|Home | About | Journals | Submit | Contact Us | Français|
For most imprinted genes, a difference in expression between the maternal and paternal alleles is associated with a corresponding difference in DNA methylation that is localized to a differentially methylated domain (DMD). Removal of a gene's DMD leads to a loss of imprinting. These observations suggest that DMDs have a determinative role in genomic imprinting. To examine this possibility, we introduced sequences from the DMDs of the imprinted Igf2r, H19, and Snrpn genes into a nonimprinted derivative of the normally imprinted RSVIgmyc transgene, created by excising its own DMD. Hybrid transgenes with sequences from the Igf2r DMD2 were consistently imprinted, with the maternal allele being more methylated than the paternal allele. Only the repeated sequences within DMD2 were required for imprinting these transgenes. Hybrid transgenes containing H19 and Snrpn DMD sequences and ones containing sequences from the long terminal repeat of a murine intracisternal A particle retrotransposon were not imprinted. The Igf2r hybrid transgenes are comprised entirely of mouse genomic DNA and behave as endogenous imprinted genes in inbred wild-type and mutant mouse strains. These types of hybrid transgenes can be used to elucidate the functions of DMD sequences in genomic imprinting.
At imprinted loci in mammalian species, the maternal and paternal alleles of a gene are distinguished from one another by different epigenetic modifications, genomic imprints, which are established during oogenesis and spermatogenesis. The transmission of these imprints is essential for normal embryonic development and leads to monoallelic gene expression in the embryo and adult (2, 11). The epigenetic modification that distinguishes the maternal and paternal alleles of most, if not all, imprinted genes is DNA cytosine methylation (10, 11, 23). The parental alleles have different levels of DNA methylation, usually concentrated in a single location within or surrounding the gene. A genomic region with this epigenetic feature is generally 1 to 5 kb in size and is called a differentially methylated domain (DMD). Within a DMD, one parental allele is highly methylated on the majority of CpG dinucleotides, and the opposite parental allele is unmethylated or methylated on a small percentage of CpG dinucleotides.
The methylation patterns of several DMDs have been extensively studied, including DMD2 of the insulin-like growth factor type 2 receptor (Igf2r) gene, the DMD of the Snrpn gene, and the DMD of the H19 gene. DMD2 is found in the second intron of Igf2r, is approximately 3 kb in size, and contains 28 CpG dinucleotides (27). The DMD2 CpGs are methylated on the maternal allele and unmethylated on the paternal allele. The DMD of Snrpn includes promoter sequences, the entire first exon, and the first intron of the gene and is approximately 6 kb in size (6, 24). Like DMD2 of Igf2r, the DMD of Snrpn is highly methylated on the maternal allele and unmethylated on the paternal allele (6, 24). These differences in methylation are established during gametogenesis; the DMDs of Igf2r and Snrpn are highly methylated in oocytes and unmethylated in sperm (7, 24). In contrast, the DMD of H19 is approximately 2 kb in size, is found 5′ of promoter sequences, and acquires extensive CpG methylation exclusively during spermatogenesis (30). Following fusion of the gametes at conception, the maternal Igf2r and Snrpn methylation patterns and the paternal H19 methylation pattern are maintained in the embryo. Many other imprinted genes have strict parent-of-origin differences in methylation, concentrated in small DMDs (2, 22), suggesting that there is an essential role for the DMDs in genomic imprinting.
Deletion of DMD sequences from an endogenous imprinted locus removes all evidence of allelic differences between the maternal and paternal alleles (4, 29, 33, 34). A 300-kb transgene containing the entire 93-kb Igf2r locus is imprinted if DMD2 is present, but not imprinted if it is removed (32). However, smaller transgenes comprised only of DMD2 sequences are not imprinted. Likewise, 130-kb transgenes, which contain both the Igf2 and H19 genes, are imprinted (1). However, small transgenes containing the entire H19 DMD plus immediately surrounding sequences are not consistently imprinted (construct XXRsdBam) (12). There are no reports of transgenes made solely of H19 DMD sequences. These observations indicate that sequences outside of DMDs are also required for gene imprinting. Therefore, the elements required for imprinting these transgenes, the imprint control elements (ICEs), most likely consist of the DMD sequences and unidentified sequence elements that are found at a distance from the DMD. The mechanism by which the ICEs function to imprint the gene is not known.
RSVIgmyc is an imprinted mouse transgene of approximately 17 kb, comprised of sequences from nonimprinted sources, that is imprinted at all chromosomal integration sites. All independently derived transgenic lines have a highly methylated maternal allele and an undermethylated paternal allele (9). Like endogenous imprinted genes, RSVIgmyc undergoes allele-specific changes in DNA methylation during development and expresses a gene from only one parental allele (10, 28). The absence of a single required RSVIgmyc sequence element for its imprinting suggests a functional redundancy for the ICEs within the RSVIgmyc transgene (16). We show here that the DMD of RSVIgmyc is the region that contains pBR322 and Rous sarcoma virus (RSV) sequences. Like the DMDs of the larger Igf2r transgenes, the DMD of RSVIgmyc alone is not imprinted, and removal of DMD sequences from the RSVIgmyc transgene results in a nonimprinted transgene. To examine the possibility that DMDs from different imprinted genes have a shared imprinting function, we introduced DMD sequences from these genes into the nonimprinted version of RSVIgmyc. Transgenes containing sequences from the Igf2r DMD2 were imprinted, with the maternal allele being more methylated than the paternal allele. Importantly, imprinted Igf2r transgenes containing only the direct repeats of DMD2 exhibit the same behavior as endogenous imprinted genes in different inbred and mutant mouse strains.
DNA was isolated from the tail of adult mice or entire embryos by proteinase K (Roche) digestion, followed by phenol-chloroform extraction and ethanol precipitation. DNA samples were resuspended in TE (10 mM Tris-HCl [pH 7.5], 1 mM EDTA). DNA was digested at the appropriate temperature with restriction endonucleases (10 U/μg of DNA), electrophoresed on an agarose gel, and transferred to GeneScreen nylon filters (NEN Research Products, Boston, Mass.) by the method of Southern (26). Filters were hybridized with one of four probes at 42°C in 40% formamide and washed in a mixture containing 0.1× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) and 0.1% sodium dodecyl sulfate at 65°C. The Cα probe is a 1.75-kb EcoRI-XbaI fragment of Cα. Probe a is a 0.6-kb PstI-BamHI fragment from the 3′ region of c-myc, and probe b is a 1.3-kb PvuII-XhoI fragment of c-myc exon 3. The RSV probe is a 0.44-kb fragment from the long terminal repeat (LTR) region of RSV (9).
All transgenic mice were created in an inbred FVB/N genetic background by pronuclear injection (21). Transgenic lines were established from these founders, and imprinting characteristics of the lines were evaluated by passage through the maternal and paternal germ lines.
DNA fragments injected into mouse zygotes for the creation of transgenic mice were derivatives of the RSVIgmyc construct (28). All constructs were created by placing plasmid vector sequences (pKS+; Stratagene) into the unique KpnI site of RSVIgmyc or into a unique NotI site created by inserting a NotI linker at the KpnI site. DMD and intracisternal A particle (IAP) sequences were inserted at the unique EcoRI site generated by the removal of the pBR322/RSV region. The inserted sequences for IAPIgmyc, Igf2rIgmyc, H19SIgmyc, SnrpnIgmyc, TR1Igmyc, and TR2+3Igmyc transgenes were derived by PCR amplification of genomic DNA with oligonucleotide primers designed to introduce flanking EcoRI sites. The H19LIgmyc construct was created with an H19 HindIII fragment (2,232 bp, accession no. U19619; nucleotides [nt] 942 to 3174); the HindIII and EcoRI sites were filled in and destroyed to subclone the H19 fragment into Ig/myc. The constructs were linearized prior to pronuclear injection by removal of all pKS+ sequences by KpnI or NotI digestion and gel isolation.
PCR amplification of each transgene insert was conducted under the following reaction conditions: 0.5 μg of mouse genomic DNA (except IAP, which used 10 ng of a plasmid containing the 5′ LTR from the Aiapy allele) (18), 1× PCR buffer (Gibco/BRL), 3 mM MgCl2 (Gibco/BRL), 0.2 mM deoxynucleoside triphosphates (dNTPs), 0.4 μM each primer, and 1.25 U of Taq DNA polymerase (Gibco/BRL) in a 50-μl reaction mixture. The cycling conditions were as follows: 94°C for 3 min; 30 cycles (25 cycles for IAP) of 94°C for 1 min, 65°C for 30 s, and 72°C for 1.5 min; and a final extension at 72°C for 10 min.
The following PCR primers were used in this study: Igf2rIgmyc, IGR2R-C (TCCGAATTCCTAGTGGGGCACCTTCATTTGCATG) andIGF2R-D (CGTGAATTCGATTTTAGCACAACTCCAATTGTGC); H19SIgmyc, H19-C (GCAGAATTCGCAAGGAGACCATGCCTATTCTTGG) and H19-D (GCAGAATTCCCTCATGAAGCCCATGACTATGGG); SnrpnIgmyc, SNRPN-A (TATGGAATTCGATATAGCCTAGAAACCAG) and SNRPN-B (AAAATTCCAAATCTAGAATGTTTTGGTC); IAPIgmyc, U5IAP (CGAGAATTCTGTTATTCGACGCGTTCTCACG) and U3IAP (GTGGGAATTCGCCCCCACATT); TR1Igmyc, IGF2R-C and TR1-R (AGGGAATTCTGATCAGGGCCAACGC); and TR2+3Igmyc, IGF2R-D and TR2+3-F (CCTGAATTCAGAACCCTTCGAATCC).
The experiments described here examine the parent-specific methylation of a number of mouse transgenes. All were derived from the imprinted RSVIgmyc transgene. The RSVIgmyc plasmid was made by combining two EcoRI DNA fragments: one was a 15-kb mouse genomic fragment from a murine translocation breakpoint (Ig/myc), and the other was a 2.5-kb fragment containing pBR322 and RSV LTR sequences (pBR/RSV) (9). Ig/myc and pBR/RSV mouse transgenic lines were created and studied. As well, a number of mouse transgenes were made by combining Ig/myc sequences with other mouse genomic sequences, either from the DMDs of endogenous imprinted genes or from an IAP LTR. The imprinting of each transgene was studied by comparing DNA methylation patterns of the maternal and paternal alleles. Allelic differences in expression were not examined, because in the absence of RSV sequences, most of the hybrid transgenes would not be transcriptionally active (9, 16).
Assuming RSVIgmyc and endogenous imprinted genes share a common imprinting mechanism, the DMD of RSVIgmyc should be a required imprinting element. To define RSVIgmyc's DMD, a series of Southern blots were performed with DNA from carriers of an RSVIgmyc transgene. Details of the structure of RSVIgmyc, including the location of methylation-sensitive restriction sites and hybridization probes used in Southern blots, are shown in Fig. Fig.1A.1A. Using double digests of EcoRI plus a methylation-sensitive restriction endonuclease, we showed that CpG dinucleotides within HpaII, HhaI, and BstUI sites (recognition sequences CCGG, GCGC, and CGCG, respectively) of pBR/RSV are methylated on the maternal RSVIgmyc allele, but unmethylated on the paternal allele (Fig. (Fig.1B).1B). In contrast, the extent of methylation difference between the maternal and paternal alleles within the adjacent IgA and c-myc sequences is negligible. IgA sequences (defined by an EcoRI-BglII digest) are heavily methylated on both the maternal and paternal alleles (Fig. (Fig.1C).1C). The CpG island sequences of c-myc, located within intron 1, have previously been shown by a HpaII digest to be unmethylated on both parental alleles (16). HhaI and BstUI sites in the CpG island (between the BglII and XbaI sites) were also unmethylated on both the maternal and paternal alleles (data not shown). From this analysis, pBR/RSV, clearly delineated by EcoRI sites, is the only element within RSVIgmyc in which one parental allele (maternal) is highly methylated and the other allele (paternal) is unmethylated. Therefore, pBR/RSV is the DMD of RSVIgmyc.
Removal of DMD sequences from imprinted Igf2r transgenes abolishes their imprinting (32). Correspondingly, removal of pBR/RSV from RSVIgmyc should negate its imprinting. However, it was previously shown that removal of pBR322 alone or removal of RSV alone from RSVIgmyc does not disrupt its imprinting (9). To address the requirement of the entire DMD (pBR/RSV) for RSVIgmyc imprinting, we created three Ig/myc transgenic lines by using a transgene from which the pBR/RSV segment of RSVIgmyc was excised (Fig. (Fig.2A).2A). For each line, the maternal allele's methylation pattern was compared to that of the paternal allele. The results for one line are shown in Fig. Fig.2B2B to toD.D. Genomic DNAs from two hemizygous transgenic carriers, one with a maternal Ig/myc allele and one with a paternal allele, were digested with the methylation-sensitive enzyme HpaII, HhaI, or BstUI. A Southern blot was made from the digested DNAs and hybridized with a probe to the 3′ c-myc region of the transgene (Fig. (Fig.2B).2B). For each restriction enzyme, the patterns of hybridizing bands on the two DNA samples are identical, consistent with the presence of equal, nonimprinted methylation patterns on the two parental alleles. Methylation patterns within other portions of the transgene were determined with the same DNA samples and restriction enzyme digests, but using hybridization probes to the IgA and 5′ c-myc regions of the transgene (Fig. (Fig.2C2C and andD,D, respectively). There were no observed differences between Southern blots made from DNA of a maternal Ig/myc carrier and those made from DNA of a paternal Ig/myc carrier. Identical maternal and paternal methylation patterns were observed for three other independent Ig/myc transgenic lines (Table (Table1).1). The consistent similarity between maternal and paternal Ig/myc methylation patterns, in comparison to the consistent differential methylation of RSVIgmyc (9), shows that removal of the pBR/RSV region from the RSVIgmyc transgene eliminates its imprinting. This observation suggests that the pBR322 region and the RSV region of the RSVIgmyc transgene have redundant functions; individually, they function to maintain imprinting, but removal of both abolishes imprinting.
Because transgenes composed entirely of Igf2r DMD2 sequences are not imprinted (32), pBR/RSV (DMD) sequences were tested for their ability to generate imprinted transgenes. Previously, two pBR/RSV transgenic lines were created, and the methylation patterns of maternal and paternal transgene alleles were compared (9). In these two lines, the pBR/RSV transgene was highly methylated on both the maternal and paternal alleles and, therefore, was not imprinted. To extend these findings, the EcoRI pBR/RSV fragment from the RSVIgmyc transgene was used to create two new transgenic lines, and DNAs from transgenic carriers of these additional lines were digested with HpaII, HhaI, or BstUI. For each of the three enzymes, the Southern blots of the maternal and paternal alleles showed identical patterns of intensely hybridizing, high-molecular-weight bands, indicating that both maternal and paternal alleles are equally methylated (Table (Table11 and Fig. Fig.2E).2E). The data presented above show that the pBR/RSV region of RSVIgmyc (the DMD) alone is not imprinted.
The imprinted RSVIgmyc transgene was generated by addition of nonimprinted pBR/RSV sequences to the nonimprinted Ig/myc genomic fragment. We tested the ability of other sequences, particularly those found in the DMDs of imprinted genes, to imprint Ig/myc. The Igf2rIgmyc transgene was created by completely replacing the pBR/RSV region of the RSVIgmyc transgene with a 668-bp fragment from the center of the DMD2 region of the Igf2r gene (Fig. (Fig.33 and and4A).4A). Three Igf2rIgmyc transgenic lines were created, and maternal and paternal methylation patterns were compared by using Southern blot analysis of HpaII-digested DNA hybridized with a probe to the IgA region of the transgene. As shown in Fig. Fig.4B,4B, the maternal Igf2rIgmyc allele is highly methylated, whereas the paternal allele is relatively undermethylated, consistent with an imprinted Igf2rIgmyc locus. Assuming c-myc CpG island sequences are unmethylated on both parental Igf2rIgmyc alleles, as seen with RSVIgmyc transgenes (16), these Southern blots indicate maternal Igf2r DMD2 sequences are methylated and paternal DMD2 sequences are unmethylated. Similar results were obtained from two additional Igf2rIgmyc lines (Table (Table1).1). These data demonstrate that Igf2r DMD2 sequences are able to restore imprinting to the Ig/myc transgene.
The DMD2 fragment in Igf2rIgmyc contains a series of repetitive sequences, organized as three short tandem repeats, each approximately 30 bp in length, and approximately three longer tandem repeats, each approximately 175 bp in length (Fig. (Fig.4A).4A). The short repeat is designated TR1, and the long repeat is designated TR2+3 (20). To determine if the TR1 repeats alone or the TR2+3 repeats alone could restore imprinting the TR1Igmyc and TR2+3Igmyc transgenes were constructed, which completely replace the pBR/RSV region of RSVIgmyc with TR1 and TR2+3 sequences, respectively. TR1Igmyc and TR2+3Igmyc transgenic lines were created, and methylation patterns were analyzed by Southern blot analysis of transgenic carrier DNA (Fig. (Fig.4B).4B). The maternal allele of TR2+3Igmyc was highly methylated, and the paternal allele was undermethylated. Conversely, both parental alleles of the TR1Igmyc transgene were highly methylated. Thus, the TR2+3Igmyc transgene retained the imprinted phenotype of the Igf2rIgmyc transgene, whereas the TR1Igmyc transgene did not.
All of the transgenes described above were analyzed in an inbred FVB/N background. In previous studies, we showed that the RSVIgmyc transgene is imprinted in an FVB/N background. However, when it was analyzed in an inbred C57BL/6 background, its imprinting was lost because the paternal allele became heavily methylated in the embryo, obtaining a level of methylation similar to that of the maternal allele. This embryonic methylation is most likely due to de novo methylation of RSVIgmyc's DMD in C57BL/6 embryos (9). To test whether the Igf2rIgmyc transgene was similarly affected, carriers of the transgene were crossed for three generations to inbred C57BL/6 mice, and the methylation patterns of maternal and paternal Igf2rIgmyc alleles were compared on a Southern blot of HpaII-digested transgenic carrier DNA. In a C57BL/6 background, the Igf2rIgmyc transgene maintains its imprinting; the maternal allele has a high level of methylation, and the paternal allele has a low level of methylation (Fig. (Fig.4C).4C). Importantly, the paternal Igf2rIgmyc allele's methylation in a C57BL/6 background is indistinguishable from that of the paternal allele in an inbred FVB/N background. A similar effect has been observed at endogenous imprinted loci, where imprinting is not dependent on strain background. Thus, in contrast to RSVIgmyc's DMD sequences, the Igf2r DMD2 sequences of Igf2rIgmyc do not become methylated in the embryo.
The Dnmt1Δ1o mutation eliminates expression of the oocyte-specific form of the Dnmt1 methyltransferase and results in loss of methylation from one-half of normally imprinted alleles in offspring of homozygous mutant females (15). These experiments examined the methylation of a number of imprinted genes; however, Igf2r was not among them. If one-half of Igf2r maternal alleles also lose their methylation in the absence of the Dnmt1o protein, we would expect that some of Igf2rIgmyc's normally methylated maternal alleles would be poorly methylated in offspring that inherited Igf2rIgmyc from homozygous Dnmt1Δ1o mutant females. To test this hypothesis, inbred FVB/N females homozygous for the Dnmt1Δ1o allele and carrying the Igf2rIgmyc transgene were mated to wild-type FVB/N males and Igf2rIgmyc methylation patterns examined in offspring at day 10.5 of embryogenesis (D10.5). As shown in Fig. Fig.5,5, the methylation patterns were quite variable among the transgenic embryos. Some transgenic embryos had methylation patterns that were a combination of the heavily methylated maternal pattern and the poorly methylated paternal pattern found on Igf2rIgmyc alleles that were maintained on a wild-type Dnmt1 background. A few embryos had a heavily methylated maternal pattern, and a few others had a poorly methylated paternal pattern. Methylation patterns of control D10.5 transgenic embryos obtained from a mating between an Igf2rIgmyc female and a wild-type FVB/N male are also shown. Except for one embryo, the maternal transgenes among control D10.5 embryos are all heavily methylated. The exceptional embryo had a poorly methylated paternal-like pattern. We conclude from this analysis that much of the maternally acquired methylation of the Igf2rIgmyc transgene is lost when the transgene is inherited from a homozygous Dnmt1Δ1o female. This could be due to a requirement of the Dnmt1o protein for establishing Igf2rIgmyc methylation in the oocyte or for maintaining oocyte-specific methylation during preimplantation development.
The creation of an imprinted transgene by addition of Igf2r DMD2 sequences to Ig/myc would suggest that different imprinted transgenes might be created by the combination of Ig/myc and other DMD sequences. This was tested with sequences from the DMDs of the imprinted H19 and Snrpn genes. Three different transgene constructs were made: H19SIgmyc, H19LIgmyc, and SnrpnIgmyc (Fig. (Fig.3).3). H19SIgmyc contains 292 bp of H19 DMD sequences, including a single CTCF binding site located approximately 4 kb 5′ of the site of transcription initiation (3, 14). H19LIgmyc contains 2,232 bp of the H19 DMD, located approximately 2 to 4 kb 5′ of the transcription start site. SnrpnIgmyc contains 529 bp of the Snrpn DMD, including the entire first exon, but excluding any direct repeats found in the first intron (6, 13, 24). Several transgenic lines of each construct were generated (Table (Table1).1). For each line of a particular transgene construct, the maternal and paternal methylation patterns were identical (Fig. (Fig.6).6). As well, all independently derived lines of a particular construct showed nearly identical levels of methylation (data not shown). In the case of the H19SIgmyc and the H19LIgmyc transgenes, both the maternal and paternal alleles had a high level of methylation. In contrast, both the maternal and paternal alleles of the SnrpnIgmyc transgene showed a low level of methylation. From this analysis, we conclude that not all DMD sequences are able to restore Ig/myc transgene imprinting, either because these DMD sequences are unable to establish gametic methylation patterns or because they do not maintain their gametic methylation patterns during embryogenesis.
In addition to the generation of a variety of transgenes containing DMD and Ig/myc sequences, we also produced a hybrid transgene by combining Ig/myc with sequences from an IAP LTR (Fig. (Fig.3).3). IAP LTR sequences are heavily methylated during much of mouse development (31), and the transcription of some genes in which an IAP element has been inserted may be influenced by parental origin (18, 19). IAPIgmyc transgenic lines were established, and the maternal and paternal methylation patterns within each line were then compared. For each of the four lines analyzed, the two parental alleles were highly methylated and indistinguishable, indicating that Ig/myc imprinting was not restored by the addition of IAP LTR sequences.
The imprinted Igf2rIgmyc and TR2+3Igmyc transgenes, both variants of the imprinted RSVIgmyc transgene, are portable imprinted mouse loci. That is, they are comprised solely of endogenous mouse sequences and are imprinted at all of the integration sites examined. As is the case with endogenous imprinted genes, the imprinting of these transgenes is disrupted by a lack of Dnmt1o protein during preimplantation development (15). These portable imprinted transgene loci are tandem arrays of unit copies 15 kb in size, with all of the required imprinting sequences contained in a unit copy. Moreover, their imprinting is absolutely dependent on the presence of Igf2r intronic DMD2 sequences, which are required for endogenous Igf2r imprinting (32, 33). Because only some sequences from imprinted genes restore Ig/myc imprinting, hybrid transgenes such as Igf2rIgmyc and TR2+3Igmyc provide a means of identifying the essential sequences for imprinting genes.
The consistent imprinting of the portable Igf2r- and Ig/myc-containing transgenes contrasts with the absence or low frequency of imprinting among many of the transgenes primarily derived from endogenous Igf2r sequences. For instance, all transgenes of less than or equal to 14 kb that contain the Igf2r DMD2 and contiguous surrounding genomic sequences are not imprinted (32). Moreover, hybrid Igf2r transgenes that incorporate DMD2 and Aprt gene sequences are infrequently imprinted (25). Only much larger Igf2r transgenes, on the order of 300 kb in size and containing DMD2, are consistently imprinted. Taken together, these observations suggest two possible explanations for the location of cis-acting imprinting signals. Besides DMD2 sequences, either a sequence remote from DMD2 is also needed for Igf2r imprinting, or a combination of genomic sequences neighboring DMD2 is also needed. Studies of the RSVIgmyc transgene favor the latter explanation; there is no single element within Ig/myc that is absolutely required for RSVIgmyc imprinting, and many different combinations of sequence elements within Ig/myc support RSVIgmyc imprinting (9, 16).
Different genomic DNA sequence features are associated with imprinted genes, including CpG islands and clustered direct repeats of short unit lengths (11, 13, 20). In the case of the imprinted Igf2r and Snrpn genes, these two features coincide with the genes' DMDs (Fig. (Fig.7A).7A). For each gene, direct repeats of a unit length between 30 and 175 nt are interspersed with unique DNA of the DMD. Remarkably, although there are no apparent sequence similarities between the Snrpn and Igf2r DMDs, the numbers and relative positions of the repeats within the two DMDs are nearly identical (Fig. (Fig.7A).7A). Either the RSV sequence element alone or the pBR322 element alone can function as RSVIgmyc's DMD (9). Notably, both of these elements contain copies of short tandem repeats (Fig. (Fig.7A7A).
Only the TR2+3 repeat sequences within the Igf2r DMD2 are required to restore Ig/myc imprinting (Fig. (Fig.44 and and7A).7A). These sequences are not contained within the Air promoter (17) and are distinct from the imprinting box defined by Birger et al. (5). In contrast, unique sequences within the 5′ end of the Snrpn DMD and outside of the repeat-containing portion of the DMD are not able to restore Ig/myc imprinting (Fig. (Fig.66 and and7A),7A), and when these sequences are deleted from the endogenous Snrpn locus, Snrpn remains imprinted (8). These observations strongly suggest that the DMD2 repeated sequences are required for imprinting the endogenous Igf2r gene. Moreover, these observations suggest that Ig/myc transgene imprinting may be restored by addition of Snrpn DMD repeated sequences and that these sequences may be important for imprinting the endogenous Snrpn gene.
At most, two copies of a repeat (from Igf2r or the RSV LTR) are required to restore transgene imprinting. If one or two unit copies of a repeat are the minimum requirement to restore Ig/myc imprinting, why does the number of repeats within the Igf2r DMD2 far exceed two? Perhaps the answer to this question lies in the analysis of Igf2r-containing hybrid transgenes. The efficiency of the establishment and/or maintenance of a maternal methylation imprint may rely on a number of features, including the length of a unit repeat, the number of repeats, and the organization of the repeats within the DMD. Three copies of the short Igf2r TR1 repeat may not have restored imprinting to Ig/myc, because the TR1 repeats do not ensure the efficient establishment or maintenance of a maternal methylation imprint. Two complete copies of the longer TR2+3 repeat (in both TR2+3Igmyc and Igf2rIgmyc transgenic mice) can more efficiently establish and/or maintain a maternal methylation imprint. For example, only an occasional Igf2rIgmyc carrier fails to establish or maintain a maternal methylation pattern (Fig. (Fig.5).5). Therefore, the entire eight repeats of DMD2, interspersed with unique DMD2 sequences, may be necessary to ensure absolute imprinting of Igf2r in every oocyte.
DMDs are necessary, but not sufficient, cis-acting elements for imprinting a gene. Only when a DMD is placed in an appropriate genomic context is the absolute requirement of DMD sequences revealed (4, 29, 33). These observations suggest that there are two fundamental cis-acting requirements for imprinting a gene. One of these requirements is furnished by sequences within a DMD, and the other requirement is furnished by the surrounding genome. Because the DMD is the site of allele-specific methylation, a likely role for a DMD is to interact with the molecular machinery that establishes methylation patterns (Fig. (Fig.7C).7C). This interaction could result in methylation of the DMD itself in both germ lines or, alternatively, could prevent its methylation in both germ lines. There is a precedent for tandem repeats from nonimprinted sources to become methylated; two copies of a mouse B1 element are much more efficient at attracting de novo methylation than a single copy (35). Repeats of DMDs may function similarly and become methylated by attracting de novo methylation.
Regardless of whether DMD sequences attract or prevent de novo methylation, the role of the neighboring genome would be to inhibit this effect in just one of the parental germ lines. In the case of RSVIgmyc, Igf2rIgmyc, TR2+3Igmyc, and many of the known imprinted genes, this mechanism would result in a higher level of methylation on the maternal allele. In the case of the H19SIgmyc and H19LIgmyc transgenes, which contain H19 DMD sequences that are normally methylated only on the paternal allele, the high level of transgene methylation is not influenced by the Ig/myc sequences and the transgenes are not imprinted. This observation may indicate a fundamental difference in the mechanism of imprinting for those genes that acquire a heritable methylation pattern on DMD sequences during spermatogenesis.
We thank J. Trasler and E. Michaud for their help and recommendations.
This work was supported by a grant from the NIH (J.R.C.).