Recent studies have demonstrated that hypothalamic mitochondrial dynamics play a critical role in regulating nutrient sensing (Dietrich et al., 2013; 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 µm², p<0.0001; Figure 1a,c), fasting resulted in a significant decrease in mitochondrial size (0.130 ± 0.005 µm², Figure 1b,c) in AgRP neurons together with a significant increase in mitochondrial density (0.551 ± 0.032 mitochondria/µm² of cytosol in fasting vs 0.423 ± 0.026 mitochondria/µm² 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 (Dietrich et al., 2013).

We next performed transcriptomic analysis of AgRP neurons using ribosomal tagging strategy to analyze AgRP neuron-specific mRNA expression levels in fed and fasted conditions using Agrp^Cre:ERT2; RiboTag mice. We observed that compared to fed state (1.036 ± 0.132, n = 5), fasting resulted in a significant increase in Agrp mRNA transcript in the input RNAs (5.248 ± 1.176, n = 5; p=0.00741, Figure 1g). Compared to the input RNAs, Agrp and Npy mRNA transcripts were enriched by 27- and 56-fold in the immunoprecipitated RNAs (IP), respectively, in fasted Agrp^Cre:ERT2; RiboTag mice (Agrp = input = 5.248 ± 1.176, n = 5; IP = 143.998 ± 22.651, n = 5; p=0.0003, Figure 1g; Npy = input = 2.680 ± 0.996, n = 5; IP = 151.286 ± 51.683, n = 5; p=0.0207, Figure 1h). Conversely, a significant decrease of Pomc mRNA transcript in the IP RNAs was found in fed Agrp^Cre:ERT2; RiboTag mice (0.203 ± 0.093, n = 5; p=0.0031, Figure 1i) compared to the input RNAs (1.062 ± 0.184, n = 5). Moreover, marginal expression of Nr5a1, encoding steroidogenic factor-1 (SF-1, highly restricted to the VMH), was detected in the input (fed = 1.175 ± 0.309, n = 5; fasted = 1.537 ± 0.246, n = 5, Figure 1j) and the IP RNAs (fed = 0.846 ± 0.293, n = 5; fasted = 1.826 ± 0.376, n = 5; p=0.074, Figure 1j). The Agrp IP/input ratio (27.439, Figure 1g), Npy IP/input ratio (56.45, Figure 1h) and enrichment were high, while Pomc (IP/input ratio, 0.191, Figure 1i) and Nr5a1(IP/input ratio, 0.72, Figure 1i) were de-enriched, validating the arcuate AgRP neuronal isolation protocol. In support of our mitochondrial morphology data, quantitative real time-PCR (qRT-PCR) analyses revealed that Dnm1l mRNA transcript (fed = 0.873 ± 0.337, n = 5; fasted = 2.095 ± 0.335, n = 5; p=0.0329, Figure 1k) was significantly upregulated in AgRP neurons of fasted mice compared to fed mice.

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.

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 β-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 μM 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.

Next, to investigate the physiological functions of DRP1 in adult AgRP neurons, we generated mice with selective and inducible deletion of Dnm1l in AgRP neurons (Figure 3—figure supplement 1a). Agrp^Cre:ERT2; tdTomato mice (kindly provided by Dr. Joel Elmquist at UTSW; Wang et al., 2014) were crossed with Dnm1l floxed mice (Dnm1l^fl/fl) (Kageyama et al., 2014; Santoro et al., 2017). As control groups, Dnm1l^+/+; Agrp^Cre:ERT2; tdTomato mice were injected with tamoxifen and Dnm1l ^fl/fl; Agrp^Cre:ERT2; tdTomato were mice injected with corn oil. Dnm1l ^fl/fl; Agrp^Cre:ERT2; tdTomato mice, referred to here as Drp1 conditional knockout mice (Drp1 cKO mice) were injected with tamoxifen to induce mature-onset deletion of Dnm1l in AgRP neurons. To validate our animal model, we analyzed and found limited pDRP1 expression in the AgRP neurons of fasted Drp1 cKO mice (14.15 ± 0.926% of AgRP neurons, n = 4; Figure 3—figure supplement 1b–h) compared to fasted Dnm1l^+/+; Agrp^Cre:ERT2; tdTomato mice (used as control to visualize AgRP neurons; 52.39 ± 3.71% of AgRP neurons, n = 4, p<0.0001, Figure 3—figure supplement 1b–h) by immunohistochemistry analysis. No difference in AgRP cell numbers was found between control (150.3 ± 4.423 neurons, n = 4) and Drp1 cKO mice (146 ± 3.082 neurons, n = 4, p=0.4605, Figure 3—figure supplement 1i).

Next, we analyzed mitochondrial morphological changes in AgRP neurons of Drp1 cKO male mice in fed and fasted states. No differences in mitochondrial size (Figure 3a–c), density (Figure 3d), aspect ratio (Figure 3e), and coverage (Figure 3f) were observed between fed and fasted Drp1 cKO male mice, indicating that selective deletion of Dnm1l in AgRP neurons prevents fasted-induced mitochondrial fission.

First, similar to control-derived cultures, about 15% of cells were tdTomato positive (Drp1 cKO, 14.70 ± 2.901%; Figure 3—figure supplement 2a–d). No significant difference was observed between the percentage of tdTomato positive cells in this group (from Dnm1l^fl/fl;Agrp^Cre:ERT2 mice) and the percentage of tdTomato positive cells derived from Dnm1l^+/+; Agrp^Cre:ERT2 mice shown in Figure 2—figure supplement 1d (p=0.9883 by two-tailed Student’s t-test).

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 (Figure 2—figure supplement 1e), primary hypothalamic neuronal cells isolated from Dnm1l^fl/fl; Agrp^Cre:ERT2; tdTomato mice showed no significant difference in the percentage of cell viability when treated either with 2 μM 4-hydroxytamoxifen or vehicle (Figure 3—figure supplement 2e). Furthermore, a significant reduction of Dnm1l mRNA expression was observed in primary hypothalamic neuronal cell cultures treated with 2 μM 4-hydroxytamoxifen derived from Dnm1l^fl/fl; Agrp^Cre:ERT2; tdTomato mice (0.4512 ± 0.2134, n = 4; Figure 3—figure supplement 2f) compared to primary hypothalamic neuronal cell cultures treated with 2 μM 4-hydroxytamoxifen derived from Dnm1l^+/+; Agrp^Cre:ERT2; tdTomato mice (1.037 ± 0.1825, n = 3, p=0.0493, Figure 3—figure supplement 2f).

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 in PA-induced maximal oxygen consumption rate was observed in high (2.5 mM) (Figure 3g,h) or low glucose (0.5 mM) (Figure 3i,j). In addition, no effects induced by etomoxir incubation were observed in primary hypothalamic neurons derived from Drp1 cKO mice (Figure 3g–j), indicating that DRP1 in the hypothalamic AgRP neurons plays an essential role in regulating PA-induced mitochondrial respiration.

To assess the effect of Dnm1l deletion on AgRP neuronal activation, we then performed and analyzed immunostaining for Fos in the hypothalamic arcuate nucleus of Drp1 cKO male mice and controls in fasting state (Figure 4a–f; Figure 4—figure supplement 1). Compared to fasted controls (bregma −1.70 mm: 37.0 ± 2.95% of AgRP neurons, n = 6, Figure 4a,c,e,g; bregma −1.46 mm: 40.6 ± 2.71% of AgRP neurons, n = 5, Figure 4g, Figure 4—figure supplement 1a–c; bregma −2.06 mm: 36.4 ± 3.97% of AgRP neurons, n = 6; total 39.14 ± 2.925% of AgRP neurons, n = 6, Figure 4g, Figure 4—figure supplement 1g–i), fasted Drp1 cKO male mice showed a significant decrease in immunoreactivity for Fos in AgRP neurons (bregma −1.70 mm: 25.1 ± 3.83% of AgRP neurons, n = 5, p=0.0336, Figure 4b,d,f,g; bregma −1.46 mm: 22.6 ± 4.15% of AgRP neurons, n = 4, p=0.0069, Figure 4g, Figure 4—figure supplement 1d–f; bregma −2.06 mm: 21.6 ± 4.82% of AgRP neurons, n = 4, p=0.0455; total, 24.2 ± 3.58% of AgRP neurons, n = 6, p=0.0097, Figure 4g, Figure 4—figure supplement 1j–l). No changes in AgRP cell number were observed between control (bregma −1.46 mm, 136.2 ± 10.71 counts, n = 5; bregma −1.70 mm, 142.3 ± 7.89 counts, n = 6; bregma −2.06 mm, 142.7 ± 8.841 counts, n = 6) and Drp1 cKO male mice (bregma −1.46 mm, 147.5 ± 1.50 counts, n = 4, p=0.3858; bregma −1.70 mm, 154.0 ± 3.48 counts, n = 5, p=0.2397; bregma −2.06 mm, 148.0 ± 3.89 counts, n = 4, p=0.6395; total, 150.833 ± 2.753 counts, n = 5, p=0.3467, Figure 4h). In agreement with reduced AgRP neuronal activation, a significant reduction in overnight fasting-induced food intake was observed in Drp1 cKO mice (5.155 ± 0.294 g, n = 11, p=0.005) compared to controls (6.391 ± 0.290 g, n = 11, Figure 4i).

Furthermore, we analyzed AgRP immunoreactive fibers in one of the major target areas of the hypothalamus, the PVN, of fasted Drp1 cKO male mice and controls. We observed that compared to controls, a significant decrease in the PVN AgRP fluorescent intensity (bregma −0.70 mm: control = 1.000 ± 0.1175, n = 4; Drp1 cKO mice = 0.4372 ± 0.0864, n = 4, p=0.0084, Figure 5a–c; bregma −0.82 mm: control = 1.000 ± 0.03906, n = 4; Drp1 cKO mice = 0.3182 ± 0.032222, n = 4, p<0.0001, Figure 5e–g; bregma −0.94 mm: control = 1.000 ± 0.1158, n = 4; Drp1 cKO mice = 0.6318 ± 0.06872, n = 4, p=0.0340, Figure 5i–k; bregma −1.06 mm: control = 1.000 ± 0.1267, n = 4; Drp1 cKO mice = 0.5641 ± 0.04275, n = 4, p=0.0173, Figure 5m–o) and particle number (bregma −0.70 mm: control = 885.25 ± 36.16 counts, n = 4; Drp1 cKO mice = 673 ± 42.15, n = 4, p=0.0087, Figure 5d; bregma −0.82 mm: control = 899.5 ± 22.15 counts, n = 4; Drp1 cKO mice = 777.6 ± 14.84, n = 4, p=0.0038, Figure 5h, bregma −0.94 mm: control = 965.6 ± 25.88 counts, n = 4; Drp1 cKO mice = 821.4 ± 24.35, n = 4, p=0.0067, Figure 5l; bregma −1.06 mm: control = 979 ± 11.64 counts, n = 4; Drp1 cKO mice = 848.5 ± 18.79, n = 4, p=0.0010, Figure 5p) were observed. Similar results were also observed in the PVN of fasted Drp1 cKO female mice compared to controls (data not shown).

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, Figure 5—figure supplement 1a–g). No changes in POMC cell number were observed between control and Drp1 cKO mice (Figure 5—figure supplement 1h).

We then analyzed α-MSH fiber immunostaining in the PVN of fasted Drp1 cKO male mice and their controls (Figure 5—figure supplement 1i,j). Significant increases in relative intensity (control = 1.000 ± 0.164 counts, n = 4; Drp1 cKO mice = 6.195 ± 0.494, n = 5, p<0.0001, Figure 5—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 α-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 5—figure supplement 1m–o).

To assess whether deletion of Dnm1l in AgRP neurons had any effect on Agrp and Pomc mRNA levels in arcuate nucleus, we next performed transcriptomic analysis of AgRP neurons using fasted control and Drp1 cKO Ribotag mice. Quantitative real time-PCR analyses revealed that Dnm1l mRNA transcript (fasted control = 4.488 ± 0.685, n = 5; fasted Drp1 cKO = 1.361 ± 0.294, n = 5; p=0.00301, Figure 5—figure supplement 2a) was significantly downregulated in AgRP neurons of fasted Drp1 cKO mice compared to fasted control mice, validating the arcuate AgRP neuronal isolation protocol and the mouse model for conditional deletion of Dnm1l in AgRP neurons. However, no significant differences in Agrp (Figure 5—figure supplement 2b) and Pomc mRNA levels (Figure 5—figure supplement 2c) in the arcuate nucleus were observed between fasted controls and Drp1 cKO mice.

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 (body weight = Dnm1l^+/+; Agrp^Cre:ERT2-TMX = 18.533 ± 0.390, n = 18; Dnm1l^fl/fl; Agrp^Cre:ERT2-TMX = 18.176 ± 0.512, n = 17; Dnm1l^fl/fl; Agrp^Cre:ERT2-Corn oil = 18.830 ± 0.308, n = 10; p=0.8649 for Dnm1l^+/+; Agrp^Cre:ERT2-TMX versus Dnm1l^fl/fl; Agrp^Cre:ERT2-TMX; p=0.9289 for Dnm1l^+/+; Agrp^Cre:ERT2-TMX versus Dnm1l^fl/fl; Agrp^Cre:ERT2-Corn oil; p=0.7045 for Dnm1l^fl/fl; Agrp^Cre:ERT2-TMX versus Dnm1l^fl/fl; Agrp^Cre:ERT2-Corn oil, 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.

Drp1 cKO mice showed significantly lower food intake compared to controls (control = 4.181 ± 0.124 g, n = 14; Drp1 cKO mice = 3.366 ± 0.139 g, n = 18, p<0.0001, total in Figure 6d,e), due to a significant reduction of food intake during the dark period (control = 2.937 ± 0.137 g, n = 14; Drp1 cKO mice = 2.471 ± 0.120 g, n = 18, p=0.0236, dark in Figure 6e). A significant increase in locomotor activity was observed in Drp1 cKO mice compared to controls (control = 50122.750 ± 5919.799 beam-break counts, n = 14; Drp1 cKO mice = 108400.917 ± 17432.685 beam-break counts, n = 18, p=0.0008, Figure 6f), which was due to a significant increase during the dark period (control = 35556.429 ± 4272.846 counts, n = 14; Drp1 cKO mice = 80081.806 ± 13085.949 counts, n = 18, p=0.0136, Figure 6f).

The difference in body weight and composition were also associated with significantly increased energy expenditure (control = 303.329 ± 5.760, n = 14; Drp1 cKO mice = 337.096 ± 7.218, n = 18, p<0.0001, Figure 6g–i), increased O₂ consumption (control = 1245.16 ± 29.94, n = 14; Drp1 cKO mice = 1345.19 ± 31.46, n = 18, p=0.0318, Figure 6j) and CO₂ production (control = 1140.4 ± 28.18, n = 14; Drp1 cKO mice = 1248.44 ± 27.15, n = 18, p=0.0105, Figure 6k) in Drp1 cKO mice compared to control mice, while no significant differences in the respiration exchange rate (RER) were observed between Drp1 cKO mice (0.8734 ± 0.00852, n = 18, p=0.3133) and control mice (0.8855 ± 0.007676, n = 14, Figure 6l). Similar to males, female Drp1 cKO mice showed significant differences in body weight and composition, feeding, energy expenditure, O₂ consumption, CO₂ production, and locomotor activity compared to controls (Figure 6—figure supplement 1).

BAT thermogenesis is a critical component of the homeostatic energy balance to maintain body temperature (Morrison and Madden, 2014). We then examined whether deletion of Dnm1l in AgRP neurons affects body temperature. We found that BAT temperature was significantly increased in Drp1 cKO mice compared to control mice (control = 33.388 ± 0.223°C, n = 8; Drp1 cKO mice = 35.038 ± 0.318°C, n = 8, p=0.0008, Figure 6—figure supplement 2a–c). Rectal temperature was also significantly increased in Drp1 cKO mice (37.300 ± 0.105°C, n = 8) compared to controls (36.654 ± 0.226°C, n = 8; p=0.0283, Figure 6—figure supplement 2d). Similar to males, the rectal temperature of female Drp1 cKO mice (37.181 ± 0.085°C, n = 7) was significantly greater than that of female controls (36.652 ± 0.142°C, n = 9; p=0.0102; Figure 6—figure supplement 2e).

Ghrelin, a gut-derived hormone secreted during food deprivation, promotes feeding behavior through NPY/AgRP neurons (Andrews et al., 2008). We found a significant decrease in Fos immunoreactivity in AgRP neurons of ghrelin-treated Drp1 cKO mice (30.22 ± 4.652% of AgRP neurons, n = 5, p=0.0001) compared to controls (75.35 ± 3.464% of AgRP neurons, n = 4, Figure 7a–g). No difference in the number of AgRP neurons in the ARC was observed between the two groups (Figure 7h). In agreement with that, ghrelin-induced hyperphagia was not observed in Drp1 cKO mice compared to controls (Figure 7i). Next, we performed patch-clamp whole-cell electrophysiological recordings in slices from Drp1 cKO mice and controls. Consistent with the Fos results, ghrelin significantly increased membrane potential (resting = −46.644 ± 0.502 mV, n = 20; ghrelin = −43.757 ± 0.678 mV, n = 20, p=0.0102; Figure 7j,l) and relative firing activity (resting = 100.000 ± 14.584, n = 22; ghrelin = 186.894 ± 20.266, n = 22, p=0.0041; Figure 7k,l) of AgRP neurons in control mice, while ghrelin-induced excitation of AgRP neurons was significantly attenuated in Drp1 cKO mice compared to controls (membrane potential, resting = −46.104 ± 0.577 mV, n = 19; ghrelin = −45.677 ± 0.644 mV, n = 19, p=0.9962, Figure 7j,l; relative firing activity, resting = 100.000 ± 15.106, n = 20; ghrelin = 119.985 ± 16.502, n = 20, p=0.9644, Figure 7k,l). Of note, no differences in the total (Figure 7—figure supplement 1a) and active form of ghrelin levels (Figure 7—figure supplement 1b) were observed between male control and Drp1 cKO mice in either fed or fasted states. Together, these data suggest that DRP1-mediated mitochondrial fission plays a critical role in regulating ghrelin-triggered AgRP neuronal activity and hyperphagia.

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.

We used the inducible Cre/loxP technology to generate mice in which DRP1 was selectively ablated in AgRP neurons (Drp1 cKO mice). First, mice expressing a tamoxifen-inducible Cre recombinase (CreER^T2) in cells expressing AgRP (Agrp^Cre:ERT2, Wang et al., 2014) were crossed with Rosa26-lox-stop-lox-tdTomato (Ai14; cre-recombinase-dependent expression) mice (Ai14 reporter mice; stock #007914; The Jackson Laboratory, Bar Harbor, ME, USA) to label AgRP-expressing cells. Agrp^Cre:ERT2; Rosa26-lox-stop-lox-tdTomato (Agrp^Cre:ERT2; tdTomato) mice have AgRP-expressing cells with the expression of tdTomato by tamoxifen administration. No observation of AgRP-tdTomato expression was found in the absence of tamoxifen administration, indicating that recombination was strictly dependent upon tamoxifen-induced Cre recombinase activation. The mice with Agrp^Cre:ERT2; tdTomato were then crossed with mice harboring conditional alleles Dnm1l floxed (Dnm1l^fl/fl; Kageyama et al., 2014) to generated mice with inducible deletion of Dnm1l specifically in AgRP neurons (Drp1 cKO mice).

Dnm1l^fl/fl; Agrp^Cre:ERT2; tdTomato mice injected with corn oil and Dnm1l^+/+; Agrp^Cre:ERT2; tdTomato mice injected with tamoxifen (TMX) were used as controls. Dnm1l^fl/fl; Agrp^Cre:ERT2; tdTomato mice were injected intraperitoneally (i.p.) with tamoxifen (0.10 mg/g BW for every 3 days with five times fasting) starting at 5 weeks of age to induce mature-onset deletion of Dnm1l in AgRP neurons of Drp1 cKO mice, and Dnm1l^+/+; Agrp^Cre:ERT2; tdTomato mice were injected with tamoxifen and Dnm1l^fl/fl; Agrp^Cre:ERT2; tdTomato were mice injected with corn oil as control groups. Because we found no differences between these two control groups, the majority of the experiments were performed using Dnm1l^+/+; Agrp^Cre:ERT2; tdTomato and Dnm1l^fl/fl; Agrp^Cre:ERT2; tdTomato mice injected with tamoxifen (to label AgRP neurons with tdTomato expression) as a control and Drp1 cKO group, unless otherwise stated.

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 MgCl₂, and 1 % NP-40 supplemented with 1 mM DTT, 1 mg/ml heparin, 100 µg/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 µl 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 µl/sample; Cat#901513, Biolegend, San Diego, CA, USA).

RNA was extracted using Qiagen RNeasy Plus Micro Kit (Cat# 74034, Qiagen, Valencia, CA, USA) according to the protocol supplied by the manufacturer. cDNA was synthesized using High Capacity cDNA Reverse transcription Kit (Cat# 4368814, Thermo Fisher Scientific, Waltham, MA, USA). qRT-PCR experiment was performed by Taqman Gene Expression Assay primers (Thermo Fisher Scientific) in triplicates using LightCycler 480 Real-Time PCR System (Roche Diagnostics, Mannheim, Germany). All genes were normalized to Actb or Rn18s. The 2(-Delta Delta C(t)) method was used to analyze the relative quantification of gene expression. The following primers were utilized: Dnm1l, Mm01342903_m1; Agrp, Mm00475829_g1; Npy, Mm01410146_m1; Pomc, Mm00435874_m1; Nr5a1, Mm00446826_m1; Actb, Mm02619580_g1; Rn18s, Mm04277571_s1.

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 (O₂ and CO₂; 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 CO₂ production to O₂ consumption. Values were adjusted by body weight to the power of 0.75 (kg−0.75) 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).

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 μm 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.

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 (diluted 1:200 in 0.1M PB, BA-1000, Vector Laboratories, 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. For double-label immunohistochemistry of Fos and POMC neurons, sections were processed using goat anti-Fos antibody (diluted 1:2000 in 0.1 M PB, sc-52-G, Santa Cruz Biotechnology, Dallas, TX, USA) overnight at RT. The following day, brain sections were washed and incubated with a Alexa Fluor 594 donkey anti-goat IgG secondary antibody (diluted 1:500 in 0.1M PB, A11058, Life Technologies, Carlsbad, CA, USA) for 2 hr at RT. Brain sections were then incubated with rabbit anti-POMC antibody (diluted 1:2000 in 0.1 M PB, H-029–30, Phoenix Pharmaceuticals, Burlingame, CA, USA). The following day, sections were washed and incubated with Alexa Fluor 488 donkey anti-rabbit IgG secondary antibody (diluted 1:500 in 0.1 M PB, A21206, Life Technologies) for 2 hr at RT. 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, Inc, Burlingame, CA, USA) and analyzed with a fluorescence microscope.

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-α-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.

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 −1.46 and −2.06 mm from bregma, PVN: −0.70 and −1.06 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).

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 μM). Hypothalamic primary neuronal cultures from Drp1 cKO mice (Dnm1l^fl/fl; Agrp^Cre:ERT2; tdTomato mice) were treated with 4-hydroxytamoxifen (2 μM). 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 μm) 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 × 10⁵ 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 μM 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.

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.

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 μL of suspended cells was loaded into each chamber of the hemocytometer. Counts were performed by triplicate by one analyst under a 40× 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 mm²) in each well (n = 5 per treatment condition) to calculate the cell viability percentage.

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-CO₂ 37°C incubator. Fifteen minutes prior to the assay, 40 μM 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 μM of oligomycin (Cat# 495455, Sigma), 10 μM of FCCP [carbonyl cyanide-p-(trifluoromethoxy) phenylhydrazone] (Cat# C2920, Sigma-Aldrich, Saint Louis, MO, USA), 10 μM of antimycin A (Cat# A8674, Sigma-Aldrich, Saint Louis, MO, USA), and 5 μM 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 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 4°C containing (in mM): sucrose 220, KCl 2.5, NaH₂PO₄ 1.23, NaHCO₃ 26, CaCl₂ 1, MgCl₂ 6, and glucose 10, pH (7.3) with NaOH. After being amputated to a small tissue block, coronal slices containing the hypothalamus (300 μm thick) were cut with a vibratome. After preparation, slices were stored in a holding chamber with an oxygenated (with 5% CO₂% and 95% O₂) artificial cerebrospinal fluid (aCSF) containing (in mM): NaCl 124, KCl 3, CaCl₂ 2, MgCl₂ 2, NaH₂PO₄ 1.23, NaHCO₃ 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.

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 µm) 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.

Mice (5 months old) were deeply anesthetized and transcardially perfused with 0.9% saline containing heparin (10 mg/l), followed by fresh fixative (4% paraformaldehyde, 15% picric acid, 0.1% glutaraldehyde in 0.1 M PB). Brain coronal sections were immunostained with rabbit anti-RFP antibody (diluted 1:1000 in 0.1 M PB, 600-401-379, Rockland Immunochemicals, Limerick, PA, USA) for AgRP neurons. After several washes with 0.1 M PB, sections were incubated with biotinylated goat anti-rabbit IgG (diluted 1:250 in 0.1 M PB, BA-1000, Vector Laboratories, Burlingame, CA, USA) for 2 hr at RT, and then rinsed in 0.1 M PB three times 10 min each time and incubated for 2 hr at RT with avidin–biotin–peroxidase (ABC; diluted 1:250 in 0.1 M PB; ABC Elite kit, Vector Laboratories). The immunoreaction was visualized with 3,3-diaminobenzidine (DAB). Sections were then osmicated (1% osmium tetroxide) for 30 min, dehydrated through increasing ethanol concentrations (using 1% uranyl acetate in the 70% ethanol for 30 min), and flat-embedded in araldite between liquid release-coated slides (Electron Microscopy Sciences, Hatfield, PA, USA). After capsule embedding, blocks were trimmed. Ribbons of serial ultrathin sections were collected on Formvar-coated single slot grids and examined using a Philips CM-10 electron microscope. Mitochondria morphology in AgRP neurons of fed and fasted mice were analyzed using ImageJ software as previously described (Toda et al., 2016).

Five-month-old mice were deeply anesthetized and decapitated. The blood was collected into a capillary tube (Microvette, CB 300 Z, Sarstedt, Nümbrecht, 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.

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.