Early-life physical activity reverses metabolic and Foxo1 epigenetic misregulation induced by gestational sleep disturbance

Abstract

Sleep disorders are highly prevalent during late pregnancy and can impose adverse effects, such as preeclampsia and diabetes. However, the consequences of sleep fragmentation (SF) on offspring metabolism and epigenomic signatures are unclear. We report that physical activity during early life, but not later, reversed the increased body weight, altered glucose and lipid homeostasis, and increased visceral adipose tissue in offspring of mice subjected to gestational SF (SFo). The reversibility of this phenotype may reflect epigenetic mechanisms induced by SF during gestation. Accordingly, we found that the metabolic master switch Foxo1 was epigenetically misregulated in SFo livers in a temporally regulated fashion. Temporal Foxo1 analysis and its gluconeogenetic targets revealed that the epigenetic abnormalities of Foxo1 precede the metabolic syndrome phenotype. Importantly, regular physical activity early, but not later in life, reversed Foxo1 epigenetic misregulation and altered the metabolic phenotype in gestationally SF-exposed offspring. Thus, we have identified a restricted postnatal period during which lifestyle interventions may reverse the Foxo1 epigenetically mediated risk for metabolic dysfunction later in the life, as induced by gestational sleep disorders.

the term “epigenetics” refers to DNA and histone modifications and noncoding RNA, which result in heritable changes in gene expression among the next generation without a change in the DNA sequence (5, 17, 29). The environment, such as diet and early life experiences, can influence the epigenome. Epigenetic abnormalities have been found to be causative factors in cancer, as well as contributing factors in autoimmune diseases, metabolic pathways, and aging. Disruption of the balance of epigenetic networks can lead to inappropriately heightened expression or silencing of genes, resulting in “epigenetic diseases”. In contrast to genetic events, the potential reversibility of epigenetic states offers exciting opportunities for new therapeutic targets (9, 19).

Sleep disorders are highly prevalent among pregnant women in late gestation and are characterized by increased awakenings during the night and sleep maintenance insomnia, which are potentially fraught with adverse pregnancy outcomes (13, 21, 38, 45). Among the disorders that impose additional disruption of sleep integrity during pregnancy, sleep-disordered breathing is particularly prevalent and has emerged as a contributor to gestational diabetes mellitus, preeclampsia, as well as adverse perinatal outcomes (28, 42). However, the impact of poor sleep quality and continuity during pregnancy could transcend the gestational period and adversely affect children, even late in life (7). Sleep disorders in adults are associated with insulin resistance, glucose intolerance, and Type 2 diabetes (52). Although a multiplicity of gestational exposures during late pregnancy is associated with an increased risk of chronic diseases in the offspring (27), little is known about the possible consequences of late-gestation pregnancy sleep disruption on offspring's metabolic function. We address this question in a murine model of gestational sleep disturbance (23) and hypothesize that excessive sleep fragmentation (SF) induces misregulation of metabolic genes in the offspring by epigenetic mechanisms, which may contribute to the “metabolic syndrome” (MetS) during adulthood. We further posited that such changes are potentially reversible by increased physical activity implemented during a selective temporally restricted developmental window.

We focused on Foxo1 (Forkhead box O1) gene, which has been identified as a transcription factor that plays critically important roles in the regulation of gluconeogenesis and glycogenolysis by insulin signaling (47) in liver. Indeed, dysregulation of FOXO1 function has been implicated in diabetes (51) and insulin resistance. In the insulin-resistant murine model, there is increased hepatic glucose production due to a loss of insulin sensitivity, and this is presumably due to unregulated FOXO1 (1). Our epigenetic analyses demonstrate that active epigenetic marks are significantly elevated, and inactive marks decreased at the Foxo1 promoter and at the putative regulatory regions upstream of the gene transcription start site (TSS) in fragmented-sleep offspring (SFo) compared with control offspring (SCo). This epigenetic signature correlated in a temporally regulated fashion with increased Foxo1 expression and elevated levels of its known gluconeogenesis gene targets. Importantly, the epigenetic changes in the Foxo1 were reversed by implementation of physical activity during early, but not later in life in the offspring. Such reversibility correlated with abrogation of the adverse metabolic consequences imposed on the offspring by excessive sleep perturbation in the mother.

MATERIALS AND METHODS

Animals.

All experiments were approved by the University of Chicago's Animal Care and Use Committee. All efforts were made to minimize the number of animals and animal suffering. C57BL/6J mice were purchased from Jackson Laboratories for breeding. The animals were housed in cages with 12:12-h light-dark cycles. Adult breeding pairs aged 3 mo were used to generate only one litter. Day 0 of gestation was defined as the day of plug observation. After birth, litter size was limited to six pups per litter to ensure adequate and standardized nutrition until weaning. Offspring mice were housed in standard conditions with 12:12-h light-dark cycles. All metabolic experiments reported herein included animals from at least six different litters, and epigenetic analyses included mice from four to six different litters. A schematic diagram of the study design is shown in Fig. 1.

Fig. 1.

Experimental design. Pregnant mice were exposed to fragmented sleep (SF) or sleep control (SC) conditions. After lactation and weaning (4 wk after birth), SCo and SFo were subdivided into three groups (n = 8 per group): 1) undergoing early physical activity (PA; 4–8 wk), 2) undergoing late PA (16–20 wk), or 3) not undergoing any PA during 4–24 wk. For the indicated time points, blood and tissue samples were collected from each group of mice, and biochemical and epigenetic analyses were performed.

Sleep fragmentation and physical activity.

The sleep fragmentation device used to induce SF in rodents has been previously described (23, 36, 41) and employs intermittent tactile stimulation of freely behaving mice in a standard laboratory mouse cage, using a near-silent motorized mechanical sweeper (model no. 80391, LaFayette Instruments, LaFayette, IN). To induce moderate to severe sleep fragmentation, we chose a 2-min interval between each sweep, implemented during the light period (7 AM to 7 PM). This paradigm is not associated with any increases in the plasma levels of the stress hormones (36, 41). Pregnant mice were sleep-fragmented from day 14 until day 19 of gestation. Physical activity (PA) was induced, as previously described (16), using a motorized forced exercise/walking wheel system for mice (Lafayette Instruments, Lafayette, IN). The PA groups were exposed to the walking wheeled system for 30 min, 3 times per week, whereas the control groups were placed in the wheeled system and remained in the immobile apparatus for the same period of time. Early PA was implemented between 4 and 8 wk of age, and late PA was implemented between 16 and 20 wk of age.

Body weight and food intake.

Body weight was assessed weekly for a period of 24 wk, always at the same time of the day (middle of the light cycle period). Food intake was carefully recorded daily for each cage starting at week 5 after birth.

Glucose tolerance test and insulin tolerance test.

Both tests were performed at week 24 after birth in a random order, except in a selected subset of mice undergoing late PA exposures that were tested at both 20 and 24 wk. In both tests, animals were fasted for 3 h with water available ad libitum. An intraperitoneal injection of sterile glucose (2 mg/g body wt for the GTT) or an intraperitoneal injection of sterile humulin (0.25 units/kg body wt for the ITT) was administered. Blood recovered from the tip of the tail at different time points (GTT: 0, 15, 30, 60, 90 and 120 min following injection; ITT: 0, 15, 30, 60, 75, 90, 105, 120 min after injection) was tested for glucose levels using an OneTouch Ultra2 glucometer (Life Scan, Milpitas, CA).

Biochemical analyses.

Blood samples were centrifuged at 2,000 g for 20 min at 4°C; subsequently, plasma was centrifuged for 5 min at 15,000 g, and immediately frozen at −80°C until further analysis. Lipid profiles, including total cholesterol and triglycerides (TG), were measured in plasma using Infinity kits (Thermo Scientific).

Gene-specific analysis of differential DNA 5-methyl-CpG and 5-hydroxymethyl-CpG.

Methylated DNA ImmunoPrecipitation (MeDIP) was performed, as previously described (33). Genomic DNA was sonicated to produce random fragments ranging in size from 200 to 1,000 bp. Denatured DNA was immunoprecipitated with antibody against 5-methylcytidine or with 5-hydroxymethyl-CpG antibody. Sonicated input DNA and immunoprecipitated DNA were analyzed by regular quantitative PCR, using specific primers at the region of interest.

Gene-specific analyses of histone modifications.

Cross-linking with formaldehyde was performed, which allowed shearing of chromatin into small-sized fragments after sonication of the cross-linked samples (chromatin fragments ranging in size from 200 to 500 bp). Immunoprecipitation of chromatin fragments with antibodies specific for different histone modifications (Table 1) was performed, as previously described (35). Purified input and immunoprecipitated DNA were amplified and analyzed by regular quantitative PCR, using specific primers at the region of interest (Table 1).

Table 1.

Gene expression assays and the ChIP and MeDIP primers used

	Assays and Primers
Cat. no. 4331182	Mm00446971_m1: Tbp, TATA box binding protein
Cat. no.4331182	Mm00490672_m1 best coverage Foxo1, forkhead box O1
Cat. no. 4331182	Mm00607939_s1 Actb, actin, beta
Cat. no. 4331182	Mm01247058_m1 Pck1, phosphoenolpyruvate carboxykinase 1, cytosolic
Cat. no. 4331182	Mm01166879_m1 Pdk4, pyruvate dehydrogenase kinase, isoenzyme 4
Cat. no.4331182	Mm01208835_m1 Ppargc1a, peroxisome proliferative-activated receptor, gamma, coactivator 1 alpha, Pgc-1α
Cat. no. 4331182	Mm00488755_m1 Fto, fat mass and obesity-associated
List of ChIP, MeDIP primers:
GAPDH cat. no. pp-1045-500, Diagenode
TSH2B cat. no. pp-1042-500, Diagenode
Tbp.F.	GCCACCTCACAAATCTCAAAGG
Tbp.R.	GACTTTCTTGGGACTCACTCTGTAGA
Foxo1.Promoter.F	AAGAAAAATACCCCACCGCC
Foxo1.Promoter.R	AATGGACGCGCGAAGTCTCC
Foxo1.P.Enhancer.F	GTCTTTCTACCTTTGGCTCC
Foxo1.P.Enhancer.R	AGATAAGCCCCAAGTGTTGG
Actb.F.	AGCAATAGCCGGAAAGCCAGATCC
Actb.R.	ACCACTGGGGCTCGCCCTATG
Fscn2.F.	CATCCAGGAGCCACTGAAAT
Fscn2.R.	GACATGGACGCTACCTGCTC

The cDNA products were amplified by real-time TaqMan PCR using the above gene expression assays from Applied Biosystems. The following list of antibodies were used: no. 06-599, anti-acetyl-histone H3 antibody (Millipore), no. 06-866, anti-acetyl-histone H4 antibody (Millipore), no. 07-473, anti-trimethyl-histone H3 (Lys4) antibody (Millipore), no. 12-370, normal rabbit IgG (Millipore), no. 07-436, anti-monomethyl-histone H3 (Lys4) antibody (Millipore), no. 07-449, anti-trimethyl-histone H3 (Lys27) antibody (Millipore), ab4729, Rb pAb to histone H3 acetyl K27 (Abcam), no. C15310210-100, 5-hmC polyclonal antibody (rabbit) (Diagenode) and no. BI-MECY-0100, anti-5-methylcytidine, (AnaSpec). GAPDH, glyceraldehyde 3-phosphate dehydrogenase; TH2B, testis/sperm-specific histone H2B type 1A.

Comparative gene expression analysis.

Total RNA from the samples was isolated and reverse transcribed. cDNA products were amplified by real-time PCR using specific primers (Table 1).

Quantitative real-time PCR data analysis.

DNA samples from input (In) and antibody-bound (IP) fractions were analyzed by real-time PCR using the SYBRGreen PCR master mix (PE Applied Biosystems) and an ABI Prism 7500 sequence detector, according to the manufacturer's protocols. Individual PCRs were carried out in triplicate to control for PCR variation, and the Ct values were collected. Quantification was performed by applying the comparative Ct method, as described by the manufacturer's protocols (PE Applied Biosystems) and in our previous epigenetic studies (34), giving the fold difference of a target sequence (t) in the IP fraction vs. a fixed amount of In DNA as a standard, where IP/In = 2 − Ct = 2 − [Ct(IP) − Ct(In)]. For each primer set, these fold-difference values were corrected by subtraction of the nonspecific signal derived from the nonimmune rabbit IgG chromatin immunoprecipitation (ChIP) (t0): (IP/In) t − (IP/In) t0. In parallel, DNA samples were amplified with primers for an internal control (c). Finally, we normalized the relative abundance of target sequences to the internal control sequence for each individual chromatin immunoprecipitation reaction using the following formula: [(IP/In) t − (IP/In)t0]/[(IP/In) c − (IP/In) c0].

Statistical analyses.

Statistical analyses were conducted using SPSS software (version 18; Chicago, IL) and consisted of either two-way ANOVA for repeated measures followed by post hoc Bonferroni corrections or unpaired t-tests as appropriate. A two-tailed P value <0.05 was considered statistically significant.

RESULTS

Physical activity during early life in SF offspring, but not during late life, reverses metabolic dysfunction induced by late-pregnancy sleep perturbations.

To investigate the consequences of SF during late gestation on metabolism and epigenetics of the offspring, pregnant mice were exposed to a SF paradigm during daylight hours (awakenings from sleep were implemented every 2 min) from day 14 until day 19 of gestation (Fig. 1) (23, 36, 41). The SF paradigm does not elicit increases in stress hormones and after short reductions in rapid eye movement sleep during the initial 12–24 h of SF, sleep recovery occurs with normalization of overall sleep duration and sleep state distribution (36, 41). Furthermore, SF during late gestation did not affect the litter size and birth weight. However, starting around 16–18 wk of postnatal age, we observed that SF-exposed male offspring (SFo; n = 6 different litters) displayed significantly higher body weight compared with sleep control-exposed offspring (SCo; n = 6 different litters; Fig. 2A and Ref. 23), as well as increased food intake (23), but no differences in metabolic expenditure (Fig. 3). At age 24 wk, SFo displayed a MetS-like phenotype, i.e., impaired glucose tolerance (Fig. 2B), reduced peripheral insulin sensitivity (Fig. 2C), higher serum triglycerides (Fig. 2D), and cholesterol (Fig. 2E), illustrating the adverse metabolic effects of gestational SF in the offspring (23).

Fig. 2.

Late-pregnancy SF induces metabolic dysfunction of the offspring. A: SFo (n = 8, ■) accrued higher body weight than SCo (n = 8; □) measured from 0 to 24 wk after birth. Glucose tolerance test (GTT; B) and insulin (ITT; C) tolerance tests in SCo (n = 8, □) and SFo (n = 8, ■) showing altered glucose tolerance and reduced insulin sensitivity in gestational SF-exposed offspring measured at age 24 wk. SFo (n = 8; solid bars) had higher serum triglycerides (D) and cholesterol (E) than the SCo (n = 8, open bars) at 24 wk after birth. Results are expressed as means ± SE; *P < 0.05.

Fig. 3.

No significant differences in energy expenditure in SFo and SCo. Oxygen consumption (V̇o₂) did not significantly differ between offspring of SFo (n = 8; solid columns) and SCo (n = 8; striped columns) mothers during a typical 24-h cycle.

We hypothesized that the MetS-like phenotype recapitulated in current experiments at age 20–24 wk in the SFo (23) may be reversed by PA. Therefore, we assessed the metabolic alterations, if any, in the SFo and SCo subjected to a PA regimen (30-min sessions, 3 times per week), implemented either early (weeks 4 to 8 of age; SFo-PA_early and SCo-PA_early) or later in life (weeks 16 to 20; SFo-PA_late and SCo-PA_late). Importantly, PA during early life reversed the MetS phenotypic alterations in SFo, as evidenced by normalization of body weight (Fig. 4A), decreased food intake (Fig. 4B), normalization of glucose tolerance (Fig. 4C), and visceral adipose tissue mass (Fig. 4D) at 24 wk of age, while late PA of SFo failed to alter the abnormal phenotype (Fig. 4, A–D).

Fig. 4.

PA during early life reversed the metabolic phenotypic alterations in SFo. Measurements of body weight (A), food intake (B), and glucose tolerance tests (C) of SFo_early (n = 8, ○), SFo-PA_early (n = 8, ■), SFo_late (n = 8, open △), and SFo-PA_late (n = 8, ▲) livers was performed as described in Fig. 1. D: PA from week 4 to week 8 of life led to a normalization of visceral fat mass in offspring of SF mothers at 24 wk of age. SCo_early (n = 8; open column), SCo-PA_early (n = 8, open column), and SFo_early (n = 8, open column), and SFo-PA_early (n = 8, open column). *P < 0.05.

SF during late gestation induces offspring misregulation of Foxo1 metabolic gene in liver by epigenetic mechanisms.

The reversibility of the SFo metabolic phenotype suggests an epigenetic mechanism. Epigenetic events play a significant role in enabling the gestational environment to influence the developmental trajectory through modulation of gene expression (5, 17, 29). Sleep disorders may interfere with the normal development of metabolic processes in organs, such as liver, through epigenetic programing, and potentially promote the development of metabolic diseases in modern society (15, 31, 49). We hypothesized that epigenetically driven changes in expression of master regulators of metabolic pathways, such as Foxo1, may partially contribute to the altered metabolic phenotype in SFo mice. FOXO1 is a transcription factor that plays critical roles in the regulation of gluconeogenesis and glycogenolysis by insulin signaling in the liver (26, 47, 51). Mechanisms governing FOXO1 regulation by posttranscriptional protein modifications have been extensively studied. However, Foxo1 regulation at the level of transcription, including chromatin modifications and the presence of any Foxo1 regulatory elements remains virtually unexplored. We assessed previously known epigenetic modifications involved in gene regulation: like histone modifications and DNA methylation marks by ChIP and MeDIP analysis, respectively, applying our previously developed approach for single gene epigenetic analysis (33, 35). We first validated this approach in both SCo and SFo using mouse liver samples (Fig. 5). For precise estimation of specific gene enrichment or depletion of each chromatin mark in both SCo and SFo samples, we performed a normalization procedure of our data based on the presence of these individual epigenetic marks in the Tbp gene, since neither its epigenetic state nor its expression were affected by SF or by physical activity intervention (data not shown). We profiled the epigenetic landscape of the Foxo1 locus (Fig. 6, A–G) in livers of 20–24-wk-old SCo and SFo mice and included animals from four to six different litters. An enrichment of DNA and chromatin marks associated with open chromatin and active transcription (histone H3 acetylated at the NH₂ terminus, hyperacetylated form of histone H3); histone H4 acetylated on lysines 5, 8, 12, and 16 (hyperacetylated form of histone H4); histone H3K4 trimethylation (H3K4m3); and 5-hydroxymethyl-CpG at the Foxo1 promoter emerged in SFo mice (Fig. 6, A–C). The inactive epigenetic mark 5-methyl-CpG was decreased in the SFo Foxo1 (Fig. 6D). We did not find any significant epigenetic differences between SCo and SFo in two control genes: Actb and Fscn2, but we did observe significant epigenetic differences that were inversely correlated with the Foxo1 epigenetic marks in another commonly employed control gene: Gapdh (Fig. 6, A–D). These results further confirmed the validity of our epigenetic analyses in SCo and SFo samples. Taken together, these observations support the presence of specific epigenetic signatures in the offspring of pregnant mice subjected to late-gestation SF. Moreover, epigenetic modifications in a region ∼2 kb upstream of the Foxo1 transcription start site (TSS) suggested the presence of a putative enhancer (Fig. 6, E–H) that was preferentially active in SFo compared with SCo (Fig. 6, A–G). Publicly available human and mouse ChIP-Seq data from ENCODE used for genome-wide identification of enhancers confirmed that this region has chromatin characteristics of an enhancer, namely, is enriched in histone H3K4 monomethylation and depleted in histone H3K4 trimethylation (Fig. 6H). The importance of enhancer elements during development provides an indication that their deregulation, including epigenetic misregulation, is likely to impose phenotypic consequences (6, 10, 44). Interestingly, the epigenetic marks in this locus significantly differed between SFo and SCo. We found an increase of 5-hydroxy-methyl-CpG (Fig. 6C), which was recently suggested to take part in the function of activated enhancers (40). In addition, by performing ChIP analysis, we observed slight increases of histone H3K4m3 (Fig. 6B), as well as histone modifications ratios of H3K4m3 (active mark)/H3K4m1 (enhancer mark) and H3K27m3 (inactive mark)/H3K4m1 (enhancer mark) (Fig. 6, E and F) that were specific for SFo and not present in SCo at this Foxo1 putative enhancer locus. Finally, we performed ChIP analyses in SCo and SFo with an antibody specific to histone H3K27 acetylation, a histone mark associated with activated gene enhancers (39) and observed an enrichment at the putative Foxo1 enhancer locus in SFo livers (Fig. 6G). This chromatin signature has been previously implicated as a switch from “poised” to “active” genome enhancers (39).

Fig. 5.

Validation of the epigenetic analysis in SCo and SFo livers, measured at age 20 wk. Methylated DNA immunoprecipitation (MeDIP) of 5-methyl-CpG and chromatin immunoprecipitation (ChIP) of H3K4m3 at the active Gapdh and transcriptionally silenced Tsh2b genes were performed in SCo (n = 4–6; open columns) and SFo (n = 4–6; solid columns) livers, as described in Fig. 6, B and D, but the Gapdh for ChIP and Tsh2b for MeDIP values of SCo and SFo levels were arbitrarily set to 1. The relative gene abundance sequence was calculated as the ratio of its concentration in the immunoprecipitated (IP) fraction to that in the input. Results are shown as means ± SD; *P < 0.05.

Fig. 6.

Late-gestation SF induces offspring misregulation of the Foxo1 in the liver by epigenetic mechanisms. ChIP of histone H3 acetylation (A), histone H3K4 trimethylation (B), MeDIP of DNA 5-hydroxymethyl-CpG (C), and 5-methyl-CpG (D) over the Foxo1 Putative Enhancer (P.Enhancer) and Promoter regions and control genes in SCo (n = 4–6; open columns) and SFo (n = 4–6; solid columns) livers, measured at age 20 wk. The relative gene abundance sequence was calculated as the ratio of its concentration in the IP fraction to that in the input DNA. These fold-difference values were corrected by subtraction of the nonspecific signal derived from the nonimmune rabbit IgG ChIP. It was then normalized to similar data obtained for Tbp (for ChIP and 5-hmC MeDIP) and Tsh2b (for 5-mC MeDIP). The values of SCo levels were arbitrarily set to 1, and results are shown as means ± SD; *P ≤ 0.05. E: Ratio of histones H3K4 tri-methylation/H3K4 monomethylation ChIPs relative abundance at the Foxo1 P.Enhancer was calculated for the SCo (n = 4–6, open column) and SFo (n = 4–6, solid column) livers. ChIP was performed as described in A and B, measured at age 20 wk. F: ratio of histones H3K27 dimethylation/H3K4 monomethylation ChIPs relative abundance at the Foxo1 P.Enhancer was calculated for the SCo (n = 4–6; open column) and SFo (n = 4–6, solid column) liver. ChIP was performed as described in A and B, measured at age 20 wk. *P < 0.05. G: ChIP of histone H3K27 acetylation at Foxo1 P.Enhancer and control gene. H: UCSC Genome Browser presentation of mouse Foxo1 region and histone H3K4m3 and histone H3K4m1 (ENCODE Project).The position of the two ChIP and MeDIP primer sets, spanning the regions of the Foxo1 promoter and putative enhancer, is shown.

The increases in transcriptionally active epigenetic marks at the Foxo1 locus were correlated with increased expression of the Foxo1 in SFo (Fig. 7). Similar correlations were observed with some of the known Foxo1 target gluconeogenesis genes (48) (Pck1, Pdk4) and its coactivator in gluconeogenesis transcription genes Pgc-1α, but not control genes (Actb, Fto). However, Gapdh, which exhibited decreases in active epigenetic marks in SFo (Fig. 6, A–C), showed significantly decreased levels of its expression in SFo compared with SCo (Fig. 7).

Fig. 7.

SF during late gestation induces upregulation in the offspring of the Foxo1 in liver. Expression analyses in SCo (n = 4–6; open columns) and SFo (n = 4–6; solid columns) livers of Foxo1, control genes, and gluconeogenesis genes were measured at age 20 wk. Each bar represents the abundance of mRNA relative to Tbp mRNA. The values of SCo levels were arbitrarily set as 1, and results are shown as means ± SD; *P ≤ 0.05.

The epigenetic abnormalities in the Foxo1 affect the putative regulatory element in early postnatal life, but not the Foxo1 promoter or expression.

To delineate the temporal window during which the aforementioned abnormal states may develop, a time-course analysis of the epigenetic landscape and expression of Foxo1 following gestational SF were assessed (Fig. 8, A–D). We analyzed the presence of histone H3 acetylation, usually distributed at both gene promoters and enhancers (44) and histone H3 trimethylation mostly enriched at gene TSS (44). We found that during early life stages (3 wk after birth) of SFo offspring, epigenetic differences between SFo and SCo mice at the Foxo1 locus are restricted to the putative regulatory element (Fig. 9, A and C), but are not detectable within the gene promoter (Fig. 8, B and D), nor do they affect the level of Foxo1 expression or one of its primary targets Pdk4 (Fig. 8, E and F). The first significant differences between SCo and SFo in the epigenetic abnormalities at Foxo1 promoter and putative enhancer, as well as its gene expression appeared at postnatal age of 8 wk (Fig. 8, A–E). Importantly, these results indicate that the epigenetic abnormalities at Foxo1 in SFo occur in a temporally regulated fashion and precede the metabolic syndrome phenotype, which develops later in life around weeks 20–24 postnatally (Fig. 2, A–E) (23).

Fig. 8.

In early life, the SFo Foxo1 epigenetic abnormalities affect the putative regulatory element. ChIP of histone H3 acetylation over the Foxo1 P.Enhancer (A) and promoter region (B). ChIP of histone H3K4 trimethylation over the Foxo1 PEnhancer region (C) and promoter region (D) in SCo (n = 4–6; open column) and SFo (n = 4–6; solid column) liver were performed as described in Fig. 6, A and B at three different time points (3, 8, and 20 wk). Expression analyses of Foxo1 (E) and its gluconeogenesis target gene Pdk4 (F) in SCo (n = 4–6; open columns) and SFo (n = 4–6, solid columns) livers were performed, as described in Fig. 4 at three different time points (3, 8, and 20 wk of age). Results are shown as means ± SD; *P < 0.05.

Fig. 9.

Late-gestation SF induces gene expression misregulation of the epigenetic chromatin modification enzymes in the offspring. A: epigenetic chromatin modification enzymes RT2 Profiler PCR Array of SCo and SFo mouse liver at age 20 wk. B: list of the most differentially expressed epigenetic chromatin modification enzymes in SFo using a Log2 scale with a cut-off of twofold changes.

To further explore specific chromatin modifications at the Foxo1 in livers of SFo mice, we assessed whether the expression of some of the major epigenetic modifying enzymes is altered. We took advantage of a commercially available array that enables concurrent analyses of 84 genes that operate as epigenetic modifying enzymes and found at least six enzymes that are differentially expressed in SFo compared with SCo (Fig. 9). Although the direct implications of these observations to Foxo1 and to the phenotype in SFo remain unclear, they reinforce, however, the concept that SF during late gestation imposes specific epigenetic alterations in the offspring. These findings may serve as guides for future mechanistic exploration of the function of these six chromatin-modifying enzymes in the metabolic dysfunction induced by gestational sleep disturbance.

Physical activity during early life in SF offspring, but not during late life, reverses Foxo1 epigenetic abnormalities and its gene expression as induced by late-gestation SF exposures.

The potential reversibility of epigenetic states offers opportunities for therapeutic interventions. Epigenetic programming can be reversed by interventions, such as treatment with epigenetic drugs, nutrition, or maternal care (9, 19). It has been recently reported that exercise appears to induce DNA methylation changes in human skeletal muscle (3) and adipose tissue, potentially affecting adipocyte metabolism (43), thereby establishing biological plausibility for potential reversibility of MetS in SFo by epigenetic mechanisms. We hypothesized that the epigenetic abnormalities induced by SF during late gestation may be reversed by PA. Therefore, we assessed changes in Foxo1 epigenetic alterations in SFo and SCo subjected to a PA regimen early (weeks 4 to 8) or late (weeks 16 to 20) in life at a fixed time interval after PA (Fig. 1). Early (Fig. 10, A–D), but not late PA (Fig. 10F), abrogated SF-induced epigenetic changes reported above. The epigenetic effect of PA on Foxo1 was specific for SFo (the control gene Actb showed the same changes in both SCo and SFo epigenetic and gene expression analysis) (Fig. 11, A–C). Foxo1 expression showed that early PA results in reductions in Foxo1 expression in SFo to levels similar to SCo (Fig. 10E) and that such reductions correlate with parallel decreases in FOXO1 downstream regulated genes (Pck1, Pdk4) and Pgc-1α, but not with control genes (Fto, Actb). However, no significant decreases in Foxo1 expression emerged after late PA (Fig. 10G), and the reversal of the epigenetic modifications also failed to occur following late PA (Fig. 10F). Most importantly, SFo Foxo1 reversal epigenetic signature by PA during early life was correlated with the disappearance of the MetS phenotype in SFo (Fig. 4, A–D). Thus, early life interventions, such as regular physical activity during a developmentally favorable window, may reverse epigenetic modifications incurred during pregnancy in the context of perturbed sleep.

Fig. 10.

Epigenetic abnormalities in Foxo1 among SFo mice are reversible by instituting PA during early life. MeDIP of 5-hydroxymethyl-CpG (A), 5-methyl-CpG (B), ChIP of histone H3K4 tri-methylation (C), and histone H3K27 trimethylation (D) over the Foxo1 PEnhancer and promoter (P) regions and expression analyses (E) were performed in SCo_early (SC) (n = 4–6; open columns), SCo-PA_early (SCPA) (n = 4–6; dotted columns), and SFo_early (SF) (n = 4–6; solid columns); SFo-PA_early (SFPA) (n = 4–6; striped columns) liver, as described in Fig. 6 after PA or no PA from 4 to 8 wk. MeDIP of 5-hydroxymethyl-CpG over Foxo1 (F) and gene expression analyses (G) were performed in SCo_late (SC) (n = 4–6; open columns), SCo-PA_late (SCPA) (n = 4–6; dotted columns) and SFo_late (SF) (n = 4–6; solid columns); SFo-PA_late (SFPA) (n = 4–6; striped columns) livers, as described in Fig. 6 after PA or no PA from 16 to 20 wk. Results are expressed as means ± SE; *P < 0.05.

Fig. 11.

The epigenetic effect of the PA on the Foxo1 is SFo-specific and was not seen at a control gene Actb. MeDIP of 5-methyl-CpG (A) and 5-hydroxymethyl-CpG (B) over the control Actb gene promoter and expression analyses (C) were performed in SCo_early (n = 4–6; open columns); SCo-PA_early (n = 4–6; dotted columns) and SFo_early (n = 4–6; solid columns); SFo-PA_early (n = 4–6; striped columns) liver as described in Fig. 6, after PA or no PA from 4 to 8 wk. Results are expressed as means ± SE. *P ≤ 0.05.

DISCUSSION

There is increasing evidence that chronic diseases like Type 2 diabetes and heart disease might originate during early life, and our previous (23) and current studies illustrate how sleep perturbations may contribute to such processes. Developmental plasticity can be viewed as the ability of one genotype to produce a range of phenotypes in response to environmental conditions (4, 14). Epigenetic processes represent major regulatory mechanisms by which the environment modulates gene transcription during development and may account for decreased cellular plasticity through the establishment of cell-specific epigenetic profiles during development (2). The complexity of histone modifications, specifically whether they constitute a “code” (50) or are consequences of gene transcription and nucleosome remodeling (20), has long been the subject of debate in the scientific literature. However, there is a general agreement in the epigenetic field that histone modifications play important roles in the regulation of several cellular processes, as well as in the development and progression of human diseases. A model for the establishment and transmission of histone methylation to the next cell generation during cell division has been proposed (18, 25). Histone acetylation associated with transcription provides one mechanism of coordinating the histone-transcription factor interactions (24). It was shown that histone H4 acetylation is not only a consequence of gene transcription, but rather plays a function in the modulation of gene expression (11). Histone modifications have been suggested as a part of the epigenetic memory that helps to stabilize gene expression in normal development (37). The specific signature illustrated by the histone modifications in SFo supports the function of histones as part of the epigenetic memory that could contribute to the transmission of sleep perturbation-induced risk for metabolic dysfunction from one generation to the next. Our observation that the epigenetic alterations in the metabolic gene master switch Foxo1 precede the emergence of Met-S in the offspring opens the question of whether gestational SF induces global or gene-specific alterations in the epigenetic landscape during early development. On the basis of the alterations in the expression of 6 of 84 epigenetic chromatin modification enzymes (Fig. 9) in SFo compared with SCo, we posit that both global and gene-specific effects (driven by specific transcription factors recruitment) occur in the epigenome and are induced by sleep perturbations.

Sleep disorders resulting in fragmented and discontinuous sleep are common during pregnancy, especially during the later phases of gestation (45). Snoring in pregnant women leads to recurrent arousals from sleep to reestablish upper airway patency and has been associated with glucose intolerance and gestational diabetes risk (42). However, it is still not clear how sleep perturbations during pregnancy may affect the developmental trajectories in the offspring and its epigenome. Our findings provide compelling evidence linking perturbations in sleep during pregnancy to a MetS-like phenotype in offspring when they reach maturity. Assuming a 30% prevalence in gestational sleep disorders, (21) a 10% prevalence of metabolic syndrome (MetS) at baseline in the adult population, and a speculative 50% penetrance in the epigenetic-associated metabolic alterations induced by SF in the subsequent generation, 50% increases in MetS prevalence would be expected within three generations. Furthermore, doubling of MetS prevalence would occur within seven generations in any given population, assuming all other environmental factors, such as diet and physical activity, remained constant. The changes in MetS prevalence during the last 75 years are compatible with such a simplistic conceptual model and further exceed such estimates (22). Notwithstanding such considerations, our findings illustrate for the comprehensive range of epigenetic modifications induced by disrupted maternal sleep on the offspring, and their metabolic phenotypic correlates. It is possible that the other phenotypic features identified in SFo, e.g., increased food consumption, may originate from deregulation of other genes, or potentially reflect specific alterations in FOXO1-dependent pathways in hypothalamic neurons governing appetite and satiety (46).

FOXO1 is a transcription factor that plays important roles in the regulation of gluconeogenesis and glycogenolysis by insulin signaling in the liver. We now show that late-gestation SF induces altered expression of the Foxo1 in the liver by epigenetic mechanisms. The epigenetic signature of these abnormalities in the Foxo1 included DNA 5-hydroxymethyl-CpG and histone modifications in the promoter region of Foxo1, as well as in regions more distal to the transcription start site. Such alterations were associated with upregulation of some of the known Foxo1 downstream gene targets involved in liver gluconeogenesis and displayed a developmentally regulated temporal trajectory, which correlated with the MetS-like phenotype, which becomes apparent in the fully mature mice. The epigenetic Foxo1 misregulation induced by SF could, therefore, encompass a generalizable and unifying mechanism for how epigenetically altered metabolic gene networks are disrupted, and ultimately contribute to abnormal function in Foxo1 target tissues. However, on the basis of changes in other metabolically involved genes in SFo, such as Pgc-1α and Gapdh, which are not Foxo1 direct gene targets, it is likely that other gene pathways are involved and merit further exploration.

Epigenetic abnormalities can be reversed by specific interventions, such as treatment with epigenetic drugs, nutrition, maternal care, or exercise (3, 9, 19, 43). Our data show for the first time that the epigenetic abnormalities in the Foxo1 in the SFo offspring, as well as the MetS-like phenotype induced by SF during late gestation were reversed by instituting physical activity, but only when such physical activity was applied during early postnatal life. In addition to the already described function of the DNA methylation in PA effects, (3, 43), we showed that PA_early can also reverse changes in other epigenetic modifications involved in Foxo1 regulation, namely, DNA 5-hydroxymethyl-CpG and histone modifications (Fig. 11, A, C, D). Institution of PA at later stages of development did not elicit changes in Foxo1 expression and epigenetic marks and also failed to reverse the MetS phenotype (Fig. 4, A–D). DNA hydroxymethylation (5-hmC), which we studied, is a novel epigenetic mark, hydroxylated from 5-mC by Ten-Eleven translocation enzymes (TET) enzymes. Recent studies have indicated that 5-hmC can act not only as an intermediate to induce 5-mC demethylation, but also that these epigenetic marks contribute to gene expression (30, 53) and to biological processes, such as tumorigenesis (32). Tet1 and Tet2 double-knockout mice showed reduced 5-hmC, increased 5-mC levels, and abnormal methylation at various imprinted loci, and animals of both sexes were fertile (8). A role for 5-hmC as an epigenetic mark distinct from 5-mC has also been proposed (12). Our work demonstrates that 5-hmC is involved in the regulation of an important metabolic regulator gene, Foxo1.

Taken together, our findings unravel the epigenetically mediated impact of altered gestational sleep on metabolic regulation in the offspring, illustrate a complex epigenetic signature in the Foxo1, a key metabolic regulator, and provide strong incentive for early detection of such sleep perturbations by identifying a restricted postnatal period during which lifestyle interventions such as physical activity may successfully reverse the risk for MetS later in life (see schematic diagram as shown in Fig. 12).

Fig. 12.

Schematic design of metabolic and Foxo1 epigenetic abnormalities reverses induced by gestational sleep disturbance. Late-pregnancy sleep perturbations induce metabolic phenotype in the SF offspring at ∼20 to 24 wk after birth. Gestational sleep disturbance causes epigenetic misregulation of Foxo1 in the SF offspring. The SFo Foxo1 epigenetic abnormalities and the metabolic phenotype in the offspring were reversed by physical activity, but only when restricted to a defined window of opportunity during postnatal life: from week 4 to week 8 after birth.

Perspectives and Significance

A high prevalence of fragmented sleep occurs during late pregnancy and is associated with adverse outcomes. Here, we show that gestational sleep perturbations induce misregulation of Foxo1, a major metabolic gene, via epigenetic mechanisms in the offspring and are associated with the emergence of a metabolic syndrome-like phenotype later in life. The Foxo1 epigenetic abnormalities and metabolic phenotype in offspring were reversed by implementation of physical activity but only when restricted to a defined window of opportunity during postnatal life. A better understanding of the epigenome plasticity in response to early nutritional and other environmental factors may have critical implications to public health.

GRANTS

This work was supported by National Institutes of Health (Grants HL-65270 and HL-86662 to D. Gozal) and by a Comer Kids Classic grant (to A. Khalyfa).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

Author contributions: V.M. and D.G. conception and design of research; V.M., A.K., A.C., and M.N. performed experiments; V.M., A.K., Y.W., A.C., M.N., and D.G. analyzed data; V.M., Y.W., M.N., and D.G. interpreted results of experiments; V.M., A.K., and D.G. prepared figures; V.M. and D.G. drafted manuscript; V.M. and D.G. edited and revised manuscript; V.M., A.K., Y.W., A.C., M.N., and D.G. approved final version of manuscript.