Orexin-driven GAD65 network of the lateral hypothalamus sets physical activity in mice

Significance

A century ago, it was noted that damage to the lateral hypothalamic (LH) brain area caused people to be motionless most of the time. Therefore, it is thought that this brain area emits essential signals for promoting physical activity. However, it remained a mystery what these signals are. Using newly developed genetic and deep-brain recording methods, we found that, unexpectedly, a critical signal for movement comes from a subset of LH cells that are molecularly distinct from neurons previously implicated in locomotion control. These movement-promoting cells were switched on by the local peptide orexin, a key signal of stress and hunger. These findings shed new light on deep-brain signals that maintain healthy levels of physical activity.

Abstract

Damage to the lateral hypothalamus (LH) causes profound physical inactivity in mammals. Several molecularly distinct types of LH neurons have been identified, including orexin cells and glutamic acid decarboxylase 65 (GAD65) cells, but their interplay in orchestrating physical activity is not fully understood. Here, using optogenetic circuit analysis and cell type-specific deep-brain recordings in behaving mice, we show that orexin cell activation rapidly recruits GAD65_LH neurons. We demonstrate that internally initiated GAD65_LH cell bursts precede and accompany spontaneous running bouts, that selective chemogenetic silencing of natural GAD65_LH cell activity depresses voluntary locomotion, and that GAD65_LH cell overactivation leads to hyperlocomotion. These results thus identify a molecularly distinct, orexin-activated LH submodule that governs physical activity in mice.

The lateral hypothalamus (LH) is thought to provide an essential drive for diverse vital behaviors, including locomotion. Peri-LH lesions in mammals disrupt context-appropriate physical activity, and are associated with the human disorder encephalitis lethargica, which has been noted to make humans “sit motionless…all day” (1–6). The LH is not a homogeneous entity, but contains many molecularly distinct classes of neurons that are thought to have different physiological roles (7). However, the interrelations of distinct LH cell classes in computing and driving context-appropriate physical activity are not fully understood.

For example, LH neurons expressing orexins/hypocretins (8, 9) become activated in diverse stressful contexts, including acute auditory stimulation, hypoglycemia, hypercapnia, and physical capture (3, 9–13). Orexin_LH cell activity may thus represent an important input variable for computing context-appropriate locomotor outputs (3, 14), and orexin peptides have been proposed to increase arousal and locomotion by actions on extrahypothalamic projections to areas such as the locus coeruleus (15, 16). However, hypoactivity phenotypes caused by orexin_LH cell deletion are milder than the hypoactivity phenotypes caused by broader LH lesions (1–3, 5, 12, 13, 17–19), whereas melanin-concentrating hormone LH (MCH_LH) cell deletion causes hyperactivity, implying that MCH_LH cells suppress locomotion (20). These findings suggest that additional drivers of physical activity may exist in the LH.

It was recently found that LH neurons expressing glutamic acid decarboxylase 65 (GAD65) are distinct from orexin and MCH neurons (21). We hypothesized that these cells may be a source of natural LH signals underlying normal levels of physical activity. Here, we investigate this hypothesis by using a combination of cell type-specific manipulation and recording techniques. We find that GAD65_LH cells operate as a stress- and orexin-activated LH module whose physiological activity is essential for normal locomotion, and whose hyperactivity causes hyperlocomotion.

Results

Orexin and Stress Rapidly Recruit GAD65_LH Neurons.

To examine whether orexin cells communicate with GAD65_LH cells, we targeted the light-activated excitatory actuator channelrhodpsin-2 (ChR2) to orexin cells, enabling selective optical activation of orexin-ChR2 cells (22). We then examined resulting optically induced postsynaptic responses in GAD65_LH neurons (Fig. 1A; group data and/or statistics for findings illustrated as figures are described in the figure legends). We found that GAD65_LH cells received time-locked excitatory glutamatergic inputs when orexin_LH cells were stimulated (Fig. 1A). To determine if orexin peptide signaling drives the GAD65_LH network, we also investigated the effect of applying exogenous orexin peptide. We used confocal network imaging of a Cre-dependent calcium indicator, GCaMP6s (23), expressed in LH slices from GAD65-Ires-Cre mice (Fig. 1B). Orexin-A peptide robustly excited the GAD65_LH network, and this excitation persisted (but was slightly reduced) in the presence of a mix of synaptic blockers (Fig. 1B). In parallel, electrophysiological recordings showed that orexin excites individual GAD65_LH cells by activating a current with a reversal potential consistent with a nonselective cationic conductance (Fig. 1C). Overall, >98% of GAD65_LH cells (64 of 65 cells) were activated by orexin peptide in vitro.

Fig. 1.

Evidence for orexin_LH → GAD65_LH excitatory circuit. (A) Optogenetic interrogation of orexin_LH → GAD65_LH signaling. (Left) Targeting strategy. (Center, Left and Right) Localization of orexin-ChR2-mCherry and GAD65-GFP expression (representative example of five brains). (Right) Orexin cell photostimulation (vertical blue line) evokes a CNQX/D-AP5–sensitive (i.e., glutamate receptor-mediated) inward current in a GAD65_LH cell. Gray lines are individual trials, and colored lines are trial averages [representative example of five cells; ≈80% cells (n = 5/6 cells) were connected]. (B) Effects of orexin peptide on GAD65_LH network activity. Scheme for targeting GCaMP6s (Left), example of GCaMP6s expression in GAD65_LH cells (representative example of five brains) (Center Left), GCaMP6s response of GAD65_LH cells to 300 nM orexin-A with corresponding mean ± SEM plot (n = 44) (Center Right), and data summary without (C) and with (B) synaptic blockers (n = 44 and n = 13 cells, respectively, mean ± SEM responses of cells during 2–20 min after orexin-A infusion) (Right). P values (italics) are from sign tests of whether the response within each group is different from zero, and P value (regular font) is from a two-tailed Mann–Whitney test comparing response amplitude between groups. (C) Effects of orexin peptide on GAD65_LH cell electrical activity. (Left) Membrane potential effect of 300 nM orexin-A on a GAD65_LH cell (representative example of six cells). syn., synaptic. (Center) Example of whole-cell current–voltage relation showing effect of orexin-A (representative example of five cells). (Right) Net orexin-activated current (mean and SEM of n = 5 cells).

If orexin cells also activate GAD65_LH cells in vivo, then stress stimuli that elicit orexin_LH cell activity should also activate GAD65_LH cells in vivo. Acute immobilization stress rapidly and robustly activates orexin_LH cells in awake mice (13). We therefore used this paradigm, combined with deep-brain fiber photometry recording of GAD65_LH cell activity and pharmacological antagonism of orexin receptors, to determine if stress and orexin drive GAD65_LH cell activity in vivo. We expressed Cre-dependent GCaMP6s, or control protein EGFP, in the LH of GAD65-Ires-Cre mice, and recorded fluorescence intensity using a fiberoptic probe implanted in the LH (Fig. 2A). We observed bursts of activity in GAD65_LH-GCaMP6s cells, but not in GAD65_LH-EGFP cells, of freely moving mice (Fig. 2A). These bursts were similar (≈5–30% ΔF/F) to the bursts recorded using photometry in other deep networks, and are thought to represent network firing bursts (24–26). Using simultaneous electrical recordings and GCaMP6s imaging from the same GAD65_LH-GCaMP6s cells in vitro, we confirmed that GCaMP6s linearly reports spike firing rate (Fig. 2 B–D), with a mean slope (≈1% ΔF/F per hertz; Fig. 2D) similar to the mean slope observed in other LH neurons (12, 13). Upon acute immobilization, we observed rapid activation of GAD65_LH-GCaMP6s cells, but not control GAD65_LH-GFP cells (Fig. 2E; peak amplitude of stress-induced GAD65_LH-GCaMP6s activity transient was 12.3 ± 2.7%; P = 0.0174 by one-sample t test: t, df = 4.782, 3; n = 4 mice). The competitive orexin receptor antagonist SB-334867 (Materials and Methods) reversibly reduced this stress-induced activation of GAD65_LH cells (Fig. 2E).

Fig. 2.

Stress-induced, orexin-dependent control of GAD65_LH cells in vivo. (A, Left) Schematic of fiber photometry recording. (A, Center) GCaMP6s expression in GAD65_LH cells, and localization of recording site and optical fiber (representative example of five brains). (A, Right) Example fluorescence trace during free cage exploration for mice expressing GCaMP6s (blue) or EGFP (black) in GAD65_LH neurons (typical examples of n = 5 and n = 5 mice, respectively). (B) Simultaneous measurement of membrane electrical activity and calcium level in GAD65_LH-GCaMP6s cells. (Top) Schematic of recording and examples of two video frames from a cell injected with depolarizing current steps. (Bottom) Example of simultaneous cell-attached and fluorescence recording from a GAD65_LH cell (typical example of n = 5 cells). (C) Examples of quantification of data in B, showing the linear relationship between calcium level and spike frequency in four different cells. (D) Group plot of slopes (taken from linear regression fits) of relations between spike frequency and calcium level (n = 5 cells). (E, Left) GAD65_LH-GCaMP6s signals aligned to immobilization stress (i.s.). Heat maps of individual trials from one mouse, in the presence and absence of orexin receptor antagonist SB-334867 (SB; 30 mg/kg i.p.; Materials and Methods), representative example of n = 4 mice. (E, Center) Averaged data for bouts such as those bouts shown on the left. Each trace is the average response of n = 4 mice (each n is a mean of four trials such as those trials shown on the left). The GFP trace shows responses of GAD65-GFP mice (negative controls). (E, Right) Mean calcium signals during i.s. (10–20 s) per mouse (each point represents one mouse, n = 4 mice). One-way repeated measures ANOVA: F(2,6) = 11.61; P = 0.0087, P values on the plot are from Tukey’s posttests.

Collectively, these data identify the orexin system as a physiologically relevant upstream activator of GAD65_LH neurons.

GAD65_LH Cell Activity Drives (and Correlates with) Voluntary Locomotion.

We next tested whether the GAD65_LH module could act as an LH effector of physical activity. To probe causal links between GAD65_LH cell activity and locomotion, we explored locomotor effects of selective silencing or activation of GAD65_LH cell activity in freely moving mice.

First, we targeted the neuroinhibitory clozapine-N-oxide (CNO) receptor hM4Di (27) to the LH of GAD65-Ires-Cre transgenic mice (Fig. 3A). This procedure enabled GAD65_LH cell inhibition by CNO (Fig. 3B). Inhibition of GAD65_LH-hM4Di cells in freely behaving mice substantially reduced their voluntary locomotion (by an average of ≈50%, as shown in Fig. 3 C and D; no such effect was seen in control GAD65_LH-ChR2 mice, in which the CNO-induced change in locomotion was 4.6 ± 8.2%, n = 5, P = 0.6016 by one-sample t test: t, df = 0.0566, 4). Interestingly, orexin receptor blockade with SB-334867 reduced voluntary locomotion under control conditions, but not when GAD65_LH-hM4Di cells were inactivated with CNO (Fig. 3D and Discussion).

Fig. 3.

Effect of natural and evoked GAD65_LH cell activity on locomotion. (A, Left) Targeting scheme for hM4Di-mCherry. (A, Center) Localization of hM4Di-mCherry. (A, Right) GAD65_LH-hM4Di-mCherry cells at high zoom (representative example of five brains). (B, Left) Effect of CNO on GAD65_LH-hM4Di cell firing. Spikes are truncated at 0 mV (representative example of n = 5 cells). (B, Right) Group data, raw values, and means (red); the on-plot P value is from a one-tailed, ratio-paired t test (t, df = 3.251, 4) (n = 5 cells). APs, action potentials. (C) Effect of CNO on locomotion of GAD65_LH-hM4Di mice. Raw data and means are shown in red (Left), and the same data are shown as % change (Right). On-plot P values are from a one-tailed paired t test (t, df = 2.767, 4; Left) and from a two-tailed one-sample t test (t, df = 3.001, 4; Right) (n = 5 mice). Corresponding negative control data are described in Results. (D) Effects of SB-334867 (30 mg/kg i.p.; Materials and Methods) on locomotion of GAD65_LH-hM4Di mice in the absence and presence of CNO. Repeated measures one-way ANOVA with Geisser–Greenhouse correction [F(1.977,11.86) = 16.15; P = 0.0004, on-plot values are P values from Tukey’s multiple comparison tests]. (E, Left) Targeting scheme for hM3Dq-mCherry. (E, Center) Localization of hM3Dq-mCherry. (E, Right) GAD65_LH-hM3Dq-mCherry cells at high zoom (representative example of five brains). (F, Left) Effect of CNO on GAD65_LH-hM3Dq cell firing. Spikes are truncated at 0 mV (representative example of n = 5 cells). (F, Right) Group data, raw values, and means (red); on-plot P value is from a one-tailed ratio-paired t test (t, df = 8.043, 4; n = 5 cells). (G) Effect of CNO on locomotion of GAD65_LH-hM3Dq mice. Raw data and means are shown in red (Left), and the same data are shown as % change (Right). On-plot P values are from a one-tailed paired t test (t, df = 2.638, 4; Left) and from a two-tailed one-sample t test (t, df = 10.31, 4; Right) (n = 5 mice). Corresponding negative control data are described in Results.

Next, we targeted the neuroexcitatory CNO receptor hM3Dq (27) to the LH of GAD65-Ires-Cre transgenic mice (Fig. 3 E and F). Stimulation of GAD65_LH-hM3Dq cells with CNO in freely-behaving mice substantially increased their voluntary locomotion (by ≈150%; Fig. 3G; no such effect was found in control WT mice injected with the same viral vector, in which the CNO-induced change in locomotion was 49 ± 48.7%, n = 5, P = 0.3684 by one-sample t test; t, df = 1.013, 4). Collectively, these data demonstrate that the activity of GAD65_LH cells is necessary for maintaining a normal level of locomotion (Fig. 3 C and D), and sufficient for stimulating locomotion (Fig. 3G).

Finally, we reasoned that if GAD65_LH cells naturally promote locomotion, then their activity should be associated with spontaneous/voluntary locomotion. To test this hypothesis, we measured GAD65_LH-GCaMP6s network activity using fiber photometry, and quantified locomotion using a sensitive rotary treadmill, which could resolve distances traveled as low as 1.8 mm (Fig. 4A and Materials and Methods). We found that spontaneously initiated locomotion bouts (defined as locomotion with a speed of >0.17 cm⋅s⁻¹; Materials and Methods) were associated with increased natural GAD65_LH cell activity (Fig. 4 B and C). Locomotion-associated GAD65_LH cell activity appeared to increase before movement was detected, and persisted throughout the locomotion bouts (Fig. 4 A–C). To estimate the temporal relationship between the natural GAD65_LH cell signals and locomotion initiation, we defined signal onset time using a third derivative method (28), and also defined the steepest point of GAD65_LH signal rise as the peak in the first derivative (Fig. 4D). The steepest rise of the perilocomotion GAD65_LH signal occurred approximately at movement onset, whereas the onset of GAD65_LH signal rise occurred ≈2 s before movement onset (Fig. 4E and Discussion).

Fig. 4.

Natural GAD65_LH cell dynamics underlying voluntary movement. (A) Scheme of experimental setup, with example raw data illustrating temporal alignment of locomotion initiation (upward deflections on the orange trace) and GAD65_LH signal (typical example of n = 4 mice). (B) Example of rotary encoder pulses converted to locomotion velocity (orange), illustrating coquantification of natural GAD65_LH cell dynamics (blue) and locomotion (typical example of n = 4 mice). (C) Peri-event plots aligned to onset of locomotion bouts [dashed black line indicates the threshold locomotion criterion (0.17 cm·s⁻¹) for bout onset]. (Top) Heat maps of individual trials from one mouse. (Bottom) Group data for bout averages in GAD65_LH-GCaMP6s and GAD65_LH-GFP (mean ± SEM). Statistics for data binned in 2-s bins: one-way repeated measures ANOVAs for control GFP group [n = 3 mice; F(9,18) = 0.8308, P = 0.5974] and for GCaMP6s group [n = 4 mice; F(9,27) = 4.630, P = 0.0009]. (D) Illustration of the third and first derivative methods used to measure the timing of GAD65_LH-GCaMP6s signal onset (ON) and steepest rise (S) relative to locomotion onset (dashed black line, defined as in B) (typical example of n = 4 mice). (E) Quantification of latencies of GAD65_LH-GCaMP6s signal onsets and steepest rises to locomotion initiation. Signal onset occurred at −2.22 ± 0.29 s (P < 0.0001 by two-tailed one-sample t test; t, df = 7.572, 11). The steepest rise occurred at −0.35 ± 0.22 s (P = 0.3139 by two-tailed one-sample t test; t, df = 1.055, 11).

Collectively, these data provide correlative (Fig. 4) and causal (Fig. 3) evidence that GAD65_LH cell activity drives locomotion in mice.

Discussion

Despite the long-recognized importance of the LH for maintaining normal physical activity in mammals, including humans (1–6, 29, 30), the functional relations of molecularly distinct LH subnetworks remained poorly understood. Our study identifies several properties of the GAD65_LH subnetwork that reveal a previously overlooked interplay between molecularly distinct LH signals relevant for locomotion control. First, GAD65_LH cells are activated by orexin neurons, and are rapidly recruited by stress in vivo in an orexin receptor-dependent manner. Second, natural GAD65_LH cell activity is necessary for maintaining normal levels of locomotion, whereas experimentally induced GAD65_LH cell activity is sufficient to promote locomotion. Third, internally initiated GAD65_LH network bursts are associated with initiation and maintenance of voluntary movement. When GAD65_LH cells were inactivated, reducing orexin receptor activity no longer affected locomotion (Fig. 3D), consistent with the idea that GAD65_LH cells act as mediators of locomotion driven by natural orexin signaling. These multiple lines of evidence demonstrate a critical role for GAD65_LH cells in driving normal locomotion, and in acting as effectors of the orexin system.

In known brain controllers of voluntary action, such as the basal ganglia, neural signals implicated in movement control appear seconds before action initiation (e.g., figure 3 E and F of ref. 24). These delays are comparable to the delays we observed between the onset of the GAD65_LH network bursts and movement initiation (≈2 s; Fig. 4E). These long delays presumably imply multiple downstream processing stages between GAD65_LH cell signals and motor commands that drive muscle activity underlying locomotion. When we attempted to determine downstream targets of GAD65_LH neurons by tracing their axon distribution in brains of GAD65_LH-ChR2-YFP mice, we found projections to multiple neural systems, extending from the LH to the thalamus, ventral tegmental area, cortical sensory areas, brainstem, and many hypothalamic nuclei (Fig. S1). A comprehensive analysis of these downstream effectors of GAD65_LH cells will be included in future studies, because it is a large task requiring the relative locomotor impacts of GAD65_LH cell targets to be deciphered. A specific question for such future work would be whether the locomotion drivers described here, and the classical basal ganglia pathways, operate in series or in parallel to promote physical activity.

Fig. S1.

Axonal projections of GAD65LH cells. (A, Left) Scheme for targeting ChR2-YFP to GAD65_LH cells. (A, Right) DAPI-stained (blue) 30-μm sagittal section of the whole brain showing GAD65_LH projections (green) to brain areas like the accessory olfactory bulb (AOB), LH, thalamus (TH), midbrain (MB), and medulla (representative example of three brains). (B) High magnification of brain areas where dense GAD65-ChR2-YFP axons are found (representative example of three brains).

Related questions for future studies are what neurotransmitters GAD65_LH cells release at their many projection targets, and what regulatory receptors they express. GAD65_LH cells do not express orexin or MCH neurotransmitters (21); note that this finding does not contradict previous reports of GAD67 expression in MCH cells, because GAD65 and GAD67 are not always coexpressed in the hypothalamus (e.g., refs. 31, 32). We also found that the GAD65_LH cells are distinct from the previously studied LH neurons expressing neuropeptide Y (NPY) or leptin receptor (33, 34) (Fig. S2 A and B). It remains to be determined whether GAD65_LH cells release excitatory and/or inhibitory transmitters, because vesicular GABA transporter (VGAT) is not always expressed in GAD65 neurons, and some hypothalamic GAD65 neurons contain glutamatergic markers (31, 32, 35). Indeed, we found that only 50% of the GAD65_LH cells contain VGAT, and that most (≈80%) VGAT_LH cells are distinct from GAD65_LH cells (Fig. S2 C and D). Thus, the previous studies of VGAT_LH cells (36–39) likely targeted larger and more heterogeneous cell populations than our study, which was selective for GAD65_LH cells. This fact may explain why the VGAT_LH cell stimulation drove different behaviors [increased eating and insignificant effect on locomotion (36, 39)] than the GAD65_LH cell stimulation (increased locomotion and no significant effect on eating; Fig. 3 and Fig. S3). The latter findings presumably imply that GAD65_LH-stimulated locomotion is unlikely to be specifically motivated by hunger or directed toward a food goal. We note that in some contexts, it may be evolutionarily advantageous to develop a locomotion drive not directed toward a specific goal (e.g., food), but toward a more general goal (e.g., change of place, physical escape). It is thus tempting to speculate that the orexin-activated GAD65_LH cells may provide such a general locomotion drive that serves, for example, to move the animal away from stresses that activate orexin neurons (3, 10, 11, 13). Such an escape-like behavioral drive would also be consistent with the observed GAD65_LH cell dynamics during capture, when the mouse is immobilized but the drive is high.

Fig. S2.

Neurochemical characteristics of GAD65_LH cells. (A) Colocalization of GAD65_LH cells and pSTAT3 (leptin receptor) LH cells. GAD65_LH cells were labeled with GFP (GAD65-GFP transgenic mouse), and pSTAT3 neurons were labeled with Alexa Fluor 555 antibodies (Materials and Methods). A total of 11.45% of pSTAT3_LH neurons contained GAD65 (analysis of 1,544 pSTAT neurons from three GAD65-GFP/pSTAT3-Alexa555 brains), and 6.89% of GAD65_LH cells contained pSTAT3 (analysis of 819 GAD65 neurons from three GAD65-GFP/pSTAT3-Alexa555 brains). The % colocalization values are averages per hemisphere. (B) Colocalization of GAD65_LH cells and NPY_LH cells. NPY cells were labeled with GFP (NPY-hrGFP transgenic mouse), and GAD65-Ires-Cre cells were labeled with ChR2-mCherry. A total of 2.06% of NPY_LH neurons contained GAD65 (analysis of 705 NPY_LH neurons from three NPY-hrGFP/GAD65-mCherry brains), and 0.78% of GAD65_LH cells contained NPY (analysis of 1,645 GAD65_LH neurons from three NPY-hrGFP/GAD65-mCherry brains). The % colocalization values are averages per hemisphere. (C) Colocalization of GAD65_LH cells and VGAT_LH cells. GAD65_LH cells were labeled with GFP (GAD65-GFP transgenic mouse), and VGAT-Ires-Cre cells were labeled with tdTomato (CAG-tdTomato;VGAT-Ires-Cre transgenic mouse) or ChR2-mCherry (Materials and Methods). (Left) Example of LH colocalization. (Right and Bottom) More examples in coronal slices from different anteroposterior LH locations (bregma coordinates indicated on the slides). (D) Quantification of data in A (combined cell counts from three brains).

Fig. S3.

Effects of GAD65_LH cell stimulation on eating and body weight. (A) Scheme for targeting the excitatory CNO receptor hM3Dq (or control protein ChR2-YFP) to GAD65_LH cells. Targeting was confirmed by histology, as shown in Fig. 3E. (B) Effect of CNO on food intake and body weight in GAD65-Ires-Cre mice expressing ChR2 or hM3Dq in GAD65_LH cells (five mice per group). Values are mean ± SEM. CNO was injected daily (0.5 mg/kg i.p.) during the indicated days. (C) Data from B, analyzed with two-way repeated measures ANOVA. pre, day 5; CNO, day 9. (Top) Interaction [F(1,8) = 0.01474, P = 0.9064]. ns, not significant. (Bottom) Interaction [F(1,8) = 13.47, P = 0.0063; numbers above bars are P values from Tukey’s multiple comparison test].

In summary, we found that physiological activity of GAD65_LH cells is vital for normal locomotion and identified orexin as a physiological upstream regulator of GAD65_LH activity. These findings establish the GAD65_LH cells as key players in intra-LH information processing. Together with recent findings on roles of defined LH neurons in eating and arousal (7, 36, 38), our results substantiate specific LH circuits as orchestrators of key voluntary actions.

Materials and Methods

Gene Transfer.

All procedures followed United Kingdom Home Office regulations and were approved by the Animal Welfare and Ethical Review Panel of the Francis Crick Institute. Adult male and female mice (at least 8 wk old) were used for in vitro experiments. Male mice were used for behavioral experiments, and kept on a standard 12-h/12-h light/dark cycle and on standard mouse chow and water ad libitum. Mice were anesthetized with isoflurane and injected with meloxicam (2 mg/kg of body weight, s.c.) for analgesia. After placement into a stereotaxic frame (David Kopf Instruments), a craniotomy was performed and a borosilicate glass pipette was used to inject viral vectors bilaterally into the LH. Two injections (each 75 nL, delivered at 75 nL⋅min⁻¹) were made into the LH in each hemisphere (bregma: −1.30 mm, midline: ±1 mm, from brain surface: 5.20 mm and 5.25 mm). Mice were allowed to recover for at least 1 wk after surgery while single-housed. Cre-dependent designer receptors exclusively activated by designer drugs (DREADD) constructs (40, 41), rAAV8/hSvn-DIO-hm3D(Gq)-mCherry [2.2 × 10¹² genome copies (gc)/mL; UNC Vector Core] or rAAV8/hSyn-DIO-hM4(Gi)-mCherry (5.3 × 10¹² gc/mL; UNC Vector Core), were used to measure the effect of cell type-specific manipulation on locomotion. For ChR-assisted circuit mapping, “FLEX switch” ChR2 constructs (42) were used. These constructs were either AAV1.EF1.flox.hChR2(H134R)-mCherry.WPRE.hGH (8.78 × 10¹² gc/mL) or AAV1.EF1.DIO.hChR2(H134R)-YFP.WPRE.hGH (6.2 × 10¹² gc/mL). For calcium imaging, we used a GCaMP6s construct (23), rAAV9.CAG.Flex.GCaMP6s.WPRE.SV40 (2.74 × 10¹³ gc/mL). Previously characterized and validated transgenic mouse lines (or their crosses) were used where indicated: GAD65-Ires-Cre mice (43) bred in homozygous (hom)-WT pairs with C57BL/6 mice, orexin-Cre mice (22) bred in heterozygous (het)-WT pairs with C57BL/6 mice, GAD65-GFP mice (21) bred in het-WT pairs with C57BL/6 mice, NPY-hrGFP mice (44) bred in het-WT pairs with C57BL/6 mice, VGAT-Ires-Cre mice (45) bred in hom-WT breeding pairs with C57BL/6 mice, and CAG-tdTomato mice (46) bred in hom-hom pairs.

Cell Type-Specific Deep-Brain Recordings in Awake Behaving Mice.

After injection of Cre-dependent GCaMP6s into GAD65-Ires-Cre mice, the fiberoptic implants were stereotaxically installed with the fiber tip above the LH (1.38 mm caudal from bregma, 0.95 mm lateral from midline, and 5 mm ventral from brain surface) and fixed to the skull as previously described (12, 13); this method is expected to capture signals only from within 500 μm of the fiber tip (12). Fiber tip locations were verified in each mouse by examining slices with a visible fiber tract. For locomotion experiments and immobilization, mice also had a custom-built, stainless-steel head-restraint implant glued to their skulls using tissue-friendly cyanoacrylate (Histoacryl; TissueSeal). Fiber photometry was performed as described by González et al. (13). Fluorescence signals were converted to ΔF/F (%) values as follows: ΔF/F = 100 * (Fr − F)/F, where Fr is the raw signal and F is the trial mean.

In Fig. 2E (immobilization stress experiments), mice were placed in a recording cage and left for 10 min; then, a trial consisting of 10 s of restraint stress (holding the mouse head stationary inside the cage by clamping the head-fixed restraint) was performed, based the method of González et al. (13). For each mouse, four trials per day were performed (with intertrial intervals >5 min). Trial blocks for drug condition (control, SB-334867, and recovery) were separated by 48 h. The competitive orexin receptor antagonist SB-334867 (or 0.9% NaCl and 10% DMSO as a control vehicle) was given i.p. 1 h before immobilization stress. This antagonist has a higher affinity for orexin type-1 receptor than for orexin type-2 receptor, but at higher concentrations, it antagonizes orexin binding to both receptors (47). We used a single high dose of SB-334867 (30 mg/kg), based on previous studies with this compound (48, 49). To control for circadian factors, all experiments were performed during the dark phase.

Quantifying Physical Activity.

Locomotor activity was assessed after a single i.p. dose of CNO or saline during the dark-phase in open home cages using either video-tracking (Anymaze or Ethovision) or a rotary treadmill. Mice of the same genotype were grouped according to vectors injected into their brains, and all groups had a similar composition based on sex (male), age, and body weight. For treadmill locomotion quantification, mice were head-restrained in a sound-attenuated chamber where they could run on a disk treadmill with a rotary encoder. The center of the animal’s head was such that 1° of rotation equated 1.8 mm of forward locomotion. The encoder generated a transistor-transistor logic (TTL) pulse every time the animal moved 1.8 mm. The pulses were collected by the same analogue-to-digital converter as the photodetector signals for accurate time-stamping of GAD65-GCaMP6s signals and locomotion. Locomotion bout onset was defined as the time of crossing a velocity threshold of 0.17 cm⋅s⁻¹, and locomotion bouts of at least 5 s in duration from this onset are illustrated in the analyses shown in Fig. 4C.

Brain Slice Optogenetics, Electrophysiology, and Imaging.

Patch-clamp recordings combined with optogenetics were carried out as reported by Schöne et al. (50). Briefly, LH slices were prepared at least 2 months after virus injection. Coronal brain slices of 250-μm thickness containing the LH were cut while immersed in ice-cold slicing solution. Slices were incubated for 1 h in artificial cerebrospinal fluid (ACSF) at 35 °C, and then transferred to a submerged-type recording chamber. Neurons containing fluorescent markers were visualized with an Olympus BX61WI microscope with an oblique condenser and fluorescence filters. Excitation light for ChR2 was delivered from a LAMBDA DG-5 beam switcher (Sutter) with a xenon lamp and ET470/40-nm bandpass filter. A 40× 0.8-N.A. objective was used to deliver 1-ms flashes of blue light (∼10 mW/mm²) onto ChR2-containing axons around the recorded cell, and postsynaptic responses were recorded at a range of potentials (0 to −90 mV) in each cell. For studying effects of orexin on membrane current (Fig. 1C), voltage-clamp ramps from 0 to −120 mV were performed at 0.1 mV⋅ms⁻¹, and orexin-evoked current was then derived in each cell by subtracting the baseline current from the orexin current at each potential and plotted at 10-mV increments for visual clarity (Fig. 1C, Right). Functional ChR2 expression was confirmed by recording light-activated action potentials in the target cells (n = 3 cells per group). For calcium imaging, brain slices were placed in a recording chamber of a BX61WI Olympus microscope controlled by Olympus Fluoview software (FV10-ASW version 4.0), perfused at 35 °C with ACSF. Confocal imaging was performed at 0.5 Hz with an Olympus 20× 0.50-N.A. objective, 488-nm argon laser excitation, and 500- to 545-nm emission collection. Motion and bleach corrections were applied if needed (StackReg plug-in, ImageJ; NIH). A region of interest (ROI) containing each GCaMP6s-positive neuron was selected via ROI manager in ImageJ. The mean fluorescence of each ROI was extracted for each frame. These raw fluorescence (Fr) values were normalized to produce ΔF/F values, where F is the mean baseline before orexin application and ΔF = (Fr − F). Simultaneous brain slice cell-attached recordings and calcium imaging (Fig. 2 B and C) were performed as reported by González et al. (13). Because the cell-attached experiments, which were necessary to preserve cytosolic integrity and thus in vivo-like GCaMP signals, do not measure transmembrane potential, it is possible that the GCaMP signal is increased by interspike depolarization as well as by the depolarization during the spikes.

Chemicals and Solutions.

For brain slice recordings, ACSF and ice-cold slicing solution were gassed with 95% O₂ and 5% CO₂, and contained the following: 125 mM NaCl ACSF, 2.5 mM KCl, 1 mM MgCl₂, 2 mM CaCl₂, 1.2 mM NaH₂PO₄, 21 mM NaHCO₃, 2 mM d-(+)-glucose, 0.1 mM Na⁺-pyruvate, and 0.4 mM ascorbic acid. The slicing solution contained 2.5 mM KCl, 1.3 mM NaH₂PO·H₂0, 26.0 mM NaHCO₃, 213.3 mM sucrose, 10.0 mM d-(+)-glucose, 2.0 mM MgCl₂, and 2.0 mM CaCl₂. For standard whole-cell recordings, pipettes were filled with intracellular solution containing the following: 120 mM K-gluconate, 10 mM KCl, 10 mM Hepes, 0.1 mM EGTA, 4 mM K₂ATP, 2 mM Na₂ATP, 0.3 mM Na₂GTP, and 2 mM MgCl₂ (pH 7.3) with KOH. Concentrations used were as follows: 300 nM orexin-A; 5 μM CNO; synaptic blocker mix: 50 μM D-AP5, 10 μM picrotoxin, 10 μM CNQX, and 10 μM CGP 35348; and 3 μM gabazine. For in vivo chemogenetic manipulations, CNO was injected i.p. at 5 mg/kg (experiments involving hM4Di) or at 0.5 mg/kg (experiments involving hM3Dq). All chemicals were from Sigma or Tocris Bioscience.

Histochemistry.

For the immunolabeling of pSTAT3 in Fig. S2, 30-μm cryosections of leptin-injected GAD65-GFP mice were pretreated with H₂O₂, NaOH, glycine, and SDS. Staining for pSTAT3 was done with rabbit anti-pSTAT3 lgG (no. 9131, 1:500; Cell Signaling Technology) sera overnight as the primary antibody and Alexa 555-conjugated donkey to rabbit IgG (A-31572, 1:500; Life Technologies) as the secondary antibody (51). In chemogenetic experiments, actuator targeting to LH (Fig. S4) was verified in fixed sections prepared as reported by González et al. (12).

Fig. S4.

Examples of coronal LH sections from four GAD65-Ires-Cre mice injected with Cre-dependent DREADD-mCherry, illustrating the typical extent of LH labeling in our study.

Statistical Analysis.

Statistical tests and descriptive statistics were performed as specified in the figure legends. In each experimental dataset at the cellular level, each n was a different cell (no repeated trials from the same cell were used as n values) and cells from at least three mice were analyzed. Before performing parametric tests, data were assessed for normality with a D’Agostino–Pearson omnibus test or Kolmogorov–Smirnov test for small sample sizes, and variances were assessed for homogeneity with a Brown–Forsyth test. To compare interactions within normally distributed data with repeated measurements, repeated measures ANOVA was used, with multiple comparison tests where appropriate. Analysis was performed with GraphPad Prism, MATLAB (The MathWorks, Inc.), and ImageJ.