Central circuit controlling thermoregulatory inversion and torpor-like state

To maintain core body temperature in mammals, the CNS thermoregulatory networks respond to cold exposure by increasing brown adipose tissue and shivering thermogenesis. However, in hibernation or torpor, this normal thermoregulatory response is supplanted by “thermoregulatory inversion”, an altered homeostatic state in which cold exposure causes inhibition of thermogenesis and warm exposure stimulates thermogenesis. Here we demonstrate the existence of a novel, dynorphinergic thermoregulatory reflex pathway between the dorsolateral parabrachial nucleus and the dorsomedial hypothalamus that bypasses the normal thermoregulatory integrator in the hypothalamic preoptic area to play a critical role in mediating the inhibition of thermogenesis during thermoregulatory inversion. Our results indicate the existence of a neural circuit mechanism for thermoregulatory inversion within the CNS thermoregulatory pathways and support the potential for inducing a homeostatically-regulated, therapeutic hypothermia in non-hibernating species, including humans.


Introduction
The core body temperature (T CORE ) of mammals is normally maintained within a narrow range that is optimal for enzymatic reactions and cellular function. The neural circuitry for normal thermoregulation alters the neural out ows to thermoeffector tissues in response to changes in skin and core temperatures to minimize deviations in T CORE 1 . In particular, brown adipose tissue (BAT) and skeletal muscle (shivering) thermogenesis are increased during exposure to a low ambient temperature (T AMB ) and inhibited in a warm T AMB 2-6 . However, when certain mammals enter torpor or hibernation [7][8][9] , their brain circuits for thermoregulation switch from a normal to an inverted state in which the response to a low T AMB is an inhibition of thermogenesis, which induces hypothermia and a reduction in energy consumption.
We have recently discovered that the central activation of A1-adenosine receptors induces a torpor-like state in rats [10][11][12] , and that the hypothermia and hypometabolism that characterize this torpor-like state arises from the induction of a novel thermoregulatory state of thermogenesis that we have called thermoregulatory inversion (TI) 10,13 . In the TI state, as in natural torpor/hibernation, the CNS control of thermogenesis is inverted, such that thermogenesis is inhibited in a cold T AMB, and stimulated in response to a warm T AMB .

Pre-dmh Transx Inverts The Normal Thermoregulatory Shivering Response
Since skeletal muscle shivering is the most signi cant source of thermoregulatory thermogenesis in humans 26 , it is important to determine if the thermoregulatory circuitry controlling shivering 22 can also be manipulated to transition to the TI state in which cooling would inhibit shivering and allow T CORE to fall as it does in torpor/hibernation.
A pre-DMH transX to -9 mm from the dorsal surface of the brain did not produce any change in nEMG in warm rats (T SKIN = 36.4 ± 0.2°C; ΔnEMG: -99.9 ± 88.4% of pre-transX control; n = 5, p = 0.5; Figs Complete pre-DMH transX inverts the thermoregulatory shivering response With T CORE and T SKIN in a warm condition and nEMG at a low, non-shivering level, a complete pre-DMH transX to -10 mm from the dorsal brain surface produced an immediate and remarkable increase in nEMG and nuchal muscle shivering (T SKIN  These results indicate that in the TI state increases in BAT and shivering thermogenesis are dependent on a glutamatergic excitation of thermogenesis-promoting neurons in the DMH. Since a complete pre-DMH transX (Fig. 2D) or a nanoinjection of muscimol in POA (Fig. 1C) induces a robust TI state, it seems unlikely that the source of the glutamatergic input to the DMH required for the skin warming-evoked activation of BAT and shivering thermogenesis in the TI state is located within the POA, as it is for normal thermoregulation 16 , but rather from neurons located caudal to the pre-DMH transX.
To provide evidence that in the TI state skin warming activates DMH neurons that project to rRPa, we examined the Fos expression (Fos-ir) in DMH neurons that were retrogradely-labeled with FluoroGold (FG) injected into the rRPa in warm-exposed rats after a pre-DMH transX (Warm-T rats  DMH-projecting neurons in the PBN are activated during skin warming in anesthetized, naïve rats and in anesthetized, pre-DMH transX rats. We compared anatomical assessments of PBN neuronal activation (Fos-ir) in 4 groups of anesthetized rats: naïve rats and pre-DMH transX rats during skin warming and during skin cooling. Our injections of the retrograde tracer, cholera toxin subunit b (CTb), in the DMH overlapped with DMH neurons retrogradely-labeled following injections of another retrograde tracer, Fluorogold (FG), in the rRPa (Extended Data Figs. 1A, 1B), and resulted in CTb retrograde labeling of neurons in the elPBN and dlPBN (Extended Data Fig. 1C). This basic anatomical result is consistent with the potential for PBN neurons to directly in uence the activity of thermogenesis-promoting neurons in the DMH. There was no difference between the number of CTb-ir neurons in the elPBN and in the dlPBN in the 4 treatment groups (p > 0.05, Fig. 5E). To analyze the extent of Fos expression in dlPBN and elPBN neurons (Fig. 5A) that were retrogradely labeled from CTb injections in DMH (Fig. 5D), we calculated the percent of CTb-labeled neurons in elPBN and in dlPBN that were also Fos-ir (% CTbFos/CTb; Fig. 5C Pattern of activation of DMH-projecting neurons in the PBN in anesthetized, naive rats and in anesthetized, pre-DMH transX rats after cold exposure Skin cooling in anesthetized, naïve rats (Cold-N), a condition in which thermogenesis is strongly stimulated, activated 27 ± 0.95% of the total elPBN neurons that project to DMH (Figs. 5B, 5C). Skin cooling in anesthetized, pre-DMH transX rats (Cold-T), a condition in which thermogenesis is inhibited, activated signi cantly fewer DMH-projecting neurons in elPBN (14.73 ± 2.31%; n = 5, p = 0.003; Figs. 5B, 5C) than in the Cold-N group. This reduction was most prominent at the Intermediate-1 level of the elPBN (Cold-N: 33.09 ± 3.65% vs. Cold-T: 14.52 ± 4.27%, n = 5, p = 0.0151; Fig. 5C). Skin cooling in anesthetized, naive rats (Cold-N) also activated DMH-projecting neurons in dlPBN (11.79 ± 0.81%). A similar fraction (12.98 ± 0.91%) of the DMH-projecting neurons in dlPBN was also activated by skin cooling in anesthetized, pre-DMH transX rats (Cold-T) (Figs. 5B, 5C). These data support the idea that in anesthetized, naive rats, activation of DMH-projecting neurons in elPBN and in dlPBN contributes to the regulation of the discharge of thermogenesis-promoting neurons in DMH during normal thermoregulatory cold defense. In addition, our nding that the activation of DMH-projecting neurons in elPBN was lower in Cold-T than in Cold-N rats would be consistent with a reduced excitation of DMH neurons from their elPBN inputs in the TI state, when skin cooling reduces thermogenesis (Figs. 1, 2, 3).
PBN neurons with direct projections to the DMH are activated during warm and cold exposure in naïve, free-behaving rats We sought to establish that PBN neurons with projections to the DMH were also activated during skin thermoreceptor stimulation in free-behaving rats. One week following CTb injections in the DMH, these naïve, free-behaving rats were exposed to either a warm or a cold T AMB , and Fos expression was quanti ed in DMH-projecting PBN neurons. Since there was no difference between the number of elPBN and dlPBN CTb-retrogradely neurons in warm-or cold-exposed free-behaving rats During warm exposure in naïve, free-behaving rats, a condition in which we expect low levels of thermogenesis, we observed similar percentages of CTbFos/CTb in DMH-projecting neurons in the dlPBN (7.68 ± 1.74%) and within elPBN (5.67 ± 0.95%) (Figs. 6A, 6C). During cold exposure, when thermogenesis should be activated, Fos-ir was also observed in DMH-projecting neurons within dlPBN (3.78 ± 0.93%) and within elPBN (19.41 ± 3.64%) (Figs. 6B, 6C). Of the dlPBN neurons that projected to DMH, a signi cantly greater percentage expressed Fos-ir in warm-exposed rats than in cold-exposed rats (n = 5 per group; p = 0.0416; Fig. 6C). In contrast, for elPBN neurons that projected to the DMH, a signi cantly greater percentage expressed Fos-ir in cold-exposed rats than in warm-exposed rats (n = 5 per group; p = 0.0032; Fig. 6C). In these same rats, we observed FG-ir neurons in the DMH that expressed Fos-ir after cold exposure (Extended Data Fig To determine if the Dyn neurons in the PBN project to the DMH, we performed immunohistochemistry (IHC) for Dyn and CTb in brains from rats that had been injected with CTb in DMH and treated with intracerebroventricular (ICV) colchicine. We observed colocalization of CTb and Dyn mainly in the dense clusters of Dyn neurons in the dlPBN (Fig. 7B). Thus, there is a concentration of VGluT2-expressing Dyn neurons in the dlPBN region, and many of these Dyn neurons project to the region of the DMH that contains thermogenesis-promoting neurons.
To determine whether the activity of DMH-projecting Dyn neurons in dlPBN could in uence the level of thermogenesis in normal thermoregulation or in the TI state, we performed a co-detection procedure to identify DMH-projecting (CTb labeling with IHC) Dyn (pDyn transcripts with ISH) neurons in PBN that were activated (c-fos with ISH) by cutaneous thermal stimuli in naïve and pre-DMH transX rats. We observed DMH-projecting (CTb) Dyn neurons in dlPBN that were activated (c-fos) during skin warming in anesthetized naïve (Warm-N) rats (Fig. 7C), a condition in which we expect thermogenesis to be inhibited.
Noticeably, fewer Dyn neurons in the dlPBN were activated in Cold-N rats than in either Warm-N or Cold-T rats (Fig. 7D). Our observation that DMH-projecting Dyn neurons in the dlPBN are activated in Warm-N rats, when inhibitory in uences on the discharge of thermogenesis-promoting neurons in DMH predominate, is consistent with a signi cant thermogenesis-inhibiting role for these DMH-projecting Dyn neurons in the dlPBN. Such a role for Dyn neurons in the dlPBN is also supported by our nding that more of them are activated in Warm-N and Cold-T rats when thermogenesis is inhibited than in Cold-N rats, when thermogenesis is active (Fig. 7D).

Dynorphin In Dmh Inhibits Normal, Cold-evoked Bat Sna And Bat Thermogenesis
Having identi ed a dynorphinergic projection from the PBN to the region of the DMH containing thermogenesis-promoting neurons, we sought to determine if Dyn in the DMH would affect normal, coldevoked BAT SNA and BAT thermogenesis. Since the degradation products of exogenous Dyn by extracellular peptidases lead to non-speci c activation of NMDA receptors 32 , we pretreated the DMH with the peptidase inhibitor Amastatin. subsequent skin cooling no longer activated BAT SNA (Fig. 8A). Additionally, skin warming had no effect on the post-Dyn completely inhibited level of BAT SNA (Fig. 8A), indicating that Dyn nanoinjection in the DMH does not induce the TI state in which skin warming activates BAT SNA (cf. Figures 1, 3).
A κ-opioid receptor antagonist in the DMH prevents the cold-evoked inhibition of BAT thermogenesis during TI Since a population of Dyn-expressing neurons in PBN projects to the DMH, and Dyn nanoinjection into the DMH inhibits normal, cold-evoked BAT SNA and reduces BAT thermogenesis, we tested the hypothesis that Dyn, acting via k-opioid 33  Thus, in the TI state, nor-BNI administration into the DMH resulted in an 82.1 ± 19.2% reduction in the cold-evoked inhibition of BAT SNA, indicating that Dyn release in the DMH is necessary for the skin cooling-evoked reduction in thermogenesis in the TI state.
Blockade of central κ-opioid receptors reduces the hypothermic response to ICV administration of an adenosine 1A receptor agonist in free-behaving rats In free-behaving rats exposed to a cool T AMB , central administration of the adenosine 1A receptor (A1A-R) agonist, CHA, produces a progressive hypothermia consistent with an inverted regulation of thermogenesis, characteristic of the TI state 13 . Since Dyn release in the DMH is required for the skin cooling-evoked inhibition of thermogenesis in the TI state in anesthetized rats (Figs. 8D, 8E), we determined if blockade of central κ-opioid receptors with ICV nor-BNI would affect the cooling-evoked hypothermia during the TI state induced by the central administration of CHA in free-behaving rats.
One hour after reducing the T AMB from 25°C to 15°C, free-behaving rats chronically instrumented for T CORE recording, received an ICV pretreatment of either 0.9% saline vehicle (5 µl) or nor-BNI, followed after 10 minutes by an ICV injection of CHA (1 mM, 5µl). During the 1 h exposure to the T AMB of 15°C prior to pretreatment, the rats maintained a normal T CORE of 36.7 ± 0.1°C (n = 3; Fig. 8G), re ecting a normal thermoregulatory cold-defense response. Following saline pretreatment, administration of CHA elicited a prompt reduction in T CORE (Fig. 8G), which reached a minimum of 22.4 ± 0.2°C (ΔT CORE = -14.2 ± 0.2°C from a baseline of 36.7 ± 0.2°C, n = 3) at 7 h:36 min ± 7 min following CHA injection (Fig. 8G). Following pretreatment with nor-BNI, injection of CHA also elicited a rapid reduction in T CORE (Fig. 8G). However, the fall in T CORE (ΔT CORE = -7.2 ± 0.8°C from a baseline of 36.7 ± 0.3°C, n = 3) following CHA administration was signi cantly less after pretreatment with nor-BNI than after saline pretreatment (p < 0.001, Bonferroni post-hoc test). Additionally, because the rate of decline in T CORE was the same in both saline and nor-BNI pretreatment conditions, the minimum T CORE of 29.5 ± 0.8°C after CHA administration was reached at a shorter time (4 h:56 min ± 50 min, p = 0.0448) after the nor-BNI pretreatment than after the saline pretreatment. The nding that during the TI state in free-behaving rats 10 pretreatment with nor-BNI reduced the maximum hypothermia but not the rate of decline in T CORE suggests that Dyn, acting via central κ-opioid receptors, plays a permissive role in sustaining the skin cooling-induced inhibition of thermogenesis that is a hallmark of the TI state 13 .

Discussion
Torpor/hibernation, naturally expressed by only a few species, is a complex behavioral phenotype 7 with a unique metabolic state in which T CORE and whole-body energy expenditure are markedly reduced through an inhibition of the normal, cold-defensive responses (BAT and shivering thermogenesis) that are essential for maintaining T CORE in a cold environment 34 . We have discovered that a torpor-like state, featuring a shifted homeostasis closely resembling the physiological alterations observed in natural torpor, can be induced in rats 10, 13 , a species that does not naturally express torpor/hibernation in a cold environment. Further, the hypothermia and hypometabolism of this torpor-like state are due to a switch (Extended Fig. 3) within the CNS circuitry regulating body temperature, such that in the TI state stimulation of cold skin thermoreceptors produces an inhibition of thermogenesis 10,13 . The present study reveals components of the principal afferent and efferent neural mechanisms underlying the TI state. We demonstrate that the well-known central thermoregulatory components, the PBN and the DMH 1-4, 6, 22 , mediate an inverted skin thermoreceptor regulation of thermogenesis during the TI state, which is independent from the integration of POA neuronal function required during normal thermoregulation. Our study also led to the discovery of a novel thermoregulatory function for a direct dynorphinergic pathway between the dlPBN and the DMH (Figs. 7, 10), playing an essential role in mediating the cold-evoked inhibition of thermogenesis that is a prominent characteristic of both the TI state and the torpor/hibernation state.
Neuraxis transection rostral to the DMH is su cient to eliminate the normal, POA-dependent, skin thermoreceptor-mediated regulation of thermogenesis and to establish the TI state, in which thermogenesis is still controlled by skin thermoreceptors, but now responds to skin warming and cooling in an inverted 13 way. This and our current nding that muscimol-induced inhibition of POA neurons also establishes the TI state are consistent with a model in which the switch from the normal to the TI thermoregulatory state requires a change in the activity of a population of POA neurons that signi cantly alters an input to the DMH. Recent studies to identify 'torpor' neurons in mice 8, 24,25 have also concluded that neurons in the POA are essential for establishing the hypothermic torpor state, including those Q neurons that project to the DMH 8 and estrogen-sensitive neurons that project to medial hypothalamic areas 25 . However, none of these studies has proposed a neural circuit through which such 'torpor neurons' would regulate thermogenesis to induce hypothermia.
Paralleling the thermore ex circuit controlling BAT thermogenesis, cold-induced shivering is dependent on the activation of shivering thermogenesis-promoting neurons in the DMH that project to shivering premotor neurons in the rRPa 22 . Remarkably, the switch to the TI state of thermoregulation inverts the skin thermoreceptor-mediated regulation of shivering (Fig. 2), just as it does for BAT thermogenesis (Fig. 3). The warming-induced activation of BAT in the TI state requires an ionotropic glutamate receptormediated excitation of DMH neurons (Fig. 3), just as the cold-induced activation of BAT does in the normal thermoregulatory state 3 . Together, these ndings strongly support our conclusions that a modulation of inputs to neurons in the DMH is necessary to induce the TI state, and that the principal thermoregulatory mechanism underlying the cooling-induced hypothermia in the TI state is a widespread inhibition of thermogenesis at the level of the thermogenesis-promoting neurons in the DMH (Extended Data Fig. 2, 3).
Our demonstration that blockade of ionotropic glutamate receptors in the PBN prevents the warminduced activation of BAT thermogenesis characteristic of the TI state indicates that ascending warm and cold thermoregulatory signaling, mediated by glutamatergic inputs to PBN neurons, is essential for the skin thermoreceptor-mediated control of thermogenesis in the TI state, as it is for normal thermoregulation 1 in both the rat 3, 4 and mouse 14 . Our discovery that the TI state occurs when the normal thermoregulatory connections between the POA and the DMH are severed (Fig. 3) or inhibited (Fig. 1), indicates the existence of POA-independent thermoregulatory pathways through which thermosensory signaling transmitted via PBN neurons can in uence the activity of thermogenesis-promoting neurons in the DMH (Extended Data Fig. 3). Further, the inversion of the effects of cold and warm skin thermoreceptors on BAT and shivering thermogenesis in the TI state must arise from a "neuronal switch" that alters the balance between the skin thermoreceptor signaling that ascends from the PBN to the POA and the thermoreceptor signaling that short-circuits the POA to directly in uence thermogenesispromoting neurons in the DMH (Extended Data Fig. 3).
We identi ed connections between neurons in the dlPBN and elPBN regions of the PBN and neurons in the region of the DMH that contains thermogenesis-promoting neurons that project to the rRPa.
Furthermore, we determined, in free-behaving and anesthetized naïve rats, that many of these DMHprojecting PBN neurons are active during normal thermoregulatory responses to skin cooling and skin warming. These ndings support the existence of novel thermoregulatory inputs from the PBN to the thermogenesis-promoting neurons in the DMH that could act in concert with those from the POA to contribute to the normal thermoregulatory control of thermogenesis. More speci cally, we identi ed a population of DMH-projecting Dyn neurons in the dlPBN region that is activated during skin warming in naïve rats, when inhibitory in uences on the discharge of thermogenesis-promoting neurons in DMH predominate. Dyn activated k-opioid receptor, has a potent inhibitory in uence on thermogenesispromoting neurons in the DMH, leading to a nearly complete reversal of the normal cold-evoked activation of BAT SNA and BAT thermogenesis (Fig. 8). These observations are consistent with a signi cant, but previously undescribed, thermogenesis-inhibiting role for these DMH-projecting Dyn neurons in the dlPBN during normal thermoregulation.
More Dyn neurons in the dlPBN were activated during skin cooling in the TI state, when thermogenesis is inhibited, than during skin cooling in naïve rats, when thermogenesis is active. In addition, blockade of the κ-opioid receptors for Dyn in the DMH markedly reduced the cold-evoked inhibition of BAT SNA in the TI state in anesthetized rats, and signi cantly reduced the maximum hypothermia induced in our torpormimicking CHA model of TI in awake rats 13  Together, these results provide strong support for our discovery of a novel thermoregulatory role for the dynorphinergic pathway from the dlPBN to the DMH in the inhibitory regulation of thermogenesis, both in the normal thermoregulatory state when skin warming inhibits thermogenesis, as well as in the TI state when skin cooling inhibits thermogenesis.
The fact that the neural circuitry underlying the TI state is present and functional in the rat, a species that does not express natural torpor, suggests that this neuronal substrate could be accessed for human clinical applications involving the management of hypothermia following ischemic incidents or for the reduction of metabolic demands during extended space ights. Further, the existence of an alternative thermoregulatory circuit mediating the TI state regulation of thermogenesis and energy expenditure could provide insight into the neural mechanisms underlying conditions in which thermoregulation (postsurgical shivering, anaphylactic or septic hypothermia) or energy metabolism (obesity) is regulated in a seemingly unphysiological manner.

Animals
Male Sprague Dawley rats (300-400 g, Charles River Laboratories) were maintained in a standard 12 hr/12 hr, light/dark cycle (lights on at 0900) with ad libitum access to standard chow and water.
Experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals, 8th Edition (National Research Council, National Academies Press, 2010) and protocols were approved by the Institutional Animal Care and Use Committee of Oregon Health and Science University.
Procedures for recording BAT sympathetic nerve activity (SNA) or muscle shivering EMG Rats were anesthetized initially with 3% iso urane in 100% O 2 and transitioned to urethane (0.8 g/kg) and chloralose (80mg/kg) following cannulation of a femoral artery and vein. Heart rate (HR) was derived from the femoral arterial pressure (AP) signal. Rats were positioned in a stereotaxic frame with the incisor bar at -4 mm below interaural zero and a spinal clamp installed on the T10 vertebra used to maintain the spine in a rigid and elevated position (detailed in 13  A similar surgical preparation was used for experiments in which a shivering EMG was recorded with a bipolar electrode inserted into a nuchal (neck) muscle. Nuchal EMG (nEMG) recordings were performed in rats anesthetized initially with 3% iso urane in 100% O2, and subsequently transitioned to a continuous intravenous infusion of inactin (85 mg/ml at 0.2 ml/h). Rats were arti cially ventilated but were not paralyzed.

Pre-DMH transection (pre-DMH transX)
A cranial window (~ 4x4 mm) was made just behind the bregma and centered on the sagittal suture. The dura mater was carefully dissected from the superior sagittal sinus and removed throughout the cranial window to allow transection of the brain without damage to the sinus or the major vessels converging on it. A transection knife (15 mm long, 2 mm wide, and 0.1 mm thick) was mounted vertically in a stereotaxic manipulator and positioned perpendicular to the sagittal sinus at -1.5 mm caudal to bregma and with the medial edge on the midline. After a slight lateral retraction of the sagittal sinus, the knife was inserted into the brain sequentially on the left and right sides of the superior sagittal sinus to a depth of either − 9 mm (partial transection) or approximately − 10 mm (complete transection).

Drug nanoinjection procedures
Intraparenchymal brain nanoinjections of drugs were performed as previously described 10, 16, 35 via glass micropipettes, using a pressure injection system (Toohey model IIe). For repeated nanoinjections at the same site, the micropipette was retracted vertically, emptied, rinsed with saline, re lled, and repositioned at the original dorsoventral coordinate. The injection sites were marked with uorescent polystyrene microspheres (1:10 dilution of FluoroSpheres F8797, F8801, or F8803, Invitrogen).
With the incisor bar positioned at -4 mm, bilateral nanoinjections (120 nl each) in the PBN were performed at -8.9 mm caudal to bregma, 2.2 mm from the midline, and 5.5 mm below the brain surface; and in the DMH at -3.2 mm caudal to bregma, 0.4 mm from the midline, and 7.8 mm below the brain surface.
Bilateral nanoinjections (180 nl each) were performed bilaterally in the medial preoptic area (MPA; -0.4 mm caudal to bregma, ± 0.4 mm from the midline, and − 7.5 mm below the brain surface), and in the median preoptic area (MnPO; at bregma on the midline, -6.5 mm below the brain surface).
Following experimental procedures, rats were perfused transcardially with isotonic saline, followed by 4% paraformaldehyde (PFA) in 10 mM sodium phosphate buffered saline (PBS; pH 7.4). The brains were removed, post xed in 4% PFA (2 h), equilibrated overnight in 30% sucrose, and sectioned (60 µm coronal sections) to localize the uorescent spots indicating the centers of the injection sites. The coordinates used for the brain intraparenchymal injections were adapted from a rat brain atlas 36 and from our previous studies involving these brain regions 1, 13, 16 .

Drugs
The

Experiments in free-behaving rats
Central administration of the A1AR agonist, CHA, produces the TI state, featuring a dramatic fall in T CORE due to an inhibition of thermogenesis in a cold T AMB 10 . Rats were anesthetized with 2% iso urane in 100% O 2 and instrumented for chronic recording of physiological variables as previously described 10 .
Rats were implanted with an intraperitoneal implantable temperature probe (Anipill®) for recording of T CORE . A guide cannula (C315G-26GA, PlasticsOne) was stereotaxically positioned in the lateral ventricle for intracerebroventricular (ICV) injection of drugs. The cannula was secured to the skull with screws and dental acrylic. Following the surgical procedure, rats were treated with buprenorphine (0.1 mg/kg), penicillin G (40,000 units/kg) and hydrated with isotonic saline (5 ml, subcutaneous). Each rat recovered for 7 days in a temperature-controlled recording chamber at an ambient temperature (T AMB ) of 25°C and received daily meloxicam (1 mg/kg orally) for the rst 3 days to reduce post-surgical in ammation. For ICV injections of CHA and nor-BNI, 5 µl of drug solution were injected over 2 minutes through an internal cannula connected to a 25 µl Hamilton syringe. Rats were brie y removed from the recording chamber, the ICV injection procedure was performed in less than 5 minutes, and the rats were immediately returned to the recording chamber for continued data acquisition.

Neuroanatomy
For anatomical tracing experiments, adult male Sprague Dawley rats (240-400 g) were anesthetized with 2-3% iso urane in 100% O2, and stereotaxically injected with cholera toxin subunit b (CTb) conjugated with Alexa-488 (1 mg/ml, 120 nl) into the right DMH (bregma: 3.2 mm caudal, 0.4 mm lateral, 7.5 mm ventral to the brain surface; incisor bar at -4 mm), and FluoroGold (FG, 2%, 30 nl) into the rRPa (relative to lambda: 3.0 mm caudal, 0.0 mm lateral, 9.2 mm ventral to the brain surface; incisor bar at -4 mm). Rats were pretreated with intramuscular injections of an antibiotic (40,000 units/kg penicillin G) and an analgesic (1 mg/kg meloxicam), and subcutaneous injection of isotonic saline (3 ml). One week after tracer injections, free-behaving rats were exposed to a cold ambient (T AMB : 10°C) or to a warm ambient ( We employed retrograde transport, combined with Fos expression, to identify DMH-projecting neurons in PBN that are activated following induction of TI in anesthetized rats. Seven days prior to the terminal experiment, rats were injected with non-conjugated CTb into the right DMH as detailed above, with some rats also receiving an injection of FG into the rRPa to identify the DMH region containing thermogenesispromoting neurons 5,35 . Following a 7-day recovery period for retrograde transport of CTb, rats were anesthetized with 2-3% iso urane, placed in the stereotaxic frame, and prepared for pre-DMH transX surgery. Four groups were studied: (a) cold naïve (Cold-N) rats were maintained with a cold skin (T SKIN < 35°C) for 1 h of baseline recordings with a stable T SKIN and T CORE , after which they received a sham brain transection surgery (transection knife lowered into the cortex) and were then maintained with a cold skin To characterize DMH-projecting (CTb-labeled) PBN neurons that express c-Fos in response to cold or warm exposure in the 4 groups of iso urane-anesthetized rats, we used the RNA-Protein Co-detection ancillary kit (Advanced Cell Diagnostics). c-fos and pDyn mRNA transcripts were detected by ISH using the RNAScope procedure, whereas CTb was labeled with immuno uorescence. Sections (20 µm) containing the PBN were rst incubated in the primary goat anti-CTb antibody (1:500) overnight at 4°C, then washed in PBST and subjected to the RNAscope Multiplex Fluorescent v2 assay following manufacturer instructions (with minor modi cations). mRNA transcripts were labeled with Opal uorophores (Opal-570, red) and (Opal-690, far-red assigned blue color). After RNAScope procedure, slides were washed in PBST and incubated for 2 h in the species-speci c secondary antibody for CTb (Alexa-488 donkey anti-goat, 1:200). Slides were washed in PBST and coverslipped with anti-fade mounting medium. CTb-labeled neurons displayed cytoplasmic green uorescence, whereas c-fos and pDyn mRNA transcripts were labeled as red and far-red (blue) punctate, respectively. In some cases, the colors were reversed for c-fos and pDyn transcripts.
All slides processed for IHC and ISH were visualized at an Olympus BX-51 uorescence microscope, and images were captured using Simple PCI software (C-Imaging Systems). Brightness and contrast were adjusted using Adobe Photoshop.
The CTb antigenicity is severely impaired by the protease step of the ISH procedure, causing a substantial decrease in CTb labeling. In addition, the pDyn and c-fos transcript signals are very strong; therefore, the number of triple-labeled neurons (CTb-pDyn-c-fos) was signi cantly underestimated with this codetection procedure.
To further visualize Dyn-containing neurons in the PBN, a group of rats (n = 3) were injected, after 7 days recovery from previous injection of CTb in DMH, by an ICV injection of colchicine, which disrupts axonal transport and concentrates the neuropeptide in the soma. Rats were injected, under general anesthesia, with 10 µl of colchicine (25 mM in 10% DMSO saline) ICV. Rats were allowed to survive for 24-36 h after colchicine injection and were perfused with PFA. Brains were removed and processed for IHC as

Data and statistical analysis
For analysis of the physiological variables, the data were averaged into 30s bins, and group data were reported as mean ± standard error of the mean (SEM). To account for slight differences in BAT SNA and nEMG recording characteristics (e.g., tissue-electrode contact, ampli er noise, etc. Neurons expressing Fos, FG, and FGFos were counted in one DMH hemi-section, located contralateral to the CTb injection in DMH, at the level where the main FG-ir cluster is located (i.e., neurons projecting to the rRPa). The counts of double-labeled (FGFos) neurons were normalized to the number of FG-labeled neurons (% FGFos/FG) to counteract variability among FG injections.
Data are reported as mean ± SEM. All statistics were performed using Prism software (version 6, GraphPad Software Inc.). Paired one-or two-tailed t-tests, one-way ANOVA followed by t-test, and a twoway ANOVA followed by post-hoc Bonferroni correction were used for statistical comparisons, as described in gure legends. Statistical results with p < 0.05 were considered signi cant.      Effect of warm and cold exposure on the activation of DMH-projecting neurons in the PBNfollowing pre-DMH TransX.
A. An example of the distribution of neurons double-labeled for warm-evoked Fos (red nucleus) and for CTb retrogradely transported from DMH (green cytoplasm) in the PBN of anesthetized naïve rats (Warm-N) and anesthetized pre-DMH TransX rats (Warm-T). The dotted yellow circles in dlPBN and elPBN represent the counting boxes used for the quantitative analysis reported in panel C.  VGluT2-expressing Dynorphinergic neurons in PBN project to DMH and are activated during TI.
A. Dynorphinergic neurons (containing pDyn transcripts; red) were observed in several PBN subdivisions along its rostro-caudal extent, but the strongest labeling was in dense clusters located in the dlPBN at the intermediate-1 and -2 levels. All Dyn neurons in these clusters express VGluT2 transcripts (blue), but none Activation of κ-opioid receptors is necessary for the cold-evoked inhibition of BAT SNA in the TI and torpor-like state.