Patterns of Hybrid Loss of Imprinting Reveal Tissue- and Cluster-Specific Regulation

Background Crosses between natural populations of two species of deer mice, Peromyscus maniculatus (BW), and P. polionotus (PO), produce parent-of-origin effects on growth and development. BW females mated to PO males (bw×po) produce growth-retarded but otherwise healthy offspring. In contrast, PO females mated to BW males (PO×BW) produce overgrown and severely defective offspring. The hybrid phenotypes are pronounced in the placenta and include PO×BW conceptuses which lack embryonic structures. Evidence to date links variation in control of genomic imprinting with the hybrid defects, particularly in the PO×BW offspring. Establishment of genomic imprinting is typically mediated by gametic DNA methylation at sites known as gDMRs. However, imprinted gene clusters vary in their regulation by gDMR sequences. Methodology/Principal Findings Here we further assess imprinted gene expression and DNA methylation at different cluster types in order to discern patterns. These data reveal PO×BW misexpression at the Kcnq1ot1 and Peg3 clusters, both of which lose ICR methylation in placental tissues. In contrast, some embryonic transcripts (Peg10, Kcnq1ot1) reactivated the silenced allele with little or no loss of DNA methylation. Hybrid brains also display different patterns of imprinting perturbations. Several cluster pairs thought to use analogous regulatory mechanisms are differentially affected in the hybrids. Conclusions/Significance These data reinforce the hypothesis that placental and somatic gene regulation differs significantly, as does that between imprinted gene clusters and between species. That such epigenetic regulatory variation exists in recently diverged species suggests a role in reproductive isolation, and that this variation is likely to be adaptive.


Introduction
Imprinted genes display allele-specific silencing based on parental origin. This phenomenon results in classes of genes with biases in expression of paternally-derived alleles as well as those preferentially transcribing maternally-derived alleles. These loci represent many gene families, and their products are involved in a variety of processes [1]. Misexpression of imprinted genes is associated with many diseases including numerous tumor types, growth dysplasias, neurological conditions, and several pregnancy-associated disorders [2][3][4][5][6][7][8]. Imprinted loci are found clustered in relatively discreet regions of mammalian genomes, implying common regulatory elements [9]. The allelic silencing of imprinted loci requires the establishment and subsequent erasure of germline-specific epigenetic marks (e.g. such that a paternally-derived allele may become maternally-derived in the following generation) [10].
The best-characterized of these gametic ''imprints'' are dense regions of methylated cytosine residues typically lying between imprinted loci and/or at promoters [11]. These regions are known as germline differentially methylated regions (gDMRs) [12][13][14], and survive the wave of demethylation that occurs during preimplantation mammalian development [15][16][17]. Germline DMRs are thought to be the primary imprint control regions (ICRs) for their associated domain. This is particularly true for those regions acquiring DNA methylation during spermatogenesis, as histones and their associated modifications are replaced by protamines [18]. Targeted deletions of gDMRs typically perturb imprinting status at the associated domain, usually dependent on which parent passes the targeted allele [19][20][21].
An increasing body of evidence indicates that these imprinting regulatory mechanisms are tissue-specific in multiple mammalian species [22][23][24][25]. These differences are particularly pronounced in comparisons of extra-embryonic vs. fetal tissues. Both individual imprinted genes and entire clusters display placenta-specific patterns of imprinting regulation [24][25][26][27][28]. Accordingly, imprinted loci have been shown to play major roles in placental development [26].
We have uncovered a naturally occurring animal model that mimics several aspects of imprinted gene associated disorders [27]. Hybrids between two recently diverged North American deer mice (genus Peromyscus) display asymmetric effects on growth and development. Female prairie deer (P. maniculatus bairdii; captive stock = BW) mated with male oldfield mice (P. polionotus; captive stock = PO) produce growth retarded offspring [28,29]. Placentas produced by this cross (designated bw6po) are particularly affected, weighing , half that of the parental strains [30][31][32].
Conversely, PO females mated to BW males produce dysmorphic overgrowth of placental and fetal tissues [30,31,33,34], and display multiple defects reminiscent of imprinted gene disorders [27]. Between 10 and 15 percent of these PO6BW conceptuses lack visible embryonic structures, typically resembling placentas and associated membranes [27]. The majority of these PO6BW conceptuses are dead by late gestation; those that survive to parturition kill the mother due to inability to pass through the birth canal [27,31].
Studies to date show perturbations of allelic usage and levels of imprinted gene expression are affected in the hybrids. The bw6po hybrids display only minor perturbations, largely confined to extra-embryonic tissues [27,35,36]. In contrast, the PO6BW offspring exhibit loss-of-imprinting (LOI) and/or significantly altered expression levels at most, but not all loci tested [27,[34][35][36]. The PO6BW LOI is mediated by a maternal effect [37]. To date, hybrid DNA methylation has only been examined at the H19 locus, where loss was associated with bi-allelic expression. Here we 1. Further assess DNA methylation-imprinted gene expression correlations, 2.Examine imprinting at other clusters, 3. Assess imprinting patterns in placental, embryonic and CNS to discern potential patterns in the cluster/gene types affected.

Results
The results are grouped by genomic region or tissue-type. We assessed at least four samples for each tissue/genotype combination in the allelic expression and at least two samples in each of the DNA methylation assays. Samples used in the latter assays are a subset of those used in the expression analyses. In the bisulfite sequence analyses, we did not include sequence reads with significant numbers of unconverted non-CpG cytosines (i.e. suggesting an incomplete reaction). See the methods section for locus-specific details.

Hybrid misregulation of the Kcnq1ot1 cluster
The Kcnq1ot1 (formerly Lit1) domain contains a number of maternally-expressed genes in addition to the paternally-expressed Kcnq1ot1 transcript. The single known gDMR in this region is maternally methylated across the Kcnq1ot1 promoter and associated with its repression on that allele [38]. We have previously shown reduced expression of two linked maternally expressed genes Cdknc1 and Phlda2 (aka Ipl, Tssc3), in the PO6BW hybrids. To date, we have been unable to identify an expressed polymorphism in these genes to assess allelic expression.
We were able to test imprinting status of another linked maternally expressed gene, Cd81(Tapa1), whose imprinting is limited to the placenta in house mice (Mus) [39]. The Cd81 paternal allele is visible in bw6po placentas, but is clearly biased in favor of the maternal allele (Fig. 1A). Expression appears to be fully bi-allelic in the PO6BW placentas. While the Cd81 data is consistent with the broad PO6BW LOI, this gene is not imprinted in cattle [40], raising the possibility that it may only be weakly imprinted in Peromyscus.
To determine if reactivation of the maternal Kcnq1ot1 allele is associated with the misexpression in this cluster, we devised an allelic usage assay in the first (59) kb of this long (,50 kb) transcript [38]. Results of this assay indicate that Kcnq1ot1 expression is biallelic in both PO6BW embryonic and placental tissues (Fig. 1A). In contrast, Kcnq1ot1 displays strict paternal expression in bw6po tissues. These data are consistent with a model in which activation of the PO6BW maternal Kcnq1ot1 transcript reduces Cdkn1c and Phlda2 expression (as we have previously shown [27]). This suggests that the Cd81 biallelic usage is more likely a lowering of maternal allele expression rather than activation of the paternal allele.
We performed bisulfite sequencing of the gDMR to determine whether the maternal Kcnq1ot1 activation was associated with loss of DNA methylation on that allele. Parental origin of sequenced clones was determined by fixed PO-BW sequence polymorphisms. Both bw6po embryonic and placental tissues displayed clones that were either largely methylated or unmethylated (Fig. 1B). As expected, the methylated cytosines appeared largely on the silenced maternal BW allele. The correlation between methylation and expression was also evident in the PO6BW placentas. That is, the maternally derived PO clones typically displayed few methylated cytosines, reflecting the observed placental bi-allelic expression.
While the maternal Kcnq1ot1 allele was also activated in PO6BW embryos, the bisulfite sequencing revealed little accompanying change in DNA methylation. While maintenance of Kcnq1ot1 cluster imprinting has been shown to be regulated by epigenetic marks other than DNA methylation, this has been demonstrated primarily in extra-embryonic tissues [22]. The retention of Kcnq1ot1 gDMR paternal methylation in PO6BW embryos may correlate with the earlier observation that Phlda2 and Cdkn1c down-regulation is placenta-specific in the cross [27].

Loss of methylation at the Peg3-Usp29 domain
We have previously shown that the paternally-expressed Peg3 gene displays LOI and increased expression in both embryonic and placental tissues. The neighboring Usp29 gene also displays PO6BW LOI (data not shown). Imprinting of Peg3 and Usp29 is maintained in the bw6po hybrids [35]. We first undertook southern analysis with methylation-sensitive restriction enzymes to determine whether reactivation of the silenced allele is accompanied by alterations in DNA methylation at the gDMR location identified in Mus. Through a length polymorphism, these data did indeed indicate loss of methylation from the maternal allele ( Figure S1).
We next performed bisulfite mutagenesis and sequencing of the Peg3 gDMR to more thoroughly examine individual gDMR cytosine methylation status. While there is no PO-BW polymorphism in the amplified region, both embryonic and placental bw6po clones displayed the expected 50:50 methylated (maternal): unmethylated (paternal) ratio (Fig. 2). In contrast, both PO6BW tissues displayed a large proportion of unmethylated sequence reads, with a greater proportion in the placenta. Thus these data indicate a correlation of DNA methylation loss at the ICR with reactivation of the Peg3 maternal allele.

Tissue-specific loss of imprinting with little DNA methylation loss at Peg10
The Peg10-Sgce cluster is organized a similar manner to the Peg3-Usp29 pair, with a gDMR lying between the two oppositely oriented paternally expressed loci. To test Peg10 allelic expression, we devised an RT-PCR/RFLP assay. Results of this assay indicated that Peg10 LOI is limited to PO6BW embryonic tissues. This is the first time this combination (placental imprinting; embryonic LOI) has been observed in the offspring of this cross (Fig. 3A). The bw6po hybrids again displayed imprinted Peg10 expression in both somatic and extra-embryonic components. Despite sequencing over 2 kb of the Sgce gene, no PO-BW polymorphism was detected to assess allelic expression.
We next performed bisulfite sequence analysis to determine if loss of cytosine methylation at the intergenic gDMR correlated with the Peg10 LOI observed in PO6BW embryos (Fig. 3B). Due to difficulty cloning the ,1 kb Peg10 gDMR amplicons, we supplemented this analysis with a methylation-specific PCR (MSP) assay. In this assay, separate primer sets are designed for the converted vs. unconverted strands after bisulfite treatment. A length polymorphism in the amplified region allowed determination of allelic parental origin (Fig. 3C). As expected, both assays revealed high levels of methylation on the Peg10 maternal allele and little methylation on the paternal allele in all cross/tissue combinations. In both cases, there were possible indications of minor methylation loss from the PO6BW maternal allele.

Lack of perturbations at the Meg3-Dlk1 domain
Several studies have suggested structural similarities between this domain and the H19-Igf2 pair. Both domains contain a maternally-expressed (H19, Meg3) and a paternally expressed (Igf2, Dlk1) locus separated by a paternally methylated IG-DMR/ICR [41]. The biallelic expression of H19 (but imprinting of Igf2) in PO6BW offspring [35] might then imply similar deregulation of Meg3 (formerly Gtl2) but imprinting of Dlk1. Consistent with this hypothesis, Dlk1 was shown to be paternally expressed in the PO6BW offspring as the result of a screen for novel imprinted loci [42].
To determine if the Meg3 locus loses imprinting analogous to that observed at H19, we developed an RT-PCR/ RFLP assay to determine allelic usage. Unlike H19, Meg3 remains tightly imprinted in the placentas and embryos of both crosses (Fig. 4A). Another transcript in this domain, Dio3, displays preferential expression of the paternal allele in Mus embryos [43], but is not imprinted in placental tissues. We examined Peromyscus hybrid Dio3 allelic expression by a similar assay. Similar to Mus, this gene displayed biallelic expression in both hybrid placenta types, but was imprinted in embryonic tissues (Fig. 4A). These data   demonstrate that, unlike the Igf2-H19 domain, imprinting in the Dlk1-Meg3 imprinted domain is not significantly influenced by interspecific hybridization.
We investigated whether DNA methylation might be perturbed at this locus in spite of the apparently undisturbed imprinting status. We carried out bisulfite sequence analysis on a portion of the 39 end of the Meg3 gDMR. The methylation patterns were not clearly allelic, and no polymorphism was available to determine parental origin. There was no unambiguous evidence for loss of methylation in this region (Fig. 4B). However, three of sixteen PO6BW embryonic clones displayed no methylated cytosine residues while this was not true of any of the bw6po derived clones.

Allelic expression of the Plagl1 and Dcn imprinted loci
To further characterize the hybrid imprinting patterns, we assessed the allelic expression status of two imprinted genes not linked to any other of those investigated. The Plagl1 gene (aka Zac1, Lot1) has an associated gDMR and one additional linked imprinted locus (HYMAI), but its regulation has been little studied [44,45]. The Decorin (Dcn) gene lacks any other known associated imprinted loci or gDMR [46]. The Mus Plagl1 gene is paternally expressed in several fetal and adult tissues, as is its human ortholog in placentas [47,48]. Dcn has been shown to be maternally expressed in Mus placentas, but exhibits biallelic expression in other tissues [49] and other species (cows, humans) [50,51].
Similarly, analysis of bw6po hybrid placentas shows imprinted expression of Dcn, and biallelic expression in all other tissues (Fig. 5A). Dcn is biallelic in all PO6BW tissues, suggesting that imprinting is lost in the placenta. Placental expression of Plagl1 is also consistently biallelic in the in PO6BW hybrids. As this is also true in the case of the bw6po hybrids, this likely also represents differences in embryonic vs. placental regulation. Note, however, that a slight amplification bias in favor of the BW allele indicates that the bw6po placentas may be closer to imprinted than those from the PO6BW cross. Allelic expression is more sporadic in PO6BW embryonic tissues, with some samples exhibiting strict imprinting (Fig. 5B), and others bi-allelic expression. Thus, the Dcn and Plagl1 data further suggest tissue-specific differences in hybrid misregulation of imprinted loci.

Brain-specific imprinting effects
Previous studies in other mammalian groups suggest that the brain represents another instance of tissue-specific regulation of imprinted loci. For example, the Igf2 gene exhibits biallelic expression in the brains of house mice, humans, and Peromyscus [30,[39][40][41]. To assess whether the hybrid misregulation also exhibited brain-specific patterns, we tested the allelic expression of several imprinted genes including H19, Peg10, Kcnq1ot1, Plagl1, and Meg3.
Unlike placental and other embryonic tissues, the brains of PO6BW embryos displayed maternal expression of H19 (Fig. 6A). Equally surprisingly, bw6po brains exhibited some reactivation of the paternal H19 allele, despite strict imprinted expression elsewhere. The Peg10 gene also showed paternal expression in the PO6BW hybrids examined (Fig. 6B), in contrast to that observed in whole embryos. The bw6po brains also displayed imprinted Peg10 expression.
In contrast, Kcnq1ot1 appears to be universally imprinted in bw6po conceptuses and to lose imprinting in all tissues of PO6BW conceptuses (Fig. 6C). Similarly, the Meg3, Plagl1 and Peg3 brain allelic expression patterns also resemble the expression patterns of the fetus as a whole (Fig. 6D and not shown).

Discussion
A logical approach to understanding the mechanisms of genomic imprinting is to categorize gene clusters by apparent  regulatory similarities. Examples include allele-specific boundaries (H19/Igf2), functional ncRNAs (e.g. Kcnq1ot1), and dual promoter methylation (e.g. Peg3/Usp29). The patterns of imprinting perturbations in the Peromyscus hybrids (summarized in Table 1) suggest functional differences in several of these analogous clusters. For example, evidence suggests that the maternally expressed Igf2r and Cd81 loci are each regulated by linked paternally-expressed ncRNAs (Air and Kcnq1ot1, respectively) [52,53]. Igf2r retains normal imprinted maternal expression in PO6BW hybrid crosses [35], while the Kcnq1ot1-Cd81 cluster is misregulated. In contrast, the paternal allele of Igf2r is reactivated in bw6po placentas [35] while both Cd81 and Kcnq1ot1 are unperturbed in this cross.
Similarly, the loss of H19 imprinting in PO6BW crosses is not mirrored by its analogue Meg3, which retains maternal monoallelic expression. Differences in hybrid regulation are also seen in the Peg10 and Peg3 domains, despite the apparent similarities in structure and regulation. While both domains exhibit loss-ofimprinting in the PO6BW cross, DNA methylation at Peg10 is apparently little perturbed.
These data suggest similarities in Peg3, Kcnq1ot1 and H19 imprinting cluster regulation not shared by more putatively similar loci (e.g. Peg 10, Meg3). We hypothesize that a major commonality among these loci will be regulation by the product of the maternal effect locus (Meil) for which genetic evidence has previously been described [37]. Specifically, we suggest that the PO allele of this gene cannot initiate and/or sustain imprinting in the presence of BW chromatin (i.e. PO6BW cross). Demonstrated imprinting within P. polionotus implies that there is natural variation at this locus, rather than a null allele or absence of the locus in this species [35]. Identification of this locus is therefore paramount in understanding the hybrid misregulation as well as normal epigenetic regulation of mammalian development.
We have shown that Meil is not identical to any of the DNA methyltransferase (Dnmt) loci, nor to several other candidate loci [33]. We suggest that the Meil product is deposited in oocytes and involved in epigenetic regulation. This hypothesis is strengthened by a study of reciprocal in vitro fertilization efficiency between PO and BW [54]. This study found that 8% of BW oocytes fertilized with PO sperm resulted in abnormalities (i.e. failure to fertilize or cleave). In contrast, PO oocytes fertilized with BW sperm resulted in a 26% abnormality rate [54].
There are two potential caveats to the observed patterns of imprinted gene misregulation. First, while we have documented imprinting within P. polionotus, it is formally possible that some loci are not imprinted in Peromyscus but bias is induced in the bw6po or both hybrid types. If this is the case, it may be a widespread problem. The house mouse crosses used to define many imprinted loci involve crosses to other species or subspecies that have also been associated with decreased fitness [55][56][57]. More broadly, crosses between species/strains that exhibit many expressed polymorphisms (used for ease of developing allelic expression assays) may also be polymorphic at cis-regulatory sequences. The second potential caveat lies in our examination of tissues from the second half of gestation; prior analyses suggest that the majority of PO6BW conceptuses are dead by this time [27]. Other studies suggest a stochastic element in the loss of epigenetic marks [58], which is also suggested by the patterns of DNA methylation loss in PO6BW conceptuses (e.g. Peg3; Fig. 2). This suggests the possibility that perturbations are typically greater in this cross, but that we have selected for less affected conceptuses. For example, Peg10 has been shown to be necessary for Mus placental development [59]. Strict imprinting of Peg10 in the examined PO6BW conceptuses may reflect that over-expression is similarly incompatible with survival to late gestation. A somatic example of Peg10 loss-of-imprinting has been found in several tumor types [60,61].
One possibility is that the PO6BW imprinting perturbations in particular are byproducts of more inclusive epigenetic perturbations. For example, DNA methylation is now known to regulate many non-imprinted/X-linked loci [62,63]. Consistent with the variability seen in the PO6BW cross, deletion of even the oocytespecific version of the maintenance DNA methyltransferase Dnmt1 (DNMT1o) also results in a large variety of defects [58]. The knockout variation is thought to be due to stochastic epigenetic variation present in the early embryo.
A stochastic element to PO6BW epigenetic mark loss may also be an important element in understanding the hybrid defects. We hypothesize that cells which lose sufficient DNA methylation at the Kcnq1ot1 and similar ICRs result in androgenetic-like expression patterns. Androgenetic (no maternally inherited genome) Mus embryos and typical androgenetic hydatidiform moles result in an early embryonic shift towards extra-embryonic cell fates. In this scenario, the PO6BW loss of DNA methylation is not due to placenta-specific factors; rather, loss of methylation results in a gene expression profile more compatible with extra-embryonic fates.
Several aspects of the PO6BW cross recall those observed in human biparental hydatidiform moles (BiHMs). Hydatidiform moles are typically androgenetic (e.g. due to loss of the oocyte pronucleus) [64]. In the last decade, molar pregnancies with a normal 1:1 parental genomic contribution have been reported [65,66], and are thought to account for the majority of recurrent familial cases of this syndrome [67]. In at least several of these families, women have borne normal children between BiHM pregnancies [67,68]. This suggests negative interactions between variants in an oocyte-associated gene product and elements within the sperm.
BiHM tissues exhibit complex patterns of imprinted gene expression and methylation perturbations via a maternal effect locus or loci [66,68,69]. These loci include PEG3, SNRPN, KCNQ1OT1, CDKN1C [69,70], and in some cases H19 [70]. Genetic studies indicate the existence of multiple BiHM susceptibility loci [67]. One of these loci maps to human chromosome 19q13.4 [71], the equivalent domain to which a paternal expressed gene involved in the Peromyscus placental hypertrophy maps [72]. While Peg3/PEG3 lies in this region, no human coding SNPs of the gene correlate with BiHM susceptibility. Instead, the NLRP7 (formerly NALP7) gene has emerged as the primary candidate in this region [69,73].
The NLRP7 product is known to be involved in apoptotic and inflammatory pathways, but the connection to the molar phenotype is unclear. House mice apparently lack a NLRP family member in this region; the Peromyscus domain has not yet been characterized. It is striking that the two phenomena (BiHM & PO6BW hybrids) display imprinting perturbations, trophoblast over-proliferation, and a genetic component mapping to corresponding genomic regions. Both identification of Meil and characterization of the early PO6BW epigenetic misregulation are likely to aid in understanding these profound shifts in cell-fate.
That such epigenetic regulatory variation exists in recently diverged species suggests a role in reproductive isolation, and that this variation is likely to be adaptive. Epigenetic phenomena by definition mediate gene-environment interactions. The importance of such interactions is increasingly being recognized in understanding biological processes. We believe this system, having the potential to link naturally differing allelic combinations with environmental and behavioral variation, has a unique potential to aid in elucidating these interactions [74].

Animals and Breeding
We purchased PO and BW stocks from the Peromyscus Genetic Stock Center (http://stkctr.biol.sc.edu/). Both parental strain and interspecific cross conceptuses were bred at UCI. Animals were kept and treated in conditions approved by the University of California Irvine Institutional Animal Care and Use Committee (IACUC), protocol #2001-230. Animals were fed a standard ad libitum high protein/fiber diet and water. The light/dark cycle was 16:8 hours. Hybrid embryos were collected at the equivalent of Mus embryonic14.5 for all assays. Embryos and placentas were split sagittally, with one half harvested for RNA, and the other for DNA.

Cloning of Peromyscus Sequences
Human, mouse, and/or other available mammalian species were aligned, and primers for the cloning of imprinted genes/ DMRs from Peromyscus. Primers were designed to conserved regions when possible. To acquire Peromyscus ICR/gDMR sequences, we probed a P. maniculatus BAC library filters (CHORI-Oakland) with PO/BW imprinted gene sequences. Positive clones were then ordered from CHORI and confirmed via PCR. Relevant DMR sequences were then cloned via PCR using BAC clones as templates, followed by primer walking. When this was not possible, BACs were sequenced to acquire sequence from the relevant regions. All sequences have been deposited in Genbank, accessions numbers EU746661-EU746681.

Allelic Expression Assays
Prenatal tissues were harvested at the equivalent of Mus embryonic day 14.5. Heads of embryos were removed prior to RNA extraction in order to avoid brain-specific assay complications. RNA was isolated from embryos, placentas, and brains with the Qiagen RNeasy kit, including a DNase step to remove genomic DNA. Superscript II reverse transcriptase (Invitrogen) was used to generate cDNA. For Kcnq1ot1, a series of Kcnq1ot1specific oligonucleotides were used as primers for cDNA synthesis. In all other cases, oligo-dT was used for synthesis. RT(-) reactions were used as controls against gDNA contamination. Allelic usage was determined by restriction fragment length polymorphism. Mixes of parental strain templates were used to assess possible allelic amplification bias [75]. Primers and annealing temperatures are listed in Table S1.

DNA Methylation Analysis via Bisulfite Treatment and Sequencing
DNA was isolated from Mus-equivalent e14.5 embryos and placentas via either phenol extraction or Qiagen DNeasy kit. Isolated DNA was then subject to bisulfite conversion via agarose bead [76] or Methylamp TM DNA Modification Kit (Epigentek) protocols. Approximately 400 ng of bisulfite-converted DNA was used per PCR reaction. PCR mixes contained 1.5 mm MgCl2, 16 PCR buffer (Applied Biosciences), 1 U Taq polymerase, and 100 pg each primer. Primers were designed from cloned sequences via Methprimer [77], and were used in nested reactions. Primary PCR reactions featured 35-40 cycles, followed by a 20 cycle nested reaction. Primers and Annealing temperatures are listed in Table S1.
Successful PCR products were cloned into a TOPO-TA vector (Invitrogen) and sequenced. As noted, sequence reads that suggested incomplete bisulfite reactions were discarded. In the case of the Peg3 and Kcnq1ot1 DMRs no sequence read with more than 3 unconverted non-CpG cytosines was included in the analysis. For the Meg3 DMR, no read with more than 2 was included. For the Peg10 DMR, no read with more than 10 was included (the region analyzed contains 321 cytosine residues).

Methylation Specific PCR (MSP)
To determine allelic methylation within the Peg10 gDMR, we designed primers specific for methylated and unmethylated bisulfite-treated DNA. The assay primers span a region where a PO/BW size polymorphism allows determination of parental origin for amplification products. MSP reactions were performed using the first round of Peg10 bisulfite PCR products as templates, and were carried out over 20 cycles. Primers and annealing temperatures are listed in Table S1.

DNA methylation Analysis via Restriction Digestion/ Southern Analysis
DNA was extracted via phenol extraction and ethanol precipitation. Ten to fifteen micrograms of each extract were digested with HpaII, MspI and or EcoRI as indicated (New England Biolabs). Digests were then electrophoresed on 1% agarose gels. For Peg3 southern analysis, the probe was random hexamer oligolabelled and hybridized in Church buffer at 65uC. Washes were done at 65uC in 0.56 SSC, 0.1% SDS. Figure S1 Southern blot analysis of PO6BW DNA methylation at the Peg3 locus. Genomic DNA was first digested with EcoRI (RI) alone, then divided into 3 aliquots. The first was not treated further; the other 2 were subsequently digested with MspI (+Msp) or HpaII (+Hpa). The probe used was a ,580 bp fragment corresponding to sequence from the Peg3/Usp29 intergenic region to Peg3 intron1. Genotype is listed at top; PO DNA is shown with all 3 enzyme combinations. BW DNA is shown cut with RI alone to illustrate the species size polymorphism (PO,11 kb, BW,16 kb). Arrows at side indicate either allelespecific bands or fully digested RI+Msp DNA (Dig). Note in the PO6BW+Hpa lane that the paternal BW allele is absent, and that the maternal (PO) band is also severely reduced. Found at: doi:10.1371/journal.pone.0003572.s001 (2.05 MB TIF)