Chemogenetic modulation of histaminergic neurons in the tuberomamillary nucleus alters territorial aggression and wakefulness

Designer receptor activated by designer drugs (DREADDs) techniques are widely used to modulate the activities of specific neuronal populations during behavioural tasks. However, DREADDs-induced modulation of histaminergic neurons in the tuberomamillary nucleus (HATMN neurons) has produced inconsistent effects on the sleep–wake cycle, possibly due to the use of Hdc-Cre mice driving Cre recombinase and DREADDs activity outside the targeted region. Moreover, previous DREADDs studies have not examined locomotor activity and aggressive behaviours, which are also regulated by brain histamine levels. In the present study, we investigated the effects of HATMN activation and inhibition on the locomotor activity, aggressive behaviours and sleep–wake cycle of Hdc-Cre mice with minimal non-target expression of Cre-recombinase. Chemoactivation of HATMN moderately enhanced locomotor activity in a novel open field. Activation of HATMN neurons significantly enhanced aggressive behaviour in the resident–intruder test. Wakefulness was increased and non-rapid eye movement (NREM) sleep decreased for an hour by HATMN chemoactivation. Conversely HATMN chemoinhibition decreased wakefulness and increased NREM sleep for 6 h. These changes in wakefulness induced by HATMN modulation were related to the maintenance of vigilance state. These results indicate the influences of HATMN neurons on exploratory activity, territorial aggression, and wake maintenance.

Histaminergic neurons in the tuberomamillary nucleus (TMN) of the posterior hypothalamus (HA TMN neurons) modulate a multitude of physiological processes and behaviours, and altered neuronal histamine signalling is associated with changes in the sleep-wake cycle, motor activity and aggression among other behavioural effects 1,2 . Several technical challenges have limited research on the specific behavioural functions of HA TMN neurons. Notably, histamine (HA) receptors are widely distributed throughout the central nervous system (CNS) and peripheral tissues, limiting the utility of pharmacologic modulation, while clusters of HA neurons may be difficult to stimulate specifically using implanted electrodes. Alternatively, chemogenetic and optogenetic techniques allow for the modulation of neuronal activities with regional, cellular and temporal specificity 3,4 . However, such techniques applied to HA TMN neurons have also yielded inconsistent effects on the sleep-wake cycle. In one recent study, chemogenetic inhibition decreased wakefulness and increased non-rapid eye movement (NREM) sleep without changing REM sleep during active periods 5 . Similarly, another optogenetic study reported that acute silencing of HA TMN neurons during wakefulness promoted NREM sleep, but not REM sleep 6 . However, Venner et al. reported that optogenetic inhibition of HA TMN neurons during wakefulness did not alter the duration of NREM sleep. Furthermore, chemogenetic activation of HA TMN neurons at Zeitgeber time (ZT) 3 did not alter the sleep/wake ratio or electroencephalogram (EEG) spectra 7 .
In addition to the sleep-wake cycle, the brain histaminergic system is involved in the regulation of locomotor activity and aggressive behaviours. Genetic deletion (knockout, KO) of histamine N-methyltransferase (Hnmt) (EC 2.1.1.8), which metabolises histamine into inactive 1-methylhistamine, induced a sustained elevation of extracellular histamine concentration in the mouse CNS and concomitantly reduced locomotor activity during the dark periods, enhanced aggressive behaviours and disrupted the normal sleep-wake cycle without changing the total amount of wakefulness 8 .
We speculated that these discrepancies in the behavioural effects of HA TMN neuron modulation could reflect relatively minor differences in behavioural assessment methods and the emotional status of the animal. Furthermore, we suggest that the histidine decarboxylase (Hdc) (EC 4.1.1.22)-Cre mouse line may provide more cell type-specific chemogenetic or optogenetic manipulation of HA TMN neuron activity 2,6 . Therefore, we reexamined histaminergic system functions in regulation of the sleep-wake cycle, locomotor activity and aggressive behaviours using designer receptor exclusively activated by designer drugs (DREADDs) in Hdc-Cre mice, and further examined the projection distribution of HA TMN fibres by immunohistology to identify the neural circuits underlying the behavioural effects of HA TMN neuron activity.

Results
Chemogenetic receptor expression by HA TMN neurons after microinjection. Cell type-specific expression of chemogenetic receptors is crucial for establishing relationships between activity changes and behaviour. We therefore first confirmed that expression of chemogenetic receptors was exclusive to HA TMN neurons in the posterior hypothalamus by scanning brain slices obtained after behavioural experiments for immunoreactivity to the vector marker mCherry. Indeed, mCherry immunoreactivity (brown) was detected only in posterior hypothalamus (TMN) in mice injected with hM3Dq ( Fig. 1A) or hM4Di (Fig. 1B). Moreover, nuclear c-Fos expression (black) was markedly higher in hM3Dq-mCherry-immunoreactive cells (Fig. 1A′) compared to hM4Di-mCherry-immunoreactive cells (Fig. 1B′) 3.0 h after administration of clozapine N-oxide (CNO) (0.3 mg/kg i.p.), consistent with greater neuronal activity specifically in CNO responsive cells 9 .
Activation of HA TMN neurons enhanced locomotor activity in a novel environment. Previous studies have yielded inconsistent or equivocal results regarding the effects of HA TMN neuron excitation or inhibition on mouse locomotor activity. Acute pharmacological inhibition of histamine decarboxylase (HDC) or deletion of the Hdc gene, interventions that would reduce brain histamine levels, have been reported to decrease locomotor activity in rodents [10][11][12][13][14] . However, a chronic increase in histamine level also decreased locomotor activity in dark periods, while an acute moderate increase in brain histamine was found to enhance locomotor activity 8,15,16 . Histamine receptor stimulation/inhibition outside the TMN, effects of histamine on other neuronal populations and (or) compensatory changes after knockout may obscure effects of HA TMN neurons on locomotion. Therefore, we established a DREADDs system to assess the effects of direct acute activation or inhibition of HA TMN neurons. The chemogenetic activation of HA TMN neurons with CNO did not significantly alter locomotor activity during the first 30-min trial compared with saline (SA) (Fig. S1A-D). Similarly, inhibition of HA TMN neurons by hM4Di-CNO did not decrease the mean locomotor activity during two 30-min trials compared with controls ( Fig. S1E-H). In the subgroup analyses, the hM3Dq-CNO group demonstrated increased total movement distance and average speed in the first trial ( Fig. 2A,B). Time-resolved analyses indicated that these indices were significantly increased in the first 5-min period ( Fig. 2A,B). In the second trial, however, hM3Dq-CNO significantly decreased movement distance and average speed (Fig. 2I,J). On the other hand, subgroup analysis of hM4Di did not reveal any significant differences ( Fig. 2E-H,M-P). These results indicate that HA TMN neuron activation enhances locomotor activity in a novel environment but actually decreases activity in a non-novel environment, suggesting that HA signalling influences exploratory behaviour.  (Fig. 3A,B) compared to saline control in the residentintruder test, and also tended to reduce the latency to first aggressive behaviour, although the difference did not reach statistical significance (Fig. 3C). The subgroup analyses of the inter-and intra-trials indicated that the specific activation of HA TMN increased the number and duration of attack in the second trial (Fig. 3D,E). Conversely, inhibition of HA TMN neurons by hM4Di-CNO had no significant influence on aggression compared to saline control ( Fig. 3A-C,G-I). These results indicate that the activation of HA TMN neurons triggers and maintains territorial aggression in mice.
Modulation of HA TMN neuron activity alters the sleep-wake cycle. Next, we examined whether specific activation of HA TMN alters the sleep-wake cycle by administering CNO or SA to AAV-hM3Dq mice at ZT3 (light period) or ZT12 (dark period). In two-arm crossover experiments, hM3Dq-CNO did not alter wake, NREM sleep, or REM sleep times within the first 6-h during light (Fig. 4A) or dark periods (Fig. S2A). Time-resolved analyses, however, revealed that CNO injection at ZT3 significantly increased total waking time We also found that wake bouts were significantly skewed toward longer duration in the hM3Dq-CNO group by cumulative probability analyses ( Fig. 4E left panel), while the number and mean duration of waking, NREM sleep and REM sleep episodes during the first 1 h were not changed ( Fig. 4C,D, and Table S1). CNO administration to AAV-hM3Dq mice at ZT3 did not alter the frequencies of vigilance state transitions (Fig. 4F). On the other hand, the latency to first NREM sleep episode was significantly extended by hM3Dq-CNO (Fig. 4G). These results indicate that the activation of HA TMN neurons contributes to arousal maintenance but not to vigilance state transitions. There were no differences in EEG power spectra during wakefulness, NREM sleep and REM sleep between CNO and SA groups (Fig. S4A). In addition, CNO injection at ZT12 did not alter sleep-wake state in sub-analyses (Figs. S2 and S4B, and Table S1).  www.nature.com/scientificreports/ We also examined whether specific inhibition of HA TMN neurons suppressed wakefulness by comparing AAV-hM4Di mouse responses to CNO and SA administered at ZT3 or ZT12. Injection of CNO decreased waking time and increased NREM sleep time during the dark periods (Fig. 5A), while REM sleep in the dark periods and all sleep and wake stages in the light periods did not differ between CNO and SA groups (Figs. 5A and S3). Analysis of sleep-wake state during the first 1 h after CNO injection revealed decreased waking duration and increased NREM sleep duration but no changes in REM sleep duration (Fig. 5B). The mean duration of NREM sleep was significantly increased by hM4Di-CNO at ZT12 (Fig. 5D and Table S2), while the number of individual sleep or waking episodes, mean durations of wakefulness and REM sleep, frequencies of vigilance state transitions, and latencies to first NREM sleep and REM sleep episodes were not significantly changed (Fig. 5C,D,F,G). The distribution of the NREM bout durations was also significantly different between CNO and SA groups during ZT12-13 ( Fig. 5E middle panel). Collectively, these findings demonstrate the HA TMN neuron inhibition during the dark period impedes arousal maintenance. EEG power densities for each episode did not differ between CNO and SA groups (Fig. S4D). Further, CNO injection at ZT3 did not significantly alter sleep-wake state (Figs. S3 and S4C, Table S2). These results indicate that HA TMN neuron activity during the dark period is crucial for maintaining arousal state.
HA TMN neurons projects to brain regions known to regulate aggression, vigilance and the sleep-wake cycle. We then investigated the potential circuits mediating these effects of HA TMN neuron chemoactivation/chemoinhibition on locomotion, aggression and wakefulness by conducting serial mCherry immunostaining to identify projection targets. mCherry-immunoreactive somata were found exclusively in the ventral and caudal TMN 6 months after AAV microinjection (Fig. 6A,B). In contrast, mCherry-positive histaminergic fibres were observed in multiple brain regions at 6 months post-AAV microinjection, including the bed nucleus of the stria terminalis (BNST) and substantia innominata (SI) regions of the preoptic area (Fig. 6C), wide regions of hypothalamus but particularly in the lateral hypothalamus (LH) (Fig. 6D), central amygdala (CeA) (Fig. 6E) and periaqueductal grey (PAG), especially the ventrolateral PAG (vlPAG) (Fig. 6F). These regions may mediate the effects of HA TMN neuron activity on exploratory locomotion, territorial aggression and the sleepwake cycle.

Discussion
In the present study, we demonstrate that the specific modulation of HA TMN neuron activity by DREADDs can influence exploratory locomotion, territorial aggression and the sleep-wake cycle of mice. Although specific activation of HA TMN neurons did not significantly alter overall locomotion in the two-arm crossover trial, activation moderately increased locomotion in trial 1 (first exposure) and moderately decreased locomotion in trial 2. In contrast to these modest effects on locomotion, activation of HA TMN neurons markedly enhanced aggression of male mice toward an intruder male. Furthermore, HA TMN neuron activation increased wakefulness for an hour in the light periods, whereas inhibition significantly decreased wakefulness during the dark periods. Transitions in sleep-wake state were also regulated by HA TMN neuron activity. Finally, we show that TMN histaminergic neurons project to brain regions implicated in fear, aggression and arousal, such as the BNST, LH, CeA and vlPAG.
The HA TMN neurons of rodents are distributed in five clusters, E1, E2 and E3 in the ventral area, E4 in the medial dorsal area and E5 more diffusely in the dorsal area 20 . At 2-3 months after injection of Hdc-hM3Dq or Hdc-hM4Di vector, mCherry-positive histaminergic neurons were localised primarily in lateral E1 and E2 (Fig. 1A,B), a distribution comparable to that of Hdc-Cre mice crossed with a Cre-dependent tdTomato reporter line 6 . In addition, however, immunostaining of brain sections 6 months after microinjection revealed additional mCherry-positive histaminergic neurons in E3 and E5 (Fig. 6A,B). A transgene downstream of the human synapsin gene (hSyn) promoter in the AAV vector was exclusively expressed in neurons, and AAV8 induced long-term upregulation of the transgene 21,22 . Therefore, a wider distribution of immunoreactive histaminergic  www.nature.com/scientificreports/ neurons in the present study may result not only from a longer time period for expression but also from use of the hdc promoter. While we assume that the observed behavioural effects are due mainly to the activities of E1 and E2 histaminergic neurons (Fig. 1), contributions of E3 and E5 neurons cannot be totally excluded.
Previous studies have clearly demonstrated that HA TMN neurons project to broad regions of the brain, including BSNT, LH, CeA and vlPAG 23,24 . Our projection study confirmed these findings and revealed a particularly dense projection to the bilateral vlPAG (Fig. 6F). Neurons of the vlPAG are involved in the regulation of locomotor activity 25 , aggressive behaviours 26 and the sleep-wake cycle [27][28][29] , consistent with our behavioural results. Antegrade and retrograde optogenetics experiments are warranted to elucidate the precise circuit mechanisms.
A positive association between hypothalamic histamine release and physiological locomotor activity was first reported in 1992 30 and thereafter evidence has accrued for important contributions of HA TMN neuronal excitation to locomotor activity, mainly based on experiments in which histamine production or H 1 and H 3 receptors were inhibited 10,13,31,32 . In these loss-of-function studies, locomotor activity was generally reduced by inhibition of histaminergic signalling in the CNS, but gain-of-function studies in which brain histamine levels were increased have yielded unexpected and inconsistent results 8,15,16 . We found no significant difference in locomotion between CNO and control groups in the two-arm crossover experiment (Fig. S1), even in subgroup intracohort analysis between trials (data not shown). However, we found that HA TMN activation increased locomotor activity in the Hdc-hM3Dq intercohort analysis for trial 1 ( Fig. 2A,B) and decreased activity in trial 2, suggesting that HA TMN neuron activation enhances locomotor activity in a novel environment 31,33 . Increased locomotor activity by specific HA chemogenetic activation was also reported by Yu et al. but they observed a greater and more sustained increase probably because of the higher dose of CNO used 34 . On the other hand, HA TMN neuron activation decreased locomotor activity in trial 2 (Fig. 2I,J). Histaminergic activation also contributes to memory retrieval [35][36][37] , so this reduced locomotor activity in trial 2 may be due to strong contextual memory for an otherwise anodyne environment, resulting in lower exploratory motivation. In contrast, HA TMN suppression by hM4Di-CNO did not significantly alter locomotor activity (Figs. 2 and S1). However, a previous study demonstrated that chemogenetic inhibition of HA TMN by 1 mg/kg CNO injection significantly decreased locomotion in dark periods 5 . We performed open-field tests with a much lower CNO dose during the light periods, the resting phase for nocturnal animals, so further studies are needed to evaluate effects of loss-of-function on locomotor activity during the dark periods.
In contrast to the subtle effects on locomotion, HA TMN neuron activation clearly enhanced territorial aggression (Fig. 3). Previous studies have suggested that elevated brain histamine levels induce aggression via H 2 receptor activation 8,17,18 , consistent with the current results. Kárpáti et al., however, found that H 1 receptors in astrocytes also play an important role in the suppression of aggressive behaviours 38 . Moreover, some HA TMN neurons contain GABA as a cotransmitter 34 , which might be involved in aggression. These observations indicate that neurotransmitters released from several subtypes of HA neurons possibly modify aggressive behaviours via multiple receptors expressed on neurons and astrocytes. Thus, further studies are necessary to elucidate the molecular and cellular mechanisms of central histaminergic system in aggressive behaviours.
We speculate that two mechanisms underlie increased aggression upon HA TMN neuron activation, direct modulation of neural pathways mediating aggression and modulation via effects on the circadian rhythm. One potential pathway is the ventromedial hypothalamus, ventrolateral subdivision (VMHvl)-vlPAG circuit 39 , as the VMHvl is innervated by the HA TMN and induces aggression upon stimulation 24,40 . The HA TMN also innervates the vlPAG (Fig. 6), which transforms higher level neuronal signals into aggressive action-specific codes 39 . A second possible pathway is the suprachiasmatic nucleus (SCN)-subparaventricular zone (SPZ)-VMHvl circuit. The SCN is the master circadian pacemaker in the mammalian brain, and sends dense axonal outputs to the SPZ, which negatively regulates aggression via the VMHvl 41 . The HA TMN also projects to the SCN 24,42 and has an inhibitory effect on SCN neuronal activity through H 1 and H 2 receptor activation [43][44][45] . Additional projection mapping and chemogenetic activation studies are required to clarify the contributions of these pathways to territorial aggression induced by HA TMN neuron excitation.  www.nature.com/scientificreports/ To our knowledge, only three previous studies have used Hdc-Cre mice to examine the functions of HA TMN neurons in regulation of the sleep-wake cycle [5][6][7] . Chemogenetic inhibition of HA TMN neurons during the dark periods decreased wakefulness and increased NREM sleep duration in a different Hdc-Cre mouse line, and these responses were accompanied by increased EEG δ power and decreased EEG power at higher frequencies   5 . A study using the same Hdc-Cre line employ in the current study found that optogenetic inhibition of HA TMN neurons by archaerhodopsin (Arch) 3.0 during wakefulness in the dark periods promoted NREM sleep and increased the number of NREM episodes but not NREM duration, and had no effect on EEG spectra 6 . Another study demonstrated that zolpidem increased NREM sleep without significant changes in EEG power in the mice whose HA TMN neurons were selectively modified to zolpidem-sensitive 46 . Although we found prolonged mean NREM duration without alteration in the number of episodes (Fig. 5C,D) and no differences in the EEG power spectra for wake and all sleep stages (Fig. S4), the inhibitory experiments in the present study revealed an important role for HA TMN neurons in maintaining wakefulness during the dark periods, in accord with the aforementioned studies. However, inhibition of HA TMN neurons with ArchT during wakefulness did not impact NREM sleep in experiments using another Hdc-Cre line 7 . These authors also induced chemogenic activation of HA TMN neurons by CNO injection at ZT3 and found no alterations in the relative durations of sleep and wake stages or EEG spectra 7 . In the present study, we also found no changes in the sleep-wake cycle following HA TMN neuron activation in the overall analysis ( Fig. 4A-C). In sub-analysis, however, Hdc-hM3Dq mice demonstrated increased wakefulness in the first hour after CNO injection at ZT3 (Fig. 4B). The increased wakefulness was a consequence of the prolonged latency to first NREM and a higher probability of longer wake bouts by HA TMN activation (Fig. 4E,G).
Previous electrophysiological studies have identified some TMN neurons as 'Waking-on' due to predominant activation during wakefulness and complete silence during NREM and REM sleep [47][48][49] . These neurons also exhibited a pronounced delay in firing during transitions from sleep to wakefulness and a long delay to an arousing stimulus, indicating important contributions to the maintenance of wakefulness and vigilance, but not to state induction 48 . Our findings from bout distributions and transition analyses, which showed prolonged wake bout duration by HA TMN activation and prolonged NREM duration by HA TMN inhibition provide further support for the role of HA TMN neurons in arousal maintenance (Figs. 4E,F and 5E,F). Steininger et al. also suggested that some histaminergic neurons in rats discharged at high rates during REM sleep, termed 'REM-related neurons' 47 . However, we found no relation between HA TMN neuron activity and REM sleep.
As previously described, the Hdc-Cre transgenic mouse line used in the present study has high specificity with minimal Cre recombinase expression outside the TMN; however, it also exhibits only ~ 50% penetrance 6 . Thus, at least half of the Hdc-neuronal population was not modulated by the chemogenetic approach, which may explain why we did not observe the significant changes in locomotor activity and REM sleep reported in previous studies that used different Cre recombinase lines with higher penetrance 5,7 . These Hdc-Cre mouse lines, one a transgenic model 7 and the other a knock-in model 34,46,50 , however, have higher ectopic expressions of Cre recombinase which were observed in various brain regions 2 . Further, several studies suggest that HA TMN neurons are functionally heterogeneous 6,51-53 and form distinct projections to influence unique physiological functions and behaviours. Further work is required to determine the individual HA TMN pathways responsible for the observed changes in locomotor activity, aggression and the sleep-wake cycle. Furthermore, we found no correlations between behavioural changes across groups (data not shown). Larger cohorts for greater statistical power and projection-specific activation methods may be required to identify such relationships.
In conclusion, the specific modulation of HA TMN neuron activity altered exploratory locomotion, territorial aggression and wake maintenance in mice. These responses may be explained by projections from HA TMN neurons to the POA, CeA, LH and vlPAG.
Sections were mounted on Superfrost glass slides (Matsunami, Osaka, Japan), dehydrated, cleared and coverslipped using Multi Mount 480 (Matsunami). The mounted brain sections were scanned using a BZ-X700 microscope (Keyence, Osaka, Japan) and composite images were generated by BZ-X Analyzer software (Keyence).
Behavioural experiments. All behavioural tests were performed with a two-arm crossover design by experimenters blinded to treatment history. Each microinjected mouse was examined first in the open-field test, followed by the resident-intruder test and sleep-wake cycle monitoring by EEG/EMG. Mice were randomly assigned to receive CNO or SA at the beginning of the first trial, then switched to the other drug for the second trial. Again, tests were analysed by experimenters blinded to mouse treatment history.
Open-field test. The open-field test was performed as previously described 34 . Briefly, mice were injected with CNO (0.3 mg/kg; Sigma, St Louis, MO) or saline vehicle (SA) as a control 30 min before the trial during the light period. Each mouse was placed into the centre of a 60 cm × 60 cm square arena and allowed to explore freely for 30 min. Total movement distance, average speed, total movement time and time spent in the central area were recorded and quantified using a video tracking system (EthoVision ® XT, Ver. 13, Noldus, Wageningen, Netherlands) by an experimenter blinded to treatment history. The time interval between first and second trials was 24 h which is long enough to wash out intraperitonially administered CNO 56 .
Resident-intruder test. The resident-intruder test was performed as previously described 8 . Briefly, resident mice were housed individually for 1 week before the testing day to increase territorial motivation. Then, the mouse was injected with CNO (0.3 mg/kg; Sigma) or SA 30 min before the trial during the light period. An unfamiliar male (intruder, C57BL6 WT, matched for age and weight) was introduced into the resident cage and the two mice were allowed to interact freely for 10 min. Aggression of the resident mouse was measured by the frequencies of 'mounting' (attempts to mount the intruder), 'chasing' and 'biting' (bites to dorsal/ventral regions of the intruder). The latency to first aggressive behaviour, number of aggressive behaviours and total duration of aggressive behaviours were measured. The resident mouse was transferred to a new clean cage immediately after the first trial and isolated to regain territorial motivation. The second trial was performed after 1 week of isolation. Another unfamiliar male mouse was used as an intruder in the second trial. All experimental data were video-recorded and analysed by an examiner blinded to treatment history.
Sleep recordings and analysis. A week after EEG/EMG electrode implantation, mice were connected to recording cables and habituated for 3 days in a cylinder chamber (22 cm in diameter) placed within a soundproof electrically shielded box (SP-BOX/S, Shinfactory, Fukuoka, Japan) maintained at constant temperature (23 °C ± 1 °C) and humidity (55.0% ± 5.0%) under a 12:12 h light-dark cycle (ZT0: 8:00, ZT12: 20:00) with food and water available ad libitum. The mice were injected with CNO (0.3 mg/kg; Sigma) or SA at 10:50 AM (10 min before ZT3) or 7:50 PM (10 min before ZT12), followed by EEG/EMG recording for 24 h. Following an interval of at least 3 days, mice were switched to the other drug intervention group for another 24 h of EEG/EMG recording. The EEG/EMG signals were amplified (Biotex, Kyoto, Japan), digitised (AD16-16U/PIC/EV, CONTEC, Osaka, Japan) and recorded by a Vital Recorder system (Kissei Comtec, Nagano, Japan). Only the first 6 h of the 24 h period were analysed. Briefly, collected data were divided into 12-s epochs and scored manually for one of three www.nature.com/scientificreports/ sleep or wake states, wake, NREM sleep, or REM sleep, using SleepSign 3 (Kissei Comtec). Time spent in each state, episode frequency and episode duration were calculated in 1-h bins. Latencies to the first NREM and REM sleep episodes were measured as the period following wakefulness evoked by intraperitoneal (i.p.) injection. To assess the differences in bout durations between CNO and SA group, we analysed the combined data of each vigilance state with cumulative probability distribution [57][58][59] . The frequencies of state transitions (Wake to NREM, NREM to Wake, NREM to REM and REM to Wake) were counted. The power spectral density of EEG signals in each stage was calculated with 0.25 Hz resolution by Fast Fourier Transform using SleepSign 3 software, then normalised for each animal by calculating the % power of each bin relative to the total power from 0 to 20 Hz.
Statistical analysis. All statistical analyses were performed using GraphPad Prism ® version 8 (GraphPad Software, La Jolla, CA). In behavioural experiments, the CNO group was compared to the SA group by twoway repeated measures ANOVA followed by Sidak's post hoc test or by Mann-Whitney U test as indicated in the figure legends. The distributions of bout duration in each vigilance state between CNO and SA groups were compared by Kolmogorov-Smirnov test. All data are presented as the mean ± standard error of the mean (S.E.M.) unless otherwise noted. Differences were considered significant at P < 0.05. All statistical analyses were performed by experiments blinded to animal treatment history.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.