Salubrinal, an inhibitor of protein synthesis, promotes deep slow wave sleep

Abstract

Previous work showed that sleep is associated with increased brain protein synthesis and that arrest of protein synthesis facilitates sleep. Arrest of protein synthesis is induced during the endoplasmic reticulum (ER) stress response, through phosphorylation of eukaryotic initiation factor 2α (p-eIF2α). We tested a hypothesis that elevation of p-eIF2α would facilitate sleep. We studied the effects of intracerebroventricular infusion of salubrinal (Salub), which increases p-eIF2α by inhibiting its dephosphorylation. Salub increased deep slow wave sleep by 255%, while reducing active waking by 49%. Delta power within non-rapid eye movement (NREM) sleep was increased, while power in the sigma, beta, and gamma bands during NREM was reduced. We found that Salub increased expression of p-eIF2α in the basal forebrain (BF) area, a sleep-wake regulatory brain region. Therefore, we quantified the p-eIF2α-immunolabeled neurons in the BF area; Salub administration increased the number of p-eIF2α-expressing noncholinergic neurons in the caudal BF. In addition, Salub also increased the intensity of p-eIF2α expression in both cholinergic and noncholinergic neurons, but this was more widespread among the noncholinergic neurons. Our findings support a hypothesis that sleep is facilitated by signals associated with the ER stress response.

much evidence suggests a link between sleep and brain protein synthesis. Sleep is associated with increased protein synthesis in several discrete brain regions as well as the whole cerebrum (30, 36). Sleep deprivation reduces the levels of certain proteins in rat basal forebrain (BF) (1) and hippocampus (15). Key components of translational control genes are preferentially expressed during sleep (7, 9). Some evidence suggests that regulation of sleep is coupled to the successful execution of protein synthesis, and that non-rapid eye movement (NREM) sleep is facilitated if protein synthesis is suppressed. Administration of protein synthesis inhibitors (PSIs) usually increases NREM sleep but reduces REM sleep (33). Recently, we reported (26) that local administration of the PSI anisomycin (Ani) induced site-specific facilitation of sleep states. NREM sleep was increased by Ani administration in the lateral preoptic area and REM sleep by administration in the perifornical hypothalamic areas of rats. The sleep state facilitated by PSI administration may depend on the predominant state of sleep normally regulated by the brain region exposed to PSI.

Arrest of protein synthesis in vivo may occur as a first line of defense against accumulation of unfolded or misfolded proteins in the endoplasmic reticulum (ER) lumen, identified as ER stress (16, 37). A current model posits that adaptive responses to ER stress are controlled by a signaling pathway led by release of the chaperone BiP/GRP78 (37, 46). An immediate adaptive response to ER stress is the phosphorylation at Ser51 of the α-subunit of eukaryotic initiation factor 2 (eIF2α), thereby blocking translation initiation and synthesis of certain classes of proteins, including membrane and secreted proteins important in brain function. The phosphorylation of eIF2α is therefore widely used as a marker of ER stress (35).

All components of the unfolded protein response (UPR) were found after 6 h of sleep deprivation in mouse neocortex, including increases in p-eIF2α as well as free BiP/GRP78 and phosphorylated protein kinase-like ER kinase (PERK), a key kinase that phosphorylates eIF2α (29). Sleep deprivation has also been shown to increase the expression of BiP in Drosophila (28, 39), white crowned sparrow (20), mice (25, 29, 41), and rats (8, 38, 42).

On basis of the findings summarized above, our study was conceived as follows. On one hand, sleep deprivation elicits the ER stress response; sleep deprivation is followed by homeostatic facilitation of sleep. On the other hand, sleep state facilitation is coupled to facilitation of brain protein synthesis. Therefore, we hypothesized that a component of the molecular pathways of the ER stress response regulating protein synthesis could also play a role in facilitation of sleep. Because induction of p-eIF2α is a key step in the ER stress response, we hypothesized that elevation of p-eIF2α would facilitate sleep.

This hypothesis was tested by administration of salubrinal (Salub), a small molecule that increases p-eIF2α by inhibiting its dephosphorylation (Ref. 4; see discussion). We administered Salub into the lateral ventricle and observed its effects on sleep. Since p-eIF2α immunolabeling was found to be more intense in the cholinergic BF area, a sleep-wake regulatory brain region, we also quantified the number of p-eIF2α-immunolabeled neurons in the BF area after Salub administration.

MATERIALS AND METHODS

Experimental subjects.

Twenty-six male Sprague-Dawley rats (250–300 g at time of surgery; Harlan) were kept in a 12:12-h light-dark cycle and provided with ad libitum access to food and water. Experiments were conducted in accordance with the National Research Council Guide for the Care and Use of Laboratory Animals. All procedures were approved by the Institutional Animal Care and Use Committee of the Department of Veterans Affairs of Greater Los Angeles Healthcare System.

Surgery.

Under aseptic conditions and anesthesia induced by intraperitoneal administration of a cocktail of ketamine + xylazine (80 mg + 8 mg/kg), rats were surgically prepared for polygraphic recording of sleep-wake parameters as previously described (27). Gold-plated stainless steel machine screws with soldered leads were threaded into the parietal and frontal bones for EEG recording; another screw electrode served as a ground. Insulated stainless steel wires were inserted into the dorsal neck muscles for electromyogram (EMG) recording. All electrodes were soldered to a standard plug. A 23-gauge guide cannula with a removable obdurator stylet was stereotaxically implanted unilaterally 1 mm above lateral ventricle (AP = −0.80 mm; DV = 3.6 mm; L = 1.4 mm from bregma) (31). The guide cannula and the plug with the electrodes were affixed on the skull with dental acrylic. Postoperative care included administration of the analgesic Buprenex (0.02 mg/kg im, twice daily for 2 days) and the application of a topical antibiotic around the incision twice daily for 3 days.

Experimental procedure.

After surgery, animals were housed in individual Plexiglas cages in an electrically shielded, sound-attenuated, temperature-controlled recording chamber (temperature 23 ± 2°C). Three days before the recordings began [postoperative day (POD) 5], cannula patency and localization were assessed by injecting the rats intracerebroventricularly with 500 ng of angiotensin II (ANG II; Sigma) in pyrogen-free saline; patency was confirmed if ANG II elicited a drinking response in <1 min (12). On POD 7, the obdurator stylet was removed from the guide cannula and replaced with a 27-gauge injector cannula that was fixed in place with dental acrylic. The injector cannula was connected by a low-dead-volume Teflon tubing (100 cm, EiCom) to a remote pump kept outside the recording chamber, permitting delivery of Salub or artificial cerebrospinal fluid (aCSF) without disturbing the rats. The rats were perfused with aCSF containing (in mM) 145 Na⁺, 2.7 K⁺, 1.0 Mg²⁺, 1.2 Ca²⁺, 1.5 Cl⁻, and 2 Na₂HPO₄ (pH 7.2) for at least 12 h before the studies began.

On POD 8, at least 18 h after the insertion of the injector cannula, the lateral ventricle of the rat was perfused beginning at zeitgeber time (ZT)12 for either 12 h (experiment 1) or 3 h (experiment 2) with either 100 μM Salub or aCSF at a flow rate of 20 nM/min (pumping speed 0.2 μl/min) by a peristaltic pump. For experiment 1, eight rats were perfused with both Salub and aCSF, in counterbalanced order, with an intervening day between the two treatments. Another set of six rats were perfused for 12 h with either 0.5% dimethyl sulfoxide (DMSO; Sigma) as vehicle control or aCSF on alternate days. Rats received only one treatment in 24 h. For experiment 2, rats were treated for 3 h with either aCSF (n = 6) or Salub (n = 6), after which they were euthanized for immunohistochemistry.

We used 100 μM Salub since this dose was shown to be nontoxic and conferred maximum cytoprotection (4). Salub (Calbiochem, La Jolla, CA) was initially dissolved in the vehicle DMSO to make a 20 mM stock solution, which was diluted with aCSF to 100 μM aliquots; the aliquots were kept at −20°C until use. The final concentration of DMSO was 0.5%.

Sleep recording and perfusion.

Recording of the sleep-wake parameters based on the EEG and EMG signals was performed as previously described (27). Briefly, EEG (band-pass filtered at 0.3–100 Hz), and EMG (10–300 Hz) were digitized at 128 Hz with a 1401 Plus interface and Spike2 software (Cambridge Electronic Design, Cambridge, UK). After completion of the recordings, under deep pentobarbital anesthesia, rats were transcardially perfused with 50 ml of PBS (pH 7.4) followed by 300 ml of 4% paraformaldehyde. The brains, after equilibrating in 30% sucrose, were cut into 40-μm-thick coronal sections on a freezing microtome. Sections were either stained for Nissl (cresyl violet) for verifying the implantation sites of the injector cannula (experiment 1) or immunohistochemically labeled for p-eIF2α and choline acetyltransferase (ChAT) (experiment 2). Data were used only from those rats in which the position of the injector cannula was histologically verified within the lateral ventricle and confirmed by drinking response to ANG II [experiment 1: n = 12 (7 rats for Salub and aCSF and 5 rats for aCSF and DMSO); experiment 2: n = 10 (5 rats each for Salub and aCSF)]. Four rats that did not show the drinking response were eliminated from the study.

Immunohistochemistry.

For immunohistochemistry, free-floating sections were treated with an antigen retrieval protocol that consisted of incubation in 0.5% sodium dodecyl sulfate (SDS) for 3 min followed by 10 min in preheated (80°C) 0.01 M sodium citrate buffer at pH 9.0. After rinsing with Tris-buffered saline (TBS) and blocking with 3% goat serum in TBS, sections were incubated in a cocktail of rabbit polyclonal anti-p-eIF2α (1:300, Stressgen, Ann Arbor, MI) and mouse monoclonal anti-ChAT (1:1,200; Chemicon, Temecula, CA) for 2 days at 4°C. Sections were then washed in TBS, blocked in goat serum, and incubated for 4 h in a mixture of goat anti-rabbit Alexa Fluor 555 and anti-mouse Alexa Fluor 488 (Invitrogen, Carlsbad, CA) at 1:500. Labeled sections were mounted and coverslipped with Vectastain (Vector Labs, Burlingame, CA) mounting medium. Immunohistochemistry was simultaneously performed on alternate sections through the BF from aCSF- and Salub-treated rats. Sections from the two groups were treated with aliquots from the same batch of antibodies.

Cell counting.

Neuronal counting was performed with the Neurolucida computer-aided plotting system (Microbrightfield) by a person blinded to the treatments of the animals. Neurons expressing p-eIF2α were counted in the BF fields containing cholinergic neurons, between stereotaxic planes of 0.26 mm and 0.92 mm posterior to bregma (31). Bilateral counts were made separately for three sections each in defined rostral, medial, and caudal regions of the BF, yielding a total of six counts for each region. For counting, section outlines were made at ×20 magnification. All cell counts were made at ×400 magnification within three rectangular grids placed over the regions containing ChAT-labeled neurons (see Fig. 2A).

Immunofluorescence intensity analysis.

Since Salub administration was found to induce a graded increase in p-eIF2α immunoreactivity in motor neurons (48), we determined the magnitude of p-eIF2α expression by measuring immunofluorescence intensity. Fluorescence intensity was measured in a randomly selected subset of neurons used for cell counting, following the method of Togo (43) modified as follows. A selection box (25 × 25 pixels) was placed under ×400 over the cytoplasmic area with the greatest staining, without overlapping the nucleus. Mean staining intensity within the selection box (defined as the sum of the gray values of all the pixels in the selection divided by the number of pixels) was measured with ImageJ (v 1.38). Cholinergic and noncholinergic neurons were separately evaluated. In addition to neurons, fluorescence intensity was also measured from the adjacent background areas. Data are reported as relative fluorescence intensity (ratio of neuronal fluorescence intensity to intensity of adjacent background). All measurements were made bilaterally in rostral, medial, and caudal BF areas, as described above. The magnitude of expression of p-eIF2α in the BF was assessed by comparing the immunofluorescence intensity of BF to that of cortex and medial preoptic area, two regions additionally shown to express p-eIF2α immunoreactivity, with the same methods.

Sleep data analysis.

Polysomnographic data were scored manually at 10-s intervals into five sleep-wake stages: active waking (AW), quiet waking (QW), slow-wave sleep 1 (SWS1), slow-wave sleep 2 (SWS2) and REM sleep, as previously described (27). Sleep-wake data of 12 h (ZT12–ZT24) were averaged in 3-h blocks. The amounts of each of the five stages as a percentage of the total recording duration were determined. The number of episodes and mean episode duration of each of the stages for the first 3 h of drug/vehicle perfusion were also calculated. In addition, REM latency was determined as the time taken for the appearance of the first REM episode of duration ≥10 s after the drug/vehicle reached the animal.

With SleepSign (Kissei Comtec), EEG spectral power of wake and NREM and REM sleep for the first 3 h of recording, divided into the first and last 90 min, was determined as follows. EEG data from an ipsilateral pair of frontoparietal EEG electrodes were digitally low-pass filtered at 50 Hz. Epochs containing artifacts were excluded, resulting in the rejection of 49% of wake, 7% of NREM, and 18% of REM sleep epochs. Spectral analyses were conducted with the fast Fourier transform on 10-s epochs tapered by Hanning window. The relative EEG spectral power, as a percentage of the total power, for wake, NREM, and REM in the first 3 h was calculated for the spectral range from 0.75 to 50.0 Hz. Averaged spectral data for each of the three stages were calculated for six frequency bands: lower delta (0.75–2.0 Hz), upper delta (2.5–4 Hz), theta (4.5–8 Hz), sigma (8.5–16 Hz), beta (16.5–30 Hz), and gamma (30.5–50 Hz). Changes in spectral power induced by Salub were calculated as a percentage of power in the respective frequency bands during the first 3 h of aCSF treatment.

Statistics.

All results are expressed as means ± SE. Differences between the effects of aCSF and Salub treatments on amounts of sleep stages, number and duration of sleep episodes, as well as p-eIF2α expression in BF neurons were analyzed by two-way ANOVA with SigmaStat v 3.0.1. The two factors used for two-way ANOVA were 1) drug (aCSF, DMSO, and Salub) vs. sleep stage (AW, QW, SWS1, SWS2, and REM) for the sleep study and 2) drug (aCSF, Salub) vs. BF region (rostral, medial, and caudal) for differences in the percentage of p-eIF2α expression in BF neurons as well as the p-eIF2α immunofluorescence intensity differences in the BF. One-way ANOVA was used to assess regional differences in the immunofluorescence intensity in cortex, preoptic, and BF areas as well as differences on the EEG spectral power bands after Salub treatment. Whenever ANOVA indicated a significant difference of P < 0.05, multiple comparisons of different conditions were made by Holm-Sidak post hoc test.

RESULTS

Sleep-wake response.

Histological analysis of brain sections showed that the injector cannula terminated in the lateral ventricle. In experiment 1, compared with vehicle, administration of 100 μM Salub for 12 h into the lateral ventricle significantly (P < 0.05) reduced AW while increasing deep slow-wave sleep (SWS2) in the first 3 h of perfusion (Fig. 1, A and B). The amounts of other sleep-wake stages including REM were unaffected for the same period. During the first 3 h of Salub perfusion, there was a significant increase in the number of episodes of SWS2 (P < 0.009) and REM (P < 0.05), although the mean duration of episodes was not affected (Fig. 1C). REM latency was not significantly affected. For the remaining 9 h of Salub perfusion, amounts of all sleep-wake stages were not significantly altered except at 10–12 h, where a significant (P < 0.05) decrease in AW and an increase in SWS2 was seen. Salub also induced differential effects on the EEG power spectrum. During the first 90 min of treatment, Salub increased the EEG power in the lower delta band by 53% (P < 0.03) while the power was reduced in the sigma (−30%; P < 0.05), beta (− 41%; P < 0.003), and gamma (−55%; P < 0.002) bands (Fig. 1D). Power in the other EEG spectral bands in wake, NREM, or REM was not significantly affected by Salub. EEG spectral power was not significantly altered during the second 90 min of SALUB administration (Fig. 1D).

Fig. 1.

Effects of salubrinal (Salub) administration on sleep parameters compared with artificial cerebrospinal fluid (aCSF) treatment. A: hypnograms and EEG delta power from a representative rat during the first 3 h of administration of aCSF (top) and Salub (bottom). Note that strongest Salub effects were in the first half of the recording. B: comparison of effects of 12 h (ZT12–ZT24) of intracerebroventricular administration of aCSF, DMSO (vehicle), and Salub on the different sleep-wake stages. Data are shown in 3-h blocks. Black horizontal bar at bottom indicates duration of treatment. In the first and last 3 h of treatment, Salub increased slow-wave sleep (SWS)2 sleep and reduced active wake (A wake, AW) compared with the vehicle (DMSO) or aCSF. DMSO induced a nonsignificant increase in quiet waking (Q wake, QW) during 4–9 h. ZT, zeitgeber. C: effects of Salub vs. aCSF on the number and duration of sleep-wake episodes in the first 3 h of treatment. D: effects of first 3 h of Salub administration on the spectral power of 6 EEG frequency bands as % difference from the aCSF treatment. NREM, non-rapid eye movement. Statistically significant difference of P < 0.05: *vs. ACSF and #vs. DMSO (2-way ANOVA for B and C and 1-way ANOVA for D; both ANOVAs were followed by Holm-Sidak post hoc test).

p-eIF2α expression in neurons.

In experiment 2, compared with vehicle 3 h of Salub administration decreased AW (28.83% vs. 41.16%) while increasing sleep, especially SWS2 (40.27% vs. 26.94%). In the caudal BF, Salub administration significantly (P < 0.05) increased the percentage of noncholinergic neurons expressing p-eIF2α (aCSF 57 ± 5%, Salub 75 ± 4%) but decreased the percentage of cholinergic neurons expressing p-eIF2α (aCSF 43 ± 5%, Salub 25 ± 4%). The neurons in the rostral and medial BF regions were not significantly affected by Salub treatment (data not shown).

The intensity of p-eIF2α expression was significantly increased in the BF neurons (1.841 ± 0.11) compared with neurons of the cortex (1.22 ± 0.03; P < 0.0003) or the medial preoptic area (1.25 ± 0.07; P < 0.0004) in Salub-treated animals (Fig. 2B). Salub treatment significantly increased p-eIF2α fluorescence intensity in noncholinergic neurons in all three regions (rostral = P < 0.03; medial = P < 0.003; caudal = P < 0.02) and in cholinergic neurons in rostral (P < 0.05) and medial (P < 0.02) BF (Fig. 2, D and E). Fluorescence intensity of the cholinergic neurons in caudal BF was not significantly different in Salub- vs. aCSF-treated animals (Fig. 2E).

Fig. 2.

A: effect of Salub administration on phosphorylated eukaryotic initiation factor 2α (p-eIF2α) expression in basal forebrain (BF) neurons. Grids represented as boxes were used for neuronal counting in rostral (a), medial (b), and caudal (c) BF. 3V, third ventricle; AC, anterior commissure; HDB, horizontal diagonal band; MCPO, magnocellular preoptic nucleus; SI, substantia innominata; OC, optic chiasm. B: examples of p-eIF2α expression in response to 3 h of Salub administration in cortex (B1), medial preoptic area (B2), and BF (B3). C: comparison of p-eIF2α expression in response to 3 h of administration of aCSF (C1) vs. Salub (C2). a and b: Magnified views of the corresponding boxes at top. D: aCSF (D1) vs. Salub (D2) treatment effect on fluorescence intensity of p-eIF2α expression in cholinergic and noncholinergic neurons of BF. p-eIF2α expression is indicated by yellow arrow in cholinergic neuron and by red arrow in noncholinergic neuron. E: regional differences in fluorescence intensity of p-eIF2α expression after aCSF vs. Salub treatment in cholinergic and noncholinergic neurons of BF (2-way ANOVA followed by Holm-Sidak post hoc test; *P < 0.05). ChAT, choline acetyltransferase.

DISCUSSION

This study found that intracerebroventricular administration of Salub increased deep slow-wave sleep (SWS2 in the rat) by 255%, while reducing AW by 49%. Delta power within NREM sleep was increased, while power in the sigma, beta, and gamma bands during NREM was reduced. Salub also increased the number of REM sleep episodes without affecting the latency or amount of REM. We note that REM sleep is often suppressed if sleep is induced by anesthetics or nonphysiological stimuli. In addition, spindles in NREM corresponding to the sigma band as well as relative power in the beta and gamma bands are increased by agents such as barbiturate or picrotoxin (24), in contrast to the effects observed in this study. These observations suggest that Salub induces normal physiological sleep and not EEG abnormalities that might be misclassified as sleep. We found that Salub increased SWS2 and decreased AW at both lights off (ZT12) and lights on (ZT24), but a possible circadian modulation by Salub requires additional experimental analysis.

Because Salub was perfused into the lateral ventricle, a site of action cannot be determined. However, we confirmed that Salub administration increased p-eIF2α expression in the BF neuronal population. Salub administration decreased the numbers of cholinergic neurons exhibiting p-eIF2α expression but increased expression in noncholinergic neurons of the caudal BF. Moreover, Salub increased the intensity of p-eIF2α expression in both cholinergic and noncholinergic neurons, but this was more widespread among the noncholinergic neurons. The intensity of the Salub-mediated increase in p-eIF2α expression was greater in the BF compared with the cortex or the medial preoptic area, suggesting that protein synthesis regulatory mechanisms in the BF might be more directly relevant to the sleep regulation. The presence of p-eIF2α in aCSF-treated rats indicates constitutive basal level of expression of p-eIF2α in our experimental paradigm. Salub's action requires the presence of p-eIF2α. We cannot now identify the kinase phosphorylating eIF2α under control conditions. Our study suggests that this kinase may play a role in sleep regulation. A strong candidate is PERK, which is upregulated by mild sleep restriction (see introduction).

Although the role of the BF in sleep-wake control is well documented, the critical importance of BF cholinergic neurons in sleep EEG patterns has been challenged (3). Instead, evidence supports the importance of noncholinergic BF neurons in generation of slow wave activity, the widely accepted index of short-term sleep homeostasis (21). This is consistent with the stereological estimates; noncholinergic neurons constitute 95% of the total BF population, of which 35% are GABAergic (14). We hypothesize that increased p-eIF2α expression in noncholinergic wake-promoting BF neurons decreases their excitability, reducing their capacity to generate behavioral and EEG arousal, and thereby contributes to increased NREM sleep and delta activity within NREM sleep. Of course, in addition, intracerebroventricular Salub may affect sleep-regulatory cells in other regions, including hypothalamus.

The maintenance of p-eIF2α in the phosphorylated state by Salub attenuates the translation of mRNAs and protein synthesis (4, 10). In this regard, the effects of Salub are directly comparable to the effects of a PSI. The findings of this study are consistent with the earlier report of peripheral PSI administration that increased deep sleep (33) as well as our recent findings (Ref. 26; see introduction). Our findings together with evidence that sleep deprivation induces ER stress support a hypothesis that sleep may be induced by signals associated with an increased demand for brain protein synthesis and accumulation of unfolded proteins. An alternative interpretation, that protein synthesis inhibition more directly facilitates sleep, is unlikely since brain protein synthesis is elevated during sleep (30, 36).

Some possible molecular pathways by which Salub may induce sleep are suggested by previous work. Phosphorylation of eIF2α activates NF-κB (11). Induction of NF-κB was shown to facilitate sleep (6, 22), and it is expressed particularly in BF neurons (2), where we found increased p-eIFα expression after Salub administration. Paradoxically, phosphorylation of eIF2α, by decreasing efficiency of recognition of AUG start codons, induces scanning of downstream AUG codons, such as that of activating transcription factor 4 (ATF4), which is rapidly induced by ER stress (17, 44). ATF4 is a repressor of cAMP response element binding protein (CREB)-mediated gene expression (5). CREB activation facilitates waking in Drosophila (18), and increasing availability of cAMP increases wakefulness in rats (23). CREB mutant mice have reduced waking (13). Thus the induction of NF-κB or ATF4 could be a mechanism linking ER stress to neurophysiological processes regulating sleep, perhaps by reducing CREB-mediated activation of arousal-promoting neurons. Alternatively, Salub may inhibit synthesis of other wake-promoting factors or their precursors (19). Further work is needed to determine which pathways mediate the sleep-promoting effects of Salub.

Salub was discovered in a screen of ∼19,000 molecules for protection against tunicamycin-induced apoptosis in vitro and was shown to act by specifically blocking serine/threonine phosphatase-dependent dephosphorylation of eIF2α (4). The central importance of eIF2α phosphorylation in normal physiology was demonstrated by the finding that deletion of the gene coding for a kinase mediating eIF2α phosphorylation (PERK^−/−) results in diabetes mellitus in mice (47). Salub was found to protect hippocampal neuronal cultures against excitotoxic neuronal injury (40) as well as brain stem motor neurons from ER stress induced by long-term intermittent hypoxia (48). It is hoped that Salub may be useful in protecting brain and other cell types from ER stress-induced injury. This regulatory process may be unusually potent in brain, since neurons in most brain regions cannot be replaced. On the other hand, it must be kept in mind that Salub-mediated suppression of protein synthesis is a double-edged sword, and may also have deleterious effects (10).

Although short-term sleep deprivation induces molecular hallmarks of ER stress (29), expression of these markers in cerebral cortex in response to chronic sleep loss is relatively less compared with acute sleep loss (8). This suggests that the sleep deprivation response may elicit ER stress only transiently. This observation is congruent with our finding that the strongest sleep-promoting effect of Salub occurred in the initial 3 h of administration. The ER stress response comprises processes that serve to diminish the stress. In addition to inhibition of folding load through inhibition of translation, the response includes mechanisms to increase degradation of unfolded or misfolded proteins and increase folding capacity (37). These counterregulatory mechanisms may account for the limited duration of Salub-induced increases in sleep. If our working model is correct, sleep itself also counteracts ER stress.

The possible role of sleep in the functioning of protein folding machinery has been previously demonstrated. Drosophila mutants deficient in the heat shock protein HSP83, a chaperone assisting in protein folding, show an exaggerated sleep homeostatic response and die after sleep deprivation, whereas activation of HSP83 rescued cyc mutant flies from the lethal effects of sleep deprivation (39). Sleep deprivation upregulates the expression of BiP, the chaperone playing a central regulatory role in sensing the UPR, in mice (29, 41) and rats (8). Experimentally induced changes in the expression of BiP affected the sleep recovery response following sleep deprivation in Drosophila: overexpression of BiP increased the sleep rebound, and its underexpression decreased the rebound (28).

Perspectives and Significance

Our finding of facilitation of sleep after induction of a key element of the ER stress signaling cascade supports hypotheses that sleep is an additional component of the compensatory response to ER stress in brain and is regulated by the ER stress pathway. Since prolongation of waking has been shown to induce ER stress, the sleep promotion by an ER stress pathway molecule observed in this study suggests that sleep may help counteract ER stress. Salub has been suggested to be used for reprogramming the ER stress pathways by balancing ER-protein load with cellular folding capacity, potentially protecting the brain against so-called conformational diseases resulting from incorrect protein folding (45). If sleep normally counteracts ER stress, the severe sleep disturbances that are known to accompany conformational diseases such as Alzheimer and Parkinson diseases (34) may contribute to the etiology of these diseases.

GRANTS

This work was supported by the Department of Veterans Affairs Research Service and National Institutes of Health Grants MH-075076 and HL-60296.