g., regulation of glucose metabolism through mitochondrial uncoupling in muscle (9), and PPARγ gene transcription and adipogenesis in adipose tissue (21). As further insights into the gene regulatory networks between miRNAs and mRNAs are uncovered, it is likely that we will further appreciate that miRNAs may not regulate the same set of genes in every cell type or under all cellular conditions, as stoichiometry and RNA-protein interactions likely affect miRNA-mRNA interactions.

Taken together, we propose a model in which colesevelam upregulates the miR-96/182/183 cluster in the livers of diabetic mice (db/db) and rats (ZDF). Inhibition of miR-182 in vivo partially reverses the beneficial effects of colesevelam on glucose tolerance. While we cannot rule out that miR-183-5p and miR-96-5p may contribute to colesevelam’s glucose-lowering effect, the low levels of expression and physiological effects observed with inhibition of miR-182-5p alone in vivo suggest that miR-182-5p is the main contributor to these effects. Nevertheless, we have identified MED1 as a novel direct target of all members of the miR-96/182/183 cluster in both mice and humans. Therefore, we propose that miR-182-5p alters hepatic gene expression, likely through Med1 and other target genes, and contributes to the beneficial effects of colesevelam in systemic glucose metabolism.

GRANTS

This work was supported by American Heart Association Grant 15PRE25090205 (to L. R. Sedgeman) and National Institute of Diabetes and Digestive and Kidney Diseases Grants T32-HL-007411 (to L. R. Sedgeman) and R01-HL-128996 (to K. C. Vickers).

DISCLAIMERS

The funding sources did not have any involvement in study design, data collection, analysis, interpretation, writing the report, or the decision to submit the article for publication.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

L.R.S., C.B., S.M.T., and K.C.V. conceived and designed research; L.R.S. and C.B. performed experiments; L.R.S., C.B., and M.A.R.S. analyzed data; L.R.S., C.B., R.M.A., and K.C.V. interpreted results of experiments; L.R.S., C.B., M.A.R.S., and K.C.V. prepared figures; L.R.S. and K.C.V. drafted manuscript; L.R.S., C.B., R.M.A., and K.C.V. edited and revised manuscript; L.R.S., C.B., R.M.A., M.A.R.S., S.M.T., and K.C.V. approved final version of manuscript.