Paraventricular hypothalamic nucleus are essential for arousal promotion and maintenance

Adequate wakefulness is fundamental for proper daytime functioning. Clinical observations indicate that the paramedian region of the hypothalamus is a critical node for controlling wakefulness. However, the specic nucleus and neural circuitry for this function remain unknown. Here, we found that inhibition of PVH vglut2 neurons induced 3-h increase of NREM sleep. Chemogenetic activation of PVH vglut2 neurons potently induced 9-h wakefulness, and PVH CRH neuronal activation also exerted wakefulness. Photostimulation of PVH vglut2 → parabrachial complex/ventral lateral septum circuits immediately drove transitions from NREM to wakefulness. Furthermore, using in vivo ber photometry or multichannel electrophysiological recordings in mice, we nd arousal-dependent increases in population activity of PVH vglut2 neurons. Most importantly, ablation of PVH vglut2 neurons dramatically led mice to hypersomnia-like behaviors. These results demonstrate that PVH vglut2 neurons are essential for physiologic arousal in the hypothalamus. AAV-hSyn-DIO-hM4Di-mCherry, AAV-hSyn-DIO-ChR2-mCherry, AAV-hSyn-DIO-mCherry, and AAV-CAG-FLEX-taCasp3-TEVp were used. AAV vectors were packaged into serotype 2/9 vectors, which consisted of AAV2 ITR genomes coupled with AAV9 serotype capsid proteins. The nal viral concentrations of the transgenes were in the range of 1–5 × 10 12 viral particles/mL. via real-time online EEG/EMG Light pulse trains various programmed and conducted during the inactive period. For chronic photostimulation, programmed light pulse trains (5-ms pulses 20 Hz 10 s and at 30-s intervals for 1 h) were used. The 473-nm laser stimulation was performed from 09:00 to 10:00. Baseline EEG/EMG recordings were acquired at the same time day on the previous day prior to laser stimulation. Sleep–wake cycle parameters durations of NREM sleep, REM sleep, and wakefulness, as well as sleep–wake transitions) were scored over an entire hour for each mouse. After receiving photostimulation, mice were sacriced at 30 min after the nal stimulation for subsequent c-Fos staining. (PBST) at anti-rabbit c-Fos (1:10000 for h); primary antibody (Millipore); and anti-mouse NeuN (1:1000 for 12 h; MAB377, Millipore). Primary antibodies were washed ve times with PBS before incubation with secondary antibodies at room temperature for 2 h (Alexa 488, 1:1000; abcam). Finally, the sections were mounted on glass slides, dried, dehydrated, and cover-slipped. Fluorescent images were collected with a confocal microscope (Nikon AIR-MP).


Introduction
Hypersomnia is characterized by an irresistible need for sleep and an inability to stay awake during major waking episodes, which results in reduced function and overall worse quality of life and even induces mental diseases, highlighting its public health importance 1 . However, few dysfunctional wake-promoting nuclei have been identi ed to induce hypersomnia. Therefore, further identi cation of key nuclei and neural circuitry for promoting wakefulness represents a common goal for clinicians and researchers.
In the last 100 years, more than 15 wake-promoting nuclei have been identi ed. Von Economo rst proposed a wake-active and promoting region located in the posterior hypothalamus from observations of marked somnolence in patients with epidemic encephalitis lethargic 2 . Furthermore, Moruzzi et al. and other studies have revealed that a brainstem ascending reticular activating system (ARAS) is responsible for wakefulness [3][4][5] . However, cell-body-speci c ablation or inhibition of components of the ARASincluding the laterodorsal tegmentum (LDT), basal forebrain (BF), pedunculopontine tegmental nucleus (PPT) cholinergic neurons, dorsal raphe nucleus (DRN) serotonergic neurons, and locus coeruleus (LC) noradrenergic neurons-yields limited alterations in sleep [6][7][8] . Additionally, the lateral hypothalamic area (LH), parabrachial complex (PB), tuberomammilary nucleus (TMN), paraventricular nucleus of the thalamus (PVT), ventral tegmental area (VTA), and supramammillary nucleus (SUM) have also been demonstrated to be involved in arousal regulation 6,9−13 . However, among these wake-promoting nuclei, only LH orexinergic and PB glutamatergic neurons have been shown to be related to hypersomnia.
Dysfunction of orexinergic neurons in the LH results in narcolepsy and sleep fragmentation 6,12,14,15 ; PB glutamatergic neurons are considered to serve as a hub, as they receive afferent chemosensory information and play a role in triggering hypercapnia-induced arousal in obstructive sleep apnea (OSA), whereas ablation of PB glutamatergic neurons decreases hypercapnia-induced arousal [16][17][18] . The further amazing research found that ablation of LH orexinergic neurons and lateral parabrachial nucleus (L-PBN) glutamatergic neurons has little effect on sleep under baseline conditions, and deletion of vesicular glutamate transporter 2 (vglut2) from the medial PB (MPB) causes only a modest (approximately 20%) reduction in wakefulness 8,14 . Clinically, patients with Parkinson's disease (PD), Alzheimer's disease (AD), Kleine-Levin Syndrome, and idiopathic hypersomnia (IH), in which LH orexinergic and PB glutamatergic neurons are thought to function normally, still show hypersomnolence 19 . Collectively, these results suggest that the key hypersomnolence control nucleus remains unidenti ed.
More than 90% of the PVH consists of glutamatergic neurons, whereas GABAergic neurons are more scarcely represented [20][21][22] . PVH vglut2 neurons co-express corticotropin-releasing hormone (PVH CRH ) 23 , arginine vasopressin (PVH AVP ) 24,25 . In the present study, we used cutting-edge techniques in transgenic mice to elucidate that activation of PVH vglut2 , PVH CRH neurons induced wakefulness. Conversely, ablation or suppression of PVH vglut2 neurons caused hypersomnia-like behaviors. Taken together, our ndings indicate that the PVH is essential for wakefulness.

Results
PVH receives direct inputs from the PVT and PB. Considering that the homologous area of the primate posterior hypothalamus in rodents is around the PVH area, which contains mainly glutamatergic neurons, we examined the role of PVH vglut2 neurons in the regulation of wakefulness. We used Cre-dependent rabies virus-mediated monosynaptic retrograde tracing in Vglut2-Cre mice (Fig. 1a, b) and found that PVH vglut2 neurons received direct inputs from the PVT, PB, ZI and VLPAG ( Fig. 1c-f), which are involved in sleep-wake control 9,[15][16][17]25,26 , suggestion that the PVH might act as a key central node for sleep-wake regulation.
PVH vglut2 neurons are preferentially active during wakefulness. We next performed in-vivo ber photometry to investigate the real-time activity of PVH vglut2 neurons across spontaneous sleep-wake cycles in freely moving mice. The recording mode for ber photometery and the expression of the Credependent AAVs expressing the uorescent calcium indicator, GCaMP6f (AAV-EF1α-DIO-GCaMP6f), in the PVH of Vglut2-Cre mice are shown in Fig. 2a, b. PVH vglut2 neuronal activities during wakefulness were signi cantly higher than those during NREM sleep (Fig. 2c-e).
We next performed in-vivo multichannel electrophysiological recordings to monitor the spike ring of individual PVH neurons in freely behaving mice (Fig. 2f). PVH vglut2 neurons exhibited a higher ring rate during wakefulness than during sleep (Fig. 2g, h). The PVH vglut2 neuronal ring rate gradually decreased before sleep onset and increased during transitions from sleep to wakefulness (Fig. 2i-k). At the onset of behavioral arousal from NREM sleep, the mean ring rate reached 13.5 Hz (Fig. 2i). Collectively, these electrophysiological results clearly indicate a mechanistic framework for the activity-dependent participation of PVH neurons in the regulation of sleep and wakefulness.
Compared with vehicle injection, chemogenetic activation of PVH vglut2 neurons signi cantly increased electroencephalographic (EEG) low delta power (0.25-1.00 Hz) and decreased high delta power (1.25-4.75 Hz) ( Supplementary Fig. 1k). No sleep rebound followed the long-lasting wakefulness, as indicated by no change in the time spent in NREM sleep during the following dark period (19:00-07:00; Supplementary Fig. 1j). Besides, there is no signi cant difference in the EEG power density of NREM sleep during the day (7:00-18:00) before/after the day of CNO injection ( Supplementary Fig. 1l). Similarly, CNO injection during the dark period also signi cantly increased wakefulness and induced high levels of arousal ( Supplementary Fig. 2a, b), further demonstrating that activation of PVH vglut2 neurons prolonged arousal even during the dark (active) period.
Next, we further explored arousal-promoting roles of subtype neurons of PVH vglut2 neurons (PVH CRH and PVH AVP neurons) and found that chemogenetic activation of PVH CRH neurons caused a potent increase in wakefulness lasting approximately 3 h and concomitantly decreased both NREM and REM sleep Optogenetic activation of PVH vglut2 neurons initiates wakefulness. Compared with the temporal precision of chemogenetic activation, optogenetic manipulations can achieve millisecond-scale control of neuronal activity. Therefore, we next employed optogenetic methods to elucidate the causal role of the PVH in controlling wakefulness. We stereotaxically injected AAVs expressing channelrhodopsin-2 (AAV-DIO-ChR2-mCherry) into the PVH (Fig. 4a). Functional expression of ChR2 was veri ed by in-vitro electrophysiology ( Fig. 4b). Next, we applied optical blue-light stimulation (10 ms, 20 Hz, 20-30 mW/mm 2 ) after the onset of stable NREM or REM sleep during the light phase (Fig. 4c). Optical stimulation of PVH vglut2 neurons during NREM sleep reliably induced transitions to wakefulness in a frequency-dependent manner ( Fig. 4d). Analysis of the probability of transitions between each pair of sleep-wake states showed that optical stimulation signi cantly enhanced the probability of wakefulness, along with a complementary decrease in the probability of NREM or REM sleep (Fig. 4e). To test whether these neurons also contributed to the maintenance of wakefulness, photostimulation was given for 1 h during the light period (09:00-10:00). Sustained activation of PVH vglut2 neurons via semi-chronic optical stimulation (10ms blue-light pulses at 20 Hz for 25 s, every 60 s for 1 h) signi cantly increased the amount of wakefulness in ChR2-mCherry mice compared with that of the baseline control between 09:00 and 10:00 (12.3 ± 1.8 min at baseline vs. 48.6 ± 1.5 min after stimulation, n = 5; Fig. 4f). These ndings demonstrate that optogenetic activation of PVH vglut2 neurons potently enhanced both the initiation and maintenance of wakefulness. PVH vglut2 neurons promote wakefulness via PB and LSv connections. We next sought to determine the downstream targets by which PVH vglut2 neurons promote wakefulness. Speci cally, AAV-hSyn-DIO-ChR2-mCherry or AAV-hSyn-DIO-mCherry constructs were injected into the PVH of Vglut2-Cre mice. We found that PVH vglut2 neurons mainly projected to two neuroanatomical sites: the PB (Fig. 5a) and LSv (Fig. 5e).
To identify the neuronal circuits mediating the wake-promoting effect of PVH vglut2 neurons, ChR2 was expressed in the PVH with optic bers targeting terminals in the PB or LSv (Fig. 5a, e). Optogenetic stimulation (10-ms pulses at 10 Hz for 2 s) of the ChR2-expressing PVH terminals evoked excitatory postsynaptic currents (EPSCs) in most of the patch-recorded PB (n = 6 cells, Fig. 5b) or LSv neurons (n = 8 cells, Fig. 5f). Moreover, 20-Hz stimulation of the bilateral PB or LSv induced a shorter transition from NREM sleep to wakefulness (latency for PB: 1.0 ± 0.8 s, latency for LSv: 1.2 ± 0.9 s) compared with that in the control (Fig. 5c, g). Analysis of the probability of transitions between each pair of sleep-wake states showed that optical stimulation signi cantly enhanced the probability of wakefulness, along with a complementary decrease in the probabilities of NREM and REM sleep (Fig. 5d, h). These results demonstrate that PVH→PB and PVH→LSv circuits mediated the wakefulness-controlling effect of PVH vglut2 neurons. PVH vglut2 neurons are necessary for the control of natural wakefulness. To determine whether PVH vglut2 neurons are necessary for natural wakefulness, we inhibited or ablated PVH vglut2 neurons with two types of AAV constructs encoding engineered Gi-coupled hM4D receptor (AAV-EF1α-DIO-hM4D(Gi)-mCherry) or caspase3 (CAG ex-taCasp3-TEVp-AAV), respectively (Fig. 6a, f). Chemogenetic inhibition of PVH vglut2 neurons decreased wakefulness during the 3 h following administration of CNO compared with that of vehicle ( Fig. 6c). At the beginning of the dark phase (ZT15; 21:00), CNO injection induced a signi cant reduction in wakefulness for approximately 2 h and resulted in a 64.0% increase in NREM sleep during the 5 h post-injection period, which was accompanied by a 26.0% decrease in wakefulness (Fig. 3d). In addition, the EEG power spectrum for each state were not affected by CNO injection (Fig. 3e). Next, in order to explore the role of subtype neurons of PVH vglut2 neurons (PVH CRH , PVH AVP neurons) in controlling sleep, we inhibited these two types of neurons with AAV constructs encoding engineered Gi-coupled hM4D receptor (AAV-EF1α-DIO-hM4D(Gi)-mCherry), respectively ( Supplementary Fig. 3a, e). There was no signi cant sleep increase found in CRH-Cre mice or AVP-Cre mice ( Supplementary Fig. 3b, f), and the EEG power spectrum for each state were not affected by CNO injection (Supplementary Fig. 3d, h). These results suggested that the PVH vglut2 neurons might act as a critical role in regulating sleep.
To further assess the functional importance of PVH vglut2 neurons controlling natural wakefulness, we speci cally ablated these neurons by bilaterally microinjecting AAV-EF1a-DIO-taCasp3-TEVp into the PVH region of Vglut2-Cre mice. This construct expressed a designer pro-caspase-3 (pro-taCasp3) in the PVH, the activation of which causes apoptosis (Fig. 6f, g). Compared with that of the control group, mice that underwent PVH vglut2 neuronal ablation showed a 28.6% decrease in the amount of wakefulness and a 74.7% increase in the amount of NREM sleep during the dark period. Similarly, ablation of PVH vglut2 neurons induced a 20.8% reduction in wakefulness and 30.6% increase in NREM sleep across an entire 24-h light/dark cycle (Fig. 6h). These results indicate that PVH vglut2 neurons are necessary for wake regulation under physiological conditions, and that dysfunction of these neurons may induce hypersomnia.

Discussion
Adequate wakefulness is essential for life and survival. In the present study, we identi ed the PVH as a critical hypothalamic nucleus for the regulation of wakefulness. In previous study, 15% reduction in baseline wakefulness is considered signi cant 6,26 . Lu et al have reported that lesion of the PPT and the ventral sublaterodorsal nucleus (vSLD) results in a 20-30% reduction in baseline wakefulness 6 . However, bidirectional chemogenetic manipulations that inhibit the PPT or activate SLD neurons have been shown to have little in uence on baseline sleep 27,28 . In our present study, three patients with lesions mostly around the PVH showed hypersomnolence lasting above 20 h per day. PSG recordings from these patients showed that stage-two NREM was strikingly dominant, indicating that these patients slept stably and were not easily awakened. Importantly, we found that following recovery from injury around the PVH in one of these patients, the proportion of stage-two NREM sleep decreased, and this patient was concomitantly better able to stay awake. Furthermore, ablation of PVH vglut2 neurons in mice induced a 30.6% reduction in wakefulness across the 24-h light/dark cycle, highlighting the signi cance of PVH vglut2 neurons in maintaining wakefulness and preventing hypersomnia. Besides, in our murine experiments, no sleep rebound was seen after PVH vglut2 -activation-induced enhancement of wakefulness. This nding is in accordance with previous studies using chemogenetics to speci cally activate wake-promoting neuronal populations 9,28−30 and indicates that chemogenetic activation of wake-promoting neuronal populations does not enhance the homeostatic drive for sleep. Taken together, our present ndings provide evidence of the su cient and necessary wake-promoting action of PVH vglut2 neurons in preventing hypersomnia.
The PVH is composed of abundant, diverse, and functionally distinct groups of neuroendocrine neurons, including CRH, AVP neurons 23-25,31−33 . The PVH is estimated to consist of approximately 56,000 neurons in humans 34 , of which 21,000 neurons express AVP, and 2,000 neurons express CRH [35][36][37] . Over 90% of PVH CRH neurons express vglut2 mRNA 22 . Morphological analysis has revealed that 50% of PVH CRH neurons colocalize with PVH AVP neurons, which regulate stress, fear, and immune responses, as well as neuroendocrine and autonomic functions 23,25,33,38,39 . There is mounting evidence that exposure to various stressors induces CRH release into the peripheral circulation 40 . Considering that PVH CRH neurons are actively involved in stress-related behaviors, our results suggest that PVH CRH neurons play an important role in stress-related insomnia.
The PVH has distinct input and output connections that participate in various brain functions. PVH vglut2 neurons are connected to the BF 41 , and PVT 42 , and project to the nucleus of the solitary tract (NTS) 43 , lateral parabrachial nucleus (L-PBN) 16,17 , pre-locus coeruleus (pLC) 31 and ventral lateral septum (LSv) 33 . Our present ndings identi ed two speci c neural circuits, among which the PVH vglut2 →PB/LSv circuit rather than the PVH vglut2 →NTS circuit was required for the maintenance of physiological wakefulness.
Glutamatergic neurons in the L-PB are necessary for arousal in response to CO 2 18 ; therefore, the PVH vglut2 →L-PB circuit may be involved in OSA. Xu et al reported that photostimulation of PVH vglut2 →LSv projections mediates stress-related self-grooming and fear-like jumping behaviors 33 . Thus, the PVH orchestrates sleep/wake states related to stress and chemicals through different circuits.
Surgery and injection of viral vectors. All mice were anesthetized with chloral hydrate (360 mg/kg, i.p.) for surgical procedures and were placed in a stereotaxic apparatus (RWD, Shenzhen, China). The skin above the skull was cut, a burr hole was made, and a small craniotomy was performed above the PVH. AAV constructs were slowly injected (30 nL/min) into the bilateral PVH (70 nL for each position; AP = -0.5 mm; ML = ± 0.2 mm; DV = -4.2 mm) for PSG recordings and brain-slice electrophysiology, or were unilaterally injected into the PVH for neuronal tracing. The glass pipette was left in the brain for an additional 10 min following injections and was then slowly withdrawn. All mice were implanted with electrodes for EEG and EMG recordings that were used for in-vivo tests at four weeks after injections under anesthesia of chloral hydrate (intraperitoneal, 360 mg/kg). The implant consisted of two stainless steel screws (1 mm in diameter), and EEG electrodes were inserted through the skull (+ 1.5 mm anteroposterior; -2.0 mm mediolateral from bregma or lambda), while two exible silver wires were inserted into the neck muscles. The electrodes were attached to a mini-connector and were xed to the skull with dental cement. The scalp wound was sutured, and the mouse was when kept in a warm environment until it resumed normal activity.
Polysomnographic recordings and analysis. After a 2-3-week recovery period, each mouse was individually housed in a recording chamber and habituated to the recording cable for 2-3 days before electrophysiological recordings. Simultaneous EEG/EMG recordings were carried out with a slip ring so that movement of the mice would not be restricted. For experiments using designer receptors exclusively activated by designer drugs (DREADDs), the recordings started at 07:00 (i.e., at the beginning of the light period), and each mouse received either vehicle or CNO (3 mg/kg, C2041, LKT) treatment for two consecutive days at 09:00 (inactive period) or 21:00 (active period). As previously described 10,48 , EEG/EMG signals were ampli ed and ltered (0.5-30 Hz for EEG, 40-200 Hz for EMG), and were then digitized at 128 Hz and recorded with SleepSign software (Kissei Comtec, Nagano, Japan). Sleep-wake states were automatically classi ed into 4-s epochs as follows: wakefulness was considered to have desynchronized EEG and high levels of EMG activity, NREM sleep was considered to have synchronized, high-amplitude, low-frequency (0.5-4 Hz) EEG signals in the absence of motor activity; and REM sleep was considered to have pronounced theta-like (4-9 Hz) EEG activity and muscle atonia. All scoring was automated based on EEG and EMG waveforms in 4-s epochs for both chemogenetic and optogenetic studies.
Optogenetic stimulation. Before the testing day, mice were given one day to adapt to optical ber cables (0.8-m long, 200-µm diameter; RWD) that were placed inside the implanted ber cannulae. On the testing day, 473-nm laser pulses (10 ms, 20 Hz) were delivered via an optic cable (Newton Inc., Hangzhou, China) using a pulse generator. Light pulse trains were generated via a stimulator (SEN-7103, Nihon Kohden, Japan) and delivered through an isolator (ss-102J, Nihon Kohden). For acute photostimulation, each stimulation epoch was applied at 20 s after identifying a stable NREM or REM sleep event via real-time online EEG/EMG analysis. Light pulse trains (5-ms pulses of various frequencies and durations) were programmed and conducted during the inactive period. For chronic photostimulation, programmed light pulse trains (5-ms pulses at 20 Hz for 10 s and at 30-s intervals for 1 h) were used. The 473-nm laser stimulation was performed from 09:00 to 10:00. Baseline EEG/EMG recordings were acquired at the same time of day on the previous day prior to laser stimulation. Sleep-wake cycle parameters (e.g., durations of NREM sleep, REM sleep, and wakefulness, as well as sleep-wake transitions) were scored over an entire hour for each mouse. After receiving photostimulation, mice were sacri ced at 30 min after the nal stimulation for subsequent c-Fos staining.
In-vitro electrophysiological recordings. At 3-4 weeks after AAV-ChR2 injections, Vglut2-Cre mice were anesthetized and transcardially perfused with ice-cold slicing buffer containing the following (in mM): 213 sucrose, 26 NaHCO 3 , 10 glucose, 0. Expression of ChR2 was con rmed by visualization of mCherry uorescence in PVH neuronal somata and axonal terminals. Neurons were identi ed and visualized with an upright microscope (Olympus, Japan) equipped with differential contrast optics, including a 40 × water-immersion objective lens (BX51WI, Olympus). Images were detected with an infrared-sensitive CCD camera (OptiMOS, Q-imaging).
Patch Fiber photometry. Following AAV-EF1α-DIO-GCaMP6f injections, an optical ber (125-µm outer diameter, 0.37 numerical aperture; Newdoon, Shanghai) was placed in a ceramic ferrule and was inserted toward the PVH. Fiber photometry 49 uses the same ber to both excite and record from GCaMP in real time. After surgery, mice were individually housed for at least 10 days to recover. Fluorescent signals were acquired with a laser beam passed through a 488-nm excitation laser (OBIS 488LS; Coherent), re ected off a dichroic mirror (MD498; Thorlabs), focused by an objective lens (Olympus), and coupled through a ber collimation package (F240FC-A, Thorlabs) into a patch cable connected to the ferrule of an upright optic ber implanted in the mouse via a ceramic sleeve (125 µm O.D.; Newdoon, Shanghai). GCaMP6 uorescence was bandpass ltered (MF525-39, Thorlabs) and collected by a photomultiplier tube (R3896, Hamamatsu). An ampli er (C7319, Hamamatsu) was used to convert the photomultiplier-tube current output to voltage signals, which were further ltered through a low-pass lter (40-Hz cut-off; Brownlee 440). The photometry voltage traces were down-sampled using interpolation to match the EEG/EMG sampling rate of 512 Hz via a Power1401 digitizer and Spike2 software (CED, Cambridge, UK). Photometry data were exported to MATLAB R2018b mat les from Spike2 for further analysis. We segmented the value of the uorescent change (ΔF/F) by calculating (F -F 0 )/F 0 , where F 0 is the baseline of the uorescent signal. We recorded data for 3-5 h per mouse for the analysis of sleep-wake transitions to calculate the averaged calcium signal of ΔF/F during all times of vigilant states. For analyzing state transitions, we determined each sleep-wake transition and calculated ΔF/F in a ± 40-s window around that time point.
Firing rate analysis. Electrophysiological data were ltered with a band-pass lter (300-6,000 Hz) to obtain neuronal spikes. Single-unit activities were sorted according to a threshold and shape detector using principal component analysis via O ine Sorter software (Plexon Co, USA). The rst two principal components of each spike on the two-dimensional plot of detected spike events were extracted. Waveforms with similar principal components were clustered via a K-means sorting method. The isolated cluster was considered as a single unit recorded from the same neuron. Spikes with inter-spike intervals < 2 ms were discarded. Cross-correlation histograms were used to eliminate cross-channel artifacts. NeuroExplorer software (version 5.0) was used for producing ring-rate rastergrams, and Prism (version 7.0) was used for producing ring-rate histograms.
Histology and immunohistochemistry. For dual immunostaining with c-Fos and mCherry, mice were deeply anesthetized with chloral hydrate (400 mg/kg) and were perfused with phosphate-buffered saline (PBS) followed by 4% PFA in 0.1-M phosphate buffer. The brain was then dissected and xed in 4% PFA at 4 °C overnight. Fixed samples were sectioned into 30-µm coronal sections using a freezing microtome (CM1950, Leica, Germany). For immunohistochemistry, the oating sections were washed in PBS and were then incubated in the following primary antibodies in PBS containing 0.3% Triton X-100 (PBST) at 4 °C: anti-rabbit c-Fos (1:10000 for 48 h); primary antibody (Millipore); and anti-mouse NeuN (1:1000 for 12 h; MAB377, Millipore). Primary antibodies were washed ve times with PBS before incubation with secondary antibodies at room temperature for 2 h (Alexa 488, 1:1000; abcam). Finally, the sections were mounted on glass slides, dried, dehydrated, and cover-slipped. Fluorescent images were collected with a confocal microscope (Nikon AIR-MP).
Statistical analysis. All data are expressed as the mean ± standard error of the mean (SEM). Sample sizes were chosen based on previous studies 48,50 . Two-way repeated-measures analysis of variance (ANOVA) was used to perform group comparisons with multiple measurements. Paired and unpaired t tests were used for single-value comparisons. One-way ANOVA was used to compare more than two groups, followed by post-hoc Tukey tests for multiple pairwise comparisons. Prism 7.0 (GraphPad Software, San Diego, CA, USA) was used for all statistical analyses. A two-tailed P < 0.05 was considered statistically signi cant.

Declarations
Data availability