Drp1 is required for AgRP neuronal activity and feeding

The hypothalamic orexigenic Agouti-related peptide (AgRP)-expressing neurons are crucial for the regulation of whole-body energy homeostasis. Here, we show that fasting-induced AgRP neuronal activation is associated with dynamin-related peptide 1 (DRP1)-mediated mitochondrial fission and mitochondrial fatty acid utilization in AgRP neurons. In line with this, mice lacking Dnm1l in adult AgRP neurons (Drp1 cKO) show decreased fasting- or ghrelin-induced AgRP neuronal activity and feeding and exhibited a significant decrease in body weight, fat mass, and feeding accompanied by a significant increase in energy expenditure. In support of the role for mitochondrial fission and fatty acids oxidation, Drp1 cKO mice showed attenuated palmitic acid-induced mitochondrial respiration. Altogether, our data revealed that mitochondrial dynamics and fatty acids oxidation in hypothalamic AgRP neurons is a critical mechanism for AgRP neuronal function and body-weight regulation.


Introduction
The central nervous system (CNS) regulates whole-body energy metabolism through multiple neuronal networks (Diano, 2013;Myers and Olson, 2012). The hypothalamus has been considered a key area of the brain in regulating metabolism via the ability of hypothalamic neurons to sense, integrate, and respond to fluctuating metabolic signals (Coll and Yeo, 2013;Sandoval et al., 2009). The hypothalamic arcuate nucleus (ARC) contains two distinct neuronal subpopulations that produce either orexigenic neuropeptides agouti-related peptide (AgRP) and neuropeptide-Y (NPY), or anorexigenic neuropeptides including alpha-melanocyte stimulating hormone (a-MSH) derived from proopiomelanocortin (POMC) (Batterham et al., 2002;Ollmann et al., 1997;Roh et al., 2016). The anatomical location of the hypothalamic ARC allows these neurons to rapidly respond to fluctuations of numerous circulating metabolic signals, including nutrients and hormones . However, the intracellular mechanisms underlying their ability to sense circulating signals, and, specifically nutrients, remain to be elucidated.
Mitochondria are the main powerhouse of the cell by producing adenosine triphosphate (ATP) (Mattson et al., 2008;Picard et al., 2016). Neurons rely on mitochondrial electron transport chain and oxidative phosphorylation to meet their high energy demands (Bélanger et al., 2011). In addition, mitochondria are highly dynamic organelles able to change their morphology and location according to the needs of the cell (Chan, 2006). The ability of mitochondria to change their morphological characteristics in response to the metabolic state to match with the needs of the cells occurs through fusion and fission events, process defined as mitochondrial dynamics. Mitochondrial morphological changes are associated with several proteins, including mitofusin 1 and 2 (MFN1 and MFN2) in the mitochondrial outer membrane and optic atrophy-1 (OPA1) in the mitochondrial inner membrane for mitochondrial fusion (Kasahara and Scorrano, 2014;Youle and van der Bliek, 2012), whereas mitochondrial fission is regulated by the activity of the dynamin-related protein 1 (DRP1, a mechanochemical protein encoded by the Dnm1l gene), which is recruited to the mitochondrial outer membrane to interact with mitochondrial fission factor (Mff) and mitochondrial fission 1 (Fis1) (Losó n et al., 2013).
Previous studies from our laboratory have shown that NPY/AgRP neuronal activation is associated with changes in mitochondrial morphology and density during fasting or after ghrelin administration (Andrews et al., 2008;Coppola et al., 2007;Dietrich et al., 2013), suggesting that changes in mitochondrial dynamics play a role in the regulation of neuronal activation of these neurons (Nasrallah and Horvath, 2014). In addition, we found that high-fat-diet-induced inactivation of NPY/AgRP neurons is associated with mitochondrial dynamics leaning towards mitochondrial fusion in this neuronal population . In the present study we interrogated the relevance of mitochondrial fission in AgRP neurons in relation to fuel availability.

Fasting induces mitochondrial fission in AgRP neurons
Recent studies have demonstrated that hypothalamic mitochondrial dynamics play a critical role in regulating nutrient sensing Santoro et al., 2017;Schneeberger et al., 2013;Toda et al., 2016). Using electron microscopy, we observed that compared to feeding (0.174 ± 0.007 mm 2 , p<0.0001; Figure 1a,c), fasting resulted in a significant decrease in mitochondrial size (0.130 ± 0.005 mm 2 , Figure 1b,c) in AgRP neurons together with a significant increase in mitochondrial density (0.551 ± 0.032 mitochondria/mm 2 of cytosol in fasting vs 0.423 ± 0.026 mitochondria/mm 2 of cytosol in feeding; p=0.0031; Figure 1d). This was associated with a decrease in mitochondrial aspect ratio (AR; the ratio between the major and minor axis of the ellipse equivalent to the mitochondrion which is indicative of mitochondrial morphological change; 1.629 ± 0.020 in fasting vs 1.769 ± 0.049 in feeding; p=0.0064; Figure 1e). However, total mitochondrial coverage in the cytosol (Figure 1f) in AgRP neurons was not altered between fed (7.237 ± 0.461% of cytosol) and fasted mice (6.830 ± 0.363% of cytosol; p=0.4853). These observations indicate that food deprivation promotes mitochondrial fission in AgRP neurons, consistent with our prior published work .

Fasting induces significant activation of DRP1 protein in AgRP neurons
Mitochondria fission is mediated by DRP1, which is recruited to the outer membrane of mitochondria to promote mitochondrial fragmentation in a GTPase-dependent manner followed by its phosphorylation at serine 616 site (Liesa et al., 2009). To examine whether food deprivation is associated with changes in activated Ser616 phosphorylation of DRP1 (pDRP1) levels, we assessed the distribution of pDRP1 immunoreactivity in AgRP neurons in fed and fasted mice. We found that percent of AgRP neurons expressing pDRP1 was significantly increased in fasting (49.2 ± 7.228% of AgRP neurons, n = 6; Figure 1m,o,q,r) compared to the fed condition (22.33 ± 3.921% of AgRP neurons, n = 6, p=0.0085, Figure 1l,n,p,r). No changes in AgRP cell number were observed between fed (152.2 ± 11.14; n = 5; Figure 1s) and fasted mice (160.8 ± 4.028, n = 6; p=0.4527, Figure 1s). These data suggest that activation of AgRP neurons in fasting state is closely associated with increased DRP1 activation and, thus, mitochondrial fission, suggesting that DRP1-mediated mitochondrial dynamics may play a role in the regulation of AgRP neuronal activity in fasting state.

Fasting triggers mitochondrial b-oxidation in the hypothalamic neurons
The hypothalamus is a key region in the control of energy metabolism via the ability of hypothalamic neurons to respond to numerous metabolic signals, including nutrients (Jin and Diano, 2018). It has been proposed that hypothalamic availability of free fatty acids controls food intake (Lam et al., 2005) and AgRP function (Andrews et al., 2008). To investigate the effect of fatty acids on mitochondrial b-oxidation in hypothalamic neurons, we assessed palmitic acid (PA)-induced mitochondrial oxygen consumption rate in primary hypothalamic neuronal cell cultures ( Figure 2).
First, we analyzed the percentage of tdTomato-expressing AgRP neurons in the cultures and found that about 15% of cells expressed tdTomato (14.75 ± 1.704%; Figure 2-figure supplement 1a-d). Next, we assessed the effect of 4-hydroxytamoxifen (to induce tdTomato expression) on neuronal cell viability by trypan blue staining method. Treatment of 2 mM 4-hydroxytamoxifen showed no significant difference in the percentage of cell viability compared to vehicle-treated primary hypothalamic neuronal cultures (Figure 2-figure supplement 1e).
A significant difference in PA-induced mitochondrial maximal oxygen consumption rate was observed in primary hypothalamic neurons according to the amount of glucose present in the culture ( Figure 2). In high glucose concentration, the rate of PA-induced oxygen consumption was significantly lower (Figure 2a,b) compared to that measured in low glucose (Figure 2c,d). Furthermore, under both high and low glucose conditions, a significant decrease in maximal oxygen consumption rate was observed by the addition of the etomoxir, inhibitor of carnitine palmitoyltransferase-1 (CPT1), transporter of fatty acids into the mitochondria.
Together, these data suggest that similar to fasting state, when glucose levels are low, hypothalamic neurons utilize fatty acids, such as palmitate, as substrates for mitochondrial respiration. mice, n = 1559 mitochondria/47 AgRP neurons/6 mice). Data are presented as mean ± SEM. **p<0.01; ***p<0.001 by two-tailed Student's t-test. ns = not significant. (g-k) Real-time PCR data showing relative mRNA levels of Agrp (g), Npy (h), Pomc (i), Nr5a1 (j), and Dnm1l (k) in total lysate of hypothalami (Input) and isolated RNA bound to the ribosomes of the hypothalamic AgRP neurons (IP) from 3-month-old fed or fasted mice (n = 5/group). Three animals were pooled for each n. Data are presented as mean ± SEM. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001 by two-tailed Student's t-test. (l-q) Representative micrographs showing immunostaining for phosphorylated DRP1 (at serine 616; pDRP1; green, l and m) and tdTomato (red, representing AgRP, n and o) and merged (p and q) in the hypothalamic ARC of 5-month-old fed and fasted male mice. Scale bar represents 100 mm. 3V = third ventricle; ARC = arcuate nucleus; ME = median eminence. (r) Graph showing the percentage of AgRP neurons immunopositive for pDRP1 (n = 6 mice/group). Data are presented as mean ± SEM. **p<0.01 by two-tailed Student's t-test. (s) Graph showing no difference in total AgRP cell number between fed and fasted male mice (n = 6 mice/group). Data are presented as mean ± SEM. p=0.4711 by two-tailed Student's t-test. The online version of this article includes the following source data for figure 1: We then assessed the effect of 4-hydroxytamoxifen on primary hypothalamic neuronal cell viability. Total viable cell number was measured by trypan blue staining method. Similar to primary hypothalamic neuronal cells isolated from Dnm1l +/+ ; Agrp Cre:ERT2 ; tdTomato mice ( We then determined PA-induced mitochondrial oxygen consumption rate in primary hypothalamic neuronal cell cultures from Drp1 cKO mice. Contrary to control mice (Figure 2), no difference

Deletion of Dnm1l in AgRP neurons affects POMC and paraventricular neuronal activation
Next, we analyzed immunostaining for Fos in POMC neurons of Drp1 cKO male mice and their controls ( Figure 5-figure supplement 1). A significant increase in POMC cells immunoreactive for Fos was observed in Drp1 cKO mice (33.683 ± 2.050% of POMC neurons, n = 4, p=0.0032) compared to controls (22.169 ± 1.297% of POMC neurons, n = 4,

(g and h)
Graphs showing the quantification of relative intensity (g) and particle number (h) of AgRP fibers in the PVN (bregma À0.82 mm) of fasted male control and Drp1 cKO male mice (n = 4 mice). (i and j) Immunostaining for AgRP (green) in the PVN (bregma À0.94 mm) of a fasted male control (i) and a fasted Drp1 cKO mouse (j). (k and l) Graphs showing the quantification of relative intensity (k) and particle number (l) of AgRP fibers in the PVN (bregma À0.94 mm) of fasted male control and Drp1 cKO male mice (n = 4 mice). (m and n) Immunostaining for AgRP in the PVN (bregma À1.06 mm) of a fasted control (m) and a fasted Drp1 cKO mouse (n). (o and p) Graphs showing the quantification of relative intensity (o) and particle number (p) of AgRP fibers in the PVN (bregma À1.06 mm) of fasted control and Drp1 cKO male mice (n = 4 mice). Scale bar represents 100 mm (a, e, i, and m). All data are presented as mean ± SEM. *p<0.05; **p<0.01; ****p<0.0001 by two-tailed Student's t-test. 3V = third ventricle; PVN = paraventricular hypothalamus. The online version of this article includes the following source data and figure supplement(s) for figure 5: Source data 1. AgRP-selective Dnm1l deficiency affects AgRP projections within the hypothalmic PVN.    figure supplement 1k) and particle number (control = 157.625 ± 13.488 counts, n = 4; Drp1 cKO mice = 393.400 ± 19.290 counts, n = 5, p<0.0001, Figure 5-figure supplement 1l) of a-MSH fibers were observed in the PVN of Drp1 cKO mice compared to their controls.
In agreement with a reduced AgRP and an increased POMC neuronal activation, we observed a significant increase in Fos immunopositive cells in the PVN of fasted Drp1 cKO mice (72.600 ± 9.092 counts, n = 5, p=0.0028) compared to their controls (22.750 ± 4.535 counts, n = 4, Figure

Deletion of Dnm1l in AgRP neurons alters energy metabolism
To determine the physiological outcome of AgRP-specific Dnm1l deletion, we assessed the metabolic phenotype of male and female Drp1 cKO mice and their controls. Before starting tamoxifen (TMX) injections at 5 weeks of age, no significant differences in body weight were observed between controls and Drp1 cKO mice in male ( Figure 6a).
A significant decrease in body weight of Drp1 cKO male mice compared to controls was observed 3 weeks after the start of TMX treatment ( Figure 6a) and was maintained through the end of the study when the mice were 20 weeks old (Figure 6a; n = 17 per group).
The decrease in body weight of Drp1 cKO male mice was associated with a significant reduction in fat mass (Figure 6b; n = 22, p<0.0001) while no significant difference in lean mass was observed (Figure 6c; n = 22, p=0.3421) compared to control mice.

Discussion
Our findings revealed a crucial role of mitochondrial fission in AgRP neurons in the regulation of hypothalamic feeding control. First, we found that activated AgRP neurons have decreased mitochondrial size accompanied by an increase in mitochondria number suggesting a mitochondrial fission process. In agreement with this, we found that Dnm1l mRNA levels and DRP1 activation (Liesa et al., 2009) are significantly increased in AgRP neurons of fasted mice compared to fed mice. These data were associated with a significant increase in FA-induced mitochondrial respiration in primary hypothalamic neuronal cells when low glucose levels (similar to fasting) were present compared to higher glucose levels. To determine the physiological relevance of mitochondrial fission in AgRP neurons, we generated a mouse model for conditional deletion of Dnm1l in AgRP neurons (Drp1 cKO mice). We found that Drp1 cKO mice, in which fasting did not induce mitochondrial fission and changes in mitochondrial function, had significant decreases in body weight, composition, and feeding that were accompanied by increases in locomotion and energy expenditure. Finally, Drp1 cKO mice also showed attenuated ghrelin-induced hyperphagia and neuronal activity of AgRP neurons. Altogether, these data revealed that DRP1-driven mitochondrial fission in AgRP neurons is an adaptive process enabling these neurons to respond to the changing metabolic environment.
Mitochondria are energy-producing organelles fundamental in support of cellular functions. Mitochondria are highly dynamic organelles able not only to move within the cell to sites where their function is required, but also to fuse (mitochondrial fusion) and divide (mitochondrial fission) in order to maintain proper cellular function.
Mitochondrial fusion and fission are highly regulated processes. Several proteins are involved in these events, including MFN1 and MFN2 and OPA1 for mitochondrial fusion, and Fis1, Mff, and DRP1 for mitochondrial fission (Pozo Devoto and Falzone, 2017). Mitochondrial dynamics through fusion and fission processes are also important in maintaining mitochondrial quality control in order to maintain optimal mitochondrial bioenergetic functions (Twig et al., 2008). Our data indicate that mitochondrial dynamics and specifically mitochondrial fission play an important role in sensing . Data are presented as mean ± SEM. (i) Food intake in 4-month-old control and Drp1 cKO female mice at 5 months of age (n = 7-9 mice/group) after either saline or ghrelin injection. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001; Two-way ANOVA with Tukey's post hoc analysis for multiple comparisons was performed. (j) Graph showing the membrane potential in AgRP neurons of 9-11-week-old control (n = 20 cells/10 mice) and Drp1 cKO male mice (n = 19 cells/10 mice) in response to ghrelin. *p<0.05; Two-way ANOVA with Tukey's post hoc analysis for multiple comparisons was performed. (k) Graph showing normalized firing rate in AgRP neurons of 9-11week-old control (n = 20 cells/10 mice) and Drp1 cKO male mice (n = 19 cells/10 mice) in response to ghrelin. Data are presented as mean ± SEM. **p<0.01 for artificial CSF-treated control versus ghrelin-treated control; *p<0.05 for ghrelin-treated control versus washed out control by two-way ANOVA with Tukey's post hoc analysis for multiple comparisons. (l) Representative tracers of AgRP neurons from a control and a Drp1 cKO mouse in response to ghrelin. Scale bar represents 100 mm. Scale bar in high magnification image represents 20 mm. 3V = third ventricle; ME = median eminence. The online version of this article includes the following source data and figure supplement(s) for figure 7: Source data 1. Deletion of Dnm1l in AgRP neurons attentuates ghrelin in induced neuronal activation and feeding.  changes of nutrients availability in AgRP neurons. First, we observed that incubation of primary hypothalamic neurons with palmitic acid induced a significant increase in mitochondrial respiration when glucose levels were low, mimicking fasting. Fasting induced increased AgRP neuronal activation and increased mitochondrial fission. However, when DRP1-induced mitochondrial fission in AgRP neurons was abolished, palmitic acid-induced mitochondrial respiration was diminished. In association with these, ghrelin-triggered changes in membrane potential and firing frequency of AgRP neurons were significantly attenuated in Drp1 cKO mice, leading to failure in inducing hyperphagia. In line with our results, Dietrich et al., 2013 have shown that in mice with AgRP-selective deletion of Mfn1 and Mfn2, mediators of mitochondrial fusion process, neuronal firing frequency was impaired in diet-induced obesity mice. The impairment of AgRP neuronal activation was reversed by increasing intracellular ATP levels , indicating that the impaired AgRP neuronal firing frequency is likely due to low intracellular ATP levels. In addition to these functions, changes in shape and size of mitochondria may also affect the ability of a cell to distribute its mitochondrial population to specific subcellular locations. This function is especially important in highly polarized cells, such as neurons. Future studies assessing mitochondrial dynamics with changes in mitochondrial subcellular distribution will address this point.
In addition to mitochondria, DRP1 has been also shown to enable peroxisomal fission. Peroxisomes are single-membrane organelles that similar to mitochondria catalyze the breakdown of long chain fatty acids through beta-oxidation and regulate the maintenance of redox homeostasis (Smith and Aitchison, 2013). Because of these shared properties and metabolic pathways, we cannot exclude a possible involvement of peroxisomes in the phenotype observed in our mice. Further studies are warranted to address this issue.
Overall, our data unmask that mitochondrial fission in hypothalamic AgRP neurons is a fundamental mechanism that allows these neurons to sense and respond to changes in circulating signals, including hormones such as ghrelin and nutrients such as glucose and palmitic acid, in the regulation of feeding and energy metabolism.

Animals
All animal care and experimental procedures done in this study were approved by the Yale University (protocol # 10670) and the Columbia University (protocols # AC-AABI0565 and AC-AABH9564) Institutional Animal Care and Use Committees. All mice were housed in a temperature-controlled environment (22-24˚C) with a 12 hr light and 12 hr dark (19.00-07.00 hr) photoperiod. Animals were provided standard chow diet (SD) (2018; 18% calories from fat; Harlan Teklad, Madison, WI, USA) and water ad libitum unless otherwise stated. All fasted mice were food deprived for 16 hr (18.00-10.00 hr) prior to the experiment. All mice studied were of the same (mixed) background.

Ribotag assays
We performed transcriptomic profiling by using ribosomal tagging strategy to analyze AgRP neurons-specific mRNA expression in vivo. To avoid the potential disadvantage that the embryonic POMC-expressing progenitor neurons differentiate into AgRP-expressing neurons, we crossed Agrp-Cre:ERT2 mice (Wang et al., 2014) with Rpl22 floxed (RiboTag, #029977, The Jackson Laboratories, Bar Harbor, ME, USA) mice to eventually generate Agrp Cre:ERT2 ; RiboTag mice, expressing a hemagglutinin A (HA)-tagged ribosomal protein in the AgRP neurons upon tamoxifen injection. Eleven-to twelve-week-old mice (1 month after the last tamoxifen injection) were used. After mice were anesthetized with isoflurane and decapitated, the brains were rapidly dissected out. To carefully collect the hypothalamic arcuate nucleus (ARC), brain tissues were sectioned in two-millimeter thick coronal sections containing mediobasal hypothalamus (MBH) in a brain matrix. The MBH ARC samples were collected under a stereomicroscope according to the brain atlas for appropriate regions and preventing differences in tissue weight. Three animals were pooled for each N. The MBH ARC samples from Agrp Cre:ERT2 ; RiboTag mice were homogenized by supplemented homogenization buffer (HB-S: 50 mM Tris, pH 7.4, 100 mM KCl, 12 mM MgCl 2 , and 1 % NP-40 supplemented with 1 mM DTT, 1 mg/ml heparin, 100 mg/ml cycloheximide, 200 U/ml RNasin Ribonuclease inhibitor, and protease inhibitor cocktail). Samples were then centrifuged at 10,000 rpm for 10 min at 4˚C. Then, 50 ml of each supernatant was transferred to a new tube serving as input fraction (containing all mRNAs). To isolate polyribosomes, we performed immunoprecipitation of ribosome-bound mRNAs in AgRP neurons. by utilizing anti-HA antibody (5 ml/sample; Cat#901513, Biolegend, San Diego, CA, USA).

Metabolic assays
Four-month-old mice were acclimated in metabolic chambers (TSE System-Core Metabolic Phenotyping Center, Yale University) for 3 days before the start of the recordings. Mice were continuously recorded for 2 days, with the following measurements taken every 30 min: food intake, locomotor activity (in the x-, y-, and z-axes), and gas exchange (O 2 and CO 2 ; The TSE LabMaster System, Chesterfield, MO, USA). Energy expenditure was calculated according to the manufacturer's guidelines (PhenoMaster Software, TSE System, Chesterfield, MO, USA). The respiratory quotient was estimated by calculating the ratio of CO 2 production to O 2 consumption. Values were adjusted by body weight to the power of 0.75 (kgÀ0.75) where mentioned. Body composition was measured in vivo by MRI (EchoMRI, Echo Medical Systems, Houston, TX, USA) monthly at 10:00 AM. Body core temperature was measured at 10:00 AM using a thermocouple rectal probe and thermometer (Physitemp instruments, Clifton, NJ, USA). Rectal temperature was measured for repeated three times, and the average was calculated. The temperature of the surface overlying BAT was measured using infrared thermography images (FLIR C2, FLIR Thermal Imaging System, Arlington, VA, USA). The infrared thermography images were taken at least three times and analyzed using FLIR Tools (FLIR Thermal Imaging System, Arlington, VA, USA).

Phosphorylated-DRP1 immunostaining
Five-month-old mice were deeply anesthetized and transcardially perfused with 0.9% saline containing heparin (10 mg/l), followed by fresh fixative of 4% paraformaldehyde in phosphate buffer (0.1 M PB, pH 7.4) as previously described (Andrews et al., 2008;Diano et al., 2011;Toda et al., 2016). Brains were post-fixed overnight at 4˚C and sliced to a thickness of 50 mm using a vibratome (#11000, PELCO easySlicer, TED PELLA Inc, Redding, CA, USA) and coronal brain sections containing the ARC were selected under the stereomicroscope (Stemi DV4, Carl Zeiss Microimaging Inc, Thornwood, NY, USA). After several washes with 0.1 M PB, brain sections were preincubated with 0.2% triton X-100 (Sigma-Aldrich, Saint Louis, MO, USA) and 2% normal goat serum in 0.1 M PB for 30 min to permeabilize tissue and cells. Brain sections were incubated with rabbit anti-phosphorylated-DRP1 (Ser-616) antibody (diluted 1:500 in 0.1 M PB, #4494, Cell Signaling, Technology, Danvers, MA, USA) overnight at room temperature (RT). The following day, brain sections were washed and incubated with a biotinylated goat anti-rabbit IgG (diluted 1:200 in 0.1M PB, BA-1000, Vector Laboratories, Inc, Burlingame, CA, USA) for 2 hr at RT. Sections were then washed and incubated in streptavidin-conjugated Alexa Fluor 488 (diluted 1:2000 in 0.1 M PB, A21370, Life Technologies, Carlsbad, CA, USA) for 2 hr at RT. No staining was performed to visualize AgRP neurons since mice were expressing tdTomato in this neuronal population, which is per se fluorescent. After several washes with 0.1 M PB, brain sections were mounted on glass slides and coverslipped with a drop of Vectashield mounting medium (H-1000, Vector Laboratories, Burlingame, CA, USA). The coverslip was sealed with nail polish to prevent drying and movement under the microscope. All slides were stored in the dark at 4˚C.

Fos immunostaining
Five-month-old mice were deeply anesthetized and transcardially perfused as described above. Immunofluorescent staining was performed using rabbit anti-Fos antibody (diluted 1:2000 in 0.1 M PB, sc-52, Santa Cruz Biotechnology, Dallas, TX, USA) overnight at RT. The following day, brain sections were washed and incubated with a biotinylated goat anti-rabbit IgG secondary antibody

AgRP and a-MSH fiber immunostaining
Five-month-old mice were deeply anesthetized and transcardially perfused as described above. Brain sections containing the hypothalamic paraventricular nucleus (PVN) were selected under the stereomicroscope. Immunofluorescence staining was performed using rabbit anti-AgRP antibody (diluted 1:1000 in 0.1 M PB, H-003-57, Phoenix Pharmaceuticals, Inc) and sheep anti-a-MSH antibody (diluted 1:1000 in 0.1 M PB, ab5087, Millipore Sigma, Burlington, MA, USA) overnight at RT. The following day, brain sections were washed and incubated with anti-rabbit Alexa Fluor 488 (diluted 1:1000 in 0.1M PB, A21206, Life Technologies) and anti-sheep Alexa Fluor 488 (diluted 1:1000 in 0.1M PB, A11015, Life technologies) for 2 hr at RT. After several washes with 0.1 M PB, brain sections were mounted on glass slides, coverslipped with a drop of vectashield mounting medium, and analyzed with a fluorescence microscope.

Fluorescent image capture and analyses
Five-month-old mice were deeply anesthetized and transcardially perfused as described above. Fluorescent images were captured with Fluorescence Microscope (Model BZ-X710, KEYENCE, Osaka, Japan). For all immunohistochemistry (IHC) analyses, coronal brain sections were anatomically matched (ARC: between À1.46 and À2.06 mm from bregma, PVN: À0.70 and À1.06 mm from bregma) with the mouse brain atlas (Franklin and Paxinos, 2019). Both sides of the bilateral brain region (ARC and PVN) were analyzed per mouse. For each mouse, three hypothalamic level-matched per mouse were used to quantify Fos immunoreactive cells in all AgRP and POMC immunostained cells observed in the ARC. The number of immunostained cells was counted manually using ImageJ software (Schneider et al., 2012) by an unbiased observer. For area measurements and particle counting, region of interest (ROI) within fluorescence images was manually selected with the mouse brain atlas for ARC, DMH, and PVN, and was then measured by ImageJ software as previously described (Jin et al., 2016).

Hypothalamic primary neuronal cell culture
Eight to ten neonatal (0-1 day old) pups were used for hypothalamic primary neuronal cell culture. For control culture, we used either Dnm1l fl/fl ; Agrp Cre:ERT2 ; tdTomato mice which neuronal cultures were treated with vehicle (ethanol) or Dnm1l +/+ ; Agrp Cre:ERT2 ; tdTomato mice which neuronal cultures were treated with 4-hydroxytamoxifen (2 mM). Hypothalamic primary neuronal cultures from Drp1 cKO mice (Dnm1l fl/fl ; Agrp Cre:ERT2 ; tdTomato mice) were treated with 4-hydroxytamoxifen (2 mM). In brief, we carefully removed the MBH of the brain and placed it onto a small culture dish that contains a small volume of Hibernate-A Medium (Cat# A1247501, Gibco-Thermo Fisher Scientific, Waltham, MA, USA). The tissues dissociated to single cells after digestion with 6 ml of Hibernate-A Medium containing 2.5% of Trypsin-EDTA for 15 min at 37˚C. Suspended cells were filtered (40 mm) and centrifuged for 5 min at 1000 rpm and the pellet was re-suspended and plated on XF96 cell culture microplates (Cat# 101085-004, Agilent Technologies, Santa Clara, CA, USA) coated with poly-D-lysine (Cat# P6407, Sigma-Aldrich, Saint Louis, MO, USA) at a density of 1 Â 10 5 cells per well, and they were cultured in Neurobasal medium (Cat# 21103049, Gibco-Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 1% penicillin-streptomycin, 2% B-27 Supplement (Cat# 17504044, Gibco-Thermo Fisher Scientific, Waltham, MA, USA), and GlutaMAX-I (Cat# 35050061, Gibco-Thermo Fisher Scientific, Waltham, MA, USA). After 10 days in culture, primary neuronal cells isolated from control (Dnm1l +/+ ; Agrp Cre:ERT2 ; tdTomato) and Drp1 cKO mice were treated with 2 mM 4-hydroxytamoxifen (H7904, Sigma-Aldrich, Saint Louis, MO, USA) for expression of a CreER recombinase while the other control group (generated from Dnm1l fl/fl ; Agrp Cre:ERT2 ; tdTomato mice) was treated with vehicle (ethanol) to assess the effect of 4-hydroxytamoxifen on cell viability. Primary neuronal cells were used for the measurement of mitochondria fatty acid oxidation 5 days later.

Cell quantification in cultures
Cells were analyzed by capturing six to eight random fields per coverslip. For the quantitative analysis of cell number, tomato expressing cells in DAPI (Cat# p36962, Thermo Fisher Scientific, Waltham, MA, USA)-stained cultures were manually counted using Image J software. Cells were visualized using Fluorescence Microscope (Model BZ-X710, KEYENCE, Osaka, Japan). Five coverslips per group were counted within an experiment.

Viability assay in cultures
Neuronal cell viability was determined by trypan blue exclusion assay in cultures maintained in each condition. The cultures were stained with 0.4% trypan blue (Cat# 302643, Sigma-Aldrich, Saint Louis, MO, USA) for 15 min at room temperature and then washed with phosphate-buffered saline (PBS). And then, 10 mL of suspended cells was loaded into each chamber of the hemocytometer. Counts were performed by triplicate by one analyst under a 40Â objective according to the standard methodology. The non-stained (live) and Trypan blue-stained (dead) cell counts were counted and calculated in three randomly selected areas (0.2 mm 2 ) in each well (n = 5 per treatment condition) to calculate the cell viability percentage.

Measurement of mitochondrial fatty acid oxidation assay
The fatty acid oxidation (FAO) was measured using a microfluorimetric Seahorse XF96 Analyzer (Agilent Technologies, Santa Clara, CA, USA) according to the protocol supplied by the manufacturer with minor modifications. Cells were starved with minimal substrate neurobasal-A medium (Cat# 10888022, Thermo Fisher Scientific) for 24 hr. The minimal substrate medium included 1% B-27 Supplement (Cat# 17504044, Gibco-Thermo Fisher Scientific, Waltham, MA, USA), 1 mM glutamine, 0.5 mM carnitine, and 2.5 or 0.5 mM of glucose. The day of the assay, 45 min prior to the assay, starved cells were washed and incubated with Seahorse XF Base medium Minimal DMEM (Cat# 102353-100, Agilent Technologies, Santa Clara, CA, USA) supplemented with 2.5 or 0.5 mM glucose and 0.5 mM carnitine in a non-CO 2 37˚C incubator. Fifteen minutes prior to the assay, 40 mM etomoxir was added to the cells to measure endogenous fatty acid uptake for FAO. Palmitate-BSA or BSA control (Seahorse XF Palmitate-BSA FAO substrate, Cat# 1102720-100, Agilent Technologies, Santa Clara, CA, USA) were added to cells right before initiating the XF assay. During the assay, cells were exposed to compounds in the following order: 5 mM of oligomycin (Cat# 495455, Sigma), 10 mM of FCCP [carbonyl cyanide-p-(trifluoromethoxy) phenylhydrazone] (Cat# C2920, Sigma-Aldrich, Saint Louis, MO, USA), 10 mM of antimycin A (Cat# A8674, Sigma-Aldrich, Saint Louis, MO, USA), and 5 mM of rotenone (Cat# R8875, Sigma-Aldrich, Saint Louis, MO, USA). Wave 2.6.0 (Agilent Technologies software, Santa Clara, CA, USA) software was used to analyze the parameters.

Electrophysiology analysis
Electrophysiology analyses were performed as previously described (Toda et al., 2016). Briefly, 11-12-week-old mice were used for recordings. After mice were anesthetized with isoflurane and decapitated, the brains were rapidly removed and immersed in an oxygenated cutting solution at 4J C containing (in mM): sucrose 220, KCl 2.5, NaH 2 PO 4 1.23, NaHCO 3 26, CaCl 2 1, MgCl 2 6, and glucose 10, pH (7.3) with NaOH. After being amputated to a small tissue block, coronal slices containing the hypothalamus (300 mm thick) were cut with a vibratome. After preparation, slices were stored in a holding chamber with an oxygenated (with 5% CO 2 % and 95% O 2 ) artificial cerebrospinal fluid (aCSF) containing (in mM): NaCl 124, KCl 3, CaCl 2 2, MgCl 2 2, NaH 2 PO 4 1.23, NaHCO 3 26, glucose 3, pH 7.4 with NaOH. The slices were eventually transferred to a recording chamber perfused continuously with aCSF at 33˚C at a rate of 2 ml/min after at least a 1 hr recovery in the storage chamber. Perforated patch recording was performed in AgRP-Tomato neurons of the ARC under voltage and current clamp. The membrane and spontaneous action potential were recorded in AgRP neurons under zero current clamp condition. For ghrelin-induced AgRP neuronal activation, baseline activity was recorded for at least 15 min. Slices were then perfused with 10 nM ghrelin, diluted in aCSF for 3 min, followed by a washout (with no ghrelin). At the end of the perforated patch recordings, the membrane of every cell was ruptured and whole-cell patch recording measured to check current-voltage relationship. All data were sampled at 5 kHz, filtered at 2.4 kHz, and analyzed with an Apple Macintosh computer using AxoGraph X (AxoGraph Scientific, Foster City, CA, USA). Statistics and plotting were performed with KaleidaGraph (Synergy Software, Inc, Reading, PA, USA) and Igor Pro (WaveMetrics, Lake Oswego, OR, USA). The average firing rate was calculated in the last 2 min of each control period or treatment application. All the experiments were performed blindly to the electrophysiologist.

Ghrelin administration
Individually housed 4-month-old mice were i.p. injected with either 0.9% saline (#0409-1966-12, Hospira Inc, Lake Forest, IL, USA) or ghrelin (10 nmol, HOR-297-B, ProSpec, Rehovot, Israel) at 9:00 AM. Immediately after injection, mice were returned to their home cages, which contained a pre-weighed amount of food. The remaining food was measured at 0.5, 1, 2, and 4 hr post-injection. For immunostaining, mice were injected with ghrelin at 9:00 AM and 1 hr later, mice were deeply anesthetized and transcardially perfused, and brains were dissected and sectioned (50 mm) using a vibratome. Brain sections were processed for Fos immunostaining. Fluorescent images were captured with a Fluorescence Microscope (BZ-X710, KEYENCE, Osaka, Japan). Fos/AgRP positive cells were counted using ImageJ software.

Measurement of circulating hormones
Five-month-old mice were deeply anesthetized and decapitated. The blood was collected into a capillary tube (Microvette, CB 300 Z, Sarstedt, Nü mbrecht, Germany) containing 0.2 mg 4-(2-aminoethyl)-benzene-sulfonyl fluoride (AEBSF, Roche, Basel, Switzerland). Serum from blood samples was obtained by centrifugation at 3000 rpm for 15 min, and each circulating hormone was determined using a commercially available ELISA kit for total ghrelin (Rat/Mouse Total Ghrelin ELISA kit, EZRGRT-91K, Millipore Sigma, Burlington, MA, USA) and active ghrelin (Rat/Mouse Total Ghrelin ELISA kit, EZRGRT-90K, Millipore Sigma, Burlington, MA, USA). Serum samples and standards were analyzed in duplicate. All procedures were performed by following the manufacturer's protocol.

Statistical analysis
Two-way ANOVA was used to determine the effect of the genotype and treatment with the Prism 7.01 software (GraphPad Software). For repeated measures analysis, ANOVA was used when values over different times were analyzed. When only two groups were analyzed, statistical significance was determined by an unpaired Student's t-test. A value of p<0.05 was considered statistically significant. All data is shown as mean ± SEM, unless otherwise stated.