|Home | About | Journals | Submit | Contact Us | Français|
A subset of mammalian genes exhibits genomic imprinting, whereby one parental allele is preferentially expressed. Differential DNA methylation at imprinted loci serves both to mark the parental origin of the alleles and to regulate their expression. In mouse, the imprinted gene Rasgrf1 is associated with a paternally methylated imprinting control region which functions as an enhancer blocker in its unmethylated state. Because Rasgrf1 is imprinted in a tissue-specific manner, we investigated the methylation pattern in monoallelic and biallelic tissues to determine if methylation of this region is required for both imprinted and non-imprinted expression. Our analysis indicates that DNA methylation is restricted to the paternal allele in both monoallelic and biallelic tissues of somatic and extraembryonic lineages. Therefore, methylation serves to mark the paternal Rasgrf1 allele throughout development, but additional factors are required for appropriate tissue-specific regulation of expression at this locus.
Genomic imprinting is a mammalian-specific phenomenon resulting in the monoallelic expression of a subset of genes. To date, approximately 100 imprinted genes have been identified in mouse.1 In order to effect imprinted gene expression, the cellular machinery requires an imprinting mark to distinguish between the parental alleles and regulate their expression accordingly. Such a mark must be maintained in somatic lineages of an individual, yet be reset in the germline in order to appropriately reflect parental origin following fertilization. Epigenetic factors, such as DNA methylation and histone modification, can be both heritable and reversible; thereby, these factors are candidates for imprinting marks. Indeed, all known imprinting control regions (ICRs) contain differentially methylated regions (DMRs) that function both to mark the parental origin of the allele and to regulate imprinted expression.2 Parental allele-specific DNA methylation is inherited directly from the gamete for many imprinted loci.3–7 In the developing germline, allele-specific DNA methylation is erased, and then reset to reflect the sex of the parent prior to the development of the mature gamete.8–11 Establishment of gametic methylation at imprinted loci relies on the activity of the DNA methyltransferase enzymes Dnmt3a and Dnmt3L.12–14 Maintenance of gametic methylation during preimplantation development is achieved via the activity of Dnmt1.15 Failure to properly establish and/or maintain differential methylation at an ICR results in abnormal expression of its associated genes and can cause developmental disorders and disease.12–14, 16, 17
While DNA methylation is frequently associated with gene silencing, such as at the imprinted genes H19 and Snrpn,3, 6 there are multiple examples of imprinted loci for which the expressed allele is methylated.18–21 In some cases, methylation of the ICR controls the expression of a non-coding RNA whose expression, in turn, results in transcriptional silencing of adjacent genes. For example, Region 2 of the Igf2r gene is methylated on the expressed maternal allele.19 This maternal allele-specific methylation directly regulates expression of an antisense non-coding RNA, Air, whose promoter is contained within the ICR. Expression of Air from the unmethylated paternal allele silences Igf2r in cis; conversely, Air is repressed on the methylated maternal allele, allowing expression of Igf2r.22, 23 In other cases, the ICR functions as an insulator, binding enhancer-blocking proteins in a methylation-sensitive manner and resulting in parental allele-specific expression. For example, the ICR located between H19 and Igf2 is occupied by the enhancer blocking protein CTCF on the unmethylated maternal allele, preventing interaction between Igf2 and its enhancers.24, 25 In contrast, the methylated paternal ICR fails to bind CTCF, allowing Igf2 to be expressed. The various mechanisms by which ICRs function illustrate that their role is complex and that the regulation of imprinted gene clusters requires both local and global modification of chromatin.
Rasgrf1 is a paternally expressed, paternally methylated imprinted gene located on mouse chromosome 9.18 The Rasgrf1 DMR and an adjacent repetitive region are located 30 kb 5’ relative to the start of transcription (Fig. 1).4 The repetitive region is critical for imprinted expression in brain; a paternally inherited deletion of the repetitive element results in silencing of the normally expressed paternal allele, coincident with a loss of paternal allele-specific methylation at the adjacent DMR.26 The repetitive element is required both for the establishment and the maintenance of methylation on the paternal allele.27 Mice that fail to maintain paternal allele-specific methylation have dramatically reduced expression of Rasgrf1, consistent with the hypothesis that paternal allele-specific methylation directs imprinting.26, 27
Investigation into the relationship between methylation at the DMR and Rasgrf1 expression has been primarily conducted in neonatal brain, where Rasgrf1 is both imprinted and abundantly expressed.18, 26, 27 In other tissues, Rasgrf1 expression is less robust, with expression reported as monoallelic in stomach and heart, biallelic in lung and thymus, and no expression in liver and kidney.18 Existing evidence strongly supports the hypothesis that methylation of the DMR is required for expression of the paternal allele in brain.26, 27 However, it is not known if expression of the maternal allele is also dependent on DMR methylation in cis in tissues exhibiting biallelic expression of Rasgrf1. Therefore, we decided to conduct a comprehensive analysis of expression and methylation in embryonic, extraembryonic and neonatal tissues to determine if expression consistently correlates with in cis methylation of the Rasgrf1 DMR. Our results illustrate that methylation at the DMR is specific to the paternal allele, regardless of the expression state of Rasgrf1. We further show that paternal allele-specific methylation is present at fertilization and is consistently maintained in all tissues throughout development, highlighting its role as a primary mark for the paternal Rasgrf1 allele.
Gametic methylation at imprinted loci has been proposed to serve as a primary imprinting mark which, following fertilization, is maintained throughout development.28 The presence of allele-specific methylation in blastocysts supports the hypothesis that differentially methylated regions can serve as imprinting marks.28, 29 Rasgrf1 is one of three known paternally methylated imprinted genes for which methylation is inherited from sperm at the time of fertilization (Fig. 2A).26 To determine if paternal allele-specific methylation is retained in the globally undermethylated blastocyst,29, 30 we used bisulfite mutagenesis followed by DNA sequencing to analyze the methylation status of the 252 bp Rasgrf1 DMR (Fig. 1).4, 26 A strain-specific polymorphism was used to distinguish between the maternal and paternal alleles in DNA isolated from F1 hybrids derived from crosses between C57BL/6J (B6) and mice bearing a portion of Mus musculus castaneus chromosome 9 (CAST9) mice (Fig. 1).
Five independent PCRs were performed on bisulfite mutagenized DNA isolated from 3.5 d.p.c. B6xCAST9 F1 blastocysts to prevent repeated analysis of a single template. Unmethylated maternal alleles and methylated paternal alleles were recovered from three of the five PCRs; in the remaining two PCRs, only methylated paternal alleles were recovered (Fig. 2B). These data illustrate that Rasgrf1 DMR methylation that is inherited from sperm is retained in the blastocyst, suggesting that it may function as the gametic imprinting mark.
We next examined the methylation status of the Rasgrf1 DMR in embryonic and extraembryonic tissues derived from an 8.5 d.p.c. B6xCAST9 F1 hybrid embryo, a stage at which the placenta is relatively free of maternally derived cells. Paternal allele-specific methylation was observed in both the embryo and the placenta (Fig. 2C), illustrating that the imprinting mark is maintained in both of these cell lineages.
We utilized RT-PCR to analyze expression of Rasgrf1 in embryonic and extraembryonic tissue. Rasgrf1 expression was detected in 8.5 d.p.c. placenta and embryo, but not in blastocysts (Fig. 3A). Blastocysts did express p57 (Fig. 3A), indicating that RNA was recovered from these embryos and suggesting that Rasgrf1 is either not expressed or is expressed at very low levels during this stage of development. To determine if differential methylation correlates with imprinted expression in these tissues we developed an allele-specific RT-PCR assay. We identified a SNP between C57BL/6 and Mus musculus castaneus in exon 15 of Rasgrf1 (position 9,561,534; GenBank accession no. NT_039476; C in B6, G in CAST). This SNP eliminates a StyI restriction endonuclease site in sequences derived from the CAST allele. RT-PCR products from F1 hybrid 8.5 d.p.c. B6xCAST9 placenta and embryo were digested with StyI to distinguish between maternal and paternal products. We observed biallelic expression of Rasgrf1 in 8.5 d.p.c. embryos, as evidenced by the presence of maternal (B) and paternal (C) products (Fig. 3B). In contrast, Rasgrf1 was expressed solely from the paternal allele in 8.5 d.p.c. placenta (Fig. 3B), demonstrating that the Rasgrf1 imprinting mark can be recognized in extraembryonic lineages.
Previous reports of Rasgrf1 identified tissue-specific expression profiles; in some tissues, Rasgrf1 expression was restricted to the paternal allele, while in other tissues expression was biallelic or absent.18 As our experiments were conducted in a different genetic background than the one used by Plass and colleagues,18 we wanted to confirm the expression profile in B6xCAST F1 hybrids.
Rasgrf1 expression was assessed in neonatal brain, lung, thymus, liver, kidney and stomach derived from B6xCAST F1 hybrids using RT-PCR. Expression was detected in all six tissues (Fig. 4A). In contrast, Plass et al.18 reported an absence of detectable Rasgrf1 expression in liver and kidney. We believe that the discrepancy between experiments reflects the level of sensitivity in the assay. While Plass et al.18 performed RT-PCR using total cDNA, we used a gene specific primer for the reverse transcription step, which may have heightened the level of sensitivity in our assay. In support of this hypothesis, brain consistently yielded the highest level of expression in our assay, while expression in thymus, liver and kidney was repeatedly low.
To determine if Rasgrf1 is imprinted in these neonatal tissues, we utilized the allele-specific RT-PCR assay described above. We detected monoallelic expression in brain and liver, and biallelic expression in lung, thymus, kidney and stomach (Fig. 4B). These data confirm that imprinted expression of Rasgrf1 is tissue-specific. Our results differ slightly from those reported by Plass et al.18 in that we observed biallelic expression in stomach rather than exclusively monoallelic expression. These findings may be reconciled by the fact that although both alleles are transcribed in stomach, the paternal allele is preferentially expressed. In addition, our results newly identify liver as a tissue with monoallelic expression of Rasgrf1.
Imprinted expression of Rasgrf1 requires a repetitive region, which directs paternal allele-specific methylation of the DMR and paternal allele-specific expression in neonatal brain.26, 27 To determine if expression depends on in cis methylation of the DMR, we examined the methylation status of the Rasgrf1 DMR in tissues with monoallelic and biallelic expression of Rasgrf1 (Fig. 5). Consistent with the hypothesis that methylation at the Rasgrf1 DMR is required for imprinted expression, we observed paternal allele-specific methylation in neonatal brain and liver. However, differential methylation was also observed at the Rasgrf1 DMR in tissues exhibiting biallelic expression: lung, thymus, kidney and stomach. These data indicate that in cis methylation of the Rasgrf1 DMR is not required for Rasgrf1 expression. Furthermore, differential methylation of the Rasgrf1 DMR is not sufficient to direct monoallelic expression of Rasgrf1, since paternal allele-specific expression is observed in tissues with biallelic expression.
The genomic region between the Rasgrf1 DMR and the Rasgrf1 transcriptional start site was examined to determine if additional CpG-rich regions were present that might influence Rasgrf1 expression via allele-specific DNA methylation patterns. We identified a 2.8 kb CpG island flanking the Rasgrf1 transcriptional start site using the CpG Island Searcher program described by Takai and Jones (http://www.uscnorris.com/cpgislands2/cpg.aspx; Fig. 1).31 We identified two polymorphisms between C57BL/6 and Mus musculus castaneus to facilitate our analysis of maternal and paternal alleles, located at +81 and +95, respectively (Fig. 1). Bisulfite mutagenesis followed by DNA sequencing was utilized to analyze a 415 bp region containing 34 CpG dinucleotides in monoallelic brain and biallelic lung and kidney. There was a striking absence of DNA methylation on both the maternal and paternal alleles in all three tissues (Fig. 6), suggesting that methylation of this region does not contribute to the regulation of Rasgrf1 expression.
We have shown that the Rasgrf1 DMR is specifically methylated on the paternal allele in tissues with imprinted and non-imprinted expression of Rasgrf1. Our analysis illustrated that tissues exhibiting monoallelic Rasgrf1 expression (brain and liver) and tissues exhibiting biallelic Rasgrf1 expression (lung, thymus, kidney and stomach) maintain differential methylation at the Rasgrf1 DMR. In addition, paternal allele-specific methylation is maintained throughout embryonic development, in both embryonic and extraembryonic lineages. Together, these data provide support for the hypothesis that the Rasgrf1 DMR functions as the primary imprinting mark for the paternal allele. However, the data also illustrate that differential methylation of this region cannot be the sole factor controlling imprinted expression, as differential methylation persists in tissues with either imprinted or biallelic expression.
The methylation status of the Rasgrf1 DMR directly influences Rasgrf1 expression in brain. The Rasgrf1 DMR contains CTCF binding sites, which have been shown to function as enhancer blockers at other imprinted loci, including H19, Igf2 and KvDMR1.24, 25, 32, 33 CTCF binds to the unmethylated Rasgrf1 DMR, where it functions in cis to silence the maternal allele via enhancer-blocking activity.32 Experimental conditions that prevent methylation of the paternal DMR result in silencing of the normally expressed paternal allele in brain, consistent with the model that the unmethylated DMR binds CTCF, blocking expression of the Rasgrf1 allele in cis.26, 32 Placement of an enhancer between the DMR and Rasgrf1 overrides the enhancer-blocking activity of the CTCF-bound DMR, implicating enhancers located upstream of the DMR in the regulation of Rasgrf1 in brain.32
In tissues with biallelic Rasgrf1 expression, the enhancer-blocking activity of the unmethylated DMR must be bypassed. As the Rasgrf1enhancer(s) have not yet been identified, it is possible that tissue-specific enhancers are located both upstream and downstream of the DMR. In this case, an upstream, brain-specific enhancer might be blocked by CTCF binding on the maternal allele, while a downstream lung-specific enhancer would continue to function. Alternatively, additional epigenetic modifications may be present at the Rasgrf1 locus in biallelic tissues, which effectively prevent CTCF from binding the unmethylated maternal allele.
Histone modifications are likely to play a role in the tissue-specific imprinted expression of Rasgrf1. In chromatin derived from adult liver, a tissue in which Rasgrf1 expression is monoallelic (Fig. 4), the maternal DMR is enriched for histone H3K4 methylation and histone H3 acetylation, while the paternal allele is enriched for histone H4K20 and histone H3K9 methylation.34 Interestingly, the modifications observed on the silent maternal allele are normally associated with transcriptionally active genes, while the modifications observed on the expressed paternal allele are generally associated with transcriptionally silent chromatin.35 The maternal allele is also characterized by an enrichment of histone H3K27 methylation,34 a modification associated with repressed chromatin. The presence of both active (H3K4me2) and repressive (H3K27me3) epigenetic modifications on the maternal Rasgrf1 allele is striking, and is reminiscent of the bivalent chromatin domains described by Bernstein et al.,36 where small regions marked with active chromatin modifications are embedded within large regions bearing repressive modifications in ES cells. These bivalent domains are associated with the transcriptional start sites for genes encoding developmental transcription factors, genes generally expressed at low levels in ES cells. During ES cell differentiation, the epigenetic modifications at these loci are altered, resulting in active or repressive patterns dependent on the locus. Bernstein et al. hypothesize that the bivalent domain functions to maintain pluripotency in ES cells, keeping expression levels low while priming the locus for gene activation. Recent data suggests that bivalent chromatin domains may play a role in the regulation of imprinted gene expression. A bivalent chromatin domain associated with Grb10 has recently been described by Sanz et al.;37 this domain is resolved in brain during neural commitment, the stage at which imprinted expression initiates in brain. Perhaps a bivalent chromatin domain is initially established at Rasgrf1 in order to prime the locus for future high expression levels in differentiated tissue, such as brain. Analysis of chromatin modifications at the Rasgrf1 DMR in monoallelic and biallelic tissues may clarify the mechanism(s) regulating tissue-specific expression patterns at this locus.
Tissue-specific imprinting can be achieved via several different mechanisms, including alternative promoters, chromatin modification and DNA methylation. Alternative promoter usage directs the tissue-specific imprinting of human ZAC/PLAGL1; expression from the differentially methylated P1 promoter is monoallelic, while expression from the unmethylated P2 promoter is biallelic.38 While there is some evidence for alternative promoters at Rasgrf1, DNA methylation does not appear to influence their activity since the CpG island flanking the Rasgrf1 transcriptional start site is unmethylated in monoallelic and biallelic tissues (Fig. 6).18, 39 At other genes displaying tissue-specific imprinting, expression appears to be controlled by a combination of DNA methylation and histone modifications. At the human imprinted NDN locus, allele-specific DNA methylation and histone H3K4 methylation mark the parental alleles in both monoallelic and non-expressing tissues.40 In contrast, histone acetylation was observed solely on the expressed paternal NDN allele in monoallelic tissues, suggesting that this epigenetic modification regulates tissue-specific transcription.40 Mouse Grb10 provides an example of an imprinted gene exhibiting multi-faceted epigenetic regulation. Expression of the paternal allele initiates at an unmethylated promoter solely in brain, although maternal allele-specific methylation of this CpG island is inherited from oocytes and is maintained in all tissues.41–43 In contrast, maternal allele-specific expression is derived from an alternate promoter displaying biallelic hypomethylation, indicating that DNA methylation is insufficient to control the tissue-specific imprinting of Grb10. There is a strong correlation between differential histone acetylation and allele-specific Grb10 expression, implicating chromatin modifications in the regulation of this gene.37, 43 Our results illustrate that, like NDN and Grb10, the tissue-specific regulation of Rasgrf1 expression cannot be explained solely by DNA methylation. Indeed, Lindroth et al.44 recently demonstrated that the presence of DNA methylation and H3K27 methylation are mutually exclusive at the Rasgrf1 DMR in MEFs. Further exploration of chromatin structure is necessary to elucidate the complex tissue-specific regulation of Rasgrf1.
C57BL/6J (B6) and Mus musculus castaneus (CAST) mice were purchased from the Jackson Laboratory (Bar Harbor, ME). Natural matings between B6 females and CAST males were used to generate F1 hybrid tissues for bisulfite and RT-PCR analyses. Neonatal brain, lung, thymus, liver, kidney and stomach were collected from 1–2 day B6xCAST neonatal mice. To facilitate the isolation of F1 hybrid mice, a strain of mice that served as the source of the M. m. castaneus Rasgrf1 allele (CAST9) was constructed. M. m. castaneus males were mated to B6 females. The resulting B/C F1 hybrid females were backcrossed to B6 males. The backcross progeny were screened for the CAST9 allele at microsatellite markers D9Mit50 (49cM), D9Mit275 (50 cM) and D9Mit35 (52 cM); Rasgrf1 is located at 50 cM. CAST9/B females were backcrossed to B6 males, and CAST9/B male and female offspring were identified among the progeny from the second backcross. CAST9/B males and females were mated, and their offspring were screened for male and female progeny homozygous for the CAST9 allele; these animals were interbred to produce the CAST9 strain. Natural matings between B6 females and CAST9 males were used to generate B6xCAST9 blastocysts (collected 3.5 days post coitum as described by Hogan et al.)45 and 8.5 d.p.c. embryos and placentae for bisulfite and RT-PCR analyses.
DNA was isolated and subjected to bisulfite modification, PCR amplification and subcloning as previously described.8 Sequencing reactions were performed using a Thermo Sequenase Cycle Sequencing Kit (USB Corporation, Cleveland, OH), and reactions were analyzed on a 4300 DNA Analyzer (LI-COR Biosciences, Lincoln, NE). All mutagenized DNAs were subjected to at least three independent PCR amplifications to ensure recovery of different strands of DNA. Subclones with identical sequences derived from a single amplification are reported as unique when variant products were derived in the reaction. The following primer pairs were used for semi-nested amplification of the mutagenized Rasgrf1 DMR, with the first nucleotide in RasBF1 corresponding to position 9,446,270 (GenBank accession no. NT_039476): Rasgrf1 DMR, RasBF1/RasBR, followed by RasBF2/RasBR. The following primer pairs were used for semi-nested amplification of the mutagenized Rasgrf1 promoter CpG island, with the first nucleotide in RasProBF1 corresponding to position 9,476,349 (GenBank accession no. NT_039476): Rasgrf1 promoter CpG island, RasProBF1/RasProBR1 followed by RasProBF1/RasProBR2. Primer sequences follow. RasBF1, 5’-GAGAGTATGTAAAGTTAGAGTTGTG-3’; RasBR, 5’-ATAATACAACAACAACAATAACAATC-3’; RasBF2, 5’-GTTAAAGATAGTTTAGATATGGAATTT-3’; RasProBF1, 5’-TTTAGTTAGTTGAGGGAGGG-3’; RasProBR1, 5’-CAAAACAAACCACTTAATTTACC-3’; RasProBR2, 5’-CTCCTCTTACTCAAATAACCTT-3’.
Embryos were disrupted using a pestle and total RNA was isolated using a High Pure RNA Tissue Kit (Roche Diagnostics Corporation, Indianapolis, IN). Reverse transcription was performed using 10% of the isolated RNA in a 20 µl reaction containing SuperScript III (Invitrogen, Carlsbad, CA) and 2 pmol of a gene-specific primer, RasR4, 5’-GAGTCCTGATGATGTTGGCT-3’. One microliter of the RT product was subjected to 20 cycles of amplification (30 seconds at 94°C, 1 minute at 60°C, 1 minute at 72°C) in the presence of 1X PCR Master Mix (Promega, Madison, WI) and 0.5 µM each of primers RasF5, 5'-GGAACTCTTGTTTTCCAGCAGC-3' and RasR5, 5'-GATTTTGGTGTTGGTGGGGAT-3'. The resulting products were diluted 1:1000 (neonatal tissues) or 1:100 (embryonic tissues) in ddH2O and subjected to 30 cycles of amplification in the presence of primers RasRT-F, 5’-CAGCCGCCGTCGGAAGCT-3’ and RasE15R, 5’-GCTGAGAGCTCCCCAGGC-3’. The products were run on a 2% agarose gel to assess expression. A separate aliquot of each product was digested with StyI and resolved on a 20% acrylamide gel to distinguish between the B6 (166 bp + 71 bp) and CAST (237 bp) alleles. The assay for p57 expression was previously described.46
We thank Marisa Bartolomei and Raluca Verona for discussion and critical reading of the manuscript. This work was supported by U.S. Public Health Service Grant HD041444 to T.L.D. and an award from the LI-COR Biosciences Genomics Education Matching Fund Program to T.L.D.