The neurobiological basis of narcolepsy

Abstract

Narcolepsy is the most common neurological cause of chronic sleepiness. The discovery about 20 years ago that narcolepsy is caused by selective loss of the neurons producing orexins (also known as hypocretins) sparked great advances in the field. Here, we review the current understanding of how orexin neurons regulate sleep–wake behaviour and the consequences of the loss of orexin neurons. We also summarize the developing evidence that narcolepsy is an autoimmune disorder that may be caused by a T cell-mediated attack on the orexin neurons and explain how these new perspectives can inform better therapeutic approaches.

Introduction

Narcolepsy affects about 1 in 2,000 people in the United States and Europe^1–3. Every individual with narcolepsy shows some degree of excessive daytime sleepiness, with a tendency to doze off when sedentary in school, at work or even when driving. Most people with narcolepsy feel rested when they wake in the morning or after a nap, but within a few hours, they feel as sleepy as a healthy person would feel if awake for the entire night.

In addition to excessive daytime sleepiness, most people with narcolepsy also have symptoms indicative of abnormal rapid eye movement (REM) sleep. REM sleep is normally characterized by dreaming and muscle paralysis that prevents an individual from acting out their dreams. In narcolepsy, REM sleep can occur at any time of day, and elements of REM sleep can mix into wake, manifesting as hypnopompic and hypnagogic hallucinations or sleep paralysis. A very distinctive symptom of narcolepsy is cataplexy: sudden muscle paralysis triggered by strong, generally positive emotions, such as when laughing or unexpectedly meeting a friend. The muscle weakness of cataplexy usually begins in the face and neck and then sometimes spreads to the trunk and limbs; in severe episodes, an individual may slump to the ground, conscious yet unable to speak or move for a minute or two. Strangely, more than half of people with the classical form of narcolepsy also have the opposite problem: an intermittent failure of REM sleep paralysis, which results in the enactment of dreams (known as REM sleep behaviour disorder)⁴.

Narcolepsy has been studied by clinicians and researchers for almost 150 years, but only in the past 20 years has the underlying cause become clear. In 1998, two research groups independently discovered orexin A and orexin B (also known as hypocretin 1 and hypocretin 2, respectively), small neuropeptides produced solely by neurons in the lateral hypothalamus^5,6. Derived from a precursor protein (prepro-orexin), orexin A and orexin B have excitatory effects on postsynaptic neurons via the orexin 1 receptor (OX1R) and OX2R⁶. Soon after, researchers found that narcolepsy is caused by a highly selective and severe loss of the orexin neurons that results in low levels of orexins in the brain and cerebrospinal fluid (CSF)^7–10. This discovery spurred the recognition of two types of narcolepsy: narcolepsy type 1 (NT1) and narcolepsy type 2 (NT2)¹¹. The classic phenotype, NT1, is characterized by chronic sleepiness plus cataplexy, and CSF orexin levels in this disorder are very low or undetectable, owing to severe loss of the orexin neurons. NT2 has generally less severe symptoms, and 90% of patients have normal CSF orexin levels. NT2 affects up to half of all narcolepsy patients^1–3 and may be caused by a partial loss of the orexin neurons, but little is known about its underlying neuropathology^12,13. In rare cases, narcolepsy results from brain injuries that damage the orexin neurons or their projections, and narcolepsy-like symptoms occur in some families, although the underlying genetics remain unknown^14,15. In NT1, the rate of orexin neuron loss is unknown, but the loss may be rapid in some patients who develop severe narcolepsy symptoms over just a few days and slower in others who develop cataplexy many years after the onset of sleepiness^16,17; some small studies have also shown a decrease in CSF orexin levels over several months from the onset of sleepiness^18,19 (BOX 1).

Box 1 | Natural history of narcolepsy.

Narcolepsy typically begins in the early teen years1¹⁸³. Over the course of a few days or weeks, sleepiness becomes quite problematic, with children falling asleep at school and when doing homework. In addition, many patients develop hypersomnia. For example, a 12-year-old who previously slept about 9 hours each night might now sleep 9–10 hours at night plus another 1–2 hours of napping in the day¹⁸⁴. Over a few years, this hypersomnia resolves, but nocturnal sleep is often fragmented, with numerous spontaneous awakenings and vivid dreams. In children, abrupt weight gain is common, with about half of children gaining 5–15 kg in just a few months, possibly owing to a reduction in basal metabolism. About 1 in 6 children with narcolepsy can develop precocious puberty⁹⁹.

Cataplexy often starts to occur in the first months after the initial increase in sleepiness but sometimes lags behind the sleepiness by years. Cataplexy in children can be quite severe and prolonged, sometimes manifesting as status cataplecticus2¹⁸⁵. Some researchers suggest that moderate loss (~50%) of the orexin neurons causes sleepiness, whereas cataplexy and rapid transitions into rapid eye movement (REM) sleep during the day probably involve severe loss (>80%) of orexin neurons^12,67,176.
Although much remains to be learned about the natural history of narcolepsy, this evolution of symptoms suggests that the orexin neurons may die over a period of weeks to months, and if the cause of orexin neuron loss can be identified, it may be possible to halt this progression.

This clear connection of NT1 to selective loss of the orexin neurons sparked tremendous advances in our understanding of narcolepsy, yet major questions and controversies remain. This article reviews the normal functions of the orexin system and how loss of the orexin neurons results in the symptoms of narcolepsy. We examine the emerging evidence suggesting that an autoimmune process may kill the orexin neurons, and present crucial outstanding questions.

Orexin neurons: anatomy and functions

Anatomy.

The orexin peptides are produced only by a cluster of neurons in the lateral hypothalamus but are released by projections of these neurons to much of the CNS, from the cortex to the spinal cord. The orexin neurons heavily innervate several regions that promote arousal and suppress REM sleep, including the basal forebrain, tuberomammillary nucleus (TMN), periaqueductal grey (PAG), dorsal raphe (DR) and locus coeruleus (LC)²⁰ (FIG. 1). They also innervate regions that regulate reward processing, such as the ventral tegmental area (VTA) and nucleus accumbens, as well as many brain areas implicated in feeding and metabolism (for example, the arcuate nucleus and lateral hypothalamus), and autonomic tone (for example, the paraventricular nucleus, parabrachial nucleus, nucleus of the solitary tract, rostral ventrolateral medulla, rostral ventromedial medulla and intermediolateral cell column of the spinal cord)²⁰. In turn, the orexin neurons receive inputs from brain regions that regulate sleep–wake states (including the DR, LC and basal forebrain), circadian rhythms, reward processing (for example, the VTA), autonomic tone (such as the paraventricular nucleus), fear and anxiety, and regions associated with visceral sensations (including the nucleus of the solitary tract and parabrachial nucleus)^21–24 (FIG. 2). This broad pattern of connections has spurred many researchers to hypothesize that the orexin neurons act as integrators of neuronal signals related to sleep–wake state, motivation and visceral needs such as hunger, to help enhance arousal and autonomic tone, especially when seeking food or other rewards, or responding to threats.

Fig. 1 |. Wake-promoting and cataplexy-suppressing orexin pathways.

a | Orexin neurons (dark blue) maintain wake by exciting various wake-promoting neurons (green), including those in the cortex, basal forebrain (BF), tuberomammillary nucleus (TMN), pedunculopontine and laterodorsal tegmental nuclei (PPT–LDT), dorsal raphe (DR) and locus coeruleus (LC). Orexin neurons may also undergo auto-excitation, helping drive sustained activity in the orexin neurons and their targets. b | Normally, orexin neurons (dark blue) block the occurrence of muscle paralysis during wake by activating rapid eye movement (REM) sleep-suppressing regions (green), such as neurons in the ventral lateral periaqueductal grey and lateral pontine tegmentum (vlPAG–LPT), DR and LC. All these nuclei inhibit the sublaterodorsal nucleus (SLD), a region which, during REM sleep, drives muscle paralysis by activating GABAergic premotor neurons that inhibit motor neurons (purple). With strong emotional stimuli, signals from the medial prefrontal cortex (mPFC) probably activate orexin neurons and neurons in the central nucleus of the amygdala, which have opposing effects on the vlPAG–LPT, DR and LC. However, in narcolepsy, the excitatory drive from the orexin neurons is absent, and signals from the amygdala can inhibit these REM sleep-suppressing regions, enabling activity in the SLD and resulting in cataplexy.

Fig. 2 |. Inputs and outputs of the orexin neurons.

The orexin neurons are influenced by signals related to sleep–wake states, circadian phase, motivational cues and visceral cues such as hunger or thirst, and they innervate many brain regions. Their activity ultimately results in long periods of wakefulness, suppression and regulation of rapid eye movement (REM) sleep, enhanced responses to rewards, increased locomotion and increased autonomic tone. In addition to the orexin neuropeptides, which bind to orexin 1 receptors (OX1Rs; red circles) and OX2Rs (dark blue circles), orexin neurons also release the fast-acting neurotransmitter glutamate and possibly dynorphin. This pattern of connections highlights how the orexin neurons are uniquely positioned to integrate a wide variety of signals to acutely and persistently promote many aspects of arousal. Arc, arcuate nucleus; BF, basal forebrain; DMH, dorsomedial nucleus of the hypothalamus; DR, dorsal raphe; LC, locus coeruleus; LPT, lateral pontine tegmentum; NAc, nucleus accumbens; NST, nucleus of the solitary tract; PB, parabrachial nucleus; POA, preoptic area; PVH, paraventricular nucleus of the hypothalamus; SCN, suprachiasmatic nucleus; TMN, tuberomammillary nucleus; vlPAG, ventrolateral periaqueductal grey; VTA, ventral tegmental area.

Activity and signalling.

Orexin neurons regulate many functions, and chief among these are stabilizing wakefulness and orchestrating REM sleep physiology^25,26. In rodents and dogs, orexin neurons are active during wake, especially during increased locomotor activity or motivated behaviour such as foraging; by contrast, these neurons fire less during quiet wake and very little during non-REM (NREM) or REM sleep^27–30. Photostimulation of the orexin neurons rouses mice from sleep, whereas their photoinhibition or chemoinhibition promotes sleep^31–34. In addition, intracerebroventricular injection of orexin A or an orexin agonist in rodents can promote wakefulness and strongly suppress REM sleep for hours^35,36, most probably by exciting neurons in the basal forebrain, PAG and monoaminergic nuclei that promote wake and suppress REM sleep. With the transition to sleep, GABAergic neurons in the ventrolateral preoptic and median preoptic areas likely inhibit the orexin neurons as well as many other wake-promoting neurons, allowing the animal to maintain sleep^37–39.

How orexins produce such long periods of wake is unclear. Little is known about orexin kinetics in vivo, but orexin A may persist in the extracellular space for long periods, whereas orexin B is likely to have a shorter half-life⁶. Orexin peptides may auto-excite orexin neurons via OX2R (FIG. 1) and, as such, once these cells start firing, they may remain active for long periods, helping sustain wake over long periods^40,41. However, this idea of auto-excitation is controversial, as some have reported that orexin neurons do not express orexin receptors⁴². Quite little is known about the firing patterns of orexin neurons, and more in vivo, long-term recordings are needed^27,28.

Importantly, the conditions that trigger release of orexins are unknown, although high-frequency firing in the orexin neurons probably triggers their release from dense core vesicles^31,43. In addition, as with other neuropeptides that are spontaneously released, some basal level of orexin tone may persist at all times; in support of this idea, orexin receptor antagonists promote sleep in humans even during the night, when orexin neurons might be expected to be inactive^44,45.

The orexin neurons produce other neurotransmitters in addition to the orexin peptides. However, whether the loss of these co-transmitters contributes to the symptoms of narcolepsy is not clear. Glutamate is released by the orexin neurons, and its effects may synergize with the excitatory effects of orexins^43,46,47. Orexin neurons probably also secrete neuronal activity-regulated pentraxin (NARP), a protein that promotes clustering of AMPA receptors at excitatory synapses^48–50. The role of NARP in the orexin system is still undetermined, but it might enhance postsynaptic responses to glutamate. Surprisingly, the orexin neurons may also co-release dynorphin, which inhibits neurons via the κ-opioid receptor^51,52. In mice, κ-opioid receptors are found in most of the same brain regions as the orexin receptors, and how the excitatory effects of orexins are integrated with the inhibitory effects of dynorphin is an active area of research^52,53. Orexin and dynorphin have synergistic effects on some neurons; for example, orexin neurons themselves can be excited by orexin B, and perhaps by orexin A, and local GABAergic inputs to the orexin neurons are inhibited by dynorphin, thus disinhibiting the orexin neurons^40,41,54. The net effect of these co-transmitters on other neurons may depend on the resting membrane potential of the cell or the specifics of its connectivity^53,55. Nevertheless, these co-transmitters are probably important, as mice lacking the orexin neurons exhibit a phenotype that is closer to human NT1 than that of mice simply lacking the orexin peptides (BOX 2). Fully understanding the consequences of orexin neuron loss in NT1 will require more work into the contributions of these co-transmitters.

Box 2 | Comparison of human type 1 narcolepsy and murine and canine models.

Mouse and canine models of narcolepsy capture many of the key features of the disease and enable researchers to study the underlying molecular and circuit-level neurobiology. Poor maintenance of wakefulness (short bouts of wake during the active period) is common in most mouse models^{26,69,71,77,186}. Mice lacking the orexin neurons, such as mice with acute expression of diphtheria toxin in the orexin neurons (orexin-tTa;TetO-DTA mice) or mice with gradual loss of the orexin neurons owing to specific expression of a toxic variant of the ataxin 3 protein (orexin–ataxin 3 mice), generally exhibit more severe symptoms than do mice with constitutive loss of orexins or orexin receptors^13,26,70,187. Mice lacking both orexin receptors have sleepiness similar to mice lacking the orexin peptides^77,188. However, mice lacking only orexin 2 receptor (OX2R) exhibit only mild sleepiness, and mice lacking OX1R are surprisingly normal, suggesting that there are important synergies between OX1R and OX2R signalling^{25,72,186,188}. Cataplexy in mice requires loss of orexins or loss of both orexin receptors^71,77,189. Obesity is more striking in mice with loss of the orexin neurons than in orexin-null mice, possibly owing to concomitant loss of co-expressed signalling molecules, such as glutamate, dynorphin and neuronal activity-regulated pentraxin (NARP)^70,190. The severity of narcolepsy symptoms, including obesity, can also vary with genetic background of the mouse strain^70,190,191.

Dogs with narcolepsy were the first animal models of narcolepsy and provided key insights into the cause of the disorder. Dogs with narcolepsy exhibit severe sleepiness, fragmented sleep, poor rapid eye movement (REM) sleep regulation and cataplexy^192–195. Some of the first links of narcolepsy to orexin signalling came with the observations that inherited canine narcolepsy is caused by loss-of-function mutations in the canine OX2R gene¹⁹⁶ and that sporadic canine narcolepsy results from the loss of production of the orexin peptides¹⁹⁷.

Disease or model	General description	Duration of wake bouts	Cataplexy	Short latency to REM sleep	Body weight	Refs
Human NT1	Loss of orexin neurons	↓↓	↑↑	Yes	↑	198,199
Orex- in-tTA;TetO-DTA mice	Rapid death of orexin neurons with acute expression of DTA upon withdrawal of doxycycline	↓↓	↑↑↑	n/a	↑↑	13,73
Orexin–ataxin 3 mice	Progressive neurotoxic death of orexin neurons starting at birth	↓	↑	Yes	↑	26,70,190
Ox−/− mice	No orexin A or orexin B	↓↓	↑↑	Yes	Small ↑	26,69
Ox1r−/−;Ox2r−/− mice	No functional orexin receptors	↓↓	↑	Yes	n/a	25,77,200,201
Ox1r−/− mice	No functional OX1R	Small ↓	None	n/a	No change	25,202
Ox2r−/− mice	No functional OX2R	↓	Rare	n/a	n/a	72,186
Sporadic canine narcolepsy	Loss of orexin A (and probably of orexin neurons)	↓	↑↑	Yes	n/a	197,203
OX2R−/− dogs	Loss of functional OX2R	↓	↑↑	Yes	n/a	196,197

DTA, diphtheria toxin A; n/a, no published data found; NT1, narcolepsy type 1; OX, prepro-orexin; TetO, tetracycline operator; tTA, tetracycline-controlled transactivator.

Orexin neurons have several other functions in addition to controlling sleep–wake behaviour. For example, through their projections to mesolimbic pathways, orexins promote reward-seeking. Mice that lack orexin neurons or that have been treated with an OX1R antagonist show much weaker conditioned preference for places associated with rewards such as morphine or cocaine^56–59. Orexins also increase home-cage locomotor activity, heart rate, brown fat thermogenesis and metabolic rate in rodents^47,60–66. Overall, orexins may help to generate a high arousal state in which an animal is motivated, attentive and autonomically primed for action.

Effects of orexin neuron loss

Poor maintenance of wakefulness.

Almost all individuals with narcolepsy feel sleepy during the day and easily transition into NREM sleep⁶⁷. Most feel rested on waking in the morning or after a nap, indicating that sleep is restorative, but sleepiness returns in just an hour or two, suggesting dysfunction in the systems that maintain wake. When sedentary or encouraged to sleep, people with narcolepsy can fall asleep very rapidly. On the Multiple Sleep Latency Test (MSLT), patients are instructed to try to sleep every 2 hours across the day. On average, across these five 20-minute nap opportunities, people with narcolepsy fall asleep in less than 8 minutes, and often in just 1–2 minutes, whereas people without narcolepsy usually fall asleep in 10–20 minutes⁶⁸. Similarly, in orexin-null mice, bouts of wakefulness are only half as long as in control mice, owing to frequent and rapid transitions into NREM sleep^69–75.

Recent research in animal models indicates that orexins drive long periods of wake via their projections to important wake-promoting brain regions, such as the TMN, LC and basal forebrain⁷⁶ (FIG. 1). Re-expression of OX2Rs, specifically in the TMN region of mice globally lacking these receptors, completely rescued their ability to produce long wake bouts⁷². Similarly, re-expression of OX1R in the LC markedly improved the maintenance of wake in mice lacking orexin receptors⁷⁷, whereas in TH::IRES-Cre mice, photoinhibition of the LC reduced the awakening effect of orexin neuron photostimulation⁷⁸. Together, these results suggest that orexin signalling through the TMN and LC regions is sufficient to promote sustained periods of wake. However, more work is needed to determine precisely which orexin projections are necessary for normal arousal.

Poor regulation of REM sleep.

Orexin neurons suppress REM sleep, and individuals with narcolepsy exhibit dysregulation of REM sleep that manifests as poor circadian timing of REM sleep, rapid transitions into REM sleep and disruption of REM sleep physiology (for example, REM sleep behaviour disorder or sleep paralysis). Normally, REM sleep occurs only during the typical sleep period, but with loss of orexin signalling, REM sleep can occur at any time of day^26,79. In fact, this pattern is central for diagnosing narcolepsy; on the MSLT, people with narcolepsy typically enter REM sleep in two or more of the five daytime naps, whereas healthy individuals rarely enter REM sleep during the day⁶⁸. Indeed, during the night, REM sleep is usually preceded by at least 60 minutes of NREM sleep, but in NT1, REM sleep often occurs within a few minutes of sleep onset^80,81.

Mechanistically, the propensity of individuals with narcolepsy to enter REM sleep during the day may arise from poor circadian control or disinhibition of REM sleep^26,82,83. REM sleep exhibits a strong circadian rhythm and is normally suppressed during the active period via circadian signals relayed from the suprachiasmatic nucleus to the dorsomedial nucleus of the hypothalamus (DMH)⁸⁴. DMH neurons signal to the orexin neurons, and lack of the orexin neurons in mice reduces the amplitude of the circadian REM sleep rhythm by about half^26,79. These results suggest that the orexin neurons, along with additional projections from the DMH, help to suppress REM sleep during the active period⁸⁵.

In addition to alterations of the timing of REM sleep in narcolepsy, the coherence of REM sleep physiology is also disrupted in narcolepsy, with dream-like hallucinations and episodes of muscle paralysis mixing into wakefulness. Just as in people with narcolepsy, orexin-null mice may suddenly transition from active wake into cataplexy for up to a minute. In individuals with narcolepsy, cataplexy is often triggered by laughter and joking, and in mice lacking orexins, such episodes are much more common with exposure to rewarding stimuli such as chocolate and running wheels^86,87.

This paralysis most probably arises from a dysfunction of brainstem pathways that produce the typical paralysis of REM sleep (FIG. 1). The sub-laterodorsal nucleus (SLD) is a key region for triggering muscle atonia during REM sleep, and the SLD is normally inhibited during wake by GABAergic neurons of the ventrolateral PAG and adjacent lateral pontine tegmentum (vlPAG–LPT)^88–92. The heavy innervation of the vlPAG–LPT by orexin neurons may typically ensure strong suppression of REM sleep during the active period^20,93. The vlPAG–LPT also receives inputs from GABAergic neurons of the central nucleus of the amygdala (CeA)⁹³. Signals related to positive emotions normally engage the medial prefrontal cortex, which excites orexin neurons and the CeA; under normal conditions, the inhibitory effects of the CeA on the vlPAG–LPT are offset by excitatory signals from the orexin neurons. However, in narcolepsy, this orexin tone is absent, and the CeA can inhibit the vlPAG–LPT, thus disinhibiting the SLD and other atonia-promoting brain regions, resulting in cataplexy. In support of this model, chemoactivation of GABAergic neurons in the CeA of orexin-null mice increases cataplexy, whereas chemoinhibition or lesioning of the CeA reduces cataplexy^93–95. Chemogenetic inhibition of the medial prefrontal cortex in orexin-null mice also reduces cataplexy that would normally be triggered by chocolate⁸⁶. Orexins may also suppress cataplexy via projections to the DR neurons, which in turn are well positioned to suppress REM sleep via their projections to the amygdala and other brain regions^77,96.

Overall, many researchers view the abnormal sleep architecture in narcolepsy as ‘behavioural state instability’, with low thresholds for transition between states and poor coherence within states that allows for frequent transitions between states and strange, intermediate states, such as cataplexy and hypnagogic hallucinations^71,75,97,98.

Effects on metabolism and feeding.

In addition to sleepiness and altered REM sleep, people with narcolepsy are prone to gain weight, possibly owing to a lower metabolic rate. Soon after NT1 onset, children can rapidly gain 5–15 kg, and adults with NT1 are often overweight or obese, despite roughly normal caloric intake and activity levels^99–104. Feeding diaries do not indicate increased caloric consumption, but in the laboratory, individuals with NT1 tend to snack more even when satiated and report more binge eating^105,106. Mice lacking orexin neurons have obesity and a reduced metabolic rate, especially during their inactive phase^13,107; however, research on metabolic rate in people with narcolepsy remains inconclusive^104,108. Moreover, there is no consistent evidence for glucose intolerance or insulin resistance in individuals with narcolepsy^108–113. More detailed metabolic studies and studies of eating and exercise habits in people with NT1, especially around NT1 onset, when weight gain is most pronounced, are required to help define the best recommendations for managing the weight of this population.

Effects on reward behaviours.

In mice and rats, considerable research indicates that loss of orexin signalling reduces drug-seeking behaviour for cocaine, amphetamine, nicotine, opiates and ethanol^114–120 (but see REFS^121–123). However, whether the same is true in individuals with NT1 is less clear. People with narcolepsy perform normally on a gambling task and use recreational drugs at typical rates^106,124,125. Any underactivity of reward pathways in NT1 could potentially manifest as depression, and indeed several studies indicate that depression is twice as prevalent among individuals with NT1 (REFS^126–129). Whether this increased rate of depression is a direct consequence of the loss of orexin neurons, or reflective of the many everyday challenges associated with narcolepsy, is unknown (BOX 3).

Box 3 | Is narcolepsy simply due to loss of the orexin neurons?

The loss of orexin signalling seems sufficient to explain the major features of narcolepsy, but some symptoms such as fragmented night-time sleep, rapid eye movement (REM) sleep behaviour disorder and resolution of hypersomnia over time cannot be easily explained by current models.

The loss of the orexin neurons might trigger compensatory responses that are helpful in some ways but harmful in others. For example, two groups have shown substantially increased numbers of histaminergic neurons in the tuberomammillary nucleus of people with narcolepsy type 1 (NT1)^204,205. Histamine is a key wake-promoting transmitter, and in contrast to the other monoamine-producing neurons, histamine neurons remain active during cataplexy²⁰⁶. Thus, increased histamine signalling might help to counter a tendency towards hypersomnia and help to maintain consciousness during cataplexy by preventing full transitions into REM sleep. However, the persistence of histamine signalling at night could also contribute to fragmented sleep. Currently, whether histamine signalling is increased in NT1 is controversial; however, if it is, it may be a useful therapeutic target.

In addition, the neuropathology of NT1 might involve more than just the hypothalamus. Volumetric and white-matter neuroimaging techniques have suggested that grey-matter volume in regions such as in the hypothalamus and inferior temporal and right prefrontal regions may be decreased in individuals with narcolepsy^207,208. Other studies have described reduced grey-matter concentration in the cortex and thalamus in NT1 (REF.²⁰⁹). However, imaging results are inconsistent, possibly owing to differences in technique and patient characteristics²¹⁰. Still, a broader search into additional brain changes would be helpful, as nearly all neuropathology research in the narcolepsy field has focused simply on the orexin neurons and a few other cell types in the hypothalamus.

Evidence for an autoimmune mechanism

Narcolepsy is caused by the selective destruction of the orexin-producing neurons. What kills the orexin-producing neurons remains a major mystery, but several lines of evidence suggest that NT1 is an autoimmune disease mediated by T cells.

T cells can be divided into two categories: CD4⁺ helper T cells secrete cytokines to regulate or assist in the active immune response, and CD8⁺ killer T cells use cytotoxic granules to lyse their target cells. The responses of both types of T cell are triggered through T cell receptors (TCRs) that recognize small, processed peptides presented to them by major histocompatibility complex (MHC) molecules on antigen-presenting cells; peptide fragments bound to MHC molecules are known as MHC–peptide complexes. CD4⁺ T cells recognize antigens bound to MHC class II molecules present on the surface of antigen-presenting cells, including astrocytes and other glial cells^130,131, and CD8⁺ T cells respond to antigens presented by MHC class I molecules, which are expressed on all nucleated cells but very rarely by neurons^132–134. A T cell is initially considered ‘naive’ before it encounters the antigen that binds its specific TCR. Once a T cell encounters its antigen alongside the appropriate co-stimulatory signals, it clonally expands to control the pathogen. Once the pathogen is cleared, the T cell numbers contract, and the remaining pathogen-specific memory T cells can respond more quickly and strongly to the same pathogen in the future.

Narcolepsy after vaccination with Pandemrix.

The most direct evidence that NT1 can be caused by an autoimmune process occurred with the H1N1 influenza pandemic in the winter of 2009–2010. Clinicians noticed a surge in individuals developing NT1 after vaccination with Pandemrix, a brand of flu vaccine mainly used in northern Europe. Pandemrix inoculation was associated with an 8–12-fold increase in new cases of NT1 in children and adolescents and a 3–5-fold increase in adults^135,136. Importantly, all these affected individuals carried the class II human leukocyte antigen (HLA) allele DQB1*06:02 and developed NT1 a few weeks to months after vaccination. Why NT1 was triggered by Pandemrix but not by other flu vaccines remains unclear, but altered viral nuclear proteins in Pandemrix may have contributed¹³⁷. This tragic event suggests that NT1 can arise from an inflammatory trigger in a genetically susceptible population, especially during childhood and adolescence. One group reported a threefold increase in new cases of NT1 in the months after the H1N1 pandemic in China¹³⁸; in this population, NT1 may have been triggered by H1N1 flu infection itself, as flu vaccination rates were very low in China.

Some researchers now hypothesize that NT1 arises from a process of molecular mimicry (FIG. 3). In addition to the increase in NT1 diagnoses after inoculation with Pandemrix, many spontaneous cases of NT1 are preceded by infection with streptococcus or influenza^138,139, and NT1 commonly first develops in the late spring and early summer¹³⁸, suggesting that it may follow immune responses triggered by winter infections. CD4⁺ T cells responding to streptococcus or influenza may cross-react with orexin peptides presented by MHC class II molecules¹⁴⁰.

Fig. 3 |. A model for T cell-mediated killing of the orexin neurons in narcolepsy.

In a possible model of T cell-mediated killing of orexin neurons, an antigen-presenting cell (APC) first takes up a pathogen and presents fragments of pathogen proteins to a naive CD4⁺ T cell using a major histocompatibility complex (MHC) class II molecule, perhaps DQB1*0602. The naive CD4⁺ T cell may secrete cytokines (circles) to help clear the infection. Memory CD4⁺ T cells are formed from the initial infection. Fever may promote the migration of pathogen-specific CD4⁺ T cells across the blood–brain barrier (BBB). The activated memory CD4⁺ T cell cross-recognizes fragments of prepro-orexin with a similar epitope as the pathogen peptide and secretes cytokines that promote destruction of the orexin neurons. Genes for which certain alleles are known to increase the risk of developing narcolepsy related to this model of T cell-mediated killing of orexin neurons are illustrated at their respective sites of action. TCR, T cell receptor.

MHC class II alleles and narcolepsy susceptibility.

HLA alleles strongly influence the development of narcolepsy. More than 90% of individuals with NT1 bear the HLA class II allele DQB1*06:02 (REFS^141–143), and crystalline structure modelling indicated that a fragment of the prepro-orexin protein fits well in the binding groove of DQB1*06:02 (REF.¹⁴⁴). The DQB1*06:02 allele increases the risk of developing NT1 200-fold¹⁴³ — the strongest known association of an HLA with any disease — and NT1 rarely develops in people lacking this allele. People who are homozygous for this allele have twice the risk of developing NT1 as those who are heterozygous^145,146. The HLA allele DQA1*01:02 is in strong linkage disequilibrium with DQB1*06:02 and carries similar risk^143,147. DPB1*05:01 also increases the risk of NT1, although the effect is smaller¹⁴⁸. By contrast, some other class II alleles, including DPB1*04:02, DQB1*06:03 and DQB1*06:09, are protective, as indicated by genome-wide association studies^149–151. Although class II alleles most strongly influence the risk of narcolepsy, two studies have shown small, independent effects of class I alleles^149,150.

Additional genetic factors.

Genome-wide association studies have also suggested that NT1 is associated with polymorphisms in additional genes that influence the response of CD4⁺ T cells to antigens. Several studies have identified a single nucleotide polymorphism in the gene encoding the TCR α-chain that roughly doubles the risk of NT1 (REFS^152–154), suggesting that NT1 might arise from an interaction of DQB1*06:02 and specific TCRs. Further, according to preliminary work, specific, rare single nucleotide polymorphisms within the TCR genes may greatly increase the reactivity of T cells to fragments of the orexin neuropeptides¹⁵⁵. NT1 risk is also associated with variants of the gene coding for cathepsin H, an enzyme involved in digesting proteins into smaller peptides that can be presented by MHC class II molecules¹⁵⁴. Variants of OX40L, which is involved in T cell differentiation, have also been implicated in NT1 risk^154,156; how polymorphisms in OX40L may contribute to NT1 is unknown, but imbalances in T helper cell subsets are commonly observed in autoimmune disorders¹⁵⁷.

Targets of the autoimmune response.

The actual target of this autoimmune process is not clear, but the orexin peptides seem to be a likely target. The orexin-producing neurons are the only cells known to be missing in NT1, and they do not produce any other unique proteins, as demonstrated by histology and gene expression arrays^158–162. Orexins are produced only in the hypothalamus⁶, although unidentified peptides antigenically similar to orexins can be found in the myenteric plexus¹⁶³. In theory, an autoimmune process might kill cells outside the brain, but there is no clinical evidence yet for cell loss in the gut or elsewhere in individuals with narcolepsy. Most importantly, the loss of the orexin neurons in the hypothalamus alone is an adequate explanation for the phenotype of the disorder, as selective killing of these cells in mice recapitulates the major features of NT1 (REFS^13,70).

Other proposed autoimmune targets include the OX2R and the intracellular protein Tribbles homologue 2 (TRIB2)^42,162–164, but OX2R and TRIB2 are expressed widely in the brain, and there is no evidence for loss of cells other than those that make orexins in NT1. Furthermore, whether orexin neurons even express OX2R is a matter of debate^40–42.

Orexin neurons may be killed by T cells.

In addition to the genetic evidence above implicating the MHC class II DQB1*06:02 allele^141–143 and a specific polymorphism in the locus encoding the TCR α-chain, several other lines of evidence suggest that a T cell-mediated immune mechanism destroys the orexin neurons in NT1 and that CD4⁺ cells are crucial. Notably, CD4⁺ cells probably do not directly destroy the orexin neurons, but they could release cytokines that could spur an attack on the orexin neurons by CD8⁺ T cells, macrophages and natural killer cells. Using a transgenic mouse model in which orexin neurons expressed haemagglutinin and T cells possessed a TCR specific for haemagglutinin, one group showed that CD8⁺ T cells are capable of destroying orexin neurons¹⁶⁵. However, a primary CD8⁺ cell attack on the orexin neurons is not likely to occur in NT1, as MHC class I molecules, which are involved in the activation of CD8⁺ T cells, are rarely expressed by neurons in humans, except very early in development and after exposure to interferon-γ^132–134.

The genetic data implicate CD4⁺ T cells in the pathology of narcolepsy, yet direct evidence is just emerging. One recent study used a sensitive method to detect rare T cell populations¹⁶⁶. The researchers polyclonally expanded rare, reactive T cell populations and detected CD4⁺ T cells reacting to epitopes along the entire prepro-orexin peptide. T cell populations reactive to TRIB2 were present in just as many controls as individuals with NT1, but those from NT1 subjects proliferated more. In some people with NT1, CD8⁺ T cells reacting to prepro-orexin were also detected, suggesting that NT1 may arise from an interaction of CD4⁺ and CD8⁺ T cells. This new, sensitive method is considered an important advance, as previous studies had been unable to identify autoreactive T cells, or found them in only a fraction of NT1 subjects^166–168. However, it still remains unclear whether these CD4⁺ T cells reactive to prepro-orexin are the primary cause of NT1, or whether they contribute once another process starts to damage the orexin neurons.

Most research on T cells in NT1 has focused on CD4⁺ or CD8⁺ T cells, but most autoimmune disorders are characterized by deficiencies in immune system regulation by CD4⁺ regulatory T cells (T_reg cells)¹⁵⁷. One report suggested that T_reg cells circulating in peripheral blood were more numerous and highly activated in individuals with NT1 than in control individuals¹⁶⁹. However, the activation of T_reg cells correlated with systemic activation of all T cells, suggesting that this activation was nonspecific. Furthermore, the antigen specificity of T_reg cells was not examined¹⁶⁹; thus, whether these cells could control orexin-specific autoimmune CD4⁺ T cells and CD8⁺ T cells remains to be determined.

Might antibodies contribute to the death of the orexin neurons?

A T cell response seems to be the primary mechanism in NT1, but this may also generate a humoral response, the possible effects of which remain highly debated. In contrast to other autoimmune disorders such as multiple sclerosis, the CSF of patients with NT1 does not show oligoclonal bands or an increase in total proteins compared with healthy controls, and researchers have found no consistent pattern of antibodies targeting the orexin peptides. In a study of 20 DQB1*06:02-positive individuals with narcolepsy who had been inoculated with the Pandemrix vaccine, 17 had antibodies that bound OX2R¹⁶⁴. This report contradicted an earlier study in which only 5% of patients with NT1 had OX2R-targeting antibodies, as compared with 3% of healthy individuals¹⁷⁰. The inconsistency of these results may be attributable to the differences in time from narcolepsy onset when the sera were collected, as OX2R antibodies were more common 500 days after disease onset than at a much later date¹⁷⁰. Further complicating matters, another group found no increase in OX2R-targeting antibodies in Pandemrix-inoculated individuals with NT1, leaving the existence of these antibodies contentious^42,171. Nevertheless, a separate study noted that whereas overall serum levels of immunoglobulin G (IgG) did not differ from controls, individuals with NT1 had higher levels of IgG complexed with orexin A¹⁷², raising the possibility of a specific antibody response to orexin A in NT1 as opposed to a nonspecific increase in systemic IgG. The inconsistencies of these clinical studies highlight the need for careful tracking of time from NT1 onset as well as consideration of HLA genotype when measuring the antibody response.

In addition, individuals with NT1 have also been reported to possess higher levels of antibodies to TRIB2, a protein expressed in orexin neurons and many other cell types¹⁶². Perhaps predictably, these autoantibodies were observed at their highest levels within the first 3 years after NT1 onset and remained slightly elevated compared with controls even 30 years after diagnosis of NT1 (REF.¹⁶²). One group reported that TRIB2-targeting antibodies from patients with NT1 caused orexin neuron loss in mice, but these mice displayed nonspecific immobility in their rest period rather than true cataplexy, and the orexin neuron loss was not replicated in a later study^173,174. Another study found that antisera from some NT1 individuals bound to non-orexin neurons in the hypothalamus, cortex or striatum, but there was no binding to orexin neurons, suggesting that these antibodies may arise only after orexin neurons are injured by another mechanism¹⁷⁵. Currently, whether antibodies contribute to the death of the orexin neurons, or are simply a non-pathogenic consequence of the cell loss, is unknown.

Future directions

In the past several years, much has been learned about the neuropathology, neural circuitry and neuroimmunology of narcolepsy; however, much remains unknown.

One of the largest mysteries is the nature of NT2. Besides a lack of cataplexy, the symptoms of NT2 are similar to those of NT1, yet almost nothing is known about its neuropathology. In NT2, CSF orexin levels are usually normal¹⁷⁶; thus, it may be caused by a modest loss of orexin neurons or a completely different process^177,178.

For patients with NT1, identifying the autoreactive T cells and the antigens that activate them will be required for the development of treatments. The selection of immunomodulatory therapies will also hinge on determining whether NT1 is a monophasic, self-limited disease or a chronic disorder that progresses over months to years. Narcolepsy is usually diagnosed years after disease onset, and for those patients, it may be too late to rescue the orexin neurons. Thus, it will be crucial to develop well-targeted therapies that can normalize activity in the key neural pathways underlying sleepiness, fragmented sleep and cataplexy. Researchers recently mapped the crystal structure of the orexin receptors¹⁷⁹, which facilitates the design of small-molecule agonists. Early preclinical research is quite encouraging, with intraperitoneal administration of one OX2R agonist increasing wake, reducing cataplexy and reducing weight in orexin-null mice³⁶. Furthermore, clinical trials are now underway with orexin agonists and unique medications that enhance signalling by histamine and other monoamine systems^180–182. Addressing the questions above and designing better treatments will be a challenge, but these goals are now achievable and should deliver great scientific and clinical benefits.

Glossary

Hypnopompic and hypnagogic hallucinations

Vivid, sometimes frightening, dream-like hallucinations that occur when falling asleep (hypnopompic) or immediately after waking (hypnagogic)

Sleep paralysis

An inability to move when falling asleep or immediately after waking

Cataplexy

Muscle weakness or full paralysis triggered by strong, generally positive, emotions

REM sleep behaviour disorder

A disorder characterized by impaired motor inhibition during rapid eye movement sleep, resulting in enactment of dreams

Major histocompatibility complex

(MHC). A set of immune molecules that bind antigens derived from pathogens and display them on the surface of antigen-presenting cells to promote acquired immune responses

Human leukocyte antigen (HLA) allele

An allele encoding a human major histocompatibility complex molecule

Molecular mimicry

A mechanism of autoimmunity in which a foreign antigen is structurally similar to ‘self’-peptides, such that immune cells targeting a pathogen accidentally target healthy tissue

Linkage disequilibrium

When the observed frequency of two alleles at two loci occurring together is more frequent than would occur by chance

Myenteric plexus

The network of sensory and motor neurons that control gut secretions and motility

Regulatory T cells

(T_reg cells). T cells that maintain tolerance to self-antigens by downregulating the activity of effector T cells using anti-inflammatory cytokines and cell-to-cell inhibition

Hypersomnia

An abnormally high total amount of sleep over 24 hours

Status cataplecticus

A prolonged period of moderate to severe weakness with low muscle tone, usually without emotional triggers