Shift workers are at increased risk of metabolic morbidities. Clock genes are known to regulate metabolic processes in peripheral tissues, eg, glucose oxidation.
This study aimed to investigate how clock genes are affected at the epigenetic and transcriptional level in peripheral human tissues following acute total sleep deprivation (TSD), mimicking shift work with extended wakefulness.
In a randomized, two-period, two-condition, crossover clinical study, 15 healthy men underwent two experimental sessions: x sleep (2230–0700 h) and overnight wakefulness. On the subsequent morning, serum cortisol was measured, followed by skeletal muscle and subcutaneous adipose tissue biopsies for DNA methylation and gene expression analyses of core clock genes (BMAL1, CLOCK, CRY1, PER1). Finally, baseline and 2-h post-oral glucose load plasma glucose concentrations were determined.
In adipose tissue, acute sleep deprivation vs sleep increased methylation in the promoter of CRY1 (+4%; P = .026) and in two promoter-interacting enhancer regions of PER1 (+15%; P = .036; +9%; P = .026). In skeletal muscle, TSD vs sleep decreased gene expression of BMAL1 (−18%; P = .033) and CRY1 (−22%; P = .047). Concentrations of serum cortisol, which can reset peripheral tissue clocks, were decreased (2449 ± 932 vs 3178 ± 723 nmol/L; P = .039), whereas postprandial plasma glucose concentrations were elevated after TSD (7.77 ± 1.63 vs 6.59 ± 1.32 mmol/L; P = .011).
Our findings demonstrate that a single night of wakefulness can alter the epigenetic and transcriptional profile of core circadian clock genes in key metabolic tissues. Tissue-specific clock alterations could explain why shift work may disrupt metabolic integrity as observed herein.
Animals studies have convincingly demonstrated that the circadian clock allows gene expression to coincide with anticipated metabolic requirements throughout day/night variations via CLOCK and BMAL1 as positive transcriptional regulators and PERIOD and CRYPTOCHROME as negative transcriptional regulators (1). The lack of clock genes, even when ablated only in skeletal muscle or adipose tissue (2, 3), results in systemic metabolic perturbations in animal models (4). These metabolic responses include hyperglycemia and insulin resistance, and can also result in obesity and type 2 diabetes in animals (3–5). As reviewed in Cedernaes et al (6) and Schmid et al (7), similar metabolic phenotypes have been observed in humans subjected to experimental paradigms mimicking night shift work, comprising reduced energy expenditure, impaired systemic glucose disposal, and increased food intake. Over time, these conditions may thus result in metabolic dysregulation and weight gain (6, 7). Although shortened sleep leads to genome-wide changes in the leukocyte transcriptome comprising clock genes (8), the influence of overnight wakefulness, as occurs in night shift work, on the circadian machinery in tissues critically involved in whole-body energy homeostasis is, however, unknown. The importance of this research is highlighted by the fact that today, at least 15% of the workforce—numbering 15 million in the United States alone—carry out shift work, with job activities scheduled during the biological night.
With this background, we characterized the effects of one night of sleep deprivation on gene expression and DNA methylation of core circadian clock genes in peripheral tissues. DNA methylation of gene promoters and promoter-interacting enhancers is one epigenetic mechanism involved in the control of gene expression (9) and is a malleable process following acute lifestyle interventions (10). We obtained subcutaneous adipose tissue and skeletal muscle biopsies from fasted healthy young men following both acute sleep deprivation and normal sleep. In addition, fasting serum cortisol and plasma glucose were measured, the latter before and 120 minutes after an oral glucose tolerance test (OGTT).
Materials and Methods
Study design
This randomized crossover within-subject trial was conducted from March through September 2013 at Uppsala Biomedical Centre, Uppsala University, Sweden. The sessions were relatively evenly distributed across the study period. Study procedures and written consent forms were approved by the Regional Ethical Review Board in Uppsala (EPN 2012/477). The study was conducted in accordance with the Helsinki Declaration. Each enrolled participant voluntarily signed the consent form.
Participants
Sixteen of 17 enrolled subjects participated in two sessions of this study. Participants were of self-reported good health, free from chronic medical conditions or chronic medication, nonsmokers, and had normal sleeping habits (7–9 h of sleep/night; Pittsburgh Sleep Quality Index score ≤ 5) (extended screening protocol in Supplement Part 1).
Study protocol and interventions
All 16 participants engaged in two conditions (acute sleep deprivation vs sleep), in which each condition was separated by at least 4 weeks. Participants came in a semifasted state (fasted since 1500 h) to the laboratory two evenings before each session's final experimental morning, and remained in the laboratory under constant supervision until the end of the experimental session (ie, approximately a 42-h laboratory stay).
Participants were provided with breakfast, lunch, and dinner during their 24-hour baseline period (each meal providing one third of the participants' individually calculated energy requirements; based on the Harris-Benedict equation factored 1.2 for light physical activity), and had an 8.5-hour sleep opportunity during the first night (2230–0700 h). During the first baseline day, participants were provided with two standardized and supervised 15-minute walks. During nonexperimental time periods, participants were confined to their rooms but were free to engage in sedentary-level activities.
Randomization to the first experimental condition (sleep or acute sleep deprivation) was generated by drawing lots, with a fixed block size of 2 and allocation ratio of 1:1. Participants were randomly assigned after having been screened by J.C. as eligible, and were scheduled in pairs for the next available session slot. The allocation sequence was only known by one of the researchers (C.B.) but was concealed from the participants, with the experimenters only notified 2 weeks in advance of each new session for experimental preparation. Participants were blinded to the experimental condition (sleep or acute sleep deprivation) until 90 minutes in advance of onset of the nighttime intervention, which took place during the second night (2230–0700 h). During this period in the sleep condition, room lights were kept off and sleep was monitored. In contrast, in the sleep deprivation condition, participants were under constant supervision 2230–0700 h to ensure wakefulness, remaining bed-restricted and fasted.
Blood sampling, biopsy collection, and OGTT
After fasting, blood samples were obtained at 0730 h. Tissue biopsies were also obtained in the fasted state, 2–3 hours after subject wake-up time, with the collection of the adipose tissue preceding that of the skeletal muscle. Following the biopsy collection, participants completed a 75g OGTT (further details provided in the Supplement).
DNA extraction and epigenetic analyses
DNA extraction and epigenetic analysis with Illumina's HumanMethylation450 BeadChip are further described in Supplement Part 1.
DNA methylation preprocessing consisted of probe filtering (removal of probes with missing β-values; probes with less than 75% of samples with detection P < .01; or nonspecific or single nucleotide polymorphism–coinciding probes), followed by adjustment of type I and type II probes using BMIQ (11), and removal of batch effects using ComBat (12). We ran four pairs of technical replicates, including at least one from each experimental condition (sleep and acute sleep deprivation), to estimate the inner variability of each probe. We only considered for further analysis the probes for which at least half of the subjects showed a methylation difference between conditions greater than the mean difference in technical replicates.
CpG sites within 1500 bp of the transcription start site of CLOCK, ARNTL, CRY1, and PER1 were analyzed (15 CpG sites for adipose tissue and 9 nine for skeletal muscle). The promoter is a key part of a gene, but enhancers also prominently contribute to the regulation of gene expression (13). To identify putative enhancers of CLOCK, ARNTL, CRY1, and PER1, we inferred chromatin states in adipose nuclei and skeletal muscle, and mapped long-range interactions in five different cells lines, with three different transcription factors (14). CpG sites located in chromatin states indicative of enhancers in adipose nuclei and skeletal muscle and in regions having long-range interactions with the promoters of CLOCK, ARNTL, CRY1, and PER1, were also analyzed (six CpG sites for adipose tissue and four for skeletal muscle).
Methylation levels are presented as β-values (ranging from zero to one, corresponding to zero and 100% methylation, respectively). P-values were adjusted for multiple testing according to the Benjamini-Hochberg method within each tissue (15).
RNA extraction and qPCR analysis of gene expression
Methods used for RNA extraction and qPCR analysis of gene expression are described in further detail in Supplement Part 1. The gene expression of CLOCK, ARNTL, CRY1, and PER1 was analyzed with qPCR in adipose tissue and skeletal muscle. All analyses were run in duplicates (primer information in Supplement Part 2). The ΔCt method was used to normalize data (16).
Statistics
Normal-distribution criteria of analyzed data were assessed with Kolmogorov-Smirnov's test of normality. Normally distributed data was analyzed with paired Student t tests, whereas nonnormally distributed variables were analyzed with Wilcoxon signed-rank test. Methylation data was analyzed using the software package R (version 3.1); we used the log2 ratio of the intensities of methylated probe vs unmethylated probe, also called M-value, which is more statistically valid for the differential analysis of methylation levels (17). All other data was analyzed using the software SPSS (version 21; SPSS Inc.) and are presented as means ± SD. Two-sided P < .05 were considered significant. For the adipose tissue, one individual was excluded for all gene expression analyses (expression values greater than mean + 2 SD for several genes). The significance values were, however, not changed when the analysis was run with or without this subject (data not shown). For PER1 in skeletal muscle, an outlier was excluded from both conditions (expression values in the sleep deprivation condition greater than mean + 2 SD), but significance values were not altered when the analysis was run with or without this subject (data not shown).
Results
Of 17 enrolled subjects, 16 completed participation in both sessions (sleep and acute sleep deprivation). One participant was excluded from later analysis due to insufficient sleep (< 7 h) in the sleep condition. Fifteen participants were therefore included in the final analysis (age, 22.3 ± 1.9 y; body mass index, 22.6 ± 1.8 kg/m2). Sleep data are presented in Supplement Part 3.
Effect of acute sleep deprivation on methylation and expression of circadian genes in adipose tissue and skeletal muscle
Methylation levels at cg04674060 (+15%; adjusted P = .036) and cg19308989 (+9%; adjusted P = .026; both CpG sites located in enhancers interacting with the promoter of PER1), and at cg20193872 (located in the promoter of CRY1; +4%; adjusted P = .026), increased after acute sleep deprivation, compared with the sleep condition, in adipose tissue (shown in Figure 1). In skeletal muscle, the investigated CpG sites were not altered (detailed probe results in Supplement Part 4).
Methylation levels after sleep and acute sleep deprivation in adipose tissue and skeletal muscle.
Methylation levels in two putative enhancers interacting with the promoter of PER1 (probes cg04674060 and cg19308989) were increased following overnight wakefulness (ie, acute sleep deprivation), in adipose tissue (SCAT) compared with after sleep. Methylation levels in the promoter of CRY1 (probe cg20193872) were also increased following overnight wakefulness (acute sleep deprivation) compared with after sleep, in adipose tissue (SCAT). No differences were seen, however, between the two conditions in skeletal muscle (VLM). Methylation levels are shown as beta values (β-value; ranging from zero to one, corresponding to zero and 100% methylation, respectively). Horizontal line represents median, box interquartile range, whiskers represent spread of remaining values. Two points that are linked by a line show the difference in methylation levels in overnight wakefulness vs sleep conditions for each individual. *, P < .05; n = 15 for all analyses. Abbreviations: SCAT, sc adipose tissue; VLM, vastus lateralis muscle.
In skeletal muscle, mRNA expression of BMAL1 and CRY1 was decreased following acute sleep deprivation (−18 and −22% compared with expression levels found after sleep; P = .033 and P = .047, respectively; see Figure 2 and Supplement Part 4). Skeletal muscle CLOCK or PER1 gene expression was unaltered. Moreover, the adipose tissue genes were unaltered following acute sleep deprivation.
mRNA expression of core clock genes after sleep and acute sleep deprivation in adipose tissue and skeletal muscle.
mRNA expression of BMAL1 and CRY1 was down-regulated in skeletal muscle (VLM) from humans following overnight wakefulness (ie, acute sleep deprivation) compared with after sleep. No differences between the two conditions were found for the other genes or in the adipose tissue (SCAT). Values are shown as expression levels relative to the control condition (sleep). Horizontal line represents median, box interquartile range, whiskers represent spread of remaining values. Two points that are linked by a line show the difference in mRNA expression in overnight wakefulness vs sleep conditions for each individual. *, P < .05; n = 14 for all analyses in adipose tissue; n = 15 for all analyses in skeletal muscle except for PER1 (see main text for descriptions of excluded values). Abbreviations: SCAT, sc adipose tissue; VLM, vastus lateralis muscle.
Effect of acute sleep deprivation on fasting cortisol and glucose tolerance
Following acute sleep deprivation, fasting serum cortisol concentrations were decreased at 0730 h (2449 ± 932 vs 3178 ± 723 nmol/L; P = .039), compared with after sleep. Plasma glucose concentrations at 120 minutes post-OGTT were higher following acute sleep deprivation, compared with the values obtained after sleep (pre-OGTT: 5.36 ± 0.30 vs 5.38 ± 0.36 mmol/l; P = .705; post-OGTT: 7.77 ± 1.63 vs 6.59 ± 1.32 mmol/l; P = .011).
Discussion
We determined the effect of one night of wakefulness, as occurs during night shift work, on DNA methylation and mRNA expression of key circadian genes (ie, BMAL1, CLOCK, CRY1, and PER1) in human skeletal muscle and adipose tissue. We provide evidence that acute sleep deprivation increases promoter methylation and reduces transcription of circadian genes in a tissue-specific manner. Our analysis reveals increased methylation of transcription-regulating regions of PER1 and CRY1 in adipose tissue and reduced gene expression of CRY1 and BMAL1 in skeletal muscle. We also observed an impaired glucose response following an OGTT after acute sleep deprivation. Our results suggest that acute sleep loss alters clock gene regulation, concomitant with deleterious metabolic effects, which are differential, rather than uniform across key peripheral metabolic tissues in healthy humans.
Our results of altered DNA methylation for promoter and promoter-interacting enhancer regions of core clock genes in adipose tissue suggest that acute sleep deprivation can cause acute epigenetic remodeling of the circadian clock. Similar acute epigenetic changes occur following other types of physiological or metabolic interventions, including acute high-intensity exercise (10). We provide additional evidence that challenges the conventional view that epigenetic regulation is largely a mitotically stable process resistant to the effect of environmental factors. Hypermethylation of core clock genes in humans is linked to insulin resistance in humans (18), and this has also partially been observed in blood of people who chronically work shifts (19). Given that the circadian clock affects key metabolic processes (1), our results suggest that sleep loss–induced hypermethylation of PER1 and CRY1 in adipose tissue may contribute to glucose intolerance as measured by the 120-minute post-OGTT glucose value.
We found that mRNA expression of the core clock genes BMAL1 and CRY1 was decreased in skeletal muscle following acute sleep deprivation. Similar changes occur in circulating leukocytes following longer periods of shortened sleep in humans (8). Skeletal muscle–specific deletion of Bmal1, or global deficiency of Cry1, impairs insulin sensitivity and glucose metabolism in mouse models (2, 20). Moreover, clock gene expression is altered in peripheral blood cells from type 2 diabetic vs nondiabetic patients (21), with an inverse correlation between clock gene expression (BMAL1, PER1, and PER3) and glycosylated protein (HbA1c) level noted. Thus, our observed transcriptional changes in circadian clock genes in skeletal muscle in response to acute sleep deprivation may impair glucose tolerance.
Although the design of our study did not allow us to ascertain the molecular cause of the observed epigenetic and transcriptional changes in skeletal muscle and adipose tissue following sleep loss, several putative candidate mechanisms can be implicated. For instance, whereas glucocorticoid levels may be slightly elevated during nocturnal wakefulness (22), glucocorticoids—as also shown in our study—are reduced during typical awakening hours (eg, between 0700 and 0800 h) (23). Glucocorticoids reset circadian rhythms of peripheral circadian clocks (24). Thus, resetting of peripheral circadian clocks may be hampered by a blunted cortisol awakening response after acute sleep deprivation.
At both the epigenetic and transcriptional level, we demonstrate tissue-specific alterations in core clock genes under conditions of acute sleep deprivation, consistent with animal studies, in which the circadian machinery exhibits tissue-specific changes in rhythm following shift-work-mimicking sleep-wake paradigms (25). Such internal desynchrony has been hypothesized to underlie metabolic effects of shift work (26–28). The physiological relevance of tissue-specific circadian clocks is further supported by aforementioned and other animal studies in which core clock genes have been ablated or rescued in a tissue-specific manner (3, 29), (eg, an adipose tissue–targeting Bmal1 deletion in mice resulting in an obese phenotype) (3). Furthermore, insulin-dependent peripheral tissues, ie, adipose tissue and skeletal muscle, shift toward a diabetes-like phenotype following sleep loss (30, 31). Our tissue-differential effects further reinforce the notion that acute circadian misalignment can produce desynchrony of peripheral circadian clocks, with possibly tissue-specific downstream metabolic effects.
Limitations
Several limitations should be kept in mind when interpreting our results. Lights were on in the TSD condition but not in the sleep condition (∼300 lux vs darkness). Given that light can entrain the human circadian clock (32), our experimental design does not allow us to disentangle if the observed effects of overnight wakefulness on core circadian genes were either driven by loss of sleep, light exposure, or both. However, it is important to note that our experiment aimed at mimicking night shift work, which is typically performed under ambient light exposure. Another limitation of our study is that expression and methylation of clock genes was measured only at a single time point, ie, under fasting conditions in the morning following each sleep intervention. Thus, our study does not allow firm conclusions on how the circadian pattern of expression and methylation pattern of clock genes is influenced by overnight wakefulness. This would, however, have required repeated tissue sampling, which is much more feasible using animal models; models that can also be maintained longer in a fasted state to avoid the entraining effect of meals on clock genes. Finally, hypermethylation of transcription-regulatory regions of core circadian genes in the adipose tissue were not paralleled by concomitant reduced expression of these genes. Given that we only sampled biopsies at one time point, a possible explanation might be that acute promoter hypermethylation altered circadian gene mRNA expression at subsequent points in the sleep deprivation condition, (ie, following biopsy collection). Supporting this assumption are results from a separate study examined the effects of an acute bout of exercise on skeletal muscle promoter methylation and corresponding gene expression (10). There, remodeling of promoter methylation of PGC-1α, a gene involved in the circadian machinery of the skeletal muscle (33), was accompanied by a delayed (ie, 3 h later), but not concomitant change in gene expression (10).
Conclusions
One night of sleep loss results in hypermethylation of regulatory regions of key clock genes. These effects are tissue specific, and occur in adipose tissue, but not in skeletal muscle. Gene expression differences were observed for the investigated clock genes in skeletal muscle, but not in adipose tissue. Shift work is associated with many of the same phenotypes observed in transgenic animal models in which the circadian clock is disrupted, (eg, glucose intolerance) (34–36). This suggests that our findings of altered peripheral clocks at the epigenetic and transcriptional level, with ensuing glucose intolerance, following acute sleep loss may contribute to metabolic disruptions typically observed in humans with activities regularly scheduled during times that produce chronic desynchrony between tissue-specific clocks.
Perspectives
Given that recurrent partial sleep deprivation decreases insulin sensitivity at the systemic and adipose tissue level in humans (30, 37), future studies to examine whether similar changes occur under conditions of recurrent partial sleep deprivation are of interest. Using repeated biopsy collection, eg, also under insulin-stimulated conditions, may decipher the time-dependent dynamics of peripheral circadian misalignment and how this might relate to metabolic perturbations, including impaired glucose tolerance. Whether our findings can be extrapolated to females or older participants is currently unknown and warrants investigation. Large interindividual differences were observed in our data for how sleep deprivation altered peripheral tissues' clock-gene methylation and gene expression. Contributing factors may be subjects' chronotype—linked to differential responses' to sleep deprivation—or seasonality; with recent studies demonstrating seasonal circadian clock gene variability in animals as well as humans (38), supported by summer-winter variation in human adipose tissue. Ambient light can influence circadian rhythms (39). Thus, light can also resynchronize peripheral circadian rhythms in the absence of a functioning central pacemaker (40); and enhance the cortisol awakening response (41). The mechanism by which different ambient light exposures influences the peripheral clock under conditions of extended wakefulness remains to be investigated. Finally, although the absence of nighttime meals in our sleep deprivation condition precludes the synchronizing influence from such a zeitgeber on peripheral clocks (42), nighttime meal intake is common in shift workers and may thereby modulate effects on tissue-specific circadian clocks.
Acknowledgments
Methylation profiling was performed by the SNP&SEQ Technology Platform in Uppsala. The platform is part of Science for Life Laboratory at Uppsala University and supported as a national infrastructure by the Swedish Research Council. We thank Dr Brendan Egan and Dr Alexander Chibalin for valuable discussions regarding experimental procedures, and Olof Ros and Jon Brandell for their help in conducting the experiments.
Author Contributions: J.C. had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. J.C. and C.B. designed the study; J.C. and C.B. wrote the protocol; J.C. and J.E.B. collected the data; J.C., S.V., J.E.B., H.V., and C.B. conducted the analyses; J.C., M.E.O., S.V., J.E.B., H.V., S.D., J.Z., H.S., and C.B. interpreted the data; J.C., M.E.O., S.V., J.E.B., H.V., S.L.D., J.R.Z., H.B.S., and C.B. contributed to writing; J.C., M.E.O., S.V., J.E.B., H.V., S.L.D., J.R.Z., H.B.S., and C.B. approved the final manuscript.
This study was registered in ClinicalTrials.gov as trial number NCT01800253.
This work was supported by the Swedish Brain Research Foundation (J.C., C.B.), AFA Försäkring (C.B.), the Novo Nordisk Foundation (C.B., J.Z.), the Swedish Society of Medicine (J.C.), Magnus Bergvall's Foundation (J.C.), Thuring's Foundation (J.C.), Tore Nilsson's Foundation (J.C.), and the Swedish Research Council (H.B.S., J.R.Z.). The funding sources had no input in the design and conduct of this study, in the collection, analysis, and interpretation of the data, or in the preparation, review, or approval of the manuscript.
Disclosure Summary: The authors have nothing to disclose.
