CTCF Is the Master Organizer of Domain-Wide Allele-Specific Chromatin at the H19/Igf2 Imprinted Region

Abstract

A paternally methylated imprinting control region (ICR) directs allele-specific expression of the imprinted H19 and Igf2 genes. CTCF protein binding in the ICR is required in the maternal chromosome for insulating Igf2 from the shared enhancers, initiation of the H19 promoter transcription, maintaining DNA hypomethylation, and chromosome loop formation. Using novel quantitative allele-specific chromatin immunoprecipitation-single-nucleotide primer extension assays, we measured the chromatin composition along the H19/Igf2 imprinted domain in cells with engineered mutations at the four ICR-CTCF binding sites. Abolishing CTCF binding in the ICR reduced normally maternal allele-specific H3K9 acetylation and H3K4 methylation at the H19 ICR and promoter/gene body and maternal allele-specific H3K27 trimethylation at the Igf2 P2 promoter and Igf2 differentially methylated regions (DMRs). Paternal H3K27 trimethylation and macroH2A1 became biallelic in the mutant cells at the H19 promoter while paternal H3K9 acetylation and H3K4 methylation became biallelic at the Igf2 DMRs. We provide evidence that CTCF is the single major organizer of allele-specific chromatin composition in this domain. This finding has important implications: (i) for mechanisms of insulation since CTCF regulates chromatin at a distance, involving repression by H3K27 trimethylation at the Igf2 locus independently of repression by DNA hypermethylation; and (ii) for mechanisms of genomic imprinting since point mutations of CTCF binding sites cause domain-wide “paternalization” of the maternal allele's chromatin composition.

Enhancer-promoter interactions may form at long distances in the genome. It is the role of insulators to ensure that only the appropriate regulatory elements interact in any given cell. CTCF insulator protein (3, 13, 27) binds to the human genome at every 2.5 genes, on average (19), and is essential for development, most likely due to the importance of its insulator activity. Nevertheless, the mechanism of insulation remains unclear. Here, we investigated the mechanism of CTCF insulation at the level of chromatin organization at the well-characterized H19/Igf2 imprinted domain (Fig. 1A).

FIG. 1.

Validation of MEFs for chromatin analysis. (A) Features of the H19/Igf2 imprinted domain in wild-type 129 × CS MEFs. H19 and Igf2 are expressed from the maternal (M) or paternal (P) allele, respectively (large arrows). In the opposite alleles both genes are silent (X). The paternal allele of the ICR is methylated (black lollipops). The maternal ICR allele is unmethylated (white lollipops) and binds CTCF insulator protein (vertical ovals) at four sites. CTCF blocks the interaction between the downstream enhancers (small horizontal ovals) and the Igf2 promoters (only P2 is indicated for simplicity). The Igf2 DMRs (gray rectangles) are paternally hypermethylated (lollipops in shades of gray). Certain regions are indicated according to their distance (−8, −4, and −3 kb) from the H19 transcription start site. (B) Summary of the changes in CTCFm × CS MEFs. (C) In vivo CTCF binding in the ICR is present in 129 × CS MEFs (white bars) but is absent in CTCFm × CS MEFs (black bars) due to the point mutations at four CTCF binding sites. Results of a ChIP real-time PCR experiment are shown. The left (−4 kb) and right (−3 kb) halves of the ICR were tested. (D) Igf2 levels are about double while H19 levels are greatly reduced in CTCFm × CS MEFs compared to normal cells. Results of a Northern hybridization are shown. Mutant versus wild-type relative values after normalization to Gapdh values are shown above the mutant bands. (E) Igf2 expression becomes biallelic in CTCFm × CS MEFs due to lack of ICR insulation. Allele-specific gene expression was assessed by RT-PCR SNuPE experiments (37). Numbers above images are relative expression levels of the presumed inactive alleles, the maternal allele of Igf2 (M) or the paternal allele (P) of H19, as a percentage of total (M + P) expression. Quantitation controls (0, 50 and 100% CS/total) are shown on the left. (F) ICR DNA methylation is biallelic in CTCFm × CS MEFs, because CTCF binding no longer protects the ICR from de novo methylation (39). Bisulfite methylation analysis results are shown. Individual, maternally (M) or paternally (P) inherited normal chromosomes or maternally inherited CTCFm chromosomes (M CTCFm) are shown. Methylated and unmethylated CpGs are indicated by closed and open squares, respectively. (G) Bisulfite methylation analysis of the H19 promoter sequences. DNA methylation is paternal allele-specific in normal and CTCFm × CS MEFs. The transcription start site (horizontal arrow) is indicated.

The paternally expressed Igf2 (insulin-like growth factor 2) and maternally expressed H19 genes are located on mouse distal chromosome 7 (4). A 2.4-kb, paternally methylated, germ line differentially methylated region (DMR) between Igf2 and H19 (1, 41, 42) is responsible for monoallelic expression of both H19 and Igf2 (22, 31, 40) and is therefore known as an imprinting control region (ICR). Igf2 expression is also regulated by two additional DMRs. Igf2 DMR1 lies upstream of the Igf2 gene and functions as a mesodermal silencer in the maternal allele (6), while DMR2, located in the sixth exon, functions as an enhancer in the paternal allele (28). Methylation-dependent interactions between the ICR and the Igf2 DMR1 or DMR2 allele result in chromosomal loop formation in the maternal or paternal allele, respectively (29).

The ICR had been suggested to function as a methylation-regulated insulator, because CTCF insulator protein is associated with the unmethylated maternal ICR allele (2, 14, 17, 18, 36), and, therefore, insulation in this position could block the Igf2 promoters from the shared enhancers (16, 23) specifically in the maternal allele. Targeted inactivation of the CTCF binding sites in the mouse has revealed four distinct roles of CTCF in the H19/Igf2 imprinted domain. One of these roles was enhancer insulation. There was a loss of enhancer-blocking activity in the mutant maternal chromosome, evident in biallelic Igf2 transcription and consequently increased body size (30, 33, 39). CTCF was also found to have a role in maintenance of the hypomethylated state of the maternal allele in somatic cells. The maternal ICR, as well as the H19 promoter and gene body became highly methylated in fetal organs when CTCF binding in the maternal ICR was inhibited by mutations. Promoter methylation silenced the H19 gene in the maternal allele in the fetus (10, 30, 33, 39). Elevated CpG methylation was also found in the maternal alleles of DMR1 and DMR2 upon maternal inheritance of the CTCF site mutations (Fig. 2C and D) (21). Transcription initiation at the H19 promoter was determined to be a third role of CTCF. CTCF site mutant embryos lacked embryonic H19 transcription (10). A fourth CTCF role was revealed to be ICR-DMR1 chromosome loop formation in the maternal chromosome. Maternal inheritance of the ICR CTCF site mutations abolished the ICR-DMR1 long-range interaction (21).

FIG. 2.

CTCF binding in the ICR is required for hypomethylation of the Igf2 DMR1 and DMR2. Bisulfite methylation analysis of the Igf2 DMR1 (A) and DMR2 (B) in 129 × CS and in CTCFm × CS MEFs is shown. DNA from reverse cross-linked input chromatin preparations (N- and X-ChIP) was analyzed to match the ChIP experiments. Other details are as described in the legend of Fig. 1F. (C) Summary of the methylation levels. Maternal (black bars) and paternal (gray bars) allele-specific CpG methylation at the Igf2 DMR1 (primary data from panel A) and DMR2 (primary data from panel B) in MEFs was compared to kidneys and livers of 17.5 dpc fetuses (primary data not shown). The maternal alleles of DMR1 and DMR2 were relatively hypomethylated compared to paternal alleles in all normal 129 × CS but not in mutant CTCFm × CS samples. (D) Summary of the methylation changes. Mutant maternal alleles in CTCFm × CS MEFs were hypermethylated compared to normal maternal alleles in response to distant ICR CTCF site mutations. The differences are calculated from data shown in panel C. pat, paternal; mat, maternal.

Our hypothesis was that CTCF insulation at the H19/Igf2 imprinted domain involves chromatin organization locally around the ICR binding sites and possibly also at a considerable distance, at the insulated Igf2 promoter and at the Igf2 DMRs. The ICR exhibits allele-specific differences in chromatin composition (5, 7, 43), but it is not known how these and putative additional differences at the Igf2 locus are established and maintained or if they are related to CTCF binding. Using cells from genetically modified mice and novel, quantitative and allele-specific chromatin immunoprecipitation-single-nucleotide primer extension (ChIP-SNuPE) assays, we found evidence that CTCF is, indeed, the single major organizer of allele-specific chromatin along the H19/Igf2 imprinted domain in somatic cells. This finding has important implications: (i) for mechanisms of insulation since CTCF regulates chromatin at a distance, involving repression by H3K27 trimethylation at the Igf2 locus independently of repression by DNA methylation; and (ii) for mechanisms of genomic imprinting since point mutations of CTCF binding sites cause domain-wide “paternalization” of the maternal allele's chromatin composition.

MATERIALS AND METHODS

Source of MEFs.

Female 129S1 (hereafter, 129) mice were mated with males homozygous for the Mus musculus castaneus form of distal chromosome 7 (hereafter, CS), as derived from strain CAST/Ei (The Jackson Laboratory, Bar Harbor, ME). These males were of strain FVB/NJ.CAST/Ei(N7) (38). Primary murine embryonic fibroblasts (MEFs) were derived from the resulting embryos at 13.5 days postcoitus (dpc). This 129 × CS and the reciprocal (CS × 129) mouse cross allowed for allele-specific analysis of H19 and Igf2 transcription, DNA methylation, and chromatin in normal cells. Female ICR-CTCF site mutant (CTCFm) (39) homozygous mice were crossed with wild-type CS males. Primary MEFs derived from the resulting CTCFm × CS embryos were used for allele-specific analysis of H19 and Igf2 transcription, DNA methylation, and chromatin composition upon maternal transmission of the ICR-CTCF site mutations. MEFs were used at early passage numbers (up to four passages).

Gene expression.

Northern blot assays were performed as described elsewhere (35). Ten micrograms of total RNA was loaded per lane. The blot was hybridized with a probe from exon 6 to visualize all possible transcripts and revealed the P2 (4.7 kb) and P3/P1 (3.9/3.7 kb) promoter bands. The membrane was rehybridized with the P1 probe (32) (provided by Peter Rotwein) to resolve the P3 and P1 promoter usage. The analysis of allele-specific transcription of the H19 and Igf2 genes was performed with reverse transcription-PCR (RT-PCR) SNuPE assays as previously described (37).

Chromatin preparation.

MEFs were grown in 25 ml of culture medium (Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 10⁻⁴ M β-mercaptoethanol, nonessential amino acids, l-glutamine, and antibiotics at standard concentrations). Confluent 15-cm plates provided about 15 to 35 million cells. Formaldehyde (1% final concentration) was added directly into the medium, and the dishes were incubated at room temperature for exactly 2 min (for native ChIP [N-ChIP] assay) or 10 min (for cross-linking and ChIP [X-ChIP]) on an orbiting platform. Cross-linking was stopped with a 0.125 M final concentration of glycine solution. Cells were washed twice with 20 ml of ice-cold phosphate-buffered saline and scraped off the plate with a rubber policeman; cells were then pelleted by centrifugation for 3 min at 250 × g and resuspended in 750 μl of ChIP lysis buffer (1% sodium dodecyl sulfate, 10 mM EDTA, and 50 mM Tris-HCl, pH 8.1) containing protease inhibitors (Complete protease inhibitor cocktail tablets, catalog number 1697498; Roche Applied Science). Sonication was carried out on ice in 1.7-ml Eppendorff tubes using a Branson Sonifer with a micro tip three or four times (N-ChIP or X-ChIP, respectively) for 10 s at 40% output. The sheared chromatin was centrifuged at 13,000 rpm for 15 min at 4°C. The supernatant was transferred to a fresh tube. An aliquot of the chromatin preparation was reverse cross-linked, proteinase K digested, phenol-chloroform extracted, precipitated, and resuspended in Tris-EDTA buffer. DNA concentration was determined using a UV spectrophotometer. One microgram of DNA was resolved on a 1.5% agarose gel to check sonication efficiency. Fragment sizes ranged from 0.2 to 1.5 kb, with a peak at 0.8 kb. Chromatin preparations were diluted to 0.4 μg/μl with ChIP lysis buffer containing protease inhibitors and were snap-frozen in liquid nitrogen in small aliquots and stored at −80 °C.

Immunoprecipitation.

We followed the protocol provided by Upstate Biotechnology. The antibodies were ChIP grade and purchased from Upstate Biotechnology (Charlottesville, VA): anti-CTCF, 06-917; anti-acetyl-histone H3 (Lys9), 07-352; anti-dimethyl-histone H3 (Lys4), 07-441; anti-trimethyl-histone H3 (Lys4), 07-473; anti-dimethyl-histone H3 (Lys9), 07-441; anti-trimethyl-histone H3 (Lys9), 07-442; anti-dimethyl-histone H3 (Lys27), 07-452; anti-trimethyl-histone H3 (Lys27) 07-449; and anti-histone macroH2A1, 07-219. Anti-normal rabbit immunoglobulin G (sc-2027; Santa Cruz Biotechnology) was used as a negative control. We used 10 μl (4 μg) of chromatin and 4 μg of specific antibody for each immunoprecipitation. The histone covalent modifications were detected using N-ChIP conditions. X-ChIP chromatin was used for detecting the macroH2A1 histone variant. The immunoprecipitated chromatin was eluted from 40 μl of protein A or AG agarose beads and reverse cross-linked, and DNA was isolated with a QIAquick kit, (Qiagen, Germany) and eluted in 100 μl of QIA Elution Buffer.

ChIP-SNuPE assay.

To quantitatively measure allele-specific chromatin modifications, we developed the ChIP-SNuPE assay (Fig. 3). The SNuPE assay takes advantage of single-nucleotide polymorphisms (SNPs) between parental alleles (37). We found these SNPs by DNA sequencing of inbred 129S1 (129) and CAST/Ei (CS) mouse strains at the specific regions of interest. In the ChIP-SNuPE assay, ChIP was performed first with a specific antibody. The maternal and paternal alleles were immunoprecipitated indiscriminately from the chromatin. Ten microliters of DNA from the ChIP elution was amplified with region-specific primers spanning the polymorphic sites. PCR fragments were isolated from agarose gels with a Qiagen gel purification kit. Equal amounts of fragments were subjected to SNuPE reactions (37), which consisted of one cycle of primer extension at the thermal denaturation midpoint temperature (T_m) of the SNuPE primer, abutting the SNP, with Taq polymerase and one of the two radionucleotides corresponding to the 129 or CS alleles. After denaturation at 98 °C for 3 min, the extended primers were run on denaturing polyacrylamide gels and quantitated with a PhosphorImager. The test samples were run side-by-side with quantitation controls (0% + 100%, 50% + 50%, and 100% + 0% of 129+ CS samples, respectively). The ratio of incorporation of allele-specific radionucleotides was measured and expressed as percent allele in the total precipitation [allele A count/(allele A + allele B count)] after adjustments to background and bias. First, the counts in the 0% 129 control sample were subtracted from all 129-type samples, and the counts in the 0% CS sample were subtracted from all CS-type samples to adjust for background incorporation. Second, the CS-specific counts were multiplied by counts obtained for the 50% 129 sample divided by the values obtained for the 50% CS counts. This adjustment eliminated bias arising from differences in incorporation efficiencies of the two radionucleotides and from differences in the number of radionucleotides that can incorporate (see below). We also adjusted the intensity of the entire set of one allele using Adobe Photoshop. This adjustment serves the same role visually as the adjustment step described above serves mathematically. This method of adjustment has been used previously for RT-PCR SNuPE (37). Amplification primers and SNuPE primers were the following (radionucleotides incorporating into 129 versus CS alleles are given in respective order [129/CS] following T_ms): H19-8kbU, GTGAGGTGGTAGCCTTCAAGAGGTCAC; H19-8kbL, GGGGACTGAGCAAACATACACAGACCC; H19-8kb SNUPE U, TTCTTCACCTTCCTGGGT (T_m of 46.2 °C; TT/C); H19-4kb II3, CGAGCATCCAGGAGGCATAAGAA; H19-4kb JJ3, CCACGAGGTACCAGCCTAGAAAATG; H19-4kb IIJJ SNUPE U, CTAAAGAGCCCCCCCACCCC (T_m of 60.3 °C; T/C); H19-2.7kb XY SNuPE PCR U, AGCAATGTCCGAAGCCGCTATG; H19-2.7kb XY SNuPE PCR L, TCGGTCTTACCAGCCACTGACGA; H19-2.7kb XY SNUPE L, ACTGGCTGGTTTTGGGGTTCAG (T_m of 58.9 °C; T/A); H19 promoter M3, TTTGGAGAATTTCAGGACGGGTGCG; H19 promoter N3, ACCCCACGACTCTCCTCCAGCTCTC; H19 PROMOTER SNUPE L, TCTTCCCCAGTTTCCCC (T_m of 49.1 °C; G/AA); H19+2kbU, GAATCCATCTTCATGGCCAACTCTGCCTGACCCGGGAG; H19+2kbL, TTGCCCTCAGACGGAGATGGACGA; H19+2kb SNUPE U, TGAATGTATACAGCGAGTGTG (T_m of 46.1 °C; T/C); Igf2DMR1ChIPU, TCAGGTGAAGGCTCTGTGGGCA; Igf2DMR1ChIPL, GATTAGGCTGCAAGCCCTCTGCTAA; Igf2DMR1SNUPE U, CCCTGGTGGCTCTTCA (T_m of 46.1 °C; A/G); Igf2DMR2ChIPU, CATGCTTGCCAAAGAGCTCAAAGAG; Igf2DMR2ChIPL, GGGGGGTGTCAATTGGGTTGTT; Igf2DMR2SNUPE U, CAAGGGGATCTCAGCA (T_m of 43.0 °C; G/A); Igf2 PR ChIPU, CCCCAAAGGCTGCTAGGAGATCCCAGGCAA; Igf2 PR ChIPL, GCCTCAGTGGTAGTGGCAGGACCTGTGCTCAGTTA; Igf2 PR SNUPE(L), GGGTAGAGGGTTCTCACAGGGACCTACTTGC (T_m of 67.0 °C; T/C).

FIG. 3.

Novel ChIP-SNuPE assays for the analysis of allele-specific chromatin modifications. (A) Outline of the SNuPE assay. The SNuPE assay takes advantage of SNPs between parental alleles (37). We found these SNPs by DNA sequencing of inbred 129 and CS mouse strains at the specific regions of interest. In the ChIP-SNuPE assay, ChIP is performed first with a specific antibody. The maternal and paternal alleles are immunoprecipitated indiscriminately from the chromatin. DNA from the precipitated chromatin is PCR amplified with region-specific primers spanning the polymorphic sites. PCR fragments are isolated and probed with a primer abutting the SNP, and the ratio of incorporation of allele-specific radionucleotides is measured and expressed as the percent allele of the total precipitation. (B) Mixing experiments at the regions of interest. Aliquots of 129- and CS-type DNA were combined in the relative ratios of the percent CS to the total (CS + 129), indicated on the top, and subjected to SNuPE with the region-specific oligonucleotides. The measured values are shown above each of the images. The samples 0, 50, and 100% were included in the experiments for quantitating alleles in samples subjected to ChIP. The locations of the region-specific assays are indicated in the legend of Fig. 1A. (C) Plot of the measured values in the mixing experiment. The assays were rigorously quantitative, using a small amount (10 ng) of total DNA. (D) Example of ChIP-SNuPE quantitation of histone composition. The ICR (−4 kb) region was analyzed. Maternal (M) or paternal (P) allele specificity was almost identical between the reciprocal mouse crosses. The numbers above the gel images are the percentages of the paternal allele's contribution to the total immunoprecipitate. CS and 129 alleles are shown on top and bottom, respectively.

Real-time PCR.

Equal aliquots (4 μl) of ChIP elution DNA were used for real-time PCR quantitation with the Bio-Rad iCycler real-time machine. Sybr Green reactions were run with RT² real-time Sybr green/fluorescein PCR master mix (Superarray Bioscience Corporation), and TaqMan reactions were run with iTaq Gold (Bio-Rad). A dilution series of known amounts of genomic DNA was used for quantitating copy numbers after a minor adjustment to relative differences between input DNA samples from the two cell types. Multiplex real-time assays were developed to quantitate several different regions in one reaction. Two sets of primers that were used in multiplex reactions are labeled with an asterisk or number sign.

Primer sequences for real-time PCR were the following: Igf2 PR U sybr green, CTGTGAGAACCCTCTACCC; Igf2 PR L sybr green, AGGACCTGTGCTCAGTTAG; Igf2DMR1 U sybr green, GCGTTTTCCTACCTGGCAAAG; Igf2DMR1 L sybr green, CTCTGCTTTCTGCCCTCCAG; Igf2 DMR2 TqmU, GTCATCGTCCCCTGATCGTG; Igf2 DMR2 TqmL, TGATGGTTGCTGGACATCTCC; Igf2 DMR2 TQPR Cy5^#, CCACCCAAAGACCCCGCCCACG; H19 −8kb TqmU, GCCTGATCTGCCAGCTTCTC; H19 −8kb TqmL, AGGAAGGTGAAGAAGGTTCTTAGG; H19 −8kb TQPR TAMRA^#, TGCACCGGGGCCACTTCTCTTGTT; H19 −4kb TqmU, CTGTGCAGCAACTGATGACC; H19 −4kb TqmL, GAACTGTAGGCAATGGCTATTTTC; H19 −4kb TQPR FAM*, ACTCAGGCTCCAGGCAGACTCAGT; H19 −3kb TqmU, TGCCCATGACAATGTCCAAGG; H19 −3kb TqmL, TCGACCACTGAGGCATAGCG; H19 −3kb TQPR TET*, TCGGGTTCGCCCACAGCAATGTCC; H19 prom TqmU, GGAGAGCTGGAGGAGAGTCG; H19 prom TqmL, CTAGCCCCTCAGTCCTTCAAC; H19 prom TQPR ROX*, CCTGCCAGACTCCAGATGCCGAGG; H19 +2kb TqmU, TACCCACCTGTCGTCCATCTC; H19 +2kb TqmL, CAGACTAGGCGAGGGGAAGG; H19 +2kb TQPR Cy5*, CCTCAAGCACACGGCCACACCCAG.

Bisulfite genomic sequencing.

Bisulfite methylation analysis at the ICR was performed as previously described (39). The DMR1, DMR2, and H19 promoter regions were sequenced as follows. For the H19 promoter, the primers were the following: H19PRbsU1, TGTTTTTGTTTTGTTTGAGTTAGTTTT (T_m of 52.0°C); H19PRbsL1, ACAAACTAAATAAAAAACAACTTCAATATA (T_m of 50.4°C); H19PRbsU2, AGTTTTTTAGTTTTTTAATATTTTTGTTAGATTTT (T_m of 54°C); and H19PRbsL2, AAAAAAATATCTAAAAACAATACCAAAC (T_m of 50.0°C). For Igf2 DMR1, the primers were the following: DMR1 bsU1, GGTTAGGTGAAGGTTTTGTGGGTAGTTATA (T_m of 58.5°C); DMR1 bsL1, ATATTCCCCTTTCAAATTCCAATCTACATCC (T_m of 59.3°C); DMR1 bsU2, TTGTGGGTAGTTATATAGAGGAAGA (T_m of 48.2°C); DMR1 bsL2, CCAACCTCTATCCCTAACTTTT (T_m of48.3°C). For Igf2 DMR2 the primers were the following: Igf2DMR2 bsU1, TATAGATATTTTAGGGAAGTTGTTTT (T_m of 47.3°C); Igf2DMR2 bsL1, CAATTAAATTATTTAAAACCAATC (T_m of 45.0°C); Igf2DMR2 bsU2, GGTTAATATGATATTTTGAAATTTG (T_m of 46.8°C); Igf2DMR2 bsL2, TAAAACCAATCAAATTTAATTTT (T_m of 45.4°C).

RESULTS

Characterization of primary MEFs carrying CTCF site mutations in the maternal ICR.

We had previously generated a mouse line (CTCFm) with engineered point mutations in four in vivo CTCF binding sites (36). Maternal inheritance of these mutations eliminated insulation function in CTCFm × CS fetuses, as evidenced from biallelic Igf2 transcription (39). CpG methylation levels became highly elevated at the ICR, the H19 promoter, and H19 gene body in CTCFm × CS fetuses (39), revealing the role of CTCF binding in the maintenance of DNA hypomethylation at the ICR and H19 sequences. Using bisulfite genomic sequencing in CTCFm × CS fetuses, we found that ICR-CTCF binding is also required for the maintenance of DNA hypomethylation at the Igf2 DMR1 and DMR2 sequences at ∼90-kb distance (Fig. 2C and D), in agreement with data from another laboratory (21). ChIP assays, in combination with real-time PCR quantitation, proved that in vivo CTCF binding in the ICR had been completely abolished in CTCFm × CS primary embryo fibroblasts (MEFs) compared to wild-type 129 × CS MEFs at the H19 ICR, promoter, and gene (Fig. 1C). CTCFm × CS MEFs exhibited important features also observed in CTCFm × CS fetuses: Igf2 transcript levels were almost twofold (Fig. 1D) and biallelic (Fig. 1E), H19 expression was greatly reduced (Fig. 1D), the maternal ICR was hypermethylated (Fig. 1F), and the DMR1 and DMR2 maternal alleles were hypermethylated compared to the normal maternal (and even compared to normal paternal) alleles (Fig. 2). These cells are ideal for testing whether CTCF binding has a role in organizing allele-specific chromatin composition at the H19/Igf2 imprinted region.

Sensitive and quantitative ChIP-SNuPE assays for measuring allele-specific chromatin differences.

Both wild-type 129 × CS MEFs and mutant CTCFm × CS MEFs harbor SNPs along the H19/Igf2 imprinted domain. These allow for allele-specific assessment of the histone composition (Fig. 3A). To quantitate allele-specific enrichment at specific subregions along the imprinted domain (Fig. 1A), we developed ChIP-SNuPE assays (Fig. 3B and C) for the Igf2 DMR1, Igf2 promoter, Igf2 DMR2, two halves of the ICR (−3 kb and −4 kb), the H19 promoter, the H19 gene body (+2 kb), and also for an intermediary region −8 kb from the H19 transcriptional start site. These assays were rigorously quantitative, as shown by DNA mixing experiments (Fig. 3B and C). Parental allele specificity of the histone modifications at the regulatory regions was nearly identical in the reciprocal CS × 129 MEFs (Fig. 3D and Fig. 4).

FIG. 4.

Allele-specific chromatin composition along the H19/Igf2 imprinted domain. Allele-specific chromatin composition was quantitated by ChIP-SNuPE assays (Fig. 3) at the specific regions indicated in Fig. 1A. ChIP was done using antibodies recognizing specific histone modifications (indicated above graphs) to precipitate chromatin from 129 × CS or reciprocal CS × 129 MEFs (indicated at the right of each row of charts). The ratio of the allele-specific histone modification at a specific region was expressed as a percentage of maternal (black bars) or paternal (gray bars) alleles in the total (maternal + paternal, or 100%) of the immunoprecipitation. Missing data points mean lack of enrichment for a specific chromatin modification at the specific region when marked as not detected (nd). (A) CTCF binding is maternal allele specific in the ICR. (B) Active chromatin modifications localize to the maternal allele at the H19 regions and to the paternal allele at the Igf2 regions. (C) Complex pattern of repressive chromatin marks along the imprinted domain (see text for details). Almost no allele-specific chromatin differences exist at a “neutral” intermediary region −8 kb upstream of the H19 promoter. Reciprocal mouse crosses exhibit nearly identical allele-specific chromatin composition. pat, paternal; mat, maternal.

Allele-specific chromatin composition at the ICR and at the H19 gene.

CTCF binding was specific to the maternal allele of the ICR at the −4 kb and −3 kb regions (Fig. 3D and 4A). At four locations studied at the H19 locus (two halves of the ICR, the H19 promoter, and H19 gene body), landmarks of active chromatin such as acetylated H3K9 (H3K9ac), dimethylated H3K4 (H3K4m2), and trimethylated H3K4 (H3K4m3) were associated with the expressed normal maternal allele in 129 × CS and CS × 129 MEFs (Fig. 3D and 4B), as measured with ChIP-SNuPE assays. Repressive histone signals, H3K9m2, H3K27m2, H3K27m3, and macroH2A1, were strongly biased toward the paternal allele at the H19 promoter and less biased at the gene body (Fig. 4C). H3K9m3 and macroH2A1 were exclusively found and strongly biased to the paternal allele of the ICR sequences, respectively. A genome-wide approach revealed an allelic bias in MEFs for H3K4m3 and H3K9m3 binding in several known ICRs, including the H19/Igf2 ICR (26). Our data are consistent with these findings, with H3K4m3 being strongly biased toward the maternal allele while H3K9m3 was exclusively found in the paternal allele in the ICR. H3K27m3 was present in both parental alleles at the ICR (Fig. 4C), in agreement with data from another laboratory (7).

Effects of ICR CTCF site mutations on the allele-specific chromatin composition at the ICR and at the H19 gene.

The ICR CTCF site point mutations caused a considerable reduction in the maternal allele-specific chromatin signals in CTCFm × CS MEFs compared to wild-type 129 × CS MEFs at the, H19 ICR, promoter, and gene (Fig. 5), as determined by real-time PCR assays. There was a concomitant increase in the level of repressive histone modifications at the H19 sequences. Most notably, both H3K27m3 and macroH2A1 levels increased considerably at the ICR sequences and at the H19 promoter and gene body. ChIP-SNuPE analyses of wild-type and mutant MEFs revealed that paternal allele-specific (Fig. 4) repressive chromatin modifications, macroH2A1, H3K9m2, and H3K27m3, became biallelic or predominantly maternal allele-specific in the mutant cells at the H19 promoter and gene body (Fig. 6A). This was the case for macroH2A1 at the ICR regions. Bialellic H3K27m3 (Fig. 4) became maternally biased in the ICR in the mutant cells (Fig. 6A). Taken together, at the H19 locus the ICR CTCF site mutations have caused the maternal allele's chromatin composition to become very similar to that of the normal paternal allele (Fig. 7). The chromatin changes corresponded to the downregulation of the H19 promoter in CTCF site mutant fetuses (Fig. 1D). While the H19 promoter is methylated in fetal organs (39), it was not yet methylated in the CTCFm × CS MEFs (Fig. 1G) that represent an earlier, embryonic development stage. CTCF binding in the ICR, therefore, is directly responsible for organizing an activating chromatin environment at the H19 promoter, which is required for H19 transcription.

FIG. 5.

CTCF is responsible for region-specific enrichment of chromatin components at the H19 and Igf2 loci. The overall enrichment for specific chromatin components was compared between normal 129 × CS MEFs (white bars) and CTCFm × CS MEFs (black bars) by real-time PCR. The location of the regions is shown in Fig. 1A. At the H19 locus, the activating chromatin marks showed reduced precipitation in CTCFm × CS MEFs compared to normal cells, but the precipitation levels of repressing marks, H3K27m3 and macroH2A1, increased. At the Igf2 regions, H3K27m3 levels were greatly decreased in the mutant cells, while activating chromatin marks were more abundant at the DMR1 and at the Igf2 P2 promoter. There was no change at the −8-kb region. Average values are shown with standard deviations.

FIG. 6.

CTCF is required for allele-specific chromatin composition locally and at a distance. ChIP-SNuPE analyses of allele-specific chromatin composition reveal the consequences of ICR CTCF site mutations. (A) Repressive chromatin marks shift toward the maternal allele at the H19 locus. (B) Activating chromatin marks shift toward the maternal allele at the distant Igf2 locus. Chromatin was precipitated in duplicates from 129 × CS and CTCFm × CS MEFs with the specific antibodies indicated. SNuPE assays were performed with region-specific oligonucleotides. Allele-specific histone modification at a specific region was expressed as a percentage of maternal or paternal (dark or light color, respectively) alleles in the total immunoprecipitate. The average values are shown with standard deviations. pat, paternal; mat, maternal.

FIG. 7.

Summary of the results. (A) Specific regions analyzed in this study. (B) Allele-specific chromatin composition at the H19 and Igf2 loci in normal MEFs. M, maternal chromosome; P, paternal chromosome. Allele-specific activating (green) and repressing (red) signals are shown in the respective chromosomes. Chromatin components with low relative abundance are in parentheses. Note that repressive chromatin at the Igf2 locus is present at the hypomethylated allele. (C) Paternalization of chromatin composition along the H19/Igf2 imprinted region. The chromatin composition of the maternally inherited CTCFm chromosome became similar to that of the normal paternal chromosome. Vertical arrows in panels B and C indicate changes (decreases and increases, respectively) in enrichment for specific histone modifications at specific regions in the CTCFm × CS compared to 129 × CS MEFs.

Allele-specific chromatin composition at the Igf2 gene and at the Igf2 DMRs.

At the Igf2 locus, we assessed the allele-specific chromatin composition using ChIP-SNuPE assays at the DMR1, DMR2, and the P2 promoter. The P2 promoter is the general promoter for Igf2 in the fetus (11); and in embryonic stem cells (35), it is expressed together with the P3 promoter in MEFs (Fig. 1D) and is hypomethylated in both alleles (12). Landmarks of active chromatin, such as histone tail covalent modifications, H3K9ac, H3K4m2, and H3K4m3, were biased toward the transcriptionally active paternal allele of the Igf2 P2 promoter (Fig. 4B). H3K4 methylation and H3K9 acetylation marks were also associated with the paternal, relatively hypermethylated (Fig. 2), alleles of DMR1 and DMR2 sequences (Fig. 4B). Landmarks of repressive chromatin showed a more complex pattern of distribution. H3K9m2, H3K27m2, and H3K27m3 were associated with the silent maternal allele at all three Igf2 regions (Fig. 4C). The heterochromatin-specific histone variant macroH2A1 is allele-specifically enriched in the methylated alleles of the DMRs at the Xist, Peg3, Gtl2/Dlk1, and Gnas imprinted domains and at the H19/Igf2 ICR (5). Here we show that macroH2A1 exhibits maternal allele-specific enrichment at the Igf2 DMR1 and DMR2 (Fig. 4B). This is the first example of the presence of macroH2A1 in a relatively CpG-hypomethylated allele. In contrast, H3K9m3 was greatly biased toward the paternal, DNA-hypermethylated allele of DMR2 (Fig. 4C). The enrichment of H3K9m3 in the CpG-methylated allele was, in fact, the common theme between the ICR and the DMR2. This histone modification might play a role in the maintenance of DNA hypermethylation in the paternal alleles of these regions.

Effects of ICR-CTCF site mutations on the allele-specific chromatin composition at the Igf2 gene and at the Igf2 DMRs.

We found greatly decreased levels of H3K27m3 at the Igf2 P2, DMR1, and DMR2 sequences in the CTCFm × CS MEFs using ChIP and real-time PCR assays (Fig. 5). H3K27m3 was specific to the maternal allele in normal cells (Fig. 4C). CTCF binding in the ICR was, therefore, required for the presence of the repressing chromatin component H3K27m3 in the maternal alleles of the Igf2 P2, DMR1, and DMR2 sequences. Levels of active chromatin signals, H3K9ac and H3K4m2, increased at the DMR1 and at the Igf2 P2 promoter (Fig. 5), and these normally paternal allele-specific marks (Fig. 4B) became biallelic in the CTCFm × CS MEFs (Fig. 6B). We found similar shifts in allele specificity for H3K9ac and H3K4m2 at the DMR2 region in the CTCFm × CS MEFs (Fig. 6B), although without apparent increases in the overall enrichment levels, which we cannot fully explain. Taken together, the chromatin composition at both the DMR1 and DMR2 regions was rearranged by the CTCF site mutations in such a way that the maternal, mutant allele became very similar to the normal paternal allele (Fig. 7). CTCF binding in the H19 ICR, therefore, organizes allele-specific chromatin composition at the Igf2 locus from a great distance. CTCF is responsible for defining the maternal allele's identity at the level of chromatin along the H19/Igf2 imprinted domain.

DISCUSSION

This study provides quantitative information on allele-specific and total levels of histone modifications along the H19/Igf2 imprinted domain in mouse cells that lack ICR-CTCF insulator protein interactions. Our data are consistent with CTCF's role as the master organizer of the maternal allele's chromatin (Fig. 7): ICR-CTCF binding is required for the maternal allele-specific active marks at the H19 locus and for the maternal allele-specific repressing mark H3K27m3 at the Igf2 DMRs at a ∼90-kb distance. ICR-CTCF binding is also essential for protecting the maternal allele at the H19 locus from the paternal allele-specific repressing chromatin signals and protecting the maternal allele at the Igf2 DMRs from paternal allele-specific activating chromatin components.

Many of the changes observed in the mutant cells corresponded to the changes in activity states of the relevant transcription units. Active chromatin was observed in transcribed regions, and repressive chromatin was observed in the silent regions in the normal as well as in the mutant cells. The DMR1 and the ICR, however, are not transcribed regions, yet their chromatin states were completely reorganized in the mutant cells. Can this newly revealed chromatin-organizing role of CTCF be related to its other known distinct roles (summarized in the introduction) in the H19/Igf2 imprinted domain? A discussion of these roles in relation to chromatin organization follows.

1. Insulation of enhancer-Igf2 promoter interaction in normal cells might be thought at first to be the most likely explanation for imposing Polycomb-repressing chromatin signals at the maternal allele of the Igf2 promoters. In the simplest scenario, the enhancers would not interact with the Igf2 promoters because of insulation; the inactivity of the promoters would initiate a repressive feedback of H3K27m3 deposition at the inactive Igf2 promoters, and H3K27m3 would spread over to the Igf2 DMRs. Alternatively, the distant ICR-CTCF binding might be responsible for H3K27m3 deposition/maintenance in the DMR1, and this repressing signal in turn might contribute to the silencing of the P2 promoter's activity. This explanation seems more likely in the case of MEFs in the light of the DMR1 knockout experiment: in the absence of DMR1, the ICR is not sufficient to inactivate the Igf2 promoters in cells of mesoderm origin (6) such as MEFs. While it may not be possible to fully discriminate between cause and effect of gene activity and chromatin composition at the Igf2 promoter, the chromatin changes observed in CTCFm × CS MEFs point to the long-range chromatin-organizing role of CTCF insulation. On the other hand, the insulation function of CTCF cannot be responsible for directing the chromatin state of the H19 promoter, because the topology requirement for insulation is not met: CTCF binding sites are not located between the H19 promoter and the enhancers.

2. CTCF-maintained DNA hypomethylation in the ICR sequences may attract or maintain active chromatin marks and may protect from repressive chromatin marks in the maternal ICR allele in wild-type cells, such as 129 × CS MEFs. In CTCFm × CS MEFs, mCpG binding proteins might recruit the NuRD complex to the hypermethylated ICR, inducing H3K9 deacetylation (Fig. 5). However, the H19 promoter's chromatin is reorganized in CTCFm × CS MEFs (Fig. 5 and 6) in the absence of DNA methylation (Fig. 1G), suggesting a more direct activating role for CTCF, such as organizing an activating chromatin environment at the H19 promoter. We should note that CpG methylation is not always incompatible with H3K9ac and H3K4m signals, as we observed in the paternal allele of the Igf2 DMR1 and DMR2 sequences in normal cells. Indeed, increased DNA methylation at the Igf2 DMR1 sequences in the maternal allele coincided with elevated H3K9ac and H3K4m2 levels. Could DNA CTCF-maintained hypomethylation at the Igf2 DMRs recruit or maintain repressing chromatin signals in the normal silent maternal allele? This might be possible if a silencer molecule preferably bound to hypomethylated DMR1 sequences and attracted the repressive chromatin components H3K27m3 and macroH2A1 to the region. The transcriptional repressor GCF2 is partially responsible for DMR1-mediated silencing activity and might play a chromatin-organizing role at the DMR1 sequences (9). The fact that macroH2A1 is associated with the maternal hypoacetylated allele is consistent with its ability to interfere with SWI/SNF nucleosome remodeling and histone acetylation (8). Based on their maternal allele-specific localization pattern at the Igf2 promoter and DMRs, the functions of H3K27m3 and macroH2A1 appear to be the maintenance of Igf2 gene repression independently of DNA methylation-mediated gene repression. It is interesting that the allele-specific CpG methylation differences at the Igf2 DMRs (Fig. 2) are not at the extremes, as in the ICR sequences. Yet the chromatin differences are remarkable between the DMR1 and DMR2 alleles (Fig. 4). This suggests that chromatin composition is a more robust parental-allelic discriminator signal at the DMRs than DNA methylation and that some (as yet unidentified) allele-specific chromatin differences may have a role in directing somatic DNA methylation patterns at the DMRs.

3. Initiating H19 promoter activity in the embryo by ICR-CTCF binding might require the recruitment of landmarks of active chromatin by CTCF to ICR sequences in early embryo development, and this active chromatin might initiate H19 promoter activity via chromatin spreading or by direct interaction. Lack of transcription initiation of H19 in embryos with microdeletions at four CTCF site binding sites (10) could be caused by the lack of active chromatin and/or the presence of repressive chromatin assembly at the ICR and at the H19 promoter, similarly to the chromatin reorganization seen in CTCFm × CS MEFs (Fig. 4). Alternatively, CTCF binding could be required for protecting the H19 promoter from de novo methylation in the 6.5 dpc embryo, as it is required for protecting the ICR and the promoter in the fetus (39) and the maternal ICR allele during the wave of de novo methylation in embryo development (10). However, we found that promoter CpG methylation in the maternal allele, although abundant in CTCFm × CS fetal livers and kidneys at 17.5 dpc (39), is not yet present in the CTCFm × CS 6.5 dpc embryo (Fig. 8) or in CTCFm × CS MEFs (Fig. 1G). Therefore, CTCF binding in the ICR has a more direct role in H19 promoter initiation, most likely via recruiting landmarks of active chromatin to the H19 promoter.

4. Because ICR-CTCF binding is required for chromosome loop formation between the ICR and the DMR1, we expected to find CTCF binding in the DMR1 sequences. We have not been able to detect this interaction using N-ChIP (Fig. 4A) or X-ChIP (data not shown). This suggests that CTCF may not directly bind DMR1, but another factor might be required for loop formation between the Igf2 locus and the CTCF-bound ICR. This interaction could mediate the transmission of a Polycomb-repressive signal from the ICR to the Igf2 DMR1. H3K27m3 is, indeed, a chromatin component shared between the maternal alleles of Igf2 DMR1 and the ICR.

FIG. 8.

The H19 promoter is unmethylated in the early embryo. The results of the bisulfite sequencing experiment are shown. The embryo proper was analyzed at 6.5 dpc. Other details are as described in the legend of Fig. 1.

Removal of the imprinting control region 2 of the Kcnq1 domain abolishes allele-specific histone modifications in the placenta (24). In the present study, introducing less drastic changes, specific point mutations, in the ICR has resulted in a complete reorganization of chromatin at the Igf2/H19 imprinted domain (Fig. 7). This is the first study to reveal the organizing effect of a single protein factor in allele-specific chromatin composition. In imprinting control region 2, Polycomb-mediated allele-specific regulation constituted a repressing mechanism independently of DNA methylation in the placenta, but in somatic tissues, DNA methylation was dominant (24). Our data suggest that imprinted Igf2 gene expression might depend on a similar H3K27m3-mediated DNA methylation-independent repression mechanism, but in this case in somatic cells: the Polycomb-repressive signal H3K27m3 is enriched in the relatively hypomethylated, silent maternal allele of the Igf2 DMRs in MEFs but not in the hypermethylated, active maternal allele in CTCFm × CS MEFs. It will be of interest to determine how CTCF achieves this long-range effect.

Chromatin composition, different covalent modifications, and the presence of histone variants (15, 20, 25) play very important roles in gene regulation. CTCF is directly involved in local chromatin composition at the beta-globin locus: enhancer-promoter loop formation by CTCF is required for local H3 hyperacetylation and H3K9 dimethylation (34). Our data provide insight into insulator action at a distance at the level of chromatin organization, revealing that CTCF plays a crucial role in the allele-specific chromatin organization at the H19 locus and also at the remote Igf2 locus. We suggest that the long-range chromatin-organizing activity of CTCF insulator binding may not be restricted to the H19/Igf2 imprinted domain but may also operate at other imprinted or nonimprinted chromosome domains genome-wide. It might be an important mechanistic component of CTCF's insulator activity.