Androgen receptor actions on AgRP neurons are not a major cause of reproductive and metabolic impairments in peripubertally androgenized mice

Excess levels of circulating androgens during prenatal or peripubertal development are an important cause of polycystic ovary syndrome (PCOS), with the brain being a key target. Approximately half of the women diagnosed with PCOS also experience metabolic syndrome; common features including obesity, insulin resistance and hyperinsulinemia. Although a large amount of clinical and preclinical evidence has confirmed this relationship between androgens and the reproductive and metabolic features of PCOS, the mechanisms by which androgens cause this dysregulation are unknown. Neuron‐specific androgen receptor knockout alleviates some PCOS‐like features in a peripubertal dihydrotestosterone (DHT) mouse model, but the specific neuronal populations mediating these effects are undefined. A candidate population is the agouti‐related peptide (AgRP)‐expressing neurons, which are important for both reproductive and metabolic function. We used a well‐characterised peripubertal androgenized mouse model and Cre‐loxP transgenics to investigate whether deleting androgen receptors specifically from AgRP neurons can alleviate the induced reproductive and metabolic dysregulation. Androgen receptors were co‐expressed in 66% of AgRP neurons in control mice, but only in <2% of AgRP neurons in knockout mice. The number of AgRP neurons was not altered by the treatments. Only 20% of androgen receptor knockout mice showed rescue of DHT‐induced androgen‐induced anovulation and acyclicity. Furthermore, androgen receptor knockout did not rescue metabolic dysfunction (body weight, adiposity or glucose and insulin tolerance). While we cannot rule out developmental compensation in our model, these results suggest peripubertal androgen excess does not markedly influence Agrp expression and does not dysregulate reproductive and metabolic function through direct actions of androgens onto AgRP neurons.

Polycystic ovary syndrome (PCOS) is the most common anovulatory disorder, characterised by elevated circulating androgen concentrations, polycystic appearance of the ovary and menstrual cycle irregularities.Two of these three hallmark features need to be present for a diagnosis of PCOS, in accordance with the widely accepted Rotterdam criteria. 1The impact of PCOS extends beyond reproductive function.[4][5] Hyperandrogenism has been identified as a critical factor in PCOS development, although the aetiology is still unclear.7][8] In animal models investigating androgen actions, dihydrotestosterone (DHT) is commonly administered rather than testosterone, due to its inability to be aromatised to oestrogens. 9Mice with chronic DHT exposure from postnatal day 21 display acyclicity, anovulation, increased body weight, adiposity and dyslipidemia. 7There is evidence for central mediation of androgen actions.Neuron-specific deletion of androgen receptor (ARKO) has been demonstrated to rescue postnatal DHTinduced reproductive and metabolic features, although not to the same extent as a global ARKO. 10 The specific neuronal population(s) which mediate the androgen-induced dysregulation is unknown.A candidate population is the agouti-related peptide (AgRP)-expressing neurons in the arcuate nucleus of the hypothalamus.Together with their cosecreted neuropeptide Y (NPY), they have a well-known role in regulation of metabolic function, promoting food-seeking behaviours and reducing energy-costly tasks.2][13] In mice, optogenetic and chemogenetic activation of AgRP neurons reduced pulsatile luteinizing hormone (LH) secretion. 14Since excessively rapid LH pulses have been reported in women with PCOS and in mouse PCOS models and are thought to drive high levels of ovarian androgen secretion, 15 suppressed AgRP neuronal activity could be responsible for this hypothalamo-pituitary gonadal (HPG) axis dysregulation.Furthermore, AgRP neurons have been reported to co-express androgen receptors in mice and ewes. 8,16In ewes, prenatal DHT or testosterone exposure increased AgRP expression in adulthood. 8Cotreatment of testosterone with flutamide (an androgen antagonist) partially blocked the increase in AgRP expression in ewes, suggestive that these androgenic actions are reversable. 8is study aimed to assess whether chronic DHT exposure from a prepubertal timepoint in mice impacts Agrp expression, and whether androgen actions in AgRP neurons are responsible for the metabolic and reproductive dysfunction evident in this model.

| Reproductive phenotyping
Estrous cyclicity in adult females was evaluated using vaginal cytology examined daily for 3-weeks from 13-to 16-weeks of age.
Uteri and ovaries were collected from perfusion-fixed animals at the end of the experimental period (16-weeks of age).Uteri were weighed.Ovaries were further post-fixed in 4% paraformaldehyde and stored in 70% ethanol until further processing.Ovaries were processed for wax embedding in paraffin wax blocks and sectioned at 5 μm thickness.Every 10th section of the ovary was collected (50 μm apart) and stained with haematoxylin and eosin.Ovarian images were captured using a light microscope (Olympus BX51) with a Â4 objective.All ovarian sections collected were used for the analysis of total corpora lutea number and preovulatory follicle number.Pre-ovulatory follicle were counted when a large antrum with well-defined outer margins, an oocyte surrounded by a corona radiata, and a well-defined theca cell layer were identifiable.Corpora lutea were only counted if they spanned across three ovary sections.

| Metabolic phenotyping
Mice were weighed weekly throughout the experiment, from 3 to 16 weeks of age, and a final body weight was recorded on the day of perfusion.
A glucose tolerance was performed at 14 weeks of age (11-weeks of DHT exposure).Mice received a subcutaneous bolus injection of glucose (1 g/kg body weight in 100 μL physiological saline) after an 8-h fast.
Blood glucose concentration measurements were taken at À15-, 0-, 15-, 30-, 45-, 60-, 90-, 120-and 180-min post-glucose injection from the tip of the tail vein using an automatic glucose monitor (i-sens Inc.; Care-SensN Premier).The area under the glucose response curve was calculated using the Prism software to determine glucose tolerance, where the average of the À15and 0-min samples were considered baseline.
Insulin tolerance at 15 weeks of age was assessed using the same protocol described for the glucose tolerance test, except mice were fasted for 3 h and injected with 0.25 U/kg body weight insulin (Sigma-Aldrich; no.15523; insulin from porcine pancreas) in 100 μL saline.
For the assessment of plasma insulin levels, at 14 weeks of age mice were fasted for 5 h and a whole blood sample ($50 μL) was collected from the tip of the tail vein into a heparinised tube.Samples were centrifuged to separate out the plasma from the red blood cells, plasma collected and stored at À20 C. A Crystal Chem ultra-sensitive mouse insulin ELISA kit (no.90080) was used for the analysis of plasma insulin concentrations in accordance with the manufacturer's instructions for the low range assay (0.1-6.4 ng/mL).Sensitivity of the assay was 0.16 ng/mL.
Abdominal visceral fat pads (parametrial and mesenteric) were weighed at the end of the experimental period (16 weeks of age, 13-weeks of DHT exposure).Parametrial fat attached to the processed ovary (processing described above, for the analysis of corpora lutea number) was analysed for average adipocyte area.From a 5-μm thick section stained with haematoxylin and eosin, images were captures using an Aperio ScanScope slide scanner at Â20 magnification.
The pixel classifier function on Qupath software was used to identify adipocytes in a randomly selected area containing between 50 and 400 cells.The average adipocyte area of individual adipocytes was determined per animal.

| Data analysis
Parametric tests were performed, based on the assumption that the data was normally distributed, had equal variances and the values were independent of each other.Two-tailed unpaired t-tests were used to test for significant differences between two groups.The statistical test used when there were two variables (genotype and DHT treatment) to account for was a two-way ANOVA.For body weight data, a repeated measures three-way ANOVA was used to compare the change in body weight over time between genotype and DHT treatment groups.If a statistically significant effect was seen in the ANOVA, Tukey's multiple comparisons test was used to determine the significant differences between groups.Results are presented as mean ± SEM, and a value of p < .05 was considered to be significant.
All analyses were performed using GraphPad Prism software (version 9.4.0).RNAscope in situ hybridisation data was not statistically analysed due to the low number of replicates.

| RESULTS
The  While there was no significant group difference in corpora lutea number between DHT CON and DHT AgRPARKO mice, the two DHT AgRPARKO mice that showed some rescued cyclicity also had a few corpora lutea present (indicated as red data points throughout graphs).These corpora lutea were smaller in appearance than the typical corpora lutea seen in blank implanted mice (Figure 3A).The number of preovulatory follicles in the ovary did not differ in response to DHT treatment or genotype (Figure 3C).

| Metabolic phenotype
Since the AgRP neurons are well-known to be involved in food intake and energy regulation, we wanted to assess whether direct actions of DHT on the androgen receptor in these neurons influences metabolic function in the DHT-treated model.Various metabolic parameters, (body weight, adiposity, glucose tolerance, insulin responsiveness and circulating insulin concentration) were assessed.
There was a significant main effect of DHT treatment on weekly body weight (F 1,34 = 32.38,p < .001).Both DHT-treated groups, irrespective of genotype, had significantly increased body weight from 5-weeks of age (2 weeks after implant exposure) that was maintained throughout the duration of the experiment compared with controls (Figure 4A).There was no significant genotype Â DHT treatment interaction or effect of genotype on weekly body weight trajectory.
There was a significant main effect of DHT treatment on adipocyte area (F 1,30 = 5.79, p = .023),where DHT implant mice had increased adipocyte area compared with blank-implant mice, but there was no effect of genotype or an interaction between variables (Figure 4B,C).Tukey's multiple comparison tests did not show significant differences between individual DHT groups compared with the respective blank implant groups.In contrast to what was observed for body weight and adipocyte area, there was no significant main effect of either DHT treatment or genotype on the percentage of visceral abdominal fat.However, there was a small, significant genotype Â DHT treatment interaction (F 1,30 = 4.53, p = .042),and post hoc analysis confirmed an increase in fat percentage in DHT AgRPARKO mice compared with DHT CON mice ( p = .049).This was not observed in blank mice (Figure 4D).
From the glucose tolerance test, there was no significant main effect of genotype or DHT treatment or an interaction between the two on the area under the curve in response to the glucose injection (Figure 5A,B).There was a significant effect of DHT treatment on fasting (8 h) blood glucose taken before the glucose tolerance test, with an overall increase in fasting blood glucose levels in DHT versus blank mice (F 1,34 = 6.80, p = .014),but no significant effect of genotype or genotype Â DHT treatment interaction and no significant post hoc (Tukey's multiple comparison test) differences.
Analysis of insulin tolerance revealed a statistically significant main effect of DHT treatment on the area under the curve in response to insulin injection (F 1,34 = 6.78, p = .014),where the DHT mice had impaired insulin sensitivity compared with blank mice (Figure 5C,D).No significant post hoc differences were observed.No effect of genotype or a significant genotype Â DHT treatment interaction was observed.Plasma insulin levels were measured at 11 weeks of age, when a significant increase in body weight was observed between blank and DHT mice, to determine if hyperinsulinemia was a contributor to metabolic dysregulation.A significant increase in plasma insulin levels was seen in DHT mice compared with blank mice (two-way ANOVA main effect of DHT treatment, F 1,34 = 6.37, p = .017).There was no significant effect of genotype on plasma insulin levels (blank CON: 0.27 ± 0.05 ng/mL, DHT CON: 0.41 ± 0.08 ng/mL, blank AgRPARKO: 0.31 ± 0.04 ng/mL, DHT, AgR-PARKO: 0.55 ± 0.1 ng/mL), and no significant genotype Â DHT treatment interaction (Figure 5E).
Prior to administration of the insulin bolus, fasting glucose levels were assessed after a mild 3 h fast.There was a significant main effect of DHT treatment (F 1,34 = 30.93,p < .001)but no effect of genotype or a genotype Â DHT treatment interaction on fasting blood glucose levels (Figure 5F).Post hoc analysis showed DHT mice had a significant increase in fasting blood glucose levels compared with blank mice in both CON and AgRPARKO genotypes (blank CON 8.6 ± 0.3 mmol/L vs. DHT CON 10.0 ± 0.3 mmol/L, p = .008;blank AgR-PARKO 8.0 ± 0.4 mmol/L vs. DHT AgRPARKO 9.9 ± 0.2 mmol/L, p < .001)(Figure 5F).shown that correcting AgRP neuronal overactivity is able to rescue diet-induced infertility 22 and pubertal delay. 23The aim of previously, 16 where only 40% of AgRP neurons were found colocalised with AR.The difference is likely to reflect the different androgenized mouse models used (prenatal vs. postnatal androgen treatments, with the latter achieving higher circulating DHT levels 7 and hence AR upregulation).It may also reflect differences in AR immunohistochemical detection (with different primary antibodies used).In the current study, the AR immunoreactivity was validated by the near complete absence of staining in AgRP neurons of AgRPARKO mice.Both experiments used the same Agrp-Cre mouse line to drive a fluorescent reporter for AgRP.
Most of the PCOS-like features previously reported in peripubertal DHT-implanted mice 7,10 were confirmed in this study, including increased body weight, elevated blood glucose levels, and anovulation.We also observed mild peripheral insulin resistance in DHT mice (not previously reported).Unexpectedly, we did not see elevated abdominal visceral adiposity in response to DHT treatment in control mice, despite the presence of significantly elevated body weight (although we did detect increased visceral adiposity that was specific to DHT AgRPARKO mice).Androgens can stimulate fatty acid, triacylglycerol production, and lipolysis. 24Others have reported an increase in fat mass as well as body weight gain in this model. 7,10,25,26It is possible that this increase in fat mass is partly accounted for by subcutaneous fat (not measured in our study), although visceral adipose tissue mass has been shown to correlate with the development of insulin resistance, while total or subcutaneous tissue mass has not. 27,28At the time of sample collection our mice were at least 2 weeks younger that those in previously-reported studies that showed increased fat mass, 7,10,25,26 so the lack of abdominal obesity in our DHT-treated controls may simply reflect insufficient time for this phenotype to development (although the reason for earlier-onset obesity in DHT AgRPARKO mice remains unknown).Increased body weight may be partly due to increased hypertrophy of the organs, as has been observed in diet-induced obese models in response to hypertriglyceridemia. 29 Furthermore, a more accurate fat measurement method such as magnetic resonance spectroscopy may have revealed a significant effect of DHT treatment.Notably, while there was no difference in adiposity as a percentage of body weight, there was an overall significant increase in the average adipocyte size in the parametrial fat in DHT versus blank-treated mice.This may explain insulin resistance in the absence of increased visceral body fat percentage.
Although AgRP expression throughout the brain and periphery is highly restricted, it has been shown to be expressed in the adrenal gland (e.g., Ref. [30]) where it is upregulated during fasting. 31This raises the possibility that deletion of androgen receptor from the adrenal AgRP-expressing cells could have influenced the metabolic outcomes in the present study.However this is unlikely, since recent immunohistochemistry and RT-PCR has clarified that AgRP is expressed in the medulla but not the cortex of mice, 32 while androgen receptors are expressed in the cortex. 33terestingly, 2 out of 10 DHT AgRPARKO mice exhibited some rescue of estrous cyclicity and the presence of corpora lutea, indicative of the occurrence of ovulation.The observed corpora lutea, however, were smaller than those observed in controls.The two partially cyclic DHT AgRPARKO animals tended toward the values of the blank-treated control animals for adiposity, fasted glucose levels and insulin responsiveness.The DHT implants of these cyclic mice were intact and looked the same as those from acyclic DHT AgRPARKO mice at the end of the experiment, so there is no evidence to suggest that the implants failed to induce were not hyperandrogenaemia.
Using sensitive liquid chromatography-tandem mass spectrometry measurements, the serum DHT concentration in female mice treated with identical implants to those in the current experiment have been reported as 1.6 mg/mL. 7Unfortunately, the sensitivity of commercially available DHT ELISAs exceeds this value by at least an order of magnitude, so we were unable to confirm the hyperandrogenaemic status of the mice in our study.Because the cyclic mice represented only 20% of the AgRPARKO group, these animals were analysed together with the 8 acyclic DHT AgRPARKO mice.Our reproductive results are similar to a recent report in which AR deletion from leptin receptor neurons was used to investigate androgen targets in a prenatally androgenized mouse model. 34In that study, leptin receptor cellspecific androgen receptor deletion, which would have included all AgRP neurons, 35,36 led to a partial but more pronounced reproductive rescue than was seen in the present study.Estrous cyclicity was  Of note, virtually all AgRP/NPY neurons co-express GABA, making up approximately a third of the arcuate GABAergic population. 40In a prenatally-androgenized PCOS-like mouse model, a rewiring of central circuitry has been shown, where GABAergic inputs are increased onto the GnRH neurons. 15In prenatally-androgenized mice the increased GABAergic innervation is hypothesised to drive the increased pulsatile LH secretion seen in this model, driving the reproductive dysregulation.
Furthermore, chronic activation of the arcuate nucleus population of GABA neurons leads to impaired reproductive parameters in female mice. 41The inhibitory role of AgRP and NPY on reproductive activity would suggest against their involvement in this androgen-induced upregulated GABAergic stimulation of LH, although it should be noted that leptin-resistant, diet-induced obese mice can sometimes exhibit high circulating LH levels and other PCOS-like features that appear to be due to AgRP over-activity, 22 and in an ovine prenatal androgen exposure PCOS model AgRP expression is increased. 8 a prenatal anti-Müllerian hormone exposure PCOS-like mouse model, it has recently been shown that kisspeptin neuron ARKO rescued reproductive dysregulation (this model exhibits a minimal metabolic phenotype). 42In a postnatal aromatase inhibitor (letrozole) PCOSlike mouse model, kisspeptin neuron ARKO improved estrous cyclicity, ovarian morphology, and insulin sensitivity (but not obesity) in comparison to controls. 43Kisspeptin neurons, therefore, are likely candidate populations upon which excess androgens act to cause infertility and insulin resistance.Notably, kisspeptin fibres come in close apposition to AgRP/NPY neurons, 44 so it remains possible that androgen dysregulation of kisspeptin neurons may indirectly affect AgRP neuron function.
Since arcuate kisspeptin neurons do not express leptin receptors during the prepubertal period, 45 they may not have been included in the aforementioned study in which AR deletion from leptin receptor neurons was used in a prenatally androgenized mouse PCOS model to investigate a broad subset of androgen targets. 34other potential neuronal population targeted by androgens is the arcuate proopiomelanocortin (POMC)-expressing neuronal group, which also regulate reproductive and metabolic function.The AR is expressed in $65% of POMC neurons in ewes, 8 although in the rat only 3% of POMC neurons were found to co-express AR. 46 They are also leptin-responsive. 35Partial loss of the MC4R function (a receptor which POMC neurons activate and AgRP neurons inhibit) results in an obese phenotype with reduced numbers of corpora lutea. 47Furthermore, female DHT-treated mice have reduced POMC expression. 48There are no studies reporting the effect of POMC AR deletion to date.
In summary, these results suggest that chronic androgen exposure induces impaired metabolic features (including insulin resistance, elevated plasma insulin, increased body weight and fasting blood glucose levels) that are not mediated through direct androgen actions in AgRP neurons.Androgen actions via AgRP neurons also appear to play a minimal role in inducing chronic androgen-induced acyclicity and anovulation.Thus, AR actions in AgRP neurons are not a major cause of PCOS-like symptoms in peripubertally androgenized mice.
These findings do not discount the role of AgRP neurons in PCOS pathology or of indirect androgen actions that involve these neurons, considering previously reported evidence of AgRP neuronal upregulation in prenatally androgenized ewes, but may reflect different androgenic actions in these two animal models.

2 | METHODS 2 . 1 |
Animals and ethical approval Transgenic Ar-flox, Agrp-Cre and tau-green fluorescent protein (τGFP) reporter mice on a predominantly C57BL/6 background strain were used in this study.Mice were group-housed in open top cages under a 12:12 h light/dark cycle at a constant temperature (21 C + 1 C) with ad libitum access to rodent chow (Envigo Teklad rodent diet 2918) and water, except for when fasted for the assessment of different metabolic parameters.All mice were handled daily to habituate them to the researcher and experimental procedures to reduce stress as a confounding variable in the results.
Ar lox/lox to generate experimental mice.Experimental mice genotypes used were: Agrp-Cre +/À , Ar lox/lox (Cre positive AgRP AR knockouts termed AgRPARKO), either expressing or not expressing τGFP in AgRP neurons, and Agrp-Cre À/À , Ar lox/lox (Cre negative controls termed CON) mice.An additional control group of Agrp-Cre: τGFP +/À , Ar wt/wt adult female mice (n = 5) was used for validation of AR expression in AgRP neurons in comparison to knockout animals, since the experimental CON animals do not express the τGFP reporter.The genotype of the transgenic mice was determined by analysis of genomic DNA (extracted from ear-punch biopsies) via polymerase chain reaction using the following primers for Ar-flox: AGC CTG TAT ACT CAG TTG GGG, AAT GCA TCA CAT TAA GTT GAT ACC (869 bp product indicating the wildtype gene and a 930 bp product indicating the Ar-floxed gene), for Agrp-Cre-recombinase: GCT TCT TCA ATG CCT TTT GC, GTG TGT GGT TCC AGC ATG AC, GG AAC TGC TTC CTT CAC GA (199 bp product indicating the wildtype gene and a 280 bp product indicating the Cre gene), for τGFP: CGA AGT CGC TCT GAG TTG TTA TC, GCA GAT GGA GCG GGA GAA AT, GCT CCT ATT GGC GTT ACT ATG (600 bp product indicating the wildtype gene and a 400 bp product indicating the τGFP gene).

2. 4 |
Effect of DHT treatment on number of hypothalamic Agrp-expressing cells using RNAscope in situ hybridization A subset of the main experimental cohort was processed for the analysis of Agrp expression in the arcuate nucleus (blank CON n = 3, blank AgRPARKO n = 2, DHT CON n = 2, DHT AgRPARKO n = 3).At 16-weeks of age mice were decapitated and brain tissue rapidly dissected, frozen on dry-ice and stored at À80 C. Brains were cryosectioned at 16 μm thickness in the coronal plane.Four sections containing the arcuate nucleus per mouse were mounted onto Superfrost Plus slides (Thermoscientific).The RNAscope 2.5 High-Definition Detection Assay-BROWN (Advanced Cell Diagnostics, catalogue number 322310) was used in accordance with the manufacturer's instructions for fresh frozen tissue.Briefly, sections were first fixed for 15 min in 4% paraformaldehyde in phosphatebuffered saline at 4 C, followed by dehydration in a series of graded ethanol solutions.Sections were pre-treated with RNAscope hydrogen peroxide solution for 10 min, followed by a 30-min treatment with RNAscope protease plus.Sections were then hybridised with the Agrp RNAscope probe (Catalogue number 400711-C2) for 2 h at 40 C.This was followed by six sequential probe amplification steps (30 min at 40 C; 15 min at 40 C; 30 min 40 C; 15 min 40 C; 30 min at room temperature and 15 min at room temperature).The signal was then developed by 3,3 0 -diaminobenzidine (DAB) for 3 min.Slides were counter stained with haematoxylin, dehydrated in graded ethanols and cover slipped with a xylene-based mounting medium.Sections were imaged on a light microscope (Olympus, BX51).Using Image J FIJI cell counting software, the number of cells stained with Agrp puncta were counted bilaterally in the medial arcuate nucleus (3-5 sections per animal were averaged to provide a single value for each mouse).
or proestrus.Tukey's multiple comparisons test confirmed that the individual DHT implant groups spent more time in met/diestrus and less time in proestrus or estrus than their respective blank implant groups ( p < .001)(Figure2A,B).Despite the absence of significant group differences in time spent in any of the cycle stages, of the 10 DHT AgRPARKO mice, 2 showed partial rescue of estrous cyclicity (Figure2A).Uterus weight (used as a proxy for circulating oestrogen F I G U R E 1 Legend on next page.concentrations) did not significantly differ in response to DHT treatment or genotype (Figure2C).There was a significant main effect of DHT treatment (F 1,34 = 105, p < .001)and genotype Â DHT treatment interaction (F 1,34 = 4.83, p = .035)on number of corpora lutea, but no main effect of genotype.Tukey's multiple comparison test confirmed the lower number of corpora lutea in individual DHT groups compared with the respective blank implant groups (Figure3A,B, p < .001).Collectively, blank CON and blank AgRPKO mice had 7.8 ± 0.7 corpora lutea per ovary while DHT CON and DHT AgRPARKO averaged 0.4 ± 0.3 corpora lutea per ovary.For the majority of DHT mice, no corpora lutea were observed over the extent of the ovary (Figure3B).

F I G U R E 1
Confirmation of AgRP neuronal androgen receptor knockout in AgRPARKO mice, and lack of effect of androgen receptor knockout on AgRP neuron number.(A) Representative images of immunoreactive androgen receptor cells (red) and GFP-labelled AgRP neurons (green) in the arcuate nucleus in control (upper panels) and AgRPARKO (lower panels) mice.White arrowheads indicate examples of AR-ir colocalised with AgRP neuron.Scale bars indicate 50 μm.(B) Percentage co-localization of AR in AgRP neurons, and (C) average number of AgRP neurons in 30 μm sections in the arcuate nucleus of control (n = 5) and AgRPARKO (n = 5) mice.(D) Average number of Agrp neurons (assessed by RNAscope in situ hybridization) in the arcuate nucleus per 16 μm coronal section from blank and DHT implant-treated mice, either control (CON) or with deletion of androgen receptors from AgRP neurons (KO) (n = 2-3).(E) Representative images of the sections quantified in D. Scale bars indicate 50 μm.*** p < .001,analysed using Student's t-test.AgRP neurons are a likely target for androgen-induced dysregulation of reproductive and metabolic function, based on their known expression of androgen receptors and involvement in regulating these processes.In this regard, recent experiments have

F I G U R E 2
Limited rescue of DHT-induced estrous cycles in AgRPARKO mice.(A) Representative estrous cycles over 3 weeks in adult blank and DHT implant-treated mice, either control (CON) or with deletion of androgen receptors from AgRP neurons (KO).The profiles within the dotted box show two mice with partial return of cyclicity in the DHT KO mice treatment group.(B) Percentage of time each treatment group spent in each estrous cycle stage over the 3-week period.(C) Uterine weight from perfusion-fixed animals at 16-weeks of age.Analysed using two-way ANOVA followed by Tukey's multiple comparison test, with main effects of genotype (G) and DHT treatment (DHT) shown above the graphs.*** p < .001.Red data points are from two DHT-treated AgRPARKO mice which had partial return of estrous cycles.the present experiment was to assess whether AgRP neurons are involved in the androgen excess-mediated development of PCOSlike traits.It was hypothesized that deletion of androgen receptors from AgRP neurons would ameliorate the metabolic and reproductive dysfunction induced in a peri-pubertal DHT-treated mouse model.Dual label immunofluorescent experiments identified that androgen receptors are expressed in most (<65%) AgRP neurons in the mouse.This is somewhat more than what has been observed Limited rescue of DHT-induced anovulation in AgRPARKO mice.(A) Representative ovarian sections from diestrus adult blank and DHT implant-treated mice, either control (CON) or with deletion of androgen receptors from AgRP neurons (KO).The dotted box shows ovarian sections from the two cyclic DHT KO mice.Corpora lutea are indicated by black asterisks and preovulatory follicles with black arrowheads.Scale bar = 500 μm.(B) Total number of corpora lutea and (C) preovulatory follicles per ovary.Analysed using two-way ANOVA followed by Tukey's multiple comparison test, with main effects of genotype (G) and DHT treatment (DHT) shown above the graphs.*** p < .001.Red data points are from two DHT-treated AgRPARKO mice which had partial return of estrous cycles.

F I G U R E 4
Body weight and adiposity of blank and DHT AgRPARKO mice.(A) Weekly body weights from 3 to 15 weeks of age of blank and DHT implanttreated mice, either control (CON) or with deletion of androgen receptors from AgRP neurons (AgRPARKO).(B) Average adipocyte area.(C) Representative images of parametrial adipocytes.Scale bar = 100 μm.(D) Abdominal visceral fat percentage of final body weight.Analysed using twoway ANOVA (with repeated measures for body weight) followed by Tukey's multiple comparison test, with main effects of genotype (G) and DHT treatment (DHT) shown above the graphs.* p < .05,*** p < .001.Red data points are from two DHT-treated AgRPARKO mice which had partial return of estrous cycles.improved, but not the anovulatory phenotype based on ovarian corpora lutea number.Taken together (and noting the important caveat of differing mouse models used), these studies suggest androgen target cells include leptin receptor-expressing cells, of which AgRP neurons appear to be a minimal component.AgRPARKO did not rescue the DHT-induced increase in body weight, insulin resistance or fasting glucose and insulin concentrations compared with control mice.From these results, direct androgen actions on AgRP neurons do not appear to be required for the development of the metabolic disturbances that result from chronic peripubertal DHT exposure, and at most they play a minor role in the development of reproductive dysfunction.The neuronal androgenic actions which mediate aspects of the DHT-induced reproductive and metabolic phenotypes 10 are therefore likely to be mediated primarily through a non-AgRP neuronal population.Cre-loxP breeding knockout approaches would be expected to result in the deletion of the desired gene from before birth, since Agrp mRNA expression develops extensively between embryonic days 12-20 in rats37 and is detectable at the time of birth in mice,38 although postnatal development is also evident.It is therefore possible that there are compensatory mechanisms that might mask a phenotype induced by AgRPARKO, especially in brain circuits that regulate functions as important as energy balance and reproduction.It may therefore be worthwhile in future studies to delete androgen receptors from AgRP neurons in adulthood, for example by using an inducible Agrp-Cre transgenic mouse line.39

F I G U R E 5
Glucose tolerance and insulin resistance of blank and DHT AgRPARKO mice.(A) Blood glucose profile in response to a 1 g/kg sc glucose challenge and (B) calculated area under the curve (AUC) at 14 weeks of age in of blank and DHT implant-treated mice, either control (CON) or with deletion of androgen receptors from AgRP neurons (KO).(C) Blood glucose profiles in response to a 0.25 U/kg sc insulin challenge and (D) AUC at 15 weeks of age.(E) Plasma insulin and (F) blood glucose concentration after a 3-h fast.Analysed using two-way ANOVA followed by Tukey's multiple comparison test, with main effects of genotype (G) and DHT treatment (DHT) shown above the graphs.* p < .05,** p < .01,*** p < .001.Red data points are from two DHT-treated AgRPARKO mice which had partial return of estrous cycles.It is unknown whether androgens act in mice to excite or silence AgRP neuron activity or to exert some other role, such as altered gene expression.In ewes, prenatal androgen exposure has been shown to upregulate the number of immunoreactive AgRP neurons.8In the initial assessment in this study, AR deletion from AgRP neurons did not alter the number of detectable AgRP neurons; although since the neurons were visualised using a Cre-driven reporter, their apparent number would not have reflected any post-developmental loss of Agrp expression.To evaluate whether there were any androgen-induced changes in mRNA levels, we analysed the number of Agrp-expressing neurons in the arcuate nucleus for DHT treatment and genotype effects, using RNAscope.Similarly to what was observed in the Cre-driven reporter expression of AgRP neurons, neither DHT treatment nor AgRPARKO appeared to affect the number of Agrp-expressing neurons compared with controls, although this conclusion is limited by the low number of replicates.Due to the high intensity of the Agrp mRNA staining within the neuron, the density of puncta of Agrp mRNA per cell could not be assessed.It therefore remains to be determined whether androgen exposure changes Agrp mRNA expression within individual cells.
Androgen receptor co-localised with AgRP neurons was quantified in the bilateral arcuate nucleus of two brain sections corresponding to the region between bregma coordinates À1.4 and À2.1 mm of control and AgRPARKO mice.The average number of stained cells per sections was a single data point for each animal.
7,10,21At the end of the experimental period (at 16-weeks of age) all implants were removed to confirm DHT powder was still contained within and that they had not ruptured or leaked.Experimental groups of the main cohort used to assess metabolic and reproductive function were as follows: blank CON n = 10, blank AgRPARKO n = 9, DHT CON n = 9, and DHT AgRPARKO n = 10.All experimental work involving animals was approved by the University of Otago Animal Ethics Committee.