The α²-Adrenoceptor Agonist Dexmedetomidine Converges on an Endogenous Sleep-promoting Pathway to Exert Its Sedative Effects

This experimental protocol was approved by the Home Office of the United Kingdom (London, UK), the Ethics Review Committee of Imperial College of Science, Technology and Medicine (London, UK), and the Institutional Animal Care and Use Committees of Beth Israel Deaconess Medical Center and Harvard Medical School (Boston, Massachusetts), and all animal procedures were carried out in accordance with the United Kingdom Animals (Scientific Procedures) Act of 1986. Adult male Fischer rats (weight, 250–300 g; n = 66) and adult male C57B6 mice (weight, 25–30 g; n = 24) purchased from B&K Universal (Grimston Aldbrough Hull, UK) and adult male D79N transgenic mice (weight, 25–30 g; n = 24; C57B6 genetic background; dysfunctional α^2A-adrenoceptor gene, kindly donated by Professor Lee Limbird, Ph.D., Department of Pharmacology, Vanderbilt University, Nashville, Tennessee) were given access to food and water ad libitum  and housed under controlled conditions (12 h of light starting at 7:00 pm; 20–22°C) in an isolated ventilated chamber. All efforts were made to minimize animal suffering and reduce the number of animals used.

Gabazine hydrobromide (SR-95531, Tocris, Bristol, UK), dexmedetomidine (donated by Orion Pharmaceuticals, Espoo, Finland), and chloral hydrate (Sigma, Poole, UK) were prepared in normal saline (NS; 0.9% NaCl w/v) solution. Halothane (Fluothane, Zeneca Ltd., Cheshire, UK) was administered via  an anesthesia machine. Atipamezole hydrochloride (5 mg/ml; Antisedan®, Pfizer, Sandwich, UK) was used as premixed solution. Pentobarbital (J.M. Loveridge Plc., BM-20111, Southampton, UK) was used as 20% solution (w/v) in NS.

Animals were anesthetized with halothane in a Plexiglas chamber (Röhm GmbH, Darmstadt, Germany), prepared for aseptic surgery, and secured in a stereotaxis frame. One hole for intracerebroventricular injection or two for TMN injections were drilled into the skull and filled with removable bone wax (Ethicon, Brussels, Belgium). Cannulae (22 gauge; 20 mm for TMN, 15 mm for intracerebroventricular; Tomlinson Tubes Ltd., Bidford-on-Avon, UK) were stereotaxically positioned (intracerebroventricular coordinates:± 1.0 mm mediolateral, −5.2 anteroposterior, −9.1 dorsoventral from Bregma; TMN coordinates:−1.0 mm mediolateral, −1.0 mm anteroposterior, −4.0 mm dorsoventral), 24and affixed with dental resin (Orthoresin, Dentsply, Surrey, UK).

Drugs were discretely administered into the TMN or intracerebroventricular space at least 4 days postsurgery. The α²-adrenoceptor agonist dexmedetomidine (2–6 μg/ 10 μl) was microinjected into the intracerebroventricular space, or the GABA^Areceptor antagonist gabazine (0.2 μg/0.2 μl per side) was microinjected directly into the TMN using a Howard CMA microinjection pump (CMA/100, CMA/Microdialysis, Stockholm, Sweden) at rates of 5 μl/min and 0.4 μl/min, respectively. The needle was removed 2 min after the completion of the injection. The locations of the injections were later confirmed histologically.

Sedation was assessed using a loss of righting reflex (LORR) endpoint defined as the inability of animals to right themselves when positioned in a supine position in a warm environment. Dose–response data were fitted according to the method of Waud 25to a logistic equation of the form

where p is the percent of the population anesthetized, D is the drug dose, n is the slope parameter, and ED⁵⁰is the drug dose for a half-maximal effect.

Rats were fitted with electroencephalogram and electromyogram electrodes according to methodology previously described. 26Briefly, they were anesthetized with intraperitoneal chloral hydrate, 350 mg/kg, prepared for aseptic surgery, and secured into a stereotaxis frame. Four electroencephalogram electrodes were screwed into the skull, two electromyogram electrodes were placed under the nuchal muscles, and all leads were connected into a pedestal socket and affixed to the skull with dental cement. Animals were allowed to recover for at least 6 days.

Sleep was scored and recorded according to methodology previously described. 26Animals were acclimated to the recording apparatus 24–48 h before experiments began. Electroencephalogram and electromyogram signals were amplified by a polygraph (GRASS, West Warwick, RI) and digitized by the ICELUS program (G-Systems, Plano, TX) for LabVIEW (National Instruments, Austin, TX). An investigator blinded to treatment manually scored sleep–wake states in 12-s epochs. Wakefulness was identified by desynchronized electroencephalogram and high electromyogram activity, NREM sleep by high-amplitude slow-wave electroencephalogram and low electromyogram, and REM sleep by a regular θ component in the electroencephalogram power spectrum coupled with very low electromyogram activity.

Animals were anesthetized with intraperitoneal chloral hydrate, 350 mg/kg, or intraperitoneal pentobarbital, 140 mg/kg, and transcardially perfused with NS (100 ml) or phosphate buffered saline (PBS; 100 ml; 0.1 m phosphate buffer, 0.9% NaCl; pH 7.4) followed by 10% formalin (500 ml; Sigma-Aldrich, Poole, UK) or 4% paraformaldehyde (500 ml; BDH, Poole, UK) in 0.1 m phosphate buffer. Whole brains were removed, postfixed in 10% formalin or 4% paraformaldehyde overnight, incubated in 20% sucrose overnight (or until tissue sank), and coronally cryosectioned (1:4 series, 30 μm).

Sections were double-immunostained using previously described methods. 14,17,27All sections were stained for c-Fos (goat polyclonal antibody; 1:20,000; Santa Cruz, Insight Biotechnology, Wembley, UK; or 1:150,000; rabbit; Oncogene, CN Biosciences, Nottingham, UK) using secondary donkey antigoat IgG (1:200; Chemicon, Harrow, UK), VectaStain® Elite ABC solution (Vector, Peterborough, UK) visualized using 3,3-diaminobenzidine (DAB; Vector) with nickel ammonium sulfate for black staining. Next, they were separated into those containing the LC, VLPO, and TMN and then counterstained for dopamine β-hydroxylase (DBH; rabbit polyclonal antibody; 1:20,000; Affiniti Research Products, Exeter, UK), galanin (rabbit polyclonal antibody; 1:50,000; Bachem, St. Helen's, UK), or adenosine deaminase (ADA; rabbit polyclonal antibody; 1:20,000, Chemicon), respectively, using donkey antirabbit IgG (1:200; Chemicon, Harrow, UK) and visualized using DAB without nickel. Galanin staining was enhanced using tyramide amplification (TSA Biotin System, PerkinElmer, Hounslow, UK) according to manufacturer's protocol.

According to previously described methodology, 28to further verify the anatomic location of the VLPO in some instances, VLPO sections were DAB-stained brown for c-Fos protein using immunohistochemistry (see previous Immunohistochemistry methodology section) and subsequently counterstained for galanin mRNA using in situ  hybridization. In brief, sections were acetylated, hybridized overnight using a ³⁵S-labeled cRNA probe synthesized from a plasmid containing the galanin coding sequence, 29and washed (1 h each in 2× sodium chloride citrate [SSC]/1 mm dithiothreitol (DTT), 50°C; 0.2× SSC/1 mm DTT, 55°C; then 0.2× SCC/1 mm DTT, 60°C). Sections were then treated with RNAase-A (Boehringer-Mannheim, Indianapolis, IN), washed in conditions of increasing stringency (including 30 min in 60°C 0.1× SCC), dehydrated in graded alcohols, and air-dried. Sections were exposed to x-ray film (2 or 3 days; Eastman-Kodak, Rochester, NY), dipped in NTB-2 emulsion (Kodak), exposed for 1 month, developed in D-19 (Kodak), fixed, dehydrated, and cover-slipped.

Using light microscopy, c-Fos positive neurons were identified by dense black nuclear staining, and the DBH-, galanin-, and ADA-positive neurons in the LC, VLPO, and TMN were identified by brown cytoplasmic staining. Locations in the brain were confirmed by staining and reference to the primary literature 14,17,26,30,31and a rat brain atlas. 24Four sections through the middle of each structure per animal were counted (blind to treatment) and averaged. Data were analyzed using one-way analysis of variance (ANOVA) and the Bonferroni test and are presented as mean ± SEM. Differences were considered significant at P < 0.05.

Using previously established methodology, 26discrete bilateral VLPO lesions were administered. Briefly, animals were secured in a stereotaxis, and the skull was exposed. After a small hole was drilled in the skull over the VLPO, the dura was incised, a fine glass micropipette (20-μm tip diameter) was stereotaxically lowered into the VLPO (coordinates:−0.6 mm anteroposterior, −8.0 mm dorsoventral, and ± 1.0 mm mediolateral from Bregma), and 15 nl of ibotenic acid (10 nmol) was microinjected by air pressure. The area of neuronal loss induced by the lesion was assessed histologically by staining for c-Fos immunoreactivity (as described by Sherin et al.  14and in previous Immunohistochemistry methodology section) followed by nonspecifically counterstaining with 0.25% thionin (Nissl stain), as previously described. 26In the region of the lesion, Nissl-stained cells that remained in the VLPO could be easily identified. The absence of c-Fos expression in the VLPO area after dexmedetomidine administration further confirmed lesion success. Lesions were quantified using cell-counting methodology described by Lu et al. , 26which used a counting box placed over the VLPO (framed by the lateral edge of the optic chiasm and extending 300 μm dorsally and 300 μm laterally), and lesions were considered successful if there was > 80% bilateral cell loss. 26Animals with lesions that missed the VLPO were used as sham-operated controls.

To assess the effects of VLPO lesions in dexmedetomidine-induced sedation, animals with bilateral VLPO lesions were administered intraperitoneal control NS or dexmedetomidine, 50 μg/kg (a dose sufficient to induce LORR for more than 2 h in naïve animals), at approximately 6:45 pm and were connected to the electroencephalogram and electromyogram recording apparatus for 2 h. After 2 h of electroencephalogram and electromyogram recording (approximately 7:00–9:00 pm, a period of maximal wakefulness in naïve animals), animals were transcardially perfused, and brains were sectioned and immunostained (see previous Immunohistochemistry methodology section). The percentages of wakefulness, rapid eye movement (REM) sleep and NREM sleep per hour were plotted from 7:00 to 9:00 pm postlesion in the absence and presence of dexmedetomidine or NS injection and compared with similar recordings made in unlesioned animals. Data were analyzed using one-way ANOVA and the Bonferroni test to determine significant differences in the paired data.

Using behavioral measures, we tested the hypothesis that the mechanism of dexmedetomidine-induced sedation (loss of consciousness) converges on the endogenous NREM sleep-promoting pathway by challenging the sedation induced by centrally administered dexmedetomidine with a systemically administered GABA antagonist. Subcutaneous administration of the GABA^Asubtype selective antagonist gabazine, 5 mg/kg, induced a rightward shift in the dose required to cause a LORR, an endpoint commonly used as a surrogate measure of loss of consciousness (fig. 2A). Dexmedetomidine alone had an ED⁵⁰of 0.32 ± 0.11 μg (mean ± SEM), whereas dexmedetomidine in the presence of gabazine pretreatment exhibited a significantly different (P < 0.05) ED⁵⁰of 0.44 ± 0.01 μg. When dexmedetomidine and gabazine were administered systemically, the presence of gabazine significantly (P < 0.05) attenuated the sedative response to dexmedetomidine. Subcutaneous gabazine, 5 mg/kg, significantly (P < 0.05) increased the latency to (fig. 2B) and decreased the duration of LORR (fig. 2C) induced by subcutaneous dexmedetomidine, 150 μg/kg (both measures indicate reduced anesthetic potency).

In the LC, where discretely administered dexmedetomidine can induce sedation, we found that local administration of the GABA^Aantagonist gabazine had no effect on dexmedetomidine-induced sedation (n = 6), whereas systemically administered gabazine significantly (P < 0.05) shifted the dexmedetomidine dose–LORR response curve to the right (fig. 2A). However, gabazine administered more distally into the arousal-promoting TMN (n = 3) attenuated the sedation induced by dexmedetomidine (P < 0.05).

We next used immunohistochemical techniques to assess whether dexmedetomidine acts on a NREM sleep-promoting pathway (fig. 1). Activation of a cellular immediate early gene c-fos , and subsequent accumulation of expressed c-Fos protein, is often used as evidence of neuronal activity. 32,33Although the link between c-Fos expression and neuronal activation is indirect, c-Fos levels have been shown to change in different parts of the brain during spontaneous sleep–wake episodes, 14,30,31,34and these changes are taken as surrogate markers for neuronal activity. Dexmedetomidine-induced sedation resulted in a pattern of c-Fos expression in the LC, VLPO, and TMN that was qualitatively and quantitatively similar to that seen previously during endogenous normal NREM sleep, 30,31 i.e. , decreased c-Fos expression in the LC, increased in the VLPO, and decreased in the TMN (fig. 3A). We further examined the effect of dexmedetomidine on c-Fos expression after administration of the α²-adrenoceptor antagonist atipamezole or the GABA^Areceptor antagonist gabazine. Fifteen-minute pretreatment with intraperitoneal atipamezole, 2 mg/kg (a dose that blocks dexmedetomidine-induced sedation), inhibited the dexmedetomidine-induced changes in c-Fos expression in the LC, VLPO, and TMN described previously. Pretreatment with intraperitoneal gabazine, 5 mg/kg, blocked dexmedetomidine-induced c-Fos changes in the TMN but not in the LC or VLPO (fig. 3A).

We examined dexmedetomidine-induced c-Fos expression in transgenic mice lacking functional α^2A-adrenoceptors to assess the mechanisms of the dexmedetomidine-induced sedation. Gene targeting was used to introduce a point mutation into the α^2A-adrenoceptor subtype in the C57B6 strain mouse genome, rendering it dysfunctional in the eponymous D79N mouse. The α^2Asubtype is known to be responsible for the sedative component of dexmedetomidine's anesthetic action, 7,35whereas α^2Band α^2Csubtypes are uninvolved in sedation but may play a role in analgesia. 36In D79N animals, dexmedetomidine cannot induce LORR at any dose. In C57B6 mice, 2 h of dexmedetomidine-induced sedation at the beginning of the dark cycle (rodent waking cycle, 7:00–9:00 pm) resulted in a c-Fos pattern that was qualitatively similar to that seen in rats (see the preceding Results section), i.e. , reduced c-Fos expression in the LC and TMN and elevated c-Fos in the VLPO relative to NS-administered (awake) animals. However, none of these changes were observed in D79N mice administered the same dose of dexmedetomidine (400 μg/kg); in this case, the pattern of c-Fos expression did not differ from that induced by NS (fig. 3B).

To confirm that the changes in c-Fos expression observed in the endogenous NREM sleep pathway (in the LC, VLPO, and TMN) and during dexmedetomidine-induced sedation are causally related to the sedative state and to assess the sequential relationship of the changes in neuronal activity in these nuclei, we tested dexmedetomidine's sedative efficacy after discretely destroying the VLPO using behavioral (sedation measured by electroencephalogram) and immunohistochemical indices. Bilateral VLPO lesions were administered by microinjection of the nonspecific excitotoxin ibotenic acid, which does not destroy fibers of passage.

Bilateral VLPO lesions reduced the percentage of time that animals spent in an electroencephalogram-assessed slow wave sleep state during the 2-h period (7:00–9:00 pm, the beginning of the dark cycle, a period of maximal wakefulness) after dexmedetomidine administration (fig. 4). Further, we found that VLPO lesions blocked dexmedetomidine-induced decreases in c-Fos expression at the level of the TMN but not at the LC. The level of NREM sleep in lesioned and unlesioned animals was consistent with the findings of Lu et al. , 26who reported that VLPO lesions decrease natural sleep in rats during the same 2-h period of the day, further indicating that the lesions performed in this study successfully destroyed VLPO neurons.

The aim of this study was to determine whether the mechanism of sedation induced by the α²-adrenoceptor agonist dexmedetomidine converges on an endogenous NREM sleep-promoting pathway. We found that pretreatment with the GABA^Areceptor agonist gabazine decreases the sedative potency of dexmedetomidine, suggesting that GABA^Areceptors may be involved downstream of the LC, the key site for initiation of dexmedetomidine sedation. We further observed that the c-Fos changes induced by 90 min of dexmedetomidine sedation were qualitatively similar to those induced during NREM sleep (a decrease at the LC, an increase at the VLPO, and a decrease at the TMN). It is likely that these changes are being initiated at α²-adrenoceptors because they can be prevented by the α²-adrenoceptor antagonist atipamezole at a dose sufficient to block dexmedetomidine-induced sedation. Further, dexmedetomidine did not induce the same changes in neuronal activity in mice genetically modified to have mutated α^2A-adrenoceptor subtype receptors (in which dexmedetomidine cannot induce sedation).

In questioning whether the activation of the VLPO during dexmedetomidine-induced sedation has a causal role, we found that slow wave electroencephalogram activity after dexmedetomidine administration (indicative of NREM sleep) was significantly attenuated by discrete bilateral VLPO lesions. This finding led us to conclude that the activation of GABAergic and galaninergic neurons in the VLPO is necessary for the sedative response. The patterns of c-Fos expression induced by dexmedetomidine in VLPO-lesioned animals demonstrate the relationship of one nucleus to another. The upstream dexmedetomidine-induced decrease in c-Fos expression in the LC was unchanged in VLPO-lesioned animals, but the downstream dexmedetomidine-induced decrease in c-Fos expression in the TMN was not observed in the absence of a functional VLPO. We further elucidated the hierarchical nature of these changes by demonstrating that discrete administration of gabazine into the TMN blocks dexmedetomidine-induced changes in the TMN but not at the level of the LC or VLPO.

It is known that discrete VLPO lesions prevent normal sleep behavior and render animals insomniac. 26Recent advances 15,17,29,37,38suggest that the VLPO is a site that uses GABA to inhibit arousal pathways. VLPO neurons are sleep-active (show c-Fos expression during sleep) and project to all of the ascending monoaminergic, cholinergic, and peptidergic arousal-promoting sites in the brain: the histaminergic TMN, noradrenergic LC, serotonergic dorsal raphe (DR), cholinergic pedunculopontine tegmental nucleus (PPTg), cholinergic laterodorsal tegmental nucleus (LDTg), and the orexinergic perifornical area (PeF). 17The VLPO neurons are identified by three distinct characteristics. First, they contain the colocalized inhibitory neurotransmitters GABA and galanin in rats (nearly 100% colocalization, whereas many surrounding neurons contain GABA only). 17Second, projections densely innervating the TMN originate from a dense cluster of VLPO neurons just lateral to the optic chiasm and a diffuse population of cells extending dorsally and medially from this cluster. Third, VLPO neurons are uniquely sleep-active (i.e. , express c-Fos during sleep and show increased firing rates during sleep). 14,38 

The inhibition of the TMN by GABA and galanin, believed to be released by VLPO neurons, 20,37is hypothesized to play a key role in causing sleep. GABA and galanin are observed in the TMN, 39–42where they have an inhibitory effect on TMN release of histamine in the cortex and subcortical regions, and are implicated in loss of wakefulness. Yang and Hatton 37recorded from the TMN while stimulating the VLPO region and found GABA^A-mediated IPSPs in the TMN. GABA^Areceptor agonist injections here cause sleepiness, and GABA^Areceptor antagonists cause wakefulness. 23,43 

Several caveats arising from our experiments are worth discussing. First, we have used rat and mouse models in this study, which raises issues of translatability between species. Transgenic experiments were introduced only as corroborating data and were conducted in mice rather than rats because rats with dysfunctional α^2A-adrenoceptor genes (mutations) are not yet available. Mice and rats react similarly to dexmedetomidine, although mice require a higher minimum dose to induce sedation, and it is believed that both species 29use similar sleep-promoting neuronal circuitry. Second, we have used two endpoints as measures of anesthetic-induced sedation, namely, behavioral LORR and telemetric electroencephalogram and electromyogram assessments. LORR assesses immobility (a feature of loss of consciousness) and equates to loss of consciousness in humans, whereas electroencephalogram and electromyogram measure slow wave electroencephalogram activity, which accompanies sedation. Third, although we used a small cohort size in the VLPO lesion experiments, our results were robust enough to show statistical significance (P < 0.05). Finally, we focused this study on the TMN as representative of arousal-promoting nuclei innervated by the VLPO but do not suggest that the TMN is the only wake promoter whose activity is modulated by dexmedetomidine. Rather, we suggest that further experiments are likely to reveal that other loci in the ascending arousal system are also indirectly inhibited by dexmedetomidine.

This is a novel demonstration of a neuronal pathway causally connected with the hypnotic response to a general anesthetic. The potential implications of using general anesthetics (or hypnotic drugs) that act via  similar mechanisms as natural sleep to induce loss of consciousness are profound. A hypnotic agent that could produce the same reparative changes as natural sleep (i.e. , hormone and immune function changes) may speed recovery time in an intensive care setting and counteract the effects of sleep deprivation, a common problem in intensive care units and for surgical patients during recovery. We suggest that α²-adrenoceptor agonist sedatives activate sleep pathways that work on nuclei upstream of the VLPO at the level of the LC, and thus may produce more restorative sleep than that induced by downstream modulation of the same pathway by GABAergic agents.

The authors thank Professor Lee Limbird, Ph.D., Department of Pharmacology, Vanderbilt University, Nashville, Tennessee, for generously donating D79N transgenic mice; Orion Pharmaceuticals, Espoo, Finland, for donating dexmedetomidine; and Quan Ha and Dr. Mann Xu, Laboratory Technicians, Department of Neurology, Harvard Medical School, Boston, Massachusetts, for excellent technical support.