|Home | About | Journals | Submit | Contact Us | Français|
Igf2 and H19 are coordinately regulated imprinted genes physically linked on the distal end of mouse chromosome 7. Genetic analyses demonstrate that the differentially methylated region (DMR) upstream of the H19 gene is necessary for three distinct functions: transcriptional insulation of the maternal Igf2 allele, transcriptional silencing of paternal H19 allele, and marking of the parental origin of the two chromosomes. To test the sufficiency of the DMR for the third function, we inserted DMR at two heterologous positions in the genome, downstream of H19 and at the alpha-fetoprotein locus on chromosome 5. Our results demonstrate that the DMR alone is sufficient to act as a mark of parental origin. Moreover, this activity is not dependent on germ line differences in DMR methylation. Thus, the DMR can mark its parental origin by a mechanism independent of its own DNA methylation.
The H19 and Igf2 genes are part of a cluster of imprinted genes on the distal end of mouse chromosome 7. The genome organization and regulation of the genes in this cluster are highly conserved on chromosome 11p15.5 in humans (32-34). IGF2 is a potent fetal mitogen (8, 31), and loss of imprinting mutations that result in increased IGF2 expression are associated with Beckwith-Wiedemann syndrome and with several types of tumors (12, 13, 30, 47). The biological function of the H19 gene product is less clear. Recent studies have suggested that Wilms' tumors frequently associated with Beckwith-Wiedemann syndrome are more likely when the loss of imprinting includes H19 in addition to IGF2 (7). These results are consistent with those of earlier cell culture studies suggesting that the H19 RNA might function as a tumor suppressor (17).
The monoallelic expression of the H19 and Igf2 genes is dependent on a common cis-acting regulatory element, the DMR (for differentially methylated region) located between kb −4.4 and −2 upstream of the H19 promoter (23, 44) (Fig. (Fig.1A).1A). This element contains a transcriptional insulator that prevents activation of the Igf2 promoters by the shared enhancers located downstream of the H19 gene. When paternally inherited, the DMR sequence is methylated and insulator activity is blocked so that Igf2 expression is permitted (2, 19, 23-25, 41). At the same time, the methylated paternal DMR induces epigenetic changes at the H19 promoter that silence H19 expression (4, 9, 40). These epigenetic changes are developmentally programmed, and once established, they can maintain repression of the paternal H19 even in the absence of the DMR (39, 40).
Beyond the DMR's role in regulating transcription of the two genes, genetic evidence is consistent with the notion that the DMR has a third distinct function: it is at least part of the imprinting control element (ICE) for the H19 and Igf2 genes. That is, the element appears to be necessary for marking the chromosomal origin of the two genes and of H19 transgenes (11, 23, 44). Furthermore, molecular studies have shown that the H19DMR is methylated in sperm but not in oocytes and the differential methylation is maintained during the global changes in methylation patterns associated with early development (3, 14, 45). These findings suggest that differential methylation of the DMR is probably the primary mark for the imprinting of the Igf2/H19 locus and, likewise, that the DMR sequences are those that contain the original or primary epigenetic mark distinguishing the maternal and paternal chromosomes. In this study, we determined that the DMR is sufficient to mark the parental origins of normally nonimprinted loci. However, this activity is not always dependent on germ line differences in methylation of DMR sequences.
All animal research was conducted in full accord with the requirements of the NICHD Animal Care and Use Committee. To generate H19R and H19F, the H19DMR carried on a 2.4-kb BglII fragment was inserted at the kb +10 EcoRI site of the H19 locus. Targeting vectors included 7 kb of 5′ homology on a SalI-EcoRI fragment and 3 kb of 3′ homology on an EcoRI-BamHI fragment, a floxed NeoR cassette for positive selection, and the Diphtheria toxin-A gene for negative selection. After electroporation into embryonic stem cells, G418-resistant clones were screened by Southern blotting with a 1.2-kb BamHI-SalI probe (5′ end) and a 2.1-kb XbaI-BamHI probe (3′ end). If correctly targeted on the 5′ end, H19R cells digested with BamHI enzyme show a 9.0-kb band in addition to the 11.3-kb band indicative of a wild-type chromosome. If correctly targeted on the 3′ end, H19R cells digested with ScaI show a 10.1-kb fragment in addition to the 19-kb fragment indicative of the wild-type chromosome. When digested with ScaI, H19F clones correctly targeted on the 5′ end yield an 11.3-kb band in addition to the 19-kb band indicative of the wild-type chromosome. Clones correctly targeted on the 3′ end show a 7.9-kb ScaI fragment in addition to the 19-kb wild-type ScaI fragment.
To generate AfpA and AfpB, the H19DMR carried on a 2.4-kb BglII fragment was inserted at the kb −0.9 XbaI site of the Afp locus. For AfpD, the H19DMR was carried on a 9-kb BamHI-XbaI fragment. Targeting vectors included 2.5 kb of 5′ homology on a BstEII-XbaI fragment and 3.3 kb of 3′ homology on an XbaI-EcoRI fragment, a floxed NeoR cassette for positive selection, and the Diphtheria toxin-A gene for negative selection. After electroporation into embryonic stem cells, G418-resistant clones were screened by Southern blotting using the 0.7-kb EcoRI-HindIII probe (5′ end) or the 1.4-kb EcoRI-XbaI probe (3′ end). If correctly targeted on the 5′ end, AfpA cells digested with EcoRI enzyme yield a 6.3-kb band in addition to the 7.8-kb band indicative of a wild-type chromosome. If correctly targeted on the 3′ end, AfpA cells digested with XbaI show a 6.8-kb fragment in addition to the 5-kb fragment indicative of the wild-type chromosome. When digested with EcoRI and with XbaI, AfpB clones correctly targeted on the 5′ end yield a 5.8-kb band in addition to the 4.3-kb band indicative of the wild-type chromosome. Clones correctly targeted on the 3′ end show a 7.4-kb XbaI fragment in addition to the 5-kb wild-type fragment. AfpD candidates were digested with EcoRI plus XbaI or with XbaI alone to analyze the 5′ and 3′ insertion sites, respectively. Clones correctly targeted at the 5′ end show an 8-kb band in addition to the 5-kb band indicative of the wild-type chromosome, while clones correctly targeted at the 3′ end show 6.8- and 5.0-kb bands indicative of the mutant and wild-type chromosomes, respectively.
Correctly targeted clones were injected into C57/BL6-J blastocysts to generate chimeric founder mice that were mated to EIIa-cre transgenic females to generate mice in which the NeoR cassette was deleted (28). These mice were identified by PCR amplification across the NeoR insertion site. To generate progeny for methylation analysis, mice carrying these mutant chromosomes and a domesticus version of the endogenous Igf2/H19 locus were crossed with Dis7CAS mice. Dis7CAS mice are mostly domesticus but are homozygous castaneus across the H19/Igf2 locus (16). Thus, the Dis7CAS mice provide a wild-type chromosome 7, albeit one that can be distinguished from the wild-type domesticus chromosome via multiple DNA polymorphisms. Alternatively, as indicated in the text, mutant chromosomes were introduced into an H19k519/H19k519 background and then backcrossed again to H19k519/H19k519 to generate mice for methylation analysis. The H19k519 chromosome carries a 9-kb deletion that removes sequences between kb −10 and −0.7, a span that includes the endogenous DMR (23).
Genomic DNA was treated with sodium bisulfite according to the manufacturer's recommendation (Intergen). Two micrograms of testes genomic DNA, pooled DNA from 100 blastocysts, or the total DNA of individual embryonic day 7.5 (e7.5) or e8.5 embryos were used in each conversion.
The DNA from approximately 20 blastocysts was used for each PCR. Each subregion was amplified using nested primers essentially as described previously (45). The sequences of the newly designed primers are as follows: subregion 1, BDMRTF5 (5′-TTAGGTATAGTATTTAATGATTTATAAGGG-3′) and BAfpBR3 (5′-AAATACACTATATTTCTAATATAAATTAT-3′), BDMRTF6 (5′ GGGGTGGTATAATATATATTTTTTGGGTAG-3′) and BAfpBR4 (5′-TTGTTTTTATAATCACATCTTTAACATAAC-3′); subregion 2, BDMRTF7 (5′-ATATGGTTTATAAGAGGTTGGAA-3′) and BDMRTR3 (5′-CTACCCAAAAAATATATATTATACCACCCC-3′), BDMRTF8 (5′-TATTTGTGTTTTTGGAGGGGGTT-3′) and BDMRTR4 (5′-CCCTTATAAATCATTAAATACTATACCTAA-3′); subregion 3, BMsp4t1 and BHha4t2, BMsp4t2 and BHha4t3; subregion 4, BHha2t1 and BMsp3t2, BHha2t2 and BMsp3t; subregion 5, BMsp2t1 and BHha1t3, BMsp2t2 and BHha1t4; subregion 6, BDMRTF3 (5′-ATGTAAGTGTGTTTTGTGTAGTAATTGATG-3′) and BMsp1t6, BDMRTF4 (5′-AGATAGTATTGAGTTTGTTTGGAGTTTGAG-3′) and BMsp1t5; subregion 7, BAfpBF1 (5′-TTAAGATGATGATGTTAATAGTAATAAATG-3′) and BDMRTR1 (5′-ACTTTTAACTACATTAAATAAACAATAAAC-3′), BAfpBF2 (5′-GGTATTGATATATTTTTTGATTTTAAGAGTG-3′) and BDMRTR2 (5′-AACTAAACTCCTAATAATTCATTTACATTT-3′). The PCR products were cloned using a TA cloning kit (Invitrogen), and the clones were sequenced on both strands. Alternatively, the PCR product was purified and restricted with AciI enzyme.
To test the ability of the DMR to mark the chromosome differentially in a parent-of-origin-dependent manner, we inserted it at two locations in the genome: at the kb +10 position of the H19 gene and at kb −0.9 of the alpha-fetoprotein gene (Afp) (Fig. (Fig.1).1). We believe that the kb +10 position is past the 3′ boundary of the imprinted cluster on chromosome 7 because no molecular marks distinguishing maternal and paternal chromosomes have been documented in the region. Moreover, the enhancer elements centered at kb +8 and at kb +24 are each fully functional on both maternal and paternal chromosomes (22, 29). Nonetheless, the kb +10 position is clearly proximal to other sequences that may normally contribute to marking the parental origin of the Igf2/H19 alleles. Insertion at the Afp locus on chromosome 5 is therefore a more stringent test of the sufficiency of the DMR to act as an ICE. The Afp gene is expressed biallelically (data not shown), and there are no known imprinted genes on mouse chromosome 5 (www.mgu.har.mrc.ac.uk/research/imprinted/imprin.html).
The Afp locus has several advantages for our study. First, mice heterozygous for loss-of-function mutations show no discernible phenotype (15). Second, the regulated expression of mouse Afp has been extensively studied in vitro and in vivo using transgenic animals (38). These studies have identified three upstream enhancers as well as promoter elements that can account for activation of the Afp gene (Fig. (Fig.1B).1B). Insertion of the DMR at the kb −0.9 position at Afp puts the DMR just upstream of the promoter, thus mimicking its location at the H19 locus. However, this same positioning also mimics the organization of the Igf2 gene in that the DMR, with its insulator activity, now separates the Afp promoter and enhancer elements.
The Afp locus is not a CpG-rich region. Specifically, there are no CpG islands in the enhancer or promoter regions or in any sequences within 50 kb upstream and 14 kb downstream of the DMR insertion site (http:ccnt.hsc.usc.edu/cpgislands). (The region 14 kb downstream of the insertion site is defined as a CpG island only when using the least stringent criteria.) At its endogenous location, the DMR is proximal to a CpG island that includes the H19 promoter and extends into the H19 RNA coding sequences. There is no equivalent sequence motif at the Afp locus. For example, the 0.9-kb sequences downstream of the insertion site on chromosome 5 (which include the Afp promoter) contain 10 scattered CpGs.
We isolated the DMR as a 2.4-kb BglII fragment. This fragment carries the 65 CpG base pairs that are methylated in sperm and escape demethylation during early embryogenesis. These sequences include all four binding sites for CTCF, a protein that is crucial for normal transcriptional regulation of the Igf2/H19 gene cluster, but the BglII fragment does not include the promoter-proximal G-rich repeats. At each chromosomal locus, the DMR was inserted in both orientations to generate mutant chromosomes H19R, H19F, AfpA, and AfpB (Fig. (Fig.11).
We generated founder lines by injecting mutated embryonic stem cells into wild-type blastocysts. The NeoR cassettes used for positive selection in vitro were removed by crossing these founders to females carrying a Cre recombinase gene under the control of the EIIa promoter (28). Male and female progeny of these crosses were then mated with wild-type tester mice to generate pups for analysis. Our crosses were set up so that pups inherited the DMR insertion and also a domesticus wild-type copy of the endogenous DMR from one parent while inheriting a castaneus wild-type version of the endogenous DMR from the other parent. Given the polymorphisms that distinguish castaneus and domesticus DMR alleles and the polymorphisms generated by the different sequences flanking the DMR in its normal and heterologous positions, we could distinguish all three DMRs in each pup: endogenous paternal DMR, endogenous maternal DMR, and heterologous DMR.
We tested for cytosine methylation of AciI sites in DNAs isolated from somatic tissues of postnatal animals and found that the inserted DMRs, just like the endogenous copy, are methylated when paternally inherited but not when maternally inherited (Fig. (Fig.2A,2A, top panel). This property is orientation independent for both insertion locations. These results demonstrate that the DMR contains sufficient information to mark its own parental origin, even on a heterologous chromosome in a nonimprinted genomic context.
At its normal position, the DMR, whether paternal or maternal, is methylated in mature sperm. In fact, demonstration of the acquisition of methylation on both chromosomes during spermatogenesis is vital empirical support for the notion that parent-of-origin-specific methylation is the primary (or gametic) imprint of the H19/Igf2 locus (46). We wished to determine whether this mechanism would apply to the mutant chromosomes and therefore isolated DNAs from the testes of sexually mature males and assayed for methylation at AciI sites in the DMR inserts (Fig. (Fig.2A,2A, bottom panel). At the H19 locus, the DMR insert is always completely methylated and thus behaves identically to the endogenous maternal and paternal DMR alleles. In contrast, the DMR insert at the Afp locus is not methylated. We confirmed and extended these results by digesting with other methylation-sensitive enzymes, including HhaI, ClaI, BspEI, and HpaII (data not shown but see Fig. Fig.55 for maps), which together assayed methylation at 24 CpG sites within the DMR. To examine all 65 CpGs, we examined the methylation status of several testis samples by direct sequencing (Fig. (Fig.3A).3A). To assay the entire insert region, we required seven nested-PCR amplifications. Each amplification was done on two independently prepared DNA samples, and multiple clones were analyzed for each reaction. The results confirmed that there is no consistent cytosine methylation of the DMR insert in adult testis.
In sum, the paternal specific cytosine methylation found on Afp::DMR inserts in differentiated tissues cannot be explained by maintenance of a difference inherited through the germ cells. Rather, our results imply that when inserted at the Afp locus, the DMR is marked differentially in the two gametes by a mechanism other than its own DNA methylation and that the differential DNA methylation across the DMR is acquired as a secondary imprint later in the development.
To determine when the methylation at the mutant Afp loci is acquired, we examined various developmental stages for parent-of-origin-specific methylation of the DMR. First, we assayed methylation at the blastocyst stage. We converted DNA isolated from pools of approximately 100 blastocysts and then used one-fifth of this DNA for each PCR. We independently amplified DNA two to four times for each PCR primer set and then analyzed multiple clones for each reaction. Our results showed that paternally inherited DMRs are largely unmethylated at this stage of development (Fig. (Fig.3B)3B) even though control sequences (the maternal Snrpn locus and the paternal endogenous H19DMR) from the same DNA samples were methylated as expected (data not shown). The lack of the DNA methylation in the blastocyst confirms that the methylation of the DMR is not the primary imprint.
We next tested restriction enzyme sensitivity of PCR-amplified samples to examine methylation of DNAs isolated from pooled blastocysts and also from individual e7.5 and e8.5 embryos. The sequences we examined included two of the four CTCF binding sites. Paternal DMRs from gastrulated embryos but not from blastocysts were always methylated at least partially (Fig. (Fig.4).4). Finally, we examined DNAs isolated from e11.5 and e12.5 embryos by using Southern analyses and determined that methylation of the DMR was complete and not distinguishable from that seen in adult tissue samples (data not shown). Thus, the acquisition of parent-of-origin-specific methylation differences occurs after implantation and around the time of gastrulation. This is the time period when large parts of the genome, including most CpG islands, are undergoing methylation (36). Our results are consistent with two possibilities: (i) the maternally inherited DMR is refractory to de novo methylation or (ii) the paternally inherited DMR attracts such methylation.
DMR insertions at the H19 locus (endogenous and kb +10) are methylated in sperm while those at the Afp locus are not. We wished to see whether we could isolate DNA sequences responsible for this difference and therefore generated a larger insertion at the Afp locus (Fig. (Fig.1).1). Specifically, this new insert, AfpD, contains sequences from kb −10 to −0.7 upstream of the H19 gene and thus carries, in addition to the DMR, the G-rich repeat elements that are common to many imprinted genes. However, this larger insert behaves similarly to the 2.4-kb DMR element. That is, the paternally inherited insert is hypermethylated in differentiated tissues (data not shown) but not methylated in sperm (Fig. (Fig.55).
Loss-of-function mutations in the mouse suggest that the sequences encompassing the H19DMR are necessary for at least three genetic functions that are each crucial for maintaining the normal monoallelic expression patterns characteristic of this locus. First, these sequences carry a methylation-sensitive transcriptional insulator whose normal function prevents activation of the Igf2 gene by the downstream enhancers it shares with the H19 gene (2, 19, 23-25, 41). Gain-of-function analyses have confirmed that the DMR sequences are also sufficient for this activity, at least in vitro (2, 19, 20, 23-25). Second, a paternally inherited and marked DMR acts as a developmentally regulated silencer that induces in cis stable changes at the H19 locus that maintain silence of the paternal allele (39, 40). Third, the results of the genetic studies were consistent with the notion that the DMR acts as an ICE. In this context, we define the ICE narrowly and mean only the cis sequences whose gametic epigenetic marking distinguishes the maternal and paternal chromosomes.
Our results demonstrate first that the 2.4-kb DMR is in fact an ICE. The DMR sequences maintain the ability to keep track of their parental origin even on a heterologous chromosome. Although the H19 and Igf2 genes are part of a very large (>1 Mb) imprinted domain, the functional differences noted between the paternal and maternal alleles of these two genes can be ascribed to a highly localized signal contained on a mere 2.4-kb sequence. The DMR sequence carries information that results in its being methylated only when paternally inherited. Because sequences within the DMR act as a methylation-sensitive insulator and silencer, this methylation induces transcriptional differences in paternal and maternal chromosomes. At its endogenous location on chromosome 7, the methylated DMR allows expression of paternal Igf2 by inactivating the transcriptional insulator that comaps with the ICE. The methylated DMR conversely blocks expression of paternal H19 by acting as a developmentally regulated silencer (37, 43). In fact, we, of course, analyzed transcription of Afp in our mutant mice and noted a fivefold parent-of-origin effect on transcription (Sangkyun Jeong and K.P., unpublished observations).
However, interpretation of these transcription results is not straightforward. Given the topology of the Afp locus, we introduced the DMR into a position where it potentially operates as an insulator on the unmethylated maternal chromosome but as a silencer on the methylated paternal chromosome (Fig. (Fig.1).1). Such a dual effect of the insertion on Afp transcription is consistent with the results we actually obtained: maternal inheritance lowering expression about fivefold and paternal inheritance lowering expression about 25-fold compared with wild-type levels. However, a real understanding of the transcriptional effects of the insertions will require the analysis of several additional control chromosomes that we are presently generating. These new mutations will allow us to distinguish silencing from insulation and also to quantitate any effects on promoter activity that are due to shifting the distance between the Afp promoter and its upstream enhancers.
Localization of the parent-of-origin identification entirely to the DMR clarifies the complex and sometimes perplexing analysis of H19 transgenes generated by pronuclear injection. The earliest transgenic studies demonstrated a critical importance for the DMR but also showed that other sequences such as the enhancer elements were essential (11, 35). These same studies also indicated that copy number was critical. More recent investigations using 5′ sequences that include the entire DMR actually showed that even single-copy H19 transgenes can be imprinted (5, 23). It seems plausible that the apparent reliance on H19 sequences outside the DMR, like the apparent reliance on multiple transgene insertions, was noted only because the so-called DMR was in fact a shortened (i.e., mutated) version.
The second major finding in this report is that the epigenetic imprinting marking of the DMR and cytosine methylation of the DMR are separable. In other words, at least with the DMR insert at the Afp locus, the primary imprint does not appear to be its DNA methylation. Rather, differential methylation is established after implantation by a mechanism that is not yet understood but presumably as a result of an interpretation of the true primary mark. Our results do not imply that DNA methylation does not play a critical role in parent-of-origin-specific expression or in imprinting even at the Afp::DMR locus but suggest that methylation of the DMR itself is not the obligatory gametic mark.
We examined Afp sequences for CpG-rich regions that might play a surrogate imprinting role on our chimeric chromosomes. As described in Results, we did not note any nearby CpG islands. The 0.9 kb between the DMR insertion and the Afp transcriptional start site contains 10 CpGs, including four dyads whose methylation status could be evaluated by restriction digestion and Southern blotting. Our initial analysis did not reveal any consistent methylation patterns to distinguish maternal from paternal chromosomes or wild-type from insertion chromosomes (Sangkyun Jeong and K.P., unpublished observations).
A key question that our present study cannot directly address is whether the same primary mark that ultimately establishes parent of origin at the Afp::DMR locus also applies to the endogenous DMR. In other words, is the difference between the functions of the DMR at its normal position and at the Afp locus only the timing of when the primary mark is converted to differential methylation? Alternatively, does the insertion of the DMR at the Afp locus create a completely novel mechanism for genomic imprinting?
Our results recall those of El-Maarri et al. (10), who examined methylation patterns for the SNRPN locus in human oocytes. Although they started with very limiting material and were not able to identify the ICE as a sufficient element for the imprinting, these investigators did not detect maternal specific methylation and thus suggested that heritable alterations other than DNA methylation might mark maternal and/or paternal alleles.
However, our results do not fit well with those of another important 2001 study (Howell et al. ), which examined the role of the oocyte-specific isoform of the DNA methyltransferase gene 1 (Dnmt1o). Dnmt1o is a maternal effect gene. Females homozygous for a Dnmt1o deletion are fine, but fetuses from such mothers do not survive and show loss of imprinting. Specifically, for example, H19 expression becomes biallelic and one half of all paternal chromosomes show complete loss of methylation while the other half show the normal methylated pattern. (Likewise, Snrpn becomes biallelic, with half of the maternal chromosomes aberrantly showing a complete loss of methylation.) Given the protein expression and localization patterns, Howell et al. explained these results by postulating that the Dnmt1o isoform is required specifically at the eight-cell morula stage to maintain methylation during cell division. This interpretation implies that the paternal marking of the endogenous H19DMR (and Snrpn) is dependent on DNA methylation even before implantation.
The nature of the primary mark on the DMR insert at Afp is presently unknown. Besides DNA methylation, additional differences in the chromatin structure of H19DMR on each chromosome have been previously reported (1, 14, 18, 26, 27, 42). However, these parent-of-origin-specific differences in nuclease sensitivity and in histone codes were only characterized in tissues which also showed differential methylation, thus making it impossible to distinguish the cause-and-effect relationships of these potential marks. Davis et al. (6) examined the acquisition of DNA methylation at the endogenous H19DMR during spermatogenesis. They noted that the distinctive methylation of the H19DMR was acquired in a two-step process. First, all methylation was removed from the paternal chromosome, and then both maternal and paternal DMRs were remethylated. Their experiments demonstrated that, even without cytosine methylation, the maternal and paternal chromosomes were functionally nonequivalent because the paternal chromosome was remethylated earlier than the maternal. However, these experiments could not clarify whether the difference implied the existence of a primary imprint other than DNA methylation or it implied only that secondary chromatin changes caused by differential DNA methylation can remain for a while even after the erasure of that methylation. Further characterization of the epigenetic modification by using the system reported here will clarify the role of nonmethylation epigenetic marks and help illuminate the general mechanisms by which the genome is imprinted.
We thank Marisa Bartolomei, Tamara Davis, and Jacquetta Trasler for providing protocols and advice for the direct sequencing assay for DNA methylation.
This work was supported by the Intramural Research program of the National Institute for Child Health and Human Development.