DNA Demethylation Reactivates a Subset of Imprinted Genes in Uniparental Mouse Embryonic Fibroblasts*

Although most imprinted genes show allelic differences in DNA methylation, it is not clear whether methylation regulates the expression of some or all imprinted genes in somatic cells. To examine the mechanisms of silencing of imprinted alleles, we generated novel uniparental mouse embryonic fibroblasts exclusively containing either the paternal or the maternal genome. These fibroblasts retain parent-of-origin allele-specific expression of 12 imprinted genes examined for more than 30 cell generations. We show that p57Kip2 (cyclin-dependent kinase inhibitor protein 2) and Igf2 (insulin-like growth factor 2) are induced by inhibiting histone deacetylases; however, their activated state is reversed quickly by withdrawal of trichostatin A. In contrast, DNA demethylation results in the heritable expression of a subset of imprinted genes including H19(H19 fetal liver mRNA), p57Kip2 ,Peg3/Pw1 (paternally expressed gene 3), andZac1 (zinc finger-binding protein regulating apoptosis and cell cycle arrest). Other imprinted genes such as Grb10(growth factor receptor-bound protein 10), Peg1/Mest(paternally expressed gene 1/mesoderm-specific transcript),Sgce (epsilon-sarcoglycan), Snrpn (small nuclear ribonucleoprotein polypeptide N), and U2af1 (U2 small nuclear ribonucleoprotein auxiliary factor), remain inactive, despite their exposure to inhibitors of histone deacetylases and DNA methylation. These results demonstrate that changes in DNA methylation but not histone acetylation create a heritable epigenetic state at some imprinted loci in somatic cells.

Normal development of mammalian embryos requires the genetic contribution of both maternal and paternal genomes (1,2). Uniparental mouse embryos, in which the entire genome is either of maternal (parthenotes) or paternal (androgenotes) origin, usually die during early stages of embryogenesis (2,3). Lethality of uniparental embryos appears to be caused by either the lack of, or overexpression of, specific genes that are imprinted and only expressed from the nonimprinted parental allele. More than 35 autosomal genes in the mouse exhibit parent-of-origin, allele-specific imprinting in embryonic and adult tissues. The imprint, once set, is stable during mitosis but is reversed and reset by passage through meiosis during gametogenesis (4,5). Given the epigenetic nature of imprinting, much attention has centered on whether DNA methylation is crucial to the establishment and maintenance of the silent imprinted allele (6,7). Most imprinted genes show parental differences in methylation patterns (7)(8)(9)(10), although the extent varies among the different genes. Furthermore, despite changes in global levels of DNA methylation in early embryogenesis, methylation of some imprinted genes remains constant even at the blastocyst stage, whereas the rest of the genome is hypomethylated (8,11,12). After implantation, most of the CpG sequences are progressively methylated, except those located in the promoter region of active "housekeeping" genes (13). These findings together with the derivation of embryos deficient in methyltransferases, in particular DNA methyltransferase 1, suggest that proper expression of some imprinted genes (10,14) requires DNA methylation. However, it is unclear whether methylation is involved in the regulation of some or all imprinted alleles. If methylation regulates the monoallelic expression of imprinted genes, demethylation should reactivate silent alleles in vitro. We tested this hypothesis by analyzing the reactivation of imprinted alleles in a series of novel nonexpressing uniparental mouse embryonic fibroblasts (MEFs) 1 treated by demethylating agents 5-azacytidine (AzaC) or 5-aza-2Ј-deoxycytidine (AzadC) alone or in combination with a histone deacetylase inhibitor, trichostatin A (TSA). Our results suggest that both DNA methylation and histone deacetylase activities regulate the differential allelic expression of some but not all imprinted genes in somatic cells and that the epigenetic modifications required for maintenance of monoallelic expression vary among different imprinted loci. Significantly, the activated state induced by DNA demethylation, but not by histone acetylation, is propagated and stably inherited during mitosis, indicating that methylation is required for long term repression of some imprinted genes.

Derivation of Wild-type (WT) and Uniparental
MEFs-WT and uniparental MEFs were derived from explanted day 13 (day of plug ϭ day 1) embryos after removing the head and internal organs. Androgenetic (AG) MEFs were generated from chimeras made by injecting androgenetic ES cells, constitutively expressing the Neo gene (15,16), into blastocysts. Parthenogenetic (PG) MEFs were generated from PG 171 WT chimeric embryos. PG embryos were derived by ethanol activation of eggs, with suppression of polar body formation, and were then aggregated at the 4-or 8-cell stage with WT embryos (17). The PG eggs were derived from mouse lines that constitutively expressed the Neo gene (15). The AG and PG MEFs were derived by culturing the primary explants in media supplemented with G418 for at least 1 week which selected against the WT cells lacking the NeoR gene. * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Treatment of Cells with Inhibitors of DNA Methylation and Histone
Deacetylases-Uniparental MEFs were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and Pen/ Strep. The day before treatment, cells were split to 50 -60% density. For the experiment shown in Fig. 2 RT-PCR Analysis of Imprinted Gene Expression-Total RNA was extracted from WT and uniparental MEFs using the RNeasy kit (Qiagen) and treated with RNase-free DNase I (Promega) to eliminate residual genomic DNA. 1.5 g of RNA was converted to cDNA using random primers and avian myeloblastosis virus reverse transcriptase in a 20-l reaction for 60 min at 42°C. PCR was performed with 1-2 l of cDNA in 25-50 l using Taq polymerase (Roche Molecular Biochemicals). Amplification consisted of 25 or 30 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 60 s performed in a PerkinElmer Life Sciences GeneAmp PCR machine 9600. Primers for all target genes are listed in Table I.
For quantitative evaluation of the expression of imprinted genes in untreated, TSA-and/or AzadC-treated AG and PG MEFS, total RNA was converted to cDNA, and quantitative PCR was performed using serial dilutions to assure a linear amplification of the target and control genes (18). The levels of target gene expression were measured relative to the housekeeping genes Hprt (hypoxanthine phosphoribosyltransferase) and Gapd (glyceraldehyde-3-phosphte dehydrogenase) as internal controls. The amplified products were analyzed by Southern blot-ting and quantified by a PhosphorImager Storm 860 (Molecular Dynamics).
RNase Protection Assays-Total RNA (10 g) isolated from untreated, TSA-and/or AzadC-treated MEFs was incubated overnight at 45°C with radiolabeled probe (ϳ10 6 cpm) in a 20-l reaction mixture containing hybridization buffer (Ambion, Inc). The reaction mixtures were digested with RNases A and T1 and subsequently analyzed by electrophoresis on polyacrylamide gel and visualized by autoradiography. The relative intensities of the target and control mRNAs were quantified by phosphorimaging.
Western Blotting of Acetylated Histones H3 and H4 -Uniparental MEFs cultured for 24 h in the absence or presence of TSA (0.2, 1, or 5 M) were harvested and nuclei prepared as described previously (19). Nuclei (3-5 ϫ 10 6 ) were resuspended in lysis buffer (10 mM HEPES, pH 7.5, 1.5 mM MgCl 2 , 10 mM KCl, 1 mM dithiothreitol, 1 mM AEBSF (ICN) and proteinase inhibitors complete (Roche)), then SDS was added to final concentration of 1%. 20 g of proteins from treated and untreated AG and PG MEFs were separated by 16.5% SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose membrane. Hyperacetylated histones were detected by anti-acetylated H3 and H4 antibodies (Upstate Biotechnology) and were visualized by chemiluminescence (Amersham Pharmacia Biotech). As control for the amount of protein loading, a parallel gel was stained with Coomassie Blue.
DNA Extraction and Methylation Assay-Genomic DNA was extracted by proteinase K/SDS digestion, phenol/chloroform extraction, and isopropyl alcohol precipitation. For p57 Kip2 (cyclin-dependent kinase inhibitor protein 2) analysis, DNA was first digested with HindIII and XbaI followed by incubation with either HpaII or MspI for at least 16 h at 37°C. For the methylation analysis of U2af1 (U2 small nuclear ribonucleoprotein auxiliary factor) and Snrpn (small nuclear ribonucleoprotein polypeptide N) genes, DNA was digested first with BamHI and HindIII followed by HpaII or MspI. Digested DNA was analyzed by Southern blotting as described previously (19).

Expression of Imprinted Genes in Uniparental
MEFs-We generated uniparental diploid MEFs containing exclusively either the paternal AG or the maternal PG genome. In these uniparental cells, the epigenetic regulation of both alleles should be identical, with an imprinted gene being either expressed or silent depending on the parent-of-origin profile of imprinting. Probes to a panel of imprinted genes showed that these cell lines stably retain parent-of-origin allele-specific imprinting status over 30 cell generations (Fig. 1). The paternally expressed genes, Igf2 (insulin-like growth factor 2), Peg1/Mest (paternally expressed gene 1/mesoderm-specific transcript), Peg3/Pw1 (paternally expressed gene 3), Snrpn, and U2af1 (20 -24) as well as the two newly identified imprinted genes Sgce (epsilon-sarcoglycan,) and Zac1 (zinc finger-binding protein regulating apoptosis and cell cycle arrest) (25) were detected only in AG and WT MEFs. In contrast, the transcripts of maternally expressed H19 (H19 fetal liver mRNA), Grb10/ Meg1 (growth factor receptor-bound protein 10/maternally expressed gene 1), and p57 Kip2 (26 -28) were detected exclusively in PG MEFs and WT MEFs. The one exception was the Igf2r/ M6Pr (insulin-like growth factor 2 receptor/mannose 6-phosphate receptor) (29). This gene is strongly expressed from the maternal allele (in PG and WT MEFs), but low levels of expression were also detected in the AG MEFs, indicating that the silencing of the paternal allele was not fully acquired in day 13 uniparental embryonic tissues. Furthermore, the tissue-specific imprinted genes Rasgrf1 (Ras protein-specific guanine nucleotide-releasing factor 1), which is expressed exclusively from the paternal allele in the brain, heart, and stomach (30), and Mash2 (mammalian achaete-scute homologue 2), a trophoblast-specific maternally expressed gene (31), were not detected in WT and uniparental MEFs. The housekeeping genes, Gapd and Hprt, were expressed in all MEFs. These results demonstrate that AG and PG MEFs retain the correct expression pattern of 12 imprinted genes and provide unique and significant advantages to analyze mechanisms leading to allele-specific silencing of these genes.

5Ј-GGCTGTACTGCTTAACCAGGGAAAG
a Primers for H19, Igf2, Sgce, and Zac1 were originally described in Refs. 21 and 25. Reactivation of Imprinted Genes in Nonexpressing Cells-To examine the roles that DNA methylation and histone deacetylation play in differential expression of imprinted genes, uniparental AG and PG MEFs were treated with increasing concentrations of either a specific inhibitor of histone deacetylase (32), TSA (0.2-5 M, Fig. 2) or with inhibitors of DNA methylation, AzaC/AzadC (1-10 M) for 24 h. RT-PCR was performed to determine whether imprinted genes were induced by drug treatment in nonexpressing cells. TSA treatment resulted in a dose-dependent induction of Igf2 and p57 Kip2 in PG and AG MEFs, respectively ( Fig. 2A). A modest induction of both Igf2 and p57 Kip2 was observed in cells treated with a low dose of TSA (0.2 M). TSA reactivation was significantly higher at 1 M and decreased at 5 M, possibly because of its adverse effect on cell proliferation and viability (33). Unlike Igf2 and p57 Kip2 , all of the other imprinted genes listed in Fig. 1 remained silent, despite induced accumulation of hyperacetylated histones H3 and H4 after culturing uniparental cells in the presence of different doses of TSA (Fig. 2B). Acetylation of core histones, in particular H3 and H4, is believed to play a role in chromatin unfolding and transcription regulation, as actively transcribed genes are enriched in hyperacetylated histones H3 and H4 (33)(34)(35). Western blot analysis showed that levels of acetylated H3 and H4 were very low (detectable only at high exposure, data not shown) in untreated AG and PG MEFs. Incubation with 0.2-5 M TSA resulted in the accumulation of acetylated H3 and H4 as determined by antibodies directed against acetylated histones ( Fig. 2B; 0.2, 1, and 5 M TSA). Under these experimental conditions, among the imprinted genes analyzed (Fig. 1), only Igf2 and p57 Kip2 were reactivated. At this point we cannot rule out whether the induction of these two genes is triggered by a locus-specific histone acetylation or by a secondary effect of a global histone acetylation.
In contrast, treatment with demethylating agents AzaC or AzadC for 24 h was not sufficient to induce detectable levels of imprinted gene mRNA in nonexpressing cells (data not shown). However, treatment with high doses of AzaC and AzadC (5-10 M) for over 24 h was toxic and inhibited cell proliferation. To minimize the adverse effects of drug treatment on cell proliferation, a protocol was elaborated in which uniparental AG and PG MEFs were treated for 3 days with low dose of AzaC, AzadC, and/or TSA (see "Materials and Methods"). As shown in Fig. 3, this resulted in the activation of a subset of imprinted genes. Peg3 and Zac1 were reactivated in PG MEFs and, p57 Kip2 and H19 in AG cells after exposure of cells to either AzaC or AzadC. Igf2 was induced in PG MEFs treated with TSA alone and was not responsive to demethylating agents, whereas both TSA and AzaC/AzadC synergistically activated p57 Kip2 in AG MEFs. In contrast, Grb10, Peg1, Rasgrf1, Sgce, Snrpn, and U2af1 remained inactive after both treatments. The effects of TSA and AzadC on the expression of Igf2r were not assessed because the gene is not fully silenced in AG MEFs.
The levels of expression of a gene paternally imprinted, p57 Kip2 , and two maternally imprinted genes, Peg3 and U2af1, were evaluated further by RNase protection assay and/or quantitative RT-PCR (Fig. 4). Measurement by phosphorimaging analysis of changes in p57 Kip2 mRNA levels revealed a significant induction of p57 Kip2 expression after treatment with either TSA or AzadC (Fig. 4A). However, the combined treatment with AzadC and TSA synergistically activated p57 Kip2 expression to about 5-fold higher levels than treatment with either AzadC or TSA alone. These results suggest that both histone deacetylase and DNA methylation activities mediate silencing of the paternal allele of p57 Kip2 in cultured MEFs. Unlike p57 Kip2 , Peg3 was activated in PG MEFs treated with AzadC alone, whereas the addition of TSA antagonized the induction by AzadC in a combined treatment (Fig. 4B), demonstrating that in this case the hyperacetylation of histones has a negative effect on gene expression. Under the same protocols, treatment of PG MEFs with AzadC and/or TSA did not significantly change the expression levels of U2af1, Hprt, or Gapd (Fig. 4, B and C). Overall the results shown in Figs. 2-4 suggest that only a subset of imprinted genes is activated in response to inhibitors of histone deacetylation and/or DNA methylation.
AzaC and AzadC Treatments Correlate with DNA Demethylation-We analyzed the methylation status of p57 Kip2 (induced by TSA and AzadC) and U2af1 (not responsive to either TSA or AzadC treatment) by comparing the digestion of genomic DNA from untreated, TSA-, AzaC-, or AzadC-treated AG and PG MEFs (Fig. 5). After treatment, cells were cultured in the absence of drugs for 3-7 days, and DNA was purified and digested with the methylation-sensitive HpaII or insensitive MspI enzymes and analyzed by Southern blotting. Blots were hybridized with probes derived from the promoter region of p57 Kip2 (Fig. 5A) and U2af1 (Fig. 5B). DNA from WT MEFs, which carry both active and silent alleles of each imprinted gene, was used as a control.
In all samples, DNA from nonexpressing cells (AG for p57 Kip2 and PG for U2af1), either untreated or TSA-treated, was relatively resistant to HpaII digestion, but was cut by MspI to produce heterogeneous fragments of various sizes (Fig. 5, A and  B). The results demonstrated that in nonexpressing cells, p57 Kip2 (in AG MEFs) and U2af1 (in PG MEFs) genes are densely methylated, whereas in the expressing cells DNA is unmethylated, as it is digested to the same extent with either HpaII or MspI (compare Fig. 5A, lanes 6 and 7, with Fig. 5B,  lanes 7 and 9). After treatment with AzadC, p57 Kip2 DNA from AG MEFs was readily digested with HpaII (Fig. 5A, compare  lanes 2 and 4), demonstrating that AzadC induced extensive demethylation at this locus, particularly with regard to the appearance of a major band at 0.4 kilobase. However, DNA from the same cells treated with TSA (which also activated p57 Kip2 ) showed only minor changes in HpaII digestion, indicating that TSA did not cause significant demethylation as seen in AzadC-treated samples. Similar results were obtained when blots were hybridized with U2af1 (Fig. 5B). DNA from AzaC-or AzadC-treated PG cells was partially digested with HpaII, whereas DNA from untreated cells was fully resistant to digestion (compare lanes [2][3][4][5]. Thus, under these experimental conditions DNA is significantly demethylated after AzaC or AzadC treatment, whereas the same sequences in untreated cells are densely methylated. However, despite significant levels of DNA demethylation, U2af1 and other imprinted genes remained silent after treatment (Figs. 3 and 4), suggesting that partial demethylation was not sufficient to reactivate these silent alleles.
Transient Inhibition of DNA Methylation, but Not Histone Deacetylase, Promotes a Heritable State of Gene Expression-In mammals, propagation of the methylation state within CpG dinucleotides by the maintenance DNA methyltransferase occurs during or shortly after replication (36). Inhibition of DNA methylation by AzaC or AzadC leads to the reactivation of some imprinted genes, specifically H19, Peg3, p57 Kip2 , and Zac1, whereas TSA treatment induces two genes, p57 Kip2 and Igf2 (Fig. 3). To determine whether these induced patterns of expression are maintained after multiple cell divisions, uniparental MEFs initially treated for 3 days with TSA, AzaC, or AzadC (Fig. 6, lanes P1) were cultured for 7 days (lanes P2) or 15 days (lanes P4) after withdrawal of the inducing agents. The results shown in Fig. 6 revealed that the expression of H19, Peg3, p57 Kip2 , and Zac1 induced by transient treatment with either AzaC or AzadC is enriched and maintained over the course of multiple cell divisions in the absence of either inhibitor (compare P1 with P4). However, the TSA-induced transient expression of p57 Kip2 and Igf2 was quickly reversed to a silent state after two to four cell cycles after drug withdrawal. These results demonstrate that changes in DNA methylation, but not levels of histone acetylation, create a dominant and heritable epigenetic state of gene expression at some imprinted loci.

DISCUSSION
In mammals, genomic imprinting results from the differential epigenetic modification to a subset of genes in the germ line, leading to parent-of-origin monoallelic expression during embryogenesis and in adult tissues. Studying the molecular basis that distinguishes the paternal and the maternal alleles of imprinted genes has been complicated by the simple fact that all somatic tissues contain both maternal and paternal genomes. This necessitated the use of a variety of techniques to distinguish between the parental alleles, such as polymorphisms generated by intercrosses or the introduction of mutations or deletions into different imprinted genes. Here we have greatly simplified the analysis of imprinted gene expression by generating novel uniparental mouse fibroblast lines retaining parent-of-origin patterns of imprinting. We showed that the parental specific expression is stably inherited during multiple cell generations in vitro, demonstrating that these uniparental cells provide unique advantages for analyzing the mechanisms of silencing of imprinted alleles. Treatment of uniparental cells  2-4, 7, 8, and 10) or MspI (lanes 5, 6, and  11) or left undigested (u, lanes 1 and 9) as indicated on the top of the panel. All DNAs in panel A were digested with Hin-dIII/XbaI and analyzed by Southern blotting using a probe that hybridizes to the promoter region of p57 Kip2  with inhibitors of histone deacetylases or DNA methyltransferases resulted in activation of only a subset of imprinted genes. In PG cells, both Zac1 and Peg3 were reactivated in cells treated with AzaC or AzadC, whereas Igf2 was induced only by TSA. In AG MEFs, p57 Kip2 was synergistically activated by combination of both AzadC and TSA, and H19 was induced after AzadC treatment but not with TSA. However, under identical experimental conditions Grb10, Peg1, Sgce, Snrpn, and U2af1, as well as the tissue-specific imprinted genes Ras-grf1 and Mash2, remained silent after all treatments, demonstrating that not all imprinted genes are responsive to changes in DNA methylation or histone acetylation levels in somatic tissues.
Recently DNA methylation and histone deacetylation have been functionally linked to transcriptional repression. Methyl-CpG-binding proteins MeCP2 (37,38), MBD2 (39), and MBD3 (40,41) reside in multiprotein complexes with histone deacetylases. These complexes assemble on methylated DNA mediating transcriptional repression through chromatin hypoacetylation with the silent state being partially reversed by TSA. Consistent with this model, we have shown that p57 Kip2 is responsive to inhibitors of both DNA methylation and histone deacetylase (Figs. [2][3][4]. This suggests that the silent paternal allele of p57 Kip2 may be associated with a repressor complex similar to those described above and may contain both histone deacetylase and methyl-binding proteins responsible for targeting the complex to methylated DNA. In contrast, H19, Peg3, and Zac1 were only responsive to loss of methylation but not to an inhibitor of histone deacetylase, suggesting that complexes containing methyl-binding proteins but not histone deacetylase may be involved in silencing at these loci. Whether the reactivation of all these genes is caused by direct alteration of their "imprint" or by indirect changes in methylation elsewhere, remains to be determined. However, simultaneous analysis of the methylation levels of imprinted genes (Fig. 5) in both AG and PG MEFs suggests that treatment affects global DNA methylation rather than some gene-specific regulatory process. Genetic evidence has implicated DNA methylation in silencing imprinted alleles of H19, Igf2, Igf2r, and p57 Kip2 because a loss-of-function mutation of the maintenance DNA methyltransferase gene (Dnmt1) resulted in biallelic expression of both H19 and p57 Kip2 (10,14) but repression of Igf2 and Igf2r expression (10). In agreement with the genetic data, we showed that in vitro demethylation caused reactivation of a subset of imprinted genes, including p57 Kip2 and H19 in nonexpressing AG MEFs, and the newly identified Zac1 gene and Peg3 from the maternal alleles in PG MEFs. Under these experimental conditions, the reactivation of Zac1 and Peg3 suggests that methylation is also required to maintain the silent state at these imprinted loci. The expression level of Snrpn also appeared to increase in Dnmt1 null mice (42); however, it is not clear whether the elevated expression is caused by an increase in the transcription from the active paternal allele or to the reactivation of the silent maternal allele. Our results show that loss in DNA methylation in PG MEFs (Fig. 5) did not result in detectable changes in expression of Snrpn (Fig.  3). Other reports (43)(44)(45)(46) have also shown that Igf2, H19, and p57 Kip2 are responsive to inhibitors of DNA methylation and histone deacetylases in a number of tissues and cell culture systems. Pedone et al. (45) suggested that Igf2 may also be induced by inhibitors of DNA methylation, and H19 may be activated in a combined treatment with AzaC ϩ TSA or sodium butyrate. In a DNA methyltransferase null background (Dnmt1 Ϫ/Ϫ) Igf2 is not expressed, with both the maternal and paternal alleles being silent and H19 being expressed from both alleles. These conflicting results suggest that the relaxation of the coregulated Igf2 and H19 imprinting is complex and that epigenetic changes elsewhere in the genome may also affect their expression under certain conditions.
Here we showed (Figs. 3 and 6) that the same treatment that led to reactivation of H19, p57 Kip2 , Peg3, and Zac1 did not change the silent state of Grb10, Peg1, Sgce, Snrpn, and U2af1, despite the paternal and maternal alleles of these genes being differentially methylated (7,42,47) and that treatment with AzadC resulted in substantial demethylation of both p57 Kip2 and U2af1 (Fig. 5). This suggests that some imprinted genes (Grb10, Peg1, Sgce, Snrpn, U2af1) are associated with repressive heterochromatin-like structures that may be propagated independently of DNA methylation and/or levels of histone acetylation. Similar mechanisms were also suggested to explain the different rates at which HPRT and PGK were reactivated on the silent X chromosome after treatment with AzaC (48). If this is the case, whatever the nature of the initial imprint is, once the silencing is set, the repressive state becomes mitotically stable and reversible only by passage through meiosis (4,5). Partial DNA demethylation in MEFs is therefore not sufficient to induce detectable expression of these imprinted genes. Also, the observed differential methylation between the expressed and silent alleles might be a consequence and not the cause of silencing. In this case, DNA demethylation would not affect the silencing. Detailed analysis of chromatin structure and composition may help elucidate the mechanisms orchestrating allele-specific silencing of this group of imprinted genes.
Propagation of epigenetic states of expression during multiple cell divisions is essential for the stability of imprinting. Remarkably, we observed that the activated state of H19, p57 Kip2 , Peg3, and Zac1 mediated by DNA demethylation was enriched and stably propagated through multiple mitoses (Fig.  6). In contrast, the activated state of Igf2 and p57 Kip2 , induced by transient inhibition of histone deacetylase, was quickly reversed after withdrawal of TSA from the culture medium, suggesting that the histone deacetylase activity associated with silencing Igf2 and p57 Kip2 is only temporarily inhibited but not disrupted by TSA treatment. These results support a previous observation (4) that changes in DNA methylation, but not histone acetylation, create a heritable epigenetic state at some, but not all, imprinted loci in somatic cells.
Here, we have shown that uniparental AG and PG MEFs provide a unique model to study the regulation of imprinted gene expression in vitro. In these lines, as in the whole mouse, imprinted genes are expressed according to their parental origin. Among the imprinted genes examined, four were responsive to partial loss of methylation (i.e. H19, p57 Kip2 , Peg3, and Zac1), suggesting that DNA methylation is the primary silencing mechanism for this set of imprinted genes and that additional mechanisms may regulate silencing of other imprinted genes in somatic cells.