• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of geneticsGeneticsCurrent IssueInformation for AuthorsEditorial BoardSubscribeSubmit a Manuscript
Genetics. Sep 2011; 189(1): 109–122.
PMCID: PMC3176116

A Survey for Novel Imprinted Genes in the Mouse Placenta by mRNA-seq

W. C. Ting, Communicating editor

Abstract

Many questions about the regulation, functional specialization, computational prediction, and evolution of genomic imprinting would be better addressed by having an exhaustive genome-wide catalog of genes that display parent-of-origin differential expression. As a first-pass scan for novel imprinted genes, we performed mRNA-seq experiments on embryonic day 17.5 (E17.5) mouse placenta cDNA samples from reciprocal cross F1 progeny of AKR and PWD mouse strains and quantified the allele-specific expression and the degree of parent-of-origin allelic imbalance. We confirmed the imprinting status of 23 known imprinted genes in the placenta and found that 12 genes reported previously to be imprinted in other tissues are also imprinted in mouse placenta. Through a well-replicated design using an orthogonal allelic-expression technology, we verified 5 novel imprinted genes that were not previously known to be imprinted in mouse (Pde10, Phf17, Phactr2, Zfp64, and Htra3). Our data suggest that most of the strongly imprinted genes have already been identified, at least in the placenta, and that evidence supports perhaps 100 additional weakly imprinted genes. Despite previous appearance that the placenta tends to display an excess of maternally expressed imprinted genes, with the addition of our validated set of placenta-imprinted genes, this maternal bias has disappeared.

GENOMIC imprinting occurs when the expression of the maternal and paternal copies of a gene differ in a parent-of-origin dependent manner (Reik and Walter 2001). Several mechanisms of genomic imprinting are shared by higher plants and therian mammals, involving differential DNA methylation, noncoding RNA, and/or histone modifications (Delaval and Feil 2004; Pauler and Barlow 2006; Hudson et al. 2010), although imprinting almost certainly arose independently in these lineages. Imprinted genes are often expressed and imprinted in a tissue- and developmental stage-specific manner. Although known imprinted genes tend to be clustered in the genome, there has been an ascertainment bias in concentrating the search among nearby genes for new imprinted candidates, motivating a need for a more balanced genome-wide scan. The occurrence of medical disorders associated with defects in imprinting provides further motivation to produce an exhaustive identification of imprinted genes. Because one allele is virtually silenced, mutations transmitted from the expressing parent behave in a dominant fashion, as is seen in human disorders associated with defects in imprinted genes (Jiang et al. 2004; Butler 2009).

To date, >100 imprinted genes have been discovered in the mouse (Morison et al. 2005), but the list is not exhaustive. Transcriptome-wide and genome-wide attempts to search for novel imprinted genes have exploited different approaches (Maeda and Hayashizaki 2006; Henckel and Arnaud 2010). Genome-wide bioinformatic predictions have successfully identified novel imprinted genes in human and mouse (Yang et al. 2003; Luedi et al. 2005; Luedi et al. 2007; Brideau et al. 2010), but the prediction power is low because the training set of known imprinted genes is small, and the genomic clustering of imprinted genes violates independence of the imprinting signals (Daelemans et al. 2010). Earlier experimental approaches such as expression microarrays on parthenogenetic and androgenetic embryos (Mizuno et al. 2002; Kuzmin et al. 2008; Sritanaudomchai et al. 2010), expression arrays on uniparental disomic (UPD) mice (Choi et al. 2001, 2005; Schulz et al. 2006), and allele-specific expression arrays on individuals with informative SNPs (Pollard et al. 2008; Brideau et al. 2010) have identified many novel imprinted genes on a larger scale than the single-gene approach. However, these methods require an abnormal configuration of the genome and can cover only a subset of genes included in the array design or the UPD region. DNA methylation-based methods have successfully identified a number of novel imprinted genes (Peters et al. 1999; Smith et al. 2003). This method first searches for differentially methylated regions (DMRs), then examines the genes in close proximity to each novel DMR. Since not all imprinted genes have an associated DMR, even this approach will likely miss some novel imprinted genes. To overcome these problems and begin to identify imprinted genes transcriptome-wide in a variety of tissues, we (Wang et al. 2008) and other investigators (Babak et al. 2008; Gregg et al. 2010a,b) have carried out mRNA-seq studies to identify novel imprinted genes through differential allele-specific expression in reciprocal F1 plants and animals. Wang et al. (2008) and Babak et al. (2008) are the first studies using RNA-seq of mouse reciprocal crosses to search for novel imprinted genes. Wang et al. (2008) performed RNA-seq of mouse neonatal day 2 (P2) brains from reciprocal crosses of AKR and PWD strains. We discovered and confirmed 14 known and 3 novel imprinted genes in P2 brains. Babak et al. (2008) did transcriptome sequencing on embryonic day 9.5 (E9.5) embryos in CAST/EiJ and C57BL/6J reciprocal crosses and they found 14 imprinted genes that are all known in mouse. No novel imprinted genes emerged from this study. Recently, Gregg et al. (2010b) published an RNA-seq study on embryonic and adult brains of CAST/EiJ and C57BL/6J reciprocal crosses. Whole E15 brain, adult cortex, and adult hypothalamus samples were sequenced and analyzed, and they claimed >1300 genes showed differential parental allelic expression in the brain. It is clear that RNA-seq provides a powerful tool for scoring parent-of-origin differential expression, and that differences in targeted tissue, developmental stage, sequence quantity, and methods of validation may contribute to differences across these studies.

In the mouse, most of the known imprinted genes are expressed and imprinted in the brain and/or placenta (Morison et al. 2005). The placenta is a mammalian-specific organ, which has important nutritional transport and immune functions for fetal growth. The placenta has been a primary target organ in studies of genomic imprinting in terms of the number and importance of known imprinted genes (Wagschal and Feil 2006; Frost and Moore 2010), motivating this RNA-seq analysis of reciprocal F1 mice to discover novel imprinted genes. Since all three previous transcriptome-wide RNA-seq studies were focused on brain or embryonic tissue, our first-pass survey in mouse placenta will complement previous studies and provide information on a tissue of particularly focused interest to the imprinting community.

Materials and Methods

Mouse strains and crosses

The four mouse strains (AKR/J, PWD/PhJ, C57BL/6, and CAST/EiJ) and the two reciprocal crosses with two strain combinations (PWD/PhJ × AKR/J, AKR/J × PWD/PhJ and C57BL/6 × CAST/EiJ, CAST/EiJ × C57BL/6) were described in Wang et al. (2010). We dissected E17.5 placenta tissues from the F1 mice from the following crosses: AKR/J female × PWD/PhJ male cross (AKR × PWD for short), PWD/PhJ female × AKR/J male (PWD × AKR for short), C57BL/6 female × CAST/EiJ male (B6 × CAST for short), and CAST/EiJ female × C57BL/6 male (CAST × B6 for short). To minimize maternal contamination, we cut only the fetal side of the placenta tissue during dissection. E17.5 was chosen because it is late enough to be able to get enough tissue in these dissections, but early enough to make it fairly easy to avoid maternal contamination.

We extracted total RNA samples from the placentas harboring the F1 progeny using Qiagen RNeasy Plus Mini kit (Qiagen). The RNA concentrations and A260 nm/A280 nm ratios were quantified with a NanoDrop ND-1000 spectrophotometer (Thermo Scientific). RNA integrity was tested using the Agilent 2100 bioanalyzer (Agilent Technologies). All of the samples had an RNA integrity number (RIN) in the range of 9.6–10.0 (RINmax = 10.0). AKR/J and PWD/PhJ genomic DNA were purchased from The Jackson Laboratory.

Illumina mRNA sequencing of the F1 placenta transcriptome

Our initial mRNA-seq was performed on total RNA samples from one AKR female × PWD male and one PWD female × AKR male placenta using an Illumina Genome Analyzer (GA) (Illumina). The mRNA-seq libraries were made with 5 µg of starting total RNA samples using the mRNA-Sequation 8-Sample Prep kit (Illumina), following the Illumina protocol for mRNA sequencing sample preparation. Eight Illumina GA lanes were sequenced for the AKR × PWD library and seven lanes for the PWD × AKR library. Image analysis and base calling were performed by the Illumina instrument software. In total, our initial screen consisted of 66.0 million short reads (read length 44 bp) for the AKR × PWD cross and 63.3 million reads for the PWD × AKR cross.

mRNA-seq alignment and quantification of total and allele-specific expression

The reads were truncated to 40 bp and aligned to the mouse reference genome (National Center for Biotechnology Information, NCBI B37) using BWA software with a maximum of four mismatches (Li and Durbin 2009). On average, 55.2% of the reads were mapped uniquely to the reference genome. Alignment counts in the exon regions were summarized by custom scripts. To identify reads that mapped to the exon–intron junctions, we built a junction database by extracting all possible junction sequences, on the basis of the gene and exon models from the Ensembl database (www.ensembl.org). A total of 4.8% of the total reads were mapped to the exon–intron junctions. The exon and junction counts were normalized by the transcript length and the total number of mapped reads to compute RPKM (Reads Per Kilobase of exon model per Million mapped reads; Mortazavi et al. 2008). We covered 12,532 unique Ensembl genes (41,110 Ensembl transcripts) with RPKM ≥ 1. If we use a more stringent cutoff of RPKM ≥ 5, then 6794 unique genes (20,026 transcripts) are retained.

To quantify the allele-specific expression in the two reciprocal crosses, at each identified SNP position we counted the reads with the reference allele as well as reads with the alternative allele (Wang et al. 2008). In addition to using the known sequence differences between the mouse strains used, we also performed de novo SNP calling from the uniquely aligned reads using SAMtools software (Li et al. 2009), followed by our own postfiltering scripts. To determine the transmission direction, we used the AKR/J allele information from the Sanger mouse genome project 2010-03 SNP release (http://www.sanger.ac.uk/resources/mouse/genomes/). Since there is imprinted X inactivation in the mouse placenta (Huynh and Lee 2005; Sado and Ferguson-Smith 2005), SNPs with an X chromosome homolog will mistakenly suggest a maternally expressed imprinted gene. To eliminate this bias, we BLATed all SNPs with sequence from 50 bp upstream and 50 bp downstream to the genome (http://genome.ucsc.edu/cgi-bin/hgBlat?command=start). We removed all SNPs with an X chromosome BLAT hit that matched with length ≥40 (equal to the read length).

In total, 43,510 high-quality autosomal SNPs with four or more counts in each of the two reciprocal crosses were identified. A total of 41,953 SNPs (96.4%) are within 1 kb upstream or downstream of Ensembl gene models. Manual annotation was performed for three known imprinted genes that are not in the Ensembl database (Peg10, Rian, and Mirg). Because both the AKR and PWD alleles in the F1 transcriptome are mapped to the reference genome, which was assembled from the B6 strain, there will be genome mapping bias toward the AKR allele if we use the same cutoff for both alleles (AKR is closer to B6 than it is to the PWD strain in terms of genetic distance). To remove this mapping bias, we generated a pseudogenome, by replacing the reference allele in the genome with the alternative allele. Then we redid the alignment with the same cutoff to the pseudogenome. We used the averaged counts from the reference genome and pseudogenome as the final SNP count summary.

Detection of significant parent-of-origin effects and identification of candidate imprinted genes

To select candidate imprinted genes for verification, we applied a formal statistical test to the 2 × 2 contingency table formed by the tally of reads of the two alleles in the two reciprocal crosses (Wang et al. 2008). In addition to using this statistical significance, we also filtered the results on the basis of the magnitude of the allelic expression bias. We define p1 as the AKR allele percentage in the AKR × PWD cross and p2 as the AKR allele percentage in the PWD × AKR cross. For a 100% maternally expressed candidate imprinted gene, we expect p1 = 1 and p2 = 0. For partially imprinted genes with preferential maternal expression, we used a cutoff of p1 > 0.65 and p2 < 0.35 (and similarly, we selected the paternally expressed imprinted candidates with a cutoff of p1 < 0.35 and p2 > 0.65). This cutoff is somewhat arbitrary and was meant to avoid the inclusion of genes with a weak allelic imbalance that would otherwise be included in our imprinted candidate set only because they are so highly expressed that they become statistically significant. For graphical presentation and discussion, the metric p2p1 quantifies the parent-of-origin effect in the range from −1 to +1. If there is no parent-of-origin effect, then p2p1 = 0. If there is preferential expression of the paternal allele, then p2p1 > 0. If there is preferential expression of the maternal allele, then p2p1 < 0. To keep the false-positive rate low, we only include candidate genes with two or more informative SNPs.

Verification of the candidate imprinted genes with allele-specific pyrosequencing

We selected three known imprinted genes (Igf2, Peg10, and Klf14) and seven candidate imprinted genes (Pde10a, Phf17, Gpsm2, Zfp64, Htra3, Phactr2, and Trim23) for allele-specific expression quantification using pyrosequencing. To exclude the possibility of stochastic expression effects and sex-specific genomic imprinting, we verified these genes in placentas harboring three female and one male F1 progeny from each of AKR–PWD reciprocal crosses. To exclude strain-specific effects, we verified three novel imprinted genes (Pde10a, Phf17, and Phactr2) in an additional three female and one male progeny bearing placentas from each of B6–CAST reciprocal crosses. The pyrosequencing assay design and sequencing protocol can be found in Wang et al. (2010).

Results

mRNA-seq alignments, transcriptome coverage, and SNP calling

This mRNA-seq study was performed on E17.5 placental tissues from reciprocal crosses of AKR and PWD mouse strains. We obtained 66 million 44-bp reads from placenta cDNA of a single AKR × PWD F1 individual and 63 million reads from the reciprocal PWD × AKR placental transcriptome. A total of 60% of the reads could be uniquely mapped to the NCBI B37 mouse reference genome, with 55.2% of reads mapping to the exons and 4.8% mapping to the exon–intron junctions. The total expression levels were quantified by RPKM, which is a normalized per-gene read count (Mortazavi et al. 2008). In the RNA-seq data, there was coverage of 12,532 Ensembl unique genes (41,110 transcripts) with RPKM >1, and 6794 unique genes had an RPKM value >5.

Informative SNP positions are needed to quantify the allele-specific expression. From de novo SNP calling based on the RNA-seq data, after quality filtering, we found 43,510 high-quality autosomal SNPs, 96.4% of which reside in known Ensembl gene models. To remove the genome mapping bias, we summarized the SNP counts by the average count when mapped to the reference genome and to a pseudogenome of the alternate strain (see Materials and Methods).

Detection of significant parent-of-origin effects

From the read counts at the informative SNP positions, we were able to determine the allele-specific expression ratio from the relative counts of the reference and alternative alleles (Wang et al. 2008). We define p1 as the expression percentage from the AKR allele in placentas from the AKR female × PWD male cross and p2 as the AKR allele percentage in the reciprocal cross. In regard to the direction of transmission, p1 is the maternal allele percentage in AKR × PWD and p2 is the paternal percentage for PWD × AKR. The Storer–Kim test was used as a formal statistical test of the null hypothesis that (p2p1) = 0. Rejections of this null hypothesis identify novel imprinted candidate genes (see Materials and Methods). To further filter the data and reduce false positives, rather than relying only on the P-value of the Storer–Kim test, we also used an arbitrary cutoff of p1 > 0.65 and p2 < 0.35 for maternally expressed candidates and p1 < 0.35 and p2 > 0.65 for paternally expressed ones. This allows for identification of partially imprinted genes when there are sufficiently many reads spanning the SNPs to make a confident call.

Of the 5557 unique genes covered with two or more informative SNPs, with the above criteria for significant parent-of-origin effect identification, we found 251 significant candidates with q-value <0.01 and SNP coverage ≥4 in each of the two reciprocal crosses (criterion 1). Of these candidate genes, 120 have preferential maternal expression, and 131 have a paternally biased expression (supporting information, Table S1). If we use RPKM > 1 and SNP coverage > 10 in both reciprocal crosses as the criteria for inclusion, 216 significant candidates are left, with 115 paternal and 101 maternal candidates (criterion 2). If we use a more stringent cutoff of RPKM > 3 and SNP coverage > 20, only 113 candidates are retained, with 60 paternal and 53 maternal candidates (criterion 3). To visualize the allelic expression ratio and the degree of parent-of-origin effect genome-wide, we made a plot for each autosome (Figure S1), and chromosome 7 is shown in Figure 1 as an example. From these figures, we observed that most of the genes show nearly 50:50 allelic expression ratios, when we scan along the chromosomes. A number of significant candidate imprinted genes emerged from the parent-of-origin effect plot.

Figure 1
Allele-specific expression ratio and the distribution along mouse chromosome 7 of parent-of-origin biased expression. (Left) Allele-specific expression levels for genes on chromosome 7 for both AKR female × PWD male and PWD female × AKR ...

Significant candidate imprinted genes that are previously known to be imprinted in mouse

Among the 251 candidate imprinted genes that we identified from criterion 1, 35 have been previously reported in the mouse literature to be imprinted. For each gene, the number of significant SNPs, total SNP counts, allelic expression ratios, and the q-value are summarized in Table 1. We compared the expression direction of these genes in our RNA-seq data and the previously reported imprinting direction, and 35 out of 35 matched. Twenty-three of the 35 genes were known to be imprinted in mouse placenta in various stages and crosses: Igf2 (Dechiara et al. 1991; Hu et al. 1995), Peg10 (Ono et al. 2003), Sfmbt2 (Kuzmin et al. 2008), Sgce (Piras et al. 2000), Plagl1 (Piras et al. 2000; Smith et al. 2002), Slc38a4 (Mizuno et al. 2002; Smith et al. 2003), Airn (also known as Igf2rAS) (Wutz et al. 1997), Rtl1 (Seitz et al. 2003), Mest (Kaneko-Ishino et al. 1995), Igf2as (Moore et al. 1997), and Dlk1(Schmidt et al. 2000) are known to be paternally expressed in the mouse placenta; H19 (Bartolomei et al. 1991), Igf2r (Barlow et al. 1991), Cdkn1c (Hatada and Mukai 1995), Grb10 (Miyoshi et al. 1998), Ppp1r9a (Ono et al. 2003), Klf14 (Parker-Katiraee et al. 2007), Nesp (Peters et al. 1999), H13 (Wood et al. 2007), Slc22a2 (Zwart et al. 2001), Asb4 (Mizuno et al. 2002), Slc22a18 (Dao et al. 1998), and Kcnq1(Gould and Pfeifer 1998) are preferentially expressed from the maternal allele. The remaining 12 genes are either known to be not imprinted in the placenta or the imprinting status in the placenta is not clear. In this study, we found that they are actually imprinted in E17.5 mouse placenta in AKR–PWD reciprocal crosses. Peg3 (Kaneko-Ishino et al. 1995) is known to be imprinted in the human placenta (Hiby et al. 2001); however, the imprinting status in the mouse placenta had not been reported. Ndn (MacDonald and Wevrick 1997) and Magel2 (Boccaccio et al. 1999) are both expressed in the mouse placenta, whereas the imprinting status was not clear. Rian (Hatada et al. 2001), Zim1 (Kim et al. 1999), Meg3 (Miyoshi et al. 2000), Mirg (Seitz et al. 2004), Usp29 (Kim et al. 2000), Impact (Hagiwara et al. 1997), Nnat (Kagitani et al. 1997), Zdbf2 (Kobayashi et al. 2009), and Zrsr1 (Hatada et al. 1993) were not previously reported to be imprinted in the mouse placenta either. Therefore, we identified 12 candidate genes with novel mouse placenta imprinting status.

Table 1
Imprinted genes identified in mouse placenta RNA-seq data that have been reported previously in the literature

The q-value rank order is presented in Table 1. We noticed that most of the known imprinted genes identified in our study have higher q-value rank relative to other genes, most of them are highly expressed in the placenta, and the imprinting status of most previously known imprinted genes is ~100%. We conclude that most of the significant imprinted genes with highest degree of parent-of-origin bias have already been identified by the genomic imprinting community. The high concordance (35 out of 35) of known imprinted genes with the significance of our test of parent-of-origin effects on allelic expression ratios provides one measure of the confidence in the results, despite the lack of replication at the RNA-seq stage.

Identification and verification of novel imprinted genes in the mouse placenta

To confirm the novel imprinted candidates identified above, we need to quantify their allele-specific expression using an independent method. We performed pyrosequencing to quantify allele-specific expression in two reciprocal F1 placenta samples. Pyrosequencing is a highly quantitative method to profile the allelic expression ratio, with a measurement coefficient of variation of 2–5% (Marsh 2007). To exclude the possibility of random monoallelic expression for specific genes (Lomvardas et al. 2006; Gimelbrant et al. 2007), and potential sex-specific imprinting status (Gregg et al. 2010a), we verified the candidates in four AKR × PWD F1 individuals (three females and one male) and four PWD × AKR F1 individuals (three females and one male). The average allelic percentage is reported in Tables 2 and and33.

Table 2
Verification of known/novel imprinted genes: RNA-seq read counts for allele-specific expression

We selected a total of 10 candidate genes for verification, including three known imprinted genes as positive controls (Igf2, Peg10, and Klf14). Among the top 20 candidates, only 2 are novel (Pde10a and Phf17), and we included both. Then we selected five additional novel candidates (Gpsm2, Zfp64, Htra3, Trim23, and Phactr2) for verification.

From the pyrosequencing results in Table 3, 8 of the 10 known and novel candidate genes we tested are verified to be imprinted; 1 candidate gene (Trim23) did not show good pyrosequencing signal due to low expression level; we observed biallelic expression for 1 candidate gene (Gspm2). Further examination of the Gspm2 gene region reveals that the different SNPs are not consistent in RNA-seq data. Careful inspection of the RNA-seq read alignments suggests that the false-positive call may have been made because of poor read mapping, as the read depth is unusually variable around this gene. Therefore, we have an empirical false discovery rate (FDR) of 1 out of 9 or 11% confirmed by our pyrosequencing verification results.

Table 3
Pyrosequencing to score differential allele-specific expression

Igf2 and Peg10 were correctly verified as paternally expressed imprinted genes, and Klf14 as maternally expressed imprinting gene, which is consistent with the results in our RNA-seq data. Among the seven novel candidates, five (Pde10a, Phf17, Zfp64, Htra3, and Phactr2) were verified to be novel imprinted genes in the mouse placenta (Table 3), one test failed due to low expression, and one failed to validate.

Pde10a is the most significant novel candidate gene (q-value rank 15). It is located on chromosome 17, 3.6 Mbp away from the known imprinted gene, Slc22a3. It is a member of the phosphohydrolyase gene family, catalyzing the hydrolysis of the cAMP and cGMP to the respective nucleoside 5′ monophosphate (Loughney et al. 1999; Soderling et al. 1999). Pyrosequencing primers were designed to target 1 of the 12 significant SNPs in this gene (Figure 2A). In the RNA-seq data, we observed expression primarily from the maternal allele in both AKR × PWD and PWD × AKR reciprocal crosses (Figure 2B). We verified it in four placentas from each of the two reciprocal crosses, and we found consistent preferential maternal expression. To exclude the possibility of strain-specific imprinting, we also tested placenta tissue from B6–CAST reciprocal crosses, and we obtained the same results (Table S2). Thus, we conclude that Pde10a is a novel imprinted gene in the E17.5 mouse placenta.

Figure 2
Verification of the novel candidate imprinted gene Pde10a, a preferentially maternally expressed imprinted gene. (A) The mouse crossing scheme used to generate the AKR–PWD reciprocal F1 placentas. One informative SNP within the gene is shown with ...

Phf17 is the second most significant novel candidate in the list. It is located on mouse chromosome 3 and it is not near any of the known imprinting cluster. Phf17 (also known as Jade1) is a component of the HBO1 complex, which has a histone H4-specific acetyltransferase activity and performs most of the histone H4 acetylation in vivo (Foy et al. 2008). Imprinted genes involved in histone modifications are particularly interesting, as they may provide a means for amplification of the imprinting signal, and for propagating the effect to other target genes. Pyrosequencing verifications confirmed preferential paternal expression in both AKR–PWD and B6–CAST crosses (Table 3 and Table S2).

Phactr2 is a phosphatase and actin regulator, and it is identified in our RNA-seq study as a maternally expressed imprinted candidate. This gene had not previously been known to be imprinted in mouse. We verified it in multiple individuals of both AKR–PWD and B6–CAST crosses, and it is confirmed to be preferentially expressed from the maternal allele (Table 2 and Table S2). In a recent Illumina ASE BeadArray survey of novel imprinted genes in human term placenta, human PHACTR2 is found to be partially imprinted, with a maternal allelic bias (Daelemans et al. 2010). Therefore, the imprinting status of Phactr2 is conserved between mouse and human. Phactr2 is on mouse chromosome 10, 104 kbp downstream of a paternally expressed known imprinted gene, Plagl1. Phactr2 is transcribed in the opposite direction to Plagl1, which could be another reciprocally imprinted sense–antisense pair (Wang et al. 2008).

Among the seven novel candidates tested, two other genes, Zfp64 and Htra3 have also been verified to be partially imprinted in the mouse placenta. Zfp64 is on mouse chromosome 2, 6 Mbps from a known imprinted gene Nesp. Zfp64 is a Krüppel family transcription factor that is under the control of Runx2 and participates in Notch signaling to regulate differentiation in mesenchymal cells (Sakamoto et al. 2008). Htra3 is a serine protease whose activity is absolutely required for its activity in TGFβ signaling inhibition (Tocharus et al. 2004). Htra3 was initially discovered to have a strong 100:0% allelic bias, but the verification results showed only partial imprinting (with a 75:25% allelic bias). This could be due to the low expression level of Htra3 in the mouse placenta (RPKM < 1).

Finally, considering the last two imprinting validation tests, the pyrosequencing signal from Trim23 was too low to determine the allelic expression percentage. Therefore we could neither confirm nor exclude the imprinting status of Trim23. Gpsm2 was shown in our pyrosequencing assay to be not imprinted in the mouse placenta (Table 3). Overall the empirical verification rate is quite high (8 out of 9 successful tests), compared to other recently published transcriptome-wide surveys.

Assessment of the degree of maternal contamination in our placenta samples

One caution about identifying novel imprinted genes in the mouse and human placenta is maternal contamination (Proudhon and Bourc’his 2010). The placenta is a complex organ that consists of many different tissue and cell types. For term and near-term placenta, the contact of maternal and fetal tissues at the interface is challenging to separate by dissection, resulting in the potential for maternal contamination (Proudhon and Bourc’his 2010). In some studies of novel imprinted genes in the placenta, the possibility for maternal contamination cannot be excluded.

Several approaches were used to minimize maternal contamination in our samples. The first was to take special precautions during the dissection. From every sample collected, tissue was only taken from the middle of the placenta and only from what was clearly the fetal side. Then we washed the tissue multiple times in PBS to remove maternal blood. Second, we quantified the degree of contamination and chose the samples for RNA-seq that had the least maternal contamination (on the basis of allelic expression ratio of several known imprinted genes with 100% paternal expression in placenta). If there were maternal contamination, paternally expressed imprinted genes would display expression from the maternal allele, and the degree of leakage could be used as a criterion to select the best samples. Third, several uterus samples near the placenta were collected at the same time, which allowed us to check the uterus expression level of a gene to determine the potential for contamination. Fourth, with the transcriptome-wide allelic expression profile, maternal contamination would be reflected by an allelic bias throughout the genome. By quantifying read counts of maternal alleles transcriptome-wide, it has been possible to estimate the degree of maternal contamination, and use this estimate to normalize SNP counts in the candidate imprinted genes.

Before we quantify the maternal bias introduced by maternal tissue contamination, we need to understand what other factors could also contribute to the deviation from 50:50 expression ratio of the two parental alleles. First, there is the possibility of global eQTL effects. As we observed from the allelic expression from a single gene, not all genes show 50:50 ratios. If the AKR allele is associated with a cis-regulatory element, it could have higher expression from the AKR allele in both reciprocal crosses. If we sum the SNP counts over all genes, it should be close to 50:50. Second, since we are aligning reads with both the AKR and PWD sequences to the B6 reference genome, there will be a mapping bias toward the AKR allele, because the mouse strain genealogy shows that the AKR strain is closer to the B6 strain. So it was important to quantify and remove the mapping bias before we could assess the degree of maternal contamination (see Materials and Methods). Finally, imprinted X inactivation takes place in the mouse placenta, which means that the X-linked genes in females will be primarily expressed from the maternal allele (Sado and Ferguson-Smith 2005). If a gene/SNP has X chromosome homology, the reads might actually be from the X chromosome, which would create a spurious maternal bias. Consequently, in this analysis the X chromosomal genes were not assessed for imprinting status.

To illustrate these confounding factors for the deviation from 50:50 allelic expression, we present an example in Table 4. Under a null model, if there is not any global eQTL effect or maternal bias or mapping bias, the allelic expression ratio will be 50:50 in both AKR × PWD and PWD × AKR crosses. Suppose there is 5% mapping bias. We would then always observe 55% expression from the AKR allele in both reciprocal crosses. If there is 5% maternal contamination, we would detect 55% expression of the AKR allele in the AKR × PWD cross, because AKR is the mother in this cross, but 45% expression of the AKR allele in the PWD × AKR cross, because PWD is the mother (Table 4). To quantify the degree of maternal contamination, we compute (p1 overallp2 overall)/2 as a metric whose expectation is zero if there is no maternal contamination (where p1 overall is the total AKR allelic expression percentage from the AKR × PWD cross summing over all genes in the transcriptome, and p2 overall is the total AKR expression percentage from the PWD × AKR cross, again summing over the transcriptome). With this metric, eQTL effects will be canceled out, leaving a bias for unimprinted genes only if there is maternal contamination.

Table 4
Quantification of global maternal contamination percentage

In our placenta data, the total AKR allelic percentages are 51.99 and 51.52% in the AKR × PWD and PWD × AKR crosses, respectively, before correcting the alignment bias (Table 4). After the mapping bias correction, the percentages are 50.50 and 50.17%, indicating that there is an ~1.5% mapping bias (Table 4). The maternal contamination is estimated to be 0.15% (Table 4), a quite tolerably low figure. For genes with moderate and high expression levels in our placenta samples, the effect of maternal contamination was negligible.

Maternally expressed placenta-only imprinted genes: Artifacts of maternal contamination?

Because of the maternal contamination problem, the imprinting status has been questioned for 13 placenta-only known imprinted genes (Proudhon and Bourc’his 2010). All are known to be maternally expressed imprinted genes. Among these genes, Gatm, Pon3, Th, Tspan32, Cd81, Tssc4, Tnfrsf23, and Osbpl5 have sufficient SNP coverage in our data to determine the imprinting status with confidence (Table 5). The genes Tfpi2, Pon2, and Dcn do not show significant parent-of-origin effect in our data, suggesting that they may not be imprinted, at least at stage E17.5 in the AKR–PWD strain combination (Table 5). Ppp1r9a is detected to be imprinted with preferential maternal expression. Nap1l4 is discovered to be a maternally expressed imprinted gene in the placenta (Engemann et al. 2000). Others have suggested that there may be leaky expression from the paternal allele (Umlauf et al. 2004). When we examined this gene in detail, we found four SNPs in the gene region, two in the exons, and two in the introns. One exonic SNP shows biallelic expression, and the other one shows preferentially maternal expression (Table S3). The parent-of-origin effect is not significant if we sum over the two SNPs (Table 5). There are also two SNPs covered by the Illumina reads in the intron, with preferential paternal expression (Table S3). This gene may be imprinted and there might be antisense noncoding transcript in the intronic region, or there may be complications from alternative splice products. Further investigation is needed to determine the imprinting status of Nap1l4.

Table 5
Coverage of known placenta-only imprinted genes whose imprinting status has been questioned

Maternal contamination could not only create false-positive calls for maternally expressed imprinted genes, but also may result in a paternally expressed imprinted gene to be a false negative. Zdbf2 could be one such example. Zdbf2 is detected in our data to be imprinted with preferential paternal expression, but it has been previously reported to be biallelically expressed in the placenta (Kobayashi et al. 2009). However, this could also be due to a different imprinting status of the same gene in different developmental stages/mouse strain combinations.

Is there a bias toward more maternally expressed imprinted genes in the placenta?

Contrasting patterns of genomic imprinting in the brain and placenta raises a series of questions about the mechanism and evolution of the control of imprinting. Previously, in a literature review of the tissue specificity and maternal vs. paternal expression of imprinted genes (Morison et al. 2005), it was noted that there is a paternal-brain/maternal-placenta bias (Wang et al. 2008; Proudhon and Bourc’his 2010). The genes imprinted in the brain but not the placenta tend to be paternally expressed, whereas the genes imprinted in the placenta but not the brain tend to be maternally expressed (Figure 3A and Table S4, P-value = 0.0001322, Fisher’s exact test). Our previous study also provided some suggestive evidence that the paternal-brain bias might be real (Wang et al. 2008). Here, we would like to ask whether the maternal-placenta bias is also true or whether there might be an artifact due to the potential maternal contamination or limited sampling. We covered 35 known imprinted genes and verified 5 additional novel imprinted genes in this study. If we break them down by the direction of imprinting, we do not see a bias toward more maternally expressed genes (Figure 3B, P-value = 0.6821, one-sided exact binomial test). If we examine all 251 candidates and classify them by their expression bias, we still see roughly equal numbers of paternally and maternally expressed candidates (Figure 3C), and the degree of allelic bias is statistically homogeneous between the two sets of reciprocal offspring.

Figure 3
Paternal vs. maternal novel and candidate imprinted genes. (A) Paternal vs. maternal imprinting status in brain and placenta in the literature. (B) The number of paternally and maternally expressed known and novel imprinted genes in our study. (C) The ...

Discussion

The number of imprinted genes in the mouse genome

Different studies present quite a wide range of estimates of the number of imprinted genes in the mouse genome, ranging from 100 genes (Luedi et al. 2007) to 600 genes (Luedi et al. 2005) to 1300 genes (Gregg et al. 2010b) to 2000 genes (Nikaido et al. 2003). There are several reasons for the broad range of these estimates. First, different studies used widely varying approaches, so they will have different false-positive rates as well as different coverage and sensitivity. Second, different studies examined different tissues and developmental stages. In our study, we found 251 candidate imprinted genes in the E17.5 placenta falling in the set with a statistical false discovery rate of 0.01, but we also show empirically that the false discovery rate is more like 11%. Most of the top genes in the list are already known to be imprinted, indicating that the genomic imprinting community has done a commendable job of identification of the imprinted genes. Exhaustive enumeration of imprinted genes will require a large community-wide effort, including multiple replicates from multiple lines, with samples of different tissues and developmental time points. If the results are to be interpreted with confidence on the basis of RNA-seq data alone, a blocked and replicated design is essential (Auer and Doerge 2010).

Our intention here was to apply RNA-seq in a simple, unreplicated design to serve as a means of nominating candidates for subsequent validation. Among our candidate imprinted genes, we selected 10 for validation with biological replication and an independent assay for allele-specific expression. One pyrosequencing assay failed, but of the remaining 9, 8 of the imprinting candidate genes were soundly confirmed. The candidates were chosen from a list with a theoretical false discovery rate of 0.01, whereas we observed that 1/9, or 11% of the candidates were false discoveries. The discrepancy between the q-value and the true verification rate could arise from several causes, most of which are expected to inflate the false-positive rate of an unreplicated RNA-seq study. First, for lowly expressed genes, with only a few mRNA copies in the transcriptome, there is a chance during library construction that only one of the two alleles might be randomly ligated to the adaptor and included in the final pool. After sequencing, the gene would resemble a monoallelically expressed gene, when in fact it is not. This is different from the random monoallelic expression that has been reported previously (Lomvardas et al. 2006; Gimelbrant et al. 2007), where single cells appear to fail to express both alleles. In applying quantitative RNA-seq for allele-specific expression, it is critical to assure that high library complexity is attained to avoid this allelic dropout caused by an insufficiently complex library. We might not get conclusive results for lowly expressed genes, so we need other independent methods to verify candidates with low expression levels. Second, sequencing bias and misalignments could also be a source of discordance. For the statistical test and subsequent inference of a q-value, several assumptions producing ideal experimental conditions are made: there is no sequencing bias, no misalignments, and the SNP-containing read counts are in proportion to the allelic expression ratio. However, in practice, these assumptions are easily violated. As a result, SNPs that truly have technical problems will be among the candidates that are found to be statistically significant by the Storer–Kim test, and these will be false-positive calls. This is another reason why we need independent verification using an orthogonal technology like pyrosequencing. To account for these factors, we used more stringent filters. With our criterion 3 (defined as RPKM > 3 and SNP coverage > 20), only 113 significant candidates were left. Among the 113 genes, most of the known ones (23/35) and the confirmed novel ones (4/5) are preserved. Thus, by applying expression level and SNP coverage cutoffs, the degree of library complexity and SNP bias problems will be reduced, resulting in a lower false discovery rate. We will reach the theoretical FDR only if we completely remove these effects and meet all the ideal experimental conditions, and the most obvious way to improve the situation is by replication. But it is important to note that even with only a single replicate of RNA-seq runs from each cross, valid, verifiable, novel, imprinted genes were discovered.

Many pairs of known imprinting genes occur as overlapping sense–antisense pairs (Morison et al. 2005; Wang et al. 2008). With a double-strand cDNA RNA-seq library, the allelic expression from the sense vs. antisense transcripts cannot be distinguished, so SNPs that fall in regions where both strands are transcribed may produce false-negative calls. By closely examining the SNPs within the candidates, we found some problematic genes with inconsistent SNPs or overlapping sense–antisense gene models. This could also contribute to the low verification rate. In the future, methods that allow preparation of strand-specific RNA-seq libraries should solve this problem (Levin et al. 2010).

Given the various limitations of RNA-seq studies, we conclude that an independent verification such as pyrosequencing or other allele-specific methods is necessary to confirm the imprinting status. It is also important to examine biological replicates, ideally from individuals from different strains to test the possibility of strain-specific effects. A much larger study, with a well-replicated and blocked design of multiple RNA-seq runs (Auer and Doerge 2010) would be needed to generate a definitive count of the number of imprinted genes. From our data, ~4.5% (251) of the 5527 genes, having sufficient data to perform the test, exhibit significant imprinting in the placenta. Given the empirical FDR of 11% for this test, 224 genes are expected to be verified. However, the 11% false-positive rate was seen among the subset of genes with the lowest q-values, and if all 251 genes were tested, it would likely be higher. On the other hand, the gene list of 251 was generated using strict selection criteria (RPKM >1, p1 > 0.65 or p1 < 0.35), and the unmeasured false-negative rate will be inflated. Therefore, while the experiment produces an estimate of 224 imprinted genes, the uncertainty in false-positive and false-negative rates suggest that a range of 100–250 genes may be the most supportable. Because this study was restricted to the E17.5 placental tissue in AKR–PWD crosses, the true number of imprinted genes across all tissues and stages is likely to be larger.

Artifacts in novel imprinted gene identification

There are various sources of artifacts in the identification of imprinted genes (Proudhon and Bourc'his 2010). First, there may be random monoallelic expression instead of genomic imprinting. We verified our candidates in multiple individuals to exclude this possibility. Second, the allelic bias could be generated by an eQTL effect. In our study, we used reciprocal F1’s, allowing us to distinguish parent-of-origin effects from the eQTL effects. Third, there may be a strain-specific PCR bias. Random primers were used in the Illumina library preparation, making PCR bias unlikely, and our confirmation method using pyrosequencing did not employ the same PCR primers. The fourth class of artifact is maternal contamination in the dissected placenta tissues. We took pains to avoid and to quantify the maternal contamination in our samples, and our quantitative analysis demonstrates that these efforts were successful (Table 4). Another artifact that might spuriously lead to allelic bias is homology to the X chromosome. Males inherit the X chromosome from the mother, so the X-linked genes in males will have 100% maternal expression. In female mouse embryos and placental tissues of fetal origin, there is imprinted X inactivation, resulting in preferential expression from the maternal allele. If an autosomal gene/SNP has X homology, there could be nonspecific amplification during RT-PCR or misalignment for the RNA-seq. Either case would result in spurious identification of a maternally expressed imprinted gene. This could happen even with zero maternal contamination. Careful attention to this possibility during read mapping should minimize its impact, although it is hard to exclude the possibility entirely.

Is there a paternal-brain and maternal-placenta bias?

Previous literature indicates that there is a maternal bias to allelic expression of imprinted genes in the placenta (Morison et al. 2005; Proudhon and Bourc'his 2010). This could be real, or it may be due to overestimation of maternally expressed imprinted genes due to maternal contamination or underestimation of the paternally expressed imprinted genes. From our results, we did not observe any bias toward maternally expressed imprinted genes in the placenta (Figure 3). We think this is simply because some paternally expressed genes are not known to be imprinted in placenta. In the 12 known imprinted genes identified in our data without prior reports of placenta imprinting, 8 of them are paternally expressed. In the list of novel candidate imprinted genes, we did not find any bias toward maternally expressed genes. This is also consistent with the minimal maternal contamination estimated in our study.

Conclusion

We have shown that even an unreplicated RNA-seq study can identify a highly informative set of genes showing parent-of-origin allelic expression differences that validated with a quite acceptable rate (89%). This provides an excellent set of candidates for genes showing genomic imprinting, including five novel genes that we validated by pyrosequencing in multiple biological samples. The finding that Phf17 shows strong paternally expressed imprinting is especially intriguing, as this gene is part of a histone H4 transacetylase complex and may specify a parent-of-origin differential histone acetylation. It is not immediately clear why Pde10a, a cAMP and cGMP phosphodiesterase should be maternally expressed and imprinted in the placenta, but the allelic expression bias is well validated. A larger-scale RNA-seq study with this reciprocal cross design, sequencing to greater coverage and using biological replication, would also be highly informative, allowing assessment of splice isoform-specific imprinting, sex difference in imprinting, interstrain variability, and more.

Acknowledgments

We thank Wei Wang and Jie Zhao from the Cornell University Microarray Core Facility for Illumina RNA-seq library construction and Peter Schweitzer and Tom Stelick from the Cornell University Sequencing Core Facility for the Illumina sequencing. We thank Qi Sun, Jarek Pillardy, and Lalit Ponnala from the Cornell Bioinformatics Service Unit for discussions and help with data analysis. We also thank Amanda Manfredo and Grace Chi for help with the experiments. This work was supported in part by a gift from the Peter and Nancy Meinig family.

Literature Cited

  • Auer P. L., Doerge R. W., 2010. Statistical design and analysis of RNA sequencing data. Genetics 185: 405–416 [PMC free article] [PubMed]
  • Babak T., Deveale B., Armour C., Raymond C., Cleary M. A., et al. , 2008. Global survey of genomic imprinting by transcriptome sequencing. Curr. Biol. 18: 1735–1741 [PubMed]
  • Barlow D. P., Stoger R., Herrmann B. G., Saito K., Schweifer N., 1991. The mouse insulin-like growth factor type-2 receptor is imprinted and closely linked to the Tme locus. Nature 349: 84–87 [PubMed]
  • Bartolomei M. S., Zemel S., Tilghman S. M., 1991. Parental imprinting of the mouse H19 gene. Nature 351: 153–155 [PubMed]
  • Boccaccio I., Glatt-Deeley H., Watrin F., Roeckel N., Lalande M., et al. , 1999. The human MAGEL2 gene and its mouse homologue are paternally expressed and mapped to the Prader-Willi region. Hum. Mol. Genet. 8: 2497–2505 [PubMed]
  • Brideau C. M., Eilertson K. E., Hagarman J. A., Bustamante C. D., Soloway P. D., 2010. Successful computational prediction of novel imprinted genes from epigenomic features. Mol. Cell. Biol. 30: 3357–3370 [PMC free article] [PubMed]
  • Butler M. G., 2009. Genomic imprinting disorders in humans: a mini-review. J. Assist. Reprod. Genet. 26: 477–486 [PMC free article] [PubMed]
  • Choi J. D., Underkoffler L. A., Collins J. N., Marchegiani S. M., Terry N. A., et al. , 2001. Microarray expression profiling of tissues from mice with uniparental duplications of chromosomes 7 and 11 to identify imprinted genes. Mamm. Genome 12: 758–764 [PubMed]
  • Choi J. D., Underkoffler L. A., Wood A. J., Collins J. N., Williams P. T., et al. , 2005. A novel variant of Inpp5f is imprinted in brain, and its expression is correlated with differential methylation of an internal CpG island. Mol. Cell. Biol. 25: 5514–5522 [PMC free article] [PubMed]
  • Daelemans C., Ritchie M. E., Smits G., Abu-Amero S., Sudbery I. M., et al. , 2010. High-throughput analysis of candidate imprinted genes and allele-specific gene expression in the human term placenta. BMC Genet. 11: 25. [PMC free article] [PubMed]
  • Dao D., Frank D., Qian N., O’Keefe D., Vosatka R. J., et al. , 1998. IMPT1, an imprinted gene similar to polyspecific transporter and multi-drug resistance genes. Hum. Mol. Genet. 7: 597–608 [PubMed]
  • DeChiara T. M., Robertson E. J., Efstratiadis A., 1991. Parental imprinting of the mouse insulin-like growth factor II gene. Cell 64: 849–859 [PubMed]
  • Delaval K., Feil R., 2004. Epigenetic regulation of mammalian genomic imprinting. Curr. Opin. Genet. Dev. 14: 188–195 [PubMed]
  • Engemann S., Strodicke M., Paulsen M., Franck O., Reinhardt R., et al. , 2000. Sequence and functional comparison in the Beckwith-Wiedemann region: implications for a novel imprinting centre and extended imprinting. Hum. Mol. Genet. 9: 2691–2706 [PubMed]
  • Foy R. L., Song I. Y., Chitalia V. C., Cohen H. T., Saksouk N., et al. , 2008. Role of Jade-1 in the histone acetyltransferase (HAT) HBO1 complex. J. Biol. Chem. 283: 28817–28826 [PMC free article] [PubMed]
  • Frost J. M., Moore G. E., 2010. The importance of imprinting in the human placenta. PLoS Genet. 6: e1001015. [PMC free article] [PubMed]
  • Gimelbrant A., Hutchinson J. N., Thompson B. R., Chess A., 2007. Widespread monoallelic expression on human autosomes. Science 318: 1136–1140 [PubMed]
  • Gould T. D., Pfeifer K., 1998. Imprinting of mouse Kvlqt1 is developmentally regulated. Hum. Mol. Genet. 7: 483–487 [PubMed]
  • Gregg C., Zhang J., Butler J. E., Haig D., Dulac C., 2010. Sex-specific parent-of-origin allelic expression in the mouse brain. Science 329: 682–685 [PMC free article] [PubMed]
  • Gregg C., Zhang J., Weissbourd B., Luo S., Schroth G. P., et al. , 2010b. High-resolution analysis of parent-of-origin allelic expression in the mouse brain. Science 329: 643–648 [PMC free article] [PubMed]
  • Hagiwara Y., Hirai M., Nishiyama K., Kanazawa I., Ueda T., et al. , 1997. Screening for imprinted genes by allelic message display: identification of a paternally expressed gene impact on mouse chromosome 18. Proc. Natl. Acad. Sci. USA 94: 9249–9254 [PMC free article] [PubMed]
  • Hatada I., Morita S., Obata Y., Sotomaru Y., Shimoda M., et al. , 2001. Identification of a new imprinted gene, Rian, on mouse chromosome 12 by fluorescent differential display screening. J. Biochem. 130: 187–190 [PubMed]
  • Hatada I., Mukai T., 1995. Genomic imprinting of p57KIP2, a cyclin-dependent kinase inhibitor, in mouse. Nat. Genet. 11: 204–206 [PubMed]
  • Hatada I., Sugama T., Mukai T., 1993. A new imprinted gene cloned by a methylation-sensitive genome scanning method. Nucleic Acids Res. 21: 5577–5582 [PMC free article] [PubMed]
  • Henckel A., Arnaud P., 2010. Genome-wide identification of new imprinted genes. Brief. Funct. Genomics 9: 304–314 [PubMed]
  • Hiby S. E., Lough M., Keverne E. B., Surani M. A., Loke Y. W., et al. , 2001. Paternal monoallelic expression of PEG3 in the human placenta. Hum. Mol. Genet. 10: 1093–1100 [PubMed]
  • Hu J. F., Vu T. H., Hoffman A. R., 1995. Differential biallelic activation of three insulin-like growth factor II promoters in the mouse central nervous system. Mol. Endocrinol. 9: 628–636 [PubMed]
  • Hudson Q. J., Kulinski T. M., Huetter S. P., Barlow D. P., 2010. Genomic imprinting mechanisms in embryonic and extraembryonic mouse tissues. Heredity 105: 45–56 [PMC free article] [PubMed]
  • Huynh K. D., Lee J. T., 2005. X-chromosome inactivation: a hypothesis linking ontogeny and phylogeny. Nat. Rev. Genet. 6: 410–418 [PubMed]
  • Jiang Y. H., Bressler J., Beaudet A. L., 2004. Epigenetics and human disease. Annu. Rev. Genomics Hum. Genet. 5: 479–510 [PubMed]
  • Kagitani F., Kuroiwa Y., Wakana S., Shiroishi T., Miyoshi N., et al. , 1997. Peg5/Neuronatin is an imprinted gene located on sub-distal chromosome 2 in the mouse. Nucleic Acids Res. 25: 3428–3432 [PMC free article] [PubMed]
  • Kaneko-Ishino T., Kuroiwa Y., Miyoshi N., Kohda T., Suzuki R., et al. , 1995. Peg1/Mest imprinted gene on chromosome 6 identified by cDNA subtraction hybridization. Nat. Genet. 11: 52–59 [PubMed]
  • Kim J., Lu X., Stubbs L., 1999. Zim1, a maternally expressed mouse Kruppel-type zinc-finger gene located in proximal chromosome 7. Hum. Mol. Genet. 8: 847–854 [PubMed]
  • Kim J., Noskov V. N., Lu X., Bergmann A., Ren X., et al. , 2000. Discovery of a novel, paternally expressed ubiquitin-specific processing protease gene through comparative analysis of an imprinted region of mouse chromosome 7 and human chromosome 19q13.4. Genome Res. 10: 1138–1147 [PMC free article] [PubMed]
  • Kobayashi H., Yamada K., Morita S., Hiura H., Fukuda A., et al. , 2009. Identification of the mouse paternally expressed imprinted gene Zdbf2 on chromosome 1 and its imprinted human homolog ZDBF2 on chromosome 2. Genomics 93: 461–472 [PubMed]
  • Kuzmin A., Han Z., Golding M. C., Mann M. R., Latham K. E., et al. , 2008. The PcG gene Sfmbt2 is paternally expressed in extraembryonic tissues. Gene Expr. Patterns 8: 107–116 [PMC free article] [PubMed]
  • Levin J. Z., Yassour M., Adiconis X., Nusbaum C., Thompson D. A., et al. , 2010. Comprehensive comparative analysis of strand-specific RNA sequencing methods. Nat. Methods 7: 709–715 [PMC free article] [PubMed]
  • Li H., Durbin R., 2009. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25: 1754–1760 [PMC free article] [PubMed]
  • Li H., Handsaker B., Wysoker A., Fennell T., Ruan J., et al. , 2009. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25: 2078–2079 [PMC free article] [PubMed]
  • Lomvardas S., Barnea G., Pisapia D. J., Mendelsohn M., Kirkland J., et al. , 2006. Interchromosomal interactions and olfactory receptor choice. Cell 126: 403–413 [PubMed]
  • Loughney K., Snyder P. B., Uher L., Rosman G. J., Ferguson K., et al. , 1999. Isolation and characterization of PDE10A, a novel human 3′, 5′-cyclic nucleotide phosphodiesterase. Gene 234: 109–117 [PubMed]
  • Luedi P. P., Hartemink A. J., Jirtle R. L., 2005. Genome-wide prediction of imprinted murine genes. Genome Res. 15: 875–884 [PMC free article] [PubMed]
  • Luedi P. P., Dietrich F. S., Weidman J. R., Bosko J. M., Jirtle R. L., et al. , 2007. Computational and experimental identification of novel human imprinted genes. Genome Res. 17: 1723–1730 [PMC free article] [PubMed]
  • MacDonald H. R., Wevrick R., 1997. The necdin gene is deleted in Prader-Willi syndrome and is imprinted in human and mouse. Hum. Mol. Genet. 6: 1873–1878 [PubMed]
  • Maeda N., Hayashizaki Y., 2006. Genome-wide survey of imprinted genes. Cytogenet. Genome Res. 113: 144–152 [PubMed]
  • Marsh S., 2007.  Pyrosequencing Protocols. Humana Press, Totowa, NJ
  • Miyoshi N., Kuroiwa Y., Kohda T., Shitara H., Yonekawa H., et al. , 1998. Identification of the Meg1/Grb10 imprinted gene on mouse proximal chromosome 11, a candidate for the Silver-Russell syndrome gene. Proc. Natl. Acad. Sci. USA 95: 1102–1107 [PMC free article] [PubMed]
  • Miyoshi N., Wagatsuma H., Wakana S., Shiroishi T., Nomura M., et al. , 2000. Identification of an imprinted gene, Meg3/Gtl2 and its human homologue MEG3, first mapped on mouse distal chromosome 12 and human chromosome 14q. Genes Cells 5: 211–220 [PubMed]
  • Mizuno Y., Sotomaru Y., Katsuzawa Y., Kono T., Meguro M., et al. , 2002. Asb4, Ata3, and Dcn are novel imprinted genes identified by high-throughput screening using RIKEN cDNA microarray. Biochem. Biophys. Res. Commun. 290: 1499–1505 [PubMed]
  • Moore T., Constancia M., Zubair M., Bailleul B., Feil R., et al. , 1997. Multiple imprinted sense and antisense transcripts, differential methylation and tandem repeats in a putative imprinting control region upstream of mouse Igf2. Proc. Natl. Acad. Sci. USA 94: 12509–12514 [PMC free article] [PubMed]
  • Morison I. M., Ramsay J. P., Spencer H. G., 2005. A census of mammalian imprinting. Trends Genet. 21: 457–465 [PubMed]
  • Mortazavi A., Williams B. A., McCue K., Schaeffer L., Wold B., 2008. Mapping and quantifying mammalian transcriptomes by RNA-Seq. Nat. Methods 5: 621–628 [PubMed]
  • Nikaido I., Saito C., Mizuno Y., Meguro M., Bono H., et al. , 2003. Discovery of imprinted transcripts in the mouse transcriptome using large-scale expression profiling. Genome Res. 13: 1402–1409 [PMC free article] [PubMed]
  • Ono R., Shiura H., Aburatani H., Kohda T., Kaneko-Ishino T., et al. , 2003. Identification of a large novel imprinted gene cluster on mouse proximal chromosome 6. Genome Res. 13: 1696–1705 [PMC free article] [PubMed]
  • Parker-Katiraee L., Carson A. R., Yamada T., Arnaud P., Feil R., et al. , 2007. Identification of the imprinted KLF14 transcription factor undergoing human-specific accelerated evolution. PLoS Genet. 3: e65. [PMC free article] [PubMed]
  • Pauler F. M., Barlow D. P., 2006. Imprinting mechanisms–it only takes two. Genes Dev. 20: 1203–1206 [PubMed]
  • Peters J., Wroe S. F., Wells C. A., Miller H. J., Bodle D., et al. , 1999. A cluster of oppositely imprinted transcripts at the Gnas locus in the distal imprinting region of mouse chromosome 2. Proc. Natl. Acad. Sci. USA 96: 3830–3835 [PMC free article] [PubMed]
  • Piras G., El Kharroubi A., Kozlov S., Escalante-Alcalde D., Hernandez L., et al. , 2000. Zac1 (Lot1), a potential tumor suppressor gene, and the gene for epsilon-sarcoglycan are maternally imprinted genes: identification by a subtractive screen of novel uniparental fibroblast lines. Mol. Cell. Biol. 20: 3308–3315 [PMC free article] [PubMed]
  • Pollard K. S., Serre D., Wang X., Tao H., Grundberg E., et al. , 2008. A genome-wide approach to identifying novel-imprinted genes. Hum. Genet. 122: 625–634 [PubMed]
  • Proudhon C., Bourc’his D., 2010. Identification and resolution of artifacts in the interpretation of imprinted gene expression. Brief. Funct. Genomics 9: 374–384 [PMC free article] [PubMed]
  • Reik W., Walter J., 2001. Genomic imprinting: parental influence on the genome. Nat. Rev. Genet. 2: 21–32 [PubMed]
  • Sado T., Ferguson-Smith A. C., 2005. Imprinted X inactivation and reprogramming in the preimplantation mouse embryo. Hum. Mol. Genet. 14(Spec No 1): R59–R64 [PubMed]
  • Sakamoto K., Tamamura Y., Katsube K., Yamaguchi A., 2008. Zfp64 participates in Notch signaling and regulates differentiation in mesenchymal cells. J. Cell Sci. 121: 1613–1623 [PubMed]
  • Schmidt J. V., Matteson P. G., Jones B. K., Guan X. J., Tilghman S. M., 2000. The Dlk1 and Gtl2 genes are linked and reciprocally imprinted. Genes Dev. 14: 1997–2002 [PMC free article] [PubMed]
  • Schulz R., Menheniott T. R., Woodfine K., Wood A. J., Choi J. D., et al. , 2006. Chromosome-wide identification of novel imprinted genes using microarrays and uniparental disomies. Nucleic Acids Res. 34: e88. [PMC free article] [PubMed]
  • Seitz H., Royo H., Bortolin M. L., Lin S. P., Ferguson-Smith A. C., et al. , 2004. A large imprinted microRNA gene cluster at the mouse Dlk1-Gtl2 domain. Genome Res. 14: 1741–1748 [PMC free article] [PubMed]
  • Seitz H., Youngson N., Lin S. P., Dalbert S., Paulsen M., et al. , 2003. Imprinted microRNA genes transcribed antisense to a reciprocally imprinted retrotransposon-like gene. Nat. Genet. 34: 261–262 [PubMed]
  • Smith R. J., Arnaud P., Konfortova G., Dean W. L., Beechey C. V., et al. , 2002. The mouse Zac1 locus: basis for imprinting and comparison with human ZAC. Gene 292: 101–112 [PubMed]
  • Smith R. J., Dean W., Konfortova G., Kelsey G., 2003. Identification of novel imprinted genes in a genome-wide screen for maternal methylation. Genome Res. 13: 558–569 [PMC free article] [PubMed]
  • Soderling S. H., Bayuga S. J., Beavo J. A., 1999. Isolation and characterization of a dual-substrate phosphodiesterase gene family: PDE10A. Proc. Natl. Acad. Sci. USA 96: 7071–7076 [PMC free article] [PubMed]
  • Sritanaudomchai H., Ma H., Clepper L., Gokhale S., Bogan R., et al. , 2010. Discovery of a novel imprinted gene by transcriptional analysis of parthenogenetic embryonic stem cells. Hum. Reprod. 25: 1927–1941 [PMC free article] [PubMed]
  • Tocharus J., Tsuchiya A., Kajikawa M., Ueta Y., Oka C., et al. , 2004. Developmentally regulated expression of mouse HtrA3 and its role as an inhibitor of TGF-beta signaling. Dev. Growth Differ. 46: 257–274 [PubMed]
  • Umlauf D., Goto Y., Cao R., Cerqueira F., Wagschal A., et al. , 2004. Imprinting along the Kcnq1 domain on mouse chromosome 7 involves repressive histone methylation and recruitment of Polycomb group complexes. Nat. Genet. 36: 1296–1300 [PubMed]
  • Wagschal A., Feil R., 2006. Genomic imprinting in the placenta. Cytogenet. Genome Res. 113: 90–98 [PubMed]
  • Wang X., Soloway P. D., Clark A. G., 2010. Paternally biased X inactivation in mouse neonatal brain. Genome Biol. 11: R79. [PMC free article] [PubMed]
  • Wang X., Sun Q., McGrath S. D., Mardis E. R., Soloway P. D., et al. , 2008. Transcriptome-wide identification of novel imprinted genes in neonatal mouse brain. PLoS ONE 3: e3839. [PMC free article] [PubMed]
  • Wood A. J., Roberts R. G., Monk D., Moore G. E., Schulz R., et al. , 2007. A screen for retrotransposed imprinted genes reveals an association between X chromosome homology and maternal germ-line methylation. PLoS Genet. 3: e20. [PMC free article] [PubMed]
  • Wutz A., Smrzka O. W., Schweifer N., Schellander K., Wagner E. F., et al. , 1997. Imprinted expression of the Igf2r gene depends on an intronic CpG island. Nature 389: 745–749 [PubMed]
  • Yang H. H., Hu Y., Edmonson M., Buetow K., Lee M. P., 2003. Computation method to identify differential allelic gene expression and novel imprinted genes. Bioinformatics 19: 952–955 [PubMed]
  • Zwart R., Sleutels F., Wutz A., Schinkel A. H., Barlow D. P., 2001. Bidirectional action of the Igf2r imprint control element on upstream and downstream imprinted genes. Genes Dev. 15: 2361–2366 [PMC free article] [PubMed]

Articles from Genetics are provided here courtesy of Genetics Society of America
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...