How Sirt1 and Per2 control the body’s internal clock and aging by regulating each other

We have previously shown that while the majority of Sirt1-deficient (Sirt1^−/−) mice (in a genetic background of 129/FVB/Black Swiss at a ratio roughtly 1:2:1) that carry targeted deletion of exons 5–6 died during embryonic development, an average of 5% of them survived to adulthood11. To study the potential impact of Sirt1 deficiency on mouse development, we followed 17 Sirt1^−/− (mutant, MT), 33 heterozygous (HET), and 30 wild-type (WT) mice over time. We observed that MT mice displayed a much shorter lifespan than HET and WT mice (Fig. 1a), and that 80% of these MT mice did not survive over a year. The MT mice appeared to suffer kyphosis and appeared much older than the WT mice of the same age (Supplementary Figure 1a). The MT mice weighed less than age-matched WT mice (Supplementary Figure 1b). In addition, peripheral blood chemistry analysis demonstrated sustained elevated numbers of granulocytes in MT mice compared with age-matched WT mice (Supplementary Figure 1c), indicating the presence of persistent inflammation. Although the granulocyte number increased over time in WT mice, the number of granulocytes in these mice was significantly lower than that found in age-matched MT mice (Supplementary Figure 1c). Further, MT mice displayed reduced bone density (Fig. 1b) and impaired wound-healing ability (Supplementary Figure 1d). Next, we examined these 2 genotypes of mice for β-galactosidase activity (an indicator for cellular senescence) and γH2AX foci (an indicator of DNA damage), which have previously been used as aging markers25,26. These markers were increased in the kidney, brain, liver, and intestine of MT mice compared with WT mice (Supplementary Figure 1e,f). Levels of hepatic Sirt1 transcripts and proteins progressively decreased from young to mid-age and aged WT mice (Fig. 1c,d). NAD⁺ levels also declined as WT mice aged (Fig. 1e), indicating diminished sirtuins activity. Reduced Sirt1 expression correlated with progressively increased protein levels of p53, p21, and p16 in the liver of WT mice (Fig. 1f), and levels of these markers were much higher in MT mice compared with WT mice (Fig. 1g). Increased expression of these genes suggests increased DNA damage and/or senescence. As aging process is accompanied with DNA damage, this data suggests that Sirt1 deficiency resulted in increased DNA damage and/or senescence, which is consistent with our data that Sirt1 mutant mice were short lived.

We next wanted to investigate hepatic Sirt1-dependent aging-related genes. To do so, livers from WT mice at 3 months (young), 12 months (middle age), and 19 months (old) of age, as well as from MT mice at 3 months of age, were taken for microarray analysis. Three mice were utilized for each group, and the mice were sacrificed at 9–11 AM. Although 3-month old MT mice exhibited no obvious morphological signs of aging (data not shown), they still displayed high levels of some molecular aging markers (Fig. 1g). When we examined gene-expression profiles in WT mice of different ages by microarray analysis, we identified 192 upregulated and 281 downregulated genes in 12-month-old WT mice compared with 3-month-old WT mice, and 450 upregulated and 308 downregulated genes in 19-month-old WT mice compared with 3-month-old WT mice. We performed one more comparison to identify common genes in these 2 groups. This comparison identified 59 upregulated and 42 downregulated genes that showed consistent changes at both 12 months and 19 months in comparison with animals at 3 months of age (Fig. 2a). We named this group of genes as hepatic aging-related genes (Supplementary Table 1).

To define Sirt1-regulated genes in the liver, we compared the gene-expression profiles of 3-month-old WT and 3-month-old MT mice. This comparison revealed 427 up-regulated and 430 down-regulated genes in MT mice compared with WT mice of the same (young) age. We hypothesized that if Sirt1 is indeed involved in aging, then this group of differentially expressed genes should possess some common genes with our previously defined aging-related gene group. Thus, we next compared the Sirt1-regulated genes with the aging-related genes. This comparison revealed a statistically significant overlap of 21 up-regulated and 9 down-regulated genes (Fig. 2b). We named this group of genes as Sirt1-dependent aging-related genes (Supplementary Table 2). We confirmed expression changes of 14 out of 16 up-regulated genes and all 9 down-regulated genes by qRT-PCR analysis (Fig. 2c,d, and Supplementary Table 3).

We next explored if the gene expression changes observed during aging might be caused by alterations of Sirt1-mediated histone modification. When we examined the important hepatic histone acetylation sites, we found that histone H4K16 displayed increased acetylation in 12-month-old WT mice compared with 3-month-old WT mice, while no obvious change in acetylation of H3K9 was observed (Fig. 3a). Similarly, when the livers of 3-month-old WT and MT mice were compared, acetylation of H4K16 but not H3K9, was increased in MT mice compared with WT mice (Fig. 3b).

To detect changes at specific chromatin regions, we applied ChIP-Seq technology to perform a genome-wide analysis of H4K16ac in WT and MT mice. We observed a global increase in H4K16ac level at promoter regions in the 3-month-old MT mice (Fig. 3c, left panel and Supplementary Figure 2), however, we did not detect any difference from the input (Fig. 3c, right panel), suggesting that Sirt1 inhibits gene transcription by decreasing the histone acetylation levels in WT mice. Consistent with this idea, genes upregulated in the 3-month-old MT mice were associated with a greater increase in H4K16ac levels at their promoters than genes that were downregulated or unchanged (Fig. 3d), and this trend was not observed from the input controls (data not shown). We chose the top two genes (Per2 and Usp2) from our Sirt1-dependent aging related gene list to validate the increase of H4K16ac in 12-month-old WT mice and 3-month-old MT mice at their promoter regions by using ChIP-qPCR (Fig. 3e,f). Both Per2 and Usp2 were found up-regulated in the Sirt1-deficient mice (Supplementary Table 3). As shown in Fig. 3e,f, the promoter regions for both genes contain higher acetylation levels on H4K16 when mice are old or Sirt1 is deleted. Interestingly, genes down-regulated in the 3-month-old MT mice were not associated with a decrease in the H4K16ac level at their promoters (Fig. 3d), suggesting that these genes may be controlled by Sirt1 through other mechanisms.

Of note, both USP227 and PER218,19,20 are involved in circadian clock regulation. We decided to first focus on the Per2 gene, not only because it is one of the genes that is most upregulated during WT mouse aging, but also because previous studies have shown that the E-box of Per2′s promoter (+500 and −500) is regulated by histone H3 K9 acetylation28. Given the observed premature-aging phenotype of Sirt1-deficient mice, we were interested in determining whether the Per2 promoter is subject to epigenetic modification during the aging process. We examined 2 candidate regions (−9194 to −9431 and −8792 to −9029) identified by our ChIP-Seq analysis (Supplementary Figure 2). Our data revealed significant enrichment of H4K16ac in these regions in both 12-month-old WT mice and 3-month-old MT mice compared with 3 months old WT mice (Fig. 4a), while no obvious changes were detected in histone H3K9 acetylation at these sites (Fig. 4b). These data suggest that H4K16ac increase is associated with aging, but H3K9ac is not.

Because acetylation and methylation are frequently intertwined, we also assessed trimethylation of H3K4 and H3K9 during aging of WT mice and in 3-month-old MT mice by regular ChIP. Trimethylation of histone H3K4, which is a marker for open-chromatin structure, was significantly enriched in the region of (−8792 to −9029) in 12-month-old WT mice, while its change was moderate or not observed in the other region (Fig. 4c). Trimethylation of histone H3K9, a marker for condensed chromatin, was reduced in region 8972 to −9029 in 12-month-old WT and 3-month-old MT mice (Fig. 4d). These data are consistent with the Per2 expression pattern reported earlier.

Next, we studied Sirt1 occupancy on the Per2 promoter in the liver of 3-month-old WT mice and data indicated that Sirt1 bound to the Per2 promoter in both regions −9194 to −9431 and −8792 to −9029; such binding was not detected in the liver from 3-month-old MT mice (Fig. 4e). This is consistent with the idea that SIRT1 affects the histone H4K16 acetylation level in these regions. We further found that Per2 protein levels were markedly increased in the liver of 3-month-old MT mice (Fig. 4f) and in aged WT (12- or 22-month-old) mice (Fig. 4g) compared with 3-month-old WT mice. Altogether, these data indicate that the promoter of Per2 is a target of Sirt1 during aging.

It has been reported that PER2 is a key molecule in maintaining the normal circadian clock29. The increased expression of Per2 in Sirt1-deficient mice that we observed prompted us to examine if Per2 could also affect Sirt1 expression. To investigate this, Sirt1 and Per2 expression patterns were examined in mouse liver samples collected according to Zeitgeber Time (ZT hour). Per2 and Sirt1 both oscillated in response to the circadian clock; however, the Sirt1 expression pattern was generally opposite to that of Per2 at both protein (Supplementary Figure 3a) and mRNA (Supplementary Figure 3b) level (i.e., when Sirt1 was high, Per2 was low, and vice versa). For instance, between ZT12 to ZT20 (Zeitgeber Time 12 hours to 20 hours), hepatic Per2 was high, but hepatic Sirt1 maintained its lowest level.

To determine the key elements regulated by PER2 in the Sirt1 promoter, we performed serial deletion of the Sirt1 promoter and generated 5 constructs covering a 1.5 kb region upstream of the transcription initiation site. We found that fragment 202-Luc is the shortest construct that was both positively regulated by Clock/Bmal1 (Supplementary Figure 3c), and negatively regulated by Per2 (Supplementary Figure 3d). Co-transfection of Clock/Bmal1 and Per2 inhibited the ability of Clock/Bmal1 to activate 202-Luc (Fig. 5a). shRNA-mediated knockdown of Per2 upregulated 202-Luc in response to Bmal1 and Clock expression (Fig. 5b). Similar data were observed in the human hepatocellular carcinoma cell line, HepG2 (Supplementary Figure 3e).

In silico analysis detected 2 possible Clock/Bmal1 binding E-box sites [−(100–116) and −(172–189)] upstream of ATG in the Sirt1 promoter (Supplementary Figure 3f and Fig. 5c). Using a biotin pull-down assay, we verified that Clock/Bmal1/Per2 bound to both sites, but displayed stronger binding ability to −(172–189). Mutation of the core domains of these 2 E-box sites abolished the binding of the Clock/Bmal1/Per2 complex (Fig. 5d and Supplementary Figure 3g). We also performed site mutagenesis on the Sirt1 promoter 202-Luc construct, and then analyzed luciferase activity in response to Clock/Bmal1/Per2 overexpression. Mutation of the E-box cores abolished the stimulatory activity of Clock/Bmal1 on 202-Luc, as well as the inhibitory effect of Per2 on this promoter region (Fig. 5e), confirming that the Clock/Bmal1/Per2 complex binds to the −(100–116) and −(172–189) regions within the Sirt1 promoter.

ChIP analysis of Per2 and Bmal1 was performed in WT mouse liver samples to detect if the binding of these 2 molecules to the Sirt1 promoter depends on a circadian rhythm. Bmal1 binds to the Sirt1 promoter in the light/dark cycle, but Per2 binds to the Sirt1 promoter more strongly at night (Fig. 5f), which is when Sirt1 expression is at its lowest level (Supplementary Figure 3a,b). In mouse embryonic fibroblasts (MEFs), the circadian complex of Clock/Bmal1/Per2 bound to the Sirt1 promoter in the same fashion following serum shock (Supplementary Figure 4a). Sixteen hours post-serum shock, the binding of Per2 to the Sirt1 promoter was significantly increased at the region of Clock/Bmal1 binding site. This increased level of binding correlated with a low wave of Sirt1 expression (Supplementary Figure 4b). Deletion of Sirt1 enhanced expression of many circadian-related genes, such as Per1, Per2 and Cry2 (Supplementary Figure 4c).

Next, we investigated if the negative reciprocal regulation between SIRT1 and PER2 exists in human primary hepatocytes. In two different preparations of freshly isolated human hepatocytes, knockdown of SIRT1 upregulated endogenous PER2 expression; overexpressing SIRT1 decreased PER2 expression. Meanwhile, knockdown of PER2 elevated endogenous SIRT1 expression, and overexpression of PER2 diminished SIRT1 expression (Fig. 5g). We also performed the same analysis in the hTERT-immortalized human primary hepatocyte cell line HHT-430, and a similar relationship between the two genes was detected (Supplementary Figure 4d).

When shPer2 lentivirus was injected into wild type mice through tail vein, SIRT1 expression is elevated at multiple circadian points during dark cycle (Fig. 5h). When SIRT1 was overexpressed in WT MEF cells, acetylation of H4K16 on Per2 promoter is significantly reduced (Supplementary Figure 4e); knocking down SIRT1 in WT MEF cells enhance acetylation of H4K16 on Per2 promoter (Supplementary Figure 4f). These data demonstrated that a negative reciprocal relationship between SIRT1 and PER2 exists in both human and mouse hepatic systems.

We showed earlier that multiple tissues in Sirt1-deficient mice displayed senescence (Supplementary Figure 1e). Based on these initial findings, we wanted to explore if dysregulation of Per2 could influence senescence. To do so, we used retroviruses containing Per2, or empty pBabe vector to infect primary WT MEF cells. Overexpression of Per2 decreased endogenous Sirt1 levels in MEF cells (Fig. 5i). Six days post-infection, cell growth was significantly decreased in Per2-infected cells (Fig. 5j). β-galactosidase activity shown by x-gal staining demonstrated that much more cells overexpressing Per2 had entered senescence than the control cells (Fig. 5k).

To further investigate the relationship among Sirt1, Per2 and senescence in a circadian fashion, we utilized liver-specific Sirt1 knockout mice (SIRT1LKO). Liver samples were collected based on the Zeitgeber Time (ZT) and gene expression was analyzed by qRT-PCR. The deletion of Sirt1 in the liver disturbed Per2 circadian expression pattern when compared with WT controls (Fig. 5l). Absence of Sirt1 also dramatically affected the expression rhythm over the light/dark cycle of some other genes, such as p16 (Fig. 5m), Cry1 (Supplementary Figure 4g), p19, and Pepck (Supplementary Figure 4h). These genes play important roles in aging, circadian rhythmicity and glucose metabolism. Taken together, the above data indicate that SIRT1 and PER2 constitute a reciprocal negative regulation loop and regulate aging process both in vitro and in vivo.

To provide evidence that Per2 may regulate Sirt1 levels in vivo, we injected virus for Per2 into the tail vein of 3-month-old mice and collected liver samples 5 days post-injection based on the circadian cycle. Overexpression of Per2 (Supplementary Figure 5a) substantially reduced Sirt1 expression at multiple time points during the ZT cycle (Fig. 6a) and influenced the expression of Sirt1 downstream genes Pepck (Fig. 6b) and G6pase (Supplementary Figure 5b), which are known to be regulated by Sirt1 and other circadian transcription factors, such as BMAL1/CLOCK. The gluconeogenesis genes Pepck and G6pase both displayed elevated expression levels in the Per2 overexpressed mice when compared with GFP retrovirus-infected mice at multiple time points, especially at ZT0, 4, 8 h. Overexpressing Per2 also altered the Cry2 circadian rhythm pattern (Fig. 6c) with advanced phases, and elevated the expression of aging-related genes p16 (Fig. 6d) and p19 (Supplementary Figure 5c) at several time points. This evidence suggests that overexpression of Per2 disrupts circadian oscillation, reduces Sirt1 levels, alters Sirt1 downstream gene expression, and increases expression of senescence markers, all of which may contribute to the shortened lifespan observed in Sirt1-deficient mice.

Next, we investigated whether reducing Per2 in SIRT1LKO mice would be able to correct the impaired glucose metabolism and the expression of aging genes. Knockdown of Per2 with shPer2 lentivirus reduced hepatic Per2 mRNA and protein levels efficiently in SIRT1LKO mice (Supplementary Figure 5d). Under these conditions, the rhythm of circadian gene expression for Cry2 shifted towards to the pattern that was observed in WT mice (Fig. 6e). At the same time, gluconeogenesis genes Pepck (Fig. 6f) and G6pase (Supplementary Figure 5e) in shPer2 lentivirus-infected LKO mice also showed a rhythmic pattern that is closer to that of wild-type mice. The level of the aging markers p16 (Fig. 6g) and p19 (Supplementary Figure 5f) was dramatically reduced at multiple time points after Per2 knockdown. In addition, SIRT1LKO/shPer2 mice displayed improved glucose tolerance ability during a glucose tolerance test (Fig. 6h); decreased hepatic oil droplet formation (Fig. 6i), and reduced expression of lipid synthesis genes SREBP (Fig. 6j) and chREBP (Fig. 6k). Notably, knockdown of Per2 also markedly reduced cell senescence revealed by decreased β-galatosidase activity (Fig. 6l). These data confirm that the negative reciprocal regulation loop between Sirt1 and Per2 observed in vitro also occurs in vivo and alteration of Per2 in SIRT1LKO mice contributes to abnormal circadian oscillation and premature aging phenotypes.

Mouse experiments were performed in accordance with the guidelines of Institutional Animal Care and Use Committee (IACUC). All animal experimental protocols were approved by either the Animal Care and Use Committee of the National Institute of Diabetes and Digestive and Kidney Diseases or the Animal Care and Use Committee of University of Macau. The usage of human samples were carried out in accordance with the guidelines of NIH Human Research Protections Program (HRPP). The de-identified normal human liver cells were obtained through the Liver Tissue Cell Distribution System (Pittsburgh, PA) after obtaining a written informed consent by a protocol approved by the Human Research Review Committee of the University of Pittsburgh, which was funded by NIH Contract # HSN276201200017C.

Male and female Sirt1 mutant mice (Sirt1^+/−)11 were intercrossed to generate Sirt1^−/− (MT), Sirt1^+/− (HET), and Sirt1^+/+ (WT) mice. Because difference in genetic background and gender might affect gene expression, we only compared gene expression within the same genetic background and the same gender (i.e. comparisons were made either among males or females in same genetic background). Specifically, the mutant mice used in Figs 1c–g and 2, 3, 4 were males in a mixed genetic background of 129/FVB/Black Swiss at a ratio roughly 1:2:1. In Figs 5h,l,m and 6, when large quantity of Sirt1 mutant mice (24 mutant mice were needed for each curve) was required for circadian studies after gene knockdown or over-expression, which could not be performed using Sirt1-deficient mice due to their poor viability, all comparisons were made using Sirt1-liver specific knockout mice (Sirt1LKO) in C57BL/6N background. To match up with this, all control mice were littermates in the same background.

Mice were maintained in a 12-h light/12-h dark cycle environment. The mouse cohorts were followed for 2 years. Because the survival rate for MT mice is very low, no any inclusion and exclusion criteria were used when MT mice were selected for use in experiments. The WT and HET mice appeared normal and were randomly picked for experiments. All animal experiments were approved by the Animal Care and Use Committee of the National Institute of Diabetes and Digestive and Kidney Diseases.

Mice were anesthetized with Avertin. Whole mice or the dissected femur/tibias were placed into a Faxitron X-ray apparatus (Specimen Radiography System) and BioMax XAR X-ray film (Kodak) was used to capture images.

Tissues were embedded in OCT compound, and 10 μm cryosections were prepared and fixed with 1% formalin. Mouse embryonic fibroblasts (MEFs) were fixed with 0.5% glutaraldehyde for 5 min at room temperature. Slides were then incubated at 37 °C overnight with a solution containing K₄ [Fe(CN)₆] 3H₂O and K₃[Fe(CN). Afterwards, the slides were counterstained with FastRed.

Mouse livers were harvested at 9 AM–11 AM. The fresh liver pieces were snap frozen with liquid nitrogen on site. Later, about 1 mg of snap-frozen liver piece were pulverized in the presence of liquid nitrogen, and the dissociated tissue was first cross linked with 1% formaldehyde. Chromatin was prepared from these formaldehyde-crosslinked tissues and fragmented to an average size of 200 bp by sonication. ChIP-Seq experiments were performed as described previously56 using antibodies against H4K16AC (Santa Cruz). Briefly, 2 μg of the antibodies bind to 20 μl of Dynabeads protein A beads at room temperature for 1 hr with rotation. Add chromatin mixture from 1 mg liver tissue to the beads + antibody, incubate at 4 °C for overnight with rotation. Then wash, reverse cross-link, precipitated ChIP DNA.

The preparation of sequencing library was conducted with the following steps modified from56: (1) The ChIP DNA ends were repaired by using the Epicentre DNA END-Repair kit (Cat. No. ER0720, Epicentre Biotechnologies). (2) “A” was added to 3′ of DNA end by using Klenow Fragment (3′− > 5′ exo-) from NEB. (3) Illumina adaptor was ligated to the DNA end by T4 DNA Ligase. (4) The ChIP DNA was amplified by using illumine index primers. (5) ChIP DNA was isolated from agarose gel. (6) The purified DNA was used directly for illumine sequence.

The mice were housed in a 12-h light/12-h dark cycle environment. During circadian-clock experiments, the livers were collected every 4 h according to the light cycle and snap-frozen with liquid nitrogen until further processing. Samples were designated according to the number of hours elapsed from the beginning of the light cycle (e.g., ZT4 represents the sample that was collected 4 h after the start of the light cycle). For Figs 5h,l,m, 6 and 7, each time point contains 3 individual mice, and each line contains 24 mice.

C57BL/6N male mice at 3 months of age were tail vein injected with virus at 1 × 10⁸ virus/200 μl PBS. Viruses used were: GFP retrovirus; shLuciferase (shLuc) lentivirus; shPer2 lentivirus; pBabe retrovirus, or Per2 retrovirus. Injections were performed twice, one week apart. The shLuciferase and shPer2 lentivirus vectors were purchased from Open Biosystems. Per2 vector was obtained from Addgene. Each virus was produced in HEK293T cells according to the manufacturer’s instructions. Virus purification was performed by ultracentrifugation.

Total RNA was isolated with STAT-60^TM (Tel-Test, Inc.) from livers or MEF cells. cDNA was synthesized with QuantiTect Reverse Transcription Kit (QIAGEN). qRT-PCR was performed using a SYBR green PCR Master Mix (Applied Biosystems, Roche) and 7500 Real Time PCR (Applied Biosystems). Primer sequences are listed at the end of the Methods section. Mouse Sirt1 and 18S primers have been published before.

pGL3B vector containing different promoter sequences were transfected into primary MEF cells. All transfections were performed with Lipofectamine^TM 2000 (Invitrogen). After a 24-h incubation, luciferase activity was assessed with the Dual-Luciferase Reporter Assay Kit (Promega). The shRNAs against Per2 (61 and 62) were purchased from Open BioSystems. Myc-Per2, Myc-Clock, and Myc-Bmal1 plasmids were obtained from Addgene.

Livers or MEF cells were cross-linked with 1% formalin for 15 min and ChIP was performed with Sirt1, Per2, Kcnk5, Bmal1, H4K16ac, H3K9ac, H3K4(Me)₃, or H3K9(Me)₃ antibodies as described previously57. Primer sequences are listed at the end of the Methods section.

Four oligonucleotides containing biotin on the 5′-nucleotide of the sense strand were used in the pull-down assays. One microgram of each double-stranded oligonucleotide was incubated with 300 μg of nuclear protein extracts for 20 min at room temperature. Streptavidin-agarose beads (30 μl, Sigma-Aldrich) that had been pre-absorbed by poly(dI-dC) (Sigma-Aldrich) were added and incubated with the extract for 4 h at 4 °C. The protein-DNA-streptavidin-agarose complexes were pelleted by centrifugation and analyzed by Western blotting with Myc-tag antibody.

To mutate the core E-box domains of 202-Luc, 202-pGL3B plasmid was used as a parental template. The PCR method (PCR SuperMix High Fidelity, Invitrogen) was applied and mutation was confirmed with sequencing. Mutagenesis was performed by BioInnovatise Inc. (Rockville, MD).

Mouse livers were harvested at 9 AM–11 AM. Fresh liver samples were immersed in RNALater (Ambion) overnight. Afterwards, the samples were extracted with STAT-60^TM (Tel-Test, Inc.) to retrieve total RNA. The RNA was purified further with RNeasy^R Mini Kit (QIAGEN). An RNA Analyzer (Agilent) was used to check RNA quality. RNAs with high quality were reverse transcribed and the cDNAs were labeled and hybridized to the Affymetrix GeneChip mouse genome 2.0 array. Gene expression data generated by the Affymetrix MOE430 chip (two samples per condition) were imported to GeneSpring for analysis. Absence and presence call for each probe was made by MAS5. To identify differentially expressed genes between any 2 conditions (fold-change >1.5 and p-value < 0.05; t-test), probes labeled by absence on both duplicates in either condition were discarded.

NAD⁺ levels were assessed with the NAD/NADH Assay Kit (Abcam, Cat. No. ab65348) according to the manufacturer’s instructions.

Primary MEF cells were derived from E14.5 embryos using a standard procedure. Cells were maintained in DMEM with 15% fetal bovine serum. For serum-shock assays, 5 × 10⁵ MEF cells were seeded into a 10-cm dish and cultured in DMEM with 15% fetal bovine serum for 7 days until confluence, then 50% horse or calf serum was used to shock the cells for 2 h. The cells were then release to normal culture condition. After 12 hours, the samples were collected every 4 h. HepG2 cells were obtained from the American Type Culture Collection and grown in DMEMde. HHT-4 cells were provided by Dr. Xin-Wei Wang and grown in described condition. All cell lines were tested and free for mycoplasma contamination.

Student’s t-test or two-way Anova (Prism) was used for analyzing the significance among different set of samples. Briefly, data (triplicated) obtained from at least 3 individual biological samples for each genotype at multiple time points were input into online student t-test calculation table (http://www.physics.csbsju.edu/stats/t-test_bulk_form.html↗), afterwards, p value was calculated. When p value is <0.05, the comparison is considered significant. For ChIP-Seq analysis, p-value was calculated by the Kolmogorov-Smirnov test. The analysis was repeated by using input read densities for comparison purposes. For Microarray, gene expression data generated by the Affymetrix MOE430 chip (two samples per condition) were imported to GeneSpring for analysis. Absence and presence call for each probe was made by MAS5. Differentially expressed genes were identified between any 2 conditions (fold-change >1.5 and p-value < 0.05; t-test), probes labeled by absence on both duplicates in either condition were discarded.

The ChIP-Seq and microarray data sets for this study have been submitted to NLM/NCBI. The data are available at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? token=kjohyouybhefnip&acc=GSE54451↗.