CTCF Elements Direct Allele-Specific Undermethylation at the Imprinted H19 Locus

Current Biology, Vol. 14, 1007–1012, June 8, 2004, 2004 Elsevier Ltd. All rights reserved.

DOI 10.1016/j .c ub . 20 04 . 05 .0 4 1

CTCF Elements Direct Allele-Specific Undermethylation at the Imprinted H19 Locus Eyal Rand, Ittai Ben-Porath, Ilana Keshet, and Howard Cedar* Department of Cellular Biochemistry Hebrew University Medical School Ein Kerem Jerusalem 91120 Israel

Summary The H19 imprinted gene locus is regulated by an upstream 2 kb imprinting control region (ICR) that influences allele-specific expression, DNA methylation [1], and replication timing [2]. This ICR becomes de novo methylated during late spermatogenesis in the male [3] but emerges from oogenesis in an unmethylated form, and this allele-specific pattern is then maintained throughout early development [4] and in all tissues of the mouse [5]. We have used a genetic approach involving transfection into embryonic stem (ES) cells in order to decipher how the maternal allele is protected from de novo methylation at the time of implantation [6]. Our studies show that CCCTC binding factor (CTCF) boundary elements within the ICR [7, 8] have the ability to prevent de novo methylation on the maternal allele. Since CTCF does not recognize its binding sequence when methylated [9, 10], this reaction does not occur on the paternal allele, thus preserving the gamete-derived, allele-specific pattern. These results suggest that CTCF may play a general role in the maintenance of differential methylation patterns in vivo. Results Previous experiments showed that most DNA sequences become de novo methylated when inserted into ES cells, while CpG islands and other specific promoter sequences [11] remain unmodified, mimicking what happens at the time of implantation during normal development [12]. This process is apparently mediated by factors that recognize cis-acting elements within CpG islands [13]. Although the mechanism for this protection is not known, it has been shown that premethylated CpG islands actually become undermethylated when transfected into ES cells, suggesting that these elements work either by directing active demethylation [12] or by selectively interfering with local maintenance methylation. The 3.8 kb ICR located upstream to the H19 gene has a GC content of ⬍3% and is therefore not considered to be a CpG island. In order to evaluate how this fragment is recognized in ES cells, we cloned it into the plasmid vector p342 [14], enzymatically methylated it in vitro, stably transfected it into ES cells, and then determined *Correspondence: [email protected]

its methylation state in pooled clones by using Southern blot hybridization analysis (Figure 1A). Strikingly, sites located in the 5⬘ region of this fragment failed to maintain methylation even though CpG residues further downstream on the same construct remained highly methylated (ⵑ50%) in the ES cell environment. These experiments suggest that the ICR contains cis-acting sequences that can direct region-specific undermethylation in embryonic cells. In order to map the cis-acting sequences responsible for this preferential undermethylation, we divided the H19 upstream region into a series of sequential fragments (C–F), cloned them into the p342 vector, methylated them with the mHpaII and mHhaI DNA methylases, and transfected them into ES cells (Figure 1B). This vector is designed to contain methyl-sensitive restriction enzyme sites positioned near the insertion locus, which serve as indicators for the presence of cis-acting sequences capable of directing undermethylation. When the vector itself was introduced into ES cells, the SmaI test sequence (subset of HpaII) remained fairly methylated (ⵑ50%), while the initially unmodified SalI site actually underwent de novo methylation. In the presence of fragment E or F, however, these same sites became highly unmethylated (80%–90%). On the other hand, fragments C and D, which are located 1 kb downstream to the ICR, were unable to induce undermethylation of the test sequence (Figure 1C). In a parallel manner, only fragments C and D underwent de novo methylation when unmethylated templates were used in the transfection assay (Figure S1). As a continuation of this genetic approach we further divided up fragments E, F, and C (Figure 2A) and by transfection identified S1/S2 and S11 as segments capable of directing undermethylation in ES cells (Figure 2B). We then carried out a fine analysis of fragment S2. To this end, small 30 bp fragments (p21–p26) were cloned into the test plasmid, methylated with HpaII and HhaI methylases, and transfected into ES cells (Figure 3A). Although all of the fragments show some undermethylation at the SmaI site in the test vector, one fragment (p23) was found to be capable of directing highlevel undermethylation (70%) (Figure 3B). We next asked whether there are any known sequence motifs located within fragment p23. A computer search indicated that this short sequence indeed includes a recognition site for the trans-acting factor CTCF, which has been implicated as a boundary protein [15]. In order to prove that CTCF is indeed involved in undermethylation, we prepared a variant p23 fragment containing mutations in the CTCF motif that prevent binding in vitro (Figure S2), and demonstrated conclusively that this abrogates its ability to induce undermethylation in the test system (Figure 3B). In addition to the CTCF site located on the p23 fragment, the ICR actually harbors three other canonical CTCF motifs, including one located within fragment S11. In this case, as well, a small 30 bp fragment (P111) containing this CTCF element was also

Current Biology 1008

Figure 1. Demethylation of the H19 ICR in ES Cells (A) ES cells were transfected with plasmid p␤H19 (map) in vitro methylated at HpaII(H) and HhaI sites. The DNA was digested with DraI/BamHI with or without HpaII, electrophoresed, and subjected to Southern blot analysis with either probe 1 or 2 (see map). Note that sites within the ICR are relatively unmethylated (probe 1), while those at the 3⬘ end of the H19 fragment remain significantly methylated. This region also underwent de novo methylation at both HpaII and HhaI sites when the unmethylated plasmid was transfected into these same cells (data not shown). (B) Fragments C–F from the H19 ICR region (map positions are according to accession number AF049091) were cloned into the PstI site of test vector p342, methylated in vitro at HpaII (H) and HhaI (C) sites (isoschizomer of CfoI), and transfected into ES cells. (B) DNA from pools of clones was cut with DraI and analyzed by Southern blot (probes 1 and 2) for methylation at the SmaI and SalI sites in the test vector. The size of the DraI digest (arrow) is 880 bp for p342 alone, plus the insert length. The smaller product (left side) of SmaI or SalI digestion (ⵑ400 bp) is the same for each construct. Note that undermethylation was very strong (80%–90% based on SmaI digestion) in the presence of fragments E or F. A background level of demethylation (ⵑ50%) was observed using the p342 vector itself or plasmids carrying fragment C or D. Conversely, de novo methylation (ⵑ40%) at the SalI site (which is not modified by HpaII or HhaI methylases) was observed only for fragments C and D or p342 itself.

able to bring about undermethylation in our system (Figure 3B). These data indicate that the H19 region contains sequences that can induce undermethylation at the time of implantation. It was thus interesting to ask why only the maternal allele remains unmethylated in vivo. It should be noted that the paternal ICR is initially methylated during spermatogenesis and then faithfully carries its methylation mark throughout preimplantation development. Since the binding of CTCF is known to be inhibited by the presence of methyl groups at its recognition sequence [10], we tested whether it might be for this reason that CTCF cannot induce undermethylation of the paternal allele at the time of implantation. In all of the transfection experiments described above, vectors were modified exclusively at HpaII and HhaI sequences, enabling us to measure methylation states at fixed test sites. Under these conditions, the CpG residues within the CTCF motifs remain unmethylated. In order to explore the effect of methyl groups on the CTCF sequence itself, we in vitro methylated all of the CpG

sites in fragment E by using the Sss1 methylase [16]. As shown in Figure 3C, in the fully modified state this fragment, as well as its derivates, S1 and S2 were unable to undergo undermethylation in ES cells. As a control, we tested the effect of full methylation on the Aprt CpG island element (IE), and this was found capable of directing undermethylation. Unlike fragment E, Sss1-methylated fragment F did undergo some undermethylation after transfection but only at HpaII and HhaI sites within the fragment itself. This may be due to the involvement of cis-acting sequences other than CTCF [17]. In order to test whether the cis-acting sequences that we have identified in the H19 ICR also operate in vivo to protect against de novo methylation, we generated a series of transgenic mouse founders containing different fragments of the H19 ICR. In this system, vector DNA is injected directly into fertilized oocytes, thus bypassing the germline. We have previously demonstrated that a large 12 kb H19 construct remains unmethylated in cells from transgenic founders (our unpublished data), indicating that this locus is indeed protected from de novo

H19 Undermethylation by CTCF 1009

Figure 2. Analysis of Elements Involved in Undermethylation (A) Subfragments S1–S12 (averaging 150 bp) were derived from fragments E, F, and C, cloned into the test vector p342, methylated in vitro at HpaII and HhaI sites, and transfected into ES cells. The positions of potential CTCF elements within fragments E, C, and F are indicated (red oval). The expected size of the SmaI digestion product is also indicated on the map. (B) DNA from pools was digested with DraI together with SmaI, ClaI (Cl), or HhaI (C) and subjected to Southern blot analysis by using probe 1 (see Figure 2). In each case, the fully methylated DraI fragment is about 1.1 kb in length. The degree of undermethylation is based on the HhaI digest. Although a CTCF sequence motif is also present in S4, it does not show any appreciable undermethylation, perhaps because it contains an element with only weak binding activity [6].

methylation in the default state even though non-CpG island DNA usually becomes de novo methylated in this system [14]. In keeping with results from ES cells, transgenic founders carrying the S1 fragment containing the CTCF element remained relatively unmethylated at the test sites, while those carrying the S3 control fragment became almost completely modified (Figure 4). These studies clearly indicate that cis-acting elements in the ICR are partially sufficient for protecting the maternal H19 allele from de novo methylation at the time of implantation. Discussion The H19 ICR initially becomes modified on the paternal alelle at a late stage of spermatogenesis, while the maternal allele remains unmethylated [3]. This pattern is maintained in the preimplantation embryo despite the massive wave of demethyation that encompasses almost the entire genome [4]. The next critical stage for DNA methylation takes place at the time of implantation when the entire genome excluding CpG islands undergoes de novo methylation. Given these conditions, the H19 ICR must know how to be recognized and maintained in its unmethylated state even though it is not a classical CpG island. In this paper, we have begun to

define the cis-acting elements and trans-acting factors that control this process. Our genetic studies indicate that there are at least three different cis-acting sequences involved in protecting the maternal H19 allele from de novo methylation. One is the CTCF site identified in fragment E. The second element must be located in fragment F. Although we did not actually map this sequence, other researchers, using an experimental strategy similar to ours, have already shown that an Oct1 motif located at position 2200 (see map) can induce local undermethylation when inserted into ES cells [17]. A third element is present in fragment S11, and this also corresponds to a known CTCF sequence. It thus appears that the region of undermethylation on the maternal H19 ICR is generated through the involvement of multiple elements, each of which may operate over a region of about 300–400 nucleotides. We propose that CTCF plays an important role in preserving allele–specific methylation of the H19 ICR. It appears that all sequences in the genome are subject to de novo methylation at the time of implantation, but recruitment of CTCF protects against this process by bringing about local undermethylation. Since the maternal allele is initially unmethylated at this stage, CTCF can bind to its recognition sequence in the ICR and, in

Current Biology 1010

Figure 3. Fine Analysis of Elements Involved in Undermethylation (A) Mini fragments P21–P26 and P111 (30 bp each) were derived from subfragments S2 and S11 and cloned into test plasmid p342, methylated in vitro, and transfected into ES cells as in Figure 3. The positions of CTCF elements are marked as red ovals. Also shown are the 30bp sequences of P23 and its mutated analog (P23*). (B) DNA from pools was digested with DraI together with SmaI and subjected to Southern blot analysis with probe 1 (see Figure 2). Only P23 and P111 were able to induce appreciable undermethylation (ⵑ70%). The mutant construct (P23*) was unable to bring about undermethylation above background. (C) The p342 test vector containing various subfragments (E, F, S1, and S2) was methylated in vitro by using SssI and was transfected into ES cells. DNA from pooled colonies was subjected to Southern blot analysis (probes 1 and 2) after digestion with DraI (see Figures 1 and 2) and the methyl-sensitive enzymes SalI, SmaI, HpaII or HhaI. Fragment Fm shows appreciable undermethylation but only at HpaII and HhaI sites within the fragment itself (see map in Figure 1). The same vector (p342) containing the Aprt IE [14] was used as a positive control for undermethylation.

this way, prevent de novo methylation. The paternal allele, on the other hand, emerges from the preimplantation embryo in its methylated form. Since CTCF cannot bind if the CpG residue in its recognition sequence is methylated, this protein does not interact with the paternal ICR, and it thus remains fully modified [9, 10]. Support for this idea comes from experiments showing that both H19 alleles are protected from de novo methylation in cloned embryos derived from primordial germ cell nuclei in which the methylation pattern had already been erased [18].

Figure 4. Undermethylation of the H19 ICR in Transgenic Mice HindIII/XbaI fragments from the p342S1 and p342S3 plasmids were injected into fertilized oocytes, and transgenic mice were identified. DNA was digested with DraI with or without SmaI and subjected to Southern blot analysis by using probes 1 and 2 (see Figure 2). We evaluated the degree of undermethylation in seven founders containing S1 (7, 50%; 10, 60%; 11, 60%; 15, 25%; 23, 50%; 25, 60%; and 29, 50%) and five containing S3 (all showed ⬍5% undermethylation). In over 20 transgenic founders tested, a similar fragment from the parent p342 plasmid (no insert) also shows ⬍5% undermethylation ([14] and our unpublished data).

Genetic support for this model was recently obtained from targeting experiments in which nine out of the ten CpG residues in the four H19 CTCF elements were mutated, a change that does not affect the ability of CTCF itself to bind its recognition sequence. Although the maternal allele was unaffected, the paternally inherited targeted allele, while initially methylated in sperm, was found to be hypomethylated in somatic tissues (N. Engel and M. Bartolomei, personal communication). The likely explanation for this surprising result is that in the absence of CpG residues, CTCF can bind its recognition element even on the methylated paternal allele, and this would then act in cis to bring about local undermethylation, as predicted from our experiments. There is now ample evidence that CTCF plays a role in the maintenance of allele-specific methylation in the H19 domain [19]. Genetic experiments show that mutation or deletion of CTCF elements within the H19 ICR causes this region to become abnormally methylated in postimplantation embryos, indicating that these sites are indeed necessary for protecting the maternal allele from de novo methylation [20, 21]. By using reversed genetics on simple, well-defined substrates, we have now demonstrated that CTCF elements may also be at least partially sufficient for directing this process. It should be noted that the region upstream to the Igf2 gene also adopts an allele-specific methylation pattern at the time of implantation [6]. Unlike the H19 ICR, this non-CpG island sequence does not carry a “primary” mark. Rather, it emerges from both gametes in a methylated form and then undergoes general demethyl-

H19 Undermethylation by CTCF 1011

ation in the preimplantation embryo. At the time of implantation, only the paternal allele undergoes de novo methylation. We suggest that this specificity may be accomplished by the distal CTCF elements located in the H19 ICR on the maternal allele. In support of this possibility, it has been shown that deletion of the H19 ICR on the maternal allele actually causes the Igf2 upstream region to become de novo methylated [22, 23]. CTCF elements are also present within differentially methylated regions associated with Igf2r, Snrpn, Meg1/ Grb10 [24], Kcnq1 [25], and Dlk1-Gtl2 [26], suggesting that this protein may play a general role in maintaining allele-specific methylation patterns during early development. CTCF elements were originally described as sequences that demarcate boundaries in the genome, presumably through the binding of the CTCF protein [7, 8, 15]. Indeed, the presence of these sites on transgenes appears to prevent position-effect inactivation [27], and it was recently shown that this might be partially mediated by protecting the transgene from de novo methylation [28]. This suggests that CTCF may actually play a more extensive role in the determination of genomic methylation patterns. Since most of these sites are located in non-CpG island regions, we propose that CTCF is designed with its own mechanism for insuring DNA accessibility during development. According to this concept, the protein binds to its target sites in the genome prior to implantation, and its continued presence serves to protect local sequences from de novo methylation, thus insuring that they remain unmethylated and available for binding throughout development. Supplemental Data Supplemental Data including Experimental Procedures and two figures are available at http://www.current-biology.com/cgi/content/ full/14/11/1007/DC1/.

6. 7.

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

Acknowledgments This research was supported by grants from the National Institutes of Health, the Israel Cancer Research Foundation, and the Belfer Foundation.

18.

Received: July 31, 2003 Revised: March 26, 2004 Accepted: March 30, 2004 Published: June 8, 2004

19.

20.

References 1. Thorvaldsen, J.L., Duran, K.L., and Bartolomei, M.S. (1998). Deletion of the H19 differentially methylated domain results in loss of imprinted expression of H19 and Igf2. Genes Dev. 12, 3693– 3702. 2. Gribnau, J., Hochedlinger, K., Hata, K., Li, E., and Jaenisch, R. (2003). Asynchronous replication timing of imprinted loci is independent of DNA methylation, but consistent with differential subnuclear localization. Genes Dev. 17, 759–773. 3. Davis, T.L., Yang, G.J., McCarrey, J.R., and Bartolomei, M.S. (2000). The H19 methylation imprint is erased and re-established differentially on the parental alleles during male germ cell development. Hum. Mol. Genet. 9, 2885–2894. 4. Kafri, T., Ariel, M., Brandeis, M., Shemer, R., Urven, L., McCarrey, J., Cedar, H., and Razin, A. (1992). Developmental pattern of gene-specific DNA methylation in the mouse embryo and germline. Genes Dev. 6, 705–714. 5. Tremblay, K.D., Saam, J.R., Ingram, R.S., Tilghman, S.M., and

21.

22.

23.

24.

25.

Bartolomei, M.S. (1995). A paternal-specific methylation imprint marks the alleles of the mouse H19 gene. Nat. Genet. 9, 407–413. Brandeis, M., Ariel, M., and Cedar, H. (1993). Dynamics of DNA methylation during development. Bioessays 15, 1–5. Bell, A.C., and Felsenfeld, G. (2000). Methylation of a CTCFdependent boundary controls imprinted expression of the Igf2 gene. Nature 405, 482–485. Hark, A.T., Schoenherr, C.J., Katz, D.J., Ingram, R.S., Levorse, J.M., and Tilghman, S.M. (2000). CTCF mediates methylationsensitive enhancer blocking activity at the H19/Igf2 locus. Nature 405, 486–489. Kanduri, C., Pant, V., Loukinov, D., Pugacheva, E., Qi, C.F., Wolffe, A., Ohlsson, R., and Lobanenkov, V.V. (2000). Functional association of CTCF with the insulator upstream of the H19 gene is parent of origin-specific and methylation-sensitive. Curr. Biol. 10, 853–856. Holmgren, C., Kanduri, C., Dell, G., Ward, A., Mukhopadhya, R., Kanduri, M., Lobanenkov, V., and Ohlsson, R. (2001). CpG methylation regulates the Igf2/H19 insulator. Curr. Biol. 11, 1128–1130. Gidekel, S., and Bergman, Y. (2002). A unique developmental pattern of Oct-3/4 DNA methylation is controlled by a cisdemodification element. J. Biol. Chem. 277, 34521–34530. Frank, D., Keshet, I., Shani, M., Levine, A., Razin, A., and Cedar, H. (1991). Demethylation of CpG islands in embryonic cells. Nature 351, 239–241. Brandeis, M., Frank, D., Keshet, I., Siegfried, Z., Mendelsohn, M., Nemes, A., Temper, V., Razin, A., and Cedar, H. (1994). Sp1 elements protect a CpG island from de novo methylation. Nature 371, 435–438. Siegfried, Z., Eden, S., Mendelsohn, M., Feng, X., Tzubari, B., and Cedar, H. (1999). DNA methylation represses transcription in vivo. Nat. Genet. 22, 203–206. Bell, A.C., West, A.G., and Felsenfeld, G. (1999). The protein CTCF is required for the enhancer blocking activity of vertebrate insulators. Cell 98, 387–396. Renbaum, P., Abrahamove, D., Fainsod, A., Wilson, G.G., Rottem, S., and Razin, A. (1990). Cloning, characterization, and expression in E. coli of the gene coding for the CpG DNA methylase from Spiroplasma sp. strain MQ1(M.Sssl). Nucleic Acids Res. 18, 1145–1152. Hori, N., Nakano, H., Takeuchi, T., Kato, H., Hamaguchi, S., Oshimura, M., and Sato, K. (2002). A dyad oct-binding sequence functions as a maintenance sequence for the unmethylated state within the H19/Igf2-imprinted control region. J. Biol. Chem. 277, 27960–27967. Lee, J., Inoue, K., Ono, R., Ogonuki, N., Kohda, T., KanekoIshino, T., Ogura, A., and Ishino, F. (2002). Erasing genomic imprinting memory in mouse clone embryos produced from day 11.5 primordial germ cells. Development 129, 1807–1817. Fedoriw, A.M., Stein, P., Svoboda, P., Schultz, R.M., and Bartolomei, M.S. (2004). Transgenic RNAi reveals essential function for CTCF in H19 gene imprinting. Science 303, 238–240. Pant, V., Mariano, P., Kanduri, C., Mattsson, A., Lobanenkov, V., Heuchel, R., and Ohlsson, R. (2003). The nucleotides responsible for the direct physical contact between the chromatin insulator protein CTCF and the H19 imprinting control region manifest parent of origin-specific long-distance insulation and methylation-free domains. Genes Dev. 17, 586–590. Schoenherr, C.J., Levorse, J.M., and Tilghman, S.M. (2003). CTCF maintains differential methylation at the Igf2/H19 locus. Nat. Genet. 33, 66–69. Leighton, P.A., Ingram, R.S., Eggenschwiler, J., Efstratiadis, A., and Tilghman, S.M. (1995). Disruption of imprinting caused by deletion of the H19 gene region in mice. Nature 375, 34–39. Shemer, R., Hershko, A.Y., Perk, J., Mostoslavsky, R., Tsuberi, B.-Z., Cedar, H., Buiting, K., and Razin, A. (2000). The imprinting box of the Prader-Willi/Angelman Syndrome domain. Nat. Genet. 26, 440–443. Hikichi, T., Kohda, T., Kaneko-Ishino, T., and Ishino, F. (2003). Imprinting regulation of the murine Meg1/Grb10 and human GRB10 genes; roles of brain-specific promoters and mousespecific CTCF-binding sites. Nucleic Acids Res. 31, 1398–1406. Kanduri, C., Fitzpatrick, G., Mukhopadhyay, R., Kanduri, M.,

Current Biology 1012

Lobanenkov, V., Higgins, M., and Ohlsson, R. (2002). A differentially methylated imprinting control region within the Kcnq1 locus harbors a methylation-sensitive chromatin insulator. J. Biol. Chem. 277, 18106–18110. 26. Takada, S., Paulsen, M., Tevendale, M., Tsai, C.E., Kelsey, G., Cattanach, B.M., and Ferguson-Smith, A.C. (2002). Epigenetic analysis of the Dlk1-Gtl2 imprinted domain on mouse chromosome 12: implications for imprinting control from comparison with Igf2–H19. Hum. Mol. Genet. 11, 77–86. 27. Burgess-Beusse, B., Farrell, C., Gaszner, M., Litt, M., Mutskov, V., Recillas-Targa, F., Simpson, M., West, A., and Felsenfeld, G. (2002). The insulation of genes from external enhancers and silencing chromatin. Proc. Natl. Acad. Sci. USA 99, 16433– 16437. 28. Mutskov, V.J., Farrell, C.M., Wade, P.A., Wolffe, A.P., and Felsenfeld, G. (2002). The barrier function of an insulator couples high histone acetylation levels with specific protection of promoter DNA from methylation. Genes Dev. 16, 1540–1554.