Epigenetic lesions are common in neoplasia and range from hypermethylation of subsets of CpG islands to loss of imprinting. By exploiting an episomal model system and the strong de novo methylation capacity of a human cancer cell line, we show that an H19 minigene rapidly becomes methylated and silenced, mimicking the inactivation of the maternal H19 allele in a range of cancers. Although the H19 imprinting control region (ICR) initially displayed methylation protection, it eventually succumbed to the pressure mounted by the de novo methylation machinery of the JEG-3 cells. Importantly, we were able to visualize the kinetics of the loss of the H19 ICR chromatin insulator function in association with chromatin compaction. Our results document that a strong de novo methylation machinery leads to loss of methylation privilege states of H19 ICR to functionally manifest loss of insulator function in a matter of only a few days in human cancer cells.
The manifestation of genomic imprinting involves the translation of gametic marks into parent of origin-dependent gene expression patterns (1, 2). The neighboring IGF2 and H19 genes emerge as paradigms of genomic imprinting because their expression is monoallelic from opposite parental alleles and governed by shared enhancers (1, 2). The repression of the maternal IGF2 and paternal H19 alleles depends on a differentially methylated ICR3 in the 5′ region of the H19 gene (3, 4, 5, 6, 7, 8) and involves allele-specific interactions between the H19 ICR and the chromatin insulator protein CTCF (8). Because the H19 ICR insulator function is methylation sensitive (9), it is perceived to normally prevent enhancers from communicating with the IGF2 promoters on the maternal allele, but not on the paternal allele (10).
Whereas epigenetic states are normally of high fidelity, epigenetic lesions are common in neoplastic cells, often leading to loss of cellular memory and alterations of gene expression patterns (11, 12). For example, the CTCF target sites of the H19 ICR are often biallelically methylated de novo in human cancers (13), presumably leading to loss of insulator function and LOI of IGF2. The inability of the maternal H19 ICR to maintain protection against de novo methylation might reflect tumor-specific mutations in CTCF that destroy its ability to interact with the H19 ICR, although the low frequency of such cases cannot provide a general explanation for IGF2 LOI (14).
In this report, we identified one cell line with an ability to rapidly de novo methylate the H19 ICR in an episomal assay. De novo methylation of the H19 ICR is accompanied by chromatin compaction and loss of insulator function within a matter of a few days. This report provides the first glimpse into the kinetics of loss of methylation privilege status of the H19 ICR accompanied by loss of insulator function.
Materials and Methods
Cells and Plasmids.
The JEG-3 and Hep3B cell lines were maintained as described previously. The episomal plasmids pREPH19A and pREPH19B, redesignated pREPH19Bm, have been described previously (Ref. 5; Fig. 1 A). The pREPH19Bh plasmid was generated by inserting a 1.3-kb fragment encompassing the human H19 ICR into a multiple cloning site in the 3′ flank of the mouse H19 reporter gene.
The Episome Insulator Assay.
The pREP4-based episomal vectors were transfected into JEG-3 cells, and extracted RNA was subjected to RNase protection expression analysis as described previously using a 365-bp H19 antisense probe and a 150-bp GAP antisense probe as control (5). All procedures were performed according to the manufacturer’s protocol of the RPAIII kit (Ambion). Quantification of individual protected fragments was done using a Fuji FLA 3000 phosphorimager. H19 expression was corrected with respect to both internal control (GAP) and episome copy number as determined by Southern blot analysis of BglII-restricted DNA hybridized with H19 and PDGFB probes (5).
Methylation and Chromatin Conformation Analyses.
Hygromycin-selected cells were equally divided for DNA and RNA extraction as described previously (5). For methylation analysis, the DNA was digested with the methylation-sensitive restriction enzyme HhaI and analyzed by standard Southern blot hybridization protocols (5).
Nuclei from cell clones and mouse fetal liver were isolated and treated with DNase I as described previously (5). Twenty μg of digested DNA were restricted with restriction enzymes, electrophoresed on a 1.7% gel, and blotted to a Hybond N+ membrane (Amersham), followed by hybridization with probes.
Probe 1 was a 3-kb BstZ171/BstZ171 fragment encompassing the 3′ part of the mouse H19 minigene. Probe 2 was a 2.4-kb BamHI/BamHI fragment encompassing 5′ part of the mouse H19 minigene. Probe 3 was a 3.8-kb EcoRV/EcoRV fragment covering the mouse H19 ICR and spacer region. Probe 4 was a 2-kb EcoRI/BglII fragment covering the mouse H19 ICR. Probe 5 was a 0.8-kb EcoRI/XbaI fragment covering the spacer region separating the H19 ICR from the H19 gene. Probe 6 was a 1.3-kb EcoRI/EcoRI fragment covering the human H19 ICR. All probe fragments were radiolabeled using a multiprime labeling kit (Amersham) and [α-32P]dCTP.
Western Blot Analysis.
The protein samples were resolved in 8% SDS-PAGE and electroblotted onto Hybond C filters (Amersham Pharmacia Biotech). Membranes were probed with the anti-CTCF polyclonal antibody (1:500 dilution), followed by incubation with the secondary antirabbit antibody. The membranes were developed with the enhanced chemiluminescence kit (Amersham Pharmacia Biotech) as described by the manufacturer.
JEG-3 Cells de Novo Methylate and Silence the H19 Gene.
We have previously noticed that episomal vectors became de novo methylated in JEG-3 cells (5, 9), but not in Hep3B cells (15). To further penetrate the de novo methylation properties of the JEG-3 and Hep3B cells, the pREPH19A episomal vector containing a mouse H19 minigene and including the ICR in its 5′ flank was transfected into JEG-3 cells. After selection with hygromycin, the methylation status of the episome vector was examined every 4-days, starting from day 15 posttransfection. The extracted DNA was digested with the methylation-sensitive restriction enzymes HhaI and HpaII. Southern blot hybridization analyses of the mouse H19 coding region digested with BstZ17I/HhaI show that the mouse H19 coding region was partially methylated at day 15 and gradually became heavily methylated during a period of 20 days (Fig. 1 B). The fragments emerging above the BstZ17I parent fragment were due to cytosine methylation sensitivity of the BstZ17I restriction enzyme (16). These results show that JEG-3 cells, but not Hep3B cells, possess a strong de novo methylation capacity for the body of the H19 gene.
The processive spreading of the methylation pattern to generate the 3.0-kb fragment suggested a gradual methylation of the promoter. To examine this possibility and how it related to its activity, we digested DNA extracted from the serial passages of the transfected JEG-3 cells as described above with BamHI and HhaI. Fig. 1,B shows, as expected, that the HhaI sites in the H19 promoter were gradually methylated during propagation of the transfected cells. The appearance of the 2.5-kb BamHI fragment in all passages suggests that cis elements at or close to the HhaI sites of the H19 promoter do not protect against methylation. Analysis of mouse H19 expression in transfected cells by RNase protection analyses revealed, not surprisingly, that the activity of the H19 promoter declined concomitant with DNA methylation (Fig. 2, A and B). However, this relationship was not straightforward, suggesting that the transcriptional activity of the reporter gene tolerated a certain degree of methylation in the early stages after transfection of the episome. Alternatively, this result might be explained by the possibility that the formation of a repressive chromatin conformation on the methylated template is time dependent.
The Gradual Erasure of the Methylation Privilege Status of the H19 ICR in JEG-3 Cells.
The kinetics of methylation of the episomal vector during the propagation of the JEG-3 cells prompted us to address whether or not the H19 ICR also succumbed to de novo methylation. Normally, the maternally derived H19 ICR allele remains unmethylated in the soma, documenting a methylation privilege status. The methylation analysis of the HhaI sites in the mouse H19 ICR was carried out by digesting DNA with EcoRI/EcoRV/HhaI. Fig. 3,A shows that the HhaI sites within the H19 ICR were not methylated as compared with the methylation of other sites in the 3′ flank of the H19gene during the initial passages. We note with interest that the one of the HhaI sites, which is part of CTCF binding sites as indicated in Fig. 3,A, eventually did not protect against de novo methylation, as judged by the presence of a parent 3.6-kb EcoRV fragment. This conclusion was further confirmed by using bisulphite sequencing of the region covering CTCF binding sites (Fig. 3,B). To rule out the possibility that the de novo methylation of the H19 ICR depended on particularly low levels of CTCF in JEG-3 cells, we performed Western blot analysis. Fig. 3 C shows that CTCF expression levels in JEG-3 cells are comparable with those in Hep3B cells and higher than those in adult mouse testis. We conclude that the de novo methylation of the H19 ICR in the JEG-3 cell line is not due to down-regulation of CTCF protein.
We have previously shown that the mouse H19 ICR harbors several multiple hypersensitive sites that map to CTCF binding sites on only the unmethylated, maternal allele. To examine the possibility that de novo methylation modified the chromatin conformation, we carried out DNase I hypersensitive studies on episomes containing the mouse H19 ICR derived from 9 and 35 days of propagation after transfection Fig. 4 B shows that the H19 ICR-specific DNase I-hypersensitive sites can be readily seen in transiently propagated episomes, whereas there were no detectable hypersensitive sites present in the longer-term transfectants. Similar results were also obtained using human ICR (data not shown). This observation suggests that the protein factors associated with CTCF target sites are present in JEG-3 cells and maintain the proper, maternal-specific chromatin conformation at the CTCF target sites in the mouse and human H19 ICRs during transient propagation in in vitro cultures. However, these associated protein factors did not protect from de novo methylation spreading over the H19 ICR and were subsequently removed during in vitro propagation of JEG-3 transfectants, as judged by disappearance of DNase-hypersensitive sites.
De Novo Methylation the Human H19 ICR Leads to Loss of Insulator Activity.
Like the mouse H19 ICR, the human H19 ICR fragment inserted in the episomes was methylated gradually at the HpaII sites. However, this effect was faster for the human H19 ICR than for the mouse H19 ICR (Figs. 4,A and 5,A), suggesting inherent differences in chromatin conformation between the human and mouse H19 ICRs. Given the fast rate of de novo methylation of the human H19 ICR in JEG-3 cells, we were interested in assessing potential effects on its insulator function. As has been described for the mouse H19 ICR above, the human H19 ICR efficiently blocks SV40 enhancer-mediated activation when positioned between the enhancer and the promoter of the reporter gene. This is exemplified in Figs. 5 B and C, which show that the insulator function is efficient at 9 days after transfection. However, by the 15th day, most of the insulator function is lost, which in all likelihood is due to the de novo methylation of the methylation-sensitive H19 ICR insulator function (9). Whereas the loss of insulator function is a fact within 6 days, the epigenetic silencing of the reporter gene is a slower event.
We show here that an H19 minigene containing an ICR in its 5′ region becomes rapidly methylated and silenced in a cancer cell line. Using this assay system, we were also able to show that the methylation privilege status of the H19 ICR was eventually eroded away, despite initial methylation protection. The de novo methylated sequences included CTCF target sites were concomitant with loss of nuclease hypersensitivity at the chromatin level. These data are compatible with the interpretation that CTCF does not protect against methylation by the DNA methylating machinery of the JEG-3 cell line and that CTCF was eventually depleted from the H19 ICR chromatin. Concomitant with this process, the insulator function was lost, leading to activation of the reporter gene in a matter of only a few days. Because the reporter gene eventually succumbed to epigenetic silencing, our model system mimics both IGF2 and H19 LOI. Interestingly, the loss of insulator function of the H19 ICR preceded epigenetic silencing of the H19 promoter by several weeks. This is not entirely surprising because methylation-mediated silencing appears to generally require a threshold density of methylated CpGs (17). Hence, it seems obvious that the methylation of the entire H19 ICR would be more time-consuming than the methylation of the few CpGs in the CTCF target sites.
Our data provide the first glimpse on the kinetics of events that are synonymous to the LOI reported at the human IGF2 and H19 loci in numerous cancer types and prove that the de novo methylation machinery of the JEG-3 cells break down methylation-protected domains by brute force. The possibility that this effect depends on deregulation of methyltransferases is supported by the observation that DNMT1 overexpression leads to biallelic H19 ICR methylation (18). In contrast to this scenario, however, there is only a limited up-regulation of DNA methyltransferase expression levels in colon cancer cells, despite widespread alterations in methylation levels (19). An alternative mechanism to manifest abnormal de novo methylation in tumor cells therefore could depend on inappropriate expression of the methyltransferase genes during the cell cycle (20). We also cannot rule out the possibility that deregulation of DNA methyltransferase-associated factors can also contribute to de novo methylation of episomal plasmids in the JEG-3 cell line.
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.
Supported by the Swedish Science Research Council, the Swedish Cancer Research Foundation, the Swedish Pediatric Cancer Foundation, and the Lundberg Foundation.
The abbreviations used are: ICR, imprinting control region; LOI, loss of imprinting; GAP, glyceraldehyde 3-phosphate dehydrogenase.