Research ArticleNeuroscience

ERK signaling pathway regulates sleep duration through activity-induced gene expression during wakefulness

See allHide authors and affiliations

Science Signaling  24 Jan 2017:
Vol. 10, Issue 463, eaai9219
DOI: 10.1126/scisignal.aai9219

ERK preps the brain for sleep

Why do we feel more tired and tend to sleep longer and more deeply after a period of sleep deprivation or after a day that was full of physical or mental activity? Findings from Mikhail et al. using mice and cultured cortical neurons suggest that sleep deprivation or activity during the waking hours stimulates gene expression dependent in part on the kinase ERK. Signaling by the ERK pathway increased the duration of sleep and affected sleep patterns. Genetic deletion of ERK in cortical neurons or intracranial injection of an ERK inhibitor reduced the amount of sleep and extended the duration of wakefulness in mice. The findings support a previously reported link between sleep and ERK in flies to provide insight into the mechanisms regulating sleep in mammals.


Wakefulness is accompanied by experience-dependent synaptic plasticity and an increase in activity-regulated gene transcription. Wake-induced genes are certainly markers of neuronal activity and may also directly regulate the duration of and need for sleep. We stimulated murine cortical cultures with the neuromodulatory signals that are known to control wakefulness in the brain and found that norepinephrine alone or a mixture of these neuromodulators induced activity-regulated gene transcription. Pharmacological inhibition of the various signaling pathways involved in the regulation of gene expression indicated that the extracellular signal–regulated kinase (ERK) pathway is the principal one mediating the effects of waking neuromodulators on gene expression. In mice, ERK phosphorylation in the cortex increased and decreased with wakefulness and sleep. Whole-body or cortical neuron–specific deletion of Erk1 or Erk2 significantly increased the duration of wakefulness in mice, and pharmacological inhibition of ERK phosphorylation decreased sleep duration and increased the duration of wakefulness bouts. Thus, this signaling pathway, which is highly conserved from Drosophila to mammals, is a key pathway that links waking experience–induced neuronal gene expression to sleep duration and quality.


Sleep is a behavior defined as a period of immobility and reduced ability to respond to external stimuli, which is homeostatically controlled (such that longer wakefulness periods lead to compensatory increase in sleep duration and intensity). Sleep is conserved across evolution (1). Caenorhabditis elegans, Drosophila, fish, reptiles, birds, and mammals satisfy all the behavioral criteria of sleep (immobility, reversibility, decreased response to stimuli, and homeostatically regulated) (27). The presence of sleep across evolution suggests that it must be regulated at a more basic cellular and molecular level than at the global complex brain structure. Evidence indicates that wakefulness is associated with increased transcription of activity-regulated (plasticity) genes. As demonstrated by several microarray experiments in mice and rats, the expression of one group of genes is consistently increased by extended wakefulness (typically synaptic activity-regulated genes such as Arc, Bdnf, Egr, and Homer1a), whereas another group (including Cirbp and Dbp) is decreased in abundance (811). Because the large majority of genes induced by prolonged wakefulness are activity-regulated genes [mostly immediate-early genes (IEGs) and stress-related genes], it is not clear whether their changes in expression merely represent neuronal activity or might directly have an effect on sleep. Nevertheless, a prominent hypothesis about sleep functions posits that sleep is “the price to pay” for waking experience–dependent plasticity (12).

Nonrapid eye movement (NREM) sleep is characterized by widespread synchronous firing of cortical neurons, and such synchronous activities occur also in isolated cortical islands, in cortical slices, under anesthesia, and even in dissociated cortical cultures (1320). Moreover, we and others have shown that primary cortical cultures can be stimulated by waking neuromodulators (including monoaminergic, glutamatergic, cholinergic, and hypocretinergic neurotransmitters or agonists) to induce tonic firing that invariably returns to the default synchronous burst firing 24 hours later (1821). Stimulated cultures show remarkably similar transcriptional changes, as do cortical tissues of mice subjected to 6 hours of sleep deprivation (18). Here, we used this simple in vitro model to dissect the molecular and cellular pathways leading to activity-regulated gene expression. We found that major activity-regulated genes were primarily induced through the extracellular signal–regulated kinase (ERK) pathway. We also found that whole-body or cortical neuron–specific deletion of Erk1 and Erk2 genes or inhibition of ERK phosphorylation in wild-type animals strongly decreased sleep duration. Thus, the ERK pathway, one of the most evolutionary conserved cellular pathways, critically controls the mammalian sleep duration, as previously evidenced in Drosophila.


Norepinephrine is the major neuromodulator for activity-regulated gene expression

The cell culture model provides a powerful tool to assess the transcriptional correlates of sleep- and wakefulness-like states. To assess the contribution of each neurotransmitter in our original waking cocktail separately, we stimulated cortical cultures with each (Fig. 1A) and assessed the expression of three major candidate genes: Homer1a, Dbp, and Arc. AMPA, norepinephrine (NE), histamine, dopamine (DA), and serotonin [5-hydroxytryptamine (5-HT)] alone had an inducing effect similar to (but less so than) that of the cocktail on Homer1a expression (Fig. 1B). A similar pattern of expression was also obtained for Arc (Fig. 1A). The expression of Dbp was assessed because it has been consistently found in several studies to decrease during wakefulness, as opposed to the above IEGs. The only neuromodulator that significantly decreased Dbp expression was NE (Fig. 1C).

Fig. 1 Candidate gene expression in cultured neurons after stimulation with each constituent of the waking cocktail.

(A to C) Three dishes from three independent cultures were stimulated for 3 hours with neuromodulators: NMDA, AMPA, kainic acid (KA), ibotenic acid (Ibot), serotonin (5-HT), histamine (Hist), DA, or NE (each 1 μM); carbachol (CCh; 10 μM); or orexin A (OrexA, 0.01 μM). Data are mean normalized mRNA expression + SD relative to sham-stimulated cultures (Control). (D to F) Gene expression after stimulation with sham (control) or NE (1 μM, 3 hours) in cultures pretreated with NE antagonist (yohimbine, prazosin, or propranolol, 10 μM for 30 min). Data are means + SD from six to nine independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 versus control, t tests with Holm correction.

Noradrenergic neurons of the locus coeruleus are the major source of the wakefulness neurotransmitter NE with diffuse projections throughout the brain and affect alertness, gene expression, and higher brain functions (2224). To understand how NE affects our candidate gene expression, each NE receptor type was blocked before NE addition to cortical cultures (Fig. 1, D to F). The α1-adrenergic antagonist prazosin did not prevent the effect of NE on Dbp expression, whereas either the α2-adrenergic antagonist yohimbine or the β-adrenergic antagonist propranolol consistently inhibited the effect of NE on gene expression (Fig. 1, E and F). α2-adrenergic and β-adrenergic receptors have opposite effects on adenylyl cyclase, with α2-adrenergic stimulation inhibiting and β-adrenergic stimulation activating adenylyl cyclase–mediated adenosine 3′,5′-monophosphate (cAMP) signaling. Therefore, we postulated that the observed effects of NE, as well as our cocktail, on gene expression might be mediated through another signaling pathway.

ERK signaling critically controls activity-regulated gene expression

A large number of signaling pathways converge to the final transcription of activity-regulated genes, among them our candidate genes (Arc, Homer1a, Dbp, and Bdnf). To find the candidate pathway (or pathways) responsible for the expression changes observed in these genes, we pretreated the cortical cultures for 30 to 60 min with pharmacological inhibitors to systematically block each of the major pathways: ERK1/2 (using U0126), protein kinase A (PKA; using H89), mitogen-activated protein kinase (MAPK) p38 (using SB203580), phosphatidylinositol 3-kinase (PI3K) (using LY294002), c-Jun N-terminal kinase (JNK; using JNK inhibitor II), calcium/calmodulin-dependent kinase II (CAMKII; using KN-93), Rho kinase (ROCK; using “Rho kinase inhibitor”), and actin polymerization (using latrunculin B). Two pathways emerged as major candidates: that mediated by PKA and the other by ERK. Blocking the PKA pathway with H89 reduced the expression of Bdnf, Homer1a, and Arc (Fig. 2, A to D). Blocking the ERK pathway with U01236 decreased the expression of Arc and Homer1a, increased that of Dbp, and also decreased that of Bdnf, although not significantly (Fig. 2C). Thus, the only pathway inhibitor that consistently affected gene expression, and the only one to alter that of Dbp, was the ERK pathway inhibitor. NE also activates the ERK pathway (25) independently of adenylyl cyclase but most probably by transactivation of epidermal growth factor receptors (EGFRs). To assess the activation of the ERK pathway by our cocktail, the phosphorylation of ERK was assessed in the cultured neurons. Cultures were either sham-treated or stimulated with the waking cocktail, and proteins were harvested 15 min later to quantify the phosphorylation of ERK1 and ERK2 (ERK1/2 or simply ERK hereafter). ERK phosphorylation was rapidly induced by the cocktail of neurotransmitters as opposed to sham (water) or no stimulation (control) (Fig. 2E). The ratio of phosphorylated ERK to total ERK (p-ERK/ERK) increased almost fourfold after cocktail stimulation as compared to sham-stimulated dishes, whereas dishes that were not stimulated showed a threefold decrease in the p-ERK/ERK ratio as compared to sham-stimulated dishes (Fig. 2F). These data suggest that wakefulness, similar to our cocktail of neurotransmitter, might activate ERK in vivo.

Fig. 2 Signaling pathway analysis.

(A to D) mRNA expression of Arc (A), Bdnf (B), Dbp (C), and Homer1a (D) in three dishes from three to four cortical cultures each pretreated with the indicated pharmacological inhibitors and then stimulated with cocktail relative to that in cocktail only–treated cultures. Data are means + SD from three dishes of three to four independent cultures. (E) Western blot of total ERK (top) and p-ERK1/2 (bottom) in cortical cultures 15 min after stimulation with cocktail, water, or nothing (control). (F) Quantification of the p-ERK/ERK ratio from (E) using ImageJ. Data are means + SD; n = 4; three independent Western blots. Blue horizontal lines connect conditions with significant differences. (G) Western blot of total ERK (top) and p-ERK1/2 (bottom) in the cortex of mice after 12 min of consolidated NREM sleep or wakefulness. (H) Quantification of the p-ERK/ERK ratio from (G) using ImageJ. Data are means + SD; n = 4; three independent Western blots. *P < 0.05, **P < 0.01, ***P < 0.001, t tests, with Holm correction in (A) to (D).

ERK phosphorylation increases and decreases with wakefulness and sleep

As we have previously reported (18), our in vitro model reliably mimics the transcriptional signature of in vivo sleep and wakefulness in the mouse cortex. To confirm our in vitro results, we assessed the in vivo phosphorylation of ERK after consolidated NREM sleep or wakefulness. We sacrificed mice after 12 min of NREM sleep at light onset [between zeitgeber time 0 (ZT0) and ZT1, first hour of the light period] and the corresponding controls that were maintained awake by gentle handling during the same interval. The amount of p-ERK was quantified in the cortex. ERK phosphorylation was strongly induced by 12 min of wakefulness as compared to the amount detected in those undergoing NREM sleep (Fig. 2G). The p-ERK/ERK ratio during NREM sleep was nearly threefold lower than it was during wakefulness (Fig. 2H), very similar to our cell culture results (Fig. 2F).

ERK1/2 loss-of-function mice have decreased amount of sleep

ERK phosphorylation may be simply a correlate of wakefulness without any direct regulation of sleep and wakefulness by the ERK pathway. To understand the role of the ERK pathway in sleep, we first investigated the effects of genetic deletion of Erk1 or Erk2. Erk1 deletion does not result in any observable phenotype, whereas Erk2 deletion results in embryonic lethality (26). Therefore, conditional Erk2 knockout (KO) mice (MAPKflox/flox) were bred into Emx1-Cre mice, resulting in deletion of Erk2 in cortical neurons. Emx1 is a marker of cortical neurons (expressed in 88% of cortical neurons) (27). These mice are viable despite important changes in brain structure (reduced cortical thickness) and behavioral deficits (28). Total wakefulness duration during the 24-hour baseline recordings was significantly increased in both Erk1 KO and Erk2 KO mice as compared to their wild-type littermates, whereas the amount of NREM sleep was significantly decreased in Erk2 KO mice (Fig. 3). Also, during the dark period after 6 hours of sleep deprivation, wakefulness was increased in both genotypes and NREM sleep was decreased (Fig. 3). These data suggest that the ERK pathway directly regulates NREM sleep duration.

Fig. 3 Effects of genetic deletion of Erk1 or Erk2 on vigilance state duration.

(Left) Hourly amounts of the various vigilance states (wakefulness, NREM sleep, or REM sleep) during a baseline period (first 24 hours; gray areas indicate the dark periods), a period of sleep deprivation (24 to 30 hours), and a period of recovery (30 to 48 hours) in Erk1 wild-type (WT; blue line, n = 6 mice), Erk1 KO (orange line, n = 7 mice), Erk2 WT (black line, n = 7 mice), and Erk2 KO (green line, n = 8 mice). (Right) Quantification of the corresponding duration of vigilance states during the 24-hour baseline and 12-hour dark recovery periods. Data are means + SD. *P < 0.05, two-way analysis of variance (ANOVA) with Holm correction for multiple tests.

Inhibition of ERK phosphorylation decreases the amount of sleep in mice

Because of the morphological and behavioral changes (in Erk2 KO mice) as well as the fact that germline inactivation might lead to compensation (between ERK1 and ERK2, as well as between ERK and other signaling pathways), we next continuously perfused a selective inhibitor of ERK phosphorylation (U0126) in the lateral ventricle of C57BL/6J mice and quantified vigilance states and the spontaneous locomotor activity. The inhibition of ERK phosphorylation significantly decreased NREM sleep, and wakefulness was increased by nearly 100 min (Fig. 4). The large increase in wakefulness was mainly distributed during the dark periods (Fig. 4). Wakefulness during the dark period was significantly increased in treated mice by almost 80 min (P < 0.01), and NREM sleep was decreased by the same amount (P < 0.01). REM sleep also showed a tendency to decrease in treated mice, and overall, nearly half of the total sleep during the dark period was lost. After a period of 6 hours of sleep deprivation, treated mice again showed significantly less NREM sleep (P < 0.001). The decrease in NREM sleep was even more marked during the dark period after sleep deprivation. More specifically, NREM sleep was decreased (P < 0.001) by 1 hour during the dark period after sleep deprivation (Fig. 4). ERK phosphorylation changes rapidly with vigilance states and spontaneous phosphorylation, and its inhibition can only be assessed in relation with the previous duration of sustained wakefulness or sleep. As an index of the inhibition of ERK phosphorylation, four mice under U0126 and four mice under dimethyl sulfoxide (DMSO) perfusion were sacrificed after 6 hours of sleep deprivation, and the amount of p-ERK was quantified by Western blotting (fig. S1). U0126 reduced total ERK phosphorylation by 29.30% (P = 0.06) and ERK2 phosphorylation by 37.85% (P < 0.05), indicating that even after 6 hours of imposed wakefulness, U0126 can reduce ERK phosphorylation (fig. S1).

Fig. 4 Effects of inhibition of ERK phosphorylation on vigilance state duration.

(Left) Average hourly amounts (mean ± SEM) of the various vigilance states (wakefulness, NREM sleep, or REM sleep) during a baseline period (first 24 hours; gray areas indicate the dark periods), a period of sleep deprivation (24 to 30 hours), and a recovery period (30 to 48 hours) in mice that were intraventricularly perfused with either U0126 (red, n = 9 mice) or solvent (DMSO; blue, n = 7 mice). (Right) Quantification of the corresponding amount of vigilance states during the 24-hour baseline (total), 12-hour baseline dark, 24-hour recovery (total), and 12-hour recovery periods. Data are means + SD. *P < 0.05, **P < 0.01, ***P < 0.001, two-way ANOVA with Holm correction for multiple tests.

To assess whether, in addition to sleep-wake distribution, the quality of sleep was also modified, we quantified the frequency and the contribution of different uninterrupted bouts of vigilance states in Erk1, Erk2 KO, and U0126-treated mice. Not only was the amount of sleep observed in the 24-hour time frame decreased in Erk2 KO and U0126-treated mice, but also wakefulness was less fragmented (Fig. 5). Sustained bouts of wakefulness were significantly increased in duration in Erk2 KO and U0126-treated mice (Fig. 5, A and B), whereas short (<16 s) wakefulness bouts were significantly less frequent in U0126-treated mice during baseline and also during recovery (Fig. 5, C and D). These differences again occurred mostly during the dark periods (active period). As a consequence of long sustained wakefulness bouts, the number of long bouts (>900 s) was significantly decreased in Erk2 KO and U0126-treated mice (Fig. 5, E and F). Together, these results indicate that the ERK pathway decreases the fragmentation of sleep and consolidates wakefulness.

Fig. 5 Effects of Erk1 and Erk2 deletion and inhibition of ERK phosphorylation on vigilance state quality.

(A and B) Duration of sustained wake bouts (>900 s) in Erk1 KO (n = 7) and Erk2 KO (n = 8) mice, their corresponding WT controls (n = 6 or 7, respectively), and U0126-treated (n = 9) or solvent (DMSO)–treated (n = 7) mice during the 24-hour baseline (A) and 18-hour recovery (B) periods. (C and D) Frequency of short wake bouts (<16 s) during the 24-hour baseline (C) and 18-hour recovery (D) periods. (E and F) Frequency of long wake bouts (>900 s) during the 24-hour baseline (E) and 18-hour recovery (F) periods. Data are means + SD from six to nine mice per condition. *P < 0.05, t test.

Inhibition of ERK phosphorylation increases spontaneous locomotor activity

Wakefulness and locomotor activity are often confounded. To assess whether the inhibition of ERK phosphorylation increases wakefulness, activity, or both, spontaneous locomotor activity of mice under U0126 perfusion was recorded by infrared sensors. U0126-treated mice were not only awake but also more active (Fig. 6A). Although the increase in activity is more evident during the dark (active) phase, even the short wakefulness bouts during the light period showed a significant increase in locomotor activity (Fig. 6, B and C). To understand whether this increase in activity is due to more locomotion or just to the increased time spent in wakefulness, we normalized the activity with the time spent in wakefulness, recorded by electroencephalogram (EEG). Treated mice not only were more awake but also displayed more than twofold increase in locomotor activity when awake (Fig. 6D).

Fig. 6 Spontaneous locomotor activity after inhibition of ERK phosphorylation.

(A) Mean locomotor activity over 2 days of baseline treatment with U0126 (red line, n = 9 mice) or DMSO (blue line, n = 7 mice). Dashed area around lines indicates ±1 SEM. (B and C) Quantification of total locomotor activity during the 12-hour light or dark period. (D) Ratio between locomotor activity counts per minute of EEG-recorded wakefulness. Data are means + SD from seven or nine mice, as indicated in (A). ***P < 0.001, t test.

ERK signaling controls vigilance state distribution but not sleep homeostasis

Changes in the quantity and quality of vigilance states affect the homeostatic regulation of sleep. Accordingly, longer wakefulness bouts and consolidated NREM sleep can be interpreted as an increase in sleep need and intensity. To evaluate the homeostatic regulation of sleep, we assessed the quantity and the time course of NREM sleep EEG slow wave activity [or delta (1 to 4 Hz) power density from spectral analysis]. Delta power was increased in U0126-treated mice during both baseline and recovery (P < 0.03 and P < 0.0004, respectively). We also performed a simulation of the time course of delta activity (see Materials and Methods) to explain the observed changes. The analysis revealed that the observed differences in delta power could be explained by changes in time spent in different vigilance states (fig. S2).


Our cell culture model provides a powerful tool to assess molecular and cellular correlates of sleep-wake regulation. One of our major discoveries (18) is that transcriptional changes induced by sleep deprivation can be reliably reproduced in vitro by chemical stimulation. We therefore followed the transcription of key candidate genes to investigate the molecular pathways activated by wakefulness and chemical stimulation. Surprisingly, NE alone could reproduce the effect of the whole cocktail on the expression of our candidate genes, confirming the essential role of NE in wakefulness and associated neuronal plasticity (29). To understand through which pathway NE stimulates gene expression, its receptors were pharmacologically blocked by antagonists. Blocking either α2-adrenoreceptors or β-adrenoreceptors inhibited the effects of NE on gene expression. Our results agree with several studies reporting the major role of NE in state-dependent gene expression (23, 30, 31). Rats depleted of NE (through locus coeruleus lesions) in the cortex show a reduction of known transcripts that increase during wakefulness, most of which are reported to be implicated in synaptic plasticity and response to cellular stress (23). Cirelli et al. (32) showed also that NE depletion reduced the homeostatic response to sleep deprivation. Together, these data confirm the importance of synaptic plasticity–related genes induced by NE during wakefulness in sleep-wake regulation.

Because α2- and β-adrenoreceptors have opposite effects on adenylyl cyclase, wherein α2 stimulation inhibits and β stimulation activates cAMP signaling, our findings indicate that another intracellular signaling pathway might be involved. Neurotransmitter-mediated stimulation of cortical cultures pretreated with relatively specific inhibitors of various signaling pathways revealed a key role for ERK phosphorylation by MAPK kinase (MEK). We also found similar changes in ERK phosphorylation in vivo as a function of sleep and wakefulness. Our results strongly suggest that transcriptional changes induced by NE and sleep deprivation occur through the ERK pathway, as already suggested (25). Our work is limited to NREM sleep, and no significant changes in REM sleep were observed. REM sleep also activates the ERK pathway, while the noradrenergic system is silenced during this state. REM sleep–induced activation of the ERK pathway depends mainly on cholinergic mechanisms and is therefore different from the effects of wakefulness (33). Obviously, the ERK pathway is not the only one activated by wakefulness or chemical stimulation. Accordingly, for instance, blocking the PKA pathway also affects the expression of several plasticity-related genes (such as Bdnf and Homer1a). It has also been shown that both ERK and PKA are required to induce changes in synaptic plasticity (34). Nevertheless, the ERK pathway was further investigated here in vivo because, as opposed to other pathways, the inhibition of its phosphorylation by U0126 affected the transcription of Dbp.

Our finding that the ERK pathway is activated both in vivo and in vitro by wakefulness and chemical stimulation confirms also the observations in Drosophila (35). Foltenyi et al. (35) found that ERK phosphorylation through EGFRs markedly increased the amount of sleep in Drosophila. Accordingly, we hypothesize that ERK phosphorylation by NE stimulation might result from transactivation of EGFR. Another study also implicated ERK phosphorylation in both the regulation of sleep and plasticity in Drosophila (36). We showed here that ERK loss of function (mainly loss of ERK2 via gene deletion) or inhibition of ERK phosphorylation in vivo reduced sleep duration and consolidated both sleep and wakefulness. More specifically, when Erk2 was deleted or when ERK phosphorylation was inhibited, longer uninterrupted wakefulness bouts were observed that might result in consolidated subsequent sleep. Changes in the homeostatic marker of sleep (EEG delta activity) are consistent with the distribution of sleep and wakefulness after inhibition of ERK phosphorylation. Given that phosphorylation is a fast process, the long-lasting changes in vigilance states observed here should be due to the downstream changes in gene transcription. ERK phosphorylation occurs quickly and reaches the maximum within 15 min (37). Therefore, sustained wakefulness is required to maintain ERK phosphorylation. This explains our finding that the effect of ERK phosphorylation inhibition was most evident during the dark periods when mice stay awake for long bouts. The phosphorylation occurs through a kinase cascade involving RAS and RAF activation, which, in turn, phosphorylate MEK (target of U0126) that phosphorylates ERK. Double-phosphorylated ERK activates MAPK-interacting kinases MNK1 and MNK2, ELK1, and ribosomal S6 kinase, leading to the phosphorylation of c-MYC and cAMP response element–binding protein (CREB) and the activation of target gene transcription. The role of phosphorylated CREB in sustained wakefulness, subsequent sleep, and memory is well documented (38, 39). Note that p-ERK also activates target protein translation by phosphorylating MNK, thus causing activation of elF4E. This downstream effector of ERK pathway has also been implicated in sleep and plasticity (40).

The fact that ERK inhibition, as opposed to other signaling pathways (including PKA), affects Dbp expression is intriguing. Detailed analysis of signaling pathways involved in synchronizing peripheral circadian rhythms found no evidence that the ERK pathway plays any role, whereas serum response factor–dependent actin dynamics (tested in our screening by inhibiting ROCK and actin polymerization without any effect) was the key pathway (41). The MAPK p38 (also inhibited in our experiments without any effect) was found to be a potential kinase of the circadian rhythm–associated protein Period (PER) that also plays a major role in stress response signaling (42). ERK phosphorylation is activated in the suprachiasmatic nucleus after a light pulse, and inhibition of this process by U0126 strongly reduces ERK phosphorylation and attenuates light entrainment of circadian rhythms without affecting clock-timing properties (43). These observations suggest that changes we report here for sleep duration are unlikely the result of the role of the ERK pathway in circadian rhythms. DBP is part of the positive feedback loop of the circadian molecular clock as opposed to PER and CRY, which are part of the negative feedback loop. Although the exact mechanism remains to be investigated, our results suggest either that activated ERK leads to the inhibition of Dbp transcription or that this inhibition occurs after activation of Per2.

Inhibiting ERK phosphorylation increased not only wakefulness but also the locomotor activity mainly during the active (dark) phase, independent of the time spent awake. Given that ERK phosphorylation can be induced by cellular stress, one could argue that the increase in ERK phosphorylation during wakefulness could be a correlate of stress response. However, several studies suggest that this hypothesis is unlikely. Mongrain et al. (44) showed that adrenalectomy that abolished increased corticosterone concentrations in three different inbred mouse strains markedly reduced the expression of many stress-related genes after sleep deprivation but not that of major activity-regulated genes, such as Homer1a. They also found that sleep deprivation in adrenalectomized mice still activated the ERK pathway, suggesting that, under these conditions, extended wakefulness per se more than the associated stress is responsible for changes in activity-regulated genes. In flies, oxidative stress or mechanical disturbances as well as waking induced by starvation did not result in ERK activation, as opposed to sleep deprivation (36).

Although U0126 used at 10 μM in vivo cannot completely inhibit ERK phosphorylation (45), the partial inhibition was still sufficient to increase wakefulness and to consolidate NREM sleep and wakefulness. Given the changes in candidate gene transcription and in vigilance states, we conclude that the accumulation of activity-regulated and sleep-related genes is principally mediated by ERK activation. Accordingly, inhibiting ERK phosphorylation reduces the accumulations of these transcripts, and therefore, longer consolidated wakefulness bouts are required to translate waking experience into sleep.

We have suggested that one of the sleep functions could be to protect the neuronal membrane homeostasis against overstimulation, as suggested by a large release of lysophospholipids after sleep deprivation or in vitro stimulation (18). Inhibition of the ERK pathway was shown to reduce the release of arachidonic acid (46). ERK has also been shown to phosphorylate and activate phospholipase A2 (47, 48). These findings suggest that ERK activation during extended wakefulness mediates the release of lysophospholipids, which, if not counterbalanced by sleep, will cause damage to cell membrane integrity.

Thus, our findings indicate that the ERK pathway, a highly evolutionary conserved signal transduction pathway from Drosophila to mammals, plays a major role in controlling plasticity and sleep-related gene transcription and translation and that this, together with other functions of ERK (metabolic stress, oxidative stress, and proliferation or apoptosis), regulates vigilance states.


Animal handling

C57BL/6J, MAPKflox/flox, and Emx1-Cre mice were purchased from The Jackson Laboratory. Erk1 KO mice were provided by B. Yalcin (University of Lausanne) and G. Pages (University of Nice Sophia Antipolis). All animals were on C57BL/6J genetic background and maintained under standard animal housing conditions with free access to food and water and a 12-hour light/12-hour dark cycle (lights on at 7:00 a.m.). All experiments were approved by the Vaud Cantonal Veterinary Office (Switzerland).

Primary cortical cultures

Cultures were prepared as detailed in (18). Briefly, embryonic day 17 to 18 mouse embryos were collected to prepare cortical cultures. Timed-pregnant mice were sacrificed by cervical dislocation, and embryos were removed and decapitated in phosphate-buffered saline at 4°C. Cortices from embryos were dissected and cut into small pieces in a dissection solution. After enzymatic digestion for 30 min with papain (10 U/ml; Roche 108014), cells were dissociated by 12 to 15 triturations using a glass pipette in Neurobasal medium (NBM) completed with B27 (Invitrogen), 0.5 mM GlutaMAX-I (Invitrogen), and penicillin-streptomycin antibiotics (1%) (complete NBM). After every four or five triturations, the cells in suspension were transferred to a separate tube. Isolated cells were centrifuged during 4 min at 150g and resuspended in 2 ml of complete NBM. One million to two million cells were harvested on petri dishes (35 mm diameter) previously coated with poly-l-lysine (0.1 mg/ml). Cultures were maintained in an incubator at 37°C with 5% CO2. Half of the medium was replaced every 7 days.

Stimulation protocol

The cocktail was prepared freshly with AMPA (1 μM, Sigma-Aldrich); noradrenaline (1 μM, Sigma-Aldrich); kainic acid (kainate), ibotenic acid, 5-HT, histamine dihydrochloride, DA hydrochloride, and N-methyl-d-aspartate (NMDA) (each 1 μM, Tocris Bioscience); carbachol (10 μM, Sigma-Aldrich); and orexin (0.01 μM, Sigma-Aldrich). Distilled water was added in sister cultures as control in all stimulation protocols. To avoid differences in gene expression caused by circadian rhythms, stimulations and RNA and protein extractions from different conditions were performed at the same circadian time. RNA was extracted 3 hours after stimulation and again 24 hours later (recovery).


For the NE study, we selectively blocked each receptor subtype with yohimbine, prazosin (each 10 μM, Sigma-Aldrich) or propranolol (10 μM, Merck Millipore) 30 min before stimulation with NE (1 μM, Sigma-Aldrich). For the pathway study, LY294002 (PI3K inhibitor, Cell Signaling Technology, 50 μM) or SB203580 (p38 inhibitor, Merck Millipore, 10 μM) was added 1 hour before simulation with the cocktail, and H89 (PKA inhibitor, Sigma-Aldrich, 10 μM), KN-93 (CAMKII inhibitor, Sigma-Aldrich, 0.5 μM), JNK inhibitor II (Merck Millipore, 10 μM), U0126 (ERK1/2 inhibitor, Cell Signaling Technology, 10 μM), latrunculin B (actin polymerization inhibitor, Merck Millipore, 0.3 μM), or ROCK inhibitor (Merck Millipore, 100 nM) was added 30 min before stimulation.

RNA extraction and analysis

RNA from cells was extracted with the RNeasy Mini Kit (Qiagen), and RNA from cortex was extracted with the RNeasy Lipid Midi Kit (Qiagen). All RNA samples were treated with deoxyribonuclease and conserved at −80°C. RNA quantities were assessed with a NanoDrop ND-1000 spectrophotometer, and the quality of RNA was controlled on Agilent 2100 Bioanalyzer chips. Three to 6 μg of RNA were obtained from cortical cultures, and 100 to 150 μg from cortices.

Gene expression analyses were performed as described by Hinard et al. (18). Briefly, to quantify relative RNA expression from cortical cultures or from mouse cortex, RNA was first reverse-transcribed: RNA (500 ng) was mixed with random hexamer (1 μg), deoxynucleotide triphosphate (1 μl of 10 mM), RNasin (1 μl of 40 U/μl), and SuperScript II (1 μl of 200 U/μl) (each from Invitrogen) completed to 20-μl volume with buffer. Denaturation was performed at 65°C for 5 min followed by reverse transcription by warming up the samples to 28°C for 10 min and then to 42°C for 60 min. The resulting complementary DNAs were diluted 1:10 in water and mixed with primers and probes of target genes by a robot (Tecan). Target genes were then amplified by real-time quantitative polymerase chain reaction in an ABI PRISM 7900HT detection system (in triplicate). A qBase v1.3.5 software was used to quantify the relative level of the RNA by the ΔCt method. Three housekeeping genes (EEF1a1, RPS9, and TBP) were measured to normalize relative expression levels by calculating their geometric means. Controls were made with a sample without SuperScript II (nonamplified control). In the cultures, the gene expression after stimulations was normalized using the mean of the control samples for each culture separately. Primers and TaqMan probes for candidate genes are provided in table S1.

Protein extraction, Western blotting, and quantification

To quantify the ratio of p-ERK/ERK, the mice were sacrificed after the experiment by cervical dislocation. Cortices were extracted and frozen in liquid nitrogen. Proteins were extracted using radioimmunoprecipitation assay (RIPA) buffer with a cocktail of protease inhibitor (Roche). Trituration in RIPA was done to homogenize the samples on ice. Samples were left on ice during 15 min and centrifuged at 4°C and 2000g for 10 min. Supernatants were collected and proteins were quantified using the BCA Protein Assay Kit (Thermo Scientific), and the absorbance was measured with Safire2 (Tecan). Protein (20 to 40 μg) was heated at 95°C for 5 min in the Laemmli buffer, loaded on a 12% SDS–polyacrylamide gel electrophoresis gel, and transferred onto a polyvinylidene difluoride membrane. Membranes were incubated with either ERK1/2 antibody or p-ERK1/2 primary antibody (each 1:2000, Cell Signaling Technology) for 2 hours followed by a secondary polyclonal goat anti-rabbit (1:2000, Dako) for 2 hours. Membranes were revealed with Amersham ECL Prime Western Blotting Detection Reagent (GE Healthcare) and Amersham Hyperfilm ECL (GE Healthcare) and developed with SRX-101A (Konica Minolta).

EEG recording and analysis

EEG and EMG (electromyogram) implantations were performed under deep anesthesia. Briefly, two EEG electrodes (frontal at 1.7 mm right from the midline and at 1.5 mm anterior to the bregma; parietal at 1.7 mm right of midline and at 1.0 mm anterior to lambda) and three anchor screws were placed on the skull. Two semirigid gold wires were used as EMG electrodes and were inserted into the neck muscle along the back of the skull. EEG and EMG wires were connected to a microconnector, cemented to the skull, and connected to recorders through a swivel. EEG and EMG signals were amplified, filtered, converted from analog to digital form (2 kHz), and down-sampled and stored at 200 Hz. Spontaneous locomotor activity was measured by infrared sensors (with ClockLab). Movements were quantified as counts per minute once every 5 min. For drug perfusion, micro-osmotic pumps from Alzet (model 1002) were prepared 1 day before the surgery following the instruction of the manufacturer and filled with U0126 (1.52 μg/μl) suspended in 40% DMSO in artificial cerebrospinal fluid (ACSF) or in 40% DMSO in ACSF as control. The pumps were placed in the back of the mice and connected to a cannula (3 to 5 mm; Brain Infusion Kit 1, Alzet) placed in the right lateral ventricle (1 mm lateral to midline; 0.3 mm anterior to lambda; 2.2-mm deep), and perfusion was made at a rate of 0.25 μl/hour.

Scoring of vigilance states was performed visually in 4-s epochs as described previously (49). Recordings included two baseline days, followed by 6 hours of sleep deprivation and 18 hours of recovery. A discrete Fourier transform analysis was performed on the EEG signal to calculate the EEG power densities. For delta power (1 to 4 Hz) calculation, baseline light and dark periods were divided into 12 and 6 intervals, respectively, with an equal number of NREM epochs. Light period of the recovery was divided into eight intervals. For sleep fragmentation and analysis, the frequency and the amount of the three different vigilance states were calculated according to Franken et al. (49). Simulation of the time course of the EEG delta power was performed on the basis of the distribution of sleep-wake epochs, as described by Franken et al. (50). Basically, the process S (EEG delta power) was considered to increase during wakefulness or REM sleep following an exponential function [St + 1 = UA − (UA − St)*edt/τi] and to decrease exponentially during NREM sleep [St + 1 = LA + (St − LA)*edt/τd] (time resolution of iteration dt = 4 s), with UA as an upper (UA = 282%) and UL as a lower (LA = 55%) asymptote with time constant of increase τi = 7.9 hours and decrease τd = 1.9 hours.

C57BL/6J mice aged 9 to 10 weeks were sleep-deprived for 6 hours at the beginning of the light period by gentle handling. Gentle handling consisted of cage tapping, introduction of paper towels in the cage, or approaching a plastic pipette next to the mouse as soon as sleeping behavior was observed. Sleep-deprived mice and undisturbed controls were euthanized by cervical dislocation. The brain and liver were collected and rapidly frozen on dry ice and stored at −80°C.


Fig. S1. In vivo inhibition of ERK phosphorylation after 6 hours of sleep deprivation.

Fig. S2. Time course of delta power during NREM sleep.

Table S1. Primers and probes for gene expression analyses.


Acknowledgments: We thank Y. Emmenegger for technical assistance. We thank P. Franken and A. Sousek (Center for Integrative Genomics, University of Lausanne, Switzerland) for help in data analysis and figure preparation, respectively. We thank B. Yalcin (Center for Integrative Genomics, University of Lausanne, Switzerland) for providing Erk1 KO mice under the authorization from G. Pages (Institute for Research on Cancer and Aging, Nice, France). Funding: This work was supported by the Swiss National Science Foundation (grant no. 146615 to M.T.) and the University of Lausanne. Author contributions: C.M. and M.T. designed all experiments. C.M., A.V., and S.J. performed the experiments. C.M. analyzed the results, prepared the figures, and helped M.T. in writing the manuscript. Competing interests: The authors declare that they have no competing interests.
View Abstract

Stay Connected to Science Signaling

Navigate This Article