A clock-dependent brake for rhythmic arousal in the dorsomedial hypothalamus

WAKE is conserved from flies to humans (Supplementary Fig. 1a), and our published and ongoing work in Drosophila and mice suggest that WAKE labels rhythmic neural networks^19,20. The DMH has been implicated in the circadian control of arousal, and mWAKE (also named ANKFN1/Nmf9)²³ is expressed in this region in mice²¹. Thus, we hypothesized that DMH^mWAKE neurons constitute a key rhythmic arousal-promoting cell population. To address the percentage of DMH neurons that are mWAKE⁺, we performed immunostaining of DMH tissue, labeling neurons with anti-NeuN antibodies and mWAKE using anti-V5 antibodies in a validated mWAKE^V5 transgenic mouse line²¹. These experiments suggest that DMH^mWAKE neurons comprise 11.6 ± 0.2% (n = 3 brains, 11 total sections) of the total population of DMH neurons (Supplementary Fig. 1b).

We next examined whether DMH^mWAKE Ca²⁺ transient peaks detected by fiber photometry also varied according to sleep/wake state and whether they differed between day vs night. The frequency of pooled transients was greater during wakefulness, compared to NREM and REM sleep (Fig. 1e), and the increased transients during wakefulness was more pronounced at night (Fig. 1f). In contrast, transient prominence was higher during REM sleep (Supplementary Fig. 1g). Finally, we assessed signal intensity changes across transitions between consolidated sleep/wake states (Fig. 1g). DMH^mWAKE GCaMP signal increased prior to the state change from NREM to wakefulness and from NREM to REM sleep and decreased prior to the state change from wakefulness to NREM sleep. In contrast, DMH^mWAKE GCaMP signal decreased following transitions from REM sleep to wakefulness, suggesting that these cells may be less important for driving those state changes. Collectively, these data demonstrate that the overall activity of DMH^mWAKE neurons is increased during wakefulness and REM sleep and that Ca²⁺ transients are elevated during wakefulness, particularly at night. These findings support the notion that DMH^mWAKE neurons are involved in the production of rhythmic arousal.

We next examined the effects of silencing DMH^mWAKE neurons on sleep/wake states. We chemogenetically inhibited DMH^mWAKE neurons, by injecting an AAV vector encoding Cre-dependent DREADD-hM4Di (AAV-DIO-hM4D-Gi) into mWake^(Cre/+) mice (Fig. 2e; Supplementary Fig. 2g, h; Supplementary Table 1). Injection of CNO at ZT10 significantly increased NREM sleep, with accompanying reductions in wakefulness and, to a lesser degree, REM sleep (Fig. 2f, g; Supplementary Fig. 2i, j). Locomotor activity was also reduced with chemogenetic inhibition of DMH^mWAKE neurons (Fig. 2h). Together, these data suggest that DMH^mWAKE neurons are both necessary and sufficient for promoting arousal and wakefulness.

While historically there has been substantial focus on the role of neuromodulatory circuits in regulating arousal^27,28, accumulating evidence has highlighted the importance of both glutamatergic and GABAergic subcortical neuronal groups in this process^28–31. Thus, we next used the INTRSECT approach³² to try to isolate GABAergic or glutamatergic DMH^mWAKE subpopulations and investigate their role in regulating arousal. We first confirmed that we could selectively manipulate GABAergic mWAKE^DMH neurons, by stereotactically injecting AAV-Con/Fon-EYFP virus into the DMH of mWake^(Cre/+); Vgat^(Flp/+) mice and performing immunostaining and RNAscope in situ hybridization (Supplementary Fig. 4a-c; Supplementary Table 1).

However, using a related approach for glutamatergic DMH^mWAKE neurons (AAV-Con/Fon-ChR2-EFYP virus with mWake^(Cre/+); Vglut2^(Flp/+) mice) resulted in little to no expression in the DMH (Supplementary Fig. 4d; Supplementary Table 1). Thus, we next tried injecting AAV-Con/Foff-EFYP virus into the DMH of mWake^(Cre/+); Vgat^(Flp/+) mice, but found in this case that viral expression was observed in substantial numbers of both glutamatergic and GABAergic DMH^mWAKE neurons (Supplementary Fig. 4f, g; Supplementary Table 1). We further tried to address this technical limitation by employing the recently developed rTARGIT system, where a tetracycline transactivator (tTA) is used to boost expression in a dual recombinase system (AAV-hSYN1-fDIO-rTTA and AAV-TRE-DIO-ChR2-EYFP with mWake^(Cre/+); Vglut2^(Flp/+) mice)³³. However, even this approach did not yield significant expression in glutamatergic DMH^mWAKE neurons (Supplementary Fig. 4e; Supplementary Table 1). Thus, we were able to selectively manipulate GABAergic, but not glutamatergic, DMH^mWAKE neurons.

We next analyzed the projection patterns for DMH^mWAKE vs GABAergic DMH^mWAKE neurons in these mice (Supplementary Fig. 6). Broadly speaking, the GABAergic subset appeared to target similar regions as DMH^mWAKE neurons as a whole, with many of these regions implicated in the regulation of arousal and sleep (i.e., preoptic area/POA, lateral hypothalamus/LH, dorsomedial hypothalamus/DMH, paraventricular thalamic nucleus/PV, lateral habenula/LHb, ventral tegmental area/VTA, posterior hypothalamus/PH, periaqueductal gray/PAG, and locus coeruleus/LC)^30,34–36 (Fig. 4b, e; Supplementary Fig. 6a). However, there were a few notable differences; GABAergic DMH^mWAKE neurons exhibited absent or substantially weaker projections to the zona incerta (ZI), medial mammillary (MM) region, and the raphe pallidus nucleus (RPA) (Supplementary Fig. 6b-g). Taken together, these data suggest that GABAergic DMH^mWAKE neurons promote arousal.

To test the function of mWAKE on sleep and arousal, we generated a null allele of mWake (mWake^(-)) using CRISPR/Cas9 (Fig. 5e and see Methods). As expected, mWake expression, as assessed by quantitative PCR (Fig. 5f) and in situ hybridization (ISH)²¹, was markedly reduced in mWake^(-/-) mice, likely due to nonsense-mediated decay. We next assessed sleep in mWake^(-/-) mice via EEG/EMG recordings. Under light:dark (LD) or constant dark (DD) conditions, there was either subtle or no differences in the amount of wakefulness, NREM, or REM sleep between mWake^(-/-) mutants and wild-type (WT) littermate controls (Supplementary Fig. 7a, b). However, there was a change in the distribution of wakefulness at night; mutants spent more daily time in prolonged wake bouts, with some (33%) exhibiting dramatically long (>6 h) bouts of wakefulness (Supplementary Fig. 7c, d). These data suggest that mWake mutants exhibit changes in the quality, but not quantity, of wakefulness at night.

Alterations in arousal level can be quantified across different parameters, including sleep/wake behavior, locomotor activity, and responsiveness to sensory stimuli³⁷. Thus, we next examined baseline homecage locomotor activity. Surprisingly, mWake^(-/-) mutants were markedly hyperactive during the subjective night compared to littermate controls, although a mild but significant increase in locomotor activity was also noted during the subjective day (Fig. 5g, h and Supplementary Movie 3). To rule out the possibility of 2nd site mutations causing this phenotype, we examined heteroallelic mWake^(Nmf9/-) mutants, which also demonstrated robust locomotor hyperactivity during the night, but not during the day (Fig. 5h). Under LD conditions, mWake^(-/-) and mWake^(Nmf9/-) mice also demonstrated increased locomotor activity during the night, although a mild, but statistically significant increase in locomotion was also seen in mWake^(-/-) mice during the daytime (Supplementary Fig. 8a, b). Locomotor activity for heterozygous mWake^(+/-) mice was not different from littermate controls (Fig. 5h). The variability of the pronounced nighttime locomotor activity in mWake^(-/-) and mWake^(Nmf9/-) mutants was driven by intense stereotypic circling behavior in a subset (~35%) of these animals (Supplementary Movie 3). These “circlers” tended to also exhibit pronounced theta activity on EEG spectrograms (Supplementary Fig. 7c), which has been associated with “active” wakefulness³⁸. We next characterized sensory responsiveness in mWake^(-/-) mutants by evaluating startle response to an acoustic stimulus during the subjective day and subjective night. mWake^(-/-) mutants exhibited an increased startle response to 100 dB and 110 dB stimuli during subjective night, but not during the subjective day (Fig. 5i, j). Taken together, these data suggest that the clock acts through mWAKE to paradoxically attenuate arousal during the active period of mice.

In Drosophila, WAKE labels rhythmic neurons and is required for their cyclical activity^19,20. Thus, we postulated that mWAKE regulates the excitability of DMH^mWAKE neurons in a time-dependent manner. To address this hypothesis using electrophysiological approaches, we needed a genetic method to visualize mWake⁺ neurons while simultaneously generating a mutant mWake allele. Our mWake^Cre line labels mWake⁺ neurons²¹ and should function as a null allele, because exon 5 is replaced by a tdTomato-P2A-Cre cassette. To test this, we assessed locomotor activity of mWake^(Cre/Cre) animals vs heterozygous controls in DD and found that the homozygous animals phenocopied the nighttime hyperactivity of mWake^(-/-) and mWake^(Nmf9/-) mutants, demonstrating that mWake^Cre is a bona fide mWake mutant allele (Supplementary Fig. 9b, c).

We performed whole-cell patch-clamp slice recordings and found that DMH^mWAKE neurons exhibited greater spiking frequency during the night vs the day, a rhythm which is aligned with the active phase of the nocturnal mouse (0.55 ± 0.10 Hz at ZT0-2 vs 1.16 ± 0.15 Hz at ZT12-14 in mWake^(Cre/+), P < 0.01). Interestingly, intrinsic excitability of these neurons was reduced during the night (Supplementary Fig. 9d). Loss of mWAKE did not significantly alter spontaneous firing of DMH^mWAKE neurons during the daytime. However, the nighttime increase in spiking frequency seen in controls was markedly enhanced in mWake^(Cre/Cre) mutants (Fig. 6e–g and Supplementary Table S2). Similarly, compared to heterozygous controls, intrinsic excitability of DMH^mWAKE neurons was greater in mWake^(Cre/Cre) mutants during the night, but not the day (Fig. 6h). Together, these findings suggest that mWAKE acts in the DMH as a brake on arousal at night, by inhibiting neural excitability specifically during that time.

All animal procedures were approved by the Johns Hopkins Institutional Animal Care and Use Committee. All animals were group-housed and maintained with standard chow and water available ad libitum. Animals were raised in a common animal facility under a 14:10 h Light:Dark (LD) cycle. Animals were acclimatized to a 12:12 LD schedule for at least 7 days or 14 days prior to timed molecular experiments or behavioral assays, respectively. Animals were housed under 68-79°F and 30-70% relative humidity conditions. Unless otherwise noted, adult male mice (2–4 months) were used in all experiments. All mouse strains were backcrossed to C57BL/6 at least seven times prior to use in behavioral experiments. Genotyping was performed either by in-house PCR and restriction digest assays, or via Taq-Man based rtPCR probes (Transnetyx). mWake^V5 mice were previously described²¹. mWake^Nmf9 were obtained from B. Hamilton (University of California, San Diego). Bmal1^– (009100), Vglut2^Flp (030212), and Vgat^Flp (029591) mice were obtained from The Jackson Laboratory.

The mWake null mutant allele (mWake^–) was generated by CRISPR/Cas9 genome editing (Johns Hopkins Murine Mutagenesis Core), using a targeting guide RNA (gRNA: CGC AGA AGA ATC CTC GCA AT) to the 4th exon of mWake and a 136 bp oligonucleotide (AGT GCG GAC TTT CTC TGG CTC CTG TCC GCA GAA GAA TCC TCG GCG GAA TTC AAT GGG CAC GTT GTT GGT CAT GAT GGC GAT GTC CAG GGG TGT CAG CCC TTC GCT GTT CGG TGT) containing two 64 bp homology arms, surrounding an 8 bp insertion (GCGGAATT), which includes an in-frame stop codon and induces downstream frameshifts. Exon 4 is predicted to be in all mWake splice isoforms. The conditional mWake knockout allele (mWake^flox) was generated via homologous recombination in hybrid (129/SvEv x C57BL/6) mouse embryonic stem cells (Ingenious Targeting Laboratory), using a construct containing two loxP sites and a Neomycin cassette flanking the 5^th exon of mWake. A transgenic mouse line expressing both Cre recombinase and tdTomato in mWake⁺ cells (mWake^Cre) was generated via homologous recombination in hybrid (129/SvEv x C57BL/6) mouse embryonic stem cells (Ingenious Targeting Laboratory). The knock-in vector was integrated 21 bp into exon 5 of the mWake locus, replacing the remainder of the exon with a tdTomato-P2A-split Cre-Neo-WPRE-BGHpA cassette, which causes frameshifted nonsense mutations downstream and producing an mWake loss-of-function allele.

To quantify mWake transcript in the mWake^(-/-) mutant mice, qPCR was performed. Control and mWake^(-/-) mutant mice were anesthetized by the isoflurane drop method and then decapitated. Hypothalami and hindbrains were dissected at ~ZT0, and RNA was extracted using Trizol Reagent (Invitrogen). qPCR was performed using a SYBR PCR master mix (Applied Biosystems) and a 7900 Real-Time PCR system (Applied Biosystems), with the following primers, which target exon 4: mWake-F: 5’-CCC TAA CGG TCA GCT TTC AAG A-3’ and mWake-R: 5’-GAC ATG CTC CAT TCC ACT TTG TAC-3’. GAPDH was used as an internal control. Ct value was compared against regression standard curve of the same primers. 3 biological replicates were performed.

mWake^(V5/V5) or mWake^(V5/V5); Bmal1^(-/-) male mice were anesthetized by the isoflurane drop method and then decapitated. Hypothalami were dissected, and protein was extracted by homogenizing the tissue in RIPA solution (Millipore Sigma, R2078) with protease inhibitor cocktail (Millipore Sigma, P8340, 1:100) and incubating for 1 h on ice. Lysates were cleared by centrifugation, and protein concentration was measured using the Pierce BCA protein assay kit (ThermoFisher Scientific, 23227). Samples were then prepared in sample buffer and, after incubating at 70 °C for 10 min, 20 μg protein was run on SDS-PAGE. Proteins were transferred to PVDF membrane for 1.5 hrs at 100 V, blocked for 1 h with Intercept Blocking Buffer (LI-COR, 927-60001), and then incubated overnight at 4 °C with rabbit V5-Tag (D3H8Q) antibody (Cell Signaling Technology, 13202, 1:2000) in blocking buffer with 0.2% Tween-20. Membranes were washed 4 times with 1xTBST (20 mM Tris, 137 mM NaCl, 0.1% Tween-20, pH 7.6) then incubated with anti-β-actin (Millipore Sigma, A5441, 1:5000) antibody for 1 h at room temperature. Following 4 additional washes with 1xTBST, membranes were incubated with anti-mouse (LI-COR, IRDye 680RD, 926-68070) and anti-rabbit (LI-COR, IRDye 800CW, 926-32211) fluorescent antibodies (1:10,000) for 1 h at room temperature. After washing 4 times with 1xTBST after the incubation, membranes were rinsed with 1xTBS and imaged using the Odyssey Fc Imaging System (LI-COR Biosciences). V5 signal was quantified using ImageJ and normalized to actin loading control.

For DMH^mWAKE neuron immunolabeling, 10–12-week-old female mWake^(V5/V5) mice were deeply anesthetized and then perfused with 4% paraformaldehyde in 1x PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄) at ~ZT10-12. Brains were removed, postfixed and sliced into 60 μm coronal sections using a VT1200S vibratome (Leica). Sections were washed in 1xPBS, blocked for 1 h in blocking buffer (PBS + 0.25% Triton X-100 + 5% normal goat serum), and then incubated with rabbit anti-V5 (D3H8Q) antibody (Cell Signaling Technology, 13202; 1:2000) and mouse anti-NeuN antibody (Abcam, ab104224; 1:500) in blocking buffer overnight at 4 °C. After washing with PBST (PBS + 0.3% Triton X-100), sections were incubated with goat anti-rabbit Alexa Fluor 488 (Invitrogen, A-11034; 1:1000) and goat anti-mouse Alexa Fluor 568 (Invitrogen, A-11004; 1:1000) secondary antibodies for 2 h at room temperature, washed and then mounted. For tdTomato and GFP/YFP images, native fluorescence of both fluorophores was visualized.

Images were acquired using a Zeiss LSM880, LSM800, or LSM700 confocal microscope under 10x-63x magnification, with post-hoc tiling as needed using Zen Black or Zen Blue (Zeiss). For DMH^mWAKE neuron quantification, images were analyzed using Imaris 9.5.

Seven week old, male mWake^(Cre/+) mice were sacrificed at ~ZT5 by cervical dislocation for single-cell RNA-Sequencing (scRNA-Seq). A modified Act-Seq⁴⁷ method was used in conjunction with a previously described dissociation protocol²⁶, with supplementation of Actinomycin D during dissociation (45 μM) and after final resuspension (3 μM), following debris removal (Debris Removal Solution (130-109-398, MACS Miltenyi Biotec)) in between. 1 mm hypothalamic sections between Bregma 0.02 mm (collecting medial and lateral preoptic area) and Bregma −2.92 mm (beginning of the supramammillary nucleus) were collected, and 2–3 mice pooled per scRNA-Seq library.

Following dissociation, tdTomato+ cells were flow-sorted using an Aria Ilu Sorter (Becton Dickinson). Between 400–1000 cells were flow-sorted per brain. Flow-sorted cells were pelleted and re-suspended in 47.6 μl resuspension media. 1 μl of flow-sorted tdTomato⁺ cells were used to quantify % of tdTomato⁺ cells with a phase-contrast microscope. Only samples containing ~99% flow-sorted tdTomato⁺ cells were processed for scRNA-Seq. The remaining 46.6 μl were used for the 10x Genomics Chromium Single Cell system (10x Genomics, CA, United States) using V3.0 chemistry per manufacturer’s instructions, generating a total of 3 libraries. Libraries were sequenced on Illumina NextSeq 500 with ~150 million reads per library (~200,000 median reads per cell). Sequenced files were processed through the CellRanger pipeline (v 3.1.0, 10x Genomics) using a custom mm10 genome (with tdTomato-P2A-Cre-WPRE-bGH sequence). All 3 libraries were aggregated together for downstream analysis.

Seurat V3⁴⁸ was used to perform downstream analysis following the standard pipeline, using cells with >200 genes and 1000 UMI counts, removing mWake-tdTomato⁺ ependymal cells and non-mWake-cells (~1% of the total cluster composed of oligodendrocytes and astrocytes) using known markers genes in the initial clustering²⁶. Louvain algorithm was used to generate different clusters, and spatial information (spatial location of different mWake-tdTomato⁺ clusters across hypothalamic nuclei) and identity of neuronal clusters were uncovered by referring to a hypothalamus scRNA-Seq database²⁶. Region-specific transcription factors expressed in mWake⁺ neurons were used to train mWake-scRNA-Seq. Spatial information of different mWake neuronal populations was further validated by matching to the Allen Brain Atlas ISH data using cocoframer⁴⁹, as well as matching to known mWake neuronal distribution across the hypothalamus. The percentage of GABAergic (Slc32a1⁺) and Glutamatergic (Slc17a6⁺) neurons within each cluster was calculated.

Animals were entrained to a 12:12 h LD cycle for at least 2 weeks before any locomotor or EEG-based behavioral experiments. Experimenters were blinded to animal genotype.

Eight to twelve week old male mice were anesthetized to surgical depth with a ketamine/xylazine mixture (100 mg/kg and 10 mg/kg, respectively). The mouse was secured into a Stoelting stereotaxis frame, and small (~0.5 mm) craniotomies were performed to allow for virus injection (50–300 nl at ~25 nl/min). Coordinates, volumes, and viruses used are listed in Supplementary Table. Post-injection, animals were allowed to recover and express viral genes for ≥2 weeks (for anatomical studies), and ≥3 weeks (for all behavior and functional manipulations). If sleep behavior was measured, EEG headmounts were implanted in a separate surgery. Locations of viral injections were confirmed by post-hoc fluorescence imaging. 1

To try to isolate glutamatergic vs GABAergic DMH^mWAKE neurons, a variety of approaches using Cre and Flp-expressing mouse lines (combined with stereotaxic AAV injections into the DMH) were utilized: mWake^(Cre/+)/Vgat^(Flp/+) (AAV-Con/Fon-ChR2 or AAV-Con/Foff-ChR2) or mWake^(Cre/+)/Vglut2^(Flp/+) (AAV-Con/Fon-ChR2 or AAV-fDIO-tTA with AAV-TRE-DIO-ChR2) (Supplementary Table 1). To validate these approaches, AAV-Con/Fon-EYFP or AAV-Con/Foff-EFYP was injected into the relevant strain (2–4 month old female mice) and 2–3 weeks later, animals were deeply anesthetized and then perfused with 4% paraformaldehyde in 1x PBS. RNAscope in situ hybridization was then performed using the RNAscope® Multiplex Fluorescent Reagent Kit V2 (Advanced Cell Diagnostics/ACD, 323100). 12 μm sections were used for RNAscope in situ hybridization according to the manufacturer’s protocol (320293). Target probes Vglut2-C2 (319171-C2), Vgat-C2 (319191-C2), tdTomato-C3 (317041-C3) were used. OpalTM Dye (AKOYA Biosciences OpalTM 4-Color Manual IHC Kit, NEL810001KT, 1:2000) was applied to develop the signal (OpalTM 570 and 690). Next, immunostaining was performed using chicken anti-GFP (Invitrogen, A10262, 1:1000), and goat anti-chicken Alexa Fluor 488 (Invitrogen, A-11039; 1:1000) antibodies. ProLong™ Gold Antifade Mountant (Invitrogen, P36930↗) was immediately placed on each slide, followed by a glass coverslip. Imaging was performed on a Zeiss LSM700 confocal microscope under 40x magnification. Data were analyzed by manually counting the number of total GFP⁺ cells and Vgat⁺/GFP⁺ or Vglut2⁺/GFP⁺ cells using Zen Blue (Zeiss) from 3 animals (3–5 slices per animal). For determining whether a cell was Vgat⁺ or Vglut2⁺, a threshold of ≥4 RNAscope signal puncta per cell was used.

AAVDJ-hSyn-Con/Fon-hChR2(H134R)-EYFP-WPRE or AAV9-EF1a-double floxed-hChR2(H134R)-EYFP-WPRE-HGHpA viruses were stereotaxically injected into the DMH of mWake^(Cre/+);Vgat^(Flp/+) or mWake^(Cre/+) mice, respectively, as described above and in Supplementary Table 1. An optical fiber (200 µm O.D., 0.39 NA, Thorlabs) within a ceramic ferrule was then inserted towards the DMH, and an EEG head mount (Pinnacle Technology) was attached to the skull with 4 screws. Dental acrylic was applied to affix the ferrule and EEG headmount on the skull. Mice were allowed to recover for at least 2 weeks after the surgery. Then, the animals were connected to the optogenetic system (Thorlabs, OGKL2) and the EEG recording rigs (Pinnacle Technology). EEG was recorded after at least 4 days of habituation, and optogenetic stimulation was randomly applied for 10 min between ZT2 to ZT7 (470 nm, 4.5–4.9 mW, 10 Hz, 20% light on). For EEG analyses, 4–9 trials per mouse were recorded over 2–3 days and were only used if the mouse was asleep at the onset of the trial. Video was acquired using a top-mounted camera (DMK 22AUC03, LORE⁺ Technology), sampling at 15 Hz. Locomotor behavior was quantified using ANY-maze software (Stoelting Company), and time spent feeding was quantified manually by an observer. For video analyses, 3–7 trials per mouse were recorded over 1–2 days and were only used if the mouse was asleep at the onset of the trial.

Male mice between 5 and 10 weeks old were deeply anesthetized with isoflurane, and the brains quickly removed and dissected in oxygenated (95% O₂, 5% CO₂) ice-cold slicing solution (2.5 mM KCl, 1.25 mM NaH₂PO₄, 2 mM MgSO₄, 2.5 mM CaCl₂, 248 mM sucrose, 26 mM NaHCO₃ and 10 mM glucose). Acute coronal brain slices (250 μm) were prepared using a vibratome (VT-1200s, Leica) and then incubated in oxygenated artificial cerebrospinal fluid (ACSF, 124 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 2 mM MgSO₄, 2.5 mM CaCl₂, 26 mM NaHCO₃ and 10 mM glucose, 290–300 mOsm) at 28 °C for 30 min and then at room temperature for 1 h. Slices were then transferred to a recording chamber, continuously perfused with oxygenated ACSF at room temperature and visualized using an upright microscope (BX51WI, Olympus). Labeled cells of interest were visualized using infrared differential interference contrast (IR-DIC) and native fluorescence. Glass electrodes (5–8 MΩ) were filled with the following internal solution (130 mM K-gluconate, 5 mM NaCl, 10 mM C₄H₈N₃O₅PNa₂, 1 mM MgCl₂, 0.2 mM EGTA, 10 mM HEPES, 2 mM MgATP and 0.5 mM Na₂GTP, pH 7.2–7.3, 300 mOsm). Whole-cell patch clamp recordings were obtained using a Multiclamp 700B amplifier (Molecular Devices). Data were sampled at 20 kHz, low-pass filtered at 2 kHz, and digitized using a Digidata 1440 A (Molecular Devices).

For baseline spontaneous and evoked firing rate measurements, recordings were performed under current clamp configuration. Baseline recordings were performed for at least 30 s to measure spontaneous firing rate. To measure evoked firing rate, current injections from −10 to 100 pA were performed; for each step, current was injected for 100 ms every 2 s. For recordings of DMH^mWAKE neurons, ~30% cells exhibited spontaneous firing, and so analyses of evoked responses were restricted to this subset of neurons.

AAV-Flex-GCaMP6s virus was injected into the DMH of mWake^(Cre/+) mice as described above and in Supplementary Table 1. An optical fiber (400 µm O.D., 0.57 NA; Doric Instruments) within a metal ferrule (1.25 mm O.D., Doric Instruments) was then inserted towards the DMH, and an EEG head mount (Pinnacle Technology) was attached to the skull with 4 screws. Dental acrylic was applied to affix the ferrule and EEG headmount on the skull. Mice were allowed to recover at least 14 days after the surgery. GCaMP6s and isosbestic fluorescence were collected from the same implant/patch cable; Ca²⁺-dependent (525 nm) and isosbestic control (430 nm) fluorescence signals (corresponding to 465 nm and 405 nm excitation, respectively) were relayed to the LED driver (Doric Instruments) through dichroic mirrors and bandpass filters within the Fluorescence Mini Cube (Doric Instruments). Doric Neuroscience Studio software (Doric Instruments) was used to control and modulate excitation light (465 nm, 572.205 Hz; 405 nm, 208.616 Hz) under lock-in mode, and fluorescence signals were recorded for 15 min followed by a 15 min break at a 12 KHz sampling rate via the fiber photometry console. Six trials were performed from ZT0-3 or ZT12-15 over 1 or 2 days, and whether recordings began at ZT0 or ZT12 was alternated between different animals. The fiber photometry console sent a 1 s TTL input to the EEG recording system every 3 min for synchronization.

Data were analyzed using custom MATLAB (MathWorks) script. Collected GCaMP, EEG and EMG data were first time-aligned using the TTL pulse sent by the fiber photometry DAQ (Doric Lenses) to the EEG/EMG data acquisition device (Pinnacle Technologies). Bleaching artifacts were removed by low-pass filtering both the reference and GCaMP fluorescence at 0.1 Hz and fitting with an exponential curve. The resulting curve was then subtracted from the reference and GCaMP fluorescence trace. Motion artifacts were corrected by subtracting the bleach-corrected reference signal from GCaMP fluorescence. Z-score was then calculated by subtracting the median from the signal and dividing it by the median absolute deviation. The final Z-score was low pass filtered at 5 Hz. For each ZT time point, six trials (15 min per trial) were averaged to calculate the average fiber photometry trace. These data were then averaged per animal to calculate average Z-scores per state per animal.

Transient peaks were then filtered by using the findpeaks function in MATLAB setting the minimum peak width to 80 ms and minimum peak prominence to 0.1. Transient prominence was defined as the height of the signal from the baseline. State transitions for each state were filtered by selecting those with +/−15 s of consolidated states prior to the transition point. The resulting traces were normalized to individual maximums and then plotted as heat maps.

DREADD receptors coupled to either Gq or Gi were expressed in a Cre-dependent fashion in mWake⁺ neurons of mWake^(Cre/+) mice via stereotaxic injection of a viral vector (AAV-DIO-hM3D-Gq or AAV-DIO-hM4D-Gi) (Supplementary Table 1). Clozapine N-oxide (CNO) (SigmaAldrich) was prepared as a stock solution of 50 mg/ml in DMSO, and then freshly diluted to 0.1 mg/ml in sterile PBS before IP injection. Solution clarity was monitored throughout dosing, and the solution was warmed to 37 °C if precipitates were observed. Vehicle control was prepared as sterile saline + 0.01% DMSO. All injections occurred at the same ZT/CT time within each experiment, and all animals were treated with vehicle or CNO each day in a cross-over design, with ≥2 days recovery between experimental recording days. To control for CNO activity on its own, 1 or 3 mg/kg CNO were IP injected into sham-injected mWake^(Cre/+) mice and locomotion and EEG data were assessed.

Statistical analyses were performed in Prism 7 and 8 (GraphPad). For comparisons of two groups of normally distributed data, unpaired Welch’s t-tests were used; if these comparisons were before and after treatment of the same animals or cells, paired t-tests were used instead. For comparisons of two groups of non-normally distributed data, Mann Whitney U tests were performed. Holm-Bonferroni corrections were used if required for multiple comparisons. For multiple comparisons of normally distributed data, one-way ANOVAs were performed with post-hoc Tukey or post-hoc Dunnett tests. For multiple comparisons of normally distributed data with 2 factors, two-way ANOVAs were performed (with repeated measures, if applicable), followed by post-hoc Sidak tests. For multiple comparisons of non-normally distributed data, Kruskal–Wallis tests were performed with post-hoc Dunn tests. Data that were not normally distributed were plotted as simplified boxplots (median with 1st and 3rd quartile boxes).

Further information on research design is available in thelinked to this article. Nature Portfolio Reporting Summary

Supplementary Information Description of Additional Supplementary Files Supplementary Movie 1 Supplementary Movie 2 Supplementary Movie 3 Reporting Summary

Source Data