Somatostatin neurons in prefrontal cortex initiate sleep-preparatory behavior and sleep via the preoptic and lateral hypothalamus

The prefrontal cortex (PFC) enables mammals to respond to situations, including internal states, with appropriate actions. One such internal state could be ‘tiredness’. Here, using activity tagging in the mouse PFC, we identified particularly excitable, fast-spiking, somatostatin-expressing, γ-aminobutyric acid (GABA) (PFCSst-GABA) cells that responded to sleep deprivation. These cells projected to the lateral preoptic (LPO) hypothalamus and the lateral hypothalamus (LH). Stimulating PFCSst-GABA terminals in the LPO hypothalamus caused sleep-preparatory behavior (nesting, elevated theta power and elevated temperature), and stimulating PFCSst-GABA terminals in the LH mimicked recovery sleep (non-rapid eye-movement sleep with higher delta power and lower body temperature). PFCSst-GABA terminals had enhanced activity during nesting and sleep, inducing inhibitory postsynaptic currents on diverse cells in the LPO hypothalamus and the LH. The PFC also might feature in deciding sleep location in the absence of excessive fatigue. These findings suggest that the PFC instructs the hypothalamus to ensure that optimal sleep takes place in a suitable place.


Tagging PFC GABA cells during SD, nesting and RS
To investigate GABAergic cells that became active following SD, nesting and RS in the PFC (prelimbic, infralimbic and medial orbital subdivisions 35 ) and, as a comparison, in the visual cortex (VC), we used c-Fos-based activity tagging 9,38,39 .Tagging was restricted to cells expressing the Slc32a1 (Vgat) gene in the PFC or the VC (Vgat-PFC-ChR2-Tag and Vgat-VC-Channelrhodopsin-2 (ChR2)-Tag mice, respectively) (see Extended Data Fig. 1a and Supplementary Table 1 for a summary of the method and a list of mouse lines).For tagging, mice were deprived of sleep for 5 h by presenting them with new objects at ZT 0 (start of the light period, greatest sleep-propensity period; ZT, zeitgeber time (hours)).Mice were then placed back in their home cages with doxycycline (Dox)-containing chow (gradually repressing the activity-tagging system) (Extended Data Fig. 1a).
Before SD, body temperature oscillated diurnally over 24 h (Extended Data Fig. 2a) 9,40 , with a lower body temperature (by about 1 °C) during the 'lights-on' (sleep time) period.During SD, the body temperature increased by about 2 °C but declined partially during the later part of SD (Fig. 1a and Extended Data Fig. 2a).After SD, mice either preferentially went to a prebuilt nest in their home cage and improved it or, if there was no nest, they built one before starting their RS (Fig. 1b).During the nesting activity, EEG showed a peak in theta frequencies, as found previously 7 , and body temperature was elevated (Fig. 1a,b and Extended Data Fig. 2a).After nesting, RS had an increase in EEG delta power within the first 2 h, as expected for sleep homeostasis 17,41 , and body temperature returned to basal levels (Extended Data Fig. 2b) (note that, during the first part of the RS period, there was intermittent nesting, and, in those times, body temperature increased, whereas, during individual NREM episodes, body temperature decreased).
Endogenous c-Fos protein was detected in GABAergic cells in both the PFC and the VC after 5 h of SD followed by 2 h of nesting and RS (Extended Data Fig. 2c).Similarly, activity tagging induced human M3 muscarinic (hM3)Dq-mCherry gene expression in about 32% of PFC GABA cells and 24% of VC GABA cells (Extended Data Fig. 2d).In control Vgat-PFC-ChR2-Tag mice, which underwent SD on Dox or experienced baseline activity but not SD off Dox, no transgene expression was present (Extended Data Fig. 3).Thus, c-Fos-driven ChR2 expression was induced in a subset of neocortical GABA cells during SD, nesting and RS in Vgat-PFC-ChR2-Tag mice.

Tagged PFC GABA neurons induce nesting and sleep
To examine the roles of these activity-tagged PFC GABA cells, parallel cohorts of Vgat-PFC-ChR2-Tag and Vgat-VC-ChR2-Tag mice were deprived of sleep and allowed nesting and RS (Fig. 2a and Extended Data Fig. 4a).We termed these mice Vgat-PFC-ChR2-Tag:SD and neurons in the LH 1,7 ; both types of cells are widely projecting, but the relevant targets for nesting are uncharacterized.
We wondered whether the neocortex exerts any top-down influence on sleep-preparatory behavior (nesting) and coordination with sleep.The neocortex does seem to contribute to sleep regulation directly.Neocortical Sst-expressing GABA cells enhance NREM-like sleep and slow-wave activity by an unknown mechanism 14,15 , and genetic silencing of layer V pyramidal and hippocampal dentate granule neurons blocks the characteristic increase in EEG delta power of recovery sleep (RS) following sleep deprivation (SD) and increases the amount of wakefulness during the active ('lights-off') period of mice 16 .The characteristic increase in delta power of NREM RS reflects a deeper sleep 17 and is part of the sleep homeostatic model.In awake and behaving animals, local delta NREM-like oscillations develop in different regions of the neocortex following use-dependent activity [18][19][20][21][22] , and this increase in local delta power depends on increasing chloride concentrations inside pyramidal neurons as wakefulness progresses 23 .
Excitability throughout the neocortex increases with time spent awake 24 .During SD and RS, particular types of neocortical GABAergic neurons, such as nitric oxide synthase 1 (Nos1)-expressing neurons, become active [25][26][27] .But the PFC seems particularly sensitive to SD [28][29][30][31] , which causes functional connectivity to degrade in the PFC more than in other neocortical areas 29,30 ; indeed, there is a selective buildup of glutamate and glutamine in the PFC relative to the VC during daylong cognitive work, which could signal tiredness 31,32 .Nearly a third of the neurons in the monkey PFC increase their firing rate during cognitive disengagement (for example, on becoming drowsy), eye closure and sleep 33,34 .
Conceptually, the PFC stores and creates combinations of purposeful actions 35 , implementing survival and autonomic processes, such as defensive responses 36 , and a selection of behavioral states in response to challenges 37 .Given the heightened sensitivity of the PFC to SD, we therefore hypothesized that the PFC could potentially link sleep pressure, which builds up as wakefulness increases (that is, during SD) 17 , with sleep-preparatory behaviors such as nesting and with sleep itself.We find that, when mice are deprived of sleep, somatostatin (SST)-expressing GABA cells in the PFC (PFC Sst-GABA cells) induce subsequent nesting, elevated theta power and body temperature increase through projections to the LPO hypothalamus while nest building is taking place.Additionally, through projections to the LH, PFC Sst-GABA cells induce NREM RS with elevated delta power and an associated body temperature decrease.This combined PFChypothalamic circuitry could ensure that, if an animal is tired, sleep takes place in a safe environment that promotes RS.Vgat-VC-ChR2-Tag:SD (with the 'SD' in these mouse names standing for the collective 'SD, subsequent nesting and RS' activities; the control groups of these mice, which had not undergone the tagging behavioral protocol, are labeled 'Ctrl').Two days after tagging, optostimulation was directed into the PFC or the VC at ZT 18 (active wake time of mice, 'lights-off' period), and behavior and sleep-wake states were recorded (Extended Data Fig. 1b).To ensure stimulating a range of GABAergic neurons that could be involved, we gave a mixed stimulation protocol: 1 min of 10-Hz, 1 min of 20-Hz and two sets of 2 min of 20-Hz light pulses with 15-min intervals, amounting to approximately 50 min of intermittent optostimulation (Extended Data Fig. 1b).
Optostimulated Vgat-PFC-ChR2-Tag:SD mice entered NREM sleep above baseline levels during the second half of the optostimulation period, coinciding with the time when nesting behavior had decreased, whereas the other groups of mice did not (Fig. 2f and Extended Data Fig. 4d).After 45 min of optostimulation, although the mean sleep latency did not change between Vgat-PFC-ChR2-Tag:SD, Vgat-VC-ChR2-Tag:SD and control mice (Fig. 2f), the amount of NREM sleep of Vgat-PFC-ChR2-Tag:SD mice increased substantially compared with that of other groups (Fig. 2f and Extended Data Fig. 4d).For the first consolidated bout of this NREM sleep, the EEG delta power for optostimulated Vgat-PFC-ChR2-Tag:SD mice increased, consistent with this sleep being recapitulated RS (Extended Data Figs.2b and 4e) 17,41 .The start of NREM sleep in Vgat-PFC-ChR2-Tag:SD mice, however, did not correlate with the final quality of the nest (Extended Data Fig. 4f).
We repeated the tagging experiments using chemogenetics with Vgat-PFC-hM3Dq-Tag mice (Fig. 3a and Extended Data Figs.1c  and 5a).Two days after tagging, an intraperitoneal (i.p.) injection of clozapine-N-oxide (CNO) (1 or 5 mg per kg) at ZT 18 elicited prolonged nesting, high nest-quality scores and increased theta power in EEG (Fig. 3b); sustained NREM sleep was induced within 1 h compared with saline-injected activity-tagged mice (Fig. 3c).Thus, the chemogenetic and optogenetic results were in the same direction.
Summarizing thus far, of the four consecutive blocks of optostimulation sessions, the first and second sessions in Vgat-PFC-ChR2-Tag:SD mice induced nesting behaviors, whereas the third and four sessions induced NREM sleep (Fig. 2f).Consistently, theta power in EEG was elevated in the first two blocks of optostimulation, whereas delta power increased in the second two blocks of optostimulation (Fig. 2f).This suggests that the reactivated tagged neurons or circuitry successfully recapitulate the behaviors (nesting and sleep) in the tagging procedure (Fig. 1a,b).

Article
https://doi.org/10.1038/s41593-023-01430-4 For Sst-PFC-hM3Dq-Tag:SD mice, after CNO injection and the nesting activity described above, the latency to NREM sleep was reduced, as found for Vgat-PFC-hM3Dq-Tag:SD mice (Fig. 3c), whereas, for Nos1-PFC-hM3Dq-Tag:SD mice, NREM sleep latency was unchanged (Extended Data Fig. 5d).For Vgat-PFC-hM3Dq-Tag:SD and Sst-PFC-hM3Dq-Tag:SD mice, sustained NREM sleep was induced above baseline compared with saline in Sst-PFC-hM3Dq-Tag:SD mice, as for the Vgat-PFC-hM3Dq-Tag:SD mice (Fig. 3c); for both sets of mice, the number of sleep attempts and episodes did not change, but only the duration of NREM episodes was prolonged (Fig. 3c).Delta power of evoked NREM sleep was increased in the first hour (Fig. 3c), consistent with this sleep being RS 41 .By contrast, for Nos1-PFC-hM3Dq-Tag:SD mice, NREM induction, although statistically significant, was not sustained (Extended Data Fig. 5d), and there were no changes in sleep latency, sleep attempts, numbers of episodes or episode duration of NREM sleep (Extended Data Fig. 5d).For both Vgat-PFC-hM3Dq-Tag:SD and Sst-PFC-hM3Dq-Tag:SD mice, the core body temperature increased within 5 min of CNO i.p. injection compared with mice treated with saline, before nesting (Fig. 3d), but, when NREM sleep started some 30 min later, temperature decreased (Fig. 3d).For Nos1-PFC-hM3Dq-Tag:SD mice injected with CNO, however, there were no temperature changes (Extended Data Fig. 5e).Therefore, of the types of Vgat-expressing cell types studied after SD, only the PFC Sst cells induced nesting, temperature changes and RS.Focusing on the PFC Sst subset of GABA cells, we confirmed the chemogenetic results using opto-activation with Sst-PFC-ChR2-Tag:SD male and female mice (Fig. 4a).After tagging, 17% of PFC Sst cells were labeled with hM3Dq (Extended Data Fig. 6a).Optostimulation was directed into the PFC of Sst-PFC-ChR2-Tag:SD animals as a session of five bouts of 2 min at various frequencies (1, 5, 10, 20 Hz) with a 10-min interstimulus interval (Fig. 4a and Extended Data Fig. 1b).The behavioral baseline evoked by optostimulation was obtained on the same animals before the tagging procedure (Sst-PFC-ChR2-Tag:Ctrl mice).For all frequencies, for both males and females, the time spent nesting during the optostimulation period was longer in Sst-PFC-ChR2-Tag:SD mice than that in Sst-PFC-ChR2-Tag:Ctrl mice (Fig. 4a,b and Extended Data Fig. 6b).At the end of the stimulations, nest scores in both sexes were higher than those achieved by control mice for all stimulation frequencies (Fig. 4c and Extended Data Fig. 6c).
As with Vgat-PFC-ChR2-Tag:SD mice, optostimulated Sst-PFC-ChR2-Tag:SD mice entered NREM sleep above baseline levels during the second half of the optostimulation period, coinciding with the time when nesting behavior had decreased (Fig. 4d and Extended Data Fig. 6d).During optostimulation, the mean NREM sleep latency decreased in Sst-PFC-ChR2-Tag:SD mice (Fig. 4e) and the amount of NREM sleep exhibited by Sst-PFC-ChR2-Tag:SD mice increased substantially compared with that of Sst-PFC-ChR2-Tag:Ctrl mice (Fig. 4f), and this sleep persisted after optostimulation at 10 and 20 Hz (Fig. 4f and Extended Data Fig. 6d).As with Vgat-PFC-ChR2-Tag:SD mice, optostimulation in Sst-PFC-ChR2-Tag:SD mice first increased theta power, followed by an increase in delta power (Fig. 4g).Therefore, PFC Sst cells that are captured by a combination of SD, nesting and RS can initiate nesting and NREM sleep similar to PFC Vgat cells, and this opto-induced behavior change does not depend on sex.

Tagged PFC Sst cells fire rapidly and target the hypothalamus
We characterized the electrophysiological identities of the tagged PFC Sst-GABA neurons in Sst-PFC-ChR2-Tag mice.Acute slices of the PFC were prepared 2 d after SD, and ChR2-mCherry + cells were whole-cell patch clamped (Fig. 5a,b).As a control, to sample the diversity of PFC Sst cells, AAV-flex-ChR2-mCherry virus was injected into the PFC of Sst Cre mice (Sst-PFC-ChR2 mice), and Sst-ChR2-mCherry + cells were patched randomly.The electrophysiological parameters of the tagged PFC Sst cells (PFC Sst-Tag:SD cells) differed from those of control PFC Sst cells from Sst-PFC-ChR2 mice (PFC Sst cells): although the resting membrane potentials of tagged cells were more hyperpolarized, their rheobase, the amount of current needed to elicit threshold firing of action potentials, was lower (Extended Data Fig. 7a).There was also a marked difference in the way that the two groups of cells (PFC Sst-Tag:SD or PFC Sst cells) were able to fire action potentials.A series of current injections were made to test cell excitability (Fig. 5b).About 70% of PFC Sst-Tag:SD cells could consistently fire action potentials at higher frequencies (15 Hz; Fig. 5b and Extended Data Fig. 7b), whereas PFC Sst cells showed a range of electrophenotypes: most were slow spiking (5 Hz), but a minority (12.5%) were fast spiking (Fig. 5b and Extended Data Fig. 7b), as reported from sampling Sst-expressing cells in the mouse VC 42 .Evoked action potentials from PFC Sst-Tag:SD cells were narrower (that is, more briefly lasting; the rising time was the same, but action potential half-width and decay time were reduced) than those from randomly sampled Sst-expressing cells (Extended Data Fig. 7c).Furthermore, these PFC Sst-Tag:SD cells differed in how they responded to optostimulation.When identified cells were given a single 10-ms light pulse, PFC Sst-Tag:SD cells produced doublet or triplet action potentials, whereas randomly sampled PFC Sst cells did not, only giving single action potentials (Fig. 5c and Extended Data Fig. 7d).When PFC Sst-Tag:SD cells were given the range of optostimulation frequencies that elicit nesting and sleep in vivo, the cells sustained multiple spikes at all stimulation frequencies (1, 5, 10, 20 Hz) (Fig. 5d and Extended Data Fig. 7e).Overall, tagging captured particularly excitable, fast-spiking Sst-expressing GABA cells, a recently discovered type of Sst-expressing neocortical cell 42 .
We examined the projections of these PFC Sst-Tag:SD cells.Although fluorescence was mainly in the PFC, both the LH (ventral part) and the LPO hypothalamus (ventral part) contained fine ChR2-enhanced yellow fluorescent protein (EYFP) + fibers (Extended Data Fig. 8a).We also examined constitutively (non-tagged) labeled cells from Sst-PFC-ChR2 mice (Fig. 5e) and found long-range projections of labeled axons from PFC Sst neurons.These axons did not cross the midline.Many fibers could be seen in the LPO hypothalamus and the LH (Fig. 5e) but not in areas of the cortex beyond the PFC, similar to Sst-PFC-Tag:SD signals (Extended Data Fig. 8b).There were no fibers, however, in the base of the brain from labeled VC Sst cells constitutively expressing ChR2-EYFP in Sst-VC-ChR2 mice (Extended Data Fig. 8c).We focused on the preoptic hypothalamus and the LH as areas potentially involved in sleep and/or nesting behaviors mediated by PFC Sst-GABA cells.

PFC Sst-GABA terminals in the LPO hypothalamus induce nesting but not sleep
We stimulated terminals in the LPO hypothalamus of both constitutively labeled PFC Sst-GABA cells (Sst-PFC-ChR2 mice) and tagged PFC Sst-GABA cells (Sst-PFC-ChR2-Tag:SD mice) (Fig. 6a).Starting at ZT 18, terminals were stimulated at 1, 5, 10 and 20 Hz in separate trials (Fig. 6a and Extended Data Fig. 9a), with the same protocol used earlier for stimulating the PFC Sst cell soma (Fig. 4a and Extended Data Figs.1b and 6b).Compared with control mice, all stimulation frequencies induced cumulative nesting behavior (Fig. 6a and Extended Data Fig. 9a,b): during the immediate 2-min optostimulation period, nesting was immediately induced and persisted (Fig. 6a,b and Extended Data Fig. 9a).For both Sst-PFC-ChR2 and Sst-PFC-ChR2-Tag:SD mice, 5-and 10-Hz stimulation frequencies gave the greatest activation of nesting behavior compared with the pre-stimulation nesting activity (Fig. 6a,b and Extended Data Fig. 9a,b).Similarly, nesting latency (time from stimulus to nesting activity) was shorter in both Sst-PFC-ChR2 and Sst-PFC-ChR2-Tag:SD mice (Fig. 6c), and nest quality was higher after optostimulation, even more so for Sst-PFC-ChR2-Tag:SD mice than for Sst-PFC-ChR2 mice, with nesting effectiveness increasing with higher optostimulation frequencies (Fig. 6d,e and Extended Data Fig. 9c).Relative theta power was also increased at the onset of optostimulation (Extended Data Fig. 9d).By contrast, these optostimulations of PFC Sst-GABA LPO terminals did not induce NREM sleep above baseline compared with controls (Fig. 6f and Extended Data Fig. 9e).Consistent with these results, the spontaneous calcium activity of tagged PFC Sst-GABA LPO terminals in Sst-PFC-GCaMP6-Tag:SD mice became elevated during spontaneous nesting (Extended Data Fig. 9f).Optostimulation of tagged PFC Sst-GABA LPO terminals increased the core body temperature (Fig. 6g).The core body temperature rose by 1 °C immediately after the first bout of optostimulation, peaking within 10 min of stimulation onset.
The temperature profile matched the time course of nesting activity under optostimulation and theta power increase (Fig. 6g and Extended Data Fig. 9d).
We looked at the postsynaptic cell types responding to PFC Sst-GABA LPO terminal inputs (Fig. 7a).In acute hypothalamic slices containing the LPO hypothalamus, tagged PFC Sst-GABA terminals from Sst-PFC-ChR2-Tag:SD mice were optostimulated, and then the subsequent mini-inhibitory postsynaptic currents (mIPSCs) were recorded randomly from cell somas adjacent to ChR2-expressing fibers (Fig. 7a).These recordings were made in the presence of action potential blockers (4-aminopyridine (4AP) and tetrodotoxin (TTX)); thus, responses were likely monosynaptically driven.In responding cells, a single 10-ms light pulse induced single mIPSCs (Fig. 7a); on the other hand, optostimulating at 10 Hz (a frequency that evoked nesting) evoked multiple mIPSCs, and their frequency and amplitude increased during optostimulation (Fig. 7a).We then analyzed cytoplasm by multiplex single-cell RT-PCR from cells that had given positive responses (Fig. 7b).All responding cells (12 cells, seven Sst-PFC-ChR2-Tag:SD mice) expressed Gad1, 92% expressed Meis2 (encoding Meis homeobox 2) and Arpp21 (encoding cyclic adenosine monophosphate (cAMP)-regulated phosphoprotein 21), and about 75% expressed Pou3f3; about half the cells expressed Sst and Nos1, and a third expressed Gal (galanin) (Fig. 7b).Of these markers, only Arpp21 expression is reasonably selective, being originally reported as enriched in the LPO hypothalamus compared with other hypothalamic areas 43 .

PFC Sst-GABA terminals in the LH induce sleep but not nesting
As for the LPO experiments, we stimulated terminals in the LH of both constitutively labeled PFC Sst-GABA cells (Sst-PFC-ChR2 mice) and tagged PFC Sst-GABA cells (Sst-PFC-ChR2-Tag:SD mice) (Fig. 8a).Starting at ZT 18, terminals were stimulated at 1, 5, 10 and 20 Hz in separate sessions (Fig. 8a and Extended Data Fig. 10a).For both the constitutively labeled PFC Sst-GABA terminals and the tagged PFC Sst-GABA terminals, all stimulation frequencies induced cumulative NREM sleep but not in control mice or mice with no light stimulation (Fig. 8a and Extended Data Fig. 10a).Sleep induction was cumulative over the 50-min session (with five bouts of optostimulation trials at 10-min intervals) (Fig. 8b,c and Extended Data Fig. 10a).Delta power was increased throughout the optostimulation (Extended Data Fig. 10b).In some instances, rapid eye-movement (REM) sleep was also induced following NREM induction (Fig. 8c).Optostimulation of PFC Sst-GABA LH terminals (constitutively labeled or tagged) did not induce nesting behavior (Fig. 8d and Extended Data Fig. 10c).Consistently, the spontaneous calcium activity of tagged PFC Sst-GABA LH terminals in Sst-PFC-GCaMP6-Tag:SD mice became elevated during the first part of NREM sleep (Extended Data Fig. 10d).Furthermore, the mean core body temperature started to decrease as soon as the first bout of optostimulation was delivered to Sst-PFC-ChR2-Tag:SD mice (Fig. 8e).This coincided with the evoked NREM induction.The lowered temperature was sustained while animals were in NREM sleep (Fig. 8e).Overall, these results show that selective stimulation of PFC Sst-GABA terminals in the LH, whether activity tagged with ChR2 or  constitutively labeled with ChR2, can initiate NREM sleep and lower core body temperature.We examined in acute LH slices the postsynaptic cell types that respond to PFC Sst-GABA terminals (Fig. 8f).Tagged Sst-expressing terminals from Sst-PFC-ChR2-Tag:SD mice were optostimulated, and mIPSCs were recorded from cell somas close to ChR2-expressing fibers (Fig. 8f).In responding cells (19 cells, six Sst-PFC-ChR2-Tag:SD mice), a single 10-ms light pulse induced single mIPSCs (Fig. 8f); optostimulating at 10 Hz evoked multiple mIPSCs; during stimulation, the frequency but not the mean amplitude of mIPSCs increased (Fig. 8f).We analyzed the cytoplasm of cells that responded by single-cell multiplex RT-qPCR (Fig. 8g).Nearly all cells expressed Gad1 (18 cells), and 37% (seven cells) of these expressed Slc17a6 (Vglut2).Of these, all Vglut2-expressing cells expressed Pmch; Gal expression was preferentially associated with Gad1-, Vglut2-and Pmch-expressing cells.For the Gad1-expressing cells that did not express Vglut2, about half expressed Sst.Many responding cells expressed Snord116 and Nrgn (Fig. 8g).Overall, a wide range of cells in the LH were inhibited by PFC Sst-GABA terminals.

Discussion
The PFC enables mammals to respond to situations, including internal states, with appropriate planning and actions 35 .We hypothesized that 'tiredness' is an internal state, and, indeed, the PFC seems particularly sensitive to fatigue 28,29,31 .An appropriate response to tiredness for mice might be nest building and sleeping.The PFC executes its planning role by sending projections to the hypothalamus and other subcortical areas 35,37,44 .We showed that, when mice (male or female) are deprived of sleep, their post-SD sleep-preparatory behavior (nesting), elevated body temperature and concurrent elevated EEG theta power and then, subsequently, RS (sleep with higher delta power 41 ) and lower body temperature are elicited by PFC Sst-GABA cells projecting to the ventral LPO hypothalamus and the ventral LH, respectively.Tagged PFC Sst-GABA terminals had enhanced calcium activity during nesting and sleep, respectively, and induced fast inhibitory postsynaptic currents (IPSCs) on target cells.Activity tagging enabled us to identify PFC Sst-GABA cells with specific properties: cells that would have been difficult to capture without using the c-Fos system.However, opto-activation of PFC Sst-GABA terminals with constitutively expressed ChR2 in the LPO hypothalamus and the LH also caused nesting, body temperature changes and NREM sleep, suggesting that these PFC cells likely contribute to baseline sleep-preparatory behavior, temperature regulation and NREM sleep as well.The PFC also might feature in deciding where to sleep even without excessive fatigue.
Many neocortical Sst-expressing GABA cells are slow spiking and primarily innervate dendrites of pyramidal cells 42,45 .About 70% of the PFC Sst-GABA cells captured by activity-tagging behaviors are different: they are more excitable and fast spiking (up to 20 Hz) and send long-range projections.All stimulation frequencies of PFC Sst-GABA terminals in the LPO hypothalamus and the LH produced nesting behavior and NREM sleep, although there was some drop off in elicited behavior at 20 Hz.Recently, around 16 subtypes of mouse neocortical Sst-expressing cells have been identified 42 , and one is fast spiking and coexpresses Parv (Pvalb) but not Nos1, partially matching the PFC Sst-GABA cells.It is also possible that, rather than discreet subtypes, there is a continuum of Sst-expressing cells 46 , and that, in this study, we enriched for the fast-spiking end of the spectrum.
Although direct optostimulation of PFC Sst-GABA terminals in the LH induced immediate NREM sleep and lower body temperature, the mechanism that underlies initial nesting with raised body temperature, followed by delayed sleep with lower temperature remains unclear.Several types of PFC Sst-GABA cells might be responsible.These cells might have been tagged at different time points during the procedure, for example, SD versus nesting versus RS, although it is striking that the majority of tagged Sst-expressing cells were the unusual, fast-spiking type and only 30% were the slow-spiking type.A limitation of our approach is that the kinetics of the Dox system do not allow the different segments of mouse behavior (sustained wakefulness, nesting, RS) to be cleanly segregated.Another limitation is that we did not examine the potential role of co-released SST peptide from PFC Sst-GABA cells; for example, SST release might lead to gradual behavioral changes and might contribute to delayed sleep effects in the natural setting.
We do not know what activates PFC Sst-GABA cells (which are relatively easy to excite) during SD-nesting-RS behaviors.PFC Sst-GABA cells might be activated indirectly by decreased activity in dopaminergic PFC-projecting VTA neurons, as decreasing VTA dopamine activity  induces nesting 1 .Intriguingly, many neocortical cells beyond the PFC were captured during the tagging procedure, including Sst-expressing cells in the VC.However, VC Sst cells did not induce nesting or sleep when reactivated and did not send axons to the hypothalamus; therefore, the role of these cells in this context is unclear.Similarly, although neocortical Nos1-expressing cells are activated during RS [25][26][27] , reactivating tagged PFC Nos1 cells did not stimulate nesting, only transiently induced NREM sleep and did not change body temperature.A rare type of neocortical Nos1-positive, Sst-negative GABAergic cell is active in the downstate of NREM sleep but does not contribute to sleep 47 .
How might the PFC Sst-GABA -to-LPO hypothalamus connection induce nesting, enhance theta power and raise body temperature?In the LPO hypothalamus, cells inhibited by PFC Sst-GABA terminals were all GABAergic and nearly all expressed Arpp21, a gene encoding a cAMP-regulated phosphoprotein that binds microRNA 48 .Many cells also expressed Nos1 and Sst, and about a third expressed the transcript for galanin.More work is needed to identify how the inhibited Arpp21-expressing cells interact with the other cells or regions that regulate nesting before sleep 7 .Activating LPO galanin-expressing cells lowers body temperature 40,49 , whereas lesioning or removing them increases body temperature 40 .Thus, the PFC SSt-GABA -to-LPO hypothalamus connection might increase body temperature by inhibiting galanin neurons.There are likely parallel routes by which nesting and sleep induction reinforce one another.Nesting provides thermal microclimates, warming the skin; this in turn, via Nos1-, Vglut2-expressing cells in the MPO hypothalamus, promotes NREM sleep and concomitant body cooling when sleep starts 3,9,50 .
How do PFC Sst-GABA -to-LH connections induce NREM sleep?VTA Sst-GABA neurons, which also project to the LH, also induce sleep [51][52][53] , suggesting that GABA projections to the LH could be a common mechanism for sleep induction.Some LH GABA cells induce wakefulness when activated [54][55][56] , and, if these are targeted by PFC Sst-GABA (and VTA Sst-GABA ) terminals, this could produce NREM sleep.The LH cells that responded with evoked IPSCs from PFC Sst-GABA terminals were GABA cells that often coexpressed Vglut2 and Pmch.Some glutamatergic Pmch cells in the LH project to and excite septal GABA cells 57 ; in principle, glutamatergic Pmch-expressing cells could also excite the wake-promoting LH GABA cells.Similar to the MPO hypothalamic Nos1-expressing, glutamate cells that co-regulate NREM induction and decreases in body temperature 9 , activating PFC Sst-GABA terminals in the LH acutely decreased body temperature, but we do not know the responsible cells.Each cell shows a different rate of mIPSCs; therefore, we used changes in percent mIPSC occurrence to normalize the changes in mIPSC frequency between conditions (that is, 5 s for before and stimulation and 10 s for after (mean of five trials per cell)).Bar graphs, mean mIPSC frequency changes (baseline versus stim, P = 0.0156; baseline versus after, P = 0.0312) and mean amplitude changes (baseline versus stim, P = 0.0156) with two-tailed Wilcoxon matched-paired signed-rank test.*P < 0.05.Mean (bar) and individual (before-after line).Rec, recording electrode.b, Gene expression matrix for LPO cells that responded to stimulating PFC Sst terminals.n = 12 neurons, N = 7 mice.Mean (line) ± s.e.m. (shading).f, mIPSCs of postsynaptic LH cells from Sst-PFC-ChR2-Tag:SD mice (4AP and TTX were present for baseline, and 5 s of 10-Hz optostimulation was used).Bar graphs: changes in percent mIPSC occurrence to normalize the changes in mIPSC frequency between conditions (that is, 5 s for pre-stimulation and 10 s for post-stimulation (mean of five trials per cell)).Mean mIPSC frequency changes (baseline versus stim, P < 0.0001, baseline versus after, P < 0.0001) and mean amplitude changes (baseline versus stim, P = 0.7609) with two-tailed Wilcoxon matched-paired signed-rank test.n = 19 neurons, N = 6 mice.g, Gene expression matrix for LH cells that responded to optostimulation of PFC Sst terminals.n = 19 neurons, N = 6 mice.glut, glutamatergic.See also Extended Data    In summary, our findings indicate that the PFC issues top-down instructions to the hypothalamus to regulate both behavioral preparation for sleep (nesting and increased body temperature) and activation of sleep-induction circuitry that induces NREM sleep (and associated lower body temperature), ensuring that optimal sleep takes place in a suitable place.
The AAV was a mixed capsid serotype (AAV1 and AAV2).To produce AAVs, the adenovirus helper plasmid pFΔ6 and the AAV helper plasmids pH21 (AAV1) and pRVI (AAV2) and the pAAV transgene plasmids were all co-transfected into HEK293 cells, and the subsequent AAV particles were collected on heparin columns 62 .This was done in house.The virus titer dilutions and volumes used for each experiment are listed in Supplementary Tables 1 and 2.
For sleep recordings, two electromyography (EMG) wire electrodes were inserted in the neck extensor muscles and two EEG screw electrodes were placed at ML (−1.5 mm), AP (+1.5 mm) and ML (−1.5 mm), AP (−2.0 mm) relative to the bregma.A third EEG electrode was placed at ML (+1.5 mm), AP (−2.0 mm) for optogenetic recording.All instrumented mice were housed singly to avoid damage to the head stage and were allowed to recover and, for the viral transgenes, to adequately express for at least 3 weeks.

Activity-tagging behavioral protocols and controls
This was carried out similarly to how we did this previously 9,38,39 .Two AAVs, AAV-P cFos -tTA and AAV-P TRE-tight -flex-'effector gene' (for example, ChR2-EYFP, hM3Dq-mCherry), were bilaterally co-injected into the PFC of Vgat Cre or Nos1 Cre or Sst Cre mice.To repress the activity-tagging system, mice were maintained on Dox-containing chow for 1 week before the surgery and at least 3 weeks after the surgery.Before SD, mice were taken off Dox for 2 d and then deprived of sleep for 5 h by introducing new objects, beginning in the new cage at the start of the 'lights-on' (ZT 0) period.Mice were gently placed back into their home cages with Dox-containing chow and allowed RS.Mice were habituated to Neurologger 2A EEG recording devices for at least 2 d before SD and RS were performed.During this time, a 24-h EEG-EMG baseline recording was obtained, and SD and RS were monitored and confirmed offline.Any mice that failed to show 5 h of clear SD and an RS accompanied by a delta power increase were discounted from the chemogenetic or optogenetic experiments.
Singly housed mice were kept in their home cage with Doxcontaining chow before and after SD.Optostimulation was carried out in the home cage, but any existing nest was destroyed, and the material was mixed with new nesting materials to reduce the habituation period.Food was purposely placed closer to the water bottle and away from the nesting materials to segregate nest-building behavior from food-and water-seeking behaviors (Extended Data Fig. 1).

EEG-EMG recordings and analysis
EEG and EMG signals were recorded using Neurologger 2A devices 63 at a sampling rate of 200 Hz, and the data were visualized with Spike2 software (Cambridge Electronic Design).EEG signals were high-pass filtered offline at 0.5 Hz (−3 dB), and EMG signals were bandpass filtered offline at 5-45 Hz (−3 dB).To define the vigilance states of wake and NREM and REM sleep, delta power (0.5-4.5 Hz), theta power (5-10 Hz) and theta/delta (T:D) ratios were calculated.Automated sleep scoring was performed using a Spike2 script, and the results were manually corrected.

Chemogenetics
Mice were allowed to habituate to the Neurologger 2A devices minimally 2 d before SD and RS were performed.Two days after SD, CNO (4936, Tocris, dissolved in saline, 1 mg per kg and 5 mg per kg) or saline was injected i.p. at ZT 18 (that is, mid-'lights-off' period, that is, at the time when mice were most active and least likely to build a nest or sleep), and vigilance states were recorded.Mice were split into random groups that received either saline (day 1) and CNO (day 2) or CNO (day 1) and saline (day 2) injections at the same circadian time https://doi.org/10.1038/s41593-023-01430-4(Extended Data Fig. 1c).Mice were habituated again to the Neurologger 2A devices at least 1 h before ZT 18 (i.p. injection, t = 0).

Consideration of clozapine-N-oxide doses.
The effects of different CNO doses (1, 5 and 10 mg per kg, injected i.p.) on sleep have been systematically tested in wild-type (C57BL/6J) mice that do not express hM3Dq receptors 64 .The injections were given at a time when mice were most sleepy, the beginning of the 'lights-on' period 64 .In the first 2 h following CNO injection, there was no significant main effect on the proportion of time spent awake or in NREM sleep or REM sleep 64 .For NREM sleep, there was no consistent effect of CNO dose on sleep latency, but, at 5 and 10 mg per kg CNO, there was a small but significant effect of prolonging individual NREM episodes and reducing their number, so that sleep architecture was slightly changed, but NREM sleep amount was unchanged.Thus, for our study, when we gave CNO during the 'lights-off' phase when mice were most awake, the difference between 1 and 5 mg per kg CNO is not likely to cause any background effects on sleep-wake dynamics.

Optogenetics
Mice were allowed to habituate to the Neurologger 2A devices minimally 2 d before SD and RS were performed.Optogenetic stimulations were generated by a 473-nm diode-pumped solid-state laser with a fiber coupler (Shanghai Laser, BL473T3T8U-100FC, Shanghai Laser & Optics Century) or a 465-nm Doric Connectorized LED (CLED_465, Doric Lenses).Stimulation protocols were programmed and controlled using Signal software (Cambridge Electronic Design) and Micro1401 (CED) for the laser and Doric Neuroscience Studio version 5.3.3.14 (Doric Lenses).Laser and LED power was kept in the range of 2-5 mW at the tip of the optic fiber (0.8-1.0 mW mm −2 at a depth of 1 mm) unless stated otherwise.
Optostimulation was carried out during ZT 18 (the mid-'lights-off' period in the animal house).Before starting the stimulation protocol, all mice were habituated for at least 30 min to the environment.For controls for Vgat-PFC-ChR2-Tag:SD and Vgat-VC-ChR2-Tag: SD mice, we used Vgat-PFC-GFP mice (AAV-flex-EGFP was injected into the PFC of Vgat Cre mice) and Vgat-PFC-ChR2-Tag mice that had had Dox removed from their diet for the same time duration as the paired experimental cohorts but that had not been deprived of sleep (Vgat-PFC-ChR2-Tag:Ctrl mice).Results from both groups of controls were pooled.For the control for Sst-PFC-ChR2-Tag: SD mice, we used the same Sst-PFC-ChR2-Tag:SD mice before the tagging procedure (Sst-PFC-ChR2-Tag:Ctrl mice).For controls for Sst-PFC-ChR2 mice, we gave no-laser or low-power (0.5-1 mW at the tip, 0.1-0.15mW mm −2 at a depth of 1 mm) optostimulation to the same Sst-PFC-ChR2 mice.

Calcium photometry
This was performed as described previously 65 .Mice were allowed to habituate to Neurologger 2A devices minimally 2 d before SD and RS were performed.Light was generated by a 473-nm diode-pumped solid-state laser with a fiber coupler (Shanghai Laser, BL473T3T8U-100FC, Shanghai Laser & Optics Century) or a 465-nm Doric Connectorized LED (CLED_465, Doric Lenses).Laser and LED power was kept in the range of 70-90 µW at the tip of the optic fiber (0.22-0.30mW mm −2 at maximum).The GCaMP6 output was filtered at 500-550 nm through the fluorescence cube, converted to voltage by a photodiode and then amplified by the lock-in amplifier (SR810, Stanford Research Systems) with a time constant of 30 ms.Photometry, EEG and EMG data were aligned offline using Spike2 software and analyzed using custom MATLAB scripts.For each experiment, the photometry signal F was normalized to the baseline using the function ΔF/F = (F − F 0 )/F 0 , where F 0 is the mean fluorescence across the signal analyzed.The baseline photometry values for photobleaching and photometry signal drift during long recording were corrected with a custom MATLAB script.

Core body temperature recording
Core body temperature was recorded using temperature loggers (DST nano, Star-Oddi) implanted abdominally as described previously 9 .A pre-defined program was set to sample the temperature data every 3 min for baseline core body temperature and during chemogenetic and optogenetic experiments.At the end of the experiments, the loggers were retrieved, and the data were downloaded and analyzed offline.For delta change against baseline analysis, the mean 24-h baseline body temperature was taken from 5 consecutive days of recording before the experimental period.

Behavioral analysis and nest scoring
All behavior was monitored with a video camera, which was placed above the test cage, and analyzed offline after the experiments.All evaluation was carried out on pre-blinded recording data by more than one experimenter.The difference was reviewed and corrected before unblinding.Videos were synchronized with stimulation protocols.Video nesting behavior over time was scored using Behavioral Observation Research Interactive Software (BORIS) 66 and aligned with sleep scoring in Spike2.Nesting behavior was defined as pushing and carrying the nesting material; or fluffing the material up or body wriggling at the center of the nest site and making space for the new nesting material.
Before the initial habituation period (starting at ZT 17) of optogenetic experiments, all previously existing nest material was removed from the home cage of the test mice, and we placed 8 g of a mixture of old and new shredding papers away from food and water.The baseline nest condition was remotely checked 5-10 min before ZT 18 (optostimulation, t = 0) without disturbing the test mice.For chemogenetic experiments, the cage was prepared 30 min to 1 h before i.p. injection at ZT 18 and monitored with an overhead video camera for 5 h.
We evaluated nest scores offline by adapting a five-point scale 1,67,68 : (1) nest materials are not noticeably touched (<10% change from baseline), (2) nest materials are partially gathered (10-50% change from baseline), (3) nest materials are sorted and gathered, but some are spread around the cage (50-90% change from baseline), (4) nest materials are sorted and gathered; identifiable but flat, (5) a perfect or near-perfect nest with a crater.
Nesting effectiveness during an optostimulation session was calculated by multiplying availability (percent, overall nesting time ÷ 25% of overall session time), performance (100%, by assuming that all scored nesting behaviors contribute to nest building) and nest quality (percent, 'good' quality nest (nest score 3-5) ÷ overall nest).

Immunohistochemistry and imaging
Mice were transcardially perfused with 4% paraformaldehyde (Thermo Scientific) in PBS.Brains were removed, and 35-µm-thick coronal sections (unless otherwise specified) were cut using a Leica SM2010R microtome or a Thermo Scientific HM 450 Sliding Microtome.Freefloating sections were rinsed once with PBS and processed for epitope retrieval by incubating sections in 0.05% Tween-20 in 10 mM sodium citrate buffer (pH 6.0) at 80-85 °C for 30 min.Sections were allowed to cool down to room temperature and then washed three times with PBS for 10 min.Sections were blocked with a solution of 20% goat serum (NGS, Vector), 0.2% Triton X-100 and PBS for 1 h at room temperature and incubated with primary antibody at an adequate dilution in 2% NGS, 0.2% Triton X-100, PBS solution overnight at 4 °C.Incubated slices were washed three times with PBS for 10 min at room temperature and incubated with a secondary antibody (Molecular Probes) at an adequate dilution in 2% NGS, 0.2% Triton X-100, PBS solution for 1.5 h at room temperature.Slices were washed three times with PBS for 10 min at room temperature and incubated with Hoechst 33342 (Life Technologies) at 1:5,000 in PBS for up to 10 min at room temperature.After a double wash with PBS, slices were mounted on slides with ProLong Gold Antifade Reagent (Invitrogen).Primary antibodies were Extended Data Fig. 1 | Experimental schematics for activity-tagging (supports Figs.1-4).a, Schematic of activity-tagging method and behavioral experiments for Vgat-expressing neurons in the PFC with ChR2-EYFP during sleep deprivation (SD).Dox chow was re-introduced immediately after 5hrs of SD (ZT0-ZT5) and animals were allowed to carry out post-SD nesting and RS without any disturbance while the activity-tagging system is gradually repressed.b, Schematic of activity-tagging opto-stimulation protocols.Prior to habituation period (1 hr), nesting materials in the home cage was reduced and mixed with new nest materials (total approx.8 g) and food pellets were moved to the far end away from the nesting materials and closer to the water bottle spout.Two separate opto-stimulation protocols were given at ZT18 after 1 hour of habituation.c, Schematic of activity-tagging chemogenic reactivation protocols.Saline and CNO were given 2 days after activity-tagging via i.p. injection with blinding.The mice had 2 days rest time between the 1 st and 2 nd injections.

Fig. 1 | 4 Fig. 2 |
Fig. 1 | SD, nesting behavior, RS and corresponding changes in core body temperature.a, Example EEG-EMG traces and sleep stage state and mean core body temperature during SD and RS and post-SD nesting activity and nest materials in the home cage.N = 7 Vgat Cre mice, baseline versus SD, P = 4.78 × 10 −4 (ZT 0-5), P = 1.12 × 10 −3 (ZT 5-6) with two-way repeated-measures (RM) ANOVA with Bonferroni correction.Freq, frequency.b, Raster plot of RS nesting, example nest image in the home cage and relative EEG spectrum.Red raster and solid black bars indicate nesting and onset of first consolidated NREM RS.Yellow dot, position of mouse; dashed red line, outline of nesting materials; solid red line, outline of nest.N, number of biologically independent mice.***P < 0.001.Mean (line) ± s.e.m. (shading).See also Extended Data Figs. 1 and 2. DP, dark period.

Fig. 7 |
Fig. 7 | PFC Sst projections to the LPO hypothalamus generate inhibitory currents on postsynaptic cells.a, mIPSCs from postsynaptic cells in the LPO hypothalamus from Sst-PFC-ChR2-Tag:SD mice.Example traces were recorded with 4AP and TTX present for baseline and with 5 s of 10-Hz optostimulation.Each cell shows a different rate of mIPSCs; therefore, we used changes in percent mIPSC occurrence to normalize the changes in mIPSC frequency between conditions (that is, 5 s for before and stimulation and 10 s for after (mean of five
https://doi.org/10.1038/s41593-023-01430-4Extended Data Fig.3| Activity-tagging controls (supports Fig.2).This Figure shows a series of controls using Vgat-PFC-hM3Dq-Tag mice (n mice = 3 each).All groups of mice were on the Dox diet prior to AAV transgene injection into the PFC.a, Experimental group.Mice were on Dox for 4 weeks (4w), then Dox was removed from the diet for 2 days (2d) and during sleep deprivation (SD), and Dox was made available straight after the SD, during the start of nesting and recovery sleep (RS).Four days later, mice were transcardially-perfused and the brain sections were immuno-stained, in this case, for induced mCherry expression in the PFC.Images in the right hand-column are magnifications of PFC cells with induced gene expression.b, First type of 'on Dox' control group.Mice were on Dox for 4 weeks, then Dox was removed from the diet for 2 days, and then Dox was re-provided during the sleep deprivation procedure and remained in the diet for 4 weeks until transcardio-perfusion.This shows that in the presence of near continuous Dox, except for 2 days interruption, no transgene expression was induced above background.c, Second type of 'on Dox' control group.Mice were on Dox continuously, including for the periods prior to SD.No transgene expression was induced above background.d, Baseline 'off Dox' control group.Mice were on Dox for 4 weeks after AAV injection and Dox was removed for 2 days plus an extra 24 hours to correspond with the experimental group, and then mice were placed back on Dox and perfused 4 days later.Although some transgene expression occurred during this time, there was no clear cell labelling at higher magnifications.n mice = number of biologically independent mice.PL, prelimbic; IL, infralimbic.Scale bars: 100 µm (white), 20 µm (yellow).

Mice
This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material.If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.All experiments were performed in accordance with the UK Home Office Animal Procedures Act (1986) and approved by Imperial College's Animal Welfare and Ethical Review Body.The following types of mice were used: Vgat-ires-Cre (Slc32a1 tm2(cre)Lowl /J) mice ( Jackson Laboratory stock 016962), kindly provided by B. B. Lowell (Beth Israel Deaconess Medical Center & Harvard Medical School, USA)