|Home | About | Journals | Submit | Contact Us | Français|
It has been known for a long time that genetic factors affect sleep quantity and quality. Genetic screens identified several mutations that affect sleep across species, pointing to an evolutionary conserved regulation of sleep. Moreover, it has also been recognized that sleep affects the expression of genes. These findings have given valuable clues about the molecular underpinnings of sleep regulation and function that might lead the way to more efficient treatments for sleep disorders.
Studies with twins in the 1930’s suggested for the first time that sleep is, to some extent, under genetic control. In 1992 PRNP, which codes for the prion protein PRNP, was the first specific gene linked to a human sleep disorder, fatal familial insomnia 1 and was later on shown to affect sleep regulation in mice 2,3. In 1999 the hypocretin/orexin system was implicated in human narcolepsy through genetic studies in dogs 4 and mice 5. Since then, other genes have been linked to human sleep disorders or have been shown to affect sleep in animals across species, suggesting that some of the core mechanisms underlying sleep and its regulation are conserved. Recently, molecular studies have also identified hundreds of brain transcripts across species that change their expression level between sleep and waking, suggesting that the functional consequences of sleep are also shared across species. Thus, specific genes can profoundly affect sleep and, conversely, sleep can influence brain gene expression.
The recent progress in identifying sleep genes and sleep-dependent expression patterns was driven by multiple experimental approaches, from classical mutagenesis screening in flies, quantitative trait loci (QTL) analysis in mice, genome-wide association studies in humans to extensive transcriptomic analysis in flies and rodents (Box 1).
Several strategies can be used. In reverse genetics a candidate gene is mutated first, and then the effects of the mutation on sleep are assessed (from genotype to phenotype). Candidate genes are chosen based on their known function, which makes them likely to be relevant for sleep. In forward genetics the starting point is the phenotype, and a significant part of the work involves going back to the genotype to identify the responsible gene (from phenotype to genotype). This approach is unbiased, and novel genes can be discovered. Forward genetic methods include quantitative trait loci (QTLs) analysis and mutagenesis screening, which can complement each other. Quantitative trait loci (QTLs) are stretches of genome closely linked to the genes that underlie the phenotype under study. QTL analysis starts with the crossing between two inbred mouse strains that differ in the trait under study, and maps a chromosomal region segregating with the phenotype in the progeny. The region may contain either a single “major” gene (that can explain >25% of the genetic variance of a trait) or several genes with small effects. In mutagenesis screenings random small mutations are induced over the entire genome. Insertional mutagenesis uses transposable elements (in flies) to induce mutations, while chemical mutagenesis uses ethylmethane sulfonate (EMS, in flies) or N-ethyl N-nitrosourea (ENU, in mice). In these studies hundreds/thousands of mutated flies or mice are usually screened for the phenotype of interest, and behavioral analysis is preferred because less time consuming and less expensive 120.
This Review will first describe the multiple aspects of sleep that can be grouped in circadian and homeostatic sleep phenotypes, and then discuss some of the genes whose mutations significantly affect sleep in flies, mice, and humans. Overall, these genes can be broadly subdivided into four major functional categories: ion channels, circadian regulation, neurotransmission, and other signaling pathways/hormones. Finally, the effects of sleep on brain gene expression will be reviewed, and the functional categories of waking-related and sleep-related transcripts will be discussed.
Studies in the 1930’s reported that some sleep phenotypes have higher concordance in monozygotic than in dizygotic twins, suggesting some genetic basis (reviewed in 6). A first requirement for studying this phenomenon is to define and characterize the phenotypes of sleep: how much, how well, and when we sleep (Fig 1). It is noteworthy, that sleep phenotypes can change independently of each other. For instance, sleep can be poorly restorative and yet of “normal” duration. Similarly, in some circadian sleep disorders the preferred bedtime occurs too early or too late, but daily sleep amount and sleep quality are preserved.
Sleep is tightly regulated by two sets of mechanisms that work partly independent of each other: the circadian and homeostatic mechanism (Fig 1A). The circadian mechanism reflects how sleep propensity changes during the 24 hours, and its function is to restrict sleep to a time of day that is ecologically appropriate. In mammals, the suprachiasmatic nucleus (SCN) of the hypothalamus is responsible for circadian (circa-diem, approximately 24 hours) rhythms. After SCN lesions sleep is no longer consolidated in one major phase (the day in rodents, the night in flies and humans), but occurs in short episodes throughout the 24-hour period. The homeostatic mechanism, instead, reflects how sleep pressure accumulates during wakefulness and is discharged when we sleep. If waking is prolonged beyond its physiological duration (~ 16 hours in humans) sleepiness increases, performance decreases, and when sleep is finally permitted its duration and/or intensity are greater than in baseline conditions. Sleep homeostasis is controlled centrally by the interaction between sleep-promoting neuronal groups in hypothalamus and basal forebrain, and waking-promoting groups in hypothalamus and brainstem. Sleep need is also locally regulated: brain regions that have undergone synaptic potentiation (e.g. after learning) show larger slow waves, reflecting deeper sleep states 7.
Sleep quantity (duration of total sleep and of its various phases) and quality (latency to sleep, brief arousals after sleep onset, amount of large slow waves) also belong to the sleep phenotypes that have to be accounted for when sleep regulation and function are assessed (Fig 1B).
Sleep quantity and quality in humans is assessed by measuring brain electrical activity with the electroencephalogram (EEG). EEG patterns change predictably depending on behavioral state, from the low voltage fast activity of waking and REM sleep to the slow waves and spindles of NREM sleep (Fig. 1B). More refined information about sleep quality comes from the analysis of the EEG power spectrum, which measures the extent to which specific frequency bands are represented in the EEG signal (Fig. 1B). Interestingly, recent studies found strong heritability of the sleep EEG power spectrum 8,9 (Box 2). Evidence for the heritability of other sleep traits is listed in Supplementary Table 1.
Two recent reports found strong heritability of the sleep EEG power spectrum 8,9, extending previous results in normal sleepers (not twins), which showed that the sleep EEG is very consistent across nights in the same subject, but varies considerably from one individual to another. Other studies found that the trait-like nature of the human sleep EEG is evident at most frequencies below 15–16 Hz121–126. Heritability is probably not as strong for delta activity, which is significantly affected by sleep/waking history. It is worth mentioning that the spectral composition of the human EEG shows striking heritability also during waking, with estimates ranging from 70 to 90% for most frequency bands 127,128. The coherence of the waking EEG is also under strong genetic influence 129,130. In fact, the EEG is among the most hereditable traits in humans, although few specific genes have been identified so far. A linkage disequilibrium between the beta frequency of the human EEG and a GABAA receptor locus has been reported in a rest state with eyes closed, but it is unknown whether this finding is specific for relaxed waking or extends to sleep 131. Thus, the challenge for the future will be not only to identify which specific genes influence the human EEG, but also to determine to which extent their effects are unique to sleep or waking. It is likely that at least some of the interindividual differences in the human EEG are related to genetic factors that are independent of behavioral state.
The most studied is the EEG delta activity, the EEG power in the 0.5–4.5Hz range during NREM sleep, also called slow wave activity (SWA). The higher the SWA, the deeper is NREM sleep, i.e. the more difficult it is to wake up when a stimulus is delivered 10. Moreover, the longer the duration of waking, the higher is the level of SWA at sleep onset (refs in 11). For these reasons delta activity is considered a marker of sleep intensity as well as of sleep need. Sleep in rodents, birds, and other animals is also customarily quantified using EEG analysis, while behavioral criteria are used as indicators of sleep in fish and invertebrates. Indeed, it is worth pointing out that in almost all animals sleep is characterized by quiescence (marine mammals are an exception) and increased arousal threshold. The latter is the single most important feature that distinguishes sleep from “rest” or quiet waking, suggesting that the reduced ability to respond to the environment is a requirement for the restorative function of sleep to occur, and can be captured without the EEG 12. Thus, the EEG is crucial for a detailed analysis of NREM and REM sleep in many vertebrates, but our understanding of sleep regulation and functions can equally benefit from behavioral analysis, as done in several newly introduced invertebrate animal models (Table 1).
Studies across in flies, mice and humans have identified genes that affect sleep. These genes can be grouped into four major functional categories (Supplementary Table 2), which will be discussed below after briefly discussing contributions from various species.
Studies over the past 8 years have demonstrated that the fruit fly shows most of the fundamental features of mammalian sleep (Table 1). Flies lend themselves ideally to forward genetic approaches (Box 1) because their genome is less redundant than mouse or human genomes, which means that a mutation in the fly is more likely to yield a strong phenotype. Recently, the naturally occurring variation in both sleep phenotypes and mRNA expression levels of wild-derived inbred lines has also been used as a promising approach to identify candidate genes 13. The genetic basis of sleep will also benefit from studies in the zebrafish Danio rerio and the roundworm Caenorhabditis elegans (Table 1), both of which have recently been shown to have a sleep-like state 14–17.
Almost all mouse genes known to affect sleep have been identified by reverse genetics studies, with more than 70 mutant lines tested so far, starting from two pioneering reports in 1996 2,18 (Supplementary Table 2). Quantative trait loci (QTL) analysis (Box 1) has recently succeeded twice, in linking a single mouse gene to a specific frequency of the sleep EEG 19,20, while no mouse gene so far has been identified using mutagenesis screenings. A few general conclusions can be drawn from these studies. First, most mouse mutant lines show effects on at least one sleep phenotype, which is perhaps not surprising since these lines carry mutations in “candidate” genes. Second, the effects on sleep quantity – if any- are usually small, with increases or, in most cases, decreases in total sleep of ~ 20% or less, due to a decline in NREM or REM sleep. Very few mutations affect both sleep phases, and then almost always in the same direction (prokineticin 2-deficient mice 21 and corticotrophin-releasing hormone overexpressing mice 22 are exceptions; Supplementary Table 2). Third, changes in the response to sleep deprivation, when present, are also usually small, although caution is needed to interpret these results, because most studies do not assess sleep homeostasis in a comprehensive manner.
In several cases the genetic basis of human sleep disorders has been clarified, and the identified candidate genes have been tested in animal models. Examples that will be discussed include the hypocretin/orexin system and its role in narcolepsy, and Period (Per) genes, whose mutations result in abnormal circadian regulation.
Over the last 3 years, mutagenesis screenings have identified two fly genes with striking effects on fly sleep, namely Shaker and Sleepless. The first gene, Shaker, which was identified with EMS mutagenesis (Box 1), codes for the alpha subunit of a tetrameric potassium channel that passes a voltage-activated fast-inactivating IA current 23. Homologous channels in vertebrates have similar properties and, in both mammals and flies, IA plays a major role in the control of membrane repolarization and transmitter release 23. Flies carrying minisleep, or other Shaker loss of function mutations, sleep only 2–4 hours every day rather than 8–10 hours, but their circadian and homeostatic regulation of sleep are normal 24. Moreover, learning and memory in these flies is impaired, and lifespan is reduced, although it is still unclear whether these deficits can be ascribed to reduced sleep 24,25. Hyperkinetic codes for a beta regulatory subunit that interacts with the alpha pore forming subunits coded by Shaker. Hyperkinetic loss of function mutations also show a short sleeping phenotype, impairment in learning and memory, and reduced life span 25. In Hyperkinetic mutants sleep is not reduced as much as in Shaker mutants (by 30–50%), which is expected since Hyperkinetic null mutations reduce, but do not abolish, the IA current.
Sleepless, was identified using insertional mutagenesis 26 (Box 1). Sleepless flies, like the most extreme Shaker null mutants, sleep only ~ 2 hours a day (~ 85% less than controls), mainly due to a decrease in sleep episode duration. Sleepless codes for a glycosyl-phosphatidylinositol-anchored protein with unknown function, and has no obvious vertebrate homolog. Quiver, however, a previously identified mutation that affects the IA current, is allelic to Sleepless, and Sleepless flies have reduced levels of Shaker. This suggests that the Sleepless short sleeping phenotype is at least in part mediated by the Shaker current. There are nevertheless some interesting differences between Shaker and Sleepless flies, most notably that only the latter show a reduced homeostatic response, as indicated by no changes in sleep duration after sleep deprivation.
Why do mutations affecting the IA current decrease sleep duration so significantly? In mammals, neurons in the cortex and thalamus are more hyperpolarized in sleep than in waking 27, and one possibility is that these mutations may impinge on the core cellular mechanisms of sleep by changing overall neuronal excitability (Fig. 2). Based on sequence similarity, the closest mammalian homologues of the Drosophila Shaker gene are the alpha subunits of the Kv1 family of potassium channels, while the Kv2, Kv3, and Kv4 families are more distantly related (“Shaker-like”). Kv1 channels activate in the subthreshold voltage range in many neurons, and can act as extremely diverse regulators of neuronal excitability 28 (Figure 2). Mice lacking the closest mammalian homologue of Shaker, Kv1.2 (Kcna2), sleep less. Their short sleeping phenotype, however, is far from being as dramatic as in Shaker flies 29, perhaps because of redundancy - there is one Shaker gene in Drosophila, but at least 16 genes code for alpha subunits of voltage-dependent potassium channels in mammals 30,31.
A striking sleep phenotype is observed in double knockout (KO) mice lacking the voltage-dependent potassium channels Kv3.1 and Kv3.332, which sleep less (by 40%), have shorter sleep episodes, an overall decrease in the EEG power spectrum more evident in NREM sleep, and no response to sleep deprivation. These mice are also hyperactive and show motor dysfunction, which may be partly responsible for their sleep fragmentation 32,33. The mechanism underlying the effect of these mutations on sleep is most likely different from those involved in Shaker and Sleepless mutants, because Kv3-type channels are mainly expressed in cortical and thalamic GABAergic interneurons, where their presence enables these cells to fire repetitively at high-frequency 34. Whether human extreme short sleepers have mutations in voltage-dependent potassium channels remains unknown, but in one case of Morvan’s syndrome, a rare autoimmune disorder with central symptoms, marked sleeplessness has been associated with the presence of autoantibodies against voltage-dependent potassium channels 35.
Another candidate sleep gene that was identified in mice is Cav3.1, which codes the alpha1G subunit of the T-type calcium channels. These channels play a crucial role in most of the neuronal oscillations displayed by thalamic and cortical neurons during NREM sleep 36. Two independent studies have reported a decrease of ~20% in NREM sleep and more frequent brief awakenings, a sign of disrupted sleep, in mutant mice carrying a Cav3.1 deletion either globally or in most of the thalamus, while cortical deletion had no effect37,38. However, detailed experiments to assess the effects of these mutations on the sleep EEG still need to be carried out.
At the molecular level, circadian genes interact and determine circadian rhythmicity. Most of these genes have been tested for their effects on sleep (Supplementary Table 2). Double KO mice of the circadian genes cryptochrome1 and cryptochrome2 sleep more and have higher NREM delta activity, but show little further increase in sleep duration and intensity after sleep deprivation 39. Other mutations of circadian mouse genes (Clock, brain and muscle ARNT-like protein 1, neuronal PAS domain protein 2, prokineticin 2) also result in an abnormal response to sleep deprivation, as do mutations of the fly circadian genes Cycle and Clock (Table 1). Overall, it appears therefore that circadian mutations not only affect the timing of sleep, as expected, but also its homeostatic regulation.
A functional circadian clock seems to require oscillations of Per protein levels and its phosphorylation, and/or its rhythmic nuclear localization (Fig. 3). Flies lacking Per have a disrupted rest/activity cycle and reduced sleep, but whether the short sleeping phenotype is due to the specific loss of Per remains unclear 40. Loss of Per2 in mice (mammals have 3 Per genes) also disrupts locomotor activity rhythms; moreover, cortical Per2 levels change depending on both time of day (high at night) and behavioral state (high in waking) 39,41–43. Since Per2 induction is sensitive to the NAD/NADH ratio 44,45 it may reflect the level of brain activity 46, which varies with sleep and waking. Thus, Per2 could be an essential part of the central clock as well as of the mechanism that regulates sleep need based on waking duration (Fig. 3). However, studies in animals remain inconclusive, because Per1, Per2, Per3, and double Per1/Per2 KO mice have normal sleep duration and sleep homeostasis.
Studies in humans, instead, point to an important role for PERIOD3, which contains variable-number tandem-repeat polymorphisms in its coding region. Relative to PER34/4 carriers, PER35/5 individuals have longer duration of SWS and higher NREM delta activity, the latter also after sleep deprivation 47. PER35/5 subjects also show higher alpha and theta EEG activity during waking and REM sleep (both signs of increased sleep pressure), and after total sleep deprivation (but not chronic sleep restriction 48) are more impaired in executive functioning tasks in the early morning 47,49. Significant interindividual differences in the cognitive impairment caused by sleep loss have been found before (reviewed in 50). Recent experiments suggest that these individual differences are also associated with different patterns of neural activation as measured using functional MRI 51. Further studies are required to confirm that PER3 plays a crucial role in modulating susceptibility to sleep deprivation.
The human familial advanced sleep phase syndrome (FASPS) is transmitted in a highly penetrant autosomal dominant manner. FASPS subjects have a normal duration of sleep but go to sleep ~ 4-hour earlier than usual (extreme early birds). Some FASPS individuals carry a serine to glycine mutation in PER2, and the mutation in vitro affects the ability of casein kinase I epsilon to phosphorylate PER2 52. The same mutation in mice recapitulates the human FASPS phenotype 53. A missense mutation in CSNK1D (human casein kinase I delta gene) also results in FASPS, and reduces enzymatic activity in vitro 54. These and other studies show that the extent to which PERIOD proteins are phosphorylated greatly affects their nuclear accumulation and thus influence the endogenous circadian period, suggesting that future proteomic studies can help our understanding of how sleep and circadian rhythms are regulated. Genetic association studies have also suggested a link between delayed sleep phase syndrome (DSPS, extreme night owls) and several genes, including PER3 (refs in 55).
In general, genetic studies have largely confirmed the arousal-promoting role of the noradrenergic, histaminergic, serotonergic, cholinergic and hypocretin/orexin systems (detailed refs in Supplementary Table 2). In some cases, these studies have also revealed a more specific role for some of these systems in regulating certain sleep phenotypes. For instance hypocretins/orexins have been shown to be essential for stabilizing sleep and wakefulness 56. Indeed narcolepsy, a neurological disorder caused by hypocretins/orexins deficiency, is characterized by the inability to maintain long waking periods, abrupt transitions into NREM sleep, and abnormal intrusions of REM sleep into waking 56. Canine narcolepsy is an autosomal recessive, fully penetrant, disorder due to a mutation in the gene coding for the hypocretin receptor 2 4. In mice, a narcoleptic-like phenotype is present in both the preprohypocretin/orexin KO mice, where the hypocretins/orexins producing cells are spared but the corresponding peptides are lost, and in the hypocretin/orexin-ataxin 3 transgenic rats and mice, where both cells and peptides are missing.
Genetic studies have also shown that the noradrenergic and histaminergic systems are important in maintaining arousal, since mice deficient in histamine (Hdc KO mice) or norepinephrine/epinephrine (Dbh KO mice) tend to sleep more, and have less trouble in going back to sleep after exposure to mild stress. On the other hand, mice lacking the serotonin transporter SERT or serotonin receptor 1A have 40–50% more REM sleep, in line with the role of serotonin in the regulation of this behavioral state.
Reverse genetics has also clarified the mechanisms by which some drugs affect sleep. For instance, adenosinergic transmission, which is blocked by caffeine, plays an important role in the homeostatic regulation of sleep, and adenosine levels increase during waking and decrease during sleep in cortex and basal forebrain of rats and cats (reviewed in 57). Mice lacking the adenosine receptor A2AR are insensitive to the wake-promoting effects of caffeine, consistent with the fact that in humans a polymorphism in A2AR contributes to individual sensitivity in the effects of caffeine on sleep 57. In humans, a functional polymorphism in the gene coding for the catabolic enzyme adenosine deaminase (ADA) results in a ~ 20–30% decrease in enzymatic activity in blood cells, ~ 50% increase in SWS duration, and ~60% increase in NREM delta activity, with no change in total sleep time 58. Whether this genetic variation also affects the response to sleep deprivation remains untested.
A variety of molecules have been tested for their effects on sleep, but no systematic analysis of all major signaling pathways has been attempted so far. While it is not possible to provide a comprehensive summary of these studies in the main text (but see Supplementary Table 2), it is clear that a few mutations produce large (40–50%) changes in sleep quantity. They include those affecting RIM1alpha, a protein involved in synaptic vesicle release 59, and UBE3A, a ubiquitin protein ligase whose null mutation, however, also causes abnormal EEG discharges 60. Other signaling pathways have been linked to a specific frequency of the sleep EEG. During REM sleep the EEG shows prominent theta activity, which varies greatly in frequency among inbred mouse strains (slow and fast theta strains). By using QTL analysis followed by fine mapping, one study was able to link Acads to theta frequency 19. Acads encodes the short-chain acyl-coenzyme A dehydrogenase, an enzyme involved in fatty acid beta oxidation, but the mechanism by which its deficiency can slow down the peak theta frequency (from ~8 to 6 Hz) remains unclear 19. Another study that used forward, molecular, and reverse genetic approaches showed that Rarb, the gene encoding the retinoid acid receptor beta, is important to determine the contribution of delta activity to the EEG during NREM sleep 20. The mechanism by which this nuclear receptor affects the amount of delta activity, and thus cortical synchrony during sleep, is unclear, but it may depend on its role in development, neural plasticity, and dopaminergic transmission. Rarb does not seem to play a role in sleep homeostasis.
A recent study using 25 recombinant inbred mice strains found that the increase in NREM delta activity after sleep deprivation depends on both duration of sleep and genotype. It was found that Dps1, a QTL locus on mouse chromosome 13 that contains more than 200 genes, accounts for 49% of the genetic variance in this trait 61,62. Two follow-up studies identified Homer1a, an immediate early gene induced by neuronal activity 63,64, as a likely candidate gene within Dps1 65,66. The conclusion was based on the consistent and strong induction of Homer1a after sleep deprivation in the whole brain of several mice strains 65, the higher expression of cortical HOMER1A in the strain with the greatest increase in delta activity, and the presence of single nucleotide polymorphisms in Homer1a upstream regulatory region 66. It remains to be proven conclusively whether Homer1a plays a role in sleep homeostasis. Future studies should show, for instance, that Homer1a KO mice have a significantly impaired response to sleep deprivation. Since Dps1 contains numerous genes, it is also possible that other major candidates will be identified within this locus. Flies lacking the single Drosophila Homer gene are hyperactive and show defects in courtship behavior, but whether their sleep regulation is abnormal is unknown 67.
Fatal familial insomnia (FFI) is a rare autosomal dominant disease due to a point mutation at codon 178 of PRNP. The same mutation is also present in patients affected by the familial form of Creutzfeldt-Jakob disease (CJD), another prion disease with extensive cortical, rather than thalamic, degeneration, and in which dementia, rather than insomnia, is the main clinical feature. In FFI patients, codon 129 on the mutated allele codes for a methionine, while in CJD patients it codes for a valine (reviewed in 68,69). FFI is characterized by a decrease in sleep spindles, sleep fragmentation, reduction in total sleep time, loss of the circadian regulation of sleep, disappearance of SWS, and intrusion of a REM-like state into wakefulness. FFI patients show accumulation of abnormal PRNP in the brain, but PRNP levels do neither correlate with the severity of the disease, nor with the extent of neuronal degeneration. It is also unclear to what extent the sleep disturbances contribute to the death of FFI patients. Mice lacking Prnp have fragmented sleep, suggesting that the normal protein may promote sleep consolidation 2,3. Sleep problems also exist in CJD patients, and transgenic mice carrying the mouse homolog of the human D178N/V129 mutation show a large decrease in REM sleep and periods of a “mixed” state that cannot be easily classified as either sleep or waking 70.
While in dogs and mice narcolepsy is genetically determined, non-genetic factors play a major role in human narcolepsy, as suggested by its low (~30%) concordance in monozygotic twins. No association has been found between human narcolepsy and polymorphisms in the genes of the hypocretin/orexin system, and so far only one case of narcolepsy with an unusually early onset has been associated with a mutation in preprohypocretin 71. Yet, most patients with narcolepsy-cataplexy have low or undetectable levels of hypocretins, and narcolepsy is strongly associated with human leucocyte antigen (HLA) alleles, in particular HLA DQB1*0602. This suggests that the deficit in hypocretinergic neurotransmission in human narcolepsy could be due to an autoimmune attack 72. Consistent with this, a recent genome-wide association study found that narcolepsy is associated with the T-cell receptor alpha locus73.
RLS is a common sleep disorder often characterized by periodic limb movements during sleep. RLS can be familial (autosomal dominant in up to 1/3 of cases). Although dopaminergic agonists are used to treat primary RLS, no association has been found between this disorder and genes involved in dopaminergic transmission. Instead, recent genome-wide case-control studies identified 4 predisposing loci on chromosomes 2p, 6p, 9p and 15q. One locus is within the homeobox gene MEIS1, which is involved in limb development. The second within BTBD9, whose function may be related to iron storage, the third between the gene coding for the kinase MAP2K5 and LBXCOR1, a homeodomain transcription factor important for the development of GABAergic interneurons in the dorsal horn of the spinal cord, and the fourth within PTPRD (protein tyrosine phosphatase receptor type delta) 74,75. A second genome-wide association study also found an association between BTBD9 and RLS with period limb movements 76. Together, the 4 loci may explain more than 50% of the risk for RLS in individuals with European ancestry.
OSAS, a common disorder characterized by recurrent episodes of apnea/hypopnea (no or reduced airflow) during sleep, is also to some extent under genetic control. The genetic basis of this disorder, however, is difficult to study, because many of the risks factors for OSAS, including obesity and alterations of the craniofacial morphology, are also under genetic control 77.
It is well established that brains of sleeping and awake animals differ in terms of patterns of neural activity, content of many neurotransmitters and neuromodulators, metabolism, and ability to react to external stimuli. Recent whole-genome transcriptomic studies have now revealed that brains of sleeping and awake animals also differ at the molecular level. In rats, mice, flies, and sparrows hundreds of brain transcripts change their level of expression between sleep and wakefulness 41,65,78–86. In the most tested species so far, rats and mice, these changes occur mainly in the cerebral cortex, cerebellum, and hypothalamus, but also other brain areas. For instance, in one study up to ~5% (752 of 15,459) of the transcripts tested in the cerebral cortex was up- or down-regulated in rats that had slept for 8 hours relative to rats that had been spontaneously awake or sleep deprived for 8 hours 41. A similar number of cortical transcripts changed their expression because of time of day independent of behavioral state, suggesting that day/night time and sleep/wakefulness influence cortical gene expression to a similar extent 41. In general, across studies, there are as many “sleep” transcripts (with higher expression during sleep relative to waking) as there are “waking” transcripts. This result suggests that, despite being usually associated with behavioral inactivity, sleep is far from being a quiet state at the cellular level. Most importantly, sleep-related and wakefulness-related transcripts belong to different functional categories, suggesting that the two behavioral states may favor different cellular processes 41. Moreover, several of the molecular correlates of sleep and wakefulness first identified in rats and mice have subsequently been found in fruit flies and/or sparrows (Fig. 4). The overlap in some major gene categories is remarkable considering that these studies differ not only with respect to the animal species that was tested, but also in terms of design (e.g. sleep deprived vs. spontaneously awake animals), duration of wakefulness (3–8 hours), brain region, and statistical approach.
Three functional categories of transcripts are most consistently increased during waking and short-term sleep deprivation relative to sleep (Fig. 4). The first group comprises genes involved in energy metabolism, including those coding for mitochondrial proteins, glucose transporters, and proteins related to glycogen metabolism 86–89. Their up-regulation may represent a mechanism by which the brain responds to the high energy requirements of wakefulness. This response, however, does not seem to persist when sleep deprivation is prolonged for more than a few hours 80,90. Imaging studies in animals 91, normal human subjects 92, and FFI patients 93 also show either no change or a decrease in cerebral metabolic rate after prolonged sleep loss.
The second group of waking-related transcripts codes for proteins involved in the response to cellular stress, including heat shock proteins and chaperones 41,78,79,82,83,85,86. This suggests that the absence of sleep may indeed represent a cellular stress for brain cells. In support of this hypothesis, in the mouse cerebral cortex a few hours of sleep deprivation induce the “unfolded protein response”, a global stress response that involves the induction of the chaperone BiP in the endoplasmatic reticulum, which promotes the degradation of misfolded proteins, and a decrease in protein synthesis 94. Importantly, there are no major signs of brain cellular damage after prolonged waking (refs in 95,96), an indication that the stress response must be protective. This is consistent with the observation that mutant flies unable to mount a strong stress response during sleep deprivation die of sleep loss earlier than controls 97.
The third group of wakefulness-related transcripts plays a role in synaptic plasticity, and more specifically in synaptic potentiation 41,65,79,86,98. Overall, the induction of these genes suggests that synaptic potentiation is favored during wakefulness relative to sleep. Consistent with this, recent experiments in flies and rats show that molecular and electrophysiological markers of net synaptic potentiation prevail in waking, while markers of net synaptic depression are present in sleep 99–101.
Some of the transcripts with increased expression during sleep are involved in protein synthesis 41,84–86(Fig. 4). These findings are in agreement with previous studies that identified a positive correlation between sleep and protein synthesis 102–107. Whether sleep favors protein synthesis globally, or enhances the synthesis of specific classes of proteins, is still unclear.
Another group of sleep-related transcripts includes CamK4 (calmodulin-dependent protein kinase IV), a gene that has recently been involved in synaptic scaling 108, and other genes that have been associated with synaptic depression and depotentiation, such those coding for calcineurin, FK506 binding protein 12, inositol 1,4,5-trisphosphate receptor and amphiphysin II 41. Thus, while wakefulness is the appropriate time for memory acquisition and synaptic potentiation, sleep may favor complementary aspects of plasticity, such as synaptic consolidation and/or downscaling. An involvement of sleep in such processes is suggested by behavioral and physiological experiments showing that sleep improves the performance of different learning tasks 109,110. At this stage, however, the mechanism by which sleep enhances performance is still debated. One idea is that sleep consolidates synapses activated by learning during the previous waking period, perhaps through mechanisms that require cortical spindle activity and the slow waves of NREM sleep 111,112. Another possibility is that, since wakefulness is associated with a diffuse potentiation of synaptic circuits that results in a net increase in synaptic weight, sleep produces a generalized depression of synapses. This downscaling would benefit the brain because it decreases the energetic cost of synaptic activity, eliminates weak and ineffective synapses, reduces cellular stress and increases signal to noise ratios 113.
Finally, a large group of sleep-related transcripts is involved in membrane trafficking and maintenance 41,84–86. Some of these transcripts are involved in exocytosis and neurotransmitter release, others in synaptic vesicle recycling, tethering/docking of vesicles to their target organelles, and cycling between trans-Golgi network and plasma membrane. Other transcripts are important for the synthesis/maintenance of membranes in general and of myelin in particular, including oligodendrocitic genes coding for myelin structural proteins, myelin-related receptors, and enzymes. Also, transcripts with higher expression in sleep code for enzymes involved in the synthesis and transport of cholesterol, a major constituent of myelin and other membranes and an important factor in regulating synaptic efficacy 114,115. Depletion of cholesterol/sphingolipid leads to instability of surface AMPA receptors and gradual loss of synapses and dendritic spines 116. Thus, it may not be by chance that sleep seems to be linked to membrane trafficking and cholesterol synthesis on one hand, and to protein synthesis and synaptic homeostasis on the other hand.
There is little doubt that more genes affecting sleep phenotypes will be identified in the near future, and it is reasonable to assume that most mutations will have less striking effects on sleep in mammals than in simpler organisms, since mammalian genomes are redundant. Even in simpler organisms, however, we should not expect single genes to completely abolish sleep or its regulation, because sleep is a complex and highly regulated behavior, present in all the animal species that have been carefully studied so far, suggesting that it has one or more fundamental functions, such as maintaining brain plasticity and saving energy 12. A similar argument applies to the brain regions that are involved in initiating and maintaining sleep and waking: there are several sleep-promoting and arousal-promoting neural systems, and the lesion of one or even a few of them does not result in the complete and permanent lack of either sleep or waking.
Sleep genetic studies are bound to contribute further to our understanding of sleep regulation and function but are also challenging. The identification of the mutated gene in animal models or human disorders is becoming faster and cheaper by the day. One major challenge lies in the choice of the phenotype: Which sleep parameters should we measure when searching for “insomnia genes”? Another issue is the large number of subjects required for genome-wide association studies, which may be a limiting factor unless extensive collaborations across sleep centers around the world are initiated. Finally, large-scale mutagenesis screening remains time-consuming and costly in mice, until – in parallel to flies - automatic, high-throughput and yet sensitive methods to assess sleep in hundreds of animals at a time become available.
Nevertheless, as reviewed in this article, some genes with profound effects on sleep have already been identified even in mammals, and others will be discovered. Neurons are more hyperpolarized during sleep than during waking, and mutations that can significantly change the resting membrane potential, the balance between inhibitory and excitatory neurotransmission, and/or the overall neuronal excitability of large sets of neurons are likely to affect quantity and quality of sleep, as Shaker, Sleepless, and Kv3.1/Kv3.3 mutants suggest. Also, in flies as well as in mammals, sleep homeostasis reflects not only the duration of prior waking, but also its intensity, and sleep need increases when waking is associated with learning 7,117,118. Thus, it is likely that mutations in many plasticity-related genes will affect the homeostatic regulation of sleep. Overall, the identification of mutations that change the need for sleep, or make subjects resistant to the negative effects of sleep deprivation, may prove crucial to further our understanding of the functions of sleep.
This work was funded by NIH grants P20 MH077967 and R01 GM075315. I thank Drs Daniel Bushey, Stephanie Maret and Giulio Tononi for critical comments on the manuscript, and Drs Daniel Bushey, Emmanuel Mignot and David Raizen for providing the pictures for Table 1.