|Home | About | Journals | Submit | Contact Us | Français|
In mammals, DNA is methylated at cytosines within CpG dinucleotides. Properly regulated methylation is crucial for normal development1,2. Inappropriate methylation may contribute to tumorigenesis by silencing tumor-suppressor genes3-10 or by activating growth-stimulating genes11-13. Although many genes have been identified that acquire methylation and whose expression is methylation-sensitive14,15, little is known about how DNA methylation is controlled16. We have identified a DNA sequence that regulates establishment of DNA methylation in the male germ line at Rasgrf1. In mice, the imprinted Rasgrf1 locus is methylated on the paternal allele within a differentially methylated domain (DMD) 30 kbp 5′ of the promoter. Expression is exclusively from the paternal allele in neonatal brain17. Methylation is regulated by a repeated sequence, consisting of a 41-mer repeated 40 times, found immediately 3′ of the DMD. This sequence is present in organisms in which Rasgrf1 is imprinted18. In addition, DMD methylation is required for imprinted Rasgrf1 expression. Together the DMD and repeat element constitute a binary switch that regulates imprinting at the locus.
We generated mice in which the Rasgrf1 repeats were replaced by a single loxP site. This allele of Rasgrf1 was designated Rasgrf1tm1Pds (Fig. 1). To evaluate whether the mutation affected methylation of the DMD, we subjected tail DNA from mice inheriting the repeat deletion from their mother or father to Southern-blot analysis using PstI and the methylation-sensitive enzyme NotI. A single NotI site is present in the DMD, and its methylation state in DNAs from all somatic tissues tested is diagnostic of the methylation state of the DMD17. PstI-digested DNA from wildtype mice (Rasgrf1+/+) gave rise to an 8.0-kbp band on Southern blots hybridized to the probe shown in Fig. 2a. The unmethylated maternal fragment was further cleaved by NotI, generating a 2.8-kbp band (Fig. 2a,b) as previously reported17,19. In heterozygous mice inheriting the mutated allele from the mother (Rasgrf1tm1Pds/+), there was no change in the methylation status of either allele. PstI digestion generated a 3.0-kbp fragment from the mutated maternal allele and an 8.0-kbp fragment from the wildtype paternal allele. Digestion with both PstI and NotI further cleaved the 3.0-kbp maternal fragment to 2.8 kbp. In contrast, animals bearing a paternally transmitted repeat deletion (Rasgrf1+/tm1Pds) lost methylation of the paternal allele. PstI digestion of DNA gave rise to an 8.0-kbp wildtype maternal allele and a 3.0-kbp mutated paternal allele band. Both were reduced to 2.8 kbp after NotI digestion, indicating that neither allele was methylated at the NotI site (Fig. 2b). Bisulfite sequencing revealed that loss of methylation was extensive throughout the DMD on the mutant paternal allele (Table 1). This indicated that the Rasgrf1 repeats are needed for paternal allele–specific DNA methylation patterns seen at the locus.
The loss of paternal allele–specific methylation may have resulted from a failure to establish the normal paternal methylation pattern in the male germ line or failure of the established pattern to be maintained in the somatic tissue of the progeny. To distinguish between these possibilities, we assayed the methylation status of the Rasgrf1 DMD in sperm DNA from males heterozygous for the repeat mutation, using three separate methods. The first assay was a Southern blot using the same digests and probes used in Fig. 2. Results showed a substantial reduction in methylation of the mutated allele (Fig. 3a). Previous studies have suggested that although the NotI site methylation state is diagnostic for paternal allele–specific methylation in somatic tissue, the germline methylation imprint may be more accurately reflected by the methylation of the five HhaI sites within the Rasgrf1 DMD19. We therefore assessed the methylation status of these sites in sperm DNA using PCR reactions specific for the mutated and wildtype alleles, with primers that spanned all five HhaI sites. Fig. 3b shows the mutant and wild-type loci and the location of primer pairs used in this analysis. Sperm DNA was digested with HhaI or left undigested, before PCR analysis, using the two allele-specific assays. We carried out a third control PCR reaction using a primer pair that spanned a region of Rasgrf1 lacking HhaI sites. In all DNA samples, the wildtype allele was amplified using primers P1 and P2 (Fig. 3c, upper gel), indicating that the HhaI sites were detectable in the methylated state on the wildtype alleles. When we amplified uncut DNA from sperm of wildtype or heterozygous mice using mutated allele–specific primers P3 and P4, a band of the predicted size was generated only in the reactions containing heterozygous DNA, indicating that the product was indeed allele-specific. However, the band was substantially diminished in reactions containing sperm DNA from heterozygotes that had been pre-digested with HhaI. This indicated that one or more of the five HhaI sites were hypomethylated at a high frequency on the mutated allele. This result was seen regardless of whether the mutated allele was paternally or maternally transmitted (data not shown for maternal transmission). An internal control PCR reaction using primers P5 and P6 showed that band loss was not due to a failure in the amplification reaction. We carried out bisulfite sequencing to evaluate whether the loss of methylation affected just one or more of the five HhaI sites and to determine whether CpGs other than those in the HhaI and NotI sites were affected. The results indicated that loss of methylation on the mutated allele of sperm DNA affected CpGs across the entire DMD and not just those in the NotI and HhaI sites (Table 1). In addition, the disruption of DMD methylation was comparable in both somatic and sperm DNA templates. These three independent means of analysis showed that loss of methylation caused by the repeat deletion was a common event affecting many CpG residues within the DMD at a time when methylation of the locus was normally established in the male germ line.
Rasgrf1 is expressed predominantly from the paternal allele in many tissues; however, expression in neonatal brain is exclusively from the paternal allele17. To determine whether the Rasgrf1tm1Pds mutation had any effect on imprinted expression of Rasgrf1, we analyzed RNA from the brain of neonatal progeny of mutant animals crossed with wildtype C57BL/6 or PWK mates. The cross facilitated separate expression analysis of the two parental Rasgrf1 alleles because of an exonic polymorphism that exists between 129S4/SvJae, the strain in which the mutation was made, and PWK17. Maternal transmission of the repeat deletion did not affect the normal expression pattern of the locus—expression was still from the paternal allele only on all strain backgrounds in amounts comparable to those seen in wild-type mice (data not shown). However, paternal transmission of the Rasgrf1tm1Pds allele resulted in substantial silencing of the normally active allele in heterozygous progeny of C57BL/6 mothers (Fig. 4, +B6/−). Notably, the silencing of the mutated paternal allele was variable in progeny of PWK mothers (Fig. 4, +PWK/−). In those animals, Rasgrf1 expression levels ranged from those seen in +B6/− mice (fourth panel from the top) to amounts seen in +B6/+ and +PWK/+ animals (bottom panel). In +PWK/− mice with wildtype levels of expression, the DMD showed extensive methylation despite the presence of a paternally transmitted Rasgrf1tm1Pds allele, whereas in animals with substantially silenced Rasgrf1 expression the paternal allele was unmethylated (Fig. 4, right panel). This indicated that it is the methylation state of the DMD rather than the presence of the repeats that has the most direct influence on expression of the locus, ruling out the trivial explanation that the repeats contain a promoter or enhancer element needed for Rasgrf1 expression. Moreover, the Rasgrf1 DMD behaves like a methylation-sensitive enhancer-blocking element, as is the case for the DMD at H19/Igf2 (refs 12,13,20; B.J. Yoon et al., unpublished data). The variable methylation seen among progeny of PWK mothers also suggested that there are strain-specific modifiers of the methylation phenotype. Analysis of the nine +PWK/− progeny revealed three animals with residual methylation on the mutated paternal allele (data not shown). Methylation analysis of a much larger number of mice on a mixed C57BL/6 ×129S4/SvJae background revealed no such variability in methylation of the mutated paternal allele. Mice on this background were used in the experiments reported in Fig. 3.
The presence of transgene repeats has been correlated with changes in local DNA methylation and gene expression in diverse systems21-23. The repetitive nature of the Rasgrf1 repeats, the presence of 39 CpGs within the 1,611-nt repeat region, or other structural or sequence motifs may be required for the Rasgrf1 repeats to regulate establishment of DNA methylation in the male germ line. These motifs may serve as sites of interaction with collaborating trans-acting factors that themselves dictate local methylation patterns. The variable loss of methylation and correlation between methylation and expression seen in mutant progeny of PWK mothers was reminiscent of previous studies of H19 imprinting24 and may be due to strain-specific differences in loci encoding such factors. Expression of these factors must be specific to the male or female germ line for Rasgrf1 and other imprinted loci to undergo allele-specific methylation.
Studies of methylation regulation at other loci have been reported. The maternally-expressed H19/Igf2 locus contains a DMD that is hypermethylated at CpGs within and adjacent to the promoter on the paternal allele25. A targeted deletion of the DMD abolished paternal allele–specific DNA methylation and Igf2 silencing when transmitted paternally26. As is the case for the Rasgrf1 DMD, the H19 DMD is an enhancer blocker (refs 12,13; B.J. Yoon et al., unpublished). However, in contrast with the Rasgrf1 repeat mutation, which caused loss of methylation establishment, the H19 mutation interfered with the maintenance or persistence of methylation patterns after their normal establishment. Although it is not known how methylation is established at the H19 locus, repeats seem not to be required. A 461-nt G-rich repetitive element at H19 was shown to be dispensable for establishment of paternal-allele methylation of an H19 transgene27. It is possible that regulation of transgene methylation differs from methylation of endogenous loci, that redundant mechanisms regulate methylation at H19 or that mechanisms regulating methylation at imprinted loci vary. These possibilities are not mutually exclusive.
At the Igf2r locus, a 3-kbp intronic sequence regulates methylation of the maternal allele before fertilization28. Transient transgenesis experiments identified a 6-nt allele-discrimination sequence (ADS) that protects the paternal allele from methylation and an 8-nt de novo methylation sequence (DNS) that facilitates methylation of the unprotected maternal allele. The ADS and DNS sequences interact with factors present in androgenetic and gynogenetic ES cells, respectively16. Dot-plot analysis shows that the Igf2r imprinting control region contains some repeated elements; however, it is not clear if these are needed for the ADS or DNS to function. Although a single copy of the 6-nt ADS is found within the Rasgrf1 repeats, it is not likely to be functionally significant, as in Igf2r transgenes the ADS sequence protects the paternal allele from methylation. Because the paternal allele is methylated at Rasgrf1, the ADS cannot be protecting the locus from methylation.
We prepared the Rasgrf1-targeting construct (pBJR3) which replaced a 2-kbp repeat–containing fragment with a PGKneor cassette flanked by loxP sites as follows. The 5′ vector arm consisted of a 2-kbp fragment immediately 5′ of the repeat that we PCR-amplified using Rasgrf1 reverse and forward primers derived from the plasmid pSPL3, which contained the 5′ Rasgrf1 fragment. The reverse primer modified four nucleotides of the wildtype Rasgrf1 sequence to generate new PstI and HindIII sites immediately 5′ to the repeat. We cloned the PstI fragment from the amplification into the PstI site of pBluescriptII (Stratagene). The 5′ arm, isolated as a SacII-HindIII fragment from the resulting clone, was inserted into the SacII and HindIII sites of pTK-NEOF which carried the loxP-neor-loxP and HSVTK markers (D. Lam and P. Aplan, National Cancer Institute). We inserted the 3-kb homologous arm, consisting of an EcoRV-SmaI fragment located 3′ of the repeats, into the SalI site of pTKNEOF to generate the pBJR3 targeting vector. We carried out ES cell culture and blastocyst microinjections according to the standard protocols. We excised the neor selectable marker by crossing germline progeny with Zp3Cre transgenic mice29.
We prepared sperm DNA from 8-wk or older heterozygous male mice inheriting a mutated allele from their fathers, mothers or wildtype control littermates by allowing sperm to swim out from the cauda epididymis into 500 μl phosphate-buffered saline (PBS). We added SDS (1.5 μl of 10% w/v) and incubated the sperm at room temperature for 15 min. Cells were centrifuged (3 min, 16,000 × g), resuspended in 100 μl PBS, diluted with an additional 900 μl PBS and 100 μl β-mercaptoethanol (1 M), and then incubated at 37 °C for 1 h. Cells were centrifuged (3 min, 16,000 × g), resuspended in 100 μl PBS, centrifuged again (5 min, 16,000 × g), resuspended in 200 ml PBS, diluted with an equal volume of 2 × lysis buffer (200 mM Tris pH 8.5, 200 mM EDTA, 400 mM NaCl, 2% SDS) and proteinase K added (400 μg/ml−1), followed by an overnight incubation at 55 °C with shaking. We then isolated DNA by phenol chloroform extraction and ethanol precipitation.
We carried out assays for NotI site methylation using Southern-blot hybridization as described17 and assays for HhaI site methylation in sperm as described19, with modifications. Using sperm DNA that was digested with HhaI or left uncut, we carried out PCR amplification. PCR product sizes were as follows: P1−P2, 263 bp; P3−P4, 228 bp; P5−P6, 148 bp.
We carried out bisulfite sequencing as described30 with some modifications, using freshly prepared solutions. Briefly, 1.0 μg of sheared DNA in 50 μl water was denatured by addition of 5 μl fresh 3 M NaOH, then incubated at 37 °C for 10 min. After adding 30 μl 10 mM hydroquinone and 520 μl 3 M sodium bisulfite, we allowed deamination to proceed at 50 °C for 16 h. We purified treated DNA using a Qiaquick gel extraction kit according to the manufacturer's instructions (Qiagen) and eluted DNA from the kit column in 50 μl elution buffer. The bisulfite reaction was completed by the addition of 5 μl fresh 3 M NaOH followed by a 5-min incubation at RT. We purified DNA again using a Qiaquick gel extraction kit and eluted it in 30 μl. For PCR, we amplified 1 μl of DNA (primers available upon request). We gel-purified all PCR products, cloned them using a TA or TopoII cloning kit (Invitrogen) and then sequenced individual clones.
We measured expression of Rasgrf1 using a quantitative PCR assay. We perpared cDNA using oligo(dT) primers and 0.5 μg total RNA. A fraction of this material (10%) was serially-diluted and used for PCR. A reaction mixture containing four primer pairs was prepared and added to each cDNA dilution. One set of the primers amplified a 378-bp product that spanned a Rasgrf1 intron–exon boundary, and another set amplified a 128-bp product specific for Rpl32 which served as an internal control.
Available upon request.
This work was made possible through grants from the NIH and the Roswell Park Alliance to P.D.S., C.P. and to the Roswell Park Cancer Institute. The authors dedicate this work to the memory of V. Chapman.