Histone Deacetylase Inhibitor Depsipeptide Activates Silenced Genes through Decreasing both CpG and H3K9 Methylation on the Promoter

Abstract

Histone deacetylase inhibitor (HDACi) has been shown to demethylate the mammalian genome, which further strengthens the concept that DNA methylation and histone modifications interact in regulation of gene expression. Here, we report that an HDAC inhibitor, depsipeptide, exhibited significant demethylating activity on the promoters of several genes, including p16, SALL3, and GATA4 in human lung cancer cell lines H719 and H23, colon cancer cell line HT-29, and pancreatic cancer cell line PANC1. Although expression of DNA methyltransferase 1 (DNMT1) was not affected by depsipeptide, a decrease in binding of DNMT1 to the promoter of these genes played a dominant role in depsipeptide-induced demethylation and reactivation. Depsipeptide also suppressed expression of histone methyltransferases G9A and SUV39H1, which in turn resulted in a decrease of di- and trimethylated H3K9 around these genes' promoter. Furthermore, both loading of heterochromatin-associated protein 1 (HP1α and HP1β) to methylated H3K9 and binding of DNMT1 to these genes' promoter were significantly reduced in depsipeptide-treated cells. Similar DNA demethylation was induced by another HDAC inhibitor, apicidin, but not by trichostatin A. Our data describe a novel mechanism of HDACi-mediated DNA demethylation via suppression of histone methyltransferases and reduced recruitment of HP1 and DNMT1 to the genes' promoter.

DNA methylation and histone modifications are critical epigenetic processes that help control chromatin structure and gene regulation (4, 6, 26, 28, 50). Although there are many kinds of histone modifications in mammalian cells (5, 33, 40, 41), up to this point it seems that it is histone acetylation and histone methylation which are associated with DNA methylation status in regulating gene expression. For example, chromatin with unmethylated CpG islands is always enriched in hyperacetylated histone, while methylated DNA has been shown to be associated with repressed chromatin (6, 26). Early evidence showed that DNA methylation might be a dominant factor in the suppression of gene expression, because the demethylating agent 5-aza-2′-deoxycytidine (5-aza-CdR) alone could activate gene expression, whereas the histone deacetylase inhibitor trichostatin A (TSA) alone could not reactivate gene expression if the genes were still methylated (9). In addition, when DNA methyltransferase was inhibited by 5-aza-CdR, promoter demethylation as well as gene reexpression occurred first, and induction of the histone code reversal lagged behind (14). However, accumulating evidence indicates that histone modifications are prerequisite for DNA methylation (2, 25, 37, 58, 63). For example, p16 silencing as well as histone methylation at lysine 9 of histone 3 (H3K9) were shown to occur prior to p16 methylation in the human cancer cell line HCT116 (2). In addition, transcription of RASSF1A was dramatically reduced in proliferating human mammary epithelial cells with increasing passage number, and histone H3 deacetylation and H3K9 trimethylation played a critical role in inducing RASSF1A silencing prior to DNA methylation (58).

In fact, DNA methylation and histone modifications collaborate to regulate gene expression. For example, the methyl-binding protein MeCP2 specifically binds to methylated CpG sequences and in turn recruits histone deacetylase 1 activity (42), through which the chromatin structure is altered, thus inducing gene silencing. In addition, DNA methyltransferase was reported to interact with SUV39H1, a histone H3K9 methyltransferase (18), and DNA methyltransferase 3b (DNMT3b)-associated CpG methylation is significantly decreased in mouse Suv39h knockout stem cells (37). Therefore, DNA methylation, histone deacetylation, and histone methylation may function together to regulate gene expression (4, 17).

Of even greater interest, both the DNA demethylating agent 5-aza-CdR and histone deacetylase (HDAC) inhibitors play overlapping roles in gene regulation and other cellular functions, further supporting an inherent correlation between DNA methylation and histone modifications (19). Based on microarray analysis, it was shown that 5-aza-CdR and TSA induced a similar gene expressional pattern in human HCT116 cells (20). It was also reported that the demethylating agent 5-aza-CdR is involved in histone modifications. For example, 5-aza-CdR was able to significantly enhance histone acetylation of H3 and H4 at multiple lysine sites induced by an HDAC inhibitor (75) and also dramatically reduced H3K9 methylation in the promoter regions of p16 and MLH1 in RKO cells (32). 5-aza-CdR treatment resulted in global decreases in H3K9 dimethylation through decreasing expression of G9A, a key enzyme responsible for H3K9 dimethylation (68). On the other hand, HDAC inhibitors are also often reported to have demethylating activity (23, 55, 60). For example, TSA and butyrate can induce promoter demethylation in Neurospora crassa (55) and mammalian cells (23). HDAC inhibitor (HDACi) caused a highly selective loss of DNA methylation, which implies that histone acetylation may direct DNA methylation. However, it is plausible that HDACi directly influences DNA methyltransferases activity. TSA and butyrate were both reported to specifically suppress DNMT3B expression by decreasing the stability of DNMT3B mRNA in human endometrial cells (70). This evidence suggests that DNA methylation and histone modifications may together comprise the regulatory machinery for control of gene expression, and any changes in the modifications to either DNA or histone may influence the other. However, although the HDAC inhibitors mentioned above exhibit a demethylating function on specific genes or globally, the exact mechanism of this demethylation is not completely understood.

Depsipeptide is another previously developed HDAC inhibitor which has much stronger activity in the inhibition of HDAC than that of TSA (75). Recent studies have indicated that depsipeptide brings about more extensive pharmacological effects on cells, including induction of DNA damage (29) and acetylation of nonhistone proteins, such as p53 (73). To further investigate whether depsipeptide has demethylating activity, the hypermethylated p16 promoter (also referred as CDKN2A), SALL3, and GATA4 in human lung cancer cell lines H719 and H23 (74), human colon cancer cell line HT-29, and the human pancreatic cell line PANC1 were examined after depsipeptide treatment. In this study, the depsipeptide-mediated demethylating activity of the p16 promoter and reactivation were evaluated. Furthermore, we also investigated the expression of histone methyltransferases SUV39H1 and G9A after depsipeptide treatment. Our data demonstrate that reduced expression of SUV39H1 and G9A results in low levels of H3K9 methylation, and this in turn results in poor recruitment of heterochromatin-associated protein 1 (HP1) and DNMT1 on the p16 promoter. These results demonstrate a possible mechanistic link between histone methylation and DNA methylation.

MATERIALS AND METHODS

Cell lines and treatments.

Human lung cancer cell lines H719 and H23, human colon cancer cell lines HT29, HCT116, and SW480, and human pancreatic cancer cell line PANC1 used in this study were purchased from the American Type Culture Collection. These cell lines were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum and 100 μg/ml streptomycin-penicillin. Cells were treated with depsipeptide (obtained from the NIH) for up to 96 h at various concentrations. Fresh medium containing depsipeptide was replaced every 24 h.

Measurement of cell viability and cell growth.

The 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay was performed to evaluate cell growth. Briefly, an equal number of cells (approximately 5,000) were seeded into a 96-well plate 24 h prior to experimental use. Cells were treated with depsipeptide at different concentrations or different intervals. After treatment, MTT dye solution (Sigma) was added into the 96-well plate. The corrected absorbance of each sample was calculated by comparison with the untreated control. Cell viability was tested by the trypan blue assay. Treated and untreated cells were harvested and stained with trypan blue (final concentration, 0.02%; Life Technologies, Gaithersburg, MD). The stained cells were then counted immediately under a microscope. At least 200 cells were counted for each time point.

RNAi.

Sequences of RNA interference (RNAi) oligonucleotides for controls (nonsilencing), DNMT3A, DNMT1, G9A, and SUV39H1 were as follows: nonsilencing small interfering RNA (siRNA), UUCUCCGAACGUGUCACGU; DNMT3A siRNA, CAUCCACUGUGAAUGAUAA; DNMT1 siRNA, CACUGGUUCUGCGCUGGGA; G9A siRNA, CCAUGCUGUCAACUACCAUGG; SUV39H1 siRNA, ACCUCUUUGACCUGGACUA. All RNAi oligonucleotides were purchased from Shanghai GeneChem Company (Shanghai). These RNAi oligonucleotides were transfected into H719 cells by using the Lipofectamine 2000 transfection kit (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions.

MS-PCR.

DNA was extracted and then treated with bisulfite as previously described with minor modifications (76). Briefly, genomic DNA (1 μg) in a volume of 50 μl was denatured with NaOH (final concentration, 0.275 M) for 10 min at 42°C. The denatured DNA was then treated with 10 μl of 10 mM hydroquinone and 520 μl of 3 M sodium bisulphate at 50°C overnight. The primers for methylation-specific PCR (MS-PCR) of the p16 promoter were designed as follows: methylation-specific primers, forward primer 5′-TTATTAGAGGGTGGGGCGGATCGC-3′ and reverse primer 5′-GACCCCGAACCGCGACCGTAA-3′; unmethylation-specific primers, forward primer 5′-TTATTAGAGGGTGGGGTGGATTGT-3′ and reverse primer 5′-CAACCCCAAACCACAACCATAA-3′. The PCR conditions were initiated with a denaturing step at 95°C for 10 min, followed by 36 cycles of 96°C for 30 s, 61°C for 20 s, and 72°C for 20 s, and were concluded with an interval of 72°C for 7 min. The PCR products were run on a 2% agarose gel, stained with ethidium bromide, and evaluated with UV light.

Bisulfite sequencing.

DNA was treated with bisulfite and purified for PCR as described previously (76). The primers for sequencing region D of the p16 promoter were as follows: forward primer, 5′-GGTAGGTGGGGAGGAGTTTAG-3′; reverse primer, 5′-CCAACCCCTCCTCTTTCTTC-3′. All primer sequences of other regions of the promoters of p16, GATA4, and SALL3 for bisulfite sequencing are available upon request. The PCR products were gel extracted (Qiagen, Valencia, CA) and ligated into the pGEM-T easy vector by using the TA cloning system (Promega, Madison, WI). Transformed Escherichia coli DH5α cells were cultured overnight, and the plasmid DNA was isolated using a kit (Qiagen). At least 10 separate clones were chosen for sequence analysis.

Measurement of DNA de novo and maintenance methyltransferase activity.

The method for measurement of DNA de novo and maintenance methyltransferase activity was as previously reported (70). Briefly, cell extracts were prepared in lysis buffer (50 mM Tris-HCl [pH 7.8], 1.0 mM EDTA [pH 8.0], 10% glycerol, 0.01% sodium azide, 10% Tween 80, 100 μg/ml RNase A, and 0.5 mM phenylmethylsulfonyl fluoride). De novo and maintenance methyltransferase activity was measured in the presence of 30 μg cellular protein, 3.0 μg of double-stranded oligonucleotides, and 2.4 μCi of S-adenosyl-l-(methyl-³H) methionine (SAM; Amersham, Piscataway, NJ). The oligonucleotide sequences used for de novo methyltransferase activity were as follows: top strand, 5′-GGGGGCCAAGCGCGCGCCTGGCGCCCGGGCCGGCTCAAGCGCGCGCCTGGCGCCCGGATC; bottom strand, 5′-GATCCGGGCGCCAGGCGCGCGCTTGAGCCGGCCCGGGCGCCAGGCGCGCGCTTG. The sequences used for maintenance methyltransferase activity were as described above, but the CGs of the primer of the bottom strand were methylated (hemimethylated strand). Following incubation at 37°C for 1 h, the reaction was terminated by adding 90 μl of stop solution (1.0% sodium dodecyl sulfate [SDS], 2.0 mmol/liter EDTA, 3.0% 4-amino salicylate, 5.0% butanol, 0.25 mg/ml calf thymus DNA, and 1.0 mg/ml proteinase K) and incubation at 37°C for 45 min. The reaction mixture was then spotted onto a Whatman GF/C filter paper disc (Fisher Scientific, East Brunswick, NJ). The disc was washed three times with 5% trichloroacetic acid, rinsed in 70% ethanol, and dried at 56°C for 20 min. The discs were submerged in UltimaGold scintillation solution (Packard, Meriden, CT), and radioactive incorporation was measured in a Beckman liquid scintillation counter (LS 5000TD). A blank control reaction was performed simultaneously using cell extracts that were heated to 80°C for 15 min to inactivate the DNMT activity. The results, expressed as counts per minute, were adjusted by subtracting the background level. Each experimental data point was performed twice in duplicate.

The assay for evaluation of the direct role of depsipeptide on DNMT1.

Determination of methyltransferase activity was performed as described above with minor modifications (51). Briefly, a reaction buffer with recombinant DNMT1 protein (purchased from New England Biolabs) was incubated with depsipeptide (5 nM) and a nonmethylated oligonucleotide or a hemimethylated oligonucleotide (sequences the same as above) with S-adenosyl-l-(methyl-³H)methionine.

Combined bisulfite restriction assay (COBRA).

A fragment of satellite 2 (sat2), which is located in a pericentric heterochromatin, was selected to test changes of methylation status in depsipeptide-treated cells. Genome DNA was extracted from H719 cells before or after depsipeptide treatment, and the DNA was treated with sodium bisulfite. A PCR fragment of sat2 was then cut with HinfI. The fragments cut by HinfI were then separated on a polyacrylamide gel electrophoresis (PAGE) gel.

RT-PCR and real-time PCR.

Cells were grown and treated with depsipeptide in a 10-cm diameter dish. Total RNA was isolated using Trizol reagent (Invitrogen, Carlsbad, CA). cDNA was synthesized from 2 μg of RNA with oligo(dT)₁₈ primers using the SuperScript kit (Invitrogen). Primer sequences used for reverse transcription-PCR (RT-PCR) and real-time PCR are available upon request.

Histone extraction and Western blot analysis.

To identify histone modifications, acid extraction of histone was performed as previously reported (75). To detect other proteins, cells were lysed with radioimmunoprecipitation assay buffer (25 mM Tris-HCl, pH 7.4, 150 mM KCl, 5 mM EDTA, 0.5% Na deoxycholate, 0.1% SDS, 1% NP-40). Equal amounts of protein (100 to 150 μg) were size fractionated on 6 to 12.5% SDS-PAGE gel. The antibodies used were anti-DNMT1 (Abcam ab13537), anti-DNMT3A (Abcam ab2850), anti-DNMT3B (Abcam ab2851), anti-G9A (Abcam ab40542), anti-SUV39H1 (Upstate 05-615), anti-HP1α (Upstate 05-689), anti-HP1β (Chemicon MAB3448), anti-HP1γ (Upstate 05-690), anti-dimethyl-H3K9 (Upstate 07-214), anti-trimethyl-H3K9 (Upstate 07-442), anti-AceH3 (Upstate 06-599), anti-H3 (Upstate 06-755), and β-actin (Huatesheng Biotechnolgy, Fushun, China).

ChIP assay and q-ChIP PCR.

A chromatin immunoprecipitation (ChIP) assay was performed as described previously (73). Briefly, 2 × 10⁷ cells treated with depsipeptide were fixed with 1% formaldehyde at 37°C for 10 min and were then lysed on ice for 15 min. These lysed extracts were subjected to shearing by sonication. After centrifugation at 14,000 rpm for 15 min, the soluble chromatin was subjected to immunoprecipitation with antibodies against different modified histones and other proteins as indicated. Then, the complexes were drawn off with protein G-agarose beads and washed sequentially with low-salt, high-salt, LiCl, and Tris-EDTA buffers and were finally extracted with freshly prepared 1% SDS-0.1 M NaHCO₃. Heating the samples at 65°C for 6 h reversed DNA and protein cross-links, and DNA was then purified with a Qiagen DNA extraction kit. The primers for all ChIPs and quantitative ChIP (q-ChIP) PCR are available upon request.

RESULTS

The HDAC inhibitor depsipeptide induces demethylation of the promoters of p16.

It has been reported that the HDAC inhibitor TSA and sodium butyrate exhibit demethylating activities in Neurospora and mammalian cells (23, 55, 60), which suggests that other members of the HDACi family may have a similar function. In this study, a novel HDAC inhibitor, depsipeptide, was chosen for examination of its role in DNA demethylation in the human lung cancer cell line H719, in which many gene loci, including the p16 promoter, are frequently methylated (74). Depsipeptide was applied to H719 cells for 0 to 96 h at low doses (1 to 5 nM). First, depsipeptide-induced inhibition of cell proliferation and changes in cell viability were tested with the trypan blue and MTT assays, respectively. As shown in Fig. 1A (left panel), depsipeptide exhibited a strong inhibitory effect on cell proliferation as assayed with MTT. However, most cells remained alive when treated with depsipeptide either in different doses or for different durations (Fig. 1A, right panel) as demonstrated by the trypan blue assay, indicating that cytotoxicity induced by depsipeptide over long-term treatment at low doses may result from inhibition of cell proliferation, but not from cell death. Since the p16 promoter in H719 cells is hypermethylated (74), MS-PCR was performed to detect changes in the methylation status of the p16 promoter after depsipeptide treatment. Figure 1B is a schematic diagram showing the promoter and distal upstream region of p16. As shown in Fig. 1C, depsipeptide exhibited duration- and dose-dependent demethylating activity, in which the band intensity of the methylated p16 (region D) was decreased and the band intensity of the unmethylated p16 was increased after depsipeptide treatment. To further confirm the demethylating effects of depsipeptide on the p16 promoter in H719 cells, DNA bisulfite sequencing was performed. Primers were designed to amplify four different fragments that are located within the CpG island and non-CpG island regions of the p16 promoter (accession number DQ406745) as shown in Fig. 1B. PCR products were then subcloned into a pGEM-T easy vector, and 10 separate subclones were sequenced. Methylation changes of each CG dinucleotide after depsipeptide treatment were analyzed in Fig. 1D. In the control samples, the p16 promoter in H719 cells was highly methylated (in 10 separate clones, the CpG methylation rates were 100% in region A, 51.1% in region B, 100% in region C, and 98.1% in region D) (Fig. 1D, upper panels). Methylated CpG in the p16 promoter was decreased to 92% in region A, 18.9% in region B, 71.3% in region C, and 54.4% in region D when cells were treated with depsipeptide at 5 nM for 96 h (Fig. 1D, lower panels). Within the four fragments, only region A, which is the distal region of the p16 promoter, was resistant to depsipeptide-induced demethylation. These data clearly indicate that depsipeptide is a potent demethylating agent. To investigate whether the depsipeptide-induced demethylation in the p16 promoter was associated with gene activation, expression of p16 was examined in H719 cells by RT-PCR. Figure 1E (upper panel) shows that p16 in H719 cells was silenced and depsipeptide was able to reactivate its expression in a dose-dependent manner. Figure 1E (lower panel) shows a quantitative expressional change in p16 mRNA in H719 cells treated with depsipeptide using real-time PCR. However, we did not observe demethylation of the hypermethylated promoters, or reexpression of the silenced genes, in cells treated with depsipeptide for short periods of time (24 h or less) (data not shown).

FIG. 1.

Depsipeptide-induced demethylation of the p16 promoter in H719 cells. (A) Measurement of inhibition of cell proliferation and cell viability in depsipeptide-treated or untreated H719 cells. The left panel shows the growth inhibition of H719 cells induced by depsipeptide at the indicated concentrations or various intervals compared to the untreated control in an MTT assay. The right panel shows changes in viability of H719 cells treated with depsipeptide in a trypan blue assay. (B) Schematic diagram of the p16 promoter and its distal upstream region. Four regions of the p16 promoter and its distal upstream region, referred to as region A (−4693 to −4135 relative to its transcriptional start site, +1), region B (−2599 to −2385), region C (−779 to −364), and region D (−144 to +83), were selected as fragments for bisulfite sequencing (BS1, BS2, BS3, and BS4, respectively) and for chromatin immunoprecipitation assay (ChIP1, ChIP2, ChIP3, and ChIP4, respectively). In addition, region D was chosen for methylation-specific PCR. Region A is the region most distal to the p16 promoter. Region B is located within a non-CpG island segment of the p16 promoter. Regions C and D are located within the CpG island of the p16 promoter. (C) Changes in DNA methylation status of the p16 promoter (region D) in H719 cells induced by depsipeptide treatment for 96 h at various concentrations (upper panel) or at 5 nM with various incubation times (lower panel) by MS-PCR. M indicates methylated DNA, and U indicates unmethylated DNA. H₂O was added into the PCR mixture as a negative control. Normal human DNA was treated with methylase SssI as a positive control. (D) Bisulfite sequencing analysis of the p16 promoter in untreated H719 cells or cells treated with depsipeptide at 5 nM for 96 h. p16 promoter regions for bisulfite sequencing include four regions (BS1, 5 CGs; BS2, 9 CGs; BS3, 8 CGs; BS4,16 CGs). Methylated CG (filled circles) and unmethylated CG (open circles) are represented. All upper panels show untreated controls, and all lower panels show depsipeptide-treated samples (5 nM for 96 h). The methylation rate in each region (as a percentage) is shown under each panel. (E) RT-PCR analysis showing p16 expression in H719 cells after depsipeptide treatment. Upper panel: p16 mRNA was extracted from H719 cells treated with depsipeptide for 96 h at the doses indicated. β-Actin served as a loading balance. Lower panel: quantitative analysis of p16 expression induced by depsipeptide was performed by real-time PCR. Untreated H719 cells were used as a control. Because there was no expression of p16 in untreated H719 cells, the value of p16 expression in cells treated with depsipeptide at 0.5 nM was assigned as 1. Expression of p16 in cells treated with depsipeptide at higher concentrations was calculated relative to the value of p16 expression in the cells treated with depsipeptide at 0.5 nM.

Depsipeptide-induced DNA demethylation occurs in multiple genes and in multiple human cancer cell lines.

To investigate whether depsipeptide-induced DNA demethylation is a generalized phenomenon, other genes, such as SALL3 and GATA4, were tested in H719 cells. As shown in Fig. 2A, two promoter regions of these genes were selected to analyze bisulfite sequencing. The promoters of SALL3 and GATA4 were also hypermethylated, as indicated in Fig. 2B and C (upper panels). It was of interest that methylated CpGs in these regions of both the SALL3 and GATA4 promoters were also decreased after depsipeptide treatment (Fig. 2B and C, lower panels).

FIG. 2.

Depsipeptide-induced DNA demethylation in different genes of H719 cells. (A) Schematic diagrams of the GATA4 promoter (left panel) and the SALL3 promoter (right panel). Region A (bp −3466 to −3115 relative to the transcriptional start site, +1) and region B (bp −461 to −120) of the GATA4 promoter or region A (bp −3762 to −3524 relative to the transcriptional start site) and region B (+700 to +908) of the SALL3 promoter were chosen for bisulfite sequencing (BS1 and BS2) and ChIP assay (ChIP1 and ChIP2), respectively. (B and C) Changes in DNA methylation of the GATA4 promoter and the SALL3 promoter in H719 cells treated with de-psipeptide at 5 nM for 96 h. All upper panels show untreated controls, and all lower panels show the depsipeptide-treated samples. (D) COBRA showing depsipeptide-induced demethylation of sat2. The upper panel shows a region of sat2 used for amplification that contains five HinfI restriction sites within the fragment. The lower panel is an image of a PAGE gel. U indicates an unmethylated fragment of sat2, and M indicates methylated fragments of sat2. A 5-aza-CdR-treated sample was chosen as a positive control.

Furthermore, in order to determine whether depsipeptide-induced demethylation operates more broadly over the genome, a fragment of sat2, which is located in a pericentric heterochromatin, was selected to test changes of methylation status in depsipeptide-treated cells. sat2 is usually hypermethylated in most mammalian cells (71). Figure 2D shows that this unmethylated fragment was obviously increased in depsipeptide-treated H719 cells, as demonstrated by a COBRA, indicating that depsipeptide induces demethylation not only in euchromatin genes but also in repetitive elements.

Depsipeptide-induced DNA demethylation is cancer cell line dependent. We selected another human lung cancer cell line, H23, human colon cancer cell lines HT-29, HCT116, and SW480, and a human pancreatic cell line, PANC1, to validate the phenomenon of depsipeptide-induced demethylation. Two genes, p16 and GATA4, were chosen to evaluate methylation changes in these depsipeptide-treated cell lines. As shown in Fig. 3A to C, depsipeptide was able to induce demethylation of both the p16 and GATA4 promoters in H23, HT-29, and PANC1 cells assayed with bisulfite sequencing (region D of the p16 promoter and region B of the GATA4 promoter). However, depsipeptide could not reduce methylation in either the p16 or GATA4 promoter in HCT116 cells and SW480 cells (data not shown). This difference in depsipeptide-induced demethylation may result from differing sensitivities to depsipeptide treatment among cell lines, as depsipeptide was able to induce a significant inhibition of cell proliferation in H23, HT-29, and PANC1 cells, but such inhibition was not seen in HCT116 cells or SW480 cells when assayed with MTT (data not shown).

FIG. 3.

Depsipeptide-induced DNA demethylation in different human cancer cell lines. Another human lung cancer cell line H23 (A), a human pancreatic cancer cell line PANC1 (B), and human colon cancer cell line HT-29 (C) were treated with depsipeptide at 5 nM for 96 h, and DNA was extracted for bisulfite sequencing analysis. Region D of the p16 promoter (BS4) and region B of the GATA4 promoter (BS2) were analyzed in these cell lines in untreated controls (all upper panels) or after treatment with depsipeptide at 5 nM for 96 h (all lower panels) with bisulfite sequencing.

Methyltransferase activities are reduced by depsipeptide treatment.

Since DNMT family members are responsible for enzymatic DNA methylation (38, 46), it is possible that depsipeptide may reduce expression of DNMTs or decrease their activities to exert its demethylating function. To determine expression of DNMTs in cells treated with depsipeptide, H719 cells were exposed to depsipeptide at concentrations ranging from 0.5 nM to 5 nM for 96 h. RT-PCR was then performed to identify expression of mRNA of DNMT1, DNMT3A, and DNMT3B. As shown in Fig. 4A, depsipeptide specifically reduced DNMT3A expression in a dose-dependent manner. The expression of DNMTs mRNA in cells treated with depsipeptide was also quantified by real-time PCR (Fig. 4B). The depsipeptide-induced reduction of DNMT3A mRNA was also consistent with its protein levels (Fig. 4C). However, depsipeptide did not downregulate expression of either DNMT1 or DNMT3B in either mRNA or protein levels in H719 cells (Fig. 4A to C).

FIG. 4.

Changes in both DNMT expression and activity in H719 cells treated with depsipeptide. (A) Total mRNA was extracted from untreated or depsipeptide-treated H719 cells, and RT-PCR was performed to determine DNMT expression. β-Actin expression was used as a loading control. (B) Real-time PCR was performed to quantify expression of DNMT induced by depsipeptide. (C) Western blotting was performed to determine DNMT expression in depsipeptide-treated H719 cells. (D and E) Measurement of de novo DNA methyltransferase activity (D) or maintenance DNA methyltransferase activity in depsipeptide treated H719 cells (E). Results are from two separate experiments, and bars represent the standard errors.

We next determined if the activity of DNMTs was also altered in depsipeptide-treated H719 cells. Cell lysates isolated from depsipeptide-treated or untreated control cells were incubated in a methylase assay buffer system to measure the global de novo or maintenance methyltransferase activity, respectively, as described previously (70). As shown in Fig. 4D and E, radioactive incorporation on substrate DNA in H719 cell extracts treated with depsipeptide was obviously decreased compared to the untreated control, indicating that depsipeptide has effects on both demethylating activity of de novo and maintenance DNA methylation. However, this depsipeptide-induced decrease in maintenance activity of DNA methyltransferase is not very consistent with changes in DNMT1 expression induced by depsipeptide, indicating that the decrease in maintenance activity of DNA methyltransferase induced by depsipeptide may be independent of DNMT1 expression.

Knockdown of DNMT3A alone is not sufficient to demethylate the p16 promoter or to reactivate p16 expression.

Since expression of DNMT3A was significantly inhibited by depsipeptide, it is possible that depsipeptide-induced demethylation of the p16 promoter and its reactivation result from downregulation of DNMT3A. To examine this hypothesis, DNMT3A was knocked down by RNAi, and the methylation status of the p16 promoter was then measured in H719 cells. Figure 5A shows that the RNAi efficacies were high, because both mRNA and protein levels of DNMT3A were significantly decreased in the DNMT3A RNAi-treated cells compared with the untreated control. MS-PCR was performed to detect methylation changes in the p16 promoter (region D) before and after cells were treated with DNMT3A RNAi. As shown in Fig. 5B, DNMT3A RNAi alone did not induce demethylation of the p16 promoter, as the unmethylated band was almost undetectable with DNMT3A RNAi treatment. With bisulfite sequencing, methylated CpG of the p16 promoter (region D) in H719 cells was slightly decreased (from 97.5% to 91.3%) in DNMT3A RNAi-treated cells compared to nonsilencing RNAi-treated cells (Fig. 5C), which indicates that DNMT3A alone does not play a dominant role in methylation of the p16 promoter. Consistent with this observation, expression of p16 mRNA was not observed in the DNMT3A RNAi-treated cells examined by RT-PCR (Fig. 5D).

FIG. 5.

Knockdown of DNMT3A alone by RNAi was not sufficient for demethylation of the p16 promoter and reactivation of its expression. (A) Efficacies of DNMT3A RNAi were determined by RT-PCR (left panel) and Western blotting (right panel). H719 cells were transfected with nonsilencing siRNA oligonucleotides (NS) as an RNAi control. (B) MS-PCR analysis of the p16 promoter (region D) in H719 cells treated with DNMT3A RNAi. (C) Methylation status of the p16 promoter in H719 cells treated with DNMT3A RNAi was quantified with bisulfite sequencing. Region D of the p16 promoter (BS4) was chosen for bisulfite sequencing. Left panel: untreated control. Right panel: DNMT3A RNAi-treated samples. (D) Changes in expression of p16 with DNMT3A RNAi were analyzed with RT-PCR in H719 cells. A positive control was made using depsipeptide-treated cells.

Reduction in binding of DNMT1 to the p16 promoter is associated with depsipeptide-induced DNA demethylation.

Although expression levels of DNMT1 and DNMT3B were unchanged in the depsipeptide-treated cells in this study, the possibility that depsipeptide may influence binding of DNMT1 or DNMT3B to the p16 promoter by changing chromatin structure could not be ruled out. To evaluate this hypothesis, ChIP assays were performed to determine the chromatin occupancy of these factors. First, region D of the p16 promoter was chosen, and the ChIP assay was performed. Figure 6A shows that DNMT1 and DNMT3A were able to bind to the p16 promoter, whereas DNMT3B was not able to bind to the p16 promoter. In order to evaluate the binding activity of DNMT3B, a Wnt1 promoter was chosen for ChIP assay because DNMT3B was reported to be able to bind to the Wnt1 promoter (67). It is clear that DNMT3B was able to bind to the Wnt1 promoter (Fig. 6A), indicating that the anti-DNMT3b was effective. After treatment with depsipeptide, binding of DNMT1 or DNMT3A to the p16 promoter was gradually decreased as the concentration of depsipeptide was increased, suggesting that a decrease in DNMT1 binding to the p16 promoter may play a role in depsipeptide-induced demethylation. q-ChIP PCR was also performed in different regions of the p16 promoter to show that binding of DNMT1 to the p16 promoter was significantly decreased in regions B, C, and D, but not in region A (Fig. 6B). In addition, the depsipeptide-induced decrease in binding of DNMT1 to the GATA4 promoter (region B) in H719 cells was also observed with the q-ChIP PCR (Fig. 6C). To further demonstrate DNMT1 was a key component in inducing DNA methylation of the p16 promoter, DNMT1 RNAi was performed. Subsequently, MS-PCR and bisulfite sequencing were used to detect methylation changes of the p16 promoters after DNMT1 RNAi treatment. As shown in Fig. 6D, the RNAi against DNMT1 was effective in reducing expression of DNMT1 as demonstrated by RT-PCR and Western blotting. Consistent with the result of DNMT1 knock down, the unmethylated p16 band (region D) was clearly seen in DNMT1 RNAi-treated H719 cells with MS-PCR (Fig. 6E). Bisulfite sequencing (10 clones tested) of four regions of the p16 promoter showed that the methylated CpG in the p16 promoter was decreased in region A (from 100% to 72%), region B (from 52.2% to 30%), region C (100% to 78.8%), and region D (from 97.5% to 73.1%) in H719 cells treated with DNMT1 RNAi (Fig. 6F). In addition, methylated CpG in the cells treated with double knockdown (double RNAi) of DNMT1 and DNMT3A by RNAi was further reduced in these four regions (Fig. 6F). Double RNAi of DNMT1/DNMT3A also reduced DNA methylation in the GATA4 promoter (region B) in H719 cells (Fig. 6G).

FIG. 6.

Decrease in binding of DNMT1 to the p16 promoter may be responsible for p16 reactivation. (A) A ChIP assay showed changes in binding of DNMTs to the p16 promoter (region D; ChIP4) in H719 cells induced by depsipeptide. Input was used as a loading control, and IgG was added into the ChIP reaction buffer as a negative control. The Wnt1 promoter was chosen as a positive control for validating the binding ability of the DNMT3B antibody. (B) Real-time PCR for the ChIP assay (q-ChIP PCR) showed changes in binding of DNMT1 to the p16 promoter (all four regions). (C) A q-ChIP PCR also showed changes in binding of DNMT1 to the GATA4 promoter (region B). (D) Efficacies of DNMT1 RNAi were tested by RT-PCR (left panel) and Western blotting (right panel). (E) MS-PCR analysis of the p16 promoter was performed in H719 cells treated with DNMT1 RNAi. (F) Bisulfite sequencing showed changes in the methylation status of the p16 promoter (all four regions) in H719 cells treated with DNMT1 RNAi or double knockdown (double RNAi) of DNMT1/DNMT3A by RNAi. All upper panels show nonsilencing (NS)-treated control. All middle panels show DNMT1 RNAi-treated samples. All lower panels show double RNAi of DNMT1/DNMT3A-treated samples. (G) Bisulfite sequencing was also performed to detect the methylation status in the GATA4 promoter (region B) in H719 cells after double RNAi of DNMT1/DNMT3A. The upper panel shows the untreated control, and the lower panel shows double RNAi of the DNMT1/DNMT3A-treated sample.

Consistent with the demethylation of its promoter, p16 expression was significantly increased with DNMT1-RNAi or double RNAi of DNMT1- and DNMT3A-treated H719 cells as observed by RT-PCR (Fig. 7A) and real-time PCR (Fig. 7B). However, p16 expression in H719 cells with double RNAi of DNMT1 and DNMT3A was less than that in depsipeptide-treated cells, which implies that other epigenetic mechanisms may be involved in p16 reactivation by depsipeptide treatment. In addition, to exclude the possibility of depsipeptide directly affecting the activity of DNMT1, an in vitro assay for detecting DNMT1 activity was performed, and Fig. 7C shows that DNMT1 activity is not directly suppressed by depsipeptide. These data suggest that depsipeptide may act on other factors to induce a decrease in binding of DNMT1 to the p16 promoter. Finally, depsipeptide has been reported to have DNA-damaging function (29), which raises the interesting question as to whether depsipeptide-induced demethylation results from a depsipeptide-induced stress reaction. We treated H719 cells with a potent DNA-damaging agent, etoposide (VP16) at 0.05 μM for 96 h, and bisulfite sequencing was performed to evaluate changes in the p16 promoter (region D). Figure 7D shows that VP16 did not obviously reduce demethylation, indicating that depsipeptide-induced demethylation does not result from its DNA-damaging function.

FIG. 7.

Reexpression of p16 induced by DNMT1/DNMT3A RNAi. (A) mRNA extracted from H719 cells treated with RNAi against DNMT1, DNMT3A, or double RNAi of DNMT1/DNMT3A. mRNA expression of p16 in cells treated with depsipeptide at 5 nM for 96 h was used as a positive control. (B) Real-time PCR was performed in H719 cells treated with DNMT1 RNAi or double RNAi of DNMT1/DNMT3A. (C) Depsipeptide has no direct role in repressing activity of DNMT1 in vitro. Recombinant DNMT1 protein was incubated with depsipeptide in a buffer containing unmethylated or hemimethylated DNA fragment and ³H-labeled SAM. Both de novo and maintenance activities of DNA methyltransferase were measured. (D) H719 cells were treated with VP16 at 0.05 μM for 96 h, and DNA was then extracted for bisulfite sequencing (region D of the p16 promoter).

Depsipeptide treatment reduces H3K9 methylation by suppressing expression of G9A and SUV39H1.

Since methylation of H3K9 is one of the primary marks of inactive chromatin structure, global levels of di- and trimethylated H3K9 (H3K9me2 and H3K9me3) were analyzed by extraction of histones and immunoblotting with specific antibodies. As shown in Fig. 8A, both global H3K9me2 and H3K9me3 were decreased in the depsipeptide-treated cells. These results led to a consideration of whether depsipeptide affects expression of histone methyltransferases and, thus in turn, reduces H3K9me2 and H3K9me3. Because G9A and SUV39H1 have been reported to catalyze dimethylation and trimethylation of H3K9 (49, 53, 61, 62), expression of G9A and SUV39H1 was evaluated in depsipeptide-treated H719 cells. Representative real-time PCR showed a depsipeptide-induced suppression of expression of G9A and SUV39H1 (Fig. 8B). For example, depsipeptide at 5 nM for 96 h suppressed expression of G9A and SUV39H1 by 6.5- and 20-fold, respectively (Fig. 8B). This depsipeptide-induced suppression of G9A and SUV39H1 was also confirmed by Western blotting for protein levels as shown in Fig. 8C. Since the global level of H3K9 methylation was decreased by depsipeptide treatment, it was expected that H3K9 methylation around specific loci of euchromatin may be also reduced. A representative ChIP assay showed that the levels of H3K9me2 and H3K9me3 on the p16 promoter (region D) were significantly decreased in a dose-dependent manner (Fig. 8D, left panels). Furthermore, a q-ChIP PCR assay was also performed to determine the level of H3K9me2 and H3K9me3 on regions A, B, C, and D of the p16 promoter (Fig. 8D, middle and right panels). Depsipeptide significantly reduced the level of H3K9me2 and H3K9me3 in regions B, C, and D of the p16 promoter but had less effect on H3K9me2 and H3K9me3 in region A of the p16 promoter. In addition, a depsipeptide-induced decrease in H3K9me2 and H3K9me3 was also observed in the GATA4 promoter (region B) demonstrated by q-ChIP PCR (Fig. 8E). These results are consistent with the trend for a depsipeptide-mediated decrease in SUV39H1 and G9A expression.

FIG. 8.

Depsipeptide treatment led to a decrease in histone methylation in H719 cells. (A) Global di- and trimethylated H3K9 in the genome were decreased in cells treated with depsipeptide. Histones were extracted from H719 cells treated with depsipeptide at 3 nM or 5 nM for 96 h. Specific antibodies against dimethylated H3K9 (H3K9me2) or trimethylated H3K9 (H3K9me3) were used to detect histone methylations. The total H3 level was used as a loading control. (B) Desipeptide treatment suppressed expression of G9A and SUV39H1 in H719 cells. Real-time PCR showed changes of mRNA in SUV39H1 and G9A in H719 cells treated with depsipeptide. (C) Representative Western blots showed changes in protein levels of G9A and SUV39H1 in H719 cells treated with depsipeptide. (D) Changes in H3K9me2 and H3K9me3 around the p16 promoter (region D) were detected by a ChIP assay in H719 cells treated with depsipeptide (left panels). Real-time PCR showed quantitative changes in H3K9me2 (middle panel) and H3K9me3 (right panel) around the p16 promoter (all four regions) in the H719 cells treated with depsipeptide. (E) Real-time PCR also showed quantitative changes in H3K9me2 (left panel) and H3K9me3 (right panel) around the GATA4 promoter (region B) in H719 cells treated with depsipeptide.

Inhibition of G9A and SUV39H1 is crucial for demethylation of both H3K9 and the p16 promoter.

Since RNAi against either G9A (13) or SUV39H1 (37), or a chemical reagent-induced decrease in G9A (35), have been reported to be effective in inducing changes in histone methylation, RNAi against SUV39H1 and G9A was performed. As shown in Fig. 9A and B, the efficacies in both RNAi experiments were high, as both mRNA and protein levels were significantly decreased in H719 cells. To further determine whether G9A and SUV39H1 bind to the p16 promoter, a ChIP assay was performed. It was clear that both G9A and SUV39H1 were able to bind to the p16 promoter (Fig. 9C, left panel), and this was quantitatively confirmed by q-ChIP PCR assay in four different regions of the p16 promoter (Fig. 9C, right panel). Binding of G9A and SUV39H1 to region A is low, but the other three regions show a greater extent of binding of G9A and SUV39H1. Next, region D was chosen to determine whether G9A RNAi or SUV39H1 RNAi decreases H3K9 methylation on the p16 promoter. As expected, G9A RNAi mainly decreased H3K9me2 on the p16 promoter (region D), whereas SUV39H1 RNAi mainly resulted in a decrease in H3K9me3 on the p16 promoter (Fig. 9D, left panels). When H719 cells were treated with double RNAi of G9A and SUV39H1, both H3K9me2 and H3K9me3 showed a more significant decrease than that with treatment with RNAi of G9A or SUV39H1 alone (Fig. 9D). A q-ChIP PCR assay showed the changes in binding of H3K9me2 and H3K9me3 to four different regions of the p16 promoter after the RNAi treatments (Fig. 9D, median and right panels). Subsequently, to evaluate whether the decrease in the expression of G9A and SUV39H1 is responsible for the CpG demethylation in the p16 promoter, MS-PCR and bisulfite sequencing were performed. As shown in Fig. 10A, although RNAi against G9A alone or against SUV39H1 alone could not induce a detectable unmethylated p16 band, a clear unmethylated band of p16 was observed when cells were treated with double RNAi of G9A and SUV39H1. Furthermore, bisulfite sequencing confirmed that double RNAi of G9A and SUV39H1 significantly reduced methylation in regions B, C, and D of the p16 promoter but not in region A (Fig. 10B; compare to 1D, upper panels). In addition, the GATA4 promoter (region B) was analyzed by bisulfite sequencing in cells treated with double RNAi of G9A and SUV39H1. It seemed that region B of the GATA4 promoter was sensitive to demethylation induced by double RNAi of G9A and SUV39H1 (Fig. 10C; compared to 2B, right upper panel).

FIG. 9.

Decrease in expression of G9A- and SUV39H1-induced reduction of H3K9 methylation and DNA methylation on the p16 promoter. (A and B) The efficacies of G9A and SUV39H1 RNAi in H719 cells were determined by mRNA levels (left panel) and protein levels (right panel). (C) Direct binding of G9A (left upper panel) and SUV39H1 (left lower panel) to the p16 promoter (region D) in H719 cells was shown in a ChIP assay. A q-ChIP PCR was performed to quantify the binding status of SUV39H1 and G9A on the p16 promoter (all four regions) (right panel). (D) Changes in H3K9me2 and H3K9me3 around the p16 promoter (region D) were detected by ChIP assay in H719 cells treated with G9A RNAi, SUV39H1 RNAi, or double RNAi (left panels). A q-ChIP PCR showed changes in H3K9me2 (middle panel) and H3K9me3 (right panel) around the p16 promoter (all four regions) in H719 cells treated with RNAi of G9A or SUV39H1 or double RNAi.

FIG. 10.

Changes in DNA methylation in the p16 or GATA4 promoter after treatment with G9A/SUV39H1 RNAi. (A) Changes in methylation status of the p16 promoter (region D) were determined by MS-PCR in H719 cells treated with RNAi against G9A or SUV39H1 or double RNAi. (B) Bisulfite sequencing showed methylation changes in the p16 promoter (all four regions) in H719 cells treated with double RNAi of G9A/SUV39H1. (C) Bisulfite sequencing showed methylation changes in the GATA4 promoter (region B) in H719 cells treated with double RNAi of G9A/SUV39H1. (D) A q-ChIP PCR showed changes in H3K9me2 (left panel) and H3K9me3 (right panel) around the p16 promoter (region D) in H719 cells treated with RNAi of DNMT1.

As another point of interest, we also evaluated whether changes in DNMT1 affect histone methylation. As shown in Fig. 10D, DNMT1 RNAi was able to reduce the level of H3K9me2 and H3K9me3 on the four regions of the p16 promoter as demonstrated by q-ChIP PCR. These data imply that there may be a feedback between expression of histone methylation and DNA methylation.

Decrease in recruitment of HP1 and DNMT1 onto the p16 promoter is responsible for depsipeptide-induced demethylation.

Generally, H3K9me2 and H3K9me3 induce DNA methylation through recruiting the HP1 family, by which they associate with DNMT1 to cause DNA methylation (37, 56). Therefore, changes in expression of the HP1 family and binding of HP1 or DNMT1 to the p16 promoter were investigated in H719 cells treated with depsipeptide. Depsipeptide did not reduce expression of total HP1α or -β as determined by mRNA (Fig. 11A) and protein levels (Fig. 11B, left panels). However, in contrast to total expression of HP1α and -β, depsipeptide obviously reduced HP1α and -β fractions in protein bound to chromatin (Fig. 11B, right panel). In addition, binding of HP1α and -β to the p16 promoter (region D) was decreased in cells treated with depsipeptide as seen in a ChIP assay (Fig. 11C, left panels). A q-ChIP PCR was used to analyze the binding of HP1β to the p16 promoter (region A, B, C, and D) in cells treated with depsipeptide, showing that depsipeptide significantly reduced binding of HP1β to the p16 promoter (there was almost no binding of HP1β whatsoever to the p16 promoter in the regions B, C, and D after depsipeptide treatment at 5 nM for 96 h) (Fig. 11C, middle panel). However, there was no obvious decrease in binding of HP1β to the p16 promoter (region A) after depsipeptide treatment (Fig. 11C, middle panel). A depsipeptide-induced decrease in binding of HP1β to the GATA4 promoter (region B) was also observed (Fig. 11C, right panel). We did not observe any binding of HP1γ to the p16 promoter in this study (data not shown). To further investigate whether the depsipeptide-mediated reduction in binding of HP1 to the p16 promoter (region D) was due to suppression of SUV39H1 and G9A, RNAi against G9A and SUV39H1 was performed and followed with the ChIP assay (Fig. 11D, left panels). q-ChIP PCR for binding of HP1β or DNMT1 to the p16 promoter in all four regions was performed in H719 cells treated with RNAi and showed a significant decrease in binding of both HP1β (Fig. 11D, middle panel) and DNMT1 (Fig. 11D, right panel) to regions B, C, and D of the p16 promoter. However, the binding of DNMT1 and HP1β to region A of the p16 promoter was not obviously changed (Fig. 11D, middle and right panels). These data suggest that G9A and SUV39H1 play a critical role in inducing DNA methylation through recruitment of HP1 and DNMT1 to the p16 promoters.

FIG. 11.

Decrease in recruitment of HP1α and HP1β on the p16 promoter was involved in depsipeptide-induced demethylation. (A) Expression of total HP1α and HP1β mRNA was evaluated with RT-PCR (left panels). Real-time PCR shows mRNA changes in expression of HP1α and HP1β in H719 cells treated with depsipeptide (right panel). (B) A Western blot shows the total amount of HP1α and HP1β in H719 cells treated with depsipeptide (left panels). HP1α and HP1β bound to chromatin were also analyzed by Western blotting (right panels). Total H3 was used as a loading control for Western blotting. (C) Binding of HP1α and HP1β to the p16 promoter (region D) was determined by ChIP assay in H719 cells treated with depsipeptide at 3 or 5 nM for 96 h (left panel). A q-ChIP PCR showed changes in binding of HP1β to the p16 promoter (all four regions) in H719 cells treated with depsipeptide (middle panel). A q-ChIP PCR showed changes in binding of HP1β to the GATA4 promoter (region B) in H719 cells treated with depsipeptide (right panel). (D) Binding of DNMT1, HP1α, and -β to the p16 promoter (region D) in H719 cells treated with RNAi of G9A or SUV39H1 or double RNAi by ChIP assay, respectively (left panel). A q-ChIP PCR shows changes in binding of HP1β (middle panel) or DNMT1 (right panel) to the p16 promoter (all four regions) in H719 cells treated with RNAi of G9A or SUV39H1 or double RNAi of G9A/SUV39H1.

Histone acetylation by HDACi is not involved in demethylation of the p16 promoter.

Because depsipeptide is a potent HDAC inhibitor, it is possible that histone acetylation status may influence DNA demethylation. To test this possibility, H719 cells were exposed separately to TSA and depsipeptide, and histone acetylation status was then examined by using specific anti-acetyl-H3. As expected, both TSA and depsipeptide were obviously able to induce histone acetylation of H3 (Fig. 12A). In addition, both TSA and depsipeptide induced obvious acetylation of H3 bound to the p16 promoter (region D) as determined by ChIP assay (Fig. 12B). It was of interest that even at higher concentrations TSA could not induce either expression of p16 as determined by RT-PCR (Fig. 12C, left panel) or demethylation of the p16 promoter (region D) as shown by bisulfite sequencing (Fig. 12C, middle panel). Consistent with these findings, TSA was not able to induce expressional changes in SUV39H1 and G9A as tested by real-time PCR (Fig. 12C, right panel). In contrast to TSA, another HDAC inhibitor, apicidin, which is similar structurally to depsipeptide, showed demethylating activity by use of bisulfite sequencing in the p16 promoter (region D) (Fig. 12D, left panel) and was able to induce p16 reexpression (Fig. 12D, middle panel). A further consistent finding was that apicidin could induce a decrease in expression of SUV39H1 and G9A in H719 cells based on real-time PCR (Fig. 12D, right panel).

FIG. 12.

Histone acetylation was not involved in DNA demethylation induced by depsipeptide treatment. (A) Western blotting was performed to detect changes of H3 acetylation in H719 cells treated with TSA or depsipeptide. (B) A ChIP assay showed an increase in binding of acetylated H3 to the p16 promoter (region D) in H719 cells treated with TSA and depsipeptide. (C) TSA (40 nM for 96 h) did not induce reexpression of p16 as demonstrated by RT-PCR (left panels) nor induce demethylation as assayed with bisulfite sequencing (region D, middle panel). Depsipeptide-treated samples were used as a positive control. TSA did not induce changes in expression of G9A or SUV39H1 in H719 cells as tested by real-time PCR (right panel). (D) Apicidin (1 μM for 96 h) induced demethylation of the p16 promoter (region D) demonstrated by bisulfite sequencing (left panel) and reactivated p16 expression as assayed with real-time PCR (middle panel). Apicidin also suppressed expression of G9A and SUV39H1 in H719 cells as tested by real-time PCR (right panel).

These data indicate that HDAC inhibitor-induced demethylation is selective and is not involved in histone acetylation, at least not in the p16 promoter.

DISCUSSION

The depsipeptide-induced decrease in expression of DNMT3A and its activity did not appear to direct its demethylating function in this study. For example, only 6.2% of methylated CpG in the p16 promoter became unmethylated when H719 cells were treated with RNAi against DNMT3A alone (Fig. 5C), and this result is consistent with several previously studies (27, 65). As has been previously demonstrated, DNMT3A may not be important in methylating whole genomic DNA but is required for methylation of repetitive DNA elements or a specific promoter that contains methylated CpG-poor regions (65). Although several studies have suggested that DNMT3B plays a more important role in inducing de novo methylation than DNMT3A (27, 52), demethylation of the p16 promoter induced by depsipeptide was not directly associated with DNMT3B in this study.

DNMT1 (38) and DNMT3A/B (46) are traditionally considered responsible for maintenance of DNA methylation and de novo methylation, respectively. However, current evidence shows that the functions of these DNMTs overlap extensively (13, 27, 39). For example, DNMT3A and DNMT3B are also reported to participate in maintenance of methylation by directly binding to DNMT1 (30). In this scenario, DNMT3A and DNMT3B can fill the methylated CpG gap omitted by DNMT1-mediated methylation in S phase and maintain a stable methylation pattern on the repetitive regions (39). On the other hand, DNMT1 also has exhibited de novo methyltransferase activity (54, 66). In particular, DNMT1 is important for inducing methylation of the p16 promoter in human cancer cells (16, 59). For example, knockdown of DNMT1 with RNAi for 9 days decreased methylation of the p16 promoter from 100% to 10% in H1299 cells (59). In this study, depsipeptide was able to induce a decrease both in de novo and maintenance activities of DNA methyltransferases (Fig. 4D and E). We found that depsipeptide did not reduce expression of DNMT1 (Fig. 4A to C), but the binding activity of DNMT1 to the p16 promoter was significantly decreased in depsipeptide-treated H719 cells (Fig. 6A and B).

Consistent with previous reports (11, 34, 72), DNMT1 activity decreases, or the binding activity of DNMT1 to its target decreased, and active demethylation occurred with resultant demethylation of hypermethylated DNA, although at the same time cell growth was severely inhibited by depsipeptide (Fig. 1A). In the presence of p53, DNMT1 was recruited to the survivin promoter and induced its methylation (12). In that case, the total amount of DNMT1 was not increased. However, DNMT1 is easily targeted to the survivin promoter and results in hypermethylation of the survivin promoter when p53 is increased by DNA damage (12).

Recent findings have further generated an exciting model wherein DNA methylation and histone methylation at lysine 9 of H3 form a mutually reinforcing epigenetic cycle for regulation of gene expression (17, 18, 37, 56). Methylation of H3K9 is one of the key components of repressed chromatin, and H3K9 methylation is often accompanied by DNA methylation in the silenced promoter (17, 33, 41). Moreover, H3K9 methylation was recently shown to be a prerequisite for DNA methylation in Neurospora crassa (63), Arabidopsis thaliana (25), and mammalian cells (15, 37). Histone methyltransferases G9A and SUV39H1 have been confirmed to catalyze dimethylation of H3K9 (61, 62) and trimethylation of H3K9 (49, 53), respectively. The recruitment of G9A and SUV39H1 to repressed genes has been documented in previous reports (44, 45, 64). For example, G9A can be targeted to the p21^waf1/cip1 promoter by associating with the CCAAT displacement protein/cut homologue (CDP/cut) (45). Although SUV39H1 is often localized with regions of heterochromatin to mediate its silenced status (53), it has also been found in specific promoters of genes or in facultative heterochromatin composed of silenced genes (44, 64). In addition, colocalization of both enzymes in the same promoter has also been previously reported. For example, the protein Gfi1b can recruit both SUV39H1 and G9A and target them to suppress gene expression or to form heterochromatin (64). In support of this observation, our data showed that both G9A and SUV39H1 could be recruited to the p16 promoter and this recruitment was abolished by depsipeptide treatment. Therefore, depsipeptide-induced suppression of G9A and SUV39H1 is directly associated with decrease in H3K9me2 and H3K9me3 and, thus, the local structure of chromatin around the p16 promoter may be altered.

Many transcriptional factors or chromatin-binding factors can recruit DNMT1 to the promoters of certain genes, such as the HP1 family (56) and UHRF1 (8). The HP1 family is a recently identified member of DNA-binding proteins which can specifically recognize repressed chromatin characterized by methylated H3K9 (3, 36) and target DNMT1 to their substrate sequence (56). Methylated H3K9 recruits HP1 (including HP1α, HP1β, and HP1γ) to their recognized promoters (21, 44), and in turn the HP1 family causes gene repression by further attracting other repressing factors, such as HDAC (43), SUV39H1 (1), or DNMT1 (56). Although depsipeptide treatment did not affect HP1α and HP1β expression in this study, binding of these factors to chromatin on the p16 promoter was decreased, probably due to the depsipeptide-induced decrease of methylated H3K9 around the p16 promoter.

Based on the above observations, we would like to suggest a mechanism to explain how depsipeptide-induced demethylation of the p16 promoter is linked with reactivation of p16 (Fig. 13). First, depsipeptide directly reduced expression of G9A and SUV39H1, with resultant significant reduction of H3K9me2 and H3K9me3 around the p16 promoter. Subsequently, recruitment of HP1α and HP1β to the p16 promoter is also decreased due to lack of sufficient H3K9me2 and H3K9me3. Finally, loss of binding of HP1α and HP1β to the p16 promoter induces a subsequent decrease in recruitment of DNMT1 and DNMT3A to their specific CpG-containing sites on the p16 promoter, and the methylation pattern of the p16 promoter thus cannot be maintained. In addition, as G9A has been reported to be strongly associated with the activity of DNMT1 (13), the depsipeptide-induced down-regulation of G9A may be one of the reasons for the decreased DNMT1 activity.

FIG. 13.

Hypothetical schematic diagram showing a possible mechanism by which depsipeptide induces demethylation of the p16 promoter and reactivation of silenced p16. All symbols are defined in the diagram.

In this study, we have evaluated several interesting phenomena that are deserving of attention. First, depsipeptide is effective in inducing demethylation in the human lung cancer cell lines H719 and H23, human pancreatic cancer cell line PANC1, and human colon cancer cell line HT29 (Fig. 1 to 3), but not in human colon cancer cell lines HCT116 and SW480 (data not shown). We believe that differences in sensitivity to depsipeptide treatment in these different cell lines are the reason for this variability in depsipeptide-induced demethylation. Here it should be pointed out that the HDAC inhibitors depsipeptide and apicidin are substrates of P-glycoprotein (a product of the multiple drug resistance gene) (31, 69). For example, depsipeptide was actively pumped out of cells by P-glycoprotein in HCT-15/FK228 cells (69), and so HCT-15/FK228 cells are resistant to depsipeptide treatment. SW480 and HCT116 cells show high expression of P-glycoprotein (22, 24, 57), and these cells were resistant to depsipeptide-induced inhibition of cell proliferation in this study, which may thus be due to the inability of depsipeptide to penetrate into these cells. As such, depsipeptide-induced acetylation or demethylation in these cell lines is not detectable. Second, both depsipeptide and TSA, which are typical HDAC inhibitors, were able to induce histone acetylation (Fig. 12A to C), but TSA was not able to induce demethylation (Fig. 12), which is consistent with what has been previously reported (10). We therefore propose that HDAC inhibitor-induced DNA demethylation may be dependent on its structure. Depsipeptide and apicidin belong to a cyclic peptide class of HDAC inhibitors, whereas TSA and SAHA belong to the hydroxamic acids class of HDAC inhibitors (7). Up to this point, there has been no clear mechanism to explain why HDAC inhibitors with different structures show different biological functions. However, different HDAC inhibitors such as depsipeptide and SAHA have been shown to induce different gene expression patterns in identical cell lines. For example, 22% of genes show no overlap of expression in cells treated with both HDAC inhibitors separately, and there is even less overlap in gene expression when comparing the action of these molecules over longer periods of treatment (47, 48). In addition, as we previously reported, depsipeptide could induce p53 acetylation, but TSA could not (73). Finally, histone acetylation is not always accompanied by gene expression. For example, TSA cannot induce gene expression if the genes are still hypermethylated, despite TSA induction of moderate histone acetylation (9, 14, 32). Therefore, induction of different patterns of gene expression by different HDAC inhibitors is not a unique phenomenon.

It was of interest that depsipeptide did not reduce methylation of the distal region of the p16 promoter (region A) (Fig. 1D), but DNMT1 RNAi was obviously able to reduce demethylation in region A (Fig. 6F). We cannot at present explain this inconsistency, but the possibility that other factors may recruit DNMT1 to region A cannot be ruled out. In fact, there was very much less binding of G9A and SUV39H1 to the distal region (region A) of the p16 promoter (Fig. 9C), and depsipeptide had no effect on binding of DNMT1 to this region A (Fig. 6B).

In conclusion, this study presents a novel mechanism for explaining induction of DNA demethylation by the HDAC inhibitor depsipeptide. Our data reveal a link between DNA methylation and histone methylation, in which histone methylation may direct DNA methylation. We believe that the novel HDAC inhibitor depsipeptide or its related HDACi will be very useful for studying the relationship between DNA methylation and histone modifications and may have potential clinical implications for the design of anticancer drugs.