Galanin neurons in the hypothalamus link sleep homeostasis, body temperature and actions of the α2 adrenergic agonist dexmedetomidine

Sleep deprivation induces a characteristic rebound in NREM sleep accompanied by an immediate increase in the power of delta (0.5 - 4 Hz) oscillations, proportional to the prior time awake. To test the idea that galanin neurons in the mouse lateral preoptic hypothalamus (LPO) regulate this sleep homeostasis, they were selectively genetically ablated. The baseline sleep architecture of LPO-ΔGal mice became heavily fragmented, their average core body temperature permanently increased (by about 2°C) and the diurnal variations in body temperature across the sleep-wake cycle also markedly increased. Additionally, LPO-ΔGal mice showed a striking spike in body temperature and increase in wakefulness at a time (ZT24) when control mice were experiencing the opposite - a decrease in body temperature and becoming maximally sleepy (start of “lights on”). After sleep deprivation sleep homeostasis was largely abolished in LPO-ΔGal mice: the characteristic increase in the delta power of NREM sleep following sleep deprivation was absent, suggesting that LPO galanin neurons track the time spent awake. Moreover, the amount of recovery sleep was substantially reduced over the following hours. We also found that the α2 adrenergic agonist dexmedetomidine, used for long-term sedation during intensive care, requires LPO galanin neurons to induce both the NREM-like state with increased delta power and the reduction in body temperature, characteristic features of this drug. This suggests that dexmedetomidine over-activates the natural sleep homeostasis pathway via galanin neurons. Collectively, the results emphasize that NREM sleep and the concurrent reduction in body temperature are entwined at the circuit level. Significance Catching up on lost sleep (sleep homeostasis) is a common phenomenon in mammals, but there is no circuit explanation for how this occurs. We have discovered that galanin neurons in the hypothalamus are essential for sleep homeostasis as well as for the control of body temperature. This is the first time that a neuronal cell type has been identified that underlies sleep homeostasis. Moreover, we show that activation of these galanin neurons are also essential for the actions of the α2 adrenergic agonist dexmedetomidine, which induces both hypothermia together with powerful delta oscillations resembling NREM sleep. Thus, sleep homeostasis, temperature control and sedation by α2 adrenergic agonists can all be linked at the circuit level by hypothalamic galanin neurons.


Introduction
It has been proposed that sleep aids metabolite clearance (1), synaptic down-scaling (2), stress reduction (3) and protection of the heart (4). Disruption of sleep causes many changes in brain gene expression and the blood plasma metabolome (5)(6)(7). Perhaps reflecting the fundamental restorative purpose(s) of sleep, the urge to sleep, known as the homeostatic drive, increases with the time spent awake and dissipates during sleep (8). Sleep deprivation causes a characteristic rebound in NREM sleep accompanied by an immediate increase in the power of delta (0.5 -4 Hz) oscillations (deeper sleep) and amount of subsequent NREM sleep, proportional to the previous time spent awake (8)(9)(10)(11). Widely expressed genes (e.g. Sik3, Adora1, clock, mGluR5, per3, reverba) have been found to modulate sleep homeostasis (9,(11)(12)(13)(14)(15)(16)(17). Astrocytes and skeletal muscle release messengers which can modulate the process (18,19). But in mammals little is known about how sleep homeostasis might work at the neuronal circuit level, or even whether the homeostatic drive is primarily locally or globally determined (20)(21)(22).
There is strong evidence that points towards the preoptic (PO) hypothalamus as playing a pivotal role (23). During sleep deprivation and recovery sleep, neurons in this area, as well as in the neighboring bed nuclei of the stria terminalis, become excited (24)(25)(26)(27). The preoptic hypothalamus also houses circuitry that regulates body temperature (28)(29)(30). cFOS-dependent activity-tagging revealed that after sleep deprivation, reactivating the tagged neurons in the preoptic area induced both NREM sleep and body cooling (26). Indeed, NREM sleep induction and core body cooling are linked by common preoptic circuitry (31), and about 80% of brain cortex temperature variance correlates with sleep-wake states (32). On entering NREM sleep, the neocortex of rats and mice cools rapidly (33,34).
These a2 adrenergic agonists are increasingly favored over benzodiazepines for long-term sedation (44). Although it used to be thought that dexmedetomidine induces sedation by inhibiting noradrenaline release from neurons in the locus ceruleus (43,45,46), there is building evidence that this is not the case (26,36,47). Dexmedetomidine induces cFOS expression in the preoptic hypothalamic nuclei (26,48), and can induce sedation even when noradrenaline release from the locus ceruleus is genetically removed (47). Using c-FOS-based activity-tagging, we found previously that dexmedetomidine requires the LPO hypothalamus to induce both NREM-like sleep and hypothermia and because we obtained similar results following sleep deprivation (see above), we suggested that dexmedetomidine-induced sleep/hypothermia probably involved the same neurons as those activated during sleep deprivation/recovery sleep (26).
Given this potential overlap, the question is whether specific cell types can be identified in the preoptic hypothalamus that are involved in recovery sleep, hypothermia and the actions of a2 adrenergic agonists. Here, we show that the neurons mediating sleep homeostasis after sleep deprivation and dexmedetomidine-induced NREM-like sleep are LPO neurons that express the inhibitory peptide galanin. In mice with selectively lesioned LPO galanin neurons, body temperature is permanently elevated and the sleepwake cycle is heavily fragmented. Without galanin neurons sleep homeostasis is blunted (no increase in delta power) and the ability of dexmedetomidine to induce high-power NREM-like oscillations and sustained hypothermia is substantially diminsihed. Thus, recovery sleep after sleep deprivation and the deep NREM-like sleep and hypothermia induced by a2a agonists depend on the same hypothalamic circuitry.

Selective genetic ablation of mouse lateral preoptic galanin neurons.
To selectively ablate LPO Gal neurons, we bilaterally injected a Cre-activatable AAV expressing Caspase 3 (AAV-FLEX-CASP3) into the LPO area of Gal-Cre mice to generate LPO-DGal mice (Fig. 1A). To confirm LPO Gal neuron ablation, we mixed AAV-FLEX-GFP and AAV-FLEX-CASP3 viruses (Fig. 1A). As controls, Gal-Cre gene-positive littermates were injected only with AAV-FLEX-GFP virus to generate LPO-Gal-GFP mice (Fig. 1A). The injection coordinates targeted galanin neurons in the LPO (and partially the edge of the MPO area). In the LPO-DGal mouse group immunohistochemistry with GFP antibodies showed that after five weeks the AAV-FLEX-CASP3 injections eliminated ~98% of LPO Gal cells, as compared with LPO-Gal-GFP littermate controls ( Fig. 1B-D).

Selective ablation of lateral preoptic galanin neurons induces a chronic increase in body temperature.
Five weeks after ablation of galanin neurons in the LPO area, LPO-DGal mice had a striking increase in their average core body temperatures compared with LPO-Gal-GFP mice ( Fig. 2A, B). In a continuous recording of body temperature over 5 days, the mice still retained a normal diurnal variation of their body temperature with a higher temperature during "lights-off" period (active phase) and lower temperature during "lightson" period (inactive phase) ( Fig. 2A, B). However, the average body temperature of the LPO-DGal mice was raised to 37°C, compared with the average 35.5°C of the LPO-Gal-GFP controls (Fig. 2C). In addition, the range of body temperature change during the 24hour cycle increased from 1°C to 2°C. In the LPO-Gal-GFP control group, the average body temperature during the day and night was around 36°C and 35°C respectively, whereas the LPO-DGal group had their average day and night body temperatures around 38°C and 36°C respectively (Fig. 2C). Thus, LPO Gal neurons must be acting chronically to induce body cooling. 7 A new feature also emerged in the diurnal temperature variation of the LPO-DGal mice. In LPO-DGal mice, a pronounced spike in body temperature appeared just prior to the transition from "lights off" to "lights on", which was not evident in the LPO-Gal-GFP control mice (Fig 2A, B, highlighted with red bars) (see Discussion). Consistent with the above findings, chemogenetic activation of LPO Gal neurons with CNO in LPO-Gal-hM3Dq mice induced hypothermia (Fig. S1A, B), as also reported by others (49). CNO had no measurable effect on baseline temperature in control mice ( Fig. S1C).

Ablation of galanin neurons in the LPO area increased sleep-wake fragmentation.
We examined how LPO Gal neuron ablation influenced the 24-hour sleep-wake cycle (12 hours lights on: 12 hours lights off) ( Fig. 3 & Fig. S2). Example EEG and EMG spectra are shown in Figure S2A. Chronic ablation of LPO galanin neurons caused a modest reduction in total wake time and an increase in total NREM time during "lights off", but no change during the "lights on" period ( Fig. 3A). The amount of REM sleep was unaffected.
Furthermore, there were no significant differences in the baseline EEG power in either the WAKE state or NREM state between LPO-Gal-GFP and LPO-DGal mice (Fig. S2B).
Sleep architecture, however, became highly fragmented following LPO Gal neuron ablation ( Fig. 3B). The number of WAKE and NREM episodes increased markedly, while their durations were shortened. These effects were most marked during the "lights off" period.
The number of REM sleep episodes and their durations were not affected (Fig. 3B). The number of WAKE to NREM and NREM to WAKE transitions were significantly increased ( Fig. 3C), but transitions between other vigilance states did not change.
Chemogenetic activation of LPO Gal neurons with CNO in LPO-Gal-hM3Dq mice induced NREM sleep (Fig. S3 A,B,C), as also reported by others (49). The power of this CNO-induced NREM sleep was higher than baseline power of NREM sleep after saline 8 injection (Fig. S3D). CNO had no measurable effect on baseline sleep after saline injection in control mice (Fig. S3E).

Ablation of LPO galanin neurons abolishes sleep homeostasis after sleep deprivation.
To examine how LPO Gal neurons regulate sleep homeostasis, a 5-hour sleep deprivation was applied to both groups of mice. In control LPO-Gal-GFP mice, there was a strong reduction in wakefulness and an increase in total sleep (NREM + REM sleep) following five hours of sleep deprivation (Fig. 4A). The main effect, compared to the baseline diurnal variation in wake and sleep times, occurred during the "lights off" period following sleep deprivation (which was carried out during the "lights on" period) (Fig. 4A). During the sleep rebound of LPO-Gal-GFP mice, the power in the delta wave band (0.5 -4 Hz) was also significantly increased compared to the delta power during baseline sleep at the equivalent zeitgeber time (Fig. 4B, E). This delta power increase is a characteristic of recovery sleep (8). In LPO-DGal mice, however, there was no change in WAKE or total sleep (NREM + REM) time following five hours of sleep deprivation (Fig. 4C). The increase in delta power seen in control animals following sleep deprivation was also abolished in LPO-DGal mice (Fig. 4D, E). Most (~80%) of the sleep lost as a result of five hours of sleep deprivation was recovered after 19 hours in LPO-Gal-GFP mice, whereas only ~22% of the sleep loss was recovered in LPO-DGal mice (Fig. 4F). Indeed, the sleep recovery rate after sleep deprivation was significantly reduced in LPO-DGal mice compared with LPO-Gal-GFP (Fig. 4F, G).

Ablation of LPO Gal neurons strongly reduced dexmedetomidine-induced
hypothermia.
An unwanted side-effect of dexmedetomidine is that it induces marked hypothermia (26,35). In control LPO-Gal-GFP mice, after injection (i.p.) of 50 µg/kg of dexmedetomidine, 9 there was a strong reduction in core body temperature from about 36 o C to 25 o C over the course of 2 hours (post-dexmedetomidine injection) (Fig. 5A, B). This hypothermia persisted beyond 4 hours post-injection. In LPO-DGal mice, however, the initial reduction in body temperature after dexmedetomidine injection lasted only for the first hour, did not reach the same nadir as in LPO-Gal-GFP control mice, and the body temperature nearly returned to baseline levels over the next hour (Fig. 5A, B).

Dexmedetomidine requires LPO Gal neurons to induce NREM-like delta power
We concurrently investigated if ablation of LPO Gal

Discussion
Previously, using cFOS-dependent activity-tagging, we found that neurons in the LPO area were sufficient to recapitulate NREM-like sleep and body cooling after both sleep deprivation and dexmedetomidine administration, but we did not identify the cells involved (26). Here, using selective genetic lesioning, we have demonstrated that these are likely to be galanin neurons. Without LPO galanin neurons, the sleep-wake cycle becomes highly fragmented, and sleep homeostasis (the enhanced delta power following sleep deprivation and the extra NREM sleep that follows) is diminished, suggesting that LPO galanin neurons track the time spent awake. Although previously, genes have been identified that modulate sleep homeostasis (see Introduction), we describe here the first neuronal cell type implicated in sleep homeostasis. Throughout the animal kingdom, the homeostatic sleep drive is reflected as changing neuronal activity with the time spent awake (23,26,51). Sleep homeostasis at the circuit level in mammals, however, has remained mysterious. It appears to be, at least in part, mediated by extracellular adenosine (9,14), released by astrocytes (18). Adenosine levels, however, only increase during wakefulness in the basal forebrain and not the PO area (52,53), and so cannot be the direct trigger for LPO galanin neurons to induce NREM sleep. Skeletal muscle can regulate, by an unknown messenger, sleep homeostasis (19), and could also activate LPO galanin neurons, for example.
We find that galanin neurons in the same LPO area are required for a substantial part of the a2 adrenergic agonist dexmedetomidine's actions in inducing its characteristic substantially high NREM-like delta power (above that of baseline NREM sleep) and sustained body cooling, characteristics which seem to be an exaggeration of recovery sleep after sleep deprivation. Thus, we suggest that the sleep homeostasis circuitry and the circuitry targeted by adrenergic sedatives are likely to be the same. 11 The circuitry in the PO area that regulates body temperature seems complex (28,30). Certain PO neurons (e.g. GABA/galanin-, Glut/NOS1-, PACAP/BDNF-, and TRPM2expressing cells), respond to immediate external or internal thermal challenge by acutely initiating body cooling or heating (28-31, 49, 54-56), but no information has been available for how genetically-specified PO neurons chronically regulate body temperature. We find that without LPO galanin neurons, the diurnal body temperature rhythms of the LPO-DGal mice are shifted permanently several degrees higher ( Fig. 2A). Thus, galanin neurons are contributing to chronic cooling of the body, correlated with the mice having considerable sleep-wake fragmentation. Lesioning of the rat VLPO area, an extremnely focal lesion, produced chronically less NREM with decreased delta power and decreased REM sleep, but body temperature was unaffected (57). This suggests the extreme ventral part of the PO does not contain temperature regulating cells, at least in the rat.
The so far unexplained active link between body cooling and NREM sleep induction seems tantalizing. It was proposed many years ago that the restorative effects of sleep homeostasis depended on lower body temperature (58). Cooling might be actively linked to sleep because cooling during sleep induces cold-induced gene expression that could remodel synapses or serve some other restorative function (33).
An extension of this process would be moving deeper into torpor and hibernation (59), where such gene products are also induced and could be involved in rebuilding synapses on arousal from hibernation (59,60).
In humans, NREM sleep induction appears when the rate of core body temperature decline is at its maximum (61). A new feature has emerged in the diurnal core body temperature variation we have observed in LPO-DGal mice: a pronounced positive spike appeared in their body temperature around the transition from "lights off" to "lights on" at ZT24 (Fig 2A, B), suggesting that LPO galanin neurons would normally be particularly active in driving down body temperature at the point in the diurnal cycle where sleep pressure is highest at the start of the "lights on" period. This would fit with their role in regulating sleep homeostasis. Increasing sleep pressure during the "lights off" wake period would result in LPO galanin neurons becoming active at this transition to lights on, inducing both sleep and driving down body temperature. This could explain why LPO-DGAL mice are actually more awake compared with LPO-GAL-GFP mice at the start of "lights on" (see red bar in Fig. 3A), further emphasizing the link between NREM sleep induction and body cooling. mating (69,70), as well as temperature and sleep (49,56). We cannot rule out if one type of LPO galanin neuron increases NREM delta power, and another promotes chronic cooling. Intersectional genetics would be needed to further target these cells to resolve this, but this will be a complex challenge. In fact, single-cell profiling and multiplex in situ labelling of the PO region found at least seven subtypes of galanin-expressing neuron (68). Most of these subtypes are GABAergic and express the vgat gene, but several are glutamatergic because they expressed the vglut2 gene, and one vgat/galanin subtype also expressed tyrosine hydroxylase and the vesicular monoamine transporter (68).
LPO galanin neurons in our study are likely to be GABAergic. We found previously that deletion of the vesicular GABA transporter (vgat) expression in LPO, preventing GABA release from LPO GABAergic neurons, abolished dexmedetomidine's ability to rapidly induce NREM-like sleep (there was no immediate increase in NREM delta power in the first 10 mins, although after an hour dexmedetomidine could still induce full sleep), 14 suggesting that LPO GABAergic neurons were critical for the initial actions of dexmedetomidine (26). Sustained galanin release would be necessary for the longer-term effects of dexmedetomidine on NREM sleep maintenance and lower body temperature.
In conclusion, based on our lesioning results, LPO galanin neurons are at the intersection of NREM sleep induction and body cooling. Although NREM sleep can still occur without these cells, they are needed for chronically cooling to maintain the normal core body temperature. Furthermore, LPO galanin are needed for sleep homeostasis. A similar result has just appeared for zebrafish, suggesting that sleep homeostasis is a primordial function of PO galanin neurons (71). We propose that the control of sleep homeostasis is actively linked to body cooling. Sustained stimulation of these galanin neurons with the a2 adrenergic agonist dexmedetomidine can induce a slide into a torpor like state (if body temperature is not corrected). Thus, these two processes, sleep homeostasis and a2a receptor sedation/torpor induction could be linked at the circuit level by hypothalamic LPO galanin neurons, which serve, in the initial phase of stimulation, to produce heightened NREM delta power above that of baseline sleep. 15

Mice. Animal care and experiments were performed under the UK Home Office Animal
Procedures Act (1986) and were approved by the Imperial College Ethical Review

Generation of recombinant AAV particles.
All AAV transgenes were packaged in our laboratory into AAV capsids with a mixed serotype 1 & 2 (1:1 ratio of AAV1 and AAV2 capsid proteins) as described previously (75).

Surgeries and stereotaxic injections of AAV.
For surgery, mice were anesthetized with an initiation concentration of 2. where temperature recordings were necessary, temperature loggers were usually inserted (abdominally) two to three weeks after mice had had their viral injection surgeries.

EEG and EMG recordings and vigilance states scoring. Non-tethered EEG and EMG
recordings were captured using Neurologger 2A devices (76). Screw electrodes were chronically inserted into the skull of mice to measure cortical EEG using the following coordinates: -1.5 mm Bregma, + 1.5 mm midline -first recording electrode; + 1.5 mm Bregma, -1.5 mm midline -second recording electrode; -1 mm Lambda, 0 mm midlinereference electrode). EMG signals were recorded by a pair of stainless steel electrodes implanted in the dorsal neck muscle. Four data channels (2 of EEG and 2 of EMG) were recorded with four times oversampling at a sampling rate of 200 Hz. The dataset was downloaded and waveforms visualized using Spike2 software (Cambridge Electronic Design, Cambridge, UK) or MATLAB (MathWorks, Cambridge, UK). The EEG signals were high-pass filtered (0.5 Hz, -3dB) using a digital filter and the EMG was band-pass filtered between 5-45 Hz (-3dB). Power in the delta (0-4 Hz), theta (6-10 Hz) bands and theta to delta band ratio were calculated, along with the root mean square (RMS) value of the EMG signal (averaged over a bin size of 5 s). All of these data were used to define the vigilance states of WAKE, NREM and REM by an automatic script. Each vigilance state was screened and confirmed manually afterwards. The peak frequency during NREM epochs were analyzed using Fourier transform power spectra to average power spectra over blocks of time.
A pre-defined program was set to sample the temperature data every two minutes for baseline core body temperature and drug/vehicle administration. At the end of the experiments, the loggers were retrieved and the data were downloaded and analyzed.

Sleep deprivation and recovery sleep. The sleep deprivation protocol was similar to
the one we used before (26), and it started at Zeitgeber time (ZT) zero (17:00), the start of the "lights-on" period when the sleep drive of the mice is at its maximum. Both experimental and control groups were sleep deprived for 5 hours by introducing novel objects or gently tapping on the cages. After sleep deprivation, mice were allowed to return back to their home cages for recovery NREM sleep. EEG and EMG signals together with temperature data were recorded for analysis.
Chemogenetics and behavioral assessment. For chemogenetic activation, clozapine-N-oxide (CNO) (C0832, Sigma-Aldrich) was used. 1 mg/kg of CNO dissolved in saline or saline in same volume was administrated by intraperitoneal injection (i.p.) and the vigilance states and core body temperature were recorded. Mice were split into random groups that either received CNO or saline injection for an unambiguous comparison.
Drugs were administrated at ZT18 (11:00, "lights-off") when the mice were in their most active period and had their highest body temperature.

Dexmedetomidine experiments. Prior to dexmedetomidine injection, animals with
implanted temperature loggers were fitted with Neurologger 2A devices, and one hour of both baseline vigilance states and core body temperature was recorded as reference. 50 18 µg per kg of dexmedetomidine (Tocris Bioscience) was dissolved in saline and delivered i.p. at ZT19 (12:00, "lights-off"). Animals were placed back to their home cage immediately after injection for a further five-hour recording and the EEG, EMG and core body temperature were simultaneously recorded. A six-hour baseline recording from the same mouse of its natural sleep-wake cycle and core body temperature between ZT18 to ZT24 (11:00 -17:00, "lights-off") was used for parallel comparison with the dexmedetomidine injection experiments.  All error bars represent the SEM.    All error bars represent the SEM.