Arcuate AgRP, but not POMC neurons, modulate paraventricular CRF synthesis and release in response to fasting

The activation of the hypothalamic–pituitary–adrenal (HPA) axis is essential for metabolic adaptation in response to fasting. However, the neurocircuitry connecting changes in the peripheral energy stores to the activity of hypothalamic paraventricular corticotrophin-releasing factor (CRFPVN) neurons, the master controller of the HPA axis activity, is not completely understood. Our main goal was to determine if hypothalamic arcuate nucleus (ARC) POMC and AgRP neurons can communicate fasting-induced changes in peripheral energy stores, associated to a fall in plasma leptin levels, to CRFPVN neurons to modulate the HPA axis activity in mice. We observed increased plasma corticosterone levels associate with increased CRFPVN mRNA expression and increased CRFPVN neuronal activity in 36 h fasted mice. These responses were associated with a fall in plasma leptin levels and changes in the mRNA expression of Agrp and Pomc in the ARC. Fasting-induced decrease in plasma leptin partially modulated these responses through a change in the activity of ARC neurons. The chemogenetic activation of POMCARC by DREADDs did not affect fasting-induced activation of the HPA axis. DREADDs inhibition of AgRPARC neurons reduced the content of CRFPVN and increased its accumulation in the median eminence but had no effect on corticosterone secretion induced by fasting. Our data indicate that AgRPARC neurons are part of the neurocircuitry involved in the coupling of PVNCRF activity to changes in peripheral energy stores induced by prolonged fasting.

obesity, metabolic syndrome, type 1 and 2 diabetes, and HIV-related lipodystrophy [5][6][7]. Despite the differences in the pathogenesis of these conditions, it has been suggested that changes in peripheral energy stores or the inability of the organism to sense them, underlie the changes in the HPA axis activity [8]. Reinforcing the coupling of the HPA axis with changes in peripheral energy stores, it was demonstrated that leptin administration reduces corticosterone in hypoleptinemic type 1 diabetic animals [9][10][11] and restores fasting-increased plasma corticosterone levels in mice [9]. The central nervous system (CNS) is considered the critical site for leptin's actions on food consumption and energy expenditure [12]. Leptin receptor (LepR) is not expressed in CRF PVN neurons [13] but it is highly expressed in brain sites sensitive to changes in peripheral energy stores, such as the hypothalamic arcuate nucleus (ARC), where LepR colocalizes with a subset of proopiomelanocortin (POMC) or agouti-related protein (AgRP) neurons (POMC ARC or Agrp ARC neurons, respectively) [14]. In the ARC, POMC is predominantly cleavaged to alpha-melanocyte stimulating hormone (α-MSH) that negatively modulates the energy homeostasis [10]. On the other hand, the activation of AgRP ARC neurons, which also produce neuropeptide Y (NPY) and GABA, increases food intake and reduces energy expenditure [10]. Additionally, pharmacological studies have demonstrated that α-MSH, AgRP and NPY signaling can modulate the HPA axis activity [15,16]. The ARC neurons are both directly and indirectly connected to the PVN [17,18]. However, in addition to the ARC, these neuropeptides/neurotransmitters are also expressed in other brain nuclei [19][20][21].
Therefore, in the present study we aimed to determine if POMC ARC and/or Agrp ARC neurons can communicate fasting-induced changes in peripheral energy stores to CRF PVN neurons.

Fasting activates the HPA axis
Initially, the effects of 24 h and 36 h fasting were compared. Both fasting periods reduced the body weight, plasma leptin and blood glucose levels, and increased plasma corticorterone concentration; however, only 36 h fasting increased mRNA expression of Crf in the PVN and Agrp and Npy in the ARC, as compared to 24 h fasting (Additional file 2: Fig. S2). Therefore, the additional studies were performed in 36 h-fasted animals.

Discussion
In the present study we demonstrated the effects of food deprivation on the HPA axis activity and the involvment of AgRP ARC neurons in the gene expression/activation of CRF PVN neurons, the major HPA axis controller.
It's well know that fasting activates HPA axis [9]. Here we demonstrated that the increased plasma corticosterone levels observed in fasted mice was associated to increased mRNA expression and activity of CRF PVN . The increased plasma glucocorticoid levels in fasted animals and humans are essential to the energy substrate utilization (fatty acids instead of carbohydrates) and to the activation of hepatic gluconeogenesis to avoid hypoglycemia [22,23]. Interestingly, it has recently been demonstrated that hyperphagia caused by various energy-depleted states, such as food deprivation, hypoglycemia, or diabetes, is also dependent on increased plasma corticosterone levels and its actions on arcuate neurons controlling food consumption [24].
The reduction in body weight and the consequent decrease in plasma glucose and leptin levels in mice fasted for 36 h can contribute to the changes in the HPA axis activity [9][10][11]. Indeed, fasting-induced hypoglycemia can recruit the counter-regulatory response and consequently increase glucagon, epinephrine and corticosterone secretion [25,26]. Additionally, the decrease in the syhthesis of leptin from the white adipose depots [27] leading to a reduction of plasma leptin levels is required for metabolic adaptations to reduce energy availability and activation of the counter-regulatory response [25].
Regarding the mRNA expression in the ARC, we observed increased mRNA expression of Agrp, Npy and Gad65/67 in the ARC, and decreased Pomc mRNA expression in mice fasted for 36 h. These data reinforce the role and the responsivity of neuropeptides/neurotransmitters expressed in ARC neurons in the control of energy homeostasis [7,8,10]. Together, our data indicate that a prolonged fasting period changes the mRNA expression and activity of CRF PVN neurons in response to changes in ARC mRNA expression/activity of neuropeptides/neurotransmitters projecting to PVN.
Leptin can reduce or increase the HPA axis activity and its response it seems to be associated with the amount of energy available, i.e. in fasted animals the most prevalent response observed in scientific studies is a decrease in plasma corticosterone levels after leptin administration [9,11,28]. Furthermore, even during food deprivation, this effect of leptin appears to be dependent on the level of the reduction of energy storage, as increased HPA activity was reported in studies using a prolonged period of fasting (e.g. 48 h) [9,24] but not in 24 h fasted animals [24,29]. In our study, we observed increased body weight loss and plasma corticosterone levels in animals fasted for 24 h, but higher body weight loss and plasma corticosterone levels after 36 h of fasting (Additional file 2: Fig. S2). The reduced mRNA expression of CRF PVN and neuronal activity of CRF PVN neurons are in accordance with the inhibitory role of leptin on the HPA axis activity [9,28].
The effects of leptin on the CRF gene expression and CRF PVN neuronal activity can not be mediated by direct leptin signaling, since LepR is not expressed in CRF PVN neurons [13]. Additionally, ARC neurons have their gene expression (Fig. 2D) and activity modulated by leptin [12,14] and, as we previously cited, icv or even site-specific administration of the neuropeptides/neurotransmiters produced by ARC modulates the HPA axis activity [30,31]. Therefore, the changes in the ARC mRNA could account for the reduction in the HPA axis activity after leptin administration. In addition, the decrease of plasma corticosterone levels in fasted mice under leptin treatment can contribute to the reduction of plasma glucose levels. However, leptin may also affect other key glucoregulatory mechanisms (i.e., sympathetic nervous system activity) [22,[32][33][34] to modulate plasma glucose levels. Together, our data reinforce the inhibitory role of leptin in the control of CRF PVN . Furthermore, a direct leptin signaling on adrenal gland [35] and/or indirect actions via ARC neurons could also be involved in the reduction of plasma corticosterone levels after leptin treatment to fasted animals.
POMC ARC neurons project to the PVN [17] and are modulated by fasting [36] and leptin [14]. Thus, we investigated the consequences of the chemogenetic activation of POMC ARC on plasma corticosterone levels in 36 h-fasted mice. Despite the effective chemogenetic activation of POMC ARC neurons in fasted animals expressing the excitatory DREADD, we observed no effect on PVN neuronal activity and on the increased plasma corticosterone levels induced by food deprivation. POMC ARC neurons are heterogenous regarding their responsivity to hormones such as leptin [37] and particularly regarding to the type of the neurotransmitter they express (e.g., AgRP, NPY and GABA) [38,39]. Therefore, the heterogeneity of these neurons must be considered before definitely excluding their participation in the control of the HPA axis activity in fasting conditions. A study using the same lineage of animals reported that POMC neurons projects to the adrenal gland, and the deletion of LepR in this neuronal population does not modify the activation of SNS induced by icv leptin [40]. We cannot exclude the possibility that the development of compensatory mechanism during development could have affected the evaluation of the role of POMC ARC neurons in the effects of leptin on ANS in the Bell and cols study [40]. Thus, our data demonstrated that the prolonged activation of the heterogenous population of POMC ARC neurons do not modify the corticosterone plasma levels during fasting.
Additionally, activation of POMC ARC reduced fasting-induced body weight loss and increased plasma leptin and glucose levels. Notably, only one-third of CNO-treated Pomc-Cre::hM3DGq mice had plasma leptin levels higher than the control group suggesting that the effect of prolonged activation of POMC ARC neurons on plasma leptin levels was secondary to body weight loss instead of a direct effect of POMC ARC neuronal modulation, as described by others [41]. Our data about body weight change and plasma glucose levels in fasted animals subjected to prolonged POMC ARC activation are contradictory to the classical role of these neurons on energy and glucose homeostasis. Despite some studies have observed the expected effect on energy balance after chemogenetic modulation of POMC ARC neurons (i.e. reduced food intake after activation) [42,43], there are divergent results about its effect on glucose homeostasis. So, it was reported that the selective activation of POMC ARC neurons can reduce pyruvate-induced hyperglycemia [44], while other study described no effect on plasma glucose levels after POMC ARC modulation [45] or even a fall in glycemia after inhibition of POMC ARC neurons [42]. Therefore, our data support the heterogeneity of these neurons and suggests that the responses observed in our study are involved with the activation of a subset of POMC ARC neuron that could also produce AgRP, NPY and/or GABA that could inhibit energy metabolism.
Given that the food consumpition-effects induced by activation of AgRP ARC neurons are mediated, at least in part, by modulating the activity of PVN neurons [46][47][48] and that these ARC neurons are inhibited by leptin [10] we also investigated whether the specific inhibition of AgRP ARC neuronal activity could modulate plasma corticosterone levels in fasted animals. Our findings show that inhibiting AgRP ARC neurons has no effect on fastingincreased plasma corticosterone levels, but it is effective in reducing PVN neuron activity. The lack of changes in corticosterone secretion after inhibiting of AgRP ARC neuronal activity is opposite to the results reported by pharmacological studies, which observed increased corticosterone secretion after intra PVN administration of NPY [15,16], the neuropeptide coexpressed with AgRP ARC . On the other hand, the present data on FosB study indicates that specific inhibition of AgRP ARC neurons reduces the activity of PVN neurons. Therefore, according to an excitatory role of NPY on PVN neurons, it was described that the NPYr2 can be coupled to Gq signaling [49]. Therefore, inhibition of AgRP ARC neurons could reduce NPY signaling via NPYr2 expressed in the PVN that in turn could lead to a decreased PVN neuronal activity in fasted conditions. However, we cannot exclude an indirect effect of AgRP ARC neurons on the PVN activity, since AgRP ARC neurons project to other brain sites that in turn are directly connected to PVN neurons [17,18].
Recently, it was demonstrated that GABAergic neurons from ARC are directly connected to CRF PVN neurons, supporting the involvement of this ARC neurotransmitter in the control of the HPA axis [18]. Interestingly, under prolonged hyperosmotic stress, GABAergic signaling in the PVN is switched from inhibitory to excitatory [50,51]. Moreover, administration of GABA A R agonist can induce CRF releasing in the median eminence [52]. Therefore, the inhibition of NPY and GABA release from AgRP ARC neurons could also contribute to the reduced neuronal activity in the PVN during fasting.
As expected inhibition of AgRP ARC reduced fastinginduced body weight loss, but didn't change the effects of fasting on plasma leptin and glucose levels. AgRP ARC neurons are known to negatively modulate the metabolism and part of these responses can be associated with a change in the thyroid axis activity and energy expenditure [53]. Therefore, reduction in energy expenditure after inhibition of AgRP ARC can be associated with the reduced body weight loss.
Since we had observed reduced PVN neuronal activity after inhibition of AgRP ARC neurons in fasted animals, we also investigated the content of CRF in the PVN and its releasing in the median eminence in these animals. Interestingly, our data indicates AgRP ARC neurons can modulate the synthesis and secretion of CRF PVN during metabolic stress. Our results corroborate the data from Kakizawa and Cols that observed increased CRF release in the ME after administration of GABA A R agonist [52]. Therefore, our findings are consistent with other studies indicating an excitatory role of AgRP ARC neurons on CRF PVN expression/activity [15,51,52,54] and highlight AgRP ARC neurons as a component of the neurocircuitry controlling the activity of CRF PVN neurons during fasting.
Despite the reduced CRF content in the medial parvocellular PVN subdivision and its accumulation in the ME, plasma corticosterone levels were not reduced in Agrp-IRES-Cre mice expressing hM4DGi treated with CNO. To conciliate these results, we need to consider the involvement of different pathways an mechanisms that underlie stress-induced increase in plasma corticosterone levels. So, considering that AgRP ARC neuronal inhibition induced a slight fall in PVN FosB expression, other PVN peptides involved in the control of pituitary-ACTH activity and ANS, such as vasopressin and even CRF expressed in PVN autonomic neurons, could act to maintain the increased plasma corticosterone levels via modulation of adrenal gland activity (neural mechanisms or dependent of ACTH) [55][56][57][58]. Additionally, mechanisms associated with changes in the sensitivity to the ACTH [59] and/or peripheral glucocorticoid metabolism [60,61] may contribute to the preserved corticosterone secretion in Agrp-IRES-Cre::hM4DGi mice treated with CNO. Therefore, the existence of redundant pathways/ neurocircuitries and mechanisms involved in the control of HPA axis are probably underlying the maintenance of increased plasma corticosterone levels in fasted animals after the selective inhibition of AgRP ARC neurons.
Finally, since CRF PVN neurons are involved in the control of endocrine, autonomic, and behavioral responses to stress [2][3][4], we speculate that its reduced content in the PVN with the inhibition of AgRP ARC neurons could be related to the modulation of other stress-related responses that were not investigated in our study. Given that AgRP ARC neurons modulate several motivational components of food consumption [54] and that CRF PVN also controls several behavioral stress-related responses [4], part of those behavioral changes elicited by AgRP ARC neurons could be associated to change in the activity of CRF PVN neurons.

Conclusion
Our study reinforces the recruitment of HPA axis during fasting and the involvement of leptin in this response. Additionally, we demonstrate that AgRP ARC , but not POMC ARC , are part of the neurocircuitry involved in the control of CRF PVN neurons activity during fasting. The uncovering of the mechanism and neurocircuitry involved in the coupling of HPA axis activity to changes in peripheral energy stores could be considered an important strategy to the development of new therapies to treat of pathologies associated with the disturbance of the HPA axis, including obesity and metabolic syndrome.

Animals
All procedures were approved by the Ethical Committee for Animal Use of the University of Ribeirao Preto (12/2017(12/ , February 2018 and Federal University of Sao Paulo (07/2020-1276210720). C57BL6 male mice were obtained from Central Animal Facility of the University of Sao Paulo-Campus Ribeirao Preto. Male Agrp-IRES-Cre, Pomc-Cre mice, Crh-IRES-Cre and the Cre-inducible tdTomato-reporter mouse were purchased from JAX mice (Bar Harbor, ME) and genotyped as previously described [62,63]. The animals were housed in groups under controlled light (12:12 h light-dark cycle; lights off at 06:00 p.m.) and temperature conditions (23 ± 2 °C), with free access to water and food, unless otherwise stated.

Experimental procedures
For all protocols, mice were single housed for at least 1 week before the beginning of the experiments and habituated to researcher manipulation. Flow charts of the experimental procedures are described in Additional file 1: Fig. S1.
For the fasting study, mice were weighed and maintained on regular diet or fasted for 36 h. At the end of the experiment (10 a.m.) they were weighed again and immediately decapitated for blood (heparinized tube) and tissue collection. In the leptin treatment experiment, fasted mice were treated with 4 i.p. injections of saline or leptin [Millipore (Burlington, MA)-doses of 0.5 µg/g or 1 µg/g)] at 6am, 2 pm, 10 p.m. of day 2 and at 6am on day 3 (last day). At 10 a.m. fasted mice treated with saline or leptin were weighed again and decapitated for blood and tissue collection. Perfusion for immunostaining studies (described below) were performed using Crh-IRES-Cre:tdTomato-reporter mouse. For
After the surgery, mice had 2 weeks to recover from the surgical procedure and to acclimate to be individually housed before the beginning of the experiments. The 2 weeks period was also necessary for the expression of the DREADDs. To confirm the correct injection and expression of the DREADDs, mCherry expression was visualized in the ARC.
We performed CRF immunofluoresce studies as described before [65]. Briefly, sections were pre-treated with blocking solution (2% normal horse serum and 0.25% Triton-X) and then incubated with a rabbit anti-CRH antibody (1:1000) for 48 h at 4 °C. This antibody recognizes mature CRF (1-41) and full pre-proCRF prohormone [66]. Next, sections were incubated with a goat anti-rabbit Alexa Fluor 488 antibody (1:1000, Thermo Fisher, Waltham, MA) for 2 h. Brain sections were sequentially mounted on glass slides and coverslipped with mounting media.

Quantitative neuroanatomical analysis
A Zeiss inverted Fluorescence motorized phase contrast microscope was used to acquire the photomicrographs associated to FosB data. Two or three images from each animal were quantified for mCherry, FosB only or double FosB + mCherry using Image J software (NIH, Bethesda, MD). Only animals mCherry + in both sides of the ARC were used in the experiments. We used mCherry as marker of specificity and efficiency of viral vector transduction and FosB as a marker of prolonged neuronal activity. FosB immunofluoresce + tdTomato (reporter animals) were used to indicate changes in CRF PVN neuronal activity.
For CRF-IR studies, fluorescence images were acquired with 20×/0.80 and 40×/0.95 objectives using a Zeiss AxioObserver D1 equipped with an Apotome.2 structured illumination module and an AxioCam 506 monochrome camera. All image processing and analysis were performed in the ImageJ-based open-source imageprocessing package Fiji (NIH, Bethesda, MD) [67]. Fluorescent signal corresponding to CRH + was blindly and bilaterally quantified in the PVN and EM. The average fluorescence intensity was quantified in low magnification (20×) images of one complete series of coronal sections per brain between bregma − 0.58 and − 0.94 mm for the PVN and between bregma − 1.58 and − 1.94 mm for the EM [68]. For each image, the histogram of signal intensity was obtained, and the tissue background level was estimated by fitting a Gaussian curve, which coincided with the dominant peak. With these parameters, a specific signal detection threshold was calculated, which was defined as the mean of the distribution plus five standard deviations. A region of interest was created according to this threshold. The mean fluorescence intensity was quantified in each image within this thresholded region.
Total RNA from PVN and ARC were extracted using RNeasy Mini Kit (QIAGEN, Hilden, Germany) and reverse transcribed (QuantiTect RT Kit, QIAGEN). For TaqMan gene expression assay, qPCR master mix Quan-tiNova kit (QIAGEN) and specific primers from Integrated DNA Technology (IDT, Coralville, IA) were used. Quantitative PCR was performed in triplicate using the QuantStudio 5 (Applied Biosystems). Relative mRNA levels were calculated using the standard curve method and normalized to the level of b-actin.

Plasma hormones and glucose determinations
Plasma corticosterone levels were determined by specific radioimmunoassays after extraction with ethanol, as previously described [69]. Plasma leptin (Millipore, Burlington, MA) and glucose (Doles Reagents, Goiania, Brazil) concentrations were measured using commercial kits.

Statistical analysis
The data are expressed as mean ± SEM. Samples were tested for regular distribution using Shapiro-Wilk normality test (Prism9.0-GraphPad software, San Diego, CA). We used one-or two-way ANOVA, followed by the appropriate post hoc tests, or an unpaired Student's t test (or the Mann-Whitney test for non-parameteric data). Differences were accepted as significant if p < 0.05. The specific statistical tests that were used are indicated in the figure legends.