Low-Dose Dihydrotestosterone Drives Metabolic Dysfunction via Cytosolic and Nuclear Hepatic Androgen Receptor Mechanisms

Abstract

Androgen excess in women is associated with metabolic dysfunction (e.g., obesity, hyperinsulinemia, insulin resistance, and increased risk of type 2 diabetes) and reproductive dysfunction (e.g., polycystic ovaries, amenorrhea, dysregulated gonadotropin release, and infertility). We sought to identify the effects of androgen excess on glucose metabolic dysfunction and the specific mechanisms of action by which androgens are inducing pathology. We developed a mouse model that displayed pathophysiological serum androgen levels with normal body mass/composition to ensure that the phenotypes were directly from androgens and not an indirect consequence of obesity. We performed reproductive tests, metabolic tests, and hormonal assays. Livers were isolated and examined via molecular, biochemical, and histological analysis. Additionally, a low-dose dihydrotestosterone (DHT) cell model using H2.35 mouse hepatocytes was developed to study androgen effects on hepatic insulin signaling. DHT mice demonstrated impaired estrous cyclicity; few corpora lutea in the ovaries; glucose, insulin, and pyruvate intolerance; and lowered hepatic insulin action. Mechanistically, DHT increased hepatic androgen-receptor binding to phosphoinositide-3-kinase (PI3K)-p85, resulting in dissociation of PI3K-p85 from PI3K-p110, leading to reduced PI3K activity and decreased p-AKT and, thus, lowered insulin action. DHT increased gluconeogenesis via direct transcriptional regulation of gluconeogenic enzymes and coactivators. The hepatocyte model recapitulated the in vivo findings. The DHT-induced hepatocyte insulin resistance was reversed by the androgen-receptor antagonist, flutamide. These findings present a phenotype (i.e., impaired glucose tolerance and disrupted glucose metabolism) in a lean hyperandrogenemia model (low-dose DHT) and data to support 2 molecular mechanisms that help drive androgen-induced impaired glucose metabolism.

Hyperandrogenemia (HA) is associated with impaired reproductive and metabolic function (1). Androgen excess in women is a public health burden (2). Animal models of HA have been developed in rodents, sheep, and primates by introducing androgens to prenatal, postnatal, and/or peripubertal animals (3–7). However, these models were not able to decipher the endocrine and metabolic features of HA from obesity. Of importance, the suprapathophysiological levels of androgens (six- to eightfold that of controls) used to induce infertility or metabolic dysfunction in some models (3, 8, 9) may not represent the pathophysiological serum androgen levels (testosterone or dihydrotestosterone [DHT]), which are only 1.7- to threefold higher in women with polycystic ovarian syndrome (PCOS) (10) or in corrected congenital adrenal hyperplasia (11) compared with that of unaffected women. These suprapathophysiological levels of androgen did not result in hyperinsulinemia (3, 9) and, thus, may lead to the destruction of pancreatic beta cells (12).

Serum androgen levels, timing of androgen exposure, and the dose of androgens in the animal models are extremely important in models attempting to mimic the pathophysiological state associated with HA. We developed a low-dose DHT mouse model displaying normal body mass/composition and pathophysiological serum androgen levels (twofold that of controls) to identify the target tissues in which androgens are inducing metabolic dysfunction and to ascertain the specific pathogenic mechanisms. To focus on androgen receptor (AR)-mediated action, DHT is an attractive androgen because it is not able to be aromatized to estrogen and its effects, therefore, are specific to AR. This is a suitable mouse model for lean HA-associated dysfunction.

The phenotypic effects of androgens on insulin action and glucose metabolism in females have been characterized and studied in animal models and humans, and it has been documented that androgen induced impaired glucose tolerance and insulin resistance (3, 4, 13–15). However, the molecular mechanisms by which androgens modulate insulin action and glucose metabolism remain largely unknown. Many studies have evaluated insulin action in HA in adipocytes (13, 15) and skeletal muscles (14, 16). Conversely, less is known about insulin action in the liver of women with HA. Interestingly, in a diet-induced, obese female mouse model of infertility, insulin resistance in the energy storage tissues was observed in association with HA (5, 17). However, the role of high androgen levels in the progression of metabolic pathologies is not well understood. Here, we evaluated the effects of HA at pathophysiological doses on hepatic insulin action.

Several prostate cancer studies have shown that AR interacts with phosphoinositide-3-kinase (PI3K) (18–20) and genes and proteins involved in the transcription of gluconeogenic genes, including forkhead box O1 (Foxo1) (21) and cAMP response element binding protein (Creb). Insulin promotes the recruitment and activation of PI3K. PI3K-p85 acts as an adaptor and is required for PI3K-p110 catalytic activity (22). PI3K is a heterodimer containing a regulatory subunit, p85 (α and β isoforms), and a catalytic subunit, p110 (α, β, and δ isoforms) (23). Insulin stimulates PI3K activity via phosphotyrosine-mediated recruitment of the heterodimer to the plasma membrane. p85 contains a Src homology 2 (SH-2) motif that interacts with p-Y residues of IRS1/2. This interaction brings the PI3K heterodimer (p85-p110) to the plasma membrane, activating its catalytic activity. p85α makes up ~80% of PI3K regulatory subunits expressed in insulin-sensitive tissues (24). The remaining 20% is mostly p85β, with <5% being composed of p50α and p55α (splice variants of p85α) (24).

FOXO1 is a transcriptional activator of gluconeogenic enzyme gene expression, primarily phosphoenolpyruvate carboxykinase (PEPCK) and glucose-6-phosphatase (G6Pase). FOXO1 is inactivated, shuttled from the cell nucleus to the cytoplasm, and targeted for ubiquitin-mediated degradation by insulin-stimulated PI3K-mediated AKT phosphorylation at serine 256 (S256), thus downregulating gluconeogenesis (25). CREB enters the nucleus when phosphorylated at S133, and increases expression of Pepck and G6p1. Hepatic glucose production (HGP) is upregulated in insulin resistance and is fueled molecularly via increased glycogenolysis and gluconeogenic enzyme activity (26). The role of AR interactions in glucose metabolic pathways has not been explored.

We demonstrated that mice receiving low-dose DHT, when compared with vehicle-treated mice, displayed anovulation and whole-body metabolic dysfunction. Of greater significance, we discovered 2 novel mechanisms by which androgens induce metabolic pathogenicity. We found that DHT-induced hepatic AR directly binds PI3K-p85, dissociating it from the catalytic PI3K-p110, and that hepatic AR directly binds to Foxo1 and CREB promoter regions, upregulating expression of these gluconeogenic enzymes and increasing hepatic glucose production.

Research Design and Methods

Generation of hyperandrogenic mouse model

Female C57BL/6 mice were sustained with normal chow and water ad libitum under a 14-h/10-h light/dark cycle. Dow Corning Silastic tubing (0.04 mm inner diameter × 0.085 mm outer diameter; Fisher Scientific, Hampton, NH) was filled with DHT to lengths of 2 mm, 4 mm, 5 mm, and 10 mm, then sealed with 2 mm of medical adhesive silicone (Factor II, Lakeside, AZ) on each side. Pellets were incubated in saline for 24 hours at 37°C for equilibration before insertion. DHT levels were measured weekly for 30 days by enzyme-linked immunosorbent assay (ELISA) (Alpha Diagnostics International, San Antonio, TX) (27) from sera of mice inserted with different lengths of DHT pellets. The sensitivity of the ELISA kit was 6 pg/μL. The 4-mm DHT pellets contained 2.0 mg of DHT. The pellets were removed and replaced with new pellets every month. DHT levels of empty and 4-mm pellets were additionally measured by mass spectrometry (courtesy of Dr. Brian Keevil, University Hospital of South Manchester, United Kingdom) (28). Other groups have used pellets containing 2.5 mg of DHT from Innovative Research of America (Sarasota, FL) (3). The company states that their 90-day continuous release pellets will release the hormone content in the specified range, thus 2.5 mg divided by 90 days equals 27.5 μg/d. One group measured DHT levels after 90 days and found a sixfold increase compared with controls (3). We designed our own DHT pellets and tested the release empirically. All experimental procedures and protocols were in accordance with and approved by the Johns Hopkins Animal Care and Use Committee.

Metabolic testing: glucose, insulin, and pyruvate tolerance; and glucose-stimulated insulin secretions

Mice that were fasted overnight (16 hours) received intraperitoneal (IP) injections of glucose (2 g/kg body weight [BW]) for a glucose tolerance test, those that fasted for 7 hours received IP injections of insulin (0.3 units/kg BW) for an insulin tolerance test (Lilly, Indianapolis, IN), and those that fasted for 12 hours received IP injections of pyruvate (2 g/kg BW) for a pyruvate tolerance test (Sigma, St. Louis, MO). Tail blood was obtained to determine blood glucose at time points between 0 and 120 minutes using a One Touch Ultra glucometer (Life Scan, Milpitas, CA). For the glucose-stimulated insulin secretions, blood samples were obtained at time 0 and at 30 minutes after 2 g/kg BW glucose IP injection to 16-h fasted mice and sera were separated by centrifugation. Glucose-stimulated insulin secretion levels were measured using Milliplex Map Mouse Serum Metabolic Hormone Panel (catalog no. MMHMAG-44K; Millipore, Billerica, MA) on a Luminex 200IS platform (Luminex, Austin, TX). IP injections, as opposed to oral administration, were used to avoid changes due to manipulation of incretins in the gut (29). Fasting times were used according to standards in the field (30). At 16 hours of fasting, hepatic glycogen content is nearly depleted, eliminating basal blood glucose variability (29).

Hormone and phosphoprotein measurements

Serum obtained weekly from 7-hour–fasted mice was analyzed for insulin, leptin, adiponectin, interleukin (IL)-6 and tumor necrosis factor-α using Milliplex Map Mouse Serum Adipokine Panel (catalog no. MADKMAG-71K; Millipore) on a Luminex 200IS platform. Serum from 16-hour–fasted mice was analyzed for cholesterol, triglycerides, and aspartate transaminase-to-alanine transaminase ratio via the Johns Hopkins University Pathology Core Facility. At 3 months postinsertion, the livers of fed and16-hour fasted mice were harvested 10 minutes after IP injection of 0.5 U/kg BW of insulin, homogenized in BioRad Cell Lysis buffer (BioRad Laboratories, Hercules, CA) and analyzed to determine phosphor-AKT and total AKT protein levels using Milliplex Map 2-Plex Total/Phospho AKT Kit (Millipore) on a Luminex 200IS platform.

Real-time quantitative polymerase chain reaction

Liver tissues from fed and fasted (control and DHT) mice were harvested and RNA was isolated using Trizol (BioRad). The RNA was reverse transcribed via an iScript cDNA synthesis kit (BioRad) and real-time quantitative polymerase chain reaction (qRT-PCR) was performed using iQ SYBR Green reagent (BioRad) and an iCycler iQ5 Q-PCR machine (BioRad). The primers used are listed in Supplemental Table 1.

Western blot

Tissues from fed and 16-hour fasted mice were harvested and equal amounts of protein were separated via sodium dodecyl sulfate polyacrylamide gel electrophoresis (Thermo Scientific, Waltham, MA) and transferred to nitrocellulose membranes. The membranes were blocked, incubated with the following primary antibodies: p-AKT, AKT, p-FOXO1 S256, FOXO1, and actin (Cell Signaling Technology, Danvers, MA); and AR (N20), PI3K p85α, PI3K p110, p-PI3K p110 Y485, p-PI3K p85α Y467, PEPCK-C (cytosolic), G6Pase-β, p-CREB1 S133, and CREB1 (Santa Cruz Biotechnology, Dallas, TX), all 1:1000 dilutions (Table 1). Then the blots were incubated in secondary antibodies (goat anti-mouse or goat anti-rabbit; BioRad), and detected using enhanced chemiluminescence (Perkin Elmer Life Sciences, Boston, MA). Densitometry was quantified using My Image Analysis Software (Thermo Scientific). Fasting times were used for reasons explained previously in this section.

Immunoprecipitation and PI3K activity assays

Liver tissue lysates were incubated with AR (Santa Cruz Biotechnology), p85, phosphotyrosine (pY), IRS1, or IRS2 antibodies (Cell Signaling Technology) overnight at 4°C, and then protein G sepharose beads (Sigma) were added to the mixture and incubated for 4 hours at 4°C. Samples were immunoprecipitated as done previously (31). Immunoprecipitates from pY, IRS1, IRS2 (Cell Signaling Technology), and AR (Santa Cruz Biotechnology) were subjected to a PI3K activity assay (catalog no. 17-493; Millipore), as detailed in the manufacturer’s protocol manual.

Chromatin immunoprecipitation

Chromatin immunoprecipitation (ChIP) was performed using ChIP-IT Express kit as indicated by the manufacturer (Active Motif, Carlsbad, CA). Briefly, liver tissue lysates were cross-linked with formaldehyde, digested via enzyme and sonication, immunoprecipitated using AR antibodies, treated with proteinase K, and DNA was isolated and then analyzed via qRT-PCR. The primers are listed in Supplemental Table 1.

Hepatocyte culture and DHT/flutamide assay

Low-dose DHT mice had a serum DHT concentration of ~300 pg/mL, or 1.02 nM [Fig. 1(C)], which was twofold that of control mice. Thus, we used 1 nM DHT in cell cultures. The H2.35 female mouse hepatocyte line was used. Cells were cultured as previously described (32). Cells were transfected using Lipofectamine 2000 (Thermo Fisher) with a pBabe-Puro vector containing a full-length, fully functional AR (33) (provided by Dr. Shuyuan Yeh, University of Rochester, Rochester, NY) or a pBabe-Puro empty vector (Addgene, Cambridge, MA). After 4 hours of transfection, growth medium was replaced with medium with or without 10 nM flutamide for 2 hours, as previously done (34). The cells were then incubated in fresh media with or without 1 nM DHT for 24 hours. The cells were serum starved for the final 3 hours of DHT treatment. For the final half hour of DHT treatment, the cells were treated with or without 100 nM insulin (32, 35). Cells were harvested using BioRad Cell Lysis buffer and processed for western, qRT-PCR, or ChIP (36) analysis. The primers used for ChIP were PEPCK promoter (37) and G6Pase promoter (38). The primer sequences are listed in Table 1.

Statistical analysis

Data were analyzed by unpaired t tests or 1-way analysis of variance using Prism software (GraphPad Software, Inc, La Jolla, CA). All results are expressed as mean ± standard error of the mean. A value of P < 0.05 was defined as statistically significant.

Results

Serum DHT levels and experimental design

Mice in which 4-mm DHT pellets were inserted displayed a twofold increase in serum DHT levels compared with control mice [Fig. 1(A) and 1(B)]. Serum DHT levels were similar between 3 and 30 days after DHT insertion (data not shown). DHT values, as measured by mass spectrometry, were similar to those obtained by ELISA [Fig. 1(B)]. Therefore, at 8 weeks of age, lean female mice had either a 4 mm-DHT pellet (DHT mice) or an empty pellet (control mice) inserted; then they were divided into 3 groups. We chose to insert DHT at 2 months to mimic adult-age HA. Over the 3 months after implantation, reproductive tests (group 1), serum hormonal assays from blood samples (group 2), and metabolic tests (group 3) were performed. Finally, the mice were killed and tissues were collected for further molecular, biochemical, and histological analyses [Fig. 1(C); Supplemental Fig. 1].

DHT mice exhibited obesity-independent impaired glucose and pyruvate tolerance and impaired insulin sensitivity

HA in females has been associated with dysfunctional glucose regulation, hyperinsulinemia, and insulin resistance (9). Female mice with DHT exhibited impaired glucose tolerance (IGT), impaired insulin sensitivity (IIS), and impaired pyruvate metabolism, shown graphically and as an area under the curve in Fig. 2(A–C). The increased glucose production following pyruvate challenge suggests an increase in hepatic gluconeogenic capacity. DHT mice exhibited an increase in basal (hyperinsulinemia) and glucose-stimulated insulin levels compared with control mice [Fig. 2(D) and 2(E)]. There was no change in body mass or composition [Fig. 2(F) and 2(G)] in DHT mice compared with control mice. Additionally, hormones and metabolites commonly associated with insulin resistance and obesity (i.e., leptin, IL-6, tumor necrosis factor-α, serum triglycerides, cholesterol, or aspartate transaminase-to-alanine transaminase levels) were unaltered (Supplemental Fig. 2).

Livers of DHT mice displayed lowered proinflammatory cytokine IL-6 mRNA expression [Supplemental Fig. 3(A)] but similar Mac-2 staining compared with control mice (data not shown). Although hyperandrogenic women with PCOS have a higher propensity for developing hepatic steatosis than nonhyperandrogenic women without PCOS (39), DHT mice displayed similar levels of lipids in hepatic tissues compared with control mice [Supplemental Fig. 3(B)]. These data suggest that DHT-induced metabolic dysfunction is independent of obesity and inflammation.

Low-dose DHT reduced hepatic insulin action

Attenuation of insulin-stimulated phosphorylation of AKT (p-AKT) in energy-storage tissues is a hallmark of insulin resistance. Whereas insulin increased p-AKT in the livers of control mice compared with noninsulin-treated control mice, the insulin-stimulated increase in p-AKT was blunted in DHT mice relative to noninsulin-treated DHT mice [Fig. 3(A) and 3(B)], indicating that low-dose DHT can lower hepatic insulin action. Insulin-stimulated pY-IRS1 and IRS2 levels were unaltered in DHT mice compared with controls [Fig. 3(A), 3(C), and 3(D)]. Total IRS1 and IRS2 levels were similarly unchanged [Fig. 3(A), 3(E), and 3(F)]. Interestingly, pY immunoprecipitates and IRS1 immunoprecipitates from DHT mice displayed decreased insulin-stimulated PI3K activity [Fig. 3(G)], indicating dysfunction at the PI3K level of insulin signaling. Also of note, p-JNK levels were unaltered in control and DHT mice at no insulin and in the presence of insulin, indicating no key role for JNK-negative regulation (25, 40) of insulin signaling in this model [Fig. 3(H) and 3(I)].

AR directly binds the regulatory subunit of PI3K, causing a reduction in the p85-p110 interaction and lowered PI3K activity

We next sought to determine the mechanism by which DHT reduced hepatic PI3K activity. AR has been shown to directly interact with PI3K (18, 19). Thus, we sought to determine if changes in AR interaction with PI3K were observed in our model. DHT mice displayed increased association of AR to p85 compared with control mice, and insulin had no effect on this interaction [Fig. 3(J) and 3(K)]. Additionally, DHT lowered the interaction of p85 with p110 at baseline and in the presence of insulin [Fig. 3(J) and 3(L)]. Several studies have shown that phosphorylation of p85 or p110 negatively regulated PI3K activity (41–43). Interestingly, insulin decreased tyrosine phosphorylation (p-Y) of both p110δ Y485 and p85α Y467. However, DHT caused an increase in p-p110δ Y485 and p-p85α Y467 in the liver of control and insulin-stimulated mice [Fig. 3(J), 3(M), and 3(N)].

Effects of low-dose DHT on gluconeogenic mRNA and protein level

IGT is associated with decreased glucose uptake (in skeletal muscle and white adipose tissue) and increased HGP via gluconeogenesis. Gluconeogenic mRNA levels (i.e., Pck1, G6P, Foxo1, HnF1a, and Creb) and gluconeogenic rate-limiting enzyme levels (i.e., PEPCK and G6Pase) were increased in DHT mice compared with control mice [Fig. 4(A–C)]. FOXO1 is a gluconeogenic coactivator repressed by insulin-stimulated AKT phosphorylation (44). DHT mice displayed increased hepatic FOXO1 protein levels and decreased p-FOXO1 S256 [Fig. 4(D–F)]. Insulin alone decreased FOXO1, but in combination with DHT, insulin no longer had its effect on FOXO1, indicating decreased insulin action [Fig. 4(D–F)].

CREB is a gluconeogenic transcription factor activated by PKA phosphorylation. DHT increased p-CREB S133 and total CREB levels compared with controls [Fig. 4(D), 4(G), and 4(H)]. Furthermore, DHT mice displayed decreased AR binding to Foxo1 gene promoter and increased binding to Creb promoter [Fig. 4(I]), suggesting that DHT-induced AR serves as a regulator of Foxo1 and an activator of Creb.

AR inhibition reverses DHT-induced hepatocyte insulin resistance

To determine if the DHT effect was specific to AR, female mouse hepatocytes (H2.35) were transfected with an AR-overexpression vector and treated with or without DHT and flutamide, a competitive AR antagonist. A comparable low dose (1 nM) of DHT was used. As in the low-dose DHT mouse model, DHT resulted in attenuated insulin-stimulated p-AKT (S473) and p-FOXO1 (S256) [Fig. 5(A–C)], indicating hepatocyte insulin resistance. This decreased insulin action was accompanied by an increased AR-p85 interaction [Fig. 5(A), 5(D), and 5(E)] and increased p-p85 (Y467) levels [Fig. 5(A) and 5(F)], as was seen in our mouse model. Interestingly, flutamide treatment blocked all the effects of DHT, thus reversing the DHT-induced decreased insulin action [Fig. 5(H–N)]. PI3K (p85α and p110β) has been shown to display nuclear localization (45). However, nuclear protein extracts of H2.35 cells were immunoprecipitated with AR and displayed no AR/p85 interaction (data not shown), indicating that AR interaction with PI3K is cytosolic.

Furthermore, in the hepatocyte model, DHT increased gluconeogenic mRNA levels (i.e., Foxo1, Creb, HnF1a, and G6P), decreased AR binding to the Foxo1 promoter, and increased AR binding to the Creb promoter compared with control [Fig. 5(O) and 5(P)]. These DHT-induced effects were prevented by flutamide. Interestingly, flutamide alone increased AR binding to the Foxo1 promoter but had no effect on Foxo1 mRNA levels [Fig. 5(O) and 5(P)]. Additionally, DHT increased Creb and FOXO1 binding to the G6P promoter and these effects were prevented by flutamide [Fig. 5(Q)]. Pck1 mRNA expression was not detected in H2.35 cells. These flutamide studies suggest that AR mediated the DHT-induced hepatic insulin resistance.

Discussion

The focus of this project was, first, to establish that low-dose DHT (twofold that of control) recapitulated metabolic dysfunction (Fig. 2) associated with HA. These are important findings in that other DHT models have used suprapathophysiological levels of DHT that do not result in hyperinsulinemia (3, 46), perhaps because of β-cell destruction. Most importantly, we sought to explore the underlying mechanism for the observed metabolic pathogenesis (Figs. 3–5).

We show that low-dose DHT in female mice recapitulated many features of lean HA-induced metabolic and reproductive dysfunction in humans. Low-dose DHT induced impaired whole-body glucose metabolism (i.e., IGT, IIS, impaired gluconeogenic capacity, and hyperinsulinemia). The liver is the main site of gluconeogenesis. Thus, the impaired gluconeogenic capacity led us to focus on the effects of DHT on hepatic glucose metabolism. Low-dose DHT resulted in hepatic insulin resistance and increased gluconeogenic protein levels. Most interestingly, we discovered a pathogenic mechanism of androgen-induced, AR-mediated hepatic insulin resistance [Fig. 6(A)]. Reproductive endocrine dysfunction (i.e., acyclicity, decreased corpora lutea, and increased atretic follicles) was documented (Supplemental Fig. 4) but was not the focus of this study.

Many PCOS animal models exist and several are discussed later in this article [Fig. 6(B)] (3–5, 7, 9,, 47). These models have benefits and have advanced the field but, as with most replicative models, including ours, they each have minor drawbacks. Androgenized sheep and primate models (48, 49); prenatal androgen exposure in rodents (50, 51); postnatal injections of dehydroepiandrosterone (a metabolic intermediate of androgens and estrogens) in rodents (6, 52); injections of letrozole, a breast cancer drug that lowers androgen-to-estrogen conversion by inhibition of aromatase, thus increasing androgen levels in rodents (4, 9); mice lacking insulin receptors and leptin receptors in proopiomelanocortin neurons (7); and monosodium l-glutamate administration in rats have all been used as models of PCOS to address certain aspects of the condition (53). Rodents receiving DHT, a nonaromatizable androgen, displayed reproductive and metabolic abnormalities associated with PCOS (3, 4, 9). The serum DHT levels in these studies were six- to eightfold higher in tested animals than in control animals, and may disrupt pancreas function (54). Our low-dose DHT model displayed many of the metabolic and reproductive features of PCOS independent of obesity. The differences observed in our DHT model compared with others (primarily hyperinsulinemia and the absence of obesity) may be due to the lower dose but could also be due to the timing of androgen exposure (i.e., postpubertal as opposed to neonatal or prepubertal) (55).

This study looks at DHT-induced AR interactions with PI3K. Free p85 (α or β) competes with heterodimer p85-p110 for binding to IRS1/2. Single knockout of p85α or β reduced the free p85 docking to and competitive inhibition of IRS1/2, thus increasing heterodimer p85-p110 docking to IRS1/2 and increasing PI3K activity via a stoichiometry compensatory effect (24, 56–58). In the p85α/p85β double knockout model, there is no compensation by the other p85 subunit, thus docking to IRS1/2 is lowered, leading to decreased PI3K activity (59). Interestingly, tyrosine phosphorylation of p110 (41) and p85 (42) have been shown to decrease PI3K activity via decreased p110-p85 dimerization (41) or increased free p85β, which competes for IRS docking (42).

A similar mechanism could be working in our model. Our findings suggest that DHT-induced cytosolic AR inhibits PI3K activity via (1) interacting with p85, (2) dissociating p85 from p110, and (3) indirectly increasing tyrosine phosphorylation of the subunits, leading to reduced PI3K activity and lowered p-AKT. Disrupted PI3K activation may be one of the pathogenic mechanisms by which DHT induces hepatic insulin resistance.

Our data do not mean that other pathways are not also regulated or involved. Free p85 has been reported to downregulate insulin signaling via p-JNK (57). However, in our DHT model, p-JNK levels were not altered [Fig. 3(H) and 3(I)]. Extracellular signal-regulated kinase (Erk1/2) has been reported as a negative regulator of insulin signaling (60, 61). Thus, Erk1/2 signaling may be involved in DHT-induced hepatic insulin resistance; however, our focus was on the AR interaction with the PI3K-AKT pathway.

AKT is the hub for insulin-mediated metabolic functions. AKT phosphorylates FOXO1, resulting in cytosolic sequestration and degradation, leading to reduced gluconeogenesis. Nonphosphorylated FOXO1 enters the nucleus and, in conjunction with other coactivators, functions to increase gluconeogenesis via increased expression of G6p and Pck1. A comprehensive study of AR targets in a prostate cancer cell line, using expression profiling and ChIP, identified 524 AR binding regions in 205 primary androgen-responsive genes (ARGs) (62). ARGs are genes regulated directly by AR-occupied androgen responsive elements AREs (15-bp partially palindromic androgen responsive elements) (63). Foxo1, Creb, and Pik3r2 (p85β) were among the 205 ARGs discovered. We used the published primers from this study for our ChIP analysis. Here, we show that AR binds to the promoter of, and regulates, Foxo1 and Creb gene transcription but not Pik3r2 [Fig. 4(I)]. According to the aforementioned AR binding-site study, Pik3r2 was the only class I PI3K gene with an AR binding site. We did not discretely explore the possibility of AR-binding sites on other PI3K genes. Pik3ca (p110α) or Pik3cb (p110β) mRNA levels were not altered by DHT; however, DHT lowered Pik3cd (p110δ) mRNA levels (data not shown) in the liver compared with controls, which indicates it may have other effects on PI3K transcription.

Under DHT stimulation, AR binds the Creb promoter region and leads to increased Creb mRNA and protein levels. We show that this effect is reversed by flutamide treatment, indicating AR specificity. Thus, DHT-induced AR binding to the Creb promoter serves as a positive regulator of Creb transcription. DHT induces dismissal of AR binding to the Foxo1 promoter in conjunction with increased Foxo1 mRNA and protein levels. This suggests that AR binds to the Foxo1 promoter and inhibits Foxo1 expression in a ligand-independent manner. AR knockdown would be needed to address this question.

The gluconeogenic enzymes G6Pase and PEPCK do not undergo posttranslational modifications and are solely regulated by their complex transcriptional machinery (44). We show that low-dose DHT in vivo upregulates G6Pase and PEPCK via AR direct binding to and regulation of Foxo1 and Creb, thus increasing HGP. These findings are validated in vitro and, furthermore, flutamide reversed all the DHT-induced effects.

In summary, our data describe a new animal model that mimics the reproductive and metabolic dysfunctions of women with HA: anovulation, IGT, IIS, hyperinsulinemia, and insulin resistance. The metabolic and reproductive dysfunction was obesity independent. Mechanistically, we demonstrated that DHT-induced AR cytosolic binding to p85, and AR nuclear binding to promoter regions of Foxo1 and Creb resulted in hepatic insulin resistance and increased hepatic gluconeogenesis.

Abbreviations:

 
• AR

  androgen receptor

 
• ARG

  androgen-responsive gene

 
• BW

  body weight

 
• ChIP

  chromatin immunoprecipitation

 
• CREB

  cAMP response element binding protein

 
• DHT

  dihydrotestosterone

 
• ELISA

  enzyme-linked immunosorbent assay

 
• ERK1/2

  extracellular signal-regulated kinase

 
• FOXO1

  forkhead box O1

 
• G6Pase

  glucose-6-phosphatase

 
• HA

  hyperandrogenemia

 
• HGP

  hepatic glucose production

 
• IGT

  impaired glucose tolerance

 
• IIS

  impaired insulin sensitivity

 
• IL

  interleukin

 
• IP

  intraperitoneal

 
• PCOS

  polycystic ovarian syndrome

 
• PEPCK

  phosphoenolpyruvate carboxykinase

 
• PI3K

  phosphoinositide-3-kinase

 
• S256

  serine 256

Acknowledgments

We thank Drs. Sally Radovick and Fred Wondisford for their guidance.

This work was supported by the National Institutes of Health (Grants R00-HD068130 to S.W. and 5T32DK007751-18 to S.A.). This work was also supported by the Baltimore Diabetes Research Center: Pilots and Feasibility Grant (to S.W.). Technical support was provided by the Integrated Physiology Core of the Baltimore DRTC (P60DK079637).

Author Contributions: S.A. and S.W. contributed to the conceptual design, performance of experiments, interpretation and analysis of data, and writing and editing the manuscript. All other authors contributed to performing some of the experiments, analyzing the corresponding data, and reviewing and editing the manuscript.

Disclosure Summary: The authors have nothing to disclose.

References

Author notes