Theoretical Review
The neuronal network responsible for paradoxical sleep and its dysfunctions causing narcolepsy and rapid eye movement (REM) behavior disorder

https://doi.org/10.1016/j.smrv.2010.08.002Get rights and content

Summary

Rapid eye movement (REM) sleep behavior disorder (RBD) is a parasomnia characterized by the loss of muscle atonia during paradoxical (REM) sleep (PS). Conversely, cataplexy, one of the key symptoms of narcolepsy, is a striking sudden episode of muscle weakness triggered by emotions during wakefulness, and comparable to REM sleep atonia. The neuronal dysfunctions responsible for RBD and cataplexy are not known. In the present review, we present the most recent results on the neuronal network responsible for PS. Based on these results, we propose an updated integrated model of the mechanisms responsible for PS and explore different hypotheses explaining RBD and cataplexy. We propose that RBD is due to a specific degeneration of a sub-population of PS-on glutamatergic neurons specifically responsible of muscle atonia, localized in the caudal pontine sublaterodorsal tegmental nucleus (SLD). Another possibility is the occurrence in RBD patients of a specific lesion of the glycinergic/GABAergic pre-motoneurons localized in the medullary ventral gigantocellular reticular nucleus. Conversely, cataplexy in narcoleptics would be due to the activation during waking of the caudal PS-on SLD neurons responsible for muscle atonia. A phasic glutamatergic excitatory pathway from the central amygdala to the SLD PS-on neurons activated during emotion would induce such activation. In normal conditions, the glutamate excitation would be blocked by the simultaneous excitation by the hypocretins of the PS-off GABAergic neurons localized in the ventrolateral periaqueductal gray and the adjacent deep mesencephalic reticular nucleus, gating the activation of the PS-on SLD neurons.

Introduction

In 1959, Jouvet and Michel discovered in cats paradoxical sleep (PS), a sleep phase characterized by a complete disappearance of muscle tone, paradoxically associated with a cortical activation and rapid eye movements (REM).1, 2 Soon after, they demonstrated that the brainstem is necessary and sufficient to trigger and maintain PS in cats. Using electrolytic and chemical lesions, it was then evidenced that the dorsal part of pontis oralis (PnO) and caudalis (PnC) nuclei contains the neurons responsible for PS onset.3, 4, 5, 6, 7 Furthermore, bilateral injections of a cholinergic agonist, carbachol (Table 1), into these structures promotes PS in cats.8 The dorsal part, where carbachol injections induce PS with the shortest latencies, was coined peri-locus coeruleus alpha nucleus (peri-LCα), pontine inhibitory area (PIA) or subcoeruleus nucleus (SubC).9, 10, 11, 12, 13, 14, 15 An experimental milestone in the field was the discovery by unit recordings in freely moving cats that many peri-LCα neurons (called “PS-on” neurons) show a tonic firing selective to PS.15, 16, 17, 18 Two types of PS-on neurons were segregated. The first ones were inhibited by carbachol, an indication that they might be cholinergic since such neurons express inhibitory autoreceptors.19 They were restricted to the rostro-dorsal peri-LCα and project to rostral brain areas including the intralaminar thalamic nuclei, the posterior hypothalamus and the basal forebrain. The second type of PS-on neurons recorded over the whole peri-LCα were excited by carbachol and projected caudally to the nucleus reticularis magnocellularis (Mc) within the ventromedial medullary reticular formation.14, 15, 18

It has been proposed that 1) the ascending PS-on neurons are cholinergic and are responsible for the cortical activation during PS and 2) the descending PS-on neurons are not cholinergic and generate muscle atonia during PS through excitatory projections to medullary glycinergic pre-motoneurons.17, 18, 20, 21, 22, 23

In contrast to data in cats, carbachol iontophoresis into the rat sublaterodorsal tegmental nucleus (SLD), the equivalent of the cat peri-LCα induces waking (W) with increased muscle activity.24 Other studies using carbachol administration in freely moving rats described either a moderate PS enhancement compared to cats25, 26, 27, 28 or no effect.29 The lack of or the weak effect of carbachol injection in rats compared to cats clearly indicates that there is a species difference. Finally, although it was initially shown that the number of pedunculopontine (PPT) and laterodorsal tegmental nucleus (Ldt) cholinergic neurons expressing the early gene c-Fos (a marker of activated neurons) increased in rats during PS recovery following PS-selective deprivation by the flower-pot technique,30 we did not confirm these findings using nearly the same approach.31 We indeed found a large number of c-Fos labeled cells located in the Ldt and SLD after PS hypersomnia but only a few of these neurons were cholinergic.31 In conclusion, our results in rats are strongly against a role of pontine cholinergic neurons in PS generation. Unit recordings of pontine and juxtacellular recordings of forebrain cholinergic neurons suggest that these cholinergic neurons are active both during W and PS and therefore could be involved in EEG activation during both states rather than in PS genesis itself.32, 33 To draw a more definitive conclusion, unit recordings combined with juxtacellular labeling of pontine cholinergic neurons, as well as local bilateral injection in the SLD of a cholinergic antagonist like atropine (Table 1) are required, because c-Fos is not a perfect marker for activated neurons34 and injection of an agonist like carbachol is not physiological.

As described above, SLD neurons activated during PS are not cholinergic. We further recently showed that they are not GABAergic. Indeed, the small number of c-Fos-positive neurons expressing a specific marker of GABAergic neurons (glutamate decarboxylase, GAD, the enzyme of synthesis of GABA (gamma-aminobutyric acid)) in the SLD did not increase in rats displaying a PS rebound compared to control or PS deprived animals.35 Lu et al.36 reported the presence in the SLD of neurons expressing a specific marker of glutamatergic neurons (vesicular glutamate transporter 2, vGlut2). Our recent preliminary results further showed that most of the c-Fos-labeled neurons localized in the SLD after PS recovery express vGlut2.37 Altogether, these results indicate that the SLD neurons triggering PS are glutamatergic.

A number of results further indicate that SLD PS-on glutamatergic neurons generate muscle atonia and sensory inhibition via descending medullary projections to GABA/glycinergic neurons. Indeed, it has been shown that the SLD sends direct efferent projections to glycinergic neurons in the ventral (GiV) and alpha (Gia) gigantocellular nuclei (corresponding to the cat magnocellular reticular nucleus, Mc) and the nucleus raphe magnus (RMg). Further, glutamate release in the GiV and Gia increases specifically during PS.38 In addition, injection of non-N-methyl D-aspartate (NMDA) NMDA glutamate agonists in these nuclei suppresses muscle tone while an increased tonus is seen during PS in cats with Gia and GiV cytotoxic lesion.39, 40 Besides, glycinergic neurons of the Gia, GiV and RMg express c-Fos after induction of PS by bicuculline (Bic, a GABAa antagonist) injection in the SLD.24 Further, glycinergic neurons of these structures monosynaptically project to lumbar spinal motoneurons41 but also to the superficial dorsal horn involved in sensory processing.42 These results combined with others showing that sensory inputs are decreased during PS43 suggest that Gia, GiV and RMg glycinergic neurons not only hyperpolarize motoneurons but also dorsal horn sensory neurons. It is likely that these neurons are also GABAergic since a large majority of the c-Fos-labeled neurons localized in these nuclei after 3:00 h of PS recovery following 72:00 h of PS deprivation express the mRNA of the enzyme of synthesis of GABA (GAD67mRNA).35 The role of these neurons has been recently challenged by results showing that some SLD neurons directly project to the spinal cord and that neurotoxic lesions of the ventral medulla have no effect on PS atonia.36 It was further reported that inactivation in mice of glycinergic and GABAergic neurons located in the GiV has no effect on muscle tone during PS.44 However, it was not determined whether the SLD neurons projecting to the spinal cord express c-Fos after PS recovery.36 Also, the lesions were rostral to the GiV and the inactivations were restricted to the medial GiV. In addition, it has been shown in cats using antidromic activation that peri-LCα PS-on neurons directly project to the medulla but not to the spinal cord, whereas peri-LCα neurons with a firing rate unrelated to PS display spinal cord projections.15 Surprisingly, it has recently been shown that co-application by microdialysis of bicuculline and strychnine (respectively GABAa and glycine antagonists) (Table 1) in the trigeminal nucleus induced no effect on PS atonia.45 However, negative results obtained with microdialysis should be interpreted with caution.46 Further, the same authors recently showed that combined microdialysis of bicuculline, strychnine and phaclophen (a GABAb antagonist) in the trigeminal nucleus restored muscle tone during PS.47 The latter data support previous results indicating that the pre-motoneurons responsible for muscle atonia of PS are localized in the GiV and co-release GABA and glycine.

By early 2000, we observed that a long-lasting PS-like episode can be pharmacologically induced with a short latency in head-restrained unanesthetized rats by iontophoretic applications of bicuculline or gabazine, two GABAa receptor antagonists, specifically into the SLD. Our results have been reproduced in freely moving rats48, 49 and in cats with pressure injection of bicuculline.50, 51 In the head-restrained rat, we also recorded neurons within the SLD specifically active during PS and activated following bicuculline or gabazine iontophoresis.52 Taken together, our data indicate that the onset of SLD PS-on neurons is mainly due to the removal during PS of a tonic GABAergic tone present during W and SWS. It is likely that such a strong tonic GABAergic inhibition is necessary to preclude in healthy subjects the occurrence of sleep onset REM sleep (SOREMS) and cataplexy. Combining retrograde tracing with cholera toxin b subunit (CTb) injected in the SLD and GAD immunostaining, we then identified neurons at the origin of the GABAergic innervation. These neurons were localized within the pontine (including SLD itself) and the dorsal deep mesencephalic reticular nuclei (dDpMe) and to a minor extent in distant areas.53 Supporting the contribution of local GABAergic neurons in the inhibition of PS-on neurons during SWS and W, a significant increase in PS is produced by administration of antisense oligonucleotides against GAD67mRNA targeted to the cat nucleus pontis oralis including the SLD.50 In addition, in rats, the number of GABAergic neurons expressing c-Fos in the rostral pontine reticular nucleus decreased following PS rebound, suggesting they are active during W and SWS and inactive during PS.54 However, we recently demonstrated that the ventrolateral part of the periaqueductal gray (vlPAG) and the dDPMe are the only ponto-medullary structures containing a large number of c-Fos-positive neurons expressing GAD67mRNA after 72h of PS deprivation.35 Furthermore, injection of muscimol in the vlPAG and/or the dDpMe induces a strong increase in PS quantities in cats*55, 56, 57 and rats.*35, 52 These congruent experimental data lead us to propose that GABAergic neurons within the vlPAG and the dDpMe are gating PS during W and SWS by tonically inhibiting PS-on neurons from the SLD. By necessity, these GABAergic neurons would be silent during PS, therefore called PS-off neurons.

In cats, the microdialysis administration of kainic acid, a glutamate agonist in the peri-LCα induces a PS-like state.58 We reproduced these experiments in rats with iontophoretic application of kainic acid in the SLD. We further showed that kainic acid (Table 1) induces an activation of PS-on neurons.24 Further, application of kynurenate, a broad glutamate antagonist, reversed the PS-like state induced by bicuculline application in the SLD.24 These results suggest that PS-on neurons are under a permanent glutamatergic excitation throughout the sleep-waking cycle, unmasked at the onset of PS by the removal of a tonic GABAergic input. It remains to be demonstrated whether this glutamatergic input is stable across all states or increases at the onset of PS. The presence of a permanent glutamatergic input on SLD neurons is not surprising since it is well accepted that all types of neurons permanently receive excitatory and inhibitory inputs. The best candidate structure for containing the glutamatergic neurons permanently activating SLD PS-on neurons is the lateral and ventrolateral PAG. Indeed, we observed that numerous non-GABAergic neurons in these two structures, both known to contain glutamatergic neurons, project to the SLD.*53, 59 In addition, although it was established by Jouvet7 that structures responsible for PS are restricted to the brainstem, numerous non-GABAergic neurons projecting to the SLD located in the primary motor area of the frontal cortex, the bed nucleus of the stria terminalis or the central nucleus of the amygdala could also use glutamate as a neurotransmitter and contribute to the activation of the SLD PS-on neurons during PS.53 Further supporting a role of the central nucleus of the amygdala in PS control, it has been shown with muscimol and bicuculline applications that this nucleus promotes PS and PS-related phenomena.60

A major achievement in the identification of the mechanisms controlling PS was the finding that serotonergic neurons from the raphe nuclei and noradrenergic neurons from the LC cease firing specifically during PS, i.e., show a PS-off firing activity, reciprocal to that of PS-on neurons.61, 62, 63, 64 Later, it has been shown that histaminergic neurons from the tuberomamillary nucleus and hypocretinergic neurons from the perifornical hypothalamic area also depict a PS-off firing activity.65, 66, *67, 68, 69, 70 These electrophysiological data were the basis for a well-accepted hypothesis suggesting that PS onset is gated by reciprocal inhibitory interactions between PS-on and PS-off monoaminergic neurons.15, 61, 71 Supporting this neuronal model, drugs enhancing serotonin and noradrenergic transmission (monoamine oxidase inhibitors and serotonin and norepinephrine reuptake blockers) specifically suppress PS.72, 73, 74 Further, applications of noradrenaline, adrenaline or benoxathian (α2 adrenergic agonist) into the peri-LCα inhibit PS but that of serotonin has no effect.75, 76, 77 In addition, noradrenaline inhibits the non-cholinergic PS-on neurons via α2-adrenoceptors but has no effect on the putative cholinergic PS-on neurons from the peri-LCα while serotonin has no effect on both types of neurons.18

Importantly, our recent data combining a marker of noradrenergic neurons (tyrosine hydroxylase, TH) and c-Fos staining after PS deprivation and recovery suggest that it is unlikely that the LC noradrenergic neurons are involved in the inhibition of PS, particularly during its deprivation. Indeed, the LC noradrenergic neurons do not display c-Fos after 72h of PS deprivation. Furthermore, Boissard et al. found no projection from the LC to the SLD.53 Nevertheless, a substantial number of noradrenergic neurons from A1 and A2 cell groups display c-Fos after PS deprivation, indicating that noradrenergic neurons from these medullary cell groups might contribute to PS inhibition.78 In summary, it is clear that noradrenaline and serotonin inhibit PS but the targeted neurons remain to be identified. We propose that rather than inhibiting SLD PS-on neurons, noradrenaline and serotonin might inhibit PS by means of a tonic excitation of the dDPMe and vlPAG PS-off GABAergic neurons. Partly supporting this hypothesis, noradrenaline but not serotonin application in the dDPMe was shown to increase W and decrease SWS and PS.57

According to the classical “reciprocal interaction” model, the inactivation of monoaminergic (PS-off) neurons at PS onset is due to a powerful inhibitory drive originating from activated cholinergic PS-on cells from peri-LCα.15, 71 However, peri-LCα neurons do not send efferent projections to monoaminergic nuclei, and acetylcholine excites noradrenergic LC neurons and is only weakly inhibitory on dorsal raphe nucleus (DRN) serotonergic neurons.79, 80 It has therefore been suggested that inhibitory amino acids such as GABA or glycine may be better candidates than acetylcholine.81, 82 To test this hypothesis, effects of iontophoretic applications of bicuculline and strychnine (a glycine antagonist) on the activity of LC and DRN cells were studied in the head-restrained rat. Bicuculline or strychnine, applied during SWS or PS when monoaminergic neurons are silent, induces a tonic firing in both types of neurons.83, 84, 85 During W, the antagonists produce a sustained increase in discharge rate compared to control. These results indicate the existence of tonic GABA and glycinergic inputs to the LC and DRN that are active during all vigilance states. Importantly, when the strychnine effect occurred during transitions between PS to W, the discharge rate further increased at W onset. In the same situation with bicuculline, the discharge rate remains unchanged at the transition between PS and W. These distinctive electrophysiological responses strongly suggest that an increased GABA release is responsible for the PS-selective inactivation of monoaminergic neurons. This hypothesis is well supported by microdialysis experiments in cats measuring a significant increase in GABA release in the DRN and LC during PS as compared to W and SWS but no detectable changes in glycine concentration.86, 87 It seems therefore that monoaminergic cells are under a tonic GABAergic inhibition gradually increasing from W to PS. GABAergic neurons located in the ventrolateral preoptic nucleus (VLPO) and the extended VLPO have been involved in this inhibition during SWS and PS, respectively.88 However, although we did also observe numerous c-Fos-positive neurons in the extended VLPO after PS rebound, they were not projecting to the LC.89 Further, it is likely that the PS-related inhibition involved primarily GABAergic neurons within the brainstem, directly projecting to monoaminergic neurons and “turned on” specifically at the onset of and during PS. Indeed, PS-like episodes spontaneously occurring in pontine cats and when pharmacologically induced by carbachol in decerebrate cats, are still associated with the silencing of serotonergic neurons.*90, 91

By combining retrograde tracing with CTb and GAD immunohistochemistry in rats, we found that the LC and DRN receive GABAergic inputs from neurons located in a large number of distant regions from the forebrain to medulla.85, 92 Two brainstem areas contained numerous GABAergic neurons projecting both to the DRN and LC, and were thus candidates for mediating the PS-related inhibition of monoaminergic neurons: the vlPAG and the dorsal paragigantocellular reticular nucleus (DPGi).85, 92 We then demonstrated by using c-Fos that both nuclei contain numerous LC-projecting neurons selectively activated during PS rebound following PS deprivation.89, 93 Since the DPGi has not previously been considered in sleep mechanisms, we studied the firing activity of the DPGi neurons across the sleep–wake cycle in head-restrained rats.94 In full agreement with our functional data using c-Fos, we found that the DPGi contains numerous PS-on neurons that are silent during W and SWS and fire tonically during PS. The PS-on neurons start discharging approximately 15 s before PS onset and become silent around 10 s before EEG signs of arousal. Further, local application of bicuculline blocked the DPGi-evoked inhibition of LC neurons95 and electrical DPGi stimulation induces an increase in PS quantities in rats.96 Taken together, these data highly suggest that the DPGi contains the neurons responsible for the inactivation of LC noradrenergic neurons during PS.94 A contribution from the vlPAG in this inhibitory mechanism is also likely. In rat brainstem slices, iontophoretic NMDA application in the vlPAG induced bicuculline-sensitive inhibitory postsynaptic potentials (IPSPs) in serotonergic neurons.97 Moreover, an increase in c-Fos/GAD immunoreactive neurons has been reported in the vlPAG after a PS rebound induced by deprivation in rats.30, *35 In summary, a large body of data indicates that GABAergic PS-on neurons localized in the vlPAG and the DPGi hyperpolarize the monoaminergic neurons during PS.

To localize brain areas activated during PS, we compared the distribution of c-Fos+ neurons in control rats, rats selectively deprived of PS for 72:00 h and rats allowed to recover from such deprivation.89, *98 Although it is likely that some populations of neurons active during PS do not express c-Fos or that c-Fos is increased in neurons activated by other factors related to the experimental condition, we believe that it is the easiest way to identify, at cellular level, populations of neurons potentially implicated in PS genesis. Supporting this claim, c-Fos+ cells with a number positively correlated to PS quantities were observed in structures such as SLD, vlPAG or DPGi, implicated in PS by other approaches (see above). More surprisingly, we observed a very large number of c-Fos+ cells in the mammillary and tuberal hypothalamus, including the zona incerta (ZI), perifornical (PeF) and lateral hypothalamic areas (LHA). A few experimental results support the notion that the hypothalamus contributes to PS regulation. Indeed, bilateral injections of muscimol in the cat mammillary and tuberal hypothalamus induce a drastic inhibition of PS.99 Further, neurons specifically active during PS were recorded in the tuberal hypothalamus of cats70, 100, 101 and head-restrained rats.69 By using double-immunostaining, we further showed that 74% of the neurons labeled for c-Fos, located in the tuberal hypothalamus after PS rebound, expressed GAD67. Our results also indicate that approximately 20% of these neurons co-contain the neuropeptide MCH.102 In contrast, almost none of the neighboring hypocretin (Hcrt or orexin) neurons were c-Fos+. Around 60% of all the MCH-immunoreactive neurons counted in the PeF, ZI and LHA were c-Fos+*98, 103 Further, it has been shown in head-restrained rats by means of juxtacellular labeling that the MCH and non-MCH GABAergic neurons fire quite exclusively during PS.104, 105 Importantly, the MCH neurons start to fire at the onset of PS while the non-MCH/GABAergic neurons start to fire before the onset of PS,105 indicating that non-MCH/GABAergic but not MCH neurons could play a role in PS induction. Nevertheless, rats receiving intracerebroventricular (ICV) administration of MCH showed a strong dose-dependent increase in PS and, to a minor extent, SWS quantities, due to an increased number of PS bouts.98 Further, subcutaneous injection of an MCH antagonist decreases SWS and PS quantities106 and mice with genetically inactivated MCH signaling exhibit altered vigilance state architecture and sleep homeostasis.107, 108 Thus, converging results indicate that MCH neurons and/or MCH signaling play a key role in PS regulation/homeostasis. Since MCH is primarily an inhibitory peptide,109 MCH neurons might promote PS by inhibiting the vlPAG/dDPMe PS-off GABAergic and the wake active aminergic and hypocretinergic neurons. Supporting this hypothesis, observations with electron microscopy indicate that MCH and Hcrt neurons are interconnected110, 111, 113, 114 and recent data from MCH-receptor 1 knockout mice demonstrated interactions between Hcrt and MCH systems, implying that MCH may exert an inhibitory influence on Hcrt signaling.112 In addition, GABA, which is present in MCH neurons, likely contributes to the cessation of activity of Hcrt neurons during SWS and PS since bicuculline application in the posterior hypothalamus (PH) induced W and c-Fos expression in hypocretinergic neurons.69, 113 Another possibility is that the Hcrt and the aminergic neurons are hyperpolarized by the non-MCH GABAergic PS-on neurons in particular at the onset of PS.

As described above, most of the populations of neurons responsible for PS control were identified, by means of c-Fos labeling induced by PS deprivation and PS hypersomnia. In the future, it will be important to employ additional experimental approaches to fully determine the role of these neurons, including tract-tracing, single unit recordings, inactivations and activations by genetic or pharmacological tools. Furthermore, several regions that contain a large number of c-Fos-labeled neurons require additional studies, including the lateral paragigantocellular nucleus, the lateral parabrachial nucleus, the nucleus raphe obscurus and the dorsal PAG.31

The observation that PS episodes in the rat start from SWS after a relatively long intermediate state during which the EEG displays a mix of spindles and theta activity, and then terminate abruptly by a short microarousal114 deserves further attention. These findings suggest that different mechanisms are responsible for the entrance in and the exit from PS. Altogether, these characteristics, as well as our current knowledge of the neuronal network, lead us to propose an updated model of the mechanisms controlling PS onset and maintenance.

PS onset would be due to the activation of glutamatergic PS-on neurons in the SLD. During W and SWS, the activity of these PS-on neurons would be inhibited by a tonic inhibitory GABAergic tone originating from PS-off neurons localized in the vlPAG and the dDpMe. These PS-off neurons would be activated during W by the Hcrt and the aminergic neurons. The onset of PS would be due to the activation by intrinsic “clock like” mechanisms of PS-on MCH/GABAergic hypothalamic neurons and PS-on GABAergic neurons localized in the DPGi and vlPAG. These neurons would inactivate the PS-off GABAergic neurons and the aminergic and Hcrt waking neurons. The disinhibited ascending SLD PS-on neurons would in turn induce cortical activation via their projections to intralaminar thalamic relay neurons in collaboration with W/PS-on cholinergic and glutamatergic neurons from the Ldt and pedunculopontine (PPT), mesencephalic and pontine reticular nuclei and the basal forebrain. Descending PS-on SLD neurons would induce muscle atonia and sensory inhibition via their excitatory projections to glycinergic neurons localized in the Gia and GiV reticular nuclei and the RMg. The exit from PS would be due to the activation of waking systems. The waking systems would inhibit the hypothalamic MCH/GABAergic and the brainstem GABAergic PS-on neurons.

RBD is characterized by the acting out of dreams that are vivid, intense, and violent. Dream-enacting behaviors include talking, yelling, punching, kicking, sitting, jumping from bed, arm flailing, and grabbing. The person may be awakened or may wake up spontaneously during the acting and vividly recall the dream that corresponds to the physical activity. RBD is usually seen in middle-aged to elderly men.115 The disorder may occur in association with various degenerative neurological conditions such as Parkinson disease (PD), multiple system atrophy (MSA) and dementia with Lewy bodies (DLB).116

RBD often precedes the development of these neurodegenerative diseases by several years. It has been reported that up to 65% of patients diagnosed with RBD subsequently developed Parkinson disease within an average time of 12–13 years from the onset of RBD symptoms. The prevalence of RBD is of 46–58% in Parkinson disease, 50–80% in dementia with DLB and 90–100% of MSA patients.116 A mild form of RBD has also been identified in narcoleptics. An acute form may also occur during alcohol or sedative-hypnotic withdrawal, tricyclic antidepressant (such as imipramine), or serotonin reuptake inhibitor use (such as fluoxetine, sertraline, or paroxetine) or other types of antidepressants (mirtazapine). Clonazepam, a benzodiazepine, is highly effective in the treatment of RBD, relieving symptoms in nearly 90% of patients with little evidence of tolerance or abuse.116 It has been further showed that clonazepam decreases the phasic, but not the tonic, electromyographic activity in the submentalis muscle during REM sleep.117 Melatonin is also effective in some patients with RBD.118 Its beneficial effect occurs within the first week of treatment. Unlike clonazepam, melatonin decreases the tonic, but not the phasic, electromyographic activity in the submentalis muscle in subjects with RBD.116

Several studies indicate that it is unlikely that RBD is due to a dysfunction of the dopaminergic nigrostriatal system. The strongest arguments are that RBD does not occur in about half of the PD patients and the use of dopaminergic agents usually does not improve RBD. In neurodegenerative diseases where RBD is frequent, neuronal cell loss was observed in the brainstem structures modulating PS, like the locus subcoeruleus, the pedunculopontine nucleus and the gigantocellular reticular nucleus, and also in their rostral afferents, especially the amygdala.116 In cats and rats, electrolytic and neurochemical lesions limited to the SLD eliminate the tonic muscle atonia and induce phasic muscle activity during PS. The phasic events include large limb twitches, locomotion, fear, attack and defensive behaviors.36, 119, 120, 121, 122 Importantly, larger lesions induce a decrease in the total quantities of PS5, 7, 36 whereas RBD patients display normal quantities of PS.123 Selective experimental lesions in the ventromedial medullary reticular nuclei have been also reported to induce a decrease of atonia during PS with an increase of phasic events.39 From these and our own experimental data, we propose that RBD in patients without atonia during PS could be due to a lesion of a sub-population of PS-on glutamatergic neurons of the SLD responsible of inducing muscle atonia via their descending projections to the premotor GABA/glycinergic neurons of the GiV. It implies that PS-on neurons of the SLD are divided in at least two subpopulations, one descending responsible for muscle atonia and the other one inducing the state of PS itself and EEG activation (Fig. 3). Data obtained in cats support the existence of these two populations of SLD PS-on cells (see above) but they have not been identified in rats. If these two populations exist, it remains to be discovered why only the descending SLD neurons would be destroyed in RBD patients. In any case, RBD patients should not have a large lesion of the SLD and surrounding nuclei since they do not display a decrease in total PS amount. Another possibility is that SLD neurons are intact and the premotor GABA/glycinergic neurons of the GiV are damaged. This better fits with the fact that only the atonia is lost in RBD and not the state of PS per se (Fig. 3). Finally, neurons located in the ventral mesencephalic reticular formation could also be implicated since neurochemical lesions in this area in cats increased muscle tone and phasic muscle activity in REM sleep.124

Another puzzling issue is the fact that some RBD patients show intense phasic motor activation without a lost of muscle atonia. These patients could have no lesion of the SLD or its GiV relay. They could present an increase in phasic activation of the motoneurons either due to an increase in excitatory phasic inputs or a decrease in phasic inhibitory inputs or both. The presence during PS of phasic glutamate excitatory and glycine/GABA inhibitory inputs on motoneurons superimposed on a tonic inhibitory input is supported by results showing an increased number of excitatory and inhibitory postsynaptic potentials during muscle twitches and the respective decrease and increase of twitches induced by the application on motoneurons of glutamate and glycine or GABA antagonists.45, 125, 126 The origin of these phasic inputs is not known. However, it is likely that they are due to the activation of the classical motor pathways arising from the glutamatergic pyramidal cells of the motor cortex directly projecting to motoneurons or indirectly via glutamatergic and GABA/glycine pre-motoneurons located in pontine and medullary reticular nuclei and intermediate spinal cord.127 An increase of excitatory phasic inputs cannot be directly attributed to a degeneration of neurons. Therefore, it is more likely that phasic inhibitory systems have been destroyed in these patients.

The RBD reported in narcoleptic patients is likely due to the absence of hypocretin although it cannot be completely ruled out that the SLD-GiV atonia pathway is lesioned in these patients. One possibility is that in normal conditions, Hcrt neurons excite the SLD-GiV pathway during PS in particular during the muscle twitches induced by a phasic glutamatergic excitation of the motoneurons (Fig. 4). Two results support this hypothesis. First, although Hcrt neurons are mainly active during active waking, they display bursts of activity during the twitches of PS.68 Second, application of hypocretin in the SLD region induces PS with atonia.128

Narcolepsy-cataplexy is characterized by two major symptoms, i.e., excessive daytime sleepiness (EDS) and cataplexy and two auxiliary symptoms, hypnagogic hallucinations and sleep paralysis. EDS occurs daily and is characterized by sleep episodes with a premature onset of REM sleep. A sudden drop of muscle tone triggered by emotional factors, most often positive, characterizes cataplexy. It can affect all striated muscles or be limited to facial muscles or to the upper or lower limbs. The monosynaptic H-reflex is suppressed, like during PS. Patients remain fully conscious during cataplexy.129 All these symptoms suggest that PS is disinhibited in narcoleptic patients. It has then been shown that disruption of type 2 hypocretin receptor induces narcolepsy in dogs and mice.130, 131 No mutation was found in human narcoleptics.132 Instead, a marked reduction in the quantities of the peptide Hcrt 1 was found in their cerebrospinal fluid and a disappearance of Hcrt staining was observed in the hypothalamus of post-mortem brain tissues.132 Importantly, it has been shown that Hcrt neurons are specifically active during W and increase their activity during muscle activation.*67, 68 They are silent during SWS or PS, except during phasic twitches during which they can fire in bursts. Interestingly, they start to fire several seconds before the onset of W at the end of PS episodes.104 Hcrt neurons, like the aminergic neurons, send projections throughout the brain, from the olfactory bulb, cerebral cortex, thalamus to the brainstem and the spinal cord.133, 134 It is therefore difficult to determine, based solely on the projections of the Hcrt neurons, which missing pathway(s) is (are) responsible for the four symptoms of narcolepsy. It has been proposed that the absence of dense Hcrt projections to the histaminergic and noradrenergic LC neurons might be responsible for narcolepsy symptoms.135 Indeed, ICV administration or local injection of Hcrt in the noradrenergic LC or the histaminergic tuberomammillary nucleus (TMN) neurons induce W and inhibits PS.136, 137, 138 Further, administration of selective noradrenaline reuptake inhibitors and alpha-1 adrenergic agonists specifically suppress cataplexy.129 Besides, the absence of the hypocretin input on serotonin neurons could also play a role, since serotoninergic reuptake blockers are effective to treat cataplexy, at least in humans.129 Finally, a missing hypocretin projection to GABAergic PS-off neurons might be implicated, since the most recent treatment for cataplexy, namely gamma-hydroxybutyrate (GHB), may act through increased GABAB transmission.139 This is an attractive hypothesis in view of our recent finding that GABAergic PS-off neurons localized in the vlPAG/dDPMe region gate the onset of PS by means of their tonic inhibition of the SLD neurons during W and SWS. Further, we found that inactivating neurons in the vlPAG/dDPMe region by means of muscimol not only induced an increase in PS quantities but also an increase in SOREMS.35 It thus can be hypothesized that, during emotions in healthy subjects, there is a phasic increase in hypocretin release on GABAergic PS-off neurons. Since the hypocretin neurons also project to aminergic neurons, they would excite these neurons, which in turn would reinforce the activation of the GABAergic PS-off neurons via direct projections (Fig. 1). The phasic increase of inhibition by GABAergic PS-off neurons would counterbalance an increased glutamatergic excitation of SLD neurons arising in the central amygdala (Fig. 5). We indeed demonstrated that non-GABAergic neurons of the central amygdala project to the SLD.53 Further, neurons increasing their activity during W and/or PS or prior to and during cataplexy were recorded in the central amygdala.140, 141 In short, cataplexy would be due to a phasic activation of SLD neurons by central amygdala neurons, normally counterbalanced by an increased phasic GABAergic inhibition from vlPAG/dDPMe neurons. Importantly, our hypothesis implies that the projection from central amygdala neurons is specific to the PS-on SLD neurons responsible for muscle atonia projecting to GABA/glycinergic pre-motoneurons and not to those responsible for PS itself or the EEG activation. Indeed, in such a case, a full PS episode would be induced. Cataplexy would be inhibited in narcoleptic patients treated with serotonin and noradrenergic reuptake blockers or alpha-1 adrenergic agonists by means of an increased excitation of the GABAergic PS-off neurons. EDS, sleep paralysis and hypnagogic hallucinations suggest that the Hcrt neurons might inhibit and delay the onset of PS and strongly contribute to the abrupt stop of PS episodes, in line with their increase in firing before the end of a PS episode.104 In their absence, the onset of PS could occur more quickly and the end of PS might take more time, leading to hypnagogic hallucinations and sleep paralysis, respectively. The inhibition of PS by hypocretin neurons would be direct by means of excitation of the PS-off GABAergic and aminergic neurons but also indirect via direct excitation of the motor system including the motoneurons which express hypocretin type 2 receptors.142

Practice points

  • 1)

    PS is induced by glutamatergic neurons localized in the SLD.

  • 2)

    SLD PS-on glutamatergic neurons are inhibited during waking and slow wave sleep by GABAergic PS-off neurons localized in the vlPAG/dDPMe region.

  • 3)

    PS is triggered by GABAergic neurons localized in the tuberal hypothalamus, vlPAG and the DPGi. These neurons inhibit the PS-off vlPAG/dDPMe GABAergic and the waking neurons at the onset of and during PS.

  • 4)

    MCH and GABA containing neurons localized in the tuberal hypothalamus are specifically active during PS. They might play a key role in the regulation of PS quantities via their projections to PS-off vlPAG/dDPMe GABAergic and the waking neurons.

  • 5)

    RBD might be due to degeneration of a sub-population of SLD neurons projecting to the GABA/glycinergic GiV pre-motoneurons. More likely, it might be due to destruction of the latter pre-motoneurons.

  • 6)

    The four main symptoms of narcolepsy are due to the lack of PS inhibition normally provided by the projections of Hcrt neurons on PS-off vlPAG/dDPMe GABAergic and the waking aminergic neurons.

  • 7)

    Cataplexy is due to the cunjunction of a lack of Hcrt excitation of PS-off vlPAG/dDPMe GABAergic neurons and a phasic glutamatergic excitation of descending SLD neurons by central amygdala neurons during emotion.

Research agenda

  • 1)

    To determine the exact role of MCH neurons in PS by means of methods such as optogenetics, transgenic mice and MCH antagonists.

  • 2)

    To demonstrate by means of local lesions, pharmacological or genetic inactivations that GABAergic PS-on neurons in the vlPAG and/or the DPGi trigger PS.

  • 3)

    To compare RBD symptoms with those obtained by lesions or inactivation of the SLD descending glutamatergic neurons and those of the glycinergic neurons in the GiV.

  • 4)

    To demonstrate by means of functional neuroanatomical studies, unit recordings combined with local pharmacology, lesions and inactivation that cataplexy is mainly due to the lack of Hcrt input on GABAergic PS-off neurons.

Section snippets

Acknowledgments

This work was supported by CNRS and University Claude Bernard of Lyon.

References (142)

  • C.J. Holmes et al.

    Importance of cholinergic, GABAergic, serotonergic and other neurons in the medial medullary reticular formation for sleep-wake states studied by cytotoxic lesions in the cat

    Neuroscience

    (1994)
  • J.C. Holstege et al.

    A glycinergic projection from the ventromedial lower brainstem to spinal motoneurons. An ultrastructural double labeling study in rat

    Brain Res

    (1991)
  • M.S. Pollock et al.

    Rapid eye movement sleep induction by microinjection of the GABA-A antagonist bicuculline into the dorsal subcoeruleus area of the rat

    Brain Res

    (2003)
  • J.P. Sastre et al.

    Importance of the ventrolateral region of the periaqueductal gray and adjacent tegmentum in the control of paradoxical sleep as studied by muscimol microinjections in the cat

    Neuroscience

    (1996)
  • L.D. Sanford et al.

    GABAergic regulation of the central nucleus of the amygdala: implications for sleep control

    Brain Res

    (2002)
  • D.J. McGinty et al.

    Dorsal raphe neurons: depression of firing during sleep in cats

    Brain Res

    (1976)
  • B.Y. Mileykovskiy et al.

    Behavioral correlates of activity in identified hypocretin/orexin neurons

    Neuron

    (2005)
  • B.E. Jones

    Paradoxical sleep and its chemical/structural substrates in the brain

    Neuroscience

    (1991)
  • D. Gervasoni et al.

    Effect of chronic treatment with milnacipran on sleep architecture in rats compared with paroxetine and imipramine

    Pharmacol Biochem Behav

    (2002)
  • P.G. Guyenet et al.

    ACh, substance P and met-enkephalin in the locus coeruleus: pharmacological evidence for independent sites of action

    Eur J Pharmacol

    (1979)
  • Y. Koyama et al.

    Mutual interactions among cholinergic, noradrenergic and serotonergic neurons studied by ionophoresis of these transmitters in rat brainstem nuclei

    Neuroscience

    (1993)
  • P.H. Luppi et al.

    Anatomical and electrophysiological evidence for a glycinergic inhibitory innervation of the rat locus coeruleus

    Neurosci Lett

    (1991)
  • B.E. Jones

    Noradrenergic locus coeruleus neurons: their distant connections and their relationship to neighboring (including cholinergic and GABAergic) neurons of the central gray and reticular formation

    Prog Brain Res

    (1991)
  • D. Nitz et al.

    GABA release in the locus coeruleus as a function of sleep/wake state

    Neuroscience

    (1997)
  • R. Goutagny et al.

    Role of the dorsal paragigantocellular reticular nucleus in paradoxical (rapid eye movement) sleep generation: a combined electrophysiological and anatomical study in the rat

    Neuroscience

    (2008)
  • R. Liu et al.

    Serotonin 5-HT(2) receptors activate local GABA inhibitory inputs to serotonergic neurons of the dorsal raphe nucleus

    Brain Res

    (2000)
  • J.S. Lin et al.

    A critical role of the posterior hypothalamus in the mechanisms of wakefulness determined by microinjection of muscimol in freely moving cats

    Brain Res

    (1989)
  • Y. Koyama et al.

    State-dependent activity of neurons in the perifornical hypothalamic area during sleep and waking

    Neuroscience

    (2003)
  • M. Jouvet et al.

    Corrélations électromyographiques du sommeil chez le chat décortiqué et mésencéphalique chronique

    CR Soc Biol

    (1959)
  • M. Jouvet et al.

    Sur un stade d’activité électrique cérébrale rapide au cours du sommeil physiologique

    CR Seances Soc Biol

    (1959)
  • G. Carli et al.

    A study of pontine lesions suppressing deep sleep in the cat

    Arch Ital Biol

    (1965)
  • M. Jouvet

    Int J Neurol

    (1965)
  • M. Jouvet

    Recherches sur les structures nerveuses et les mécanismes responsables des différentes phases du sommeil physiologique

    Arch Ital Biol

    (1962)
  • R. George et al.

    A cholinergic mechanism in the brainstem reticular formation: induction of paradoxixal sleep

    Int J Neuropharmacol

    (1964)
  • H.A. Baghdoyan

    Cholinergic mechanisms regulating REM sleep

  • G. Vanni-Mercier et al.

    Mapping of cholinoceptive brainstem structures responsible for the generation of paradoxical sleep in the cat

    Arch Ital Biol

    (1989)
  • K. Sakai et al.

    C R Seances Acad Sci D

    (1979)
  • K. Sakai et al.

    State-specific neurones in the ponto-medullary reticular formation with special reference to the postural atonia during paradoxical sleep in the cat

  • K. Sakai

    Neurons responsible for paradoxical sleep

  • K. Sakai et al.

    Pontine structures and mechanisms involved in the generation of paradoxical (REM) sleep

    Arch Ital Biol

    (2001)
  • K. Sakai et al.

    Are there cholinergic and non-cholinergic paradoxical sleep-on neurones in the pons?

    Neuroreport

    (1996)
  • M. El Mansari et al.

    Responses of presumed cholinergic mesopontine tegmental neurons to carbachol microinjections in freely moving cats

    Exp Brain Res

    (1990)
  • M.H. Chase et al.

    Evidence that glycine mediates the postsynaptic potentials that inhibit lumbar motoneurons during the atonia of active sleep

    J Neurosci

    (1989)
  • P.H. Luppi et al.

    The nuclei of origin of monoaminergic, peptidergic, and cholinergic afferents to the cat nucleus reticularis magnocellularis: a double-labeling study with cholera toxin as a retrograde tracer

    J Comp Neurol

    (1988)
  • P. Fort et al.

    C R Acad Sci III

    (1990)
  • P. Fort et al.

    Glycine-immunoreactive neurones in the cat brain stem reticular formation

    Neuroreport

    (1993)
  • R. Boissard et al.

    The rat ponto-medullary network responsible for paradoxical sleep onset and maintenance: a combined microinjection and functional neuroanatomical study

    Eur J Neurosci

    (2002)
  • P. Bourgin et al.

    Induction of rapid eye movement sleep by carbachol infusion into the pontine reticular formation in the rat

    Neuroreport

    (1995)
  • S. Deurveilher et al.

    Pontine microinjection of carbachol does not reliably enhance paradoxical sleep in rats

    Sleep

    (1997)
  • K.J. Maloney et al.

    Differential c-Fos expression in cholinergic, monoaminergic, and GABAergic cell groups of the pontomesencephalic tegmentum after paradoxical sleep deprivation and recovery

    J Neurosci

    (1999)
  • Cited by (218)

    • Functional roles of REM sleep

      2023, Neuroscience Research
    • Peripheral Nervous System and Sleep

      2023, Encyclopedia of Sleep and Circadian Rhythms: Volume 1-6, Second Edition
    View all citing articles on Scopus

    The most important references are denoted by an asterisk.

    View full text