Hypothalamic miR-1983 Targets Insulin Receptor β and the Insulin-mediated miR-1983 Increase Is Blocked by Metformin

Abstract

MicroRNAs (miRNAs) expressed in the hypothalamus are capable of regulating energy balance and peripheral metabolism by inhibiting translation of target messenger RNAs (mRNAs). Hypothalamic insulin resistance is known to precede that in the periphery, thus a critical unanswered question is whether central insulin resistance creates a specific hypothalamic miRNA signature that can be identified and targeted. Here we show that miR-1983, a unique miRNA, is upregulated in vitro in 2 insulin-resistant immortalized hypothalamic neuronal neuropeptide Y-expressing models, and in vivo in hyperinsulinemic mice, with a concomitant decrease of insulin receptor β subunit protein, a target of miR-1983. Importantly, we demonstrate that miR-1983 is detectable in human blood serum and that its levels significantly correlate with blood insulin and the homeostatic model assessment of insulin resistance. Levels of miR-1983 are normalized with metformin exposure in mouse hypothalamic neuronal cell culture. Our findings provide evidence for miR-1983 as a unique biomarker of cellular insulin resistance, and a potential therapeutic target for prevention of human metabolic disease.

The hypothalamus functions as the key modulator of nutritional status and energy homeostasis. Neuropeptide Y (NPY)/agouti-regulated peptide (AgRP) and pro-opiomelanocortin (POMC)-expressing neurons located adjacent to the third ventricle, and projecting to the median eminence that has a fenestrated blood-brain barrier, are ideally situated to sense changes in circulating factors, such as peripheral hormones and dietary components. The dysregulation of insulin- and leptin-sensitive NPY/AgRP and POMC neuronal populations have been linked to the development of obesity, insulin resistance, and type 2 diabetes mellitus (T2DM) (1).

In the hypothalamus, insulin, an anorexigen, acts on NPY/AgRP neurons to decrease food intake, and ultimately body weight (2). Hypothalamic actions of insulin are also necessary for insulin-induced suppression of hepatic gluconeogenesis and adipose lipolysis. The inability of the hypothalamus to respond to insulin to mediate these effects is defined as central insulin resistance. This insulin resistance in the brain is known to develop prior to peripheral tissues, preceding adipocyte accumulation on a high-fat diet (HFD) (3). Decreased insulin receptor (InsR) levels in NPY/AgRP neurons of the arcuate nucleus alone have been shown to be sufficient to induce central insulin resistance and hyperphagia (4). Of interest, metformin treatment, a first-line therapy for insulin resistance, has been linked to changes in NPY expression in the hypothalamus of mice and humans (5-7). Thus, impairment of insulin signaling in NPY neurons potentially acts as a precipitating factor for the development of both central and peripheral insulin resistance; however, the cellular mechanisms by which this occurs remain undefined.

Previous work in our laboratory has demonstrated that an NPY/AgRP neuronal model exposed to high levels of insulin develops cellular insulin resistance, the inability to respond to insulin, between 8 and 24 hours, as measured by a reduction in phosphorylated protein kinase B (Akt) activation with subsequent insulin treatment (8). Both InsR and InsR substrate 1 protein were diminished with the development of insulin resistance but did not involve any changes at the messenger RNA (mRNA) level, implicating posttranscriptional mechanisms. Lysosomal degradation of the InsR was partially responsible for the downregulation of InsR, yet, blocking this pathway did not completely restore insulin signaling as measured by phosphorylated Akt. This finding suggested that another layer of regulation could exist, potentially through microRNAs (miRNAs), small RNA molecules of approximately 20 to 22 nucleotides that can posttranscriptionally regulate mRNAs, preventing protein translation (9).

Many studies have implicated miRNAs in metabolic disease, acting individually or in concert with one another, but most have focused on peripheral insulin resistance and metabolism (10-14). Knockout of the key endoribonucleases in miRNA biogenesis, Drosha DGCR8 (10, 12) or Dicer (13, 15-18), causes detrimental changes in energy metabolism. In particular, specific miRNAs, such as miR-143, miR-802, miR-103, miR-107, miR-28-1, and the let-7 family, among many others, have been linked to metabolic changes, including insulin sensitivity, in adipocytes, liver, and the pancreas (19-22). In the brain, miR-200a, miR-103/107, miR-7, and miR-17-92 have been found to mediate insulin sensitivity, specifically in hypothalamic tissue (23, 24), whereas miR-7a has been linked to the hypoglycemic response (25). Therefore, miRNAs of the central nervous system have a key role to play in regulating whole-body energy homeostasis, and need further analysis given the increasing numbers of miRNAs being described.

In this study, our established in vitro neuronal models were used to investigate whether NPY-expressing neurons exhibit a hallmark miRNA expression profile on the induction of cellular insulin resistance and to assess whether this is reflective of the changes occurring in HFD-fed mice. We sought to examine the targets of the top identified candidate miRNA, miR-1983, for links to insulin signaling to further understand the mechanism underlying the onset of insulin resistance in NPY/AgRP neurons. We also cotreated with insulin and metformin to determine if this insulin sensitizer could also alleviate the changes in miRNAs. Finally, we measured the levels of miR-1983 in the blood serum of obese humans to determine its diagnostic potential.

Materials and Methods

Cell Models and Culture Conditions

The murine mHypoE-46 (embryonic-derived male) (26) and mHypoA-NPY/GFP (adult-derived female) (27) cell lines were cultured in 5.5-mM glucose Dulbecco’s modified eagle medium (DMEM; Sigma-Aldrich) supplemented with 5% fetal bovine serum (FBS; Gibco, Thermo Fisher Scientific), and 1% penicillin-streptomycin (Gibco) at 37 °C with 5% CO₂, as described previously (8, 28).

Microarray Experiment: Analysis and Validation

The mHypoE-46 cell line was treated with sterile 1X phosphate-buffered saline (PBS, vehicle control) or insulin (NovoRapid Insulin, Novo Nordisk) for 4 (10 nM) and 24 hours (100 nM), after which miRNAs were harvested and isolated with the Mirvana miRNA isolation kit (Ambion, Thermo Fisher Scientific) according to manufacturer instructions. Total RNA was assessed for quantity and quality on a Nanodrop 2000c spectrophotometer (Thermo Fisher Scientific) before submission to the Princess Margaret Genomics Centre, Toronto, Ontario, Canada, for miRNA analysis on the NanoString nCounter analysis system for mouse miRNA and mouse-associated viral miRNAs, based on miRbase v15 containing 620 probes. Initial assessment of quality and standardization of data were conducted using nSolver Analysis software NanoString v1.0 by NanoString Technologies. Code counts were normalized to the positive controls using the geometric mean method. Next, CodeSet content was performed using the top 100 genes and the geometric mean method. The normalized code counts were converted to log2 values for statistical analysis. The data set obtained after the analysis is available on NCBI GEO accession GSE125951. On receipt of microarray results, experiments in the mHypoE-46 cells were repeated and RNA isolation conducted using Purelink Mini RNA isolation columns (Ambion, Thermo Fisher Scientific). The mHypoA-NPY/GFP cells were also treated with 100-nM insulin for 16 hours to examine miR-1983 expression, a time point sufficient to cause insulin resistance in this line based on prior work. DNase treatment was performed during RNA isolation using on-column Purelink DNase as per directions from the manufacturer (Thermo Fisher Scientific). RNA was quantified and assessed for purity on the Nanodrop 2000c spectrophotometer (Thermo Fisher Scientific). mHypoE-46 cells were also treated with H₂O (vehicle) or 20-μM metformin hydrochloride (Tocris Bioscience, Cedarlane) for 1 hour before the addition of PBS or 100-nM insulin for 24 hours. RNA was isolated using the Norgen Total RNA Purification Kit with an on-column DNAse I treatment (Norgen Biotek Corp).

Reverse Transcription, Primer Design, and Quantitative Polymerase Chain Reaction for Messenger RNA Analysis

Total RNA (500 ng to 1 μg) was used for first-strand complementary DNA (cDNA) synthesis via the Applied Biosystems (ABI) high-capacity cDNA archive kit (Thermo Fisher Scientific) following the manufacturer’s instructions. Quantitative polymerase chain reaction (qPCR) was conducted using the Platinum SYBR Green qPCR SuperMix-UDG w/ROX (Thermo Fisher Scientific), and primers were designed using Primer Quest online tool (Integrated DNA Technologies; IDT). Primers included insulin receptor, forward 5’TCCCATCAAATATTGCCAAAATT3’, reverse 5’CAGAAATAGATAAATACTTCCAATCAC3’; and histone 3a, forward 5’CGCTTCCAGAGTGCAGCTATT3’, and reverse 5’ATCTTCAAAAAGGCCAACCAGAT3’. For all cell lines, samples of 12.5 or 25 ng of cDNA were loaded in triplicate for qPCR and a dose curve of 50, 25, 12.5, 6.25, and 3.125 ng was used to calculate absolute RNA values. Total mRNA levels of all genes were normalized to histone 3a for analyses. Samples in 384-well plates (ABI) were run on an ABI 7900HT Thermal Cycler under the following conditions: 50 °C for 2 minutes, 95 °C for 2 minutes, then 40 cycles of 95 °C for 15 seconds, and 60 °C for 1 minute followed by a melt curve at 95 °C for 15 seconds, 60 °C for 15 seconds, and 95 °C for 15 seconds. Output from the ABI 7900HT thermal cycler was analyzed using ABI Prism 7000 SDS 2.4 software (ABI).

Reverse Transcription–Polymerase Chain Reaction for MicroRNA analysis for Cell Line Experiments and Animal Tissues

Total RNA (100-250 ng) was used to generate cDNA using the TaqMan miRNA Reverse Transcription Kit (Thermo Fisher Scientific) and TaqMan miRNA assay–specific primers for RT and qPCR reactions for miR-1983 (ID:121204_mat), miR-20a-3p (ID:002491), miR-296-5p (ID:000527), Let-7d-5p (ID:002283), and internal control snoRNA202 (ID:001232). qPCR was conducted using 25 ng of cDNA in triplicate with TaqMan Fast Advanced Master mix (Thermo Fisher Scientific). For the metformin experiment, total RNA (10 ng) was used to generate cDNA using the TaqMan Advanced miRNA cDNA synthesis kit, and qPCR was conducted in triplicate using TaqMan Advanced miRNA assays for miR-1983 (ID: mmu482701_mir) and internal control miRNA, mir-221-3p (ID: mmu481005_mir), and TaqMan Fast Advanced master mix according to the manufacturer’s protocol. The advanced cDNA kit was used for this experiment because it is the current technology used to detect miRNAs, and small nucleolar RNAs (snoRNAs) are incompatible with this kit. Plates were run on an ABI 7900HT thermal cycler under the following conditions: 95 °C for 20 seconds, 40 cycles of 95 °C for 1 second, and 60 °C for 20 seconds. The ΔΔCT method was used to calculate relative fold change between groups (29). miRNA levels were normalized to snoRNA202 or to mir-221-3p levels as an internal control, depending on the compatibility with the cDNA kit used.

MicroRNA Target Identification

Targets were identified using miRWalk 2.0 software(30). Potential targets appearing in all 4 algorithms (miRWalk, miRanda, RNA22, and TargetScan mouse) were selected for classification into pathways of interest using pantherdb.org. The mRNA targets found to be involved in the insulin signal transduction pathway were selected for further investigation. miRNA targets were validated as having good miRSVR scores according to the accepted literature (31), and target sites similar between human and mouse mRNA sequences were identified using TargetScan mouse.

Cloning of the Insulin Receptor 3’ Untranslated Region Into pmirGLO and Cotransfections of Construct and MicroRNA-1983 Mimic in mHypoE-46 Cells

The 3’ untranslated region (UTR) of the Insr gene was amplified from mouse genomic DNA (Genomic DNA Purification Kit, Thermo Fisher Scientific) with primers designed to add restriction enzyme sites complementary to the multiple cloning site of the pmirGLO vector (Promega Corp). Primers: Insr 3’UTR + SacI site (bold letters) forward 5’TTGTCTGCATGAGCTCTCAGAAGTCTTGCTCAGGTG3’ and Insr 3’UTR + SalI site (bold letters) reverse 5’ CGGTCTGACCCGTCGACACCATCATTCATTTACCAAG3’. After restriction digest of both the vector and 3’UTR with SacI and SalI, the 2 were ligated with T4 Ligase (Thermo Fisher) and transformed into HB101-competent cells. Bacterial colonies that were ampicillin resistant were selected and grown to isolate DNA. The presence of the insert in the correct orientation was confirmed by running restriction-digested products on a 1% agarose gel (Thermo Fisher). Restriction enzymes and buffers were purchased from Cell Signaling Technology Inc. Vector DNA containing the 3’Insr UTR insert was sent for sequencing (for primers see Table 1) at the Centre for Applied Genomics. For transfections, the mHypoE-46 cells were cultured to 70% to 75% confluency in 6-well plates (Sarstedt). Twenty-four hours later, 25 nM of miRNA-1983 mimic or negative control (as used earlier) and 1 μg of pmiRGLO-Empty or 3’ Insr UTR was transfected using 6 μL of Turbofect per well (Dharmacon, GE) according to the manufacturer’s specifications for 48 hours. Cells were washed once with ice-cold 1 × PBS, after which cells were lysed in 450-μL passive lysis buffer per well and assayed using a dual luciferase assay kit (Biotium Inc) according to the manufacturer’s instructions.

Table 1.

Candidate microRNAs that were significantly altered by 100-nM insulin at 24 hours vs 10-nM insulin for 4 hours or by 100-nM insulin at 24 hours vs 24 hours of vehicle phosphate-buffered saline

miRNA	Insulin: 24 h vs 4 h	P	24 h: insulin vs PBS	P	4 h: insulin vs PBS	P
mmu-miR-1983	1.45	.03	1.35	.06	1.06	.44
mmu-miR-20a+mmu-miR-20b	1.83	.03	1.34	.23	–1.11	.54
mmu-miR-296-5p	1.95	.03	1.29	.30	1.27	.36
mmu-let-7d	1.64	.05	1.21	.39	–1.02	.82
mmu-miR-361	–1.17	.01	–1.08	.03	–1.2	.32
mmu-miR-96	–1.53	.005	–1.17	.10	–1.06	.48
mmu-miR-30b	–1.46	.007	–1.18	.07	1.03	.46
mmu-miR-1198	1.06	.44	1.3	.001	–1.11	.83
mmu-miR-33	–1.44	.08	–1.31	.01	–1.19	.42
mmu-miR-450a-5p	–1.16	.18	–1.14	.05	–1.31	.29

None of these miRNAs were altered by 4-hour treatment with insulin compared to vehicle PBS. P less than .05 is considered statistically significant.

Abbreviations: miRNA, microRNA; PBS, phosphate-buffered saline.

MicroRNA-1983 Mimic Transfections in mHypoE-46 Cells

mHypoE-46 cells were cultured to 70% to 75% confluency in 100-mm tissue culture dishes (Sarstedt). Approximately 24 hours later, 25 nM of miRNA-1983 mimic or negative control (Thermo Fisher Scientific) was transfected using 5 μg/mL of Dharmafect 3 (Dharmacon, GE) according to the manufacturer’s specifications for 24 or 48 hours. Cells were washed with ice-cold 1 × PBS, after which total protein and RNA (containing miRNAs) were isolated using the Ambion PARIS Kit according to specifications of the manufacturer.

Protein Isolation, Sodium Dodecyl Sulfate–Polyacrylamide Gel Electrophoresis and Western Blotting

Protein was isolated as previously described (28) or using the Ambion PARIS Kit according to the manufacturer’s specifications and quantified using bicinchoninic acid protein assay kit (Pierce, Thermo Fisher Scientific). For Western blotting, 15 to 25 μg of protein was resolved on 10% bis-acrylamide-tris gels and transferred onto 0.2-μm polyvinylidene fluoride using the iBlot 2 dry blotting system (Invitrogen, Thermo Fisher Scientific). Antibodies rabbit antimouse for IRβ-subunit (32), α-tubulin (33), with rabbit secondary antibody immunoglobulin G (34), were used with SignalFire Elite ECL reagent (purchased from Cell Signaling Technology Inc). The β-actin antibody (35) was purchased from Sigma-Aldrich. All primary antibodies were diluted 1:1000 and were incubated overnight at 4 °C in 5% nonfat milk in tris-buffered saline (TBS) with 0.1% Tween-20 (TBS-T), except for α-tubulin, which was diluted in 5% BSA in 0.1% TBS-T. Blots were then washed in 0.1% TBS-T and incubated with an antirabbit secondary antibody diluted 1:7500 in 5% nonfat milk in 1X TBS-T for 1 to 2 hours at room temperature. Blots were washed in 0.1% TBS-T and developed on a Kodak Imaging 3000 station (Eastman Kodak Company). Western blotting for mouse hypothalamic tissue was conducted so that all groups were processed, run on gels, incubated with antibodies, and imaged together at the same time. Densitometry was conducted using Image J software (National Institutes of Health, http://imagej.nih.gov/ij/).

Animal Models

Male and female mice of the CD-1 strain were purchased from Charles River Laboratories at age 7 weeks and allowed to acclimatize for 1 week before the commencement of experiments. Mice were kept on 12:12 light-dark schedule (7 am-7 pm). All animal experiments were conducted with the approval of the animal care committee at the University of Toronto. CD-1 mice were fed either 60% kilocalories from fat (60% HFD), control (7% sucrose matched) diet (D12492 and D12450J respectively, Research Diets Inc) or a rodent chow diet (Teklad Diet 8664; Envigo) as indicated in the figure legends. The MKR mice were bred in house and genotyped as described previously (36-38). MKR mice were fed Teklad Diet 8664. Where indicated, mice were weighed weekly to track their weight gain over the course of the 5 weeks of diet exposure. All animals were euthanized with isoflurane, and tissues were then collected and stored at –80 °C until processed for analysis. RNA and protein were isolated from tissues using the Ambion mirVana PARIS Kit (Thermo Fisher Scientific). Blood collected to assay insulin levels was retrieved using heparin-coated syringes, after which tubes were spun at 6000g at 4 °C for 10 minutes wherein plasma was isolated and stored at –80 °C. For serum miRNA analysis, blood was harvested under isoflurane anesthesia via cardiac puncture into uncoated syringes, and allowed to clot on ice for 1 hour, then centrifuged at 3000g for 10 minutes before storage at –80 °C (39).

Primary Culture Experiments

For primary culture experiments, cells from the hypothalamii of four 8-week-old male CD-1 mice were dispersed by trituration and cultured in neurobasal A media (Gibco, Thermo Fisher Scientific) supplemented with 10% FBS, 5% horse serum (Gibco), 1% PS (Gibco), 1 × B27 serum-free supplement (Gibco), and 1 × GlutaMAX supplement (Gibco) for 7 to 9 days. Each hypothalamus was divided into 2 culture plates and each plate received 10-ng/mL ciliary neurotrophic factor (CNTF, R&D Biosystems) every 3 days to induce proliferation. Cultures were then treated with PBS or 100-nM insulin for 24 hours in DMEM containing 5.5 mM glucose, 5% FBS, and 1% PS before protein collection as described next. All animal procedures were conducted in accordance with the regulations of the Canadian Council on Animal Care and approved by the University of Toronto’s animal care committee.

Human Patient Study

This study was approved by the University Health Network Research Ethics Board and was conducted in compliance with the Declaration of Helsinki. Individuals with varying body mass indices (BMIs) and insulin sensitivities were recruited. Serum was collected in the overnight-fasted state, in uncoated (no EDTA) tubes, and spun at 3000g for 15 minutes at 4 °C. Serum was then immediately frozen at –20 °C, then moved to –80 °C for long-term storage. Measurements of circulating glucose were measured by a bedside glucometer (StatStrip, Nova Biomedical Corp). Circulating insulin in the serum samples was measured via enzyme-linked immunosorbent assay on the Abbott Architect i2000 system at the Toronto General Hospital core laboratory or by the HI-14K Human Insulin-Specific radioimmunoassay from EMD Millipore according to the manufacturer’s directions (Etobicoke). The core laboratory at Toronto General Hospital assayed all other circulating blood parameters listed. A total of 12 participants (3/9 female/male; all values as mean ± SEM; BMI, 35.8 ± 4.2; age, 46.5 ± 3 years; fasted glucose levels, 5.2 ± 0.2) were included in the pilot serum study for the Pearson coefficient (R) and linear regression analysis, where insulin and glucose levels, as well as the other listed metabolic parameters, were available for all individuals. Homeostatic model assessments of insulin resistance (HOMA-IR) scores were determined for the serum group using the HOMA2 model (40). Exclusions were individuals who had undergone bariatric surgery, had extremely high insulin levels (outside range), and/or were taking other potentially confounding drugs. No participants included in the regression analysis were taking insulin or insulin sensitizers, unless used in the comparative data with metformin-treated (8) vs drug-free (19) individuals. A further 17 samples were included in the metformin-treated (7/1 female/male; all values as mean ± SEM; BMI, 48.7 ± 10.2; age, 51.9 ± 8.0 years; glucose levels not available) vs drug-free participants, (6/1 female/male; all values as mean ± SEM; BMI, 53.0 ± 6.1; age, 42.1 ± 11.7 years; glucose levels not available).

MicroRNA Isolation From Mouse and Human Serum and Reverse Transcription–Quantitative Polymerase Chain Reaction

Serum samples were isolated using the Ambion mirVana PARIS Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. After addition of the 2× denaturing solution (mirVana kit) or lysis buffer A (Norgen kit), cel-miR-39-3p RNA (25 fmol; Sequence 5’→3’; UCACCGGGUGUAAAUCAGCUUG; IDT, currently use Norgen) was added to each sample to serve as a control. The isolation was then continued as per the manufacturer’s instructions (41). cDNA was synthesized using the TaqMan Advanced miRNA cDNA synthesis kit, and qPCR was conducted in triplicate using TaqMan Advanced miRNA assays for miR-1983 (ID: mmu482701_mir) and cel-mir-39 (ID: 478293_mir) and TaqMan Fast Advanced master mix according to the manufacturer’s protocol. Human serum samples were also assayed for the presence of known circulating human miRNAs (42) miR-142-3p (TaqMan ID: 477910) and miR-222-3p (TaqMan ID: 477982). Samples were run on an ABI 7900HT thermal cycler under the following conditions: 95 °C for 20 seconds, 40 cycles at 95 °C for 1 second, and 60 °C for 20 seconds. Comparison of mouse experimental groups was conducted via the ΔΔCT method.

Statistical Analysis

All data were analyzed using GraphPad Prism version 6.0c. A t test or 1- or 2-way analysis of variance followed by Bonferroni or Fisher least significant difference post hoc analyses were conducted, as described. Pearson correlation or linear regression analyses were conducted where indicated. Data were considered statistically significant where the P value was found to be less than .05. All data are presented as ± SEM unless indicated otherwise.

Results

To identify changes in the miRNA profile with the onset of insulin resistance, mHypoE-46 neurons were treated with 10- and 100-nM insulin for 4 and 24 hours, respectively, to capture periods of normal and aberrant insulin signaling, as defined in our previous studies (8). Microarray analysis revealed significant changes in miRNAs across all groups; however, only 10 miRNAs were found to be significantly different at 24-hour insulin exposure vs 4-hour insulin or 24-hour vehicle (see Table 1). Our top candidates, based on the level of statistical significance of changes at 4- vs 24-hour insulin, were miR-1983, -20a-3p -20-b-5p, -296-5p, and let-7d-5p. Among the top 4 miRNAs, qRT-PCR validation of the array data found that only miR-1983 (Fig.1A) and miR-20a-3p (Fig. 1B) displayed an upregulation at 24 hours in the same cell line used in the array, mHypoE-46. To determine if the changes in miR-1983 and miR-20a-3p were unique to the mHypoE-46 neurons, the mHypoA-NPY/GFP cell line was treated with insulin for 16 hours, a time point sufficient to elicit cellular insulin resistance in this cell line (28). With this validation, we found that only miR-1983 was increased on insulin exposure in the mHypoA-NPY/GFP cells, whereas miR-20a-3p was not changed in these neurons. Thus, our top candidate for further study was miR-1983, an isoleucine (Ile) pretransfer RNA (tRNA)-derived miRNA. Since metformin is used as the first-line therapy for insulin resistance, we also pretreated the mHypoE-46 neurons with 20-μM metformin with or without insulin. The insulin-induced increase in the levels of miR-1983 were completely normalized with metformin exposure (Fig. 1C).

Figure 1.

High insulin levels upregulate miR-1983 in models of hypothalamic neurons. A, Microarray (n = 3) and quantitative polymerase chain reaction (qPCR) validation (n = 4-5) of miR-1983 in the mHypoE-46 cell line after 4-hour (10 nM for microarray) and 24-hour (100 nM) treatment with insulin. miR-1983 levels in the mHypoA-NPY/GFP (n = 6-7) were also quantified on induction of insulin resistance at 16 hours. B, Microarray (n = 3) and qPCR validation (n = 4-5) of miR-20a-3p/20b-5p, miR-296-5p, and let-7d-5p in the mHypoE-46 cell line at 4 and 24 hours, as well as miR-20a-3p in mHypoA-NPY/GFP neurons at 16 hours (n = 4-5). C, Metformin (20 μM) treatment of the mHypoE-46 neurons with and without 100 nM insulin for 24 hours. Data shown are mean ± SEM. *P less than .05, **P less than .01, ***P less than .001.

Next, we sought to determine whether miR-1983 targets any genes involved in insulin signaling. Target analysis with current software identified the InsR as a potential target of miR-1983, and a closer in silico analysis of the 3’ UTR of the Insr gene revealed a conserved binding site (Fig. 2A). To assess if miR-1983 could bind to the 3’UTR of Insr, transient transfections of the pmirGLO vector containing this region were conducted with an miR-1983 mimic or negative control. The miR-1983 mimic reduced the luciferase output by approximately 15% after 48 hours (Fig. 2B), suggesting that miR-1983 binds the Insr 3’UTR and subsequently reduces translation of the reporter. We also determined that transfection of the neurons with a mimic of miR-1983 was sufficient to decrease the protein levels of the IRβ subunit by approximately 20% after 24 hours, and by 35% after 48 hours (Fig. 2C). While Insr mRNA levels were not changed by the miR-1983 mimic at 24 hours, 48 hours of transfection decreased Insr mRNA levels by approximately 12% (see Fig. 2C). These results confirm that miR-1983 downregulates IRβ levels in NPY/AgRP neurons. Hallmarks of insulin-induced cellular insulin resistance were confirmed by quantifying Insr mRNA and IRβ subunit protein levels after a 24- and 48-hour insulin exposure (Fig. 2D). As reported previously, we found that the IRβ subunit, but not Insr expression, was significantly decreased on longer-term exposure to 100-nM insulin. However, miR-1983 levels in the mHypoE-46 cells returned to normal after 48 hours despite the continued decrease in IRβ subunit (see Fig. 2D), indicating that other mediators, including lysosomal degradation, are also involved. Taken together, these data suggest that the Insr mRNA is targeted by miR-1983, preventing protein translation and may contribute to InsR downregulation by high insulin exposure.

Figure 2.

miR-1983 targets and decreases IRβ protein in mHypoE-46 neurons. A, TargetScan analysis of the Insr-3’ untranslated region (3’ UTR) in mouse for the miR-1983 binding site and conserved cross-species sequence alignment. B, Transfection of the pmirGLO vector (1 μg) containing approximately 3000 bp 3’ Insr-UTR sequence concurrently with the miR-1983 mimic (25 nM) (n = 3-4 independent experiments/group, each run in triplicate). C, Messenger RNA (mRNA) and protein levels of Insr and IRβ after transfection of a miR-1983 miRNA mimic (25 nM) into mHypoE-46 cells for 24 (n = 4) or 48 (n = 4) hours. D, mRNA and protein levels of Insr and IRβ after exposure of mHypoE-46 cells to insulin (100 nM) for 24 h (n = 4) with miR-1983 and IRβ levels at 48 h (n = 4). Data shown are mean ± SEM. *P less than .05, ***P less than .001.

Our next question was to determine if the increase in miR-1983 in the hypothalamus could be induced by hyperinsulinemia in a mouse model similar to our cell model paradigm. We collected the hypothalamii from male and female MKR mice, known to have hyperinsulinemia, severe insulin resistance, and decreased fatty acid oxidation, despite being lean (36,37). These mice displayed pronounced elevation of miR-1983 in their hypothalamus (Fig. 3A). These results are independent of an HFD feeding that could potentially induce other confounding factors in the mice, such as high plasma lipid levels (or hyperlipidemia). A second strain of mouse, CD-1, was used to assess miR-1983 levels after 5 weeks of an HFD. While the changes in the MKR mice were much more pronounced, the CD-1 hypothalamii also had a modest increase in miR-1983 (see Fig. 3A). Of interest, male and female MKR mice both have high insulin levels, but only the male CD-1 mice have increased insulin, and this is associated with the increased levels of miR-1983 that are not seen in female CD-1 mice (see Fig. 3A). Significant alterations were observed in miR-1983 in visceral fat, but not the soleus muscle (Fig. 3B). This miRNA was not detectable in the liver. There was a significant decrease in the kidney of miR-1983 in male mice fed a 60% HFD (see Fig. 3Bi). This particular finding is in line with findings by Edinger et al, as miR-1983 is downregulated by aldosterone in the kidney (43), possibly elicited by increasing aldosterone content within this tissue documented with HFD exposure (44). Primary culture of hypothalamic neurons from 8-week-old CD-1 mice treated with 100-nM insulin or vehicle for 24 hours demonstrated that the IRβ subunit is downregulated in a nonimmortalized and heterogeneous population of hypothalamic neurons (Fig. 3C). Furthermore, CD-1 mice exposed to a 60% HFD for 5 weeks showed decreased hypothalamic IRβ subunit levels compared to chow-fed mice (see Fig. 3C). Both findings provide evidence of hypothalamic insulin resistance in the whole hypothalamus that correlate to high insulin levels and increased miR-1983 levels in these mice.

Figure 3.

miR-1983 is elevated in the hypothalamus of male female mice concomitant with hyperinsulinemia, while serum miR-1983 levels are expressed in other metabolic tissue and positively correlated with circulating insulin levels and decreased IRβ. A, Hypothalamic miR-1983 and fasting insulin levels of male CD-1 mice on sucrose-matched control (n = 7-8) or 60% high-fat diet (HFD) (n = 8) compared to MKR (n = 8) mice. Hypothalamic miR-1983 and fasting insulin levels of female CD-1 mice on sucrose-matched control (n = 7-8) or 60% HFD (n = 8) compared to MKR (n = 3) mice. B, (i) miR-1983 levels in the muscle, kidney and fat (n = 5-8/group) and (ii) serum miR-1983 levels (n = 7-8) of male CD-1 mice fed a sucrose-matched control vs a 60% HFD. C, Western analysis of IRβ subunit protein levels in male mouse–derived primary hypothalamic cultures treated with 100-nM insulin for 24 hours vs vehicle control (n = 4), and male CD-1 mice fed chow control or a 60% HFD for 5 weeks (n = 11-12). Full Western blot images are shown beside each graph. Data shown are mean ± SEM unless otherwise indicated. *P less than .05, ** P less than .01, ***P less than .001.

miRNAs are detectable in the plasma and serum of mice and humans and could be putative biomarkers of metabolic disease (45). As miR-1983 had previously been detected in the circulation of HFD-fed mice (39), and miR-1983 were increased with high insulin levels in vitro, we assessed serum levels of miR-1983 in HFD-fed male CD-1 mice. The levels of miR-1983 were significantly increased in HFD-fed mouse serum compared to chow-fed mice (Fig. 3Bii). Therefore, we also assessed if miR-1983 could be correlated with these parameters in human overweight and obese patients. Serum levels of miR-1983 were significantly correlated with a number of blood parameters (Fig. 4A and 4B), with the most statistically significant correlation being with increasing insulin levels and HOMA-IR scores.

Figure 4.

miR-1983 is elevated in the serum of obese humans, and is highly correlative to hyperinsulinemia and homeostatic model assessment of insulin resistance (HOMA-IR). A, Pearson correlation analysis of serum miR-1983 levels with specific blood parameters in the serum of the 12 patients included in the cohort after exclusions, and B, graphical representation of the data in the table where solid circles are statistically significant. Linear regression analysis of C, serum insulin, and D, HOMA-IR scores with miR-1983 levels in human serum samples. E, Levels of miR-1983 in the blood of patients either taking metformin (n = 8) or drug free (n = 19). Data shown are mean ± SEM unless otherwise indicated. *P less than .05, **P less than .01, ***P less than .001.

As miRNAs have been proposed to serve as biomarkers, we next asked if miR-1983, along with miR-223-3p and miR-142-3p, both of which have reported associations with obesity and diabetes (42), could distinguish between humans that were hyperinsulinemic (fasting blood insulin > 50 pmol/L) at the time of sample collection. Both miR-1983 and miR-222-3p were significantly increased in the serum of hyperinsulinemic individuals (R² = 0.80; P < .001 and .51; P < .008, respectively), unlike miR-142-3p (see Fig. 4C), as demonstrated by linear regression analysis. Importantly, miR-1983 was also highly statistically significantly (R² = 0.84; P < .001) correlated with the HOMA-IR score, while the R² score for miR-222-3p was 0.53 with P less than .01, also statistically significant (see Fig. 4C). This same cohort of patients was analyzed for the levels of miR-1983 in the blood against any individuals being treated with metformin who were excluded from the regression analysis. While there were only 8 individuals taking metformin on the list, there appears to be a difference from those not yet prescribed metformin although the data did not reach statistical significance in this pilot study (Fig. 4D). Clearly, more individuals will have to be analyzed to draw general conclusions in humans regarding the correlation to miR-1983. Nonetheless, the human data support our proposition that miR-1983 can serve as a circulating indicator of insulin resistance and should be pursued further.

Discussion

Our study has uncovered a novel association between an miRNA and high insulin levels and linked it to the potential onset of insulin resistance due to a downregulation of the IRβ protein subunit in NPY neurons from the hypothalamus. Investigating central insulin resistance is relevant since insulin resistance in NPY neurons leads to hyperphagia and increased fat mass, and thus changes in this brain region may contribute to the initiation of metabolic disease. We demonstrate that miR-1983 is not an abundantly expressed miRNA under normal physiologic conditions in mouse tissues or in human blood. In fact, it is almost undetectable in the serum of individuals with normal insulin levels. However, the significant increases detected are highly correlated to the progression to insulin resistance, with pathological implications for prediabetes.

The role of miRNAs in regulating the diverse facets of metabolism is becoming increasingly evident (14). Individual miRNAs are regulated in a tissue-specific manner (22, 24); however, it is currently unknown if the development of central insulin resistance changes specific miRNAs within the hypothalamus. This study indicates that there are specific miRNAs induced in NPY neuronal models with prolonged insulin treatment, and that this miRNA induction is conserved in the hypothalamus of animals with hyperinsulinemia. The MKR mice are a model of high insulin, without the confounding factors of an HFD-fed model (36, 37). Since the results are more robust, without the concomitant increase in weight with an HFD, this miRNA marker may prove useful for cases of nonobese insulin resistance. While requiring more detailed studies, this could have implications in obesity- and HFD-related T2DM vs the obesity-independent, hyperinsulinemic development of the disease both in lean and overweight humans.

Our finding that miR-1983 is increased with high levels of insulin and targets IRβ describes an additional mechanism by which high insulin downregulates InsR and induces cellular insulin resistance in NPY neurons. However, the differences in the magnitude of downregulation induced by insulin vs the miR-1983 mimic suggest that other mechanisms are involved, including lysosomal degradation of InsR, as previously identified (8). Furthermore, in the neuronal cell line, 48-hour insulin treatment no longer increased miR-1983, despite a sustained downregulation of IRβ. This may indicate that miR-1983 upregulation may contribute to the initial stages of IRβ downregulation or that the effects of the initial miR-1983 upregulation may be maintained. However, since the insulin was applied only once at the beginning of the experiment, it is possible that the insulin degraded over time. The MKR mouse experiment represents a situation in which the hypothalamus would be constantly exposed to a hyperinsulinemic state and supports the association between high insulin exposure and increased hypothalamic miR-1983 levels. Exactly how high insulin increases miR-1983 is currently unknown; however, a possible mechanism is implicated in the process by which miR-1983 is uniquely derived from the pre-Ile tRNA to become a fully functional miRNA (46) The low abundance of basal miR-1983 levels may be linked to the fact that it is formed from a pre-Ile tRNA and its cleavage is repressed by the RNA binding protein La (46). Phosphorylation of the La protein at the threonine 389 site by Akt dramatically impairs its action (47). Therefore, we hypothesize that prolonged activation of the insulin-initiated signal transduction pathway, and concomitant Akt activity, may also repress the action of the La protein. This would create a window in which there is increased formation of miR-1983 observed in our study in response to high levels of insulin. This hypothesis remains to be tested in future studies.

miRNAs are found in the serum both of mice and humans where they are highly stable due to either their enclosure in exosomes or being bound to proteins, such as argonaute (48, 49). As prior global miRNA analysis in obese states detected miR-1983 in the circulation (39), we assessed the levels in the serum of male CD-1 mice after 5 weeks on an HFD. The levels of miR-1983 were increased in the HFD-fed mice serum. Whether the miR-1983 in the blood originates from the hypothalamus or from the visceral fat is not known, as both were found to have increased levels and may release the miRNAs via exosomes into the circulation. Great interest in exosome biology is currently at the forefront of metabolic disease studies (50, 51). Exosomes are potential intercellular communication vesicles that can travel vast distances to alter physiological functions. Recently, miRNA-containing exosomes from hepatocytes differed between early-onset obesity and longer-term obesity, with the early exosomes promoting insulin sensitivity (52). In our case, circulating miRNA, potentially from the fat, could enhance the effects of miR-1983 to decrease InsR levels (53, 54). miRNAs have been detected in the serum of mice and humans, and are considered to be putative biomarkers of metabolic disease (45). We determined that serum miR-1983 elevations in diet-induced obese mice coincide with those in the hypothalamus, creating the opportunity that serum levels may be used as a surrogate for hypothalamic changes. As miR-1983 has now been detected in humans (43, 46), this has favorable implications for future clinical investigation into whether miR-1983 could be a marker for prediabetes.

With this in mind, metformin is often the first-line treatment on the diagnosis of increased glucose levels, and indication for the reduction of insulin resistance (55). There are many studies on the action of metformin at the level of peripheral metabolic tissues, such as fat and muscle (10-14), but not as much is known about metformin action in the brain. However, metformin can cross the blood-brain barrier and act directly on neurons (56). We have previously demonstrated that metformin can increase the phosphorylation of adenosine 5′-monophosphate kinase (AMPK) and S6 kinase in the mHypoE-46 neurons (57). Further, metformin can normalize palmitate- and tumor necrosis factor α–mediated increases in NPY (57). This evidence correlates well with what is currently known from the literature. Metformin acts in the liver through AMPK to increase insulin sensitivity (58, 59), and when injected directly into the third ventricle of the brain in mice, activates S6 kinase and decreases feeding (60). Of interest, metformin has been reported to alter certain circulating miRNAs in patients with T2DM (61, 62). Further, metformin directly increases Dicer, and its anti-inflammatory effects in macrophages are partially dependent on Dicer (63, 64). However, it has not yet been reported whether metformin can normalize levels of miRNAs in hypothalamic neurons, thus our finding both in isolated neurons and in human blood samples that metformin normalizes miR-1983 is particularly interesting. Metformin likely controls miR-1983 independently of insulin levels, as in the neuronal cell line, miR-1983 was restored with metformin despite the presence of insulin in the medium. Indeed, in further support of this idea, the patients taking metformin, who may have lower miR-1983 levels, unexpectedly did not have lower insulin levels. Since our patient cohort is relatively small and mainly consists of a distinctive set of individuals attending an obesity clinic for potential bariatric surgery, it would be prudent to expand these studies to a prospective cohort of individuals. It would also be worthwhile to treat hyperinsulinemic mice with metformin and determine the effects on miR-1983 in the circulation and hypothalamus. Future studies will investigate the mechanisms by which metformin can regulate miRNA levels, especially miR-1983, and that may potentially lead to novel therapy aimed at early detection and rescue of insulin resistance.

The worldwide RNA-based therapeutics market is one of the fastest emerging technologies. Currently, there are a number of ongoing clinical trials based on miRNA therapeutics for several diseases ranging from cancer to glioblastoma (65). In preclinical trials, miRNA formulations show great promise because of the low toxicity profiles and the high efficacy of delivery (66). The putative market for an miRNA therapy to treat prediabetes is high given the prevalence of the disease worldwide. The potential to use artificial exosomes for transport of important miRNAs involved in metabolism is a possibility (50, 51). These are sufficient reasons for further investigation of the need and use of stable miR-1983 inhibitors for prevention of insulin resistance. The overall findings of this paper illuminate the role of hypothalamic and circulating miR-1983 in the progression to insulin resistance, both cellular and whole-body as measured by HOMA-IR. The potential of miRNA-1983 as a biomarker molecule for detection of insulin resistance can prevent further development of T2DM and its complications.

Glossary

Abbreviations

ABI

Applied Biosystems

AgRP

agouti-regulated peptide

Akt

protein kinase B

AMPK

adenosine 5′-monophosphate kinase

BMI

body mass index

cDNA

complementary DNA

FBS

fetal bovine serum

HFD

high-fat diet

HOMA-IR

homeostatic model assessment of insulin resistance

Ile

isoleucine

InsR

insulin receptor

mRNA

messenger RNA

miRNA

microRNA

NPY

neuropeptide Y

PBS

phosphate-buffered saline

POMC

pro-opiomelanocortin

qPCR

quantitative polymerase chain reaction

RT

reverse transcription

snoRNA

small nucleolar RNA

T2DM

type 2 diabetes mellitus

TBS

tris-buffered saline

tRNA

transfer RNA

UTR

untranslated region

Additional Information

Disclosures: The authors have nothing to disclose.

Data Availability

Data are available on request from the corresponding author.