An Unexpected Function of the Prader-Willi Syndrome Imprinting Center in Maternal Imprinting in Mice

Genomic imprinting is a phenomenon that some genes are expressed differentially according to the parent of origin. Prader-Willi syndrome (PWS) and Angelman syndrome (AS) are neurobehavioral disorders caused by deficiency of imprinted gene expression from paternal and maternal chromosome 15q11–q13, respectively. Imprinted genes at the PWS/AS domain are regulated through a bipartite imprinting center, the PWS-IC and AS-IC. The PWS-IC activates paternal-specific gene expression and is responsible for the paternal imprint, whereas the AS-IC functions in the maternal imprint by allele-specific repression of the PWS-IC to prevent the paternal imprinting program. Although mouse chromosome 7C has a conserved PWS/AS imprinted domain, the mouse equivalent of the human AS-IC element has not yet been identified. Here, we suggest another dimension that the PWS-IC also functions in maternal imprinting by negatively regulating the paternally expressed imprinted genes in mice, in contrast to its known function as a positive regulator for paternal-specific gene expression. Using a mouse model carrying a 4.8-kb deletion at the PWS-IC, we demonstrated that maternal transmission of the PWS-IC deletion resulted in a maternal imprinting defect with activation of the paternally expressed imprinted genes and decreased expression of the maternally expressed imprinted gene on the maternal chromosome, accompanied by alteration of the maternal epigenotype toward a paternal state spread over the PWS/AS domain. The functional significance of this acquired paternal pattern of gene expression was demonstrated by the ability to complement PWS phenotypes by maternal inheritance of the PWS-IC deletion, which is in stark contrast to paternal inheritance of the PWS-IC deletion that resulted in the PWS phenotypes. Importantly, low levels of expression of the paternally expressed imprinted genes are sufficient to rescue postnatal lethality and growth retardation in two PWS mouse models. These findings open the opportunity for a novel approach to the treatment of PWS.


Introduction
Genomic imprinting regulates gene expression only from one allele that is inherited either from the mother or from the father. Genomic imprinting is important as defects in this process often result in human diseases. Human chromosome region 15q11-q13 represents an imprinted domain referred as the PWS/AS domain, because paternal deletions cause Prader-Willi syndrome (PWS) and maternal deletions cause Angelman syndrome (AS) (for review [1]). The symptoms of PWS include neonatal feeding difficulties and hypotonia, morbid obesity developing in early childhood, and mild mental retardation. AS is characterized by ataxia, absence of speech, seizures, and mental retardation.
The PWS/AS imprinted domain contains a number of paternally expressed genes, including MKRN3, MAGEL2, NDN, C15ORF2, SNURF-SNRPN, and C/D box small nucleolar RNAs (snoRNAs) SNORD107, SNORD64, SNORD108, SNORD109A, SNORD116, SNORD115, and SNORD109B [1]. Mouse chromosome 7C has a conserved PWS/AS imprinted domain with exception of presence of Frat3 and absence of C15orf2, Snrod108, and Snord109a/b [2]. SNURF-SNRPN/Snurf-Snrpn (hereafter abbreviated SNRPN/Snrpn) encodes two different proteins within a single transcript [3]. Many upstream exons of SNRPN/Snrpn were identified [4,5,6]. With SNRPN/Snrpn exon 1 associated with the major promoter and upstream exons with weaker promoter activity, there are alternative transcripts starting from these SNRPN/Snrpn exons and span the UBE3A/Ube3a antisense transcript [7,8]. SnoRNAs are encoded within these large SNRPN sense/UBE3A antisense and Snrpn sense/Ube3a antisense transcripts derived from both SNRPN/Snrpn major and upstream alternative promoters. SNORD116/Snord116 and SNORD115/ Snord115 are present as multiple copy gene clusters. Among these paternally expressed imprinted genes, SNORD116 plays a major role in PWS etiology, because deficiency of this gene caused the key characteristics of the PWS phenotype in human [9,10,11]. Evidence from two mouse models with different targeted deletions of Snord116 supports the Snord116 function in early postnatal growth, motor learning, and feeding regulation [12,13]. Deficiency of other genes in the PWS/AS domain may also contribute to the PWS phenotype [2,14,15]. On the other hand, UBE3A is the AS gene and encodes E6-AP ubiquitin-protein ligase expressed preferentially from the maternal chromosome in brains [16,17]. Mutations of the Ube3a in mice resulted in the phenotype resembling human AS [18,19].
Imprinted genes at the PWS/AS domain are coordinately regulated through a cis-acting imprinting center that contains two functional elements, the PWS-IC and AS-IC. Numerous studies in humans patients and mouse models have led to the suggestion that on the paternal chromosome, the PWS-IC is a positive regulatory element required for establishment and maintenance of paternal imprinting [6,20,21], whereas on the maternal chromosome, the AS-IC is suggested to function in allele-specific repression of the PWS-IC to prevent a paternal imprinting program [22,23,24]. However, the equivalent of the human AS-IC element has not yet been identified in mice. Three mouse models for the PWS-IC deletions have been generated by the introduction of deletions at the Snrpn promoter. In first model, a deletion of a 35-kb fragment at the paternal Snrpn promoter led to a maternal pattern of DNA methylation and gene expression on the paternal chromosome, resulting in perinatal lethality [20]. In another study, paternal inheritance of a 4.8-kb deletion (D4.8) at the Snrpn exon 1 caused a mosaic imprinting defect, resulting in partial lethality and growth retardation [6]. In the third study, a 6-kb deletion extending 1 kb further upstream of the D4.8 region at the Snrpn promoter exhibits a complete PWS-IC deletion phenotype [25]. On the other hand, attempts to define the equivalent of the human AS-IC element by targeted mutations in mice so far have been unsuccessful [26,27]. However, an insertion/duplication mutation 13 kb upstream of Snrpn exon 1 resulted in an AS imprinting defect [26], and a targeted replacement of mouse PWS-IC with human PWS-IC caused a maternal imprinting defect with an AS phenotype [14].
In previous study, we have used the D4.8 mouse model to demonstrate the function of the PWS-IC in the regulation of paternal imprinting [6]. In current study, we used this mouse model to investigate the imprinting features on the maternal chromosome regulated by the PWS-IC. If the PWS-IC is required only for paternal imprinting and is suppressed on the maternal chromosome, then, deletion of the PWS-IC should not affect the maternal imprint. Unexpectedly, we found that maternal inheritance of the PWS-IC deletion resulted in a maternal imprinting defect accompanied by an altered maternal pattern of gene expression and epigenetic modifications, toward a paternal state. Our findings provide the first evidence that the PWS-IC is required for maternal imprinting in mice.

Results
Maternal inheritance of the PWS-IC D4.8 mutation activated the upstream alternative Snrpn promoter in cis To study the maternal imprinting features regulated by the PWS-IC, we investigated the maternal pattern of gene expression in the mouse model with the PWS-IC D4.8 mutation at Snrpn exon 1. Snrpn is paternally expressed from exon 1 with the major promoter and from alternative upstream exons with weaker promoter activity [6]. Although exon 1 of Snrpn was removed by the D4.8 mutation, Snrpn is still able to transcribe from the upstream exon promoter, initiating at alternative upstream exons splicing to Snrpn exon 2 [6] ( Figure 1A). Using RT-PCR and quantitative RT-PCR analyses, the alternative Snrpn transcripts could be detected by primers specific for upstream exon 1 and exon 3 (u1-ex3) to measure the weaker promoter activity ( Figure 1A and 1B), and by primers specific for the downstream exon 7 (ex7) to measure both major and weaker promoter activity ( Figure 1A and 1C). The analysis of maternal-specific expression of the Snrpn transcripts was accomplished in mice with paternal inheritance of a deletion spanning from exon 2 of Snrpn to Ube3a (DS-U) [28]. Because Snrpn (from exon 2 to exon 10) was removed by the DS-U mutation on the paternal chromosome, the Snrpn u1-ex3 and ex7 transcripts measured, if any, could only be expressed from the maternal chromosome ( Figure 1A). As the paternally expressed imprinted gene Snrpn was not expressed from the wildtype maternal chromosome in the m + p DS-U mice ( Figure 1B, 1C, and 1F, e, m + p DS-U ), it is surprising that Snrpn was partially expressed from the maternal D4.8 chromosome in the m D4.8 p DS-U mice ( Figure 1B and 1F, c, m D4.8 /p DS-U , 21% of the u1-ex3 transcripts; Figure 1C and 1F, c, m D4.8 /p DS-U , 35% of the ex7 transcripts), compared with that in the wild-type mice ( Figure 1B, 1C, and 1F, a, m + /p + ). These results suggested that maternal inheritance of the PWS-IC D4.8 mutation partially activated the paternally expressed imprinted gene Snrpn on the maternal chromosome. On the other hand, when the DS-U mutation was on the maternal chromosome, Snrpn is fully expressed from the wild-type paternal allele in the m DS-U p + mice ( Figure 1B and 1F, d, m DS-U /p + , 97% of the u1-ex3 transcripts; Figure 1C and 1F, d, m DS-U /p + , 108% of the ex7 transcripts), and is partially repressed on the paternal D4.8 chromosome in the m DS-U p D4.8 mice ( Figure 1B and 1F, b, m DS-U /p D4.8 , 39% of the u1-ex3 transcripts; Figure 1C and 1F, b, m DS-U /p D4.8 , 27% of the ex7 transcripts), consistent with the previous report [6].
The paternally expressed imprinted genes Snord116 and Snord115 are encoded within the large Snrpn sense/Ube3a antisense transcripts whose expression is driven by the Snrpn promoter ( Figure 1F, top). To further confirm activation of the Snrpn promoter on the maternal D4.8 chromosome, we examined expression of Snord116 and Snord115. Since both Snord116 and Snord115 were also deleted by the DS-U mutation, maternalspecific expression of Snord116 and Snord115 was analyzed in mice with paternal inheritance of the DS-U mutation, so that the detected Snord116 and Snord115 transcripts could only be from the maternal chromosome. We found that Snord116 and Snord115 was not expressed from the maternal wild-type chromosome in the m + p DS-U mice ( Figure 1D-F, e, 0%), but a small amount of the Snord116 and Snord115 transcripts was expressed from the maternal D4.8 chromosome in the m D4.8 p DS-U mice ( Figure 1D and 1F, c, m D4.8 /p DS-U , 8% of the Snord116 transcripts; Figures 1E and 1F, c, m D4.8 /p DS-U , 10% of the Snord115 transcripts). These suggested that maternal inheritance of the D4.8 mutation partially activated the Snrpn promoter resulting in expression of Snord116 and Snord115. On the other hand, Snord116 and Snord115 were also partially expressed from the paternal D4.8 chromosome in the m DS-U p D4.8 mice ( Figure 1D and 1F, b, m DS-U /p D4.8 , 12% of the Snord116 transcripts; Figure 1E and 1F, b, m DS-U /p D4.8 , 14% of the Snord115 transcripts).
Activation of the Snrpn promoter by maternal inheritance of the D4.8 mutation was demonstrated not only when paternal inheritance of the DS-U mutation in the m D4.8 /p DS-U mice but also when paternally inheriting the wild-type allele in the m D4.8 /p + mice, since there were significant increases of the Snrpn u1-ex3, Snrpn ex7, Snord116, and Snord115 transcripts in the m D4.8 /p + mice compared with those in wild-type mice ( Figure S1).
The paternally expressed imprinted gene Ndn was expressed from the maternal chromosome carrying the D4.8 mutation The PWS-IC plays a dual role as the Snrpn promoter and as an IC in the PWS/AS region [6]. Given that the PWS-IC D4.8 mutation affected the maternal imprinting of the adjacent Snrpn upstream promoter (Figure 1), we investigated whether the D4.8 mutation also perturbs its IC function in maternal imprinting. Transcription of a paternally expressed imprinted gene Ndn was analyzed as an index of the extent of any affected gene expression within the PWS/AS region, since the Ndn locus is located about 1 Mb upstream of the Snrpn promoter. Maternal-specific expression of the Ndn transcripts was analyzed in mice with paternal inheritance of a deletion on Ndn (DNdn) [29]. In these mice, any detected Ndn transcripts were expressed exclusively from the maternal chromosome and not from the paternal DNdn chromo-some, since the primer pair used for RT-PCR and qRT-PCR analyses is located at the region deleted in the DNdn mutation ( Figure 2A). Our results showed that Ndn was not transcribed from the maternal wild-type chromosome in the m + p DNdn mice ( Figure 2B and 2D, b), but was partially expressed from the The Snrpn sense/Ube3a antisense transcripts initiated from Snrpn exon 1 with the major promoter activity and from Snrpn upstream exons with weaker promoter activity are marked as bold and thin arrows, respectively. SnoRNAs are encoded within these large Snrpn sense/Ube3a antisense transcripts derived from both Snrpn major and upstream exon promoters. Snord116 and Snord115 are multiple copy gene clusters. The centromeric (cen) and the telomeric (tel) positions are indicated. Paternally and maternally expressed genes are marked as blue and red boxes, respectively. DS-U indicates a large deletion from Snrpn exon 2 to Ube3a. D4.8 indicates a 4.8-kb deletion at Snrpn exon 1. The levels of the Snrpn u1-ex3, Snrpn exon 7, Snrod116, and Snord115 transcripts from wild-type mice were set as 100%. Mat, maternal chromosome; Pat, paternal chromosome. doi:10.1371/journal.pone.0034348.g001 maternal D4.8 chromosome in the m D4.8 p DNdn mice ( Figure 2B and 2D, c, 28%). These results suggested that maternal inheritance of the D4.8 mutation partially activated the paternally expressed imprinted gene Ndn on the maternal chromosome. Therefore, the maternal PWS-IC D4.8 mutation disturbed not only the maternal imprint of the adjacent Snrpn upstream promoter but also the maternal imprint of the Ndn promoter which is 1 Mb away from the D4.8 region.
In addition, the abundance of the Ndn transcripts was reduced when the D4.8 mutation was on the paternal chromosome ( Figure 2C and 2D, d, m + /p D4.8 , 27%), as previously reported [6]. Importantly, mice with the D4.8 mutation on both maternal and paternal chromosomes expressed a level of the Ndn transcripts comparable with that in the wild-type controls ( Figure 2C and 2D, a, m + /p + , 100%; f, m D4.8 /p D4.8 , 101%). These results suggested that maternal inheritance of the D4.8 mutation compensated the loss of Ndn expression due to paternal inheritance of the D4.8 mutation. Accordingly, we found a small but consistent increase in Ndn expression in mice with only the maternally inherited D4.8 mutation ( Figure 2C and 2D, e, m D4.8 /p + , 120%). Expression of the maternally expressed imprinted gene Ube3a was reduced when the D4.8 mutation was on the maternal chromosome While the paternally expressed imprinted genes were activated, it was of interest to examine whether the maternally expressed imprinted gene was repressed. Ube3a, known as the AS gene, encodes E6-AP ubiquitin-protein ligase whose expression derives preferentially from the maternal chromosome in brains [16,17], and is negatively regulated by the paternal expressed Snrpn sense/ Ube3a antisense transcripts derived from the Snrpn promoter [14,30,31]. Since the maternal Snrpn promoter was activated by maternal inheritance of the D4.8 mutation (Figure 1), we found a decreased level of E6-AP in the m D4.8 p + mice ( Figure 3A and 3B, d, 78% of E6-AP), compared with that in wild-type mice ( Figure 3A and 3B, a, 100% of E6-AP). Maternal-specific reduction of Ube3a expression by maternal inheritance of the D4.8 mutation was further analyzed in mice with paternal inheritance of the DS-U mutation. Because Ube3a was deleted by the DS-U mutation on the paternal chromosome, Ube3a could only express from the maternal chromosome. Compared to the maternal wild-type chromosome in the m + p DS-U mice ( Figure 3A and 3B, b, 72% of E6-AP), the maternal D4.8 chromosome expressed a reduced level of E6-AP in the m D4.8 p DS-U mice ( Figure 3A and 3B, c, 46% of E6-AP). These results suggested that maternal expression of Ube3a was partially repressed by maternal inheritance of the D4.8 mutation.
Together, we found that maternal inheritance of the PWS-IC D4.8 mutation changed the maternal pattern of gene expression toward a paternal state: the paternally expressed imprinted genes were partially activated and the maternally expressed imprinted genes were partially repressed on the maternal D4.8 chromosome.
Maternal inheritance of the D4.8 mutation complemented a postnatal lethality phenotype in PWS mouse models paternally inheriting the D4.8 or DS-U mutations The functional significance of the acquired paternal pattern of gene expression on the maternal chromosome was further investigated by genetic complementation experiments in two PWS mouse models paternally inheriting the D4.8 or DS-U mutations. The D4.8 mutation did not affect survival when inherited maternally (Table 1, mating II), but caused postnatal lethality in 56% of the mice when inherited paternally (Table 1, mating III), as previously reported [6]. Interestingly, there was an almost complete rescue of the lethality in mice inheriting the D4.8 mutation from both parents, with 96% survival rate for mice observed up to 2 months of age (Table 1, mating IV). In addition, while all mice with only paternal inheritance of the DS-U mutation died (Table 1, mating V), double heterozygous mice inheriting the DS-U mutation paternally and the D4.8 mutation maternally survived close to the expected Mendelian ratios (Table 1, mating VI). These results suggested that maternal inheritance of the D4.8 Maternal inheritance of the D4.8 mutation complemented a growth retardation phenotype in PWS mouse models paternally inheriting the D4.8 or DS-U mutations Paternal inheritance of the D4.8 or DS-U mutations resulted in not only postnatal lethality but also growth retardation in surviving mice [6,28]. The heterozygous pups paternally inheriting the D4.8 mutation were smaller compared to the age-matched wild-type mice ( Figure 4A, m + p + and Figure 4B, m + p D4.8 ), as previously reported [6]. Interestingly, homozygous pups inheriting the D4.8 mutation from both parents had an average body size indistinguishable from age-matched wild-type mice ( Figure 4A, m + p + , and Figure 4C, m D4.8 p D4.8 ). Furthermore, the m D4.8 p DS-U double heterozygous pups obtained by mating female mice carrying the D4.8 mutation to male mice carrying the DS-U mutation have body size similar to age-matched wild-type littermates and littermates with only the maternally inherited D4.8 mutation (m D4.8 p + ) ( Figure 4A and 4D). Measurements of body weight from groups of mice with those five different genotypes for up to 6 weeks clearly showed that maternal inheritance of the D4.8 mutation complements a growth retardation phenotype caused by paternal inheritance of the D4.8 or DS-U mutations ( Figure 4E).
Maternal inheritance of the D4.8 mutation altered histone modifications at the Snrpn and Ndn promoters Parent-of-origin specific epigenetic modifications on the PWS-IC correlate with transcriptional status and parent-of-origin specific epigenotypes of the imprinted genes spread over the PWS/AS domain. Given the paternal pattern of gene expression on the maternal D4.8 chromosome (Figures 1-3), we determined whether epigenetic status at the PWS/AS domain was altered by maternal inheritance of the D4.8 mutation. First, we analyzed histone modification profiles within the PWS/AS imprinted domain by ChIP-on-chip assays that combined chromatin immunoprecipitation to identify regions enriched with trimethylation on histone H3 lysine 4 (H3K4me3) and mouse genomic tiling array including sequences of the PWS/AS region from Peg12 to ATP10A (Chr7: 64,846,543-69,740,076) ( Figure 5A). Within the Snrpn locus, wild-type mice displayed a cluster of ChIP peaks surrounding the Snrpn promoter with the highest H3K4me3 enrichment located within a region corresponds to the D4.8 mutation ( Figure 5B and 5D, m + p + , Snrpn peak 1). Surprisingly, while paternal inheritance of the D4.8 mutation resulted in disappearance of the entire ChIP cluster ( Figure 5B, m + p D4.8 ), maternal inheritance of the D4.8 mutation in the m D4.8 p D4.8 mice partially restored H3K4me3 modification with the highest peak shifted to the right upstream of the D4.8 region ( Figure 5B and 5E, m D4.8 p D4.8 , Snrpn peak 2). Accordingly, the m D4.8 p + mice had a combinatorial pattern of the m D4.8 p D4.8 and wild-type mice showing two ChIP peaks with high H3K4me3 enrichment ( Figure 5B, 5D, and 5E, m D4.8 p + , Snrpn peak 1 and 2). Notably, when maternal inheritance of the D4.8 mutation, the acquired H3K4me3 enrichment (Snrpn peak 2) was located within 1 kb further upstream of the D4.8 region ( Figure 5B, m D4.8 p D4.8 and m D4.8 p + , Snrpn peak 2), where the full PWS-IC functional element must include with the D4.8 region [25].
Within the Ndn locus, paternal inheritance of the D4.8 mutation in the m + p D4.8 mice reduced overall amplitude of H3K4me3 by half, whereas maternal inheritance of the D4.8 mutation in the m D4.8 p D4.8 mice restored the amplitude of H3K4me3 to the extent comparable with that in wild-type mice ( Figure 5C and 5F), suggesting that maternal inheritance of the D4.8 mutation increased H3K4me3 at Ndn. Since H3K4me3 is a mark for an active state of gene expression, these findings are consistent with the results that expression of Ndn was partially inhibited in the m + p D4.8 mice, but was restored to a normal level in the m D4.8 p D4.8 mice ( Figure 2C), due to activation of Ndn on the maternal D4.8 chromosome ( Figure 2B).
The alteration of H3K4me3 modification was further confirmed by ChIP combined with quantitative PCR analysis (ChIP-qPCR). We designed a primer pair to amplify the Snrpn promoter right upstream of the D4.8 region for qPCR ( Figure 6A). The qPCR fragment includes the Snrpn peak 2 region shown in Figure 5B. Another primer pair was designed to amplify the Ndn locus including the Ndn peak 1 region shown in Figure 5C (Figure 6B). ChIP-qPCR analyses showed a marked reduction of H3K4me3 at both Snrpn and Ndn in the m + p D4.8 mice when compared with those in wile-type mice, whereas the m D4.8 p + mice had significant increases in H3K4me3 ( Figure 6C and 6D). These results support ChIP-on-chip analyses which showed reductions of H3K4me3 enrichment at Snrpn and Ndn when paternal inheritance of the D4.8 mutation and increases of H3K4me3 when maternal inheritance of the D4.8 mutation ( Figure 5). To determine whether the effect of maternal inheritance of the D4.8 mutation is on the maternal allele, we used mice with paternal inheritance of the DNdn mutation. In this assay, only the wild-type Ndn allele on the maternal chromosome could be detected because the primer pair designed for ChIP-qPCR is located at the region deleted in the DNdn mutation ( Figure 6B). Compared with wild-type mice, the m + p DNdn mice showed a dramatic reduction of H3K4me3 ( Figure 6E). This result clearly indicates paternal-specific H3K4me3 at Ndn and paternal deletion contributes to the significant reduction of H3K4me3. This is similar to the human NDN promoter showing paternal-specific association with H3K4me3 [32]. When maternally inheriting the D4.8 mutation, an elevated level of H3K4me3 was detected in the m D4.8 p DNdn mice compared with the m + p DNdn mice ( Figure 6E). Since the paternal copy of Ndn was deleted in these mice, the elevated H3K4me3 was derived from the remaining maternal copy of Ndn in the m D4.8 p DNdn mice. We next examined acetylation of histone 3 (H3Ac) as an additional marker of an active gene expression state. ChIP-qPCR analyses showed reductions of H3Ac at both Snrpn and Ndn in the m + p D4.8 mice ( Figure 6F and 6G, respectively). On the other hand, maternal inheritance of the D4.8 mutation did not affect H3Ac at the Snrpn promoter in the m D4.8 p + mice ( Figure 6F), but did increase H3Ac at Ndn ( Figure 6G). When the paternal Ndn was deleted in the m + p DNdn mice, a marked reduction of H3Ac was observed ( Figure 6H), suggesting the paternal copy is the one preferentially modified with H3Ac. This is similar to the human NDN promoter showing paternal bias with H3Ac [32]. Similar to increased H3K4me3 ( Figure 6E), m D4.8 p DNdn mice showed an increase of H3Ac on the maternal copy of Ndn compared to the m + p DNdn mice ( Figure 6H).
Finally, we examined H3K9me3 which is a mark of a repressive chromatin state. ChIP-qPCR analyses showed that the m + p D4.8 mice had marked increases in H3K9me3 at both Snrpn and Ndn, whereas the m D4.8 p + mice showed reductions of H3K9me3 compared with wild-type mice ( Figure 6I and 6J). Furthermore, the m D4.8 p DNdn mice showed the reduction of H3K9me3 on the maternal copy of Ndn when compared with the m + p DNdn mice ( Figure 6K). There was approximately 2-fold enrichment of H3K9me3 in the m + p DNdn mice compared with wild-type mice ( Figure 6K). It should be noted that in ChIP-qPCR analysis, the level of ChIP was normalized against the level of input in each sample: ChIP from the m + p DNdn mice was normalized against the input with only one copy of the maternal Ndn allele, while ChIP from the m + p + mice was normalized against the input with two Ndn copies from both parents. It is possible that the maternal copy of Ndn could be preferentially modified with H3K9me3, which is similar to the human NDN promoter with H3K9me3 towards maternal bias [32]. Therefore, after normalization with input, the ChIP-qPCR result might show a 2-fold enrichment of H3K9me3 in the m + p DNdn mice compared with the m + p + mice, even though both mice could have similar levels of H3K9me3 enrichment on their maternal wild-type copies of Ndn. However, we can not rule out the possibility that paternal inheritance of the DNdn mutation acts in trans to increase H3K9me3 on the maternal chromosome.
Maternal inheritance of the D4.8 mutation altered DNA modification at Ndn and Mkrn3 Next, we analyzed whether maternal inheritance of the D4.8 mutation affects the DNA methylation status at the PWS/AS domain. Silencing of the maternal alleles of Snrpn, Ndn, and Mkrn3 is associated with maternal-specific CpG methylation on their promoters [33,34,35,36]. Since the D4.8 mutation deletes the CpG island at the Snrpn promoter, we examined CpG methylation at the Ndn and Mkrn3 loci. Consistent with an earlier report [6], Southern blot analysis using the methylation-sensitive SacII enzyme revealed that CpG methylation on the SacII site at the Ndn locus was not affected by maternal inheritance of the D4.8 mutation ( Figure S2). However, use of a more sensitive analysis with sodium bisulfite sequencing revealed a lesser degree of methylation of the 42 CpGs on the Ndn promoter in the m D4.8 p + mice compared with that in wild-type mice ( Figure 7B, top). Since CpG methylation at Ndn is maternal-specific [35], we further used mice with paternal inheritance of the DNdn mutation to demonstrate maternalspecific reduction of CpG methylation. To enable detection of methylation only on the maternal Ndn allele, the reverse primers used in PCR to amplify the bisulfite-treated DNA were positioned within the region deleted in the DNdn mutation ( Figure 7A). Sodium bisulfite sequencing analyses showed decreased CpG methylation at the Ndn promoter on the maternal D4.8 chromosome in the m D4.8 p DNdn mice ( Figure 7B, bottom, right), when compared to the maternal wild-type chromosome in the m + p DNdn mice ( Figure 7B, bottom, left). These results suggest that maternal inheritance of the D4.8 mutation decreased CpG methylation on the maternal Ndn allele.
Finally, methylated DNA immunoprecipitation with 5-methylcytidine specific antibody (MeDIP) followed by quantitative PCR analysis (MeDIP-qPCR) confirmed a reduction of methylated DNA in mice with maternal inheritance of the D4.8 mutation (m D4.8 p + ) ( Figure 7C, left). In contrast, an increase of methylated DNA was found in mice with paternal inheritance of the D4.8 mutation (m + p D4.8 ) ( Figure 7C, left), which is consistent with the previous report [6]. Using the primer pair located at the region deleted in the DNdn mutation for MeDIP-qPCR analyses ( Figure 7A), we confirmed maternal inheritance of D4.8 mutation contributes to reduction of DNA methylation at the maternal Ndn allele, when the m D4.8 p DNdn mice was compare with the m + p DNdn mice ( Figure 7C, right).
In addition to Ndn, we found that Mkrn3 locus exhibited a similar alteration of DNA methylation by sodium bisulfite sequencing and MeDIP-qPCR analyses ( Figure 7E and 7F), despite Southern blot analysis from an earlier report showed CpG methylation on the NotI site was not affected by maternal inheritance of the D4.8 mutation [6]. When sodium bisulfite sequencing were used to analyze 22 CpG sites at the Mkrn3 promoter ( Figure 7D), the m D4.8 p + mice showed a lesser degree of CpG methylation when compared with that in wild type mice ( Figure 7E). Similarly, MeDIP-qPCR analyses demonstrated a reduction of methylated DNA in the m D4.8 p + mice ( Figure 7F). In contrast, an increase of methylated DNA was found in the m + p D4.8 mice ( Figure 7F), which is consistent with the previous report [6]. These results suggested that maternal inheritance of the D4.8 mutation decreased DNA modification on Ndn and Mkrn3.
Taken together, maternal inheritance of the D4.8 mutation have a role in controlling allelic differential modifications at Ndn with increased H3K4me3 and H3Ac, decreased H3K9me3, and reduced DNA methylation, by which the maternal allele is changed toward a more paternal epigenotype. This was correlated with activation of the paternally expressed imprinted gene Ndn on the maternal chromosome by maternal inheritance of the D4.8 mutation (Figure 2).

Discussion
Studies of the PWS-IC deletions in human patients and mouse models have suggested that the PWS-IC positively regulates paternal-specific gene expression and is responsible for establishment and maintenance of the paternal imprint [6,20,21]. On the other hand, it has been proposed that the AS-IC functions in allele-specific repression of the PWS-IC to prevent the paternal program on the maternal chromosome, and this interpretation is based on genetic analysis showing that the maternal AS-IC imposes a silent epigenetic state on the neighboring SNRPN/Snrpn promoter [22,23,24]. However, the equivalent of the human AS-IC element has not been identified in mice. In contrast to the current hypothesis that the PWS-IC as a positive regulator for the paternal imprinting program must be repressed on the maternal chromosome, our findings suggest another dimension that the PWS-IC is also required for a maternal chromosome to have the maternal pattern of gene expression. The maternal chromosome carrying the PWS-IC D4.8 mutation failed to properly silence the paternal imprinting program, suggesting that the maternal PWS- IC negatively regulates the paternally expressed imprinted genes, in stark contrast to its known function on the paternal chromosome as a positive regulator for paternal-specific gene expression.
The maternally expressed gene UBE3A/Ube3a is the AS gene and is negatively regulated by the paternal expressed SNRPN sense/UBE3A antisense and Snrpn sense/Ube3a antisense transcripts derived from the SNRPN and Snrpn promoters, respectively [14,30]. On the wild-type maternal chromosome, silencing of the Snrpn promoter results in expression of Ube3a ( Figure 8A, c). Previously, we demonstrated that maternal transmission of an insertion/duplication mutation 13 kb upstream of Snrpn exon 1 (AS-IC an , an anchor mutation on the AS-IC) activates the Snrpn promoter, resulting in severely decreased expression of Ube3a, causing AS phenotypes [26] (Figure 8A, e). In this report, we found that when the main Snrpn promoter was deleted, the maternal PWS-IC D4.8 mutation activates the weaker upstream alternative Snrpn promoter and expresses a low level of the Snrpn sense/Ube3a antisense transcripts, resulting in mild reduction of Ube3a expression ( Figure 8A, d). Phenotype effects of the D4.8 mutation are being studied further for the symptoms of AS. In both cases of the D4.8 mutation and the AS-IC an mutation, activation of paternally expressed imprinted genes on the maternal chromosome leads to the ability to complement the lethality and growth retardation phenotypes in mouse models of PWS. In addition, the acquisition the paternal gene expression pattern was correlated with alteration of DNA methylation on the maternal chromosome toward to a more paternal epigenotype: the AS-IC an mutation causes loss of Snrpn methylation and decreased Ndn methylation ( Figure 8A, e) and the D4.8 mutation causes decreased Ndn methylation on the maternal chromosome ( Figure 8A, d), while the Ndn and Snrpn promoters are fully methylated on the maternal wild-type chromosome ( Figure 8A, c) The PWS-IC has a dual function, one as the Snrpn promoter and the other as an IC regulator of the PWS/AS domain. Maternal transmission of a targeted replacement of mouse PWS-IC with human PWS-IC (PWS-IC Hs ) expressed the Snrpn sense/Ube3a antisense transcripts from the inserted human SNRPN promoter, but did not affect any other paternally expressed imprinted transcripts on the maternal chromosome ( Figure 8A, f) [14], suggesting that the IC function was not lost. In our mouse model, maternal inheritance of the PWS-IC D4.8 mutation disrupts not only maternal imprinting of Snrpn but also maternal imprinting of Ndn which is 1 Mb away from the D4.8 region ( Figure 8A, d), suggesting that this D4.8 mutation perturbs the IC function on the maternal imprint at the PWS/AS region. In addition, maternal inheritance of the PWS-IC Hs rescues lethality in a PWS mouse model inheriting the PWS-IC 35-kb deletion (PWS-IC del ) paternally, but the PWS-IC Hs/del mice still have a growth deficiency [14,20]. Maternal inheritance of the PWS-IC D4.8 mutation rescues both lethality and growth retardation phenotypes in PWS mouse models. The lethality and growth retardation phenotypes seem to correlate with the dual role of the PWS-IC as the Snrpn promoter and as the IC regulator for imprinted genes at the PWS/AS domain. Mouse models of PWS have failure to thrive which results in postnatal lethality and growth retardation. Maternal expression of the Snrpn sense/Ube3a antisense transcripts from the inserted human SNRPN promoter complements one failure to thrive locus to rescue lethality, but is not able to complement a second failure to thrive locus which contributes to a growth retardation phenotype [14]. In our mouse model, maternal inheritance of the PWS-IC D4.8 mutation perturbs the IC function of the maternal imprint at the PWS/AS region, and thereby activates the paternally expressed imprinted genes spread over the PWS/AS domain, which could complement all of the failure to thrive loci, resulting in rescues of both lethality and growth retardation phenotypes in PWS mouse models. Furthermore, maternal inheritance of the PWS-IC D4.8 mutation only caused a mild reduction of Ube3a expression ( Figure 8A, d), whereas maternal inheritance of the PWS-IC Hs resulted in severely decreased expression of Ube3a ( Figure 8A, f), leading to AS phenotypes [14].
On the paternal wild-type chromosome, the Snrpn and Ndn promoters are unmethylated and the paternally expressed imprinted genes are fully activated ( Figure 8A, a). When paternal inheritance of the D4.8 mutation deletes the Snrpn exon 1, the Snrpn sense/Ube3a antisense partially transcribes from the upstream exons ( Figure 8A, b). Although mice with paternal inheritance of the D4.8 mutation expressed relatively similar levels of the Snrpn, Snord116, and Snord115 transcripts as mice with maternal inheritance of the D4.8 mutation ( Figure 8A, b and d), different phenotypic effects of the D4.8 mutation were found depending on the origin of inheritance: paternal transmission of the D4.8 mutation caused PWS phenotypes showing postnatal lethality and growth retardation [6]; maternal inheritance of the D4.8 mutation is able to complement postnatal lethality and growth retardation phenotypes in the PWS mouse models. These results raise a possibility that gene(s) other than Snord116, and Snord115, and Snrpn are also involved in these PWS phenotypes, although deficiency of SNORD116 in human or Snord116 in mice has been demonstrated to contribute to PWS pathogenesis [9,10,11,12,13]. This hypothesis is also supported by analyses of several mouse models for PWS [2,15]. We noticed that the above parent-oforigin specific effects of the D4.8 mutation appeared to correlate with the levels of the Ndn transcripts, in the condition with Snrpn, Snord116, and Snord115 expressed only from the upstream alternative Snrpn promoters. Specifically, the m + /p D4.8 mice showed partial expression of Ndn with 50% lethality and growth retardation [6], whereas the m D4.8 /p DS-U mice could express normal or increased levels of total Ndn transcripts and appeared Figure 6. ChIP-qPCR analyses for H3K4me3, H3Ac, and H3K4me3 at Snrpn and Ndn. (A) Schematic diagram of the Snrpn promoter. Gene structure is shown at the top, where the white box represents the Snrpn exon 1 with the +1 as the major transcriptional start site. The region deleted in the D4.8 mutation started from 22,702 is indicated as a gray line. A primer pair (half-arrows) was used for qPCR to amplify the Snrpn promoter from 22,919 to 22,705 right upstream of the D4.8 region (black box). This qPCR region includes the Snrpn peak 2 region shown in Figure 5B. (B) Schematic diagram of Ndn. The gene structure is shown at the top, where the white box represents the Ndn exon with the +1 as the transcriptional start site. The region deleted in the DNdn mutation started from +228 is indicated (gray line). A primer pair (half-arrows) was used for qPCR to amplify the region from +773 to +976 (black line). This qPCR region includes the Ndn peak 1 region shown in Figure 5C. (C-E) Quantification of H3K4me3 at Snrpn (C) and Ndn (D, E) in the wild-type m + p + mice (C-E), the m D4.8 p + mice (C, D), the m + p D4.8 mice(C, D), the m + p DNdn mice (E), and the m D4.8 p DNdn mice (E). (F-H) Quantification of H3Ac at Snrpn (F) and Ndn (G, H) in the wild-type m + p + mice (F-H), the m D4.8 p + mice (F, G), the m + p D4.8 mice (F, G), the m + p DNdn mice (H), and the m D4.8 p DNdn mice (H). (I-H) Quantification of H3K9me3 at Snrpn (I) and Ndn (J, K) in the wild-type m + p + mice (I-H), the m D4.8 p + mice (I, J), the m + p D4.8 mice(I, J), the m + p DNdn mice (K), and the m D4.8 p DNdn mice (K). The level of ChIP was normalized against the level of input in each sample. The normalized level of ChIP from the wild-type mouse was set as 1. m + p + , n = 6; m + p D4.8 , n = 4; m + p D4.8 , n = 4; m + p DNdn , n = 3; m D4.8 p DNdn , n = 3. doi:10.1371/journal.pone.0034348.g006 Although these results suggest that Ndn is also a potential candidate gene responsible for the PWS phenotypes, it should be pointed out that targeted deletions of Ndn in mice had reported contradictory results, ranging from no to severe effects on lethality [29,37,38]. The reason for the differences is not clear, genetic backgrounds are suspected to be a contributing factor. However, growth retardation has not been reported in surviving mice with Ndn deficiency [29,37,38]. Loss of another gene or more than one gene regulated by the maternal PWS-IC might contribute to the lethality and growth retardation phenotype. Two mouse models with different targeted mutations of Magel2 have been created. The first study indicated reduced embryonic viability and postnatal growth retardation from birth until weaning [39]. The second study showed neonatal lethality (around 50%) and postnatal growth retardation due to the suckling deficit [40].
The partial imprinting defect caused by maternal or paternal inheritance of the PWS-IC D4.8 mutation indicates that one or more elements outside the D4.8 region are additionally required for full PWS-IC activity [6]. Recently, paternal transmission of a deletion extended 1 kb further upstream of the D4.8 region results in fully penetrant imprinting defects, suggesting that this 1-kb interval contains functional elements that confer full PWS-IC activity with the D4.8 region [25]. We found that maternal inheritance of the D4.8 mutation obtained H3K4me3 enrichment and reduced H3K9me3 located within this 1-kb region just upstream of the D4.8 mutation. These epigenetic changes are being studied further for their parent of origin and function as a potential IC or promoter. If present only on the maternal allele, it is possible that the maternal PWS-IC D4.8 mutation could activate the remaining portion of the PWS-IC by creating an active chromatin hub on the maternal chromosome. Thereby, partial expression of the paternally expressed imprinted genes on the maternal chromosome could be due to activation of this potential IC element or be the direct effects of a partial loss of the PWS-IC by the D4.8 mutation. On the other hand, when paternally inheriting the D4.8 mutation, H3K4me3 enrichment is not present at the PWS-IC, although the paternally expressed imprinted genes are also partially expressed with the remaining portion of the PWS-IC.
Together, our findings provide evidence for the first time that the PWS-IC functions not only in paternal imprinting but also in maternal imprinting at the PWS/AS domain in mice ( Figure 8B). The PWS-IC controls expression of imprinted genes accompanied by parent-specific epigenetic modifications. On the paternal chromosome, the PWS-IC positively regulates the paternally expressed imprinted genes with the Snrpn and Ndn promoters are unmethylated on the CpG islands and modified with active chromatin marks H3K4me3 and H3Ac [41] (Figure 8B). On the maternal chromosome, the PWS-IC represses expression of the paternally expressed imprinted genes with the Snrpn and Ndn promoters are methylated on the CpG islands and modified with a repressive chromatin mark H3K9me3 ( Figure 8B). Furthermore, we demonstrated PWS phenotypic rescues by maternal inheritance of the PWS-IC deletion, in contrast to paternal inheritance of the PWS-IC deletion causing the PWS phenotypes. We identified a previously unrecognized and important role of the PWS-IC at the PWS/AS domain.

Ethics statement
All of the mice were bred and maintained according to a protocol (protocol number: AN772) approved by the Baylor College of Medicine Animal Care and Use Committee at the institution's specific pathogen-free mouse facility, which is approved by the American Association for Accreditation of Laboratory Animal Care and is operated in accordance with current regulations and standards of the US Department of Agriculture and the Department of Health and Human Services.

Mouse models
Mutant mice with a 4.8-kb deletion (D4.8) at Snrpn exon 1were generated as described [6]. Mutant mice carrying a deletion from Snrpn exon 2 to Ube3a (DS-U) were previously described [28]. Mutant mice with a deletion at Ndn (DNdn) were described [29]. Mice with the D4.8 mutation and mice with the DS-U mutation are maintained on a C57BL6/J genetic background. Mice with the DNdn mutation are maintained on a hybrid C57BL6/J and 129/ SvEv genetic background.

RT-PCR analysis
Total RNA was purified from mouse whole brain dissected from pups at day 1 of age using an RNeasy plus kit (Qiagen, Hilden, Germany). 2 mg of DNase I-treated total RNA was used for reverse transcription to synthesize the first strand cDNA by the Superscript III First-strand synthesis system (Invitrogen, Carlsbad, CA). Quantitative RT-PCR analysis was performed using Light-Cycler Fast-Start DNA Master SYBR Green I (Roche). PCR conditions and primer sequences are listed in Table S1 [6,13,26,42]. Hprt and Gapdh transcripts were amplified as controls for gene expression. For quantification experiments, we used a least 3 sets of mice with every genotypes. In each group of mice with different genotypes, the levels of gene expression were normalized against the levels of an endogenous control in each (half-arrows) were used for nested PCR to amplify the bisulfite-treated DNA at the Ndn promoter from 269 to +470 (black line). A third primer pair (half-arrows) was used for the MeDIP-qPCR analysis to amplify the region from +237 to +470 (black line). (B) Sodium bisulfite sequencing analyses of methylation status of 42 CpG dinucleotides across the Ndn promoter (269 to +470) in the wild-type m + p + mice, the m D4.8 p + mice, the m + p DNdn mice, and the m D4.8 p DNdn mice. Each line represents an individual clone with open and closed circles corresponding to unmethylated and methylated CpGs, respectively. (C) MeDIP-qPCR analyses of DNA methylation at the Ndn promoter (+237 to +470) in the wild-type m + p + mice, the m D4.8 p + mice, and the m + p D4.8 mice (left), as well as in the m + p DNdn mice and the m D4.8 p DNdn mice (right). The level of MeDIP DNA was normalized against the level of input DNA in each sample. m + p + , n = 3; m + p D4.8 , n = 3; m + p D4.8 , n = 3; m + p DNdn , n = 3; m D4.8 p DNdn , n = 3. (D) Schematic diagram of the Mkrn3 promoter. Gene structure is shown at the top, where the white box represents the partial Mkrn3 exon with the +1 as the transcriptional start site. A NotI site at +139 is indicated. Locations of CpG dinucleotides are shown as vertical bars. Two primer pairs (half-arrows) were used for nested PCR to amplify the bisulfitetreated DNA at the Mkrn3 promoter from 2469 to +91 (black line). A third primer pair (half-arrows) was used for the MeDIP-qPCR analysis to amplify the region from +21 to +324 (black line). (E) Sodium bisulfite sequencing analyses of methylation status of 22 CpG dinucleotides across the Mkrn3 promoter (2469 to +91) in the wild-type m + p + mice and the m D4.8 p + mice. Each line represents an individual clone with open and closed circles corresponding to unmethylated and methylated CpGs, respectively. (F) MeDIP-qPCR analyses of DNA methylation at the Mkrn3 promoter (+21 to +324) in the wild-type m + p + mice, the m D4.8 p + mice, and the m + p D4.8 mice. The level of MeDIP DNA was normalized against the level of input DNA in each sample. The normalized level of MeDIP DNA from the wild-type mouse was set as 1. m + p + , n = 3; m + p D4.8 , n = 3; m + p D4.8 , n = 3. doi:10.1371/journal.pone.0034348.g007 sample. In each set of experiments, the normalized level of gene expression from the wild-type mouse was always set as 1.

Western blot analysis
Mouse whole brain was dissected from pups at day 1 of age. Brain samples were homogenized and lysed in NP40/SDS buffer {1% Nonidet P-40, 0.01% SDS, 0.1 M Tris-HCl, pH 7.2, and complete Protease Inhibitor Cocktail Tablet (Roche Applied Sciences, Indianapolis IN)}. Sixty micrograms of mouse brain protein were used for electrophoresis on 10% Tris-Cl ready gels (Bio-Rad, Hercules CA). The proteins were transferred to nitrocellulose membrane (Bio-Rad). The membranes were then incubated with the appropriate antibodies as follows: rabbit antihuman E6-AP was diluted 1:1000 (A300-352A; Bethyl Labs, Montgomery TX) or goat anti-human actin IgG was diluted 1:500 (ac-1616 Santa-Cruz Biotechnology, Santa Cruz CA). The membranes were then incubated with either goat anti-rabbit IgG horseradish peroxidase (HRP) or donkey anti-goat HRP (ac-2004 or ac-2020, respectively, Santa Cruz). The signals of western blotting were detected by enhanced chemiluminesence (ECL, GE healthcare) exposed to X-ray films. For quantification experiments, the X-ray films were scanned and the intensity of the signals was quantified by densitometry. We used a least 3 sets of mice with every genotypes. In each group of mice from different genotypes, the levels of E6-AP were normalized against the levels of actin in each sample. In each set of experiments, the normalized level of E6-AP from the wild-type mouse was always set as 1.

Chromatin immunoprecipitation (ChIP)
For ChIP-on-chip assays, mouse whole brain dissected from pups at day 1 of age was used for MNase chromatin immunoprecipitation assay as described [43]. Brain samples were homogenized in Douncing buffer {10 mM Tris-Cl at pH 7. For ChIP-qPCR analyses, mouse whole brain dissected from pups at day 1 of age was used for ChIP assays as described by Millipore/Upstate Biotechnology (available at http://www. millipore.com). For chromatin modification analysis, chromatin extracted from mouse brain was immunoprecipitated with antitrimethyl H3K4 antibodies (Abcam, clone ID: ab8580), anti-acetyl Histone 3 (Millipore/Upstate, catalog #: 06-599), or antitrimethyl H3K9me3 (Millipore/Upstate, clone ID: 6F12-H4). Then, qPCR analyses were performed using the primer sets to amplify co-precipitated DNA from Snrpn and Ndn. The primers used are listed in Table S1.

Mouse genomic tiling array
We designed a custom mouse genomic tiling array using the Agilent E-array platform. The array included 44,000 oligonucleotides covering sequences of the mouse imprinted gene clusters from ATP10a to Peg12 at the PWS/AS region (Chr7: 64,846,543-69,740,076). DNA products from chromatin immunoprecipitation were labeled and applied to the genomic tiling arrays as described by the protocol of Agilent Technologies (available at http://www. agilent.com).

Sodium bisulfite sequencing analysis
Genomic DNA was purified from mouse brain dissected from pups at day 1 of age using the DNeasy blood & Tissue Kit (Qiagen, Hilden, Germany). Bisulfite treatment of genomic DNA was carried out using the EZ DNA methylation kit (ZYMO Research, Irvin, CA) as described (available at http://www. zymoresearch.com). From bisulfite-treated DNA, the CpG-rich regions of Ndn and Mkrn3 were amplified by nested PCR with primer sets listed in Table S1. For the Ndn CpG region, a 700-bp first-round PCR product was amplified with the forward primer NEC43F [44] and the reverse primer Ndn-bis1R. Then, a 628-bp second-round PCR product was amplified with the forward primer Ndn-bis2F and the reverse primer Ndn-bis2R. For the Mkrn3 CpG region, a 625-bp first-round PCR product was amplified with the forward primer Mkrn3-bis1F and the reverse primer W28 [27]. Then, a 606-bp second-round PCR product was amplified with the forward primer Mkrn3-bis2F and the reverse primer Mkrn3-bis2R. Nested PCR reactions were performed with ZymoTaq DNA polymerase (ZYMO Research) with first-round PCR as follows: 10 min at 94uC; 1 min at 94uC, 2 min at 58uC, and 2 min at 72uC cycled 5 times; 45 s at 94uC, 1 min at 58uC, and 1 min at 72uC cycled 35 times; 10 min at 72uC, and secondround PCR as follows: 10 min at 94uC; 45 s at 94uC, 1 min at 58uC, and 1 min at 72uC cycled 35 times; 10 min at 72uC. PCR products from two independent bisulfite conversion reactions were Figure 8. Schematic representation of genetic architecture at the PWS/AS domain. (A) Gene expression and DNA methylation associated with the D4.8 mutation, the AS-IC an mutation, or the PWS-IC Hs mutation were shown. On the paternal wild-type chromosome, the Snrpn and Ndn promoters are unmethylated and the paternally expressed imprinted genes are activated (a). When the PWS-IC D4.8 mutation deletes the CpG island at the Snrpn promoter on the paternal chromosome, the Snrpn sense/Ube3a antisense is transcribed only from the upstream exons, but is not transcribed from the major promoter Snrpn exon 1, resulting in partial activation of Ube3a. The Ndn promoter was partially methylated with decreased gene expression (b). On the maternal wild-type chromosome, silencing of Ndn and Snrpn is associated with DNA methylation at their promoters. Ube3a is activated (c). When the PWS-IC D4.8 mutation deletes the CpG island at the Snrpn promoter on the maternal chromosome, the Snrpn promoter at the upstream exons is activated and transcribes the Snrpn sense/Ube3a antisense, resulting in partial reduction of Ube3a expression. The Ndn promoter was partially activated with decreased DNA methylation (d). Maternal inheritance of an insertion/duplication mutation 13 kb upstream of Snrpn exon 1 (AS-IC an ) causes loss of Snrpn methylation, decreased Ndn methylation, activation of the maternally repressed genes, and silencing of Ube3a due to expression of the maternal copy of the Snrpn sense/Ube3a antisense (e). Maternal transmission of a targeted replacement of mouse PWS-IC with human PWS-IC (PWS-IC Hs ) expressed the Snrpn sense/Ube3a antisense transcripts from the inserted human SNRPN promoter, resulting in silencing of Ube3a. The PWS-IC Hs does not affect any other paternally expressed imprinted transcripts on the maternal chromosome (f). (B) A model on how PWS-IC controls both paternal and maternal imprint at the PWS/AS domain. On the paternal chromosome, the PWS-IC functions as the major promoter for the Snrpn sense/Ube3a antisense transcripts. The paternal PWS-IC also acts at long distances to activate the Snrpn upstream exons and Ndn gene. The active Snrpn and Ndn promoters are unmethylated on the CpG islands and modified with H3K4me3 and H3Ac. On the other hand, the maternal PWS-IC acts in cis to silence the paternally expressed imprinted genes with the Snrpn and Ndn promoters methylated on the CpG islands and modified with H3K9me3, a mark of a repressive chromatin state. The maternally expressed imprinted genes Ube3a is expressed due to lack of the Snrpn sense/Ube3a antisense transcripts on the maternal chromosome. doi:10.1371/journal.pone.0034348.g008 cloned into the pGEM-Teasy vector (Promega Corp.). DNA sequencing was performed using forward and reverse primers T7 and SP6 (MCLAB, South San Francisco, CA).

Methylated DNA immunoprecipitation (MeDIP)
Genomic DNA was purified from mouse brain dissected from pups at day 1 of age using the DNeasy blood & Tissue Kit (Qiagen, Hilden, Germany). MeDIP assays were performed as described [45]. Briefly, 5 mg of genomic DNA in MeDIP buffer (10 mM sodium phosphate at pH 7.0, 140 mM NaCl, 0.05% Triton X-100) was sonicated to produce random fragments ranging in size from 300 bp to 1,000 bp. The DNA was immunoprecipitated with antibody against 5-methylcytidine (Eurogentec, Belgium, clone ID: 33D3) at 4uC for overnight and washed with MeDIP buffer three times. The precipitated DNA was treated with proteinase K at 50uC for 4 h and recovered by QIAprep Spin Miniprep kit (Qiagen, Hilden, Germany). Then, qPCR analyses were performed to amplify the precipitated DNA from Ndn and Mkrn3. The primers used are listed in Table S1. Figure S1 Expression analyses of Snrpn, Snrod116, and Snord115 in the m D4.8 p + mice. (A-D) The Snrpn u1-ex3 (A), Snrpn exon 7 (B), Snrod116 (C), and Snord115 (D) transcripts were analyzed by qRT-PCR. Total RNA was isolated from brains of wild-type mice (a, m + p + ) (n = 4) and mice inheriting the D4.8 mutation maternally (b, m D4.8 p + ) (n = 4). For qRT-PCR, the levels of gene expression from wild-type mice were set as 1. (E) Schematic representation of the mouse PWS/AS domain (top) and summary of gene expression in the m + p + and m D4.8 p + mice (bottom, a and b). The Snrpn sense/Ube3a antisense transcripts initiated from Snrpn exon 1 with the major promoter activity and from Snrpn upstream exons with weaker promoter activity are marked as bold and thin arrows, respectively. SnoRNAs are encoded within these large Snrpn sense/Ube3a antisense transcripts derived from both Snrpn major and upstream exon promoters. Snord116 and Snord115 are multiple copy gene clusters. The centromeric (cen) and the telomeric (tel) positions are indicated. Paternally and maternally expressed genes are marked as blue and red boxes, respectively. D4.8 indicates a 4.8-kb deletion at Snrpn exon 1. The levels of the Snrpn u1-ex3, Snrpn exon 7, Snrod116, and Snord115 transcripts from wild-type mice were set as 100%. Mat, maternal chromosome; Pat, paternal chromosome. (TIF) Figure S2 Methylation analysis of Ndn in mice carrying the D4.8 deletion. Southern blotting was performed to analyze the methylation patterns of the Ndn CpG island in mice with paternal inheritance of the D4.8 deletion (m + p D4.8 ), mice with maternal inheritance of the D4.8 deletion (m D4.8 p + ), mice with the D4.8 deletion from both parents (m D4.8 p D4.8 ), and wild type mice (m + p + ). Genomic DNA isolated from brains was digested with HindIII (H) alone or in combination with SacII (SH), and hybridized with a probe from the Ndn 59 flanking region. Fragment sizes: me, 3.4 kb; unme, 1.9 kb. me, methylated; unme, unmethylated. (TIF)