Agouti‐related peptide neuronal silencing overcomes delayed puberty in neonatally underfed male mice

Abstract Agouti‐related peptide (AgRP) neurons are thought to indirectly regulate the activity of hypothalamic gonadotrophin‐releasing hormone neurons which control fertility. AgRP neurons also drive caloric intake and are modulated by metabolically‐relevant hormones, providing a link to the hypothalamic–pituitary‐gonadal axis. In mice expressing Cre‐dependant designer receptors (DREADDs) in AgRP neurons, we activated or silenced these neurons in vivo using the synthetic ligand clozapine‐N‐oxide (CNO) to observe the effect of AgRP neuron activity on timing of puberty. To validate these animals, we chronically treated both stimulatory (hM3Dq) and inhibitory (hM4Di) DREADD × AgRP‐Cre mice with CNO, observing a pronounced increase and decrease of food intake, respectively, consistent with the known orexigenic effects of these neurons. RNAscope was performed to visually confirm the activation of AgRP neurons. Puberty onset was assessed in males and females. There was no effect on preputial separation in males or vaginal opening and first oestrus in females after CNO treatment from day 26 to 30 to chronically modulate AgRP neurons. Next, to determine whether the delay in puberty onset occurring in response to neonatal underfeeding could be overcome by inhibiting AgRP neuronal activity, mice were raised in large (neonatally underfed) or normal litter sizes. The delay in puberty from underfeeding was completely reversed in CNO‐treated AgRP‐hM4Di male mice. These data highlight the inhibitory role of AgRP neurons to delay puberty onset when undernutrition occurs during the neonatal period, at least in male mice. Trail registration number JNE‐22‐0081‐OA.R2


| INTRODUCTION
Reproduction is an energy demanding function and thus, nutritional status and reproductive function are very closely associated. During times of low food abundance, animals must choose to allocate increased time to foraging at the expense of reproductive activity.
Metabolic status and fertility are affected by many neuronal populations, one of which is the arcuate population of agouti-related peptide (AgRP) neurons. These neurons drive food intake in response to situations of caloric need, and are modulated by metabolically-relevant hormones such as leptin and insulin. 1 They also provide a link to the hypothalamic-pituitary-gonadal (HPG) axis, possibly through functional synaptic contact with GnRH neurons [2][3][4] and/or with arcuate and anteroventral kisspeptin neurons. 5,6 It is likely that a complex network exists between AgRP neurons and other neurons of the GnRH neuronal network to coordinate reproductive activity.
AgRP neurons, located in the arcuate nucleus of the hypothalamus, coexpress the highly potent orexigenic peptides AgRP and neuropeptide Y (NPY), as well as gamma-aminobutyric acid (GABA). [7][8][9] AgRP is an inverse agonist of the melanocortin 3 and 4 receptors (MC3R and MC4R), inhibiting the anorexigenic effects of the melanocortin system neuropeptide α-melanocyte-stimulating hormone (MSH) and stimulating caloric intake. 7,9,10 Intracerebroventricular administration of AgRP leads to an increase in bodyweight and food intake. 11,12 In leptin-deficient, infertile male and female ob/ob mice, ablation of AgRP neurons lead to a restoration of fertility. 13 Moreover, AgRP neuron-specific leptin receptor knockout mice exhibited delayed onset of oestrous cyclicity, indicating that at least part of the actions of leptin occur directly on AgRP neurons. Remarkably, rescue of leptin receptors only in AgRP neurons is sufficient to rescue the disturbances seen in leptin receptor knockout mice. 14 This provided evidence that the increased AgRP neuron signalling that results from the lack of inhibitory leptin actions leads to the suppression of the HPG axis, and that leptin signalling directly in AgRP neurons is sufficient and partially required for normal reproductive function.
Despite these findings regarding the role of leptin in AgRP neuronal actions, little is known about the role of AgRP neurons per se in reproductive function, particularly pubertal timing. In order to study this, we used chemogenetic designer receptors exclusively activated by designer drugs (DREADD) technology to inhibit or stimulate AgRP neuronal activity over relatively long time periods. After validating the AgRP-Cre Â Cre-dependant DREADD mouse models, we examined the effect of AgRP neuronal modulation on puberty onset. We then attempted to overcome the delay in puberty onset that is known to occur in response to neonatal underfeeding 15 by inhibiting AgRP neuronal activity.

| Mouse care
C57BL/6J mice were group housed (3)(4)(5) per cage) at a University of Otago animal facility under standard laboratory conditions (12 h on/off; lights on at 0630 h) and temperature-controlled environment (22 ± 1 C). Experiments performed during the animal's dark phase were completed under red light to ensure that the circadian rhythm of these animals was not disturbed. Unless otherwise specified, food (standard rodent chow) and water were provided ad libitum. All experimental protocols were approved by the University of Otago Animal Ethics Committee.

| Mouse model
DREADDs are mutated muscarinic acetylcholine G protein-coupled receptors (GPCR) that are reported to respond specifically to nanomolar concentrations of clozapine-N-oxide (CNO), an otherwise pharmacologically inactive compound. When expressed only in specific cell types, this chemogenetic technology enables precise activation and inhibition of these neuronal populations noninvasively in vivo.
We utilized two different types of Cre-dependant DREADD mice.
The first of these was the stimulating hM3Dq-flox mouse line (Jackson Laboratories, # 026220) that enables control of Gq-mediated signalling, and the second was the inhibiting hM4Di-flox mouse line (Jackson Laboratories, # 026219) that enables control of Gi-mediated signalling. 16 LoxP sites flank a stop codon upstream to the hM3Dq and hM4Di gene sequences. To specifically target the AgRP population, both mouse strains were individually crossed with an AgRP-Cre mouse (Jackson Laboratories, # 012899), where Cre recombinase is exclusively expressed in cells encoding the AgRP gene. 17 When crossed, the sequence between the LoxP sites (i.e., the stop-codon) is excised by Cre-recombinase, allowing the hM3Dq or hM4Di receptors to be expressed exclusively in AgRP-expressing cells.
Control mice received the same CNO treatment in each experimental protocol as the AgRP-hM3Dq or AgRP-hM4Di mice in order to control for any possible off-target effects of CNO.

| Treatments
For chronic manipulation of AgRP neurons, CNO was dissolved in 0.2% DMSO/water at 0.01 mg/ml concentration. The mice were administered CNO (approximatively 5 mg/kg/day) through their drinking water, assuming that a mouse drinks 5-10 ml per day. 18 Water was changed daily to ensure that the drug was fresh, and the liquid was protected from the light.
For acute manipulation of AgRP neurons, CNO was dissolved in 0.5% DMSO/saline at 0.1 mg/ml concentration. Mice were administered CNO (1 mg/kg) via a subcutaneous injection.

| Bodyweight and food intake measurements
AgRP neurons are orexigenic, therefore, to validate both the AgRP-hM3Dq and AgRP-hM4Di mice, as well as the dose of CNO used, bodyweight was measured in the presence of chronic CNO administration and food intake was measured in the presence of acute CNO administration. A significant increase in these measures in the AgRP-hM3Dq mice and a significant decrease in AgRP-hM4Di mice would confirm AgRP neuronal activation and inhibition, respectively, indicating that the activity of these neurons was indeed being altered as expected in response to the dose of CNO used. 6,19,20 To this end, AgRP-hM3Dq (n = 5), AgRP-hM4Di (n = 14), and control male mice

| Glucose and insulin tolerance tests
To assess glucose homeostasis, a glucose tolerance test (GTT) was performed in adult AgRP-hM3Dq (n = 7), AgRP-hM4Di (n = 6) and control (n = 9) male mice. Mice were fasted for 8 h and received an acute subcutaneous CNO injection (1 mg/kg) 30 min before the GTT test. Blood glucose levels were measured in tail vein blood using a glucose meter (CareSensN, model GM505PAD) at baseline, 15, 30, 60, and 90 min after intraperitoneal challenge of glucose (1 g/kg).

| Experiment 1: Effect of chronic AgRP stimulation and inhibition on puberty onset
AgRP-hM3Dq (n = 7), AgRP-hM4Di (n = 6), and control (n = 9) male mice, and AgRP-hM3Dq (n = 6), AgRP-hM4Di (n = 7), and control (n = 10) female mice, from litter size ranges 4-10, were chronically treated with CNO ($5 mg/kg/day) via their drinking water from PND 26 to 30. Timing of puberty was determined using daily visual assessment of preputial separation and vaginal opening as the anatomical indications of puberty onset in male and female mice, respectively. Once vaginal opening had occurred, the age at first oestrus was determined by vaginal cytology analysis as an additional marker of puberty in females.
2.9 | Experiment 2: Effect of chronic AgRP neuron inhibition on puberty onset in neonatally underfed and neonatally normally fed pups We were interested in determining whether chronic AgRP inhibition was sufficient to counteract the delay in puberty onset observed under neonatal underfeeding conditions. 15 To test this, AgRP-Cre mice were crossed with Cre-dependant hM4Di-flox mice and litter sizes were manipulated on PND 2 by randomly distributing pups among mothers such that large litters (neonatally underfed) had 12 pups and normal litters (neonatally normally fed) had six pups. All mice were weaned at PND 21 and given ad libitum access to food and water.
Female pups from large litters were used if their bodyweight was under 12 g at PND 25, and females from normal litters were used if their bodyweight was over 12 g at this age. For males, the inclusion criteria was below or above 13 g for large and normal litters, respectively.
No more than two pups per group were excluded by these criteria. In experiment 2, a three-way ANOVA was used for repeated measures followed by the Tukey's multiple comparison test (applied to bodyweight data in Figures 4A and 5A, where time, genotype and neonatal nutritional condition were the factors). Then, to analyse the interaction between the genotype and the neonatal nutritional condition, a two-way ANOVA was used, followed by the Holm-Šídák multiple comparison post-hoc test (applied to data from Figure 4B-D and 5B-E).
Nonparametric data (RNAscope experiment, age at preputial separation, at vaginal opening and at first oestrus, area under the curves for GTT, and area over the curves for ITT) was analysed using Kruskal-Wallis test followed by the Dunn's multiple comparison post-hoc test.
The assumptions to perform a Kruskal-Wallis test were that the data distribution did not fit a Gaussian distribution or have equal variances, or the data were essentially ordinal (i.e., puberty age).

| Effect of acute AgRP neuron stimulation and inhibition on glucose levels
To investigate the effect of AgRP modulation on glucose levels, GTT and ITT were performed in AgRP-hM3Dq, AgRP-hM4Di and control mice after CNO administration. There was no significant difference in the increase of glucose levels after glucose challenge between control, AgRP-hM3Dq and AgRP-hM4Di mice ( Figure 1D). AgRP neuronal modulation did not affect overall blood glucose responses during the GTT as assessed by the areas under the curves, which were not different between groups ( Figure 1E). significantly higher in AgRP-hM4Di mice compared to control mice ( Figure 1G).
Blood glucose levels were unaffected by CNO administration in the sample collected prior to glucose or insulin treatments (data not shown).

| Immunohistochemistry and RNAscope
The Cre-dependant DREADDs hM3Dq and hM4Di have an HA tag that gets turned on when the DREADD is expressed. Immunohistochemistry for HA was used as a marker of hM3Dq and hM4Di expres-

| Experiment 1: Effect of chronic AgRP neuron stimulation and inhibition on puberty onset
CNO was chronically administered from PND 26 to 30. The age at preputial separation and at vaginal opening and first oestrus were used as an anatomical indicator of puberty onset in male mice ( Figure 3A) and female mice ( Figure 3B,C), respectively.
No significant difference was observed in the age at preputial separation in CNO-treated AgRP-hM3Dq and AgRP-hM4Di mice compared to controls, and no significant differences were observed in the age at vaginal opening and at first oestrus in CNO-treated AgRP-hM3Dq and AgRP-hM4Di mice compared to controls.

| Experiment 2: Effect of chronic AgRP inhibition on puberty onset in neonatally underfed pups compared to neonatally normally fed pups
We next wanted to determine whether the delay in puberty onset observed in neonatally underfed mice 15 could be overcome by AgRP neuron inhibition.

| Males
There were a significant age effect (F There was a significant genotype effect on age at preputial separation (F (1,28) = 5.57, p < .05). The age at preputial separation was delayed by 3 days in neonatally underfed control mice compared to neonatally normally fed control mice (p < .01) ( Figure 4C). This effect was completely reversed in neonatally underfed AgRP-hM4Di mice compared to neonatally underfed controls, with a similar age at preputial separation observed in neonatally underfed AgRP-hM4Di and both neonatally normally fed groups, despite the markedly compromised bodyweight caused by the additive effects of neonatal underfeeding and AgRP neuronal silencing ( Figure 4D). The bodyweight at which preputial separation was observed was significantly lower in CNO-treated neonatally underfed AgRP-hM4Di mice compared to neonatally underfed control mice ( Figure 4D).

| Females
There were a significant age effect (F   Figure 5E).

| DISCUSSION
The neuropeptides secreted by AgRP/NPY neurons are well-known to be highly orexigenic 8 and critically involved in the drive to forage for food at the expense of other activities. 22 Their involvement in reproductive processes is less well understood. AgRP neurons have been shown to make contact with GnRH neurons in lactating rats. 2 AgRP (as well as NPY) can influence GnRH neuronal activity. Roa et al. reported primarily inhibitory, but also some excitatory effects of these two neuropeptides. 4 Recently, viral addition of activating (hM3Dq) DREADDs into AgRP neurons was used to demonstrate the ability of these neurons to impair estrous cycles and also the time taken to conceive after mating in female mice. 6 Based on brain slice recordings from GnRH and kisspeptin neurons, the authors suggested that it is the latter cell type that is the direct target of AgRP neurons.
The present study is the first to use Cre-dependant silencing DREADDs to investigate the effects of manipulating AgRP neuronal activity on puberty onset. Our results show that inhibition of AgRP neuronal activity can overcome the adverse effects of nutritional restriction on puberty onset in males.
AgRP is well known to have a strong orexigenic effects, driving food intake behaviour. Chronic CNO administration for 4 days led to a significant increase in bodyweight in AgRP-stimulated adult mice, in accordance with previous observations. 6,19,20 A dramatic increase in cumulative food intake was observed in AgRP-hM3Dq mice compared to control adult mice, from 30 min after acute CNO administration, performed at the beginning of the light phase when the food intake is the least. 21 The increased orexigenic behaviour of these animals is in accordance with the significant increase in food intake observed after an intracerebroventricular injection of AgRP, 12 and with previous evaluation of CNO treatment on AgRP neuron-stimulated animals. 6,20 Conversely, no significant decrease in bodyweight was observed in mice that were AgRP-silenced, during CNO treatment, compared to control adult mice. The absence of bodyweight suppression in AgRP neurons had no effect on glucose tolerance but decreased insulin sensitivity in mice, 29,30 although others showed a significant effect of chemogenetic AgRP neuronal activation and no effect of optogenetic activation on glucose tolerance. 29 The latter chemogenetic effect may reflect the fact that those authors used double the glucose dose and fasting period than we did in the current experiment. Our rationale for treating with CNO from PND 26 rather than an earlier age was to avoid adaptation or compensatory mechanisms, as well as to lessen the likelihood of obtaining an effect on puberty that was secondary to altered bodyweight caused by an extended period of treatment.
Considering the increase of Agrp mRNA expression during undernutrition, 8 we decided to inhibit AgRP neurons to attempt to rescue the delay in puberty onset observed during underfeeding conditions. 15 We used cross-fostering to alter the nutritional status of pups. This led to the expected differences in bodyweight between the neonatally underfed and neonatally normally fed mice in both sexes.
Before CNO treatment, at day 26, neonatally underfed mice exhibited a $20%-30% lower bodyweight compared to neonatally normally fed mice, in both sexes. AgRP neuronal silencing from day 26 to 30 exacerbated this effect of neonatally underfeeding male mice, but had no effect on bodyweight of neonatally normally-fed mice. This is consistent with the idea that the effect of AgRP silencing is most evident when the activity of these neurons is naturally high (i.e., in underfed animals). By contrast, the inability of AgRP silencing to cause a bodyweight reduction in neonatally normally-fed male mice that were approaching their adult body size (see Figure 4), may simply reflect the greater drive to eat of growing animals. No catch-up growth occurred during the D26-D35 period in male mice that were neonatally underfed, which is consistent with other studies. 33 However, female mice that were neonatally underfed exhibited catch-up growth during the D26-D35 period. This sexual dimorphism in catch-up growth capacity has already been shown in rats. 34 In this study, AgRP neuronal silencing consistently overcame the puberty-delaying effect of neonatal underfeeding in males, but not in females. Little is known about sex differences in how AgRP neurons modulate puberty timing. In both males and females, ablation of AgRP neurons restores fertility in leptin-deficient mice 13 and rescue of leptin receptors only in AgRP neurons enables puberty onset and fertility in leptin receptor knockout mice of both sexes. 14 However, sexually dimorphic kisspeptin neurons, rather than GnRH neurons, may be the primary direct target of AgRP neurons, 6 and recent experiments have demonstrated male-specific susceptibility of puberty onset to metabolically-relevant kisspeptin neuronal manipulation. 37 On the other hand, the significant effects observed in males but not in females might simply be due to the timing of CNO administration relative to puberty timing. The PND 26-30 CNO treatment period may have been optimal for males but not for females with regard to the relative ages at which the HPG axis matures and become responsive to AgRP neuronal inputs.
The chemogenetically silenced AgRP neurons presumably attenuated their secretion of the neuropeptides AgRP and NPY, as well as the neurotransmitter GABA. It is not possible to determine from this study which of these neurochemicals is the primary effector for withholding puberty onset. Evidence from other studies shows that GABA tially rescues the infertility phenotype of leptin signalling-deficient mice. 5,13,25,40 In the present study, we confirmed a delay of puberty onset when undernutrition occurs during neonatal period, which would serve to withhold the capacity to reproduce until energy reserves and perhaps environmental conditions are more favourable. We showed that inhibition of AgRP neuronal activity was able to overcome the adverse effects of nutritional restriction on puberty onset in males. These results suggest an inhibitory role for AgRP neurons, acting as a brake on prepubertal HPG activity at least in male mice.

AUTHOR CONTRIBUTIONS
George ADP Connolly: Formal analysis; investigation; methodology.