Neuronal Androgen Receptor Regulates Insulin Sensitivity via Suppression of Hypothalamic NF-κB–Mediated PTP1B Expression

All animal study protocols were reviewed and approved by the Animal Care and Use Committee of the University of Rochester Medical Center, in accordance with National Institutes of Health guidelines. The floxAR mice, targeting vector construction, and chimera founder generation have been described previously (17). The synapsin I–Cre mice were provided by Dr. David Rempe (University of Rochester) and bred in a C57BL/6J background. NARKO mice were identified by genomic DNA PCR, as described previously (17). Animals were housed in pathogen-free facilities, maintained on a 12-h light-dark cycle schedule, and had access to standard laboratory chow (no. 5010, Laboratory Diet; PMI Nutrition International) and water ad libitum. The mouse neuronal cell line, Neuro-2A (N2A) was obtained from American Type Culture Collection (Manassas, VA). The hypothalamic GT1-7 cell line was provided by Dr. Pamela Mellon (University of California San Diego, San Diego, CA).

Fasting blood samples were taken from mice 16–18 h after withdrawal of food. Blood glucose was measured using a glucometer (One Touch Ultra; LifeScan). Serum insulin and leptin were determined using insulin and leptin ELISA kits (Crystal Chem) according to the manufacturer’s protocol. Serum triglyceride was determined by GPO-Trinder assay (Sigma-Aldrich). Serum free fatty acid was measured using a NEFA-Kit-U (Wako). Serum testosterone was determined using an ELISA kit (Diagnostic Systems Laboratories). Tissue triglyceride content was determined in the extracts from 50–100 mg fresh, frozen livers as described previously (9).

Insulin tolerance tests were performed on 6-h fasted mice after intraperitoneal injection of 1 unit/kg body weight recombinant bovine insulin (Sigma Aldrich). Blood glucose concentrations were determined at 0, 15, 30, 45, 60, and 90 min after insulin administration. Blood glucose was measured, and the percentage of reduction in blood glucose was compared with the zero time within groups. For the pyruvate tolerance test (PTT), after an 18-h fast, mice received an intraperitoneal injection of 2 g/kg body weight sodium pyruvate (Sigma-Aldrich). Blood glucose concentrations were measured at time points indicated after injection.

Total RNA was prepared from cells or tissues with TRIzol (Invitrogen) according to the manufacturer’s instructions. cDNA synthesis was carried out by RT-PCR with Superscript RNase H-reverse transcriptase and cDNA cycle kit (Invitrogen) using 4 μg total RNA according to the manufacturer’s instructions. Expression levels of RNA were determined by quantitative real-time PCR performed in an iCycler real-time PCR amplifier (Bio-Rad Laboratories) using iQ SYBR Green Supermix reagent (Invitrogen). The relative copy number of Gapdh RNA was quantified and used for normalization. The ΔΔC_T method was used to calculate relative differences between wild-type (WT) and knockout mice.

Data are presented as mean ± SEM. Differences between two means were assessed by unpaired, two-tailed Student t test. Data involving more than two means were evaluated by one-way ANOVA followed by Tukey post hoc tests (SigmaStat [SyStat] and GraphPad Prism [GraphPad Software, Inc.]). P values <0.05 are considered statistically significant.

Immunohistochemical analysis showed that AR was highly expressed within the arcuate nucleus of the hypothalamus, ventromedial hypothalamus, and dorsomedial hypothalamus, while accounting for a relatively small fraction of cells in the paraventricular nucleus (Fig. 1A). These results suggested the putative role of AR in the hypothalamus, the critical site of insulin’s action on glucose homeostasis in the brain. To address whether AR regulated insulin signaling in neuronal cells, we manipulated AR expression in N2A cells and found that overexpression of AR enhanced insulin-dependent phosphorylation of AKT (Fig. 1B). In contrast, knocking down AR expression in N2A cells resulted in decreased insulin-stimulated AKT phosphorylation (Supplementary Fig. 1A). To further investigate the role of AR in regulating insulin signaling in neuronal cells in vivo, we generated NARKO mice by crossing floxAR mice with synapsin I–Cre mice (18). Cre-mediated genomic DNA recombination of floxed AR allele was detected in various areas of the brain, including the cerebrocortex, hypothalamus, hippocampus, and, to a relatively less extent, in the cerebellum. In contrast, AR DNA deletion was not detected in peripheral tissues, indicating a restricted AR disruption in the central nervous system (Fig. 1C). Disruption of AR in the hypothalamus of NARKO mice was examined using AR mRNA and protein expression analysis (Fig. 1D and E). NARKO mice exhibited a >85% decrease in AR expression in neurons in the arcuate nucleus of the hypothalamus, ventromedial hypothalamus, dorsomedial hypothalamus, and paraventricular nucleus (Supplementary Fig. 1B and C). Circulating testosterone levels in NARKO mice were measured and were not statistically different from controls (Fig. 1F), supporting our strategy to determine the function of brain AR without the complication of reduced testosterone signaling occurring with global AR deletion. The presence of normal external genitalia and internal reproductive organs in NARKO mice indicated normal androgen action in male sexual development (Fig. 1G).

To determine the importance of neuronal AR deficiency in the regulation of energy homeostasis, we monitored the body weights of NARKO mice and found there was a significant increase beginning at 28 weeks of age and thereafter (Fig. 2A). Consistent with the growth curves, NARKO mice at 36 weeks of age had increased visceral adiposity, reflected by enlargement of both epididymal and retroperitoneal fat pads, which was primarily accounted for by the increased size of the adipocytes (Fig. 2B–D). These data indicated the development of obesity and impaired energy homeostasis, which was also revealed by changed metabolic parameters in the serum (Table 1). Elevated serum triglycerides and free fatty acid levels suggested excessive hepatic lipid synthesis. Increased expression of hepatic lipogenesis genes sterol regulatory element-binding protein 1c (Srebp1c), stearoyl-CoA desaturase 1 (Scd1), and fatty acid synthase (Fas) in NARKO mice was noted, whereas acetyl CoA carboxylase expression was not significantly elevated (Fig. 2E). Hepatic macrovesicular steatosis, although moderate, was noted in the livers of NARKO mice using histological examinations demonstrating numerous intermediate lipid vacuoles surrounding and displacing the nuclei of hepatocytes (Fig. 2F). The hepatic lipid accumulation was further revealed by Oil Red O staining to capture lipid droplets within hepatocytes (Fig. 2G). Consistent with the histological examination, we detected increased triglycerides in the liver of NARKO mice (Fig. 2H).

Insulin resistance in NARKO mice at 36 weeks of age was suggested by insulin tolerance testing (Fig. 2I). The presence of elevated fasting blood glucose and insulin levels provides further evidence to support the notion of insulin resistance (Table 1). Increased gluconeogenesis is a characteristic of insulin resistance and fasting hyperglycemia. PTTs were performed by administration of the gluconeogenic substrate pyruvate to promote gluconeogenesis in the fasting state. NARKO mice exhibited significantly higher blood glucose concentrations at 30, 60, and 90 min after pyruvate administration than control mice (Fig. 2J). The glucose area under the curve of PTT showed a 20% increase in NARKO mice, indicating increased gluconeogenesis (data not shown). In fasted NARKO mice, the expression of the hepatic gluconeogenic gene phosphoenolpyruvate carboxykinase (Pck1) was also significantly increased, and glucose-6-phosphatase trended upwards (Fig. 2K).

Expression of the gluconeogenic genes is tightly controlled by the liver and the brain. Hepatic signal transducer and activator of transcription 3 (STAT3) has been shown to mediate insulin action in the brain to suppress hepatic glucose production, beyond direct activation of insulin signaling in the liver (19,20). We examined the induction of hepatic STAT3 phosphorylation in young and nonobese NARKO mice. The phosphorylation of hepatic STAT3 in control mice was significantly induced by glucose injection as expected. However, STAT3 phosphorylation in livers of NARKO mice was significantly reduced (Fig. 3A). These data suggest blunted action of insulin in the brain on the suppression of hepatic glucose production.

Insulin action in agouti-related peptide (AgRP) neurons was demonstrated to be required for suppression of hepatic glucose production (21). Upregulated gene expression of AgRP neuropeptide was noted in the hypothalamus of NARKO mice, further indicating that the dampened action of insulin in the hypothalamus resulted from AR deficiency (Fig. 3B). Hypothalamic insulin signaling was examined, and the ability of insulin to stimulate downstream AKT phosphorylation was impaired in the hypothalamus of NARKO mice (Fig. 3C). Moreover, NARKO mice exhibited reduced IR phosphorylation in response to insulin administration (Fig. 3D). Taken together, these results indicate impaired hypothalamic insulin signaling and hypothalamic insulin resistance in NARKO mice. Intracellular insulin signaling is executed by the tyrosine phosphorylation cascade, and the extent of tyrosyl phosphorylation is regulated by PTPs. While impaired hypothalamic insulin signaling in NARKO mice was observed, we also found an increase of PTP1B expression, a known physiological negative regulator of insulin signaling (Fig. 3E). These results suggest that elevated PTP1B levels may contribute to the development of hypothalamic insulin resistance in NARKO mice.

To address whether AR deficiency attenuates hypothalamic insulin signaling through increasing PTP1B-mediated IR dephosphorylation, we stably increased AR expression in murine hypothalamic GT1-7 neuron cells and validated AR expression by transactivation assays (Fig. 4A). Overexpression of AR was able to downregulate the expression of PTP1B protein, whereas the expression of PTP1B was slightly increased after knocking down AR expression (Fig. 4B). The tyrosine phosphorylation of IR was enhanced in AR-overexpressing cells when stimulated with 1 nmol/L insulin, and it was detected with as low as 0.1 nmol/L insulin in AR-overexpressing cells, compared with control cells (Fig. 4C). These data indicate that AR may modulate insulin signaling through regulating PTP1B expression. PTP1B mRNA expression was downregulated by activating AR with its ligand, dihydrotestosterone (DHT) (Fig. 4D). In contrast, PTP1B mRNA expression was upregulated when the AR expression was suppressed (Fig. 4E). To further address whether AR regulates PTP1B promoter activity, we established a luciferase reporter construct driven by the mouse PTP1B promoter (pGL3-PTP1B) and found that the promoter activity of PTP1B was repressed by AR (Fig. 4F). These data suggest that PTP1B expression was regulated by AR-mediated transcriptional suppression.

PTP1B expression is induced by inflammation in vivo through NF-κB–mediated transcriptional activation (22). We found that NF-κB induces PTP1B promoter activity in a dose-dependent manner in pBabe GT1-7 control cells, but not in pBabe-AR GT1-7 cells with AR overexpression (Fig. 5A). The promoter activity of PTP1B was induced when the AR was knocked down; whereas this induction was blocked by inhibiting NF-κB activity through overexpression of NF-κB inhibitor α (IκBα) (Fig. 5B). These results indicate that AR suppresses PTP1B expression by inhibiting NF-κB–mediated transcriptional induction, and that AR can suppress NF-κB activity in hypothalamic neurons. Suppression of the transactivation ability of NF-κB by AR in neuron cells was further observed using reporter constructs responsive to active NF-κB (Fig. 5C and Supplementary Fig. 2A). Activated hypothalamic NF-κB induced insulin and leptin resistance after HFD consumption was demonstrated to be a critical mediator of chronic overnutrition with energy imbalance and obesity (23). Results showing AR suppression of hypothalamic NF-κB activity suggest that loss of AR suppression may accelerate HFD-induced hypothalamic NF-κB activation.

To further test our hypothesis, we challenged NARKO mice with a short-term HFD feeding paradigm for 14 days and found that HFD significantly induced hypothalamic PTP1B expression in NARKO mice compared with controls where induction was moderate (Fig. 5D). The relatively increased hypothalamic NF-κB activation in NARKO mice after short-term HFD was shown by increased NF-κB in the nucleus (Fig. 5E and Supplementary Fig. 2B). These results indicate a heightened activity of hypothalamic NF-κB responding to HFD challenge in NARKO mice. In addition, an increase of SOCS3 mRNA expression in the hypothalamus of NARKO mice supported the increase of hypothalamic NF-κB activation (Supplementary Fig. 2C). The exhibited decrease of hypothalamic insulin receptor substrate 1 (IRS1) protein indicated potentially impaired hypothalamic insulin signaling in NARKO mice under HFD challenge (Supplementary Fig. 2D). The short-term HFD challenge hastened activation of hypothalamic NF-κB in young-adult NARKO mice at 10 weeks of age and promoted accelerated weight gain, subtle but significant, compared with controls (Fig. 5F). We did not observe a significant increase of food intake in NARKO mice during the short-term HFD feeding period (Fig. 5G). Enhanced NF-κB activation may favor a chronic inflammatory status. An increased inflammatory state was noted as more reactive astrogliosis in the hypothalamus and hippocampus of NARKO mice fed chow diets (Supplementary Fig. 2E). These results support the notion that decreased AR function may predispose the brain to acute and chronic inflammation via AR regulation of NF-κB activity.

Functional AR deficiency contributes to the development of insulin resistance in aging males. In humans, a dose-response relationship between testosterone levels and the odds of developing the metabolic syndrome is observed in men across different ages. Aging men with a 25% decrease of circulating testosterone levels tend to have a twofold increase of metabolic syndrome (24). In rodents, global ARKO male mice start developing insulin resistance at as early as 20 weeks of age (8). However, the mechanisms by which AR in individual organs coordinately regulates insulin sensitivity and contributes to insulin resistance remain largely unexplored. The current study has revealed a previously unrecognized role of neuronal AR in regulating systemic insulin sensitivity. The finding that neuronal AR deficiency is sufficient to reduce insulin sensitivity locally and systemically suggests that the brain neuron may be an initiator of the metabolic consequence resulting from disrupted AR function.

Insulin resistance is strongly associated with sustained inflammatory changes from challenges of nutritional excess (25). Diet-induced metabolic inflammation involves altered intracellular homeostasis, which may atypically trigger NF-κB–mediated inflammation in nonimmune cells to negatively affect intracellular insulin signaling (26,27). Given more susceptibility to intracellular stress, neurons may develop insulin resistance more rapidly than peripheral tissues. This concept is supported by recent findings that hypothalamic insulin resistance is an early event leading to hepatic insulin resistance rather than muscle or adipose tissue insulin resistance under short-term HFD challenge (28). The finding that AR suppresses NF-κB activation at the neuronal level highlights the importance of AR in prevention of the earliest stage of hypothalamic inflammation. The more reactive astrogliosis in the hypothalamus of NARKO mice on a chow diet before development of obesity further suggests that chronic hypothalamic inflammation is induced by functional AR deficiency. Hypothalamic inflammation induced by NF-κB activation can directly promote the expression of hypothalamic PTP1B, which negatively affects insulin signaling within cells (22).

The cross-talk between AR and NF-κB has been shown in PCa cells in vitro, where AR interrupts NF-κB signaling through ligand-dependent induction of IκBα (29). It is also reported that testosterone regulates the inflammatory response through the inhibition of the NF-κB–dependent expression of adhesion molecules and chemokines in human endothelial cells (30). Testosterone inhibition of NF-κB–dependent induction of vascular cell adhesion molecule-1 in human aortic endothelial cells depends on AR function. The suppression effect of testosterone is shown to be completely blocked by the concomitant administration of the AR blocker cyproterone acetate (31). Detailed molecular mechanisms of AR interference with NF-κB remain quite unclear, although induction of IκBα by AR to negatively regulate NF-κB activity was shown in PCa in vivo (32). In the same study, the author also reported DHT-dependent AR function to decrease the NF-κB immunoreactivity in PCa, which provides another mechanism of how AR interrupts NF-κB signaling. In addition, AR/NF-κB interactions or competitive binding to cis regulatory DNA elements, between AR and NF-κB, provide different mechanisms of AR interference with NF-κB signaling via ligand-independent or ligand-dependent AR function (33,34).

In the current study, we observed ligand-dependent and ligand-independent AR suppression of NF-κB activity on PTP1B promoter or on the synthetic NF-κB element reporter construct in neuron cells. Induction of IκBα mRNA expression via ligand-dependent AR activation was observed in AR-overexpressed GT1-7 neuron cells (data not shown). We speculated that the interaction between AR and NF-κB might promote stronger interference of NF-κB activation than endogenous AR in a ligand-independent manner in the AR-overexpressed cell culture system. Moreover, posttranslational modifications, such as phosphorylation or acetylation, leading to ligand-independent activation of AR to interrupt NF-κB activity, might contribute to the phenomena we observed. The ligand-independent activation of AR through posttranslational modifications is indeed commonly observed in metastatic PCa cells and contributes to the development of castration-resistant/metastatic PCa (35). Molecular mechanisms of the antagonistic modulation of NF-κB signaling by AR await future studies.

We propose that loss of functional AR in discrete hypothalamic neurons increases the susceptibility to nutrient-induced hypothalamic activation of NF-κB, leading to impairment of insulin action in the brain. Impaired insulin signaling influences the ability of hypothalamic neurons to regulate hepatic gluconeogenesis, through phosphoenolpyruvate carboxykinase. Elevated glucose production in the liver, in turn, leads to the deterioration of hepatic insulin sensitivity and results in insulin resistance. The ensuing insulin resistance induces the hepatic lipogenic regulator Srebp1c and promotes de novo lipogenesis. This sequence of events consequently raises triglyceride levels in the blood and increases the delivery of free fatty acids to adipose tissue, expanding the fat storage in adipocytes and worsening the insulin-resistant state. Moreover, progressively elevated nutrients and free fatty acids in the circulation can further promote hypothalamic NF-κB activation and lead to chronic hypothalamic inflammation (Fig. 6). Therefore, a vicious cycle is set up for disturbing glucose homeostasis and stimulating further insulin secretion from the pancreas. The net result is the classic triad of metabolic syndrome and type 2 diabetes, hyperglycemia, hyperinsulinemia, and hypertriglyceridemia. The consumption of excessive nutrients may accelerate this vicious cycle, as suggested by short-term HFD challenge in AR-deficient animals showing accelerated weight gain, although no obvious increase of food consumption was observed. More studies, including a longer period of HFD feeding, detailed energy expenditure and food intake examinations, and hypothalamic inflammatory pathway dissections, would be needed in the future to answer the remaining questions.

The significance of our findings to human biology closely relates to the metabolic syndrome association with declined testosterone levels via ADT treatment for PCa. Men undergoing ADT have a higher prevalence of metabolic syndrome, with more than half of the men (55%) in the ADT group developing the metabolic syndrome compared with 22% and 20% of men in the non-ADT and control groups, respectively (36,37). Current studies, although by artificially generated neuronal AR deletion in animal models, link the functions of neuronal AR to the hypothalamic insulin signaling of hepatic glucose production. We suggest that ADT treatment may dampen AR function in the hypothalamus at early stages to influence insulin sensitivity. The increase of serum insulin levels observed in short-term ADT, 1 month after the initiation of ADT, is likely to reflect increased insulin secretion to balance the elevation of hepatic glucose production from impaired hypothalamic insulin signaling.

Besides ADT treatment, it is clear that aging is associated with the decline in testosterone levels in men (38,39). Longitudinal observations demonstrate that late-onset hypogonadism is a factor in the etiology of metabolic syndrome in elderly men, which can be observed to begin at as early as 40 years of age (24,40,41). Evidence from interventional studies indicates a beneficial effect of testosterone supplementation on the improvement of metabolic manifestations in aging men (42–44). Although showing the benefits of testosterone therapy to ameliorate parameters of the metabolic syndrome, such studies must also take into account the adverse effects of AR overactivation in the prostate with systemic testosterone supplementation. Concerns about long-term effects on the prostate have limited the testosterone supplement as a therapy (45,46). Moreover, men with marked hyperglycemia and insulin resistance after ADT for PCa may not be suitable candidates for testosterone substitution. Correspondingly, selective AR modulators, synthetic ligands that bind to AR and display a tissue-selective activation of AR signaling, may provide a better choice for treatment (47–49). However, more studies are needed to determine whether selective AR modulators specifically targeting neuronal AR can induce meaningful outcomes.

In conclusion, the present findings demonstrate that loss of or decreased functional AR in neurons directly interferes with hypothalamic insulin signaling through enhancement of NF-κB activation. AR suppression of hypothalamic NF-κB activation provides the potential for tissue-selective treatments rather than global testosterone supplementation in PCa patients undergoing ADT. Although challenges still remain, pharmacologically targeting neuronal AR and hypothalamic NF-κB activation may represent a novel strategy to combat obesity and insulin resistance in aging men and patients receiving ADT.

This work was supported by the George Whipple Professorship Endowment, National Institutes of Health grants CA-155477 and CA-156700, and Department of Health Clinical Trial and Research Center of Excellence Grant DOH99-TD-B-111-004 to China Medical University.

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

I-C.Y. researched data and wrote the manuscript. H.-Y.L., N.-C.L., and L.-Y.F. researched data. J.D.S. and S.Y. contributed to discussion and reviewed and edited the manuscript. L.C. contributed to discussion. C.C. wrote the manuscript and contributed to discussion. C.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Dr. Margot Mayer-Pröschel, Dr. Michael Kerry O’Banion, and Dr. Lisa Opanashuk (University of Rochester Medical Center) for fruitful discussion, critical review, and suggestions. The authors acknowledge Karen Wolf (University of Rochester Medical Center) for manuscript proofreading and preparation.