Genomic analysis of sleep deprivation reveals translational regulation in the hippocampus

Abstract

Sleep deprivation is a common problem of considerable health and economic impact in today's society. Sleep loss is associated with deleterious effects on cognitive functions such as memory and has a high comorbidity with many neurodegenerative and neuropsychiatric disorders. Therefore, it is crucial to understand the molecular basis of the effect of sleep deprivation in the brain. In this study, we combined genome-wide and traditional molecular biological approaches to determine the cellular and molecular impacts of sleep deprivation in the mouse hippocampus, a brain area crucial for many forms of memory. Microarray analysis examining the effects of 5 h of sleep deprivation on gene expression in the mouse hippocampus found 533 genes with altered expression. Bioinformatic analysis revealed that a prominent effect of sleep deprivation was to downregulate translation, potentially mediated through components of the insulin signaling pathway such as the mammalian target of rapamycin (mTOR), a key regulator of protein synthesis. Consistent with this analysis, sleep deprivation reduced levels of total and phosphorylated mTOR, and levels returned to baseline after 2.5 h of recovery sleep. Our findings represent the first genome-wide analysis of the effects of sleep deprivation on the mouse hippocampus, and they suggest that the detrimental effects of sleep deprivation may be mediated by reductions in protein synthesis via downregulation of mTOR. Because protein synthesis and mTOR activation are required for long-term memory formation, our study improves our understanding of the molecular mechanisms underlying the memory impairments induced by sleep deprivation.

in today's society, people obtain insufficient sleep for many reasons, such as busy schedules, sleep disorders, or psychiatric disturbances (reviewed in Refs. 41, 86). This sleep loss is in turn associated with deleterious effects on cognitive function (reviewed in Refs. 4, 15). One of the brain regions whose function appears to be compromised by sleep deprivation is the hippocampus. The hippocampus is crucial for the formation of spatial, contextual, and declarative memories (reviewed in Refs. 1, 59), and hippocampus-dependent memory consolidation is particularly susceptible to disruption by sleep deprivation (30–32, 55, 70, 74, 88). Even relatively brief periods of sleep deprivation (5–6 h) impair consolidation of hippocampus-dependent associative (31) and spatial learning in rodents (74), without affecting hippocampus-independent versions of these tasks. Brief sleep deprivation also disrupts hippocampal synaptic plasticity (48, 83), a cellular model of memory.

However, the underlying mechanisms by which sleep deprivation impairs hippocampal function are not well understood. One clue has come from studies showing that hippocampus-dependent memory is most strongly impacted by sleep deprivation when animals are deprived of sleep during the first 5–6 h following learning (31, 63). This window coincides with the period of memory stabilization called consolidation, which depends critically on waves of gene expression and protein synthesis (10, 36, 42, 43). There also appears to be a time window immediately following spatial learning in humans during which sleep can improve memory consolidation, whereas sleep deprivation prevents this enhancement (22, 23). Thus, a prediction from this body of research is that sleep deprivation may affect signaling mechanisms that regulate transcription and translation, thus disrupting the mechanisms of memory consolidation in the hippocampus.

Recent work has identified a handful of signaling pathways and molecules affected by sleep deprivation in the hippocampus (reviewed in Ref. 35), but this list is likely far from complete. One method to identify novel molecular targets of sleep deprivation is to determine how gene expression is affected by sleep loss. Although several wide-scale gene expression studies after sleep deprivation have been performed (12, 51, 52, 80), thus far only one has focused on the hippocampus (14). Because previous studies have seen sizable differences in the gene expression responses to sleep deprivation across brain regions (33, 77), it is important to study each brain area of interest directly. Therefore, in this study, we performed a genome-wide microarray to assess the effects of 5 h of sleep deprivation on gene expression in the mouse hippocampus. We then used bioinformatic analysis of the resulting patterns of gene expression to identify particular cellular signaling disruptions that might underlie the negative effects of sleep deprivation on hippocampal function.

METHODS

Subjects.

C57BL/6J adult male mice (2–4 mo of age) were housed individually on a 12 h/12 h light-dark schedule with lights on at 7 AM [Zeitgeber time (ZT) 0]. Food and water were available ad libitum throughout the experiment. To acclimate the mice to the experimenter and to the techniques utilized during sleep deprivation, each animal was handled daily for 3–6 days prior to sleep deprivation. Handling consisted of the same interventions used during sleep deprivation, for 2–3 min per mouse. Mice were not removed from their cages during handling. For microarray studies, sleep deprivation began between ZT4 and 6, and for qPCR validation, sleep deprivation began between ZT3 and 6. For simplicity, we refer to these groups as SD ZT5. For early sleep deprivation qPCR experiments, sleep deprivation began at ZT0. In experiments on recovery sleep, sleep deprivation began at ZT3–5, and was followed by 2.5 h of recovery. Sleep deprivation was carried out in the animals' home cages for 5 h by gentle handling. This consisted of making mild noises or tapping or jostling the animal's cage, disturbing the animal's nesting material, or stroking the animal. These interventions were only carried out when animals settled and attempted to go to sleep, and direct contact with the animals was kept to a minimum. This technique has been shown to be highly effective at inducing total sleep deprivation (56), without being a strong stressor (34, 57, 82). Nonsleep-deprived mice were left undisturbed in their home cages. Hippocampal dissections were performed immediately following the behavioral treatment, and alternated between SD and NSD animals. All experiments were approved by the Institution of Animal Care and Use Committee of the University of Pennsylvania and were carried out in accordance with all National Institutes of Health guidelines.

Microarrays.

RNA extraction was performed as previously described (84), except that DNase treatment and both precipitation steps were omitted. Instead, phase separation was carried out using 1-bromo-3-chloropropane instead of chloroform, an equal volume of 70% ethanol was added before beginning RNeasy cleanup, and following elution from the RNeasy column in 50 μl RNase-free water, 1 μl of Superase-In (Ambion) was added to each sample and samples were concentrated to ∼20 μl by SpeedVac. RNA was submitted to the University of Pennsylvania Microarray core for cDNA preparation and hybridization to Mouse 430_2 Affymetrix chips. Target preparation and hybridization protocols were conducted as described in the Affymetrix GeneChip Expression Analysis Technical Manual. Each sample was hybridized to its own chip. A confocal scanner was used to collect fluorescence signal at 3 μm resolution after excitation at 570 nm. The average signal from two sequential scans was calculated for each microarray feature. Robust multiarray average normalization and statistical analysis were performed using the affy and limma packages from R/Bioconductor (29). Multiple testing corrections were performed using the method of Benjamini and Hochberg (8, 9). Microarray data generated in this study will be made publicly available through Gene Expression Omnibus (GSE33302). Of the 29,479 probe-sets on each chip, 22,689 were expressed in at least one condition, sleep deprivation or nonsleep deprivation (as defined by an average log expression value >4).

Quantitative real-time RT-PCR.

RNA preparation, cDNA synthesis, and quantitative real-time RT-PCR (qPCR) analysis was performed as previously described (84). RNA concentration and purity were quantified by NanoDrop spectrophotometry (Thermo Fisher Scientific, Wilmington, DE). Generation of cDNA was carried out by the RETROscript kit (Ambion) with 1 μg of RNA as template. For quantitative real-time RT-PCR, reactions were prepared in 96-well optical reaction plates (ABI, Foster City, CA) with optical adhesive covers (ABI). Three technical replicates were used. Reactions were carried out in the ABI Prism 7000. Primer sequences can be found in Table 1. Data were normalized to Actg1, Hprt, and Tuba4a prior to calculation of differences, using the same primers as described previously (84). Relative quantification of gene expression was performed according to ABI's User Bulletin #2. Fold change was calculated from the delta Ct values with corrections for standard curve data from each gene and housekeeping gene expression levels for each sample based on the relative standard curve method described in the Applied Biosystems manual. Because corrections were made for primer efficiency, we have presented the data as fold change. The data presented are the calculated means for the biological replicates with n being equal to the number of biological replicates (i.e., the number of mice examined). We used t-tests to compare fold change values for each gene in each comparison of interest. For validation experiments, one-tailed P values are reported because of our initial prediction about the direction of each fold change based on microarray data.

Table 1.

Summary of qPCR validation

Gene Name	Probe-set ID	Microarray FC	qPCR FC	qPCR P Value	SD, n	NSD, n	Forward Primer	Reverse Primer
Fos	1423100_at	2.24	1.46	2E-03	9	9	CCGACTCCTTCTCCAGCA	TGTCACCGTGGGGATAAAGT
Arc	1418687_at	1.87	1.38	8E-03	8	9	AGCAGCAGACCTGACATCCT	GGTGTCATTCTCCTGGCTCT
Hspb1	1425964_x_at	1.77	2.07	1E-04	10	9	CACAGTGAAGACCAAGGAAGG	GGTGAAGCACCGAGAGATGT
Adamts2	1455720_at	1.61	1.38	3E-02	8	9	GAACCACAGGCTCTGGACAC	AACTGGACCACCGAGTCATC
Hspa8	1431182_at	1.57	1.18	2E-03	9	8	CGGAAGAGCTGATGGTAAAGA	CACCTTCCTGCTGCTTCTCT
Tsc22d3	1420772_a_at	1.56	1.67	1E-06	19	19	TTCTCTGCTTGGAGGGGATT	GCTCACGAATCTGCTCCTTT
Hspa5	1416064_a_at	1.50	1.66	1E-05	10	9	GGAGACTGCTGAGGCGTATT	TGGGCATCATTGAAGTAAGC
Nr4a1	1416505_at	1.43	1.08	1E-01	9	9	AAAATCCCTGGCTTCATTGAG	TTTAGATCGGTATGCCAGGCG
Prkab2	1435874_at	1.39	1.39	3E-04	10	10	GGGAAAGGAGCACAAGATC	CTGCTGCCAGGGTACAAAC
Htr1a	1450219_at	1.39	1.20	5E-02	9	9	CTGTTTATCGCCCTGGATGT	ATGAGCCAAGTGAGCGAGAT
Lats2	1419678_at	1.39	1.23	3E-02	14	13	ACCAAATTACATCGCTCCGG	TCGTCTCCCAGTTGATCACCT
Elk1	1421897_at	1.38	1.35	2E-03	15	15	TCAAGTTGGTGGATGCAGAG	GCAACCTCTGGGTAGGACAC
Prkaa2	1434766_at	0.79	0.88	5E-02	10	10	GGAATCCCGAAGCCAGAGCAA	CGATGCTTTTGAAGTCCAGA
Prkab1	1424119_at	0.79	0.81	3E-02	10	10	CATCCTCCCGCCACACCTGC	GAGCACCATCACTCCATCCT
Kcnv1	1439333_at	0.78	0.79	8E-03	8	9	GTCTGGTCCAGGTCTTGAGG	CACCATGCACAAGGAACACT
Hnrpdl	1449039_a_at	0.72	0.75	3E-03	13	14	AAGAACCAGCAGGATGACGGT	TGCAGTCTACCACTTCCCCAA
Usp2	1417168_a_at	0.66	0.82	7E-03	8	9	AAAGAGTGCCCAGGGTCTG	TCTCTCAGCTCTCGGGTGTT
Sirt7	1457279_at	0.65	0.84	3E-02	10	10	GAGCCTACCCTCACCCACA	GTGGAGCCCATCACAGTTC
Rbm3	1429169_at	0.54	0.72	2E-02	14	13	AGGACTTGCCTTCTGCCATGT	ATAGGCCCAAAGCTGCTGAA

Genes detected as differentially regulated in our sleep deprivation microarray (multiple testing corrected P value <0.05) tested by qPCR. All comparisons between sleep-deprived (SD) and nonsleep-deprived (NSD) samples measured by qPCR are significant at P < 0.05 using 1-tailed t-tests, except Nr4a1. For each gene, qPCR expression is represented as the fold change (FC) in SD mice relative to NSD mice, normalized to the average expression of the housekeeping genes Actg, Hprt, and Tuba4a.

Transcription factor binding site enrichment analysis.

Bioinformatic analysis of enrichment of transcription factor binding sites (TFBS) in the promoters in lists of genes whose expression was altered by sleep deprivation was carried out as previously described (49) using the 5 kb regions upstream of transcription start sites. Genes were searched for the presence of 584 vertebrate positional weight matrices (PWMs) obtained from the TRANSFAC database v8.4 (http://www.gene-regulation.com) for known TFBSs at P < 0.0002. Matches reaching criterion were then filtered using human-mouse conservation and were retained if there was at least 80% conservation or they had P < 0.00002. Enrichment of TFBSs in each gene list was calculated by dividing the frequency of occurrence of a given TFBS PWM in the gene list of interest by the frequency of occurrence of the same PWM in the background list of 16,757 RefSeq annotated mouse genes using a false discovery rate (FDR) of <0.01 as a cutoff. Overrepresentation of multiple binding sites in the same promoters was analyzed by χ².

Functional clustering analysis.

Enrichment of functional annotation was assessed using the Database for Visualization and Integrative Discovery (DAVID) (17) and the following sources: Gene Ontology (GO) biological process, GO molecular function, KEGG pathways, and SwissProt and Protein Information Resource keywords. Enrichment for each term was defined relative to the all mouse probe-sets present in the microarray, and was defined as a P < 0.05 with at least three genes per term per dataset. Fuzzy heuristical clustering was performed using kappa similarity >0.3 and requiring an enrichment score >1.5 (P value geometric mean <0.05).

Western blot analysis.

Frozen hippocampal tissue was homogenized in RIPA buffer with protease and phosphatase inhibitors. Proteins were separated by 4–20% Tris-glycine SDS-PAGE and transferred to polyvinylidene difluoride membranes. Membranes were blocked in 5% BSA-TBST and incubated overnight at 4°C in primary antibody (phospho-mTOR, mTOR, 1:1,000; Cell Signaling). They were washed and incubated with appropriate horseradish peroxidase-conjugated goat anti-mouse or anti-rabbit IgG (1:5,000, Santa Cruz) for 1 h. Blots were exposed on film by ECL and quantified using ImageJ, and the density of signal was normalized to β-tubulin levels (1:20,000, Sigma).

RESULTS

Sleep deprivation induces widespread changes in gene expression in the hippocampus.

We first asked whether mice that were sleep deprived for 5 h by gentle handling showed genome-wide changes in gene expression in the hippocampus compared with nonsleep-deprived controls. A schematic of our experimental approach can be found in Fig. 1. Of the 22,689 probe-sets expressed in our samples, we detected 616 differentially expressed probe-sets, a 2.7% hit rate. These probe-set hits corresponded to 533 genes, 214 upregulated and 319 downregulated, at a multiple testing-corrected P value of <0.05, (Supplemental Table S1).¹

Fig. 1.

Analysis of gene expression following 5 h of sleep deprivation. Microarray experimental design and results. An initial comparison was made between genome-wide mRNA expression patterns in hippocampal tissue taken from young male sleep-deprived (SD) and nonsleep-deprived (NSD) mice, using Mouse 430_2 Affymetrix microarray chips. Following normalization by robust multiarray average (RMA), microarray data were analyzed using the affy and limma packages in R/Bioconductor. BH, Benjamini-Hochberg; DAVID, Database for Annotation Visualization and Integrated Discovery; EASE, Expression Analysis Systematic Explorer.

We next used qPCR to validate our microarray studies using hippocampi from separate groups of sleep-deprived and nonsleep-deprived mice. We examined the expression of 12 upregulated (Fos, Arc/Arg3.1, Hspb1, Adamts2, Hspa8, Tsc22d3, Hspa5/Bip, Nr4a1, Prkab2, Htr1a, Lats2, and Elk1) and seven downregulated (Prkaa2, Prkab1, Kcnv1, Hnrpdl, Usp2, Sirt7, and Rbm3) genes. These genes were chosen because of potential ties to the regulation of synaptic plasticity, because many were present in the biological function clusters identified by the bioinformatic analysis described below (see Fig. 3), and to assess genes that spanned a wide range of fold changes. Analysis by qPCR validated the gene expression changes of 18 of the 19 genes, a validation rate of ∼95% (Fig. 2 and Table 1). The one gene that did not validate, Nr4a1, was modestly upregulated in the qPCR data, but this increase was not statistically significant. We also examined the expression of a set of these genes following a 5-h period of sleep deprivation shifted to begin immediately after lights-on (ZT0), to determine if the regulation of these genes by sleep deprivation depended on differences in sleep drive across the day. All genes showed similar changes in expression regardless of when sleep deprivation began (data not shown).

Fig. 3.

Enriched functions regulated by sleep deprivation. Sleep deprivation downregulates translation and upregulates transcription. Functional annotation terms from the following databases: Gene Ontology (GO) biological process and molecular function, KEGG pathways and protein information resource keywords, were clustered based on similarity using the Database for Annotation Visualization and Integrated Discovery (DAVID). Clusters of functional terms enriched in SD down- or upregulated gene lists compared with the genome as a whole (P value <0.05) are represented as bars. Height of bars represents the enrichment score of each cluster, with the scores of downregulated clusters shown as negative numbers for visualization purposes. Enrichment score is calculated as −log(10) of the geometric mean P value among all clustered terms. Only clusters with enrichment score >1.5 (average P value of functional terms within the cluster <0.05) were considered. Examples of genes found within each cluster are shown, with qPCR-tested genes in boldface. Note that there were significant clusters of ion-binding functional terms found within both the up- and downregulated gene lists. For details of the functional terms included in these clusters, see Supplemental Table S2.

Fig. 2.

Quantitative RT-PCR validation of genes upregulated or downregulated by sleep deprivation in the hippocampus. Quantitative RT-PCR (dark gray) was used to validate the expression level of genes identified by microarray analysis (light gray) as being changed in the hippocampus by sleep deprivation. For each gene, expression is represented as the fold change in SD mice relative to NSD mice, normalized to the average expression of housekeeping genes Actg, Hprt, and Tuba4a. The fold change values from the microarray for SD/NSD are shown for each gene for comparison. All SD/NSD qPCR comparisons are significant at P < 0.05, except Nr4a1 induction (see Table 2). Bars indicate ± SE.

Bioinformatic analysis identifies translational regulation as a prominent target of sleep deprivation in the mouse hippocampus.

Identification of the specific genes whose expression is altered by sleep deprivation is useful, but it is often more informative to determine how multiple genes altered by sleep deprivation are functionally related. To do so, we used functional clustering analysis to determine if genes that were up- or downregulated by sleep deprivation were enriched in particular cellular functions or pathways using DAVID (17). This analysis identified 11 functional annotation clusters that were significantly enriched in our data (enrichment score >1.5, mean P value <0.05), five unique to downregulated genes, six unique to upregulated genes, and one (ion-binding) that was significantly enriched in both up- and downregulated genes (Fig. 3).

The most prominent clusters of downregulated genes (enrichment score >2, mean P value <0.01) were enriched in genes involved in the regulation of ubiquitination/proteolysis, including several ubiquitin-specific peptidases (Usp2, Usp24, Usp3, and Usp34), translation, which includes translation initiation factors (Eif2a, Eif3s6ip, Eif4el3, and Eif5), as well as genes linked to mRNA processing and transport (Rbm3 and Denr), and RNA-binding, containing the nuclear mRNA shuttle Hnrpdl and cold-induced RNA-binding proteins Cirbp and Rbm3. Other moderately enriched functional clusters among genes downregulated by sleep deprivation included cholesterol metabolism, containing a subunit of the energy sensor AMP-activated kinase (AMPK) (Prkaa2) and the very low-density lipoprotein receptor (Vldlr), negative regulation of transcription, including multiple genes containing histone deacetylase activity (Sirt5, Sirt7, Hdac3, and Hdac9), as well as ion binding, encompassing a wide array of protein classes that require ion cofactors such as Prkaa2, Kcnv1, Kcnk2, Camk4, Zswim1, and Nfx1.

Among upregulated genes, the most highly enriched clusters (enrichment score >2) are related to nucleosomes/chromatin assembly, containing the transcription factors Elk1 and Fos and multiple histone family members such as H2afj, Hist1h2bc, and Hist3h2a, RAS/RAF signaling genes, including three members of the RAS oncogene family (Rab8b, Rab15, and Rab21), and the unfolded protein response (UPR), including multiple heat shock proteins (Hspa8, Hsp110, and three Hsp40 homologs). Other enriched functional clusters among upregulated genes were associated with positive regulation of transcription, including several transcription factors (Fos, Elk1, Nr4a1, Creb1, and Crem), ion-binding (Adamts2 and Lats2), negative regulation of kinase activity (Nr4a1, Lats2, Dusp19), and ATP/nucleotide binding (Hspa8, Lats2). Details of the individual genes found within each functional cluster can be found in Supplemental Table S2. The only functional category enriched in both up- and downregulated genes was ion binding, which likely reflects a general need for ion cofactors for the function of several gene products. The combination of downregulated translation initiation genes, downregulated RNA-binding genes, many of which function to shuttle mRNA from the nucleus to ribosomes for translation, and upregulated UPR genes, which act to stall protein synthesis, suggests that repression of translation may be a major effect of sleep deprivation.

Analysis of the gene list altered by sleep deprivation using DAVID also identified seven enriched signaling pathways (Table 2). Ketone metabolism, splicing, and prostate cancer were uniquely enriched pathways among genes downregulated by sleep deprivation, whereas MAPK signaling, antigen processing and presentation, and systemic lupus/nucleosome function were enriched among upregulated genes. Genes involved in the insulin signaling pathway were enriched among both up- and downregulated gene sets. Interestingly, the insulin-related genes in each set map to distinct components of the pathway (Fig. 4). Downregulated genes map primarily to protein synthesis and other anabolic processes, including the mTOR pathway, whereas upregulated genes map mostly to MAPK activity and transcriptional regulation, thus mirroring the results of the functional enrichment analysis described above (Fig. 3). The enrichment in the insulin signaling pathway predicts that the translation regulation may be mediated by mTOR.

Table 2.

Enriched pathways regulated by sleep deprivation

Enriched Pathways (KEGG)	SD vs. NSD	Enrichment P Value	Genes
Synthesis and degradation of ketones	−	0.008	Acat1, Bdh, Hmgcs1
Insulin signaling	−	0.01	Foxo1, D130043N08Rik, Eif4e, Srebf1, Prkab1, Exoc7, Prkaa2
Splicing	−	0.03	Sfrs5, Ddx46, Cdc40, Tcerg1, Rbmx
Prostate cancer	−	0.04	Foxo1, Pdgfrb, Pdgfb, Pten, Rb1
MAPK signaling	+	0.0007	Stk4, 2900078C09Rik, Hspb1, Dusp1, Mknk2, Elk1, Nr4a1, Crkl, Hspa8, Fos
Antigen processing and presentation	+	0.01	Calr, Nfya, Hspa8, Hspa5, Creb1
Systemic lupus/nucleosome	+	0.02	H2afj, Hist1 h2bb, HIST1H3B, HIST1H3C, HIST1H3D, HIST1H3E, HIST1H3F, Hist2 h2aa1, Hist2 h2aa2, HIST2H3B, Hist2 h3c1, Hist2 h3c2, Hist3 h2a, Gm11277, Gm13646, HIST1H2BF, HIST1H2BJ, Hist1 h2bl, HIST1H2BN, LOC100046213
Insulin signaling	+	0.04	Prkab2, Mknk2, Elk1, Crkl, Ppp1r3c

Biological pathways from the KEGG database enriched in SD down- or upregulated gene lists compared with the genome as a whole (P value <0.05) are listed. Note that the insulin signaling pathway is overrepresented among genes that are both up- and downregulated by sleep deprivation.

Fig. 4.

The effects of sleep deprivation on the insulin signaling pathway. Genes regulated by sleep deprivation mapped to the insulin signaling pathway (adapted from KEGG and Wikipathways). Upregulated gene names are written in lower case, in bold and italics, and downregulated gene names are written in bold with underlining. Note that genes from several functional clusters and enriched pathways shown in Fig. 3 and Table 2 are contained within this signaling network.

We next performed bioinformatic analysis to determine if particular TFBS were overrepresented in the promoter regions of the up- and downregulated gene lists relative to a list of 16,757 annotated mouse promoters (49). Twelve binding sites, corresponding to eight known transcription factors (AP-2, E2F, HIF-1, Nrf-1, IPF1, HIC1, Egr-2, and ETF), were significantly overrepresented in the downregulated genes (FDR < 0.01, Supplemental Table S3), and the simultaneous presence of all binding sites except IPF1 was particularly overrepresented (P = 3.1E-8). None of these transcription factors were significantly altered at the mRNA level in our microarray, suggesting that their activity is regulated by sleep deprivation at a posttranscriptional level. In contrast, no binding sites were significantly enriched in the upregulated genes. These findings suggest that sleep deprivation downregulates gene expression through coordinated regulation of transcription, whereas the upregulation of gene expression has no common transcriptional regulatory component.

Sleep deprivation reduces levels of the translational regulator mTOR.

Results from the bioinformatics analysis described above suggested that sleep deprivation downregulates translation in the hippocampus, and implicated the evolutionarily conserved TOR pathway as a potential mediator of this effect via its involvement in the insulin signaling network. Long-term memory formation is known to rely on protein synthesis (10, reviewed in Ref. 37), and one of the key regulators of protein synthesis during memory consolidation is mTOR (reviewed in Refs. 28, 38, 45). Increased mTOR phosphorylation generally leads to an increase in translation (87). Therefore, we investigated if sleep deprivation affects mTOR levels and/or the levels of mTOR phosphorylation in the hippocampus using Western blot analysis. Sleep deprivation led to a decrease in total mTOR levels and an even larger decrease in phosphorylated mTOR (Fig. 5). We also found that allowing mice to sleep for 2.5 h after sleep deprivation reversed the effects of sleep deprivation on both total and phosphorylated levels of mTOR (Fig. 5). Activity monitoring indicated that the mice were asleep for 95% of the recovery period (data not shown). This finding suggests that the decreases in total and phosphorylated mTOR are related to sleep loss itself, because previous work has shown that 2.5 h recovery is sufficient for C57BL6/J mice to dissipate nearly all of the homeostatic sleep drive built up during a 6-h period of sleep deprivation (25, 40). Interestingly, the same period of rebound following sleep deprivation is also sufficient for the mouse hippocampus to regain the capacity for long-term synaptic plasticity (48, 83). We also tested the effects of 2.5 h of recovery following sleep deprivation on a subset of validated genes. We found that Arc, Tsc22d3, Prkab2, Hnrpdl, and Usp2 showed significant recovery, whereas Adamts2, Htr1a, Elk1, and Kcnv1 did not (Fig. 6). This demonstrates that the molecular effects of gene expression have varied time-courses of recovery and may indicate that different pathways regulated their expression during sleep deprivation. In summary, these data identify mTOR as a novel target of sleep deprivation and are consistent with an overall downregulation of protein synthesis by sleep deprivation.

Fig. 5.

Sleep deprivation reduces levels of mTOR and mTOR phosphorylation, and recovery sleep reverses these effects. A: representative Western blots of phosphorylated mTOR (p-mTOR, left) and total mTOR (mTOR, right) from hippocampus homogenates of SD animals (n = 9) and NSD controls (n = 9). β-Tubulin reactivity is shown as a loading control (bottom). Quantitation is shown with protein levels represented as a fold change in SD mice relative to NSD mice, which are normalized to the β-tubulin loading control. B: representative Western blots of p-mTOR (left) and mTOR (right) from hippocampus homogenates of SD and NSD mice that were allowed to sleep for 2.5 h post-SD (SD+R, n = 7). β-Tubulin reactivity is shown as a loading control (bottom). Quantitation is shown with protein levels represented as a fold change in SD+R mice relative to NSD mice, which are normalized to the β-tubulin loading control. Bars represent ± SE. **P < 0.005, *P < 0.01, 2-tailed t-test.

Fig. 6.

qPCR analysis of hippocampal gene expression following 2.5 h of recovery sleep after sleep deprivation. For each gene, expression is represented as the fold change in SD mice allowed to sleep for 2.5 h (SD+R) relative to NSD mice, normalized to the average expression of housekeeping genes Actg, Hprt, and Tuba4a. Black line denotes no change in gene expression between SD and NSD mice (fold change = 1). Bars indicate ± SE. *Significant differences between SD+R relative to SD (P < 0.05).

DISCUSSION

This is just the second microarray study to focus on the effects of sleep deprivation on the hippocampus (14), a brain area whose function appears to be particularly susceptible to disruption by sleep loss (30–32, 55, 70, 74, 88), and is the first study to do so in mice. We show that a period of sleep deprivation that produces deficits in hippocampus-dependent memory and synaptic plasticity (83) causes widespread changes in hippocampal gene expression. We identify several genes that have not previously been found to be regulated by sleep or sleep deprivation. These include the validated genes Tsc22d3, Prkab2, Adamts2, Htr1a, Kcnv1, and Sirt7 (Fig. 1, Supplemental Tables S1 and S2). This study highlights novel sleep deprivation target genes that are likely to have functional impact. For example, Tsc22d3 has been shown in other systems to negatively regulate the memory- and synaptic plasticity-related signaling molecule extracellular signal-regulated kinase (ERK) (76), levels of which peak during sleep (21) and are reduced in the hippocampus following sleep deprivation (32, 69), and Prkab2 is a subunit of the energy-sensing molecule AMP-activated kinase (AMPK), which undergoes increased phosphorylation following short-term sleep deprivation (11, 20, 62) and plays a role in homeostatic sleep regulation (11). We also show that several genes whose expression is altered by sleep deprivation in the cortex are similarly affected in the hippocampus, including Arc/Arg3.1, Fos, Hnrpdl, Rbm3, and the chaperones Hspa5/Bip and Hspa8 (see for example Refs. 58, 85).

In contrast, our hippocampal microarray study did not find induction of either Homer1a or Zif268/Egr1/NGFI-A, known markers of sleep deprivation in cortex (52, 85), suggesting that there may be important differences in the patterns of gene expression induced by sleep deprivation in different areas of the brain. This is not surprising, given a microarray study showing differential effects of sleep deprivation on gene expression in cortex, basal forebrain, and hypothalamus (79), and a previous study showing in rat brain that sleep deprivation upregulates Zif268 in cortex but downregulates it in the hippocampus (67). Recent meta-analysis of the available genomic data following sleep deprivation in the cortex has revealed a core set of overlapping genes among studies, consisting of just 91 genes (85). We assessed the agreement of our study in the hippocampus with these 91 consensus genes that were differentially expressed in the cortex following sleep deprivation (85). We observed 40 exact matches (44% agreement) and 15 members of the same gene family (60% agreement) (Supplemental Table S4). We show higher agreement with the consensus list than do previous microarray studies (12, 51, 52) or the Allen Brain Institute (80). Thus, our analysis identified a more reliable set of sleep deprivation target genes than has any other previously available dataset. It is likely that the matches with the consensus list represent genes that are induced by sleep deprivation in multiple brain areas, whereas the remaining genes from the consensus list may include genes that are not regulated by sleep deprivation in the hippocampus.

A principal finding of our study is that sleep deprivation appears to inhibit protein synthesis and that this may occur in two ways. First, our microarray results show that, at the transcript level, sleep deprivation downregulates genes involved in translational control. This includes translation initiation factors (Eif4e2 and Eif5) and genes linked to mRNA processing and transport (Rprd2, Rbm3, Hnrpdl, Cirbp, RbmX, and Denr) (19, 44, 53, 54, 73). Functional annotation clustering of gene expression results supports the conclusion that a prominent effect of sleep deprivation is to regulate transcript levels of genes involved in both RNA binding and translation (Fig. 3, Supplemental Table S2). Enriched pathway analysis identified insulin signaling as a key network affected by sleep deprivation, which included components of the mTOR translation regulatory pathway (Table 1, Fig. 4). TFBS analysis further shows that transcription factors may specifically regulate downregulated genes in a coordinated fashion (Supplemental Table S3). In addition to these effects on gene expression, sleep deprivation also appears to impact translation initiation via posttranscriptional alterations in translation regulatory mechanisms. This is supported by the observations that levels of total and phosphorylated mTOR protein decrease after sleep deprivation (Fig. 5), whereas our microarray data show that mTOR transcript levels are unchanged (see Supplemental Table S1).

Our findings are consistent with earlier observations that sleep promotes brain protein synthesis (13, 51, 60, 61, 68, 89). Protein synthesis is a crucial step in both the consolidation of hippocampus-dependent memory and the maintenance of long-lasting hippocampal synaptic plasticity (reviewed in Refs. 2, 37, 45). Inhibition of the regulator of protein synthesis mTOR impairs long-lasting forms of plasticity and several forms of memory in the rodent (7, 64, 71, 75), and enhanced mTOR function has been linked to improved memory (16, 39). Of note, researchers studying a developmental form of visual cortex plasticity in cats have found that sleep helps consolidate synaptic plasticity in vivo (24) and that pharmacological mTOR inhibition specifically prevents the consolidation of plasticity that occurs during sleep (72). Hence, it is possible that sleep-dependent memory consolidation is mediated in part by mTOR-dependent protein synthesis. Interruption of this process might therefore contribute to the effects of sleep deprivation on hippocampal plasticity and memory. Future investigation will be needed to determine the molecular mechanisms by which sleep deprivation reduces total mTOR protein and phosphorylation, and what downstream targets of mTOR are affected.

We performed meta-analysis that found little overlap between the genes regulated by sleep deprivation in our microarray study and the proteins identified as being regulated in the mouse cortex after sleep deprivation using proteomics (65). Only two of the 43 proteins match our list of sleep deprivation-regulated genes, one coding for a common fragment on heat shock protein 8 (NP_112442.2) and similar to heat shock cognate 71 kDa protein (XP_483871.1), and one coding for Secernin 1 (NP_081544.1). Comparison of our data with available proteomic studies in rats following sleep deprivation (5, 66) also shows no overlap. It is difficult to say if this lack of correspondence could be due simply to limits of protein detection by proteomics, because the three proteomic studies mentioned above only attempted to identify spots with differential expression. Similarly, proteomic studies of the effect of sleep deprivation carried out in rats only analyzed a limited number of spots with higher abundance in sleep-deprived animals (5, 66). Therefore, a hit in our microarray might not have been detected in these proteomic studies because it was not detectable as a spot on the gel, because it was present in a spot with other peptides that occluded its change in expression, or because it was present but was not significantly altered by sleep deprivation at the protein level. If the third case is true, the minimal overlap between available transcriptomic and proteomic studies could support the conclusion that sleep deprivation stalls translation via mTOR, creating a lack of correspondence between transcript and protein levels. mTOR regulates cap-dependent translation initiation, which involves the majority of eukaryotic transcripts. However, the exact subset of genes that are regulated at the translational levels by mTOR activation is not known. It is interesting to note that the limited overlap between our microarray and the Pawlyk et al. (65) mouse proteomic study corresponds to proteins belonging to the UPR. It is known that the translation of proteins that allow the cell to cope with transient stress can be cap-independent (50) and thus mTOR-independent. This could explain why that overlap exists. As proteomic approaches improve (78), it would be interesting to compare the effects of sleep deprivation on mRNA and protein levels on a broad scale in hippocampal tissue.

The current study focused primarily on gene expression changes at the end of a 5-h period of sleep deprivation, with additional testing of select genes after 2.5 h of recovery. Therefore, in future studies it will be of interest to examine a time-course of these effects, to determine at what point during sleep deprivation particular genes are targeted, and for what duration. For example, are the genes that are induced after 5 h of sleep deprivation upregulated for the full 5 h? And would they return to baseline with continuing sleep deprivation? The data shown in Fig. 6 also demonstrate that not all genes recover in the same time following sleep deprivation, and it would be interesting to expand on this finding in future studies. Because some genes had not yet recovered in the time it takes for sleep debt to dissipate (25, 40), it could indicate that genes that do not recover in that time frame contribute to other more long-lasting consequences of sleep deprivation. It could also be useful to extend the current analysis by comparing sleep-deprived and control samples to a tether point at the start of the deprivation period. This protocol could answer whether mRNA levels for particular genes are rising or falling in sleep-deprived and control animals relative to the absolute level where they started, rather than just relative to each other.

We have focused this article on the effects of sleep deprivation on protein synthesis, but our data point to regulation of additional cellular processes and signaling pathways that will be of interest to study in more detail. To give one example, it is becoming evident that metabolism is deeply affected by sleep and sleep disturbances (reviewed in Refs. 3, 26, 46, 47, 81), and the insulin signaling network identified by our enriched pathway analysis is crucial in metabolic control. In fact, studies in humans have found connections between short sleep duration and diabetes onset (6, 27), and even one night of sleep restriction can affect insulin resistance (18). Based on our identification of a set of individual disrupted components of this signaling pathway, future studies may be able to determine how sleep deprivation disrupts insulin signals.

In conclusion, this is the first study to perform a genome-wide analysis on the effects of sleep deprivation on gene expression in the mouse hippocampus, and we have identified many genes that had not been previously linked to either sleep or sleep deprivation. Altered genes were significantly clustered by function, with one of the primary regulated cellular processes being protein synthesis. Supporting this bioinformatic approach were our novel findings that levels and activation of the translational regulator mTOR were downregulated by sleep deprivation in the hippocampus. This work identifies a crucial signal molecule in plasticity and memory as a target of sleep deprivation in the hippocampus, potentially explaining why a brief period of sleep deprivation specifically disrupts protein synthesis-dependent forms of plasticity and memory storage.

GRANTS

This research was supported by National Institutes of Health Grants GM-07517 (to C. G. Vecsey; M. Nusbaum, PI), HL-07953 (to C. G. Vecsey; A. I. Pack, PI), MH-090711 (to C. G. Vecsey), NS-007413 (to L. Peixoto; M. Robinson, PI), T32HL-007953 (to M. Wimmer; A. I. Pack, PI), K12GM-081259 (to J. H. K. Choi; Y. Paterson, PI), P50AG-017628 (to T. Abel; A. I. Pack, PI), and R01GM-085226 (to S. Hannenhalli).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: C.G.V., L.P., J.H.K.C., M.W., and T.A. conception and design of research; C.G.V., L.P., J.H.K.C., M.W., D.J., P.J.H., J.B., K.M., and A.J.P. performed experiments; C.G.V., L.P., J.H.K.C., M.W., D.J., and S.H. analyzed data; C.G.V., L.P., J.H.K.C., M.W., D.J., and T.A. interpreted results of experiments; C.G.V., L.P., and J.H.K.C. prepared figures; C.G.V. and L.P. drafted manuscript; C.G.V., L.P., J.H.K.C., M.W., D.J., and T.A. edited and revised manuscript; C.G.V., L.P., J.H.K.C., M.W., D.J., and T.A. approved final version of manuscript.