A Non-Coding RNA Within the Rasgrf1 Locus in Mouse Is Imprinted and Regulated by Its Homologous Chromosome in Trans

Background Rasgrf1 is imprinted in mouse, displaying paternal allele specific expression in neonatal brain. Paternal expression is accompanied by paternal-specific DNA methylation at a differentially methylated domain (DMD) within the locus. The cis-acting elements necessary for Rasgrf1 imprinting are known. A series of tandem DNA repeats control methylation of the adjacent DMD, which is a methylation sensitive enhancer-blocking element. These two sequences constitute a binary switch that controls imprinting and represents the Imprinting Control Region (ICR). One paternally transmitted mutation, which helped define the ICR, induced paramutation, in trans, on the maternal allele. Like many imprinted genes, Rasgrf1 lies within an imprinted cluster. One of four noncoding transcripts in the cluster, AK015891, is known to be imprinted. Methodology/Principal Findings Here we demonstrate that an additional noncoding RNA, AK029869, is imprinted and paternally expressed in brain throughout development. Intriguingly, any of several maternally inherited ICR mutations affected expression of the paternal AK029869 transcript in trans. Furthermore, we found that the ICR mutations exert different trans effects on AK029869 at different developmental times. Conclusions/Significance Few trans effects have been defined in mammals and, those that exist, do not show the great variation seen at the Rasgrf1 imprinted domain, either in terms of the large number of mutations that produce the effects or the range of phenotypes that emerge when they are seen. These results suggest that trans regulation of gene expression may be more common than originally appreciated and that where trans regulation occurs it can change dynamically during development.


Introduction
Genes inherited in two copies, one from each parent, are normally functionally equivalent, however, notable exceptions to this have been known for many years [1][2][3][4][5]. Among the genes that are functionally distinguishable according to their parent of origin are the imprinted genes, for which one parental copy is silenced while the other is expressed. Because these genes have unique modes of regulation and are often implicated in disease syndromes, much effort has been devoted to their identification and study.
The first imprinted locus identified was an artificial transgene carrying elements of the RSV LTR and a translocated c-myc gene. This transgene was expressed in the heart, but only when inherited from the father. Furthermore, the silencing that occurred upon maternal transmission was accompanied by DNA methylation of the transgene, which has been known to regulate gene expression [6]. Shortly after this discovery, two naturally occurring genes, H19 and Igf2r, were shown to exhibit parent of origin specific expression [7,8] and DNA methylation [9,10]. Since these discoveries, the role of DNA methylation in the regulation of imprinted genes was firmly established [11]. Besides DNA methylation, additional factors are known to regulate imprinted gene expression, including the covalent modifications to histone proteins associated with DNA [12][13][14][15][16], the CTCF protein, [17][18][19][20][21][22], and physical interactions between the two parental chromosomes [23][24][25][26][27][28].
DNA sequences that control imprinting lie within the ICR, which often regulates imprinting of several nearby genes. Rasgrf1 is paternally expressed in brain [29] and its ICR is located 30 kb upstream of the Rasgrf1 transcription start site between Rasgrf1 and four upstream noncoding (nc) RNAs ( Figure 1A). One of these ncRNAs, A19 (also called AK015891), located ,10 kb upstream of the Rasgrf1 ICR, is also imprinted in brain, where expression is solely from the paternal allele [30]. As ICRs have only been identified for seven imprinted clusters [31], the Rasgrf1 system provides an uncommon opportunity to study regulation of imprinted gene expression.
The Rasgrf1 ICR contains a differentially methylated domain (DMD) as well as a series of tandem DNA repeats, consisting of 40 copies of a 41 bp repeat unit. The DMD acquires DNA methylation exclusively on the paternal allele and placement of DNA methylation is controlled by the tandem repeats ( Figure 1B) [32,33]. The repeats are required for both establishment and maintenance of allele-specific DNA methylation during spermatogenesis and through pre-implantation development. Only a few other sequences have been implicated as cis-acting regulators of DNA methylation [34][35][36]. Imprinted expression is a result of both allele-specific binding of the methylation-sensitive enhancer blocker-binding protein, CTCF, and allele-specific DNA methylation. CTCF binds the unmethylated maternal DMD and disrupts enhancer to promoter interaction, resulting in silencing of Rasgrf1. On the paternal allele, the DMD is methylated, preventing binding of CTCF and allowing paternal allele expression [22]. The presence of the DNA repeats on the paternal allele is necessary for the establishment and for the maintenance of DNA methylation at the DMD until implantation at embryonic day 5.5 (e5.5). Deletion of the paternal repeats before this time point leads to loss of DNA methylation on the paternal allele, resulting in silencing of Rasgrf1 [33]. The DNA repeats, in combination with the DMD, constitute a binary switch that is necessary for imprinted expression of Rasgrf1.
We previously generated a Rasgrf1 mutation in which the Rasgrf1 repeats were replaced with sequences referred to as Region 2 (R2) that were implicated in the control of Igf2r imprinting [37]. When paternally inherited, this modification allowed both methylation and Rasgrf1 expression, in cis, of the paternal allele, indicating R2 could substitute for the endogenous Rasgrf1 repeats. Interestingly, the wild-type maternal allele also became methylated and expressed in trans, and once modified, the altered epigenetic state persisted in the next generation, an example of paramutation [37]. This was our first observation of a trans-expression effect in the Rasgrf1 genomic region.
Here, we report that a second ncRNA, AK029869, located ,5 kb upstream of the Rasgrf1 ICR is imprinted and paternally expressed in neonatal brain. We also show that deletion of the tandem repeats within the Rasgrf1 ICR perturbs imprinting of AK029869, and has the ability to do so in trans. Finally, we demonstrate that several ICR mutations on the maternal copy of Rasgrf1 lead to silencing of the normally expressed AK029869 paternal allele in trans. These constitute additional examples of trans-expression effects within the Rasgrf1 imprinted cluster and demonstrate the robust communication between Rasgrf1 alleles across chromosomal boundaries.

AK029869 is imprinted in brain
To test the imprinting status of AK029869, we first identified useful SNPs between the polymorphic mouse strains PWK and C57BL/6 (B6) to distinguish the expression from the two alleles. We chose one SNP that overlapped with an AluI restriction site. AluI digestion produces distinct banding patterns for the two strains. We established reciprocal crosses between PWK and B6, extracted RNA from brains of post partum day 10 (P10) progeny and then performed RT-PCR followed by AluI digestion to identify the expressed alleles. We found that AK029869 expression was exclusively from the paternal allele at P10 (Figure 2A). Though there was a bias in the efficiency of amplification of the two alleles, paternal allele specific expression was quite clear. As imprinted expression can be developmental time point specific, we repeated the analysis using brain samples collected between embryonic day 16.5 (e16.5) through P42. We found that  [32,33], which blocks H3K27me3 placement [14] as well as CTCF binding and enhancer blocking activities of the DMD, allowing enhancer-driven expression [22]. Tandem repeats stimulate H3K27me3 on the maternal DMD (pink), which blocks DNA methylation [14], allowing CTCF binding and enhancer-blocker activities, which silences Rasgrf1 transcription [22]. Black and white circles over the DMD respectively depict methylated and unmethylated CpGs. doi:10.1371/journal.pone.0013784.g001 AK020869 was paternally expressed throughout this interval ( Figure 2B). In addition to brain, we performed the analysis on testes taken between embryogenesis and adulthood. AK029869 expression was biallelic in testes at all stages tested (not shown).
Allele-specific expression of AK029869 depends on the Rasgrf1 repeats Next, we wanted to test whether paternal allele-specific expression of AK029869 depended upon the presence of the tandem DNA repeats in the ICR, as is the case for Rasgrf1. To do this, we established reciprocal crosses between PWK animals and mice carrying a targeted deletion of the tandem DNA repeats (Drep, Figure 3A, [32]). The Drep allele, and all other Rasgrf1 ICR mutations described below were prepared on the 129S4Jae background, which shares the same AK029869 AluI site polymorphism as B6. As with Rasgrf1, paternal inheritance of a Drep allele led to silencing, in cis, of AK029869. Expression of the normally silent maternal allele was unaffected, as was DNA methylation at the maternal DMD [32]. Unexpectedly, maternal inheritance of the Drep allele led to silencing, in trans, of the paternal AK029869 allele ( Figure 3B). This trans-silencing was independent of DNA methylation, as maternal inheritance of a Drep allele did not change the DNA methylation status of either DMD, or expression of Rasgrf1 [32].

Effect of timing of repeat deletion on AK029869 expression
After fertilization, there is a developmental time before which the repeats are required to maintain methylation at the Rasgrf1 DMD, and after which they are dispensable [33]. Therefore, we wondered if the cisor trans-silencing we observed at P10 in Rasgrf1 repeat-deficient mice was also sensitive to the developmental stage at which the repeats were deleted. To control deletion of the maternal or paternal repeats after fertilization, we used mice homozygous for an allele with loxP sites flanking the Rasgrf1 DNA repeats (flox-rep, Figure 4, [33]). This allele is functionally wildtype, preserving normal imprinted expression of both AK029869 and Rasgrf1 as well as imprinted DNA methylation [33]. We crossed mice bearing the flox-rep allele with different Cre transgenes to delete the repeats. To facilitate allele-specific expression analysis, we bred the Cre transgenes onto the PWK mouse background. The breeding strategy is shown in Figure S1. AK029869 is imprinted in brain at many developmental stages. (A) Brains from post-natal day 10 (P10) or (B) embryonic day 16.5 (e16.5) mice and older (P11-P42) were assayed for AK029869 imprinted expression. Animals were progeny of C57BL/6 (B6) and PWK parents that carry the AluI polymorphism in AK029869 shown below the gel in (A), where numbers represent AluI fragment sizes in nucleotides. Maternal strain is shown first in red followed by paternal strain in blue. Where no cross is shown, the mice were inbred. (A) RT-PCR products from six independent progeny of two reciprocal crosses were digested with AluI before gel analysis (left panel). Inbred strains and mixtures of inbred cDNAs in the ratios indicated were included as markers and controls for amplification bias (right panel). Band sizes and the strain source are to the left of the gel. (B) cDNAs were analyzed as in (A) with the exception that undigested (U) RT-PCR products were also analyzed and PCR was done using mRNAs that were (+) or were not (-) reverse transcribed (RT). M identifies marker lanes. doi:10.1371/journal.pone.0013784.g002 We used the Zp3 Cre transgene ( Figure 3C), which is active at e0.0 and can delete the repeats at the one-cell stage [38], and the Meox2 Cre transene ( Figure 3D, E), which is active at e5.5 and deletes the repeats in the embryonic ectoderm [39]. Depending on the direction of the cross, we were able to delete the repeats at these two time points from the maternally or the paternally transmitted allele.
From these crosses, we collected P10 brain and testes from 50 litters of mice. The potentially informative animals were those bearing the Cre transgene, one 129S4Jae allele and one PWK allele at AK029869. We also genotyped each tissue sample to determine the extent of deletion of the loxP-flanked Rasgrf1 repeats because only mice that had complete Cre-mediated deletion of the repeats were informative. We prepared cDNA from all informative brain samples and performed allele-specific expression analysis of AK029869.
This analysis revealed multiple patterns of AK029869 expression control by the Rasgrf1 repeats that changed according to developmental stage. When the paternal repeats were deleted by the Zp3 Cre transgene near the time of fertilization, AK029869 expression was both silenced in cis on the paternal allele and activated in trans on the maternal allele ( Figure 3C). Interestingly, this pattern changed when the paternal repeats were deleted later in the embryonic ectoderm of the e5.5 epiblast by a maternal Meox2 Cre transgene. Deletion at this later stage caused only paternal allele silencing but no maternal allele activation ( Figure 3D). When the maternal repeats were deleted by a paternally transmitted Meox2 Cre transgene, paternal expression of AK029869 expression was silenced in trans ( Figure 3E). As Zp3 Cre is not expressed in sperm, it could not be used to delete the maternal repeats upon fertilization. These results demonstrated that germ line inheritance of a Rasgrf1 repeat deletion allele silenced the normally active paternal allele in cis and in trans, whereas somatic deletion of the repeats after fertilization produced multiple cis and trans regulatory effects that included silencing and activation and that varied by developmental stage.
Additional Rasgrf1 ICR mutations result in cis and trans silencing of AK029869 In addition to the Drep allele, we tested several other ICR alleles for their ability to influence AK029869 imprinting, when transmitted as germ line mutations. Figure 4 shows the alleles, their effects on AK029869 expression, and Figure 5 summarizes the results.
The first additional ICR mutation had an extra enhancer (enh) in place of the tandem DNA repeats [22]. When paternally transmitted, expression of AK029869 was from the paternal allele, as it is in wild-type mice ( Figure 4B, lanes 4-7). However, when The wild-type Rasgrf1 allele (top) includes a 1.6 kbp repeat element (black triangles) needed to program DNA methylation at the DMD (white rectangle). In the repeat deficient allele (bottom), the repeats were replaced with a single loxP site (white triangle). The repeat deficient allele was prepared as a germ line mutation [32] and as a conditional allele whereby the repeats can be deleted by Cre recombinase [33]. Deletion of the repeats from the conditional allele leaves an frt site adjacent and in addition to the loxP site. (B) Allele-specific expression of AK029869 was assayed as in Figure 2 using P10 brain cDNAs from progeny of reciprocal crosses between wild-type PWK animals and B6 mice carrying the repeat deletion in their germ line. The mutation was made on a 129S4Jae background, which has the same polymorphism as B6. The analysis was repeated using mice from which the repeats were removed by Cre transgenes at different times after fertilization. (C) Paternal repeats were deleted by a maternally transmitted Zp3-Cre transgene that can delete the repeats upon fertilization [38] or (D) by a maternally transmitted Meox2-Cre transgene that can delete the paternal repeats in the embryonic ectoderm of the e5.5 epiblast [39]. (E) The maternal repeats were deleted by a paternally transmitted Meox2-Cre transgene. Animals that lacked complete deletion of the repeats were excluded from the analysis (not shown). Rpl32 was included as a control for samples lacking detectable AK029869 expression. The strain origins of bands are shown to the left of the gels with the maternal and paternal strains shown in red and blue, respectively. doi:10.1371/journal.pone.0013784.g003 maternally transmitted, the paternal allele was silenced in trans and the maternal allele was inappropriately activated in cis ( Figure 4B,  lanes 8-11). The second additional ICR mutation (flox-rep-enh) carried loxP sites on either side of the repeats, with an extra enhancer 39 of the repeats [33]. Like the enh allele, maternal transmission inappropriately activated the maternal allele in cis, and expression of the paternal allele was strongly diminished in trans (Figure 4, lanes 12-17).
The extra enhancer on the enh and flox-rep-enh alleles was from the housekeeping gene Pgk and is known to cause over expression of Rasgrf1 [40]. We wondered whether the enhancer was so strong that it boosted AK029869 expression from the mutated allele high enough to mask otherwise normal levels of expression from the homologous wild-type PWK allele. This could lead us to conclude, erroneously, that maternal transmission of the mutated enhancer carrying alleles silenced the paternal allele in trans. To explore this possibility, we mixed brain cDNAs from a wild-type PWK mouse and a mouse with the enh allele in varying ratios and looked for evidence of amplification of both alleles. In each case, even when the mixture contained a three fold excess of  [22], flox-rep-enh [33], flox-rep [33] and R2 [37]. Enh-neo is a neomycin phosphotransferase cassette carrying the PGK enhancer; Igf2 R2 is the Region 2 DMR from Igf2r; black, white and grey triangles respectively depict wild-type repeats, loxP sites, and frt sites. All mutated alleles were on the 129 background, which shares the same AluI restriction polymorphism with B6. (B) Allele-specific expression of AK029869, and total expression of Rpl32 were assayed as in Figure 2 using brain cDNAs from P10 progeny of reciprocal crosses between PWK mice and mates carrying various germ line mutations. Allele specific analysis of Rasgrf1 expression was as described [29]. doi:10.1371/journal.pone.0013784.g004 cDNA from enh-bearing mice, we were able to easily detect PCR products diagnostic of both alleles. This provides confidence that PCR artifacts did not confound our conclusion that maternally transmitted extra enhancer alleles silenced the paternal allele in trans ( Figure S2).
The third additional ICR mutation we tested lacked the Rasgrf1 repeats and replaced them with the differentially methylated domain, referred to as Region 2, from the imprinted and maternally expressed Igf2r gene (R2 allele). When the R2 allele was maternally inherited, AK029869 demonstrated normal Figure 5. Summary of Rasgrf1 and AK029869 expression phenotypes of ICR mutants. Each of the allele combinations represented in rectangles on the left hand side were tested for their effect on both AK029869 and Rasgrf1 expression. Abbreviated names for the alleles used here as well as their formal names are shown. Green and red lines respectively indicate positive or negative effects on expression that do not occur in wildtype mice. The = and R symbols in rectangles on the right hand side indicate whether expression, if any, is from the maternal (R) or the paternal (=) alleles. Black circles below the DMD indicate that the DMD is methylated and white circles indicate lack of methylation. Grey circles and dashed lines in mice with a paternal R2 allele indicate partial methylation and expression. doi:10.1371/journal.pone.0013784.g005 expression from the paternal allele only (Figure 4, lanes 20-23). However, when paternally inherited, the R2 allele silenced the paternal allele in cis. Interestingly, this differed from the effect of the paternal R2 allele on Rasgrf1 expression. When the R2 allele was paternally inherited, the paternal copy of Rasgrf1 was activated in cis and the normally silent maternal Rasgrf1 allele was activated in trans ( [37] and Figure 4, lanes [25][26]. This is because Region 2 positively regulates DNA methylation at the Rasgrf1 DMD, in both cis and trans, when paternally inherited. A caveat is that paternal Rasgrf1 expression is reduced, but still detectable, with an R2 paternal allele [37]. Because AK029869 expression is low in wildtype animals, it is possible that paternal transmission of the R2 allele reduced AK029869 expression below the threshold of detection. Maternal transmission of the R2 allele has no effect on Rasgrf1 expression or DNA methylation at the DMD [37].

Discussion
Although the role of DNA methylation in the control of imprinted expression is relatively well understood, advances in the understanding of other factors regulating imprinted expression have been made more slowly. The body of evidence for the involvement of trans-expression effects in imprinted gene regions is growing. At the imprinted gene, Ins2, a mutation on the paternal allele silenced expression of the normally expressed maternal allele in trans [41]. A mutation to the paternal allele of the paternally expressed gene, Snrpn, activated expression of the normally silenced maternal allele in trans [42]. At a nearby locus, a mutation on the maternal allele of the maternally expressed gene Ube3a, silenced expression of the paternally expressed antisense transcript, Ube3a-ATS, in trans [43]. At the H19/Igf2 imprinted locus, a maternal allele deletion at the H19 allele led to a reduction in the level of DNA methylation at the wild-type Igf2 paternal allele [44]. Also at the H19/Igf2 locus, replacement of a differentially methylated domain between Ins2 and Igf2 on the paternal allele led to activation of the normally silent Igf2 maternal allele [45]. Similarly, when this mutation was present on the maternal allele, the normally silent H19 paternal allele was activated [45]. Consistent with this trans chromosomal communication that affects expression at imprinted genes is evidence that physical interactions occur between imprinted loci and both homologous and non-homologous loci [23,27].
In addition to imprinted genes, trans-effects occur at other loci including the X-chromosome inactivation center [46], odorant receptors [47], and T-helper cytokines [48]. As with imprinted genes, the expression of these genes is also epigenetically regulated [49][50][51]. During mammalian X chromosome inactivation, one of the female X chromosomes is silenced to effect equal levels of X chromosome gene expression between XX and XY individuals. X chromosome inactivation is regulated by the genes Xite, Tsix, and Xist. Two of these genes, Tsix and Xite, also function in trans to regulate X chromosome counting and choice through physical pairing of the two X chromosome homologues [46,52]. Similar interactions occur between non-homologous chromosomes in Thelper cytokine gene expression and allelic exclusion of odorant receptor expression [47,48], however, the significance of these interactions for gene expression are not clear [53]. Recent data using the Hi-C method show such trans chromosomal interactions are quite common in humans [54]. Little is known about the significance of these long-range genomic interactions or if they are responsible, mechanistically, for the trans expression effects we describe. The Hi-C data were collected for human cells and The Rasgrf1 imprinted cluster in mouse is uniquely positioned as a model to study trans-expression effects for several reasons. First, the Rasgrf1 ICR has been well characterized. Sequences controlling DNA methylation have been identified [32] and the mechanism by which methylation controls expression has been determined [22]. Such details are lacking for other imprinted genes. Second, a variety of ICR mutation alleles are available for this region. Third, the effect that each of these alleles has on the methylation state of the ICR (and Rasgrf1 imprinted expression) is known.
Here, we characterized a novel noncoding RNA within the Rasgrf1 imprinting cluster, AK029869, and showed it is paternally expressed in brain and that this imprinting is subject to trans regulation by various ICR mutations. Imprinted expression of AK029869 and the trans effects that regulate it are distinct in several ways from control of Rasgrf1 imprinting, which we previously described [22,32,33,37]. First, the tandem DNA repeats within the Rasgrf1 ICR were important for proper imprinted expression of AK029869, as is the case for Rasgrf1 expression. But whereas Rasgrf1 imprinted expression depends only on the paternal repeats [32], proper AK029869 imprinting depended on both parental repeats. Loss of either set silenced paternal expression of AK029869, with maternally deleted repeats causing silencing in trans ( Figure 3B). Second, loss of the paternal Rasgrf1 repeats upon fertilization caused silencing of Rasgrf1 imprinted expression [33]. Strikingly, deletion of the paternal repeats during this time led to a reversal of imprinted expression of AK029869 in brain, silencing the normally active paternal allele and activating the normally silent maternal allele ( Figure 3C). Third, proper imprinted expression of Rasgrf1 required the paternal repeats only prior to the epiblast stage. Deleting the repeats at the epiblast stage or later had no effect on Rasgrf1 expression [33]. In contrast, when the repeats were deleted from either the maternal or paternal allele at the epiblast stage, the paternal allele was silenced (Figures 3D and 3E).
Regarding how the expression of AK029869 is regulated, we considered three possibilities. We first considered that DNA methylation levels at the DMD might regulate AK029869. However, maternal inheritance of a Drep allele, an enh allele, or a flox-rep-enh allele all preserved DNA methylation at the paternal DMD, but led to silencing in trans of the paternal AK029869 allele ( [22,32,33], Figures 3 and 4). This indicates that paternal DMD methylation is not sufficient for imprinted AK029869 expression. In contrast, whenever the DMD was methylated, we observed expression of Rasgrf1 [22,32,33].
Second, we considered that proper imprinted expression of AK029869 may occur as long as both the maternal and the paternal alleles carry some form of the tandem DNA repeats, but this is not the case. Mice inheriting a maternal copy of a flox-repenh allele had both the maternal and the paternal tandem DNA repeats, but underwent an inversion of the normal imprinting pattern, with the maternal allele becoming activated and the paternal allele becoming silenced. In addition, mice with a maternally inherited R2 allele lacked one copy of the DNA repeats but exhibited apparently proper paternal allele specific expression of AK029869 (Figure 4).
Third, we considered that sequence spacing within the ICR might be critical for proper expression of AK029869. However, paternal inheritance of the R2 allele retained wild-type sequence spacing, as Region 2 and the tandem DNA repeats are both approximately 2kb, but this led to silencing in cis of the paternal allele ( Figure 4). Therefore, in contrast to the binary switch model for expression of Rasgrf1, there is a distinct mechanism for AK029869 imprinting that is far more complex than the relatively simple binary switch mechanism for Rasgrf1 imprinting.
Despite the complexity of AK029869 regulation, there are three patterns that emerged from our data. First, any changes in spacing within the ICR on the maternal allele led to silencing in trans of the paternal allele. For example, the R2 allele contains an approximately 2kb deletion of the tandem DNA repeats but, since Region 2 is roughly 2kb, insertion of Region 2 retains normal sequence spacing. When maternally inherited, the R2 allele allowed expression of the paternal allele. On the other hand, the floxrep-enh allele changed the sequence spacing of the region, and led to paternal allele silencing. The one exception is the enh allele, which kept the maternal allele spacing but silenced the paternal allele in trans. The enhancer insertion may produce additional changes to local chromatin structure that over ride otherwise normal sequence spacing on the ICR.
A second pattern is that deletion of the paternal repeats, regardless of the resulting sequence spacing, led to silencing in cis of the paternal allele. For example, paternal inheritance of the R2 allele deleted the tandem DNA repeats and preserved sequence spacing, but silenced the paternal allele. Again, the one exception was if the repeats were replaced with the extra enhancer. The enh allele also deleted the tandem DNA repeats and preserved sequence spacing, but it allowed expression of the paternal allele. This represents the third pattern we observed, any allele with an extra enhancer led to activation of that allele in cis, regardless of sequence spacing or the presence of the repeats on that allele.
Therefore, it appears that sequence spacing in the region of the repeats, but not presence of the repeats, on the maternal allele is important, while the presence of the repeats, but not sequence spacing, on the paternal allele is important. Also, the presence of an extra enhancer may override these two requirements to allow expression in cis, but not in trans, indicating that access to an enhancer is necessary for expression of AK029869.
Other imprinted loci, including Igf2, undergo allele specific differences in three-dimensional conformation [24][25][26]28]. One consequence of the maternal allele conformation is that it might prevent distant enhancers from interacting with the Igf2 promoters, enforcing silence of the maternal copy of Igf2. This mechanism may control Rasgrf1 imprinting. Additional and more complex CTCF dependent interchromosomal interactions occur between Igf2 and other genes [27], and such regulation may be occurring to control AK029869 imprinting. If this is the case, it could be dependent on sequence spacing, repeat-content, or a combination of the two, which could in turn be influenced by the presence of enhancers. Experiments such as 3C or FISH are needed to address this question. Nevertheless, the results discussed above clearly demonstrate an abundance of trans-expression effects within the Rasgrf1 imprinted cluster.

SNP and restriction site identification
SNPs were identified in AK029869 using the Jackson Laboratory Mouse Genome Informatics website (http://www. informatics.jax.org). SNPs available between C57/BL6 and 129 versus PWK mouse strains were noted. In each case, potential SNPs were analyzed using NEB cutter (http://tools.neb.com/ NEBcutter2/index.php) to select SNPs overlapping a restriction enzyme recognition site for identification of allele-specific expression.

Ethics statement
The Institutional Animal Care and Use Committee at Cornell University approved all research involving animals described here, as required by the United States National Institutes of Health and Department of Agriculture. Cornell University is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care.

Tissue collection
Crosses were set up between either wild-type B6 and PWK mice or B6 mice homozygous for loxP-flanked copies of the Rasgrf1 DNA repeats and PWK mice carrying either Zp3 Cre or Meox2 Cre alleles. In each case, crosses were set up as reciprocal pairs to either rule out expression differences due to strain QTLs or to examine the effect of inheritance of both maternal and paternal repeat deletions. The progeny of each of these crosses were sacrificed at P10 (except for the wild-type imprinting time course experiment) and a small portion of the brain was collected for genotyping. The remainder of the brain was snap frozen in liquid nitrogen for later use. P11 and younger mice were killed by decapitation; P21 and older animals were killed by CO 2 asphyxiation.

Genotyping
Brain DNA samples were prepared by lysing in Laird's lysis buffer plus proteinase K overnight followed by ethanol precipitation. Brain DNA was genotyped for the presence of one B6 allele and one PWK allele using primers either AKnewFWD

RNA preparation
Brain samples were isolated from F1 progeny of reciprocal crosses at P10 (except in the case of the imprinting timecourse experiment) and total RNA was prepared. For each neonatal brain, 2 mls of GTC RNA lysis buffer was used (4M guanidium thiocyanate, 25 mM pH 7.0 sodium citrate, 100 mM betamercaptoethanol, 0.5% sarcosyl, 0.2M pH 4 sodium acetate, and 50% acidic phenol) and each brain was homogenized for 45 seconds at 18,000 rpm. Following homogenization, RNA was extracted with 0.2 volumes chloroform followed by isopropanol precipitation. RNA was resuspended in 10 mM Tris-EDTA. cDNA analysis cDNA was prepared from 5 ug of RNA treated with 2.5 ul of DNaseI (Invitrogen). Amplification was done using random primers (Invitrogen) and Superscript II reverse transcriptase (Invitrogen). Following cDNA synthesis, nested PCR was performed using 0.5 ul cDNA as template. First round PCR was done with primers PDS155 (59 -ATT CAC CGC TGC TGC TTA AA -39) and AKR1-KPK (59 -TAG GAA AAT GGC TCG GTG TC -39) for 19 cycles. 2 ul of first round PCR product was diluted into 18 ul of water, and 1.5 ul of this dilution was used as template for 35 cycles of second round PCR. Second round PCR primers were AKnewFWD (59 -CTT TCT CCA GCA ACC TAT C) and AKnewREV (59 -AAG GAC CTG CCG CTT AAC T -39). In each case, cycling conditions were 94uC for 30 seconds, 60uC for 1 minute, 72uC for 2 minutes. 10 ul of second round PCR product was digested 5 hours to overnight with 1U AluI (NEB). Digests were heat inactivated and run on a 3% agarose gel. Figure S1 Crossing scheme for developmental time point specific repeat deletions. To delete the DNA repeats as specific times after fertilization, we used an allele containing a loxP-flanked copy of the Rasgrf1 DNA repeats in combination with the Cre transgenic mice. To facilitate allele-specific expression analysis, we bred specific Cre alleles onto the PWK/PhJ mouse background. These mice (PWK Cre) were mated with mice homozygous for a loxP-flanked version of the Rasgrf1 DNA repeats (Floxed), which was created on the 129S4Jae strain background and backcrossed to C67BL/6. In the presence of Cre recombinase expression, the loxP-flanked repeats can be deleted at specific time points. Zp3 Cre is active at e0.0 and deletes the repeats at the one-cell stage, while Meox2 Cre is active at e5.5 and deletes the repeats in the embryonic ectoderm of the e5.5 epiblast around the time of implantation into the uterine wall. Depending on whether the loxP-flanked repeats were inherited maternally or paternally, we were able to delete the repeats at these time points on either the maternally or the paternally inherited allele. After genotyping to ensure that the animals were not mosaic for deletion of the DNA repeats, and to ensure that the animals carried one PWK and one 129S4Jae allele at the AK029869 locus, we carried out allelespecific expression analysis.

Supporting Information
Found at: doi:10.1371/journal.pone.0013784.s001 (0.42 MB TIF) Figure S2 The presence of an extra enhancer allele does not mask detection of expression from wild-type alleles in trans. cDNAs from a wildtype and an extra enhancer allele were mixed in varying ratios. The mixed cDNAs were subjected to allele specific expression analysis of AK029869 as described in the main text. Strains and band sizes are shown to the right (see main text, Figure 2A). In each case, even when the extra enhancer allele (enh) was present in a 3:1 ratio to the wild-type allele (WT), banding patterns from both alleles were present indicating that expression from the extra enhancer allele did not obscure expression from the wild-type allele during amplification. '' + '' = with reverse transcriptase, '' -'' = without reverse transcriptase, (-) = water, B6 = C57Bl6/J, PWK = PWK/PhJ. Found at: doi:10.1371/journal.pone.0013784.s002 (0.17 MB TIF)