Journal of Applied Physiology

Invited Review: Molecular genetic studies on sleep-wake regulation, with special emphasis on the prostaglandin D2 system

Osamu Hayaishi


To elucidate the exact role of the PGD2 system in sleep-wake regulation in vivo, the sleep behavior of knockout mice, generated in the author's and other laboratories, was examined for lipocalin-type PGD synthase (L-PGDS), PGD receptor, adenosine A2A receptor, and histamine H1 receptor; transgenic mice overexpressing the human L-PGDS gene, generated in the author's laboratory, were also examined. The circadian profiles of sleep patterns of wild-type and the genetically manipulated mice were essentially identical, indicating the possibility that the deficiency of one system may be effectively compensated by some other systems during development. Available evidence indicated that the PGD2 system is involved in the homeostatic regulation of non-rapid eye movement sleep and that the arousal effect of orexin A is mediated by the histamine H1receptor system.

  • knockout mice
  • transgenic mice
  • adenosine
  • histamine

during the last several decades,molecular genetic approaches have become an increasingly popular and powerful tool in the study of the molecular mechanisms underlying sleep-wake regulation. The fruits of these investigations have been the subject of extensive reviews in recent years. For example, in a special issue of the Journal of Sleep Research entitled “Genetic approaches to Sleep and Sleep Disorders”, 10 contributors presented state-of-the-art reviews up to mid-1999 (36). A symposium entitled “Genetics of Sleep” was held at the Third International Congress of World Federation of Sleep Research Societies in Dresden in October 1999 (38). Because the orexin/hypocretin system was suggested as having a role in the cause of narcolepsy, this subject has also been reviewed in depth (19). In the present article, therefore, focus is placed mainly on the ongoing and recent experimental studies in my laboratory related to the molecular genetics of sleep and wakefulness, with special emphasis on the PGD2and related systems. In addition, the possible involvement of phosphatidylinositol 4,5-bisphosphate, another lipid mediator in the brain function, in rapid eye movement (REM), but not non-rapid eye movement (NREM), sleep has been studied with phospholipase C (PLC)-β4 knockout (KO) mice (17).


PGD2 has been implicated as a physiological regulator of sleep because PGD2 is the major prostanoid in the mammalian brain and the intracerobroventricular infusion of femtomolar amounts per minute of PGD2 induced both NREM and REM sleep in rats, mice, and monkeys. Sleep promoted by PGD2 was indistinguishable from natural sleep as judged by several electrophysiological and behavioral criteria, whereas sleep induced by hypnotic drugs manifested itself differently from natural or PGD2-induced sleep.

Lipocalin-type PGD synthase (L-PGDS), the enzyme that produces PGD2 from PGH2 in the brain, is a monomeric glycoprotein with a molecular weight of ∼26,000. The enzyme has been purified from the brains of rats, frogs, and humans, and its crystal structure, as well as catalytic properties, was determined. When selenium chloride, a potent, specific, and reversible inhibitor of L-PGDS was infused into the third ventricle of a rat during daylight hour, slow-wave sleep and REM sleep were inhibited time and dose dependently. After ∼2 h from the start of the infusion, both slow-wave sleep and REM sleep were almost completely but reversibly inhibited, indicating that L-PGDS plays a crucial role in sleep regulation. These earlier studies up to 1999 have been reviewed previously (10-13, 37).

Whereas our studies were mainly carried out with rats and monkeys, Roberts and co-workers (28) reported in 1980 that the endogenous production of PGD2 increased up to 150-fold in patients with systemic mastocytosis during deep sleep episodes. Subsequently, the PGD2 concentration was shown to be elevated progressively and selectively up to 1,000-fold in the cerebrospinal fluid (CSF) of patients with African sleeping sickness (25). These clinical observations are consistent with the concept that excessive endogenous production of PGD2 is responsible for sleep also in humans under certain pathological conditions. However, experiments to verify this assumption have not yet been carried out in animals.

To examine how the L-PGDS gene and endogenously produced PGD2 function in vivo, we generated transgenic (TG) mice by incorporating the human L-PGDS gene into mice (26). Northern blot analyses clearly revealed that human L-PGDS mRNA was overexpressed in almost all tissues, including the brain of these TG mice. The TG mice were therefore expected to sleep most of the time. However, contrary to expectations, the animals appeared to be quite healthy and to grow and to sleep normally. In fact, there was no significant difference in the overall sleep pattern between the wild-type (WT) and TG mice. However, when the tails of the mice were clipped for DNA sampling at 8:00 PM, the amount of NREM sleep, but not that of REM sleep, of the TG mice increased sharply and significantly. This effect seemed to last for several hours, after which the amount of NREM sleep returned to the control level after 5–6 h. The maximum increment was almost as high as the maximum amount of sleep during the daytime. However, the sleep pattern of the WT mice was essentially unaffected by the tail clipping. These somewhat unexpected results may be explained by assuming that L-PGDS is not a rate-limiting enzyme in the arachidonic cascade system and that the rate-limiting enzyme located upstream of the L-PGDS-catalyzed step may be induced or activated by a noxious stimulus such as tail clipping. In fact, the enzyme cyclooxygenase, so-called COX, is generally believed to be the rate-limiting enzyme rather than the individual synthases. If this were the case, the sleep patterns of WT and TG mice would be essentially identical. It is possible, however, that the pain stimulus activated COX, which then produced an excessive amount of PGH2, the substrate for L-PGDS. The L-PGDS step would then become the rate-limiting step under these conditions, thus leading to the production of a quantity of PGD2 larger in the TG mice than in the WT mice and ultimately to an increased amount of NREM sleep in the TG mice. To test the validity of this interpretation, the amount of PGD2 produced in the brains of WT and TG mice was measured before and after the tail clipping. As expected, the amount of PGD2 in the brains of WT and TG mice before the tail clipping were essentially identical, even though the amount of L-PGDS in the brains of TG mice was far greater than that in the WT mice. The amount of PGD2 in the brains of the TG mice increased sharply and significantly for ∼3 h after the tail clipping and then started to decrease thereafter. These changes almost exactly paralleled the time course of changes in NREM sleep. In contrast, the PGD2 content in the brains of the WT controls remained essentially the same. Thus it seems reasonable to conclude that the increase in NREM sleep in the TG mice after tail clipping was probably due to the inflicted pain causing the induction of COX or possibly some other rate-limiting enzyme(s) upstream of the L-PGDS step, resulting in an increased level of PGH2 in the brain, and prostanoid being converted to a large amount of PGD2 by the excessive amount of PGDS in these TG mice.

L-PGDS KO mice were also generated. Circadian profiles of NREM and REM sleep in WT and L-PGDS KO mice were almost identical, suggesting that the absence of L-PGDS does not affect basal sleep patterns (8). Sleep is regulated as a function of prior wakefulness, and sleep propensities increase during waking or sleep deprivation (SD). Eguchi et al. (8) investigated the effect of 6 h of SD on the amounts of NREM and REM sleep in WT and L-PGDS KO mice. In the WT mice, after SD, the amount of NREM sleep increased about twofold compared with that before SD, indicating that a strong rebound of NREM sleep occurred during the recovery period. However, in the KO mice, the amount of NREM sleep after SD was almost the same as that before SD, indicating that the NREM rebound did not occur after SD. In contrast, REM sleep rebound after SD was observed in WT mice as well as in KO mice. Eguchi et al. (8) further analyzed the electroencephalogram (EEG) power spectra of each type of sleep. The spectrum distribution pattern of NREM sleep was almost the same between WT and KO mice before SD and also after SD. The spectrum distribution pattern of REM sleep was also identical between WT and KO mice before and after SD. These results showed that SD did not affect the distribution pattern of the EEG power spectrum of either type of sleep in KO or WT mice. The PGD2 content in the brains of the WT mice after SD was approximately twofold higher than that before SD, whereas the amount of PGD2 in the brains of the KO mice was unchanged after SD. These results indicate that a deficiency of L-PGDS did not increase sleep propensity for NREM sleep during SD and that endogenous PGD2 is probably involved in the homeostatic regulation of NREM sleep (8).

These results taken together indicate that 1) PGD2 is an endogenous sleep-promoting substance in the brain of mammals, 2) L-PGDS plays a crucial role in sleep regulation, and 3) the PGDS gene is the first gene to be shown to be involved in the homeostatic control of NREM sleep.


L-PGDS is mainly located in the membrane system surrounding the brain rather than in the brain parenchyma, such as leptomeninges and choroid plexus except in oligodendrocytes, and is secreted into the CSF to become a β-trace protein (37). β-Trace was originally discovered by Clausen in 1961 to be the second most abundant CSF protein next to albumin. PGD2 produced by L-PGDS also circulates in the CSF and exhibits circadian fluctuation in parallel with the sleep-wake cycle. The infusion of PGD2 via a microdialysis probe showed that PGD2 did not induce sleep in most parts of the brain parenchyma but did effectively promote sleep when it was infused into the subarachnoid space underlying the rostral basal forebrain of rats, the so-called PGD2-sensitive zone, indicating the presence of a cluster of PGD receptors (DPR) in this area (23). The DPR gene was originally cloned and characterized by Hirata et al. (15); thereafter, DPR-deficient mice were generated (22). Immunofluorescence staining of the adult mouse brain revealed that DPR immunoreactivity (IR) was mainly localized in the leptomeninges of the basal forebrain, and electron microscopic observation indicated that DPR-IR particles were predominantly located on the plasma membranes of arachnoid trabecular cells of the leptomeninges (23). To find out how the DPR-mediated signal is transmitted into the brain parenchyma, a number of neurotransmitters, peptides, and hormones were applied to the PGD2-sensitive zone. Adenosine, especially A2Aagonists but not A1 agonists, was found to be the only compound that could mimic the somnogenic activity of PGD2(30).

Administration of CGS-21680, a specific adenosine A2Aagonist but not an agonist of the A1 subtype, into the subarachnoid space induced sleep, suggesting that PGD2-induced sleep may be mediated by adenosine through the adenosine A2A receptor (A2AR) system. Both CGS-21680- and PGD2-induced sleep were attenuated by pretreatment of rats with KF-17837, an A2AR antagonist, in a dose-dependent manner. When PGD2 was applied to this area, the amount of adenosine in the extracellular space increased time and dose dependently. However, in the DPR-deficient mice, the amount of NREM sleep did not increase after PGD2 infusion; the extracellular adenosine level was also unchanged in the subarachnoid space of the rostral basal forebrain (27). Therefore, DPR-expressing cells are considered to be important for the production and release of adenosine into the subarachnoid space of the rostral basal forebrain in mice, which results in sleep induction. Because the baseline sleep-wake patterns of WT and DPR-deficient mice were essentially identical, these results also indicate that DPR and PGD2 are involved in the homeostatic regulation of NREM sleep. Quite recently, another type of receptor for PGD2, CRTH2, was discovered in the T helper type 2 cells, eosinophils, and basophils in humans and mice (1, 14). The CRTH2 receptor is involved in the chemotactic activity of PGD2 toward these cells and is also expressed in the mouse brain. However, the expression level of mRNA for CRTH2 in the brain was essentially unchanged between the WT and DPR-deficient mice, suggesting that CRTH2 may be unrelated to the PGD2-induced sleep and adenosine release.

My group has also employed A2AR KO mice to confirm our hypothesis that PGD2-induced sleep is mediated by the A2AR system. Preliminary evidence indicated that PGD2 exerted its somnogenic effects in a manner at least partially dependent on the A2AR system (27), somewhat analogous to the interaction between A2AR and D2 dopamine receptor (7).


Wakefulness, REM sleep, and NREM sleep are driven by highly complicated patterns of neuronal and humoral activities in specific wake- and sleep-generating systems, but the target cells are usually difficult to identify in in vivo experiments. The immediate early gene product Fos is a useful marker of neuronal activation that has facilitated the identification of wake- and sleep-active neurons. Using Fos immunohistochemistry, Sherin and co-workers (35) recently identified a discrete cluster of neurons within the ventrolateral preoptic area (VLPO) that may play a critical role in the generation of sleep. The VLPO sends specific GABAergic and galaninergic efferents to the core of the tuberomammillary nucleus (TMN) (34), the source of the ascending histaminergic arousal system and an area considered to be a wake center.

To determine the neural regions involved in the response to PGD2 or adenosine, Fos immunohistochemistry was used to identify neurons activated by infusion of PGD2 or an A2A agonist into the subarachnoid space, in collaboration with Scammell and co-workers (31). PGD2increased NREM sleep and induced striking expressions of Fos in the VLPO, the basal leptomeninges, and several other brain regions that may be related to sleep (31).

Scammell et al. (31) found that infusion of PGD2 into the subarachnoid space just anterior to the preoptic area induced Fos IR in the VLPO in association with an increase in NREM sleep. This neuronal activation was accompanied by a decrease in the Fos level in the putative wake-active neurons of the TMN. These observations suggest that PGD2 may induce sleep via meningeal DPR and probably adenosine A2AR with subsequent activation of the VLPO. PGD2 does not promote sleep when infused below the posterior hypothalamus, and it is thus unlikely that wake-promoting neurons of the TMN are directly inhibited by PGD2. PGD2 induces sleep most effectively when infused into the subarachnoid space just anterior to the preoptic area, and PGD2 can increase the firing rates of sleep-active preoptic neurons. Within the preoptic area, sleep-active neurons are most abundant in the VLPO region. Scammell et al. showed that PGD2 induces Fos IR in the VLPO in proportion to the production of sleep. Considered together, these observations indicate that PGD2 may stimulate sleep-active VLPO neurons.

The VLPO may induce sleep through inhibition of wake-promoting TMN neurons (34). The VLPO projects heavily to the proximal dendrites and soma of TMN neurons; most of these axons contain GABA and galanin, i.e., inhibitory neurotransmitters. Neurons of the TMN are tonically active during the waking state, are less active during NREM sleep, and cease firing during REM sleep. GABA levels are elevated in the posterior hypothalamus during sleep, and electrical stimulation of the lateral preoptic area can elicit GABAA-mediated inhibitory postsynaptic potentials in TMN neurons. Galanin also hyperpolarizes and decreases the firing rate of TMN neurons. Furthermore, insomnia caused by preoptic neuron lesions was reversed by muscimol injection into the posterior hypothalamus in the cat (29). The highly sensitive Fos antiserum used in the Scammell et al. study (31) has permitted anatomic identification of wake-active TMN neurons and the demonstration of an inverse relationship between Fos IR in the VLPO and TMN. These findings indicate that when the VLPO neuronal activity is increased that of TMN neurons is decreased, further supporting the hypothesis that GABAergic or galaninergic projections from the VLPO contribute to the inhibition of the TMN, thus generating NREM sleep by a “flip-flop” mechanism.

Although more experiments are needed to work out the detailed mechanisms, my laboratory's current tentative conclusion is summarized by the scheme given in Fig. 1. PGD2 is produced in the trabecular cells of the arachnoid membrane and choroid plexus and circulates in the CSF and promotes sleep by inducing meningeal cells to release paracrine signaling molecules such as adenosine, which subsequently excite nearby sleep-active neurons via A2AR-expressing neurons projected to VLPO. These cells then send inhibitory signals to downregulate the neuronal activity involved in the maintenance of wakefulness in the TMN, which contributes to arousal through the histamine H1receptor (H1R).

Fig. 1.

Schematic representation of the molecular mechanisms of sleep-wake regulation by PGD2. Solid horizontal arrow represents excitatory; open horizontal arrow represents inhibitory.


Narcolepsy is a unique neurological disorder characterized by a persistent daytime sleepiness and abnormal REM sleep. Dysfunction of the orexin/hypocretin system in narcoleptic dogs, mice, and humans suggests that this system plays an important role in the aetiology of narcolepsy (19). For example, mice lacking the orexin peptide display an increased propensity for REM, and also NREM sleep, and a decrease in awake time during the active period of normal rodents (5). Canine narcolepsy is caused by a mutation in the orexin 2 receptor (OX2R) gene (21). Furthermore, human narcolepsy is associated with a deficiency in the orexin system. These findings indicate that orexin-OX2R interaction is involved in pathological sleep regulation in humans and animals. However, the role of orexin in physiological sleep and the mechanisms involved in vigilance control are still unknown. The orexin neurons are exclusively localized in the lateral hypothalamus and project their fibers to most aminergic nuclei, including the histaminergic TMN, where OX2R are abundant. It is well known that activation of the histaminergic system promotes wakefulness through activation of H1R (4, 20,24). Furthermore, administration of modafinil, an increasingly popular wake-promoting drug used for the treatment of narcolepsy, produces wakefulness in rats in association with activation of the TMN (32). These findings suggest that the histaminergic system may play a role in the orexin system in narcolepsy.

Recently, Huang et al. (16) demonstrated that infusion of orexin A into the third ventricle of WT mice increased arousal from the second 2 h onward but did not significantly change the sleep-stage distribution during the first 2 h of infusion. A similar delay for 1 h was also observed in rats after the intracerobroventricular administration of orexin A. Because the TMN was recently shown to be a putative wake center, the investigators used a microdialysis probe to directly deliver orexin A to the TMN of rats. Orexin A (5 and 25 pmol/min) promptly increased wakefulness after the start of perfusion, and the arousal effect was prolonged for 1 h after the perfusion had ended, clearly showing that the TMN or its vicinity is a direct action site for the arousal effect of orexin A (16).

The histaminergic neurons are located in the TMN, where a dense distribution of orexin A-IR processes and OX2R expression are observed. Histamine promotes cortical wakefulness probably either through direct cortical projections or by tonic control over the sleep-generating mechanisms in the preoptic/anterior hypothalamus. To examine whether orexin activates the histaminergic system, we employed an in vivo microdialysis method and found that perfusion of orexin A into the TMN of rats increased histamine release from both the medial preoptic area and the frontal cortex, both of which have been implicated in the arousal effect of histamine, suggesting that application of orexin A into the TMN activated the histaminergic system in the brain. The duration of the increase in histamine release from the medial preoptic area and the frontal cortex almost paralleled that of the arousal effect. These results are supported by the observation that orexin A markedly increased the basal firing rate of histaminergic neurons in vitro and suggest that activation of the histaminergic system is involved in the orexin A-induced wakefulness.

Histamine has been reported to promote wakefulness through activation of H1R. H1R KO mice provided a valuable tool to study the role of the histaminergic system in orexin A-induced arousal. H1R KO mice were initially generated by Inoue et al. in 1996 (18), and their behavior and their neuropharmacological characteristics have been investigated in detail. Although these authors reported that mice lacking H1R showed a significant decrease in ambulation in an open field and on an activity wheel, electrophysiological studies have not been carried out so far.

Results from my laboratory (16) showed that mice of each genotype displayed essentially the same amounts of sleep and wakefulness under basal conditions, yet orexin A infusion into the TMN significantly increased the wakefulness in WT mice but not at all in the H1R KO mice, clearly indicating that H1R plays a crucial role in mediating orexin A-induced arousal. Nevertheless, contrary to the narcoleptic symptoms manifested in orexin KO mice, these symptoms were not detected in any of the H1R KO mice examined, as judged either by electrophysiological criteria or from the results of infrared video recordings. The detailed mechanisms underlying the interaction between orexinergic and histaminergic systems are currently under investigation.


PLC catalyzes the formation of two second messengers, inositol trisphosphate and diacylglycerol, from phosphatidylinositol 4,5-bisphosphate. Inositol trisphosphate mobilizes intracellular Ca2+, and diacylglycerol activates protein kinase C. Because phosphatidylinositol 4,5-bisphosphate is another lipid mediator involved in the signal transduction in the brain and Ca2+has recently been implicated in sleep regulation (2), my group decided to study the sleep behavior of mutant mice lacking PLC-β4. PLC-β4 KO mice were generated and found to show arrhythmic REM sleep with unusual REM/wake repeats and a faster thetalike EEG wave during REM sleep. These changes appear to be rather specific for REM sleep because the NREM sleep of these KO mice was essentially identical to that of WT mice (17).

To date, gene KO mice have been used to demonstrate the involvement of orexin (5), the serotonin 1β-receptor (3), and albumin D binding protein (9) in REM sleep. Respective roles of these compounds in REM sleep regulation and their possible interaction need further investigation.


Although the molecular genetic approach to sleep research has become an increasingly powerful tool and has yielded numerous interesting and sometimes unexpected experimental results, the molecular mechanisms underlying these phenotypes are not yet clear in most instances. It is somewhat surprising that in most, if not all, of these genetically manipulated mice, the circadian sleep patterns are usually essentially unchanged. This is probably due, in large part, to the postgenomic events such as the complex and diverse nature of the neuronal and humoral regulatory circuits as well as the compensatory changes during development. Many prominent changes in phenotypes became visible after postgenomic conditioning, such as pain stimulus, SD, and so forth, as exemplified by several instances in this review.

Gene targeting, which either introduces specific mutations or creates null mutations (KO) or excessive expression (TG), is obviously a powerful tool in biology. However, most experiments are carried out with mice, and it is not possible to extrapolate these results or even speculate about human neural systems on the basis of nonhuman (and usually nonprimate) neural systems. Furthermore, the data are sometimes difficult to interpret because of polymorphism of the genetic background. Nevertheless, it is gratifying to see the increasing numbers of genetically manipulated models being produced and tested in numerous laboratories, yielding a wealth of new insights into the mystery of sleep.


The author expresses deep gratitude to Y. Urade, N. Eguchi, J. F. Chen, S. Narumiya, K. Kitahama, F. Tsuji, T. Yoshioka, and I. Tobler for their collaboration and/or help in the preparation of this manuscript.


  • The work from this laboratory has been mainly supported by a Health Science Research Grant of the Ministry of Health, Labor, and Welfare of Japan (100107 to O. Hayaishi) and by grants from the Special Coordination Funds for Promoting Science and Technology of the Ministry of Education, Culture, Sports, Science and Technology of Japan (to O. Hayaishi and Y. Urade).

  • Address for reprint requests and other correspondence: O. Hayaishi, Osaka Bioscience Institute, 6-2-4 Furuedai, Suita, Osaka 565-0874, Japan (E-mail: hayaishi{at}

  • 10.1152/japplphysiol.00766.2001


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 7.
  7. 8.
  8. 9.
  9. 10.
  10. 11.
  11. 12.
  12. 13.
  13. 14.
  14. 15.
  15. 16.
  16. 17.
  17. 18.
  18. 19.
  19. 20.
  20. 21.
  21. 22.
  22. 23.
  23. 24.
  24. 25.
  25. 26.
  26. 27.
  27. 28.
  28. 29.
  29. 30.
  30. 31.
  31. 32.
  32. 33.
  33. 34.
  34. 35.
  35. 36.
  36. 37.
  37. 38.
View Abstract