In 1998, orexin A and B were identified from rat brain extracts as ligands of an orphan GPCR, HFGAN72 (orexin receptor-1; OX₁R) (Sakurai et al., 1998). Orexins constitute a novel peptide family with no significant structural similarities to known families of regulatory peptides. Orexin A is a 33-amino-acid peptide with two sets of intrachain disulfide bonds. It has an N-terminal pyroglutamyl residue and C-terminal amidation (Sakurai et al., 1998). The primary structure of orexin A predicted from the cDNA sequences is completely conserved among several mammalian species (human, rat, mouse, cow, sheep, dog, and pig). On the other hand, rat orexin B is a 28-amino-acid, C-terminally amidated linear peptide, which has 46% (13/28) sequence identity to orexin A. Orexin B also has a high degree of sequence similarity among species, although substantial species variants were found (Sakurai et al., 1998; Shibahara et al., 1999; Alvarez and Sutcliffe, 2002) (Figure 1A) (Shibahara et al., 1999; Alvarez and Sutcliffe, 2002; Sakurai, 2005). Orexin A and B are produced from a common single precursor polypeptide, prepro-orexin, through usual proteolytic processing presumably by prohormone convertases (Figure 1B). An mRNA encoding the same precursor peptide was independently isolated by de Lecea et al. as a hypothalamus-specific transcript. They predicted that the precursor encoded two neuropeptides, hypocretin-1 and -2 (de Lecea et al., 1998).

HFGAN72, now called OX₁R, was initially identified as an expressed sequence tag (EST) from human brain (Sakurai et al., 1998). Subsequently, orexin receptor-2 (OX₂R) was identified by searching EST databases (Sakurai et al., 1998). Both receptor genes are highly conserved among species (Sakurai et al., 1998). OX₁R has one order of magnitude greater affinity for orexin A than for orexin B. In contrast, OX₂R has similar affinity for both orexin A and orexin B (Sakurai et al., 1998) (Figure 1B). OX₁R is coupled to the G_q/11 class of G proteins, which results in activation of phospholipase C (PLC) with subsequent triggering of the phosphatidylinositol cascade. OX₁R also stimulates cAMP synthesis in primary rat astrocyte culture (Woldan-Tambor et al., 2011). OX₂R is shown to be coupled to both G_q/11 and inhibitory G_i proteins when expressed in cell lines (Zhu et al., 2003) (Figure 1B). Food deprivation was reported to exert a differential effect on coupling between orexin receptors and G proteins (Karteris et al., 2005).

Orexin-producing neurons (orexin neurons) are exclusively localized to the perifornical area and lateral and posterior hypothalamic area in the rat brain (Peyron et al., 1998; Date et al., 1999; Nambu et al., 1999) (Figure 2A). This distribution has been confirmed in human (Elias et al., 1998). The number of orexin neurons is assumed to be around 3000 in the rat brain, and 70,000 in the human brain (Peyron et al., 1998; Thannickal et al., 2000), and these cells diffusely project to the entire neuroaxis (Peyron et al., 1998; Date et al., 1999; Nambu et al., 1999) (Figure 2B). This anatomical structure suggests that the activity of orexin neurons influences multiple brain areas. The strongest staining of orexin-immunoreactive nerve endings in the brain was found in the tuberomammillary nucleus (TMN), paraventricular thalamic nucleus (PVT), arcuate nucleus (Arc) of the hypothalamus, and most notably, monoaminergic nuclei in the brain stem, such as the locus ceruleus (LC) which receives the densest oreinergic fibers in the brain stem, and raphe nuclei.

Orexin colocalizes with dynorphin (Chou et al., 2001), galanin (Hakansson et al., 1999), prolactin (Risold et al., 1999), neuronal activity-regulated pentraxin (NARP) (Reti et al., 2002), glutamate (Abrahamson et al., 2001), and neurotensin (unpublished data). Many orexin neurons are glutamatergic (Rosin et al., 2003; Torrealba et al., 2003) but are not GABAergic (Rosin et al., 2003). A recent study using an optogenetic technique confirmed that orexin neurons released glutamate on TMN histaminergic neurons (Schone et al., 2012).

In the central nervous system, both orexin receptor mRNAs are expressed in regions that receive dense orexin innervations as described above. OX₁R and OX₂R mRNAs show partial overlap, but largely distinct and complementary distribution patterns, suggesting that each receptor subtype plays different physiological roles. OX₁R is expressed in many brain regions such as the prefrontal and infralimbic cortex (IL), hippocampus (CA2), amygdala, bed nucleus of the stria terminalis (BST), PVT, anterior hypothalamus, dorsal raphe (DR), ventral tegmental area (VTA), LC, and laterodorsal tegmental nucleus (LDT)/pedunculopontine nucleus (PPT) (Trivedi et al., 1998; Lu et al., 2000; Marcus et al., 2001). OX₂R is expressed in the amygdala, TMN, Arc, dorsomedial hypothalamic nucleus (DMH), paraventricular nucleus (PVN), LHA, BST, PVT, DR, VTA, LDT/PPT, CA3 in the hippocampus, and medial septal nucleus (Lu et al., 2000; Marcus et al., 2001). Double in situ hybridization studies have revealed the expression patterns of OX₂R and OX₁R mRNA more precisely. In the TMN, all vesicular monoamine transporter 2 (VMAT2)-positive histaminergic neurons expressed OX₂R mRNA, while OX₁R mRNA was not detected. In the DR and median raphe nuclei (MnR), the majority of VMAT2-positive serotonergic neurons expressed both OX₁R and OX₂R mRNA. In the LC, all VMAT2-positive noradrenergic neurons exhibited OX₁R, whereas OX₂R mRNA was exclusively detected in VMAT2-negative non-noradrenergic neurons. In the LDT and PPT, all vesicular acetylcholine transporter (VAChT)-positive cholinergic neurons expressed OX₁R but not OX₂R mRNA, but many OX₁R-positive and/or OX₂R-positive noncholinergic neurons were intermingled with cholinergic neurons (Mieda et al., 2011). These histological findings suggest that orexins and their receptors are likely to play a broad regulatory role in the monoaminergic/cholinergic systems.

In mice with a genetically encoded retrograde tracer and in rats with combined antero- and retrograde tracers, upstream neuronal populations that make innervations to orexin neurons were revealed (Sakurai et al., 2005; Yoshida et al., 2006). These studies showed that orexin neurons are innervated by the lateral parabrachial nucleus (LPB), ventrolateral preoptic nucleus (VLPO), medial and lateral preoptic areas, basal forebrain (BF), posterior/dorsomedial hypothalamus, VTA, DR nucleus, and MnR. Many upstream neurons were identified in regions associated with emotion including the IL, amygdala, shell region of the nucleus accumbens (NAc), lateral septum (LS) and BST. Orexin neurons were also shown to receive innervations from regions associated with energy homeostasis including NPY-, agouti-related peptide (AgRP)-, and α-melanin-stimulating hormone-immunoreactive fibers, which presumably originate in the arcuate nucleus (Broberger et al., 1998; Elias et al., 1998).

A number of factors that influence firing rate or membrane potential of orexin neurons have been identified (Table 1).

Importantly, both noradrenaline and serotonin (5HT), both of which are wake-active substances, hyperpolarize and inhibit activity of orexin neurons through activation of G protein-regulated inwardly rectifying K⁺ channels via α₂-adrenoceptors and 5HT_1A-receptors, respectively (Li et al., 2002; Muraki et al., 2004; Yamanaka et al., 2003b, 2006).

In addition, the cholinergic agonist carbachol activates some orexin neurons through M3 muscarinic receptors (Yamanaka et al., 2003b; Sakurai et al., 2005). However, histamine appeared to have no effect on orexin neurons. These observations suggest that serotonin and noradrenaline neurons might send inhibitory feedback projections to orexin neurons. These feedback mechanisms might stabilize the activity of both orexin neurons and monoaminergic neurons. Furthermore, although orexin neurons do not express functional dopamine receptors, dopamine can inhibit orexin neurons by acting on α₂-adrenoceptors (Yamanaka et al., 2003b, 2006). It was also shown that agonists of ionotropic glutamate receptors excite orexin neurons, whereas glutamate antagonists (AP-5, CNQX, or NBQX) reduce their activity (Li et al., 2002; Yamanaka et al., 2003b). These results indicate that orexin neurons are tonically excited by glutamatergic neurons. At the same time, GABAergic and glycinergic input to orexin neurons strongly inhibits the activity of orexin neurons (Xie et al., 2006; Matsuki et al., 2009; Hondo et al., 2011; Karnani et al., 2011b).

In addition, a sulfated octapeptide form of cholecystokinin (CCK-8S), neurotensin, oxytocin, vasopressin, orexin, and a mixture of amino acids activate orexin neurons (Tsujino et al., 2005; Tsunematsu et al., 2008; Yamanaka et al., 2010; Karnani et al., 2011a), whereas glucose and leptin inhibit them (Table 1). These factors are implicated in energy homeostasis. It was also shown that adenosine inhibits orexin neurons via the A₁ receptor. This pathway might be related to the sleep-promoting effect of adenosine (Liu and Gao, 2007). In addition, orexin neurons are affected by physiological fluctuations in acid and CO₂ level (Williams et al., 2007). Since orexin affects respiratory function, this mechanism might play a role in the regulation of respiratory function (Nakamura et al., 2007; Sunanaga et al., 2009). A recent report suggested that acid-sensing ion channels 1a (ASIC1a) located on orexin neurons contribute to the regulation of breathing by sensing local acidity (Song et al., 2012). These results suggest that orexin neurons are specialized sensors of the internal environment.

Sleep-active neurons in the POA, especially the ventrolateral preoptic area (VLPO), appear to play a critical role in the initiation of NREM sleep and maintenance of both NREM and REM sleep (Sherin et al., 1998). Neurons in the VLPO fire at a rapid rate during sleep, with attenuation of firing during the waking period (Figures 3A,B). These neurons mostly contain GABA and/or galanin, and send descending inhibitory projections to wake-active neurons that produce wake-promoting neurotransmitters, including histamine, noradrenaline, 5-HT, and acetylcholine (Sherin et al., 1998; Lu et al., 2002) (Figures 3A,B). On the contrary, these sleep-promoting neurons are inhibited by wake-active transmitters such as noradrenaline, acetylcholine, and 5-HT (Gallopin et al., 2000). These reciprocal interactions of inhibition constitute the flip-flop switching of wake/sleep states (Saper et al., 2001). Importantly, GABAergic neurons in the POA also densely innervate orexin neurons (Sakurai et al., 2005; Yoshida et al., 2006). This pathway might be important to turn-off orexin neurons during sleep (Figure 3B). In fact, orexin neurons are strongly inhibited by both GABA_A and GABA_B receptor agonists (Yamanaka et al., 2003a; Xie et al., 2006). Moreover, selective deletion of the GABA_B receptor gene in orexin neurons resulted in highly unstable sleep/wake architecture in mice (Matsuki et al., 2009).

As already mentioned, narcolepsy patients and animals with defects of orexin or OX2R cannot maintain a consolidated wakefulness state. It is recognized that monoaminergic neurons in the hypothalamus and brain stem, including neurons in the TMN, LC, and DR, play crucial roles in maintaining arousal (Saper et al., 2005). Firing rates of these neurons are known to be synchronized and strongly associated with sleep/wake states. They fire tonically during the awake state, less during NREM sleep, and are virtually quiescent during REM sleep (Vanni-Mercier et al., 1984). Orexin neurons were also reported to discharge during active wakefulness and cease firing during both NREM and REM sleep (Lee et al., 2005), or displayed transient discharge in REM, in vivo (Gerashchenko and Shiromani, 2004; Blanco-Centurion et al., 2007). Furthermore, many studies also suggested that noradrenergic cells of the LC (Hagan et al., 1999; Bourgin et al., 2000), dopaminergic cells of the VTA (Nakamura et al., 2000), serotonergic cells of the DR (Brown et al., 2002; Liu et al., 2002), histaminergic cells of the TMN (Yamanaka et al., 2002), and cholinergic neurons in the BF (Eggermann et al., 2001) are all excited by orexins in vitro (Alam et al., 1999). These results suggest that orexin neurons fire during the wakeful period, and excite these wake-active neurons to sustain their activity.

On the other hand, as described previously, orexin neurons are regulated by acetylcholine and monoamine. Constantly, orexin neurons are innervated by BF-cholinergic neurons (Sakurai et al., 2005). A cholinergic agonist, carbachol, activates some populations of orexin neurons. Furthermore, serotonergic and noradrenergic neurons send inhibitory projections to orexin neurons (Muraki et al., 2004; Sakurai et al., 2005; Yamanaka et al., 2006). These findings indicate that orexin neuron maintain an awake state by activation of wake-active neurons. On the other hand, during a sleep state, sleep-active neurons inhibit both orexin neuron and wake-active neurons to maintain a sleep state. If orexin neurons were deleted as in narcolepsy, sleep-active neurons and wake-active neurons would make a mutually inhibitory system. This might result in behavioral instability, which is a major symptom of narcolepsy (Figure 3C).

Additionally, optogenetic activation of orexin neurons expressing channelrhodopsin-2 increased the probability of transition to wakefulness from sleep (Adamantidis et al., 2007). This effect was observed throughout the entire light/dark period, but was diminished with sleep pressure (Carter et al., 2009). Carter et al. suggested that the orexin system promotes wakefulness throughout the light/dark period, but the downstream targets are inhibited with increased sleep pressure overriding the effect of orexin. Optogenetic inhibition of orexin neurons expressing halorhodopsin resulted in induction of slow wave sleep during the light period, although it had no effect during the dark period (Tsunematsu et al., 2011). Pharmacogenetic modulation of orexin neurons also revealed that excitation of orexin neurons increased wakefulness time and inhibition of orexin neurons decreased wakefulness time and increased NREM sleep (Sasaki et al., 2011). These selective modulations of orexin neurons in vivo suggest that activity of orexin neurons actually influences an animal's vigilance states. Furthermore, optogenetic modulation of orexin neurons and LC neurons revealed that the wake-promoting influence of orexin is mediated by LC neurons (Carter et al., 2012).

Intracerebroventricular (icv) injection of orexin during the light period potently increases the awake period in rats, and this effect is markedly attenuated by a histamine H₁ antagonist (Yamanaka et al., 2002). The pharmacological effect of orexin A on waking time in mice is almost completely absent in H₁-receptor-deficient mice (Huang et al., 2001). Moreover, focal restoration of OX₂R in neurons of the TMN and adjacent parts of the posterior hypothalamus in mice lacking OX₂R completely rescued the sleepiness of these mice (Mochizuki et al., 2011). These results suggest that the TMN-histaminergic pathway might be an important effector site of orexin for sleep/wake regulation.

Consistently, OX₂R knockout mice show characteristics of narcolepsy (Willie et al., 2003), while OX₁R knockout mice exhibit only very mild fragmentation of the sleep-wake cycle (Willie et al., 2001). However, notably, the phenotype of OX₂R knockout mice is far less severe than that found in prepro-orexin knockout mice and double receptor knockout mice. Especially, OX₂R knockout mice are only mildly affected by cataplexy and direct transitions to REM sleep from an awake state. Furthermore, the effects of orexin A on wakefulness and NREM sleep were significantly attenuated in OX₂R knock-out mice as compared with wild-type mice, but OX₁R knockout mice also showed an impaired response (Mieda et al., 2011). These observations suggest that although the OX₂R-mediated pathway has a pivotal role in the promotion of wakefulness, OX₁R also has additional effects on sleep-wake regulation. These findings suggest that loss of signaling through both OX₁R- and OX₂R-dependent pathways is necessary for emergence of a complete narcoleptic phenotype.

Cataplexy has been proposed to be controlled by mechanisms similar to those producing atonia during REM sleep. Recent reports suggested that the REM-related atonia is induced by activation of REM-on neurons in the sublaterodorsal tegmental nucleus (SLD) (Fuller et al., 2007; Luppi et al., 2011). One of the activation pathways of REM-on neurons is likely to be mediated by input from the amygdala, which is well-known to be activated during REM sleep (Luppi et al., 2011). Strong, generally positive emotional stimuli, which might activate the amygdala, are known to trigger cataplexy in narcolepsy-cataplexy patients. These suggest that the amygdala-induced REM-on neuron activation might play a role in cataplexy. During wakefulness and NREM sleep, the SLD REM-on neurons are inhibited by REM-off neurons in the ventrolateral periaquaductal gray and dorsal deep mesencephalic reticular nucleus (vlPAG/dDPMe) (Luppi et al., 2011). These findings suggest a possibility that excitatory input from orexin neurons to vlPAG/dDPMe supports the activity of these REM-off neurons to avoid muscle atonia. This hypothesis explains why narcoleptic patients exhibit cataplexy during emotional events. Moreover, neural input from the limbic system, especially the amygdala, to orexin neurons might also be implicated in the pathophysiology of cataplexy. Some report also suggested that cholinergic neurons in the LDT/PPT are implicated in REM-related atonia (Shiromani et al., 1988), and the same pathway is implicated in cataplexy. A local injection of orexin into the PPT strongly inhibited REM-related atonia in cats (Takakusaki et al., 2005). It is possible to speculate that emotional stimuli may increase orexin release in the PPT, which indirectly inhibits cholinergic neurons, to prevent muscle atonia in wild-type animals. Another report also showed that orexin activates lateral vestibular nucleus neurons and promotes vestibular-mediated motor behavior. This report suggests that orexin is critical when an animal is facing a major motor challenge. The orexin system participates not only in sleep and emotion but also in motor regulation directly. These findings may also account for the mechanism of cataplexy (Zhang et al., 2011).

It is known that food perception often evokes cataplexy in narcoleptic dogs and orexin knockout mice (Reid et al., 1998; Clark et al., 2009), suggesting that orexin signaling is physiologically activated upon perception of food, and that this system is necessary to evoke proper feeding behavior. This suggests that regulation of feeding behavior might also be controlled by input to orexin neurons from the limbic system, because some of the affective content of the perception of food is thought to be processed in the amygdala and limbic system (Berthoud, 2004), and this information may be passed on to orexin neurons.

The LHA has been implicated in the reward system by both lesion experiments and the intracranical self-stimulation paradigm (Anand and Brobeck, 1951; Olds and Milner, 1954). Recently, several lines of evidence have suggested that orexins are involved in modulation of the reward system.

Narcoleptic patients are often treated with highly addictive amphetamine-like drugs (Nishino and Mignot, 1997) but they rarely become addicted to these drugs (Akimoto et al., 1960; Guilleminault et al., 1974). Consistently, orexin knockout mice are less susceptible than wild-type mice to developing morphine dependence as measured by physical withdrawal response (Georgescu et al., 2003; Narita et al., 2006). Moreover, abnormal activity in reward brain circuits was observed in human narcolepsy (Ponz et al., 2010).

These observations suggest that orexin neurons play important roles in reward processing. In fact, VTA dopaminergic neurons receive input from orexin neurons composed of both synaptic terminals (Peyron et al., 1998; Marcus et al., 2001; Fadel and Deutch, 2002) and en-passant fibers (Balcita-Pedicino and Sesack, 2007), and orexin directly activates VTA dopaminergic neurons (Nakamura et al., 2000; Korotkova et al., 2003). Icv orexin-induced hyperlocomotion and stereotypy were blocked by a dopamine receptor antagonist (Nakamura et al., 2000). Moreover, icv or local VTA infusion of orexin was shown to reinstate drug-seeking or food-seeking behavior in rodents (Boutrel et al., 2005; Harris et al., 2005).

On the contrary, orexin neurons receive projections from the VTA, NAc, and LS—regions involved in reward systems (Yoshida et al., 2006). Consistently, dopamine has an inhibitory influence on food intake and reward pathways when injected into the LHA/PFA (Yang et al., 1997). These reciprocal interactions might constitute regulatory mechanisms of reward systems (Figure 4).

Furthermore, many reports suggest a critical role of orexin signaling in neural plastic effects at glutamatergic synapses in the VTA. In vivo administration of an OX₁R antagonist, SB334867, inhibited cocaine- or high fat-induced potentiation of excitatory currents in VTA dopaminergic neurons (Borgland et al., 2006, 2009). Orexin A input to the VTA potentiates N-methyl-D-aspartate receptor (NMDAR)-mediated neurotransmission via PLC/PKC-dependent recruitment of NMDA receptors in dopamine neuron synapses (Borgland et al., 2006). Moreover, orexin B increased presynaptic glutamate release in addition to postsynaptic potentiation of NMDA receptors in the VTA (Borgland et al., 2008). These findings suggest roles of orexin in the mechanisms of reward systems and drug addiction (Figure 4).

Behavioral studies have suggested that orexin neurons play an important role in motivation, feeding, and a variety of reward-seeking behaviors. For example, orexins are involved in self-administration behavior. Under fixed ratio schedules, SB334867 reduced self-administration of nicotine, alcohol, high-fat pellets and sucrose in food-restricted rats, and heroin (Hollander et al., 2008; Nair et al., 2008; Cason et al., 2010; Jupp et al., 2011; Smith and Aston-Jones, 2012). Under a progressive schedule that requires the animal to lever-press progressively more times to obtain a reinforcer, SB-334867 also reduced the motivation to self-administer sucrose in food-sated, but not food-restricted rats, and cocaine (Borgland et al., 2009; Espana et al., 2010). These results show that OX₁R signaling is important for motivation for a variety of highly salient, positive reinforcement. Orexin is also involved in reward-based feeding. Furthermore, orexin is also involved in cue-induced reinstatement of seeking for rewards. SB-334867 blocked foot shock-induced reinstatement of cocaine-seeking behavior (Boutrel et al., 2005; Smith and Aston-Jones, 2012) and an olfactory cue-induced reinstatement of alcohol-seeking behavior (Jupp et al., 2011). These findings suggest that orexin serves to motivate the animal to engage in goal-directed behavior interacting with reward systems.