|Home | About | Journals | Submit | Contact Us | Français|
Conceived and designed the experiments: JPH CC. Performed the experiments: JPH BLO. Analyzed the data: JPH. Contributed reagents/materials/analysis tools: JPH CLS SVK. Wrote the paper: JPH.
Genomic imprinting is an exception to Mendelian genetics in that imprinted genes are expressed monoallelically, dependent on parental origin. In mammals, imprinted genes are critical in numerous developmental and physiological processes. Aberrant imprinted gene expression is implicated in several diseases including Prader-Willi/Angelman syndromes and cancer.
To identify novel imprinted genes, transcription profiling was performed on two uniparentally derived cell lines, androgenetic and parthenogenetic primary mouse embryonic fibroblasts. A maternally expressed transcript termed Imprinted RNA near Meg3/Gtl2 (Irm) was identified and its expression studied by Northern blotting and whole mounts in situ hybridization. The imprinted region that contains Irm has a parent of origin effect in three mammalian species, including the sheep callipyge locus. In mice and humans, both maternal and paternal uniparental disomies (UPD) cause embryonic growth and musculoskeletal abnormalities, indicating that both alleles likely express essential genes. To catalog all imprinted genes in this chromosomal region, twenty-five mouse mRNAs in a 1.96Mb span were investigated for allele specific expression.
Ten imprinted genes were elucidated. The imprinting of three paternally expressed protein coding genes (Dlk1, Peg11, and Dio3) was confirmed. Seven noncoding RNAs (Meg3/Gtl2, Anti-Peg11, Meg8, Irm/“Rian”, AK050713, AK053394, and Meg9/Mirg) are characterized by exclusive maternal expression. Intriguingly, the majority of these noncoding RNA genes contain microRNAs and/or snoRNAs within their introns, as do their human orthologs. Of the 52 identified microRNAs that map to this region, six are predicted to regulate negatively Dlk1, suggesting an additional mechanism for interactions between allelic gene products. Since several previous studies relied heavily on in silico analysis and RT-PCR, our findings from Northerns and cDNA cloning clarify the genomic organization of this region. Our results expand the number of maternally expressed noncoding RNAs whose loss may be responsible for the phenotypes associated with mouse pUPD12 and human pUPD14 syndromes.
Unlike other vertebrates, mammalian embryos have an absolute dependency on both parental genomes for successful development. Mouse embryos, in which the entire genome is of maternal (parthenogenetic) or paternal (androgenetic) origin, do not develop past embryonic day 10 , . The basis for this developmental failure is that some genes are subject to genomic imprinting and are expressed from only one parental allele. Consequently, loss of the expressed allele will render the embryo null for the imprinted gene's function, or conversely, duplication of the expressed allele may lead to overabundance of the imprinted product leading to detrimental effects. The primary goal of our research was to identify and characterize novel imprinted genes that may contribute to mammalian phenotypes displaying parent of origin effects.
In addition to its developmental roles, appropriate imprinting is important to human health into adulthood, as misexpression of imprinted genes has been implicated in a range of diseases. Imprinted genes that serve roles in growth regulation are often aberrantly expressed in cancers , . For example, loss of imprinting that leads to enhanced expression of the insulin-like growth factor 2 (Igf2) is frequently observed during cancer progression , . Imprinting has also been implicated in a variety of neurological disorders, of which the best characterized are Angelman and Prader-Willi syndromes . More recently, it was shown that the major autism susceptibility locus (AUTS1) on human 7q may have a bias towards paternal inheritance, suggesting the involvement of imprinting , . In rodents, female mice deficient in Paternally expressed gene 1/Mesoderm specific transcript (Peg1/MEST)  or Paternally expressed gene 3 (Peg3)  have aberrant maternal nurturing behavior with low incidences of pup retrieval, nest building, and milk ejection. These findings clearly illustrate the necessity to gain a better understanding of imprinting in mammalian biology. To achieve this goal, an extensive screen for novel imprinted genes has been undertaken.
In the mouse, most imprinted genes are found in clusters that are distributed among eleven regions on eight chromosomes . These areas were defined by using Robertsonian translocations to generate embryos that are uniparental for specific chromosomal regions that cause obvious defects in development, post-natal growth, and behavior . Many imprinted genes have been identified within these regions , . Although the translocation studies may have been exhaustive in defining regions whose uniparental inheritance resulted in overt phenotypes, not all imprinted mouse genes map to these regions. The imprinted mouse genes, Zac1 , Ins1 , Grf1 , Peg1/MEST , Nnat , Htr2 , and Impact , are known to lie outside these regions on seven different chromosomes forming “microimprinted domains,” revealing that imprinted genes are more widely distributed than previously anticipated. Given the significance of imprinting, the identification of other imprinted genes is both necessary and informative, as these findings may provide further insights into the roles of imprinting in mammalian development, behavior, and its contribution to human pathological conditions. Moreover, the elucidation about why and how this unusual form of mammalian epigenetic gene regulation evolved requires additional insights from a more complete analysis of all imprinted genes and regions.
The derivation of androgenetic and parthenogenetic mouse embryonic fibroblast (PMEF) lines that stably retain the parent-of-origin pattern of imprinted gene expression has been previously described. These lines were used as source of material for a subtractive screen for maternally imprinted genes that led to the identification of two novel imprinted genes, ε-sarcoglycan (Sgce) and Zac1, mapping to mouse chromosomes 6 and 10, respectively . Concomitantly, experiments to determine the effects of DNA methylation and histone acetylation on the maintenance of the imprinted status of gene expression employed these cell lines . Here, we utilized these cell lines in conjunction with DNA microarray technology to search for additional novel imprinted transcripts. We report the isolation and characterization of a maternally expressed RNA called Irm (Imprinted RNA near Meg3) that is located on the distal region of mouse chromosome 12 near the known imprinted genes, Delta-like homolog 1 (Dlk1) and Maternally expressed gene 3/Gene trap locus 2 (Meg3/Gtl2). This chromosomal segment has been identified in three different mammalian species - humans, mice, and sheep - as having significant parent of origin effects on embryogenesis, muscle growth and possibly post-natal behavior. In both sheep and humans, this region contains six described genes: Dlk1, Dlk1 associated transcript “DAT”, Meg3/Gtl2, Peg11, Peg11 antisense transcript, and Meg8 . In mice, there are six imprinted transcripts for which cDNAs have been cloned. Meg3/Gtl2 and Dlk1 were the first two imprinted genes found in this gene cluster , , , , . Two noncoding maternally expressed RNAs, Rian  and the C/D snoRNA MBII-343 , were discovered later that map to this region, although their precise chromosomal location was unknown at the time. Dio3 displays a bias for paternal expression, although the maternal allele is expressed at roughly one-quarter of the level of its paternal counterpart , , . Finally, MicroRNA containing gene (Mirg) was identified as another maternally expressed noncoding RNA . To date, cDNA clones have not been isolated for any other imprinted genes in this region.
MicroRNAs are an abundant class of endogenously expressed noncoding RNAs that are 19–24 nucleotides in length. Previously, we have shown that microRNAs can act as cancer classifiers ,  and that microRNAs that are differentially regulated are important in regulating the levels of both oncogenes and tumor suppressors in this disease , . Recently implicated directly in cancer initiation and/or progression using mouse models (reviewed in ), microRNAs regulate gene expression usually through imperfect base-pairing interactions with the 3′ UTR of its target mRNAs leading to mRNA degradation and/or translational inhibition. Large numbers of mRNAs, often numbering in the hundreds, were shown to be most likely downregulated directly by each of twenty-six microRNAs investigated , . As would be expected for a potent gene regulator, mouse knockouts of miR-155 ,  and miR-1-2  demonstrate that that these microRNAs are essential for normal immune function and cardiac development, respectively. Also, the miR-17/92 cluster has been knocked out in mice, resulting in a neonatal death characterized by with lung hypoplasia and a ventricular septal defect .
129S1 and CzechII/Ei mice were purchased from Jackson laboratory and maintained according to the NIH animal care and use guidelines. All experiments involving animals were approved by the institutional review board. Androgenetic (AG) and parthenogenetic (PG) PMEFs were prepared as described previously . For the current study, passage six PG PMEFs and passage seventeen AG PMEFs were utilized. For the imprinting analysis, PMEFs were generated from day 13 mouse embryos (day of plug=d1) from appropriate 129S1 and CzechII/Ei crosses. Embryos were dissected and washed twice in sterile phosphate buffered saline solution. Decapitated and eviscerated embryos were passed several times through an 18-gauge needle and resuspended on a per embryo basis in 1 ml 10% heat inactivated fetal bovine serum (FBS) in DMEM supplemented with 100 µg/ml DNase and 500 µg/ml collagenase. Triturated embryos were incubated at 37°C for thirty minutes, centrifuged at 1200 g for five minutes, and ultimately, plated with two disassociated embryos per a 150 mm dish in 10% FBS/DMEM with penicillin and streptomycin. PMEF cells were grown for two passages prior to RNA isolation.
Total RNA was prepared from PMEF cells grown in a monolayer in 150 mm dishes using the Trizol reagent (Invitrogen) according to the manufacturer's directions. For DNA microarray analysis, mRNA was isolated from total RNA using the Oligotex mRNA purification system (Qiagen). Incyte Genomics reverse transcribed AG and PG polyA+ RNA with either Cy3 or Cy5 label. The labeled cDNA was then hybridized to the Incyte mouse GEM1 cDNA microarray according to their established protocol.
Total RNA from PMEFs or mouse tissues was treated with RQ1 RNase-free DNase (Promega) to remove residual genomic DNA, followed by heat inactivation of the endonuclease. 5 µg RNA was reverse transcribed by Superscript II RNase H- RT (Invitrogen) in 60 µl reaction volumes according to the manufacturer's procedure, except that the reaction contained both 25 ng/µl Oligo dT and 5 ng/µl random hexamers. PCR was performed using 1 µl of the RT reaction as a starting material according to standard procedures. PCR cycling parameters were typically 4 minutes at 94°C, followed by thirty cycles of 30 sec at 94°C, 30 sec at 58°C, 90 sec at 72°C, and a final seven minute extension at 72°C. Supplemental Table S1 has the sequences of oligonucleotide primers used in this study.
RT-PCR products were generated from PMEF or brain RNA that was obtained from appropriate embryos or adult mice, respectively, derived from 129S1 and CzechII/Ei matings. Subsequently, RT-PCR products were cleaned using the MinElute PCR purification kit (Qiagen). RT-PCR samples were then treated with and without the indicated restriction enzymes obtained from New England Biolabs. Digests were analyzed by agarose gel electrophoresis with ethidium bromide staining. Supplemental Table S2 has 129S1 and CzechII/Ei restriction site polymorphisms discovered during the course of this study.
An arrayed mouse day 19 embryo cDNA library (Origene MEA-1001) was screened by PCR to isolate Irm, Meg8, Meg9, Peg11, and anti-Peg11 clones, while an adult mouse brain cDNA library was used for Dlk1/“DAT” cloning (Origene MAB-1001). The gene-specific primers for cDNA library screening are denoted by “#” in Supplemental Table S1. RLM-Race (Ambion) was performed for Irm according to the manufacturer's directions, with the initial PCR utilizing the 5′ RACE Outer and Irm 197Dn primers. The subsequent amplification used the same upstream primer and the Irm 120Dn primer.
Northern blots were purchased from Origene (MB-2020) and Clontech (7763-1). Blots were either hybridized with cDNA probes corresponding to the Irm sequence 19-555, Dlk1 830-1274), Dlk1 Alt. pA (“DAT”) 1533-2044, Meg3 267-578, Meg8 1-294 (AF498299), or Meg9 2614-3009. Importantly, the Irm sequence that was selected contains sequences from exon 1 of Rian, thereby detecting both RNAs simultaneously. For Peg11 and Anti-Peg11, three Northerns were performed. Specifically, a double stranded DNA probe corresponding to Anti-Peg11 477-1052 (Peg11 2977-3552) was used to detect both transcripts simultaneously. To distinguish signals derived from sense and antisense Peg11, two single stranded DNA probes corresponding to the same region were made and then used separately for Northerns. Blots were hybridized in ExpressHyb (Clontech) at 68°C for at least two hours. Blots were then washed twice in 2× SSC/0.1% SDS at room temperature for thirty minutes and twice in 0.1× SSC/0.1% SDS at 50°C for thirty minutes. Whole mount in situ hybridization was performed on dissected mouse embryos at different gestation stages, according to the conventional technique  with the following modifications: hybridization was performed at 68°C; an overnight wash at 68°C was added to limit non-specific RNA-RNA hybridization. To synthesize sense and antisense cRNA probes, a segment of Irm complimentary DNA spanning nucleotides 19-555 that contains part of the first Rian exon was labeled by DIG-UTP by run-off transcription (Roche).
The identification of novel imprinted genes is often hampered by the difficulty in obtaining sufficient quantities of uniparental embryos or embryos with a particular uniparental disomy to screen for imprinted genes. To circumvent this problem, androgenetic (AG) and parthenogenetic (PG) PMEFs that maintain appropriate imprinted gene expression were established , . Poly A+ RNA purified from both AG and PG PMEFs were subjected to RNA expression profiling using an Incyte mouse GEM1 array, containing 8,638 cDNA probes corresponding to identified genes . In this screen, we expected to find both maternal and paternal imprinted genes to be differentially expressed. Genes that are modulated by passage number as well as genes whose expression depends on the levels of an imprinted gene product would also behave as differentially expressed. Table 1 shows the forty most differentially expressed genes as found by the DNA microarray analysis. Three known maternally expressed genes (Grb10, Igf2r, and Meg3/Gtl2) were identified, along with two paternally expressed transcripts (Peg3 and U2af35-rs1), validating the experimental approach.
For each of the forty ESTs that are listed in Table 1, RT-PCR was performed on total RNA from wild-type (WT), AG, and PG PMEFs to determine if expression was consistent with being imprinted. Figure 1 shows a typical RT-PCR result and demonstrates that appropriate imprinted gene expression is maintained in these uniparental cell lines. Grb10, Meg3/Gtl2, Peg3, and Dlk1 were selected as imprinted gene controls. As expected, the biallelic gene histone acetyltransferase 1 (Hat1) and imprinted genes (Grb10, Meg3/Gtl2, Peg3 and Dlk1) showed appropriate expression by RT-PCR in the three cell lines (Figure 1). Ultimately, nine of the forty ESTs (denoted by stars in Table 1) were highly differentially expressed between AG and PG PMEFs by RT-PCR (data not shown). Of these nine, five ESTs corresponded to imprinted genes that were known at the time (Grb10, Igf2r, Meg3/Gtl2, Peg3, and U2af35-rs1). This result suggested the possibility that a subset of the four remaining ESTs (corresponding to Postn, Mpeg1, Racgap1, and Irm) may in fact be imprinted. To demonstrate conclusively the imprinting status of a gene, an assessment of allele specific transcription was performed using an informative polymorphism for the four ESTs whose imprinting status was unknown. We found that Postn, Mpeg1, and Racgap1 were biallelically expressed (data not shown), suggesting that the genes corresponding to these ESTs may be regulated by passage number or by an imprinted gene product. Intriguingly, the final EST W89392 (Irm) with its high RNA levels in PG versus AG PMEFs connected to other ESTs that mapped to the same BAC (AC040981.2) as the known mouse imprinted gene Meg3/Gtl2. Meg3/Gtl2 is the founding member of a much larger imprinted gene region that has been implicated as having parent of origin effects in three mammalian species. Hence, an exhaustive screen was undertaken for a 1.96 Mb region around Irm to investigate the imprinting status of all contained mRNAs. For the sake of clarity, results will be presented from centromere to telomere. Figure 2 shows the genomic organization of this region in mice.
Six mRNAs (Yy1, Slc25a29, Hdcmp, Wars, AK147993, and Begain) as depicted in Figure 2 that are on the centromeric side of Dlk1 were tested by RT-PCR in WT, AG, and PG PMEFs to determine if they were differentially expressed between these samples. Similarly, two mRNAs, Ppp2r5c and Dync1h1, on the telomeric side of Dio3 were also tested. For seven of the eight genes, RT-PCR products were obtained from the three cell lines at similar levels, indicating that these genes are biallelically expressed in PMEFs (data not shown). Mouse Hdcmp (accession number AI132487) was undetectable by RT-PCR. Since the neighboring genes, Yy1, Slc25a29, Wars, and AK147993, are not imprinted in PMEFs, it is unlikely that Hdcmp is imprinted (see Figure 2).
For Begain RT-PCR, our original primer set fell within the region that is common to Begain mRNA irrespective of which of the two mouse Begain promoters is used. Our results were consistent with biallelic expression. In sheep, Begain has been recently shown to be biallelically expressed in all tissues when transcription is initiated at one of its two promoters . Transcripts derived from the other promoter in sheep are characterized by imprinted expression in a tissue specific manner. Since our RT-PCR product would detect transcripts derived from both promoters, multiple primer sets were designed to investigate expression from each mouse promoter. Unfortunately, the alternative first exons of Begain are quite GC rich and specific amplifications products were not observed. Hence, we can only conclude that at least one of the two Begain promoters is biallelic.
Dlk1 is paternally expressed in mice ,  and in other mammalian species with the exception of marsupials , . A DraIII restriction site polymorphism between CzechII/Ei and 129S1 mice in the Dlk1 cDNA was identified and used to confirm the imprinted status of Dlk1 (Figure 3A). Northern analysis showed that Dlk1 is expressed in many adult mouse tissues, including brain, kidney, testis, and thymus. Mouse brain expressed the most Dlk1 mRNA and is characterized by an additional and abundant 4.5 kb transcript that was not detected in the other investigated tissues (Figure 3B).
In sheep and humans, a neighboring gene “DAT” (DLK1 Associated Transcript) is also paternally expressed . In addition to their common paternal expression, DLK1 and “DAT” lie in the same transcriptional orientation with the polyA of DLK1 roughly 3.1 kb upstream of the polyA of “DAT” in humans. In mice, several ESTs from Unigene cluster Mm.458824 map to the syntenic region of “DAT”. An EcoO109I restriction site polymorphism demonstrated that “DAT” is paternally expressed in mice (Figure 3C). Northern analysis detected “DAT” only in the brain and its 4.5 kb length was identical to the brain-specific long form of mouse Dlk1 (Figure 3D). This finding in conjunction with the fact that the genomic distance between the polyAs of Dlk1 and “DAT” in mice is ~2.7 kb indicated that “DAT” was most likely an alternatively polyadenylated transcript derived from the Dlk1 gene. To test this hypothesis, RT-PCR was performed with primers before and after the canonical Dlk1 polyA site and yielded a product only in mouse brain (Figure 3E). Moreover, four independent cDNA clones from the Origene mouse brain cDNA library (MAB-1001) were isolated and sequenced, demonstrating unequivocally that the “DAT” sequences result from an alternative polyadenylated product from the Dlk1 gene (Accession numbers: EU434914-EU434917). The lengths of our two longest clones that contain the entire Dlk1 ORF are consistent with the size of the 4.5 kb form of alternatively polyadenylated Dlk1. The significance of this alternative pre-mRNA processing event is unknown, but EST analysis indicates that is conserved across mammals (data not shown). In addition to alternative polyadenylation, our isolated cDNA clones are characterized by alternative splicing within the Dlk1 ORF. The existence of these alternatively spliced products was independently confirmed by several mouse Dlk1 mRNAs and ESTs in GenBank.
Figure 4A shows that Meg3/Gtl2 is maternally expressed using an identified Alw26I restriction site polymorphism between CzechII/Ei and 129S1 mice. This result confirms several earlier studies , . By Northern analysis, Meg3/Gtl2 is highly expressed in the brain and testis (see Figure 4B). In adult mouse brain, the predominant form is 2.4 kb, while in testis, a 6.5 kb transcript is most abundant. In addition, Meg3/Gtl2 contains in its last intron (Relative to Accession number Y13832) the evolutionarily conserved microRNA miR-770 (see Figure 2 and Table 2). This observation is intriguing since H19, a maternally expressed noncoding RNA near Igf2, also is characterized by having an evolutionarily conserved microRNA miR-675 . Table 2 shows several mRNAs that include Bmp1, Bmp15, Fosb, Lmna, Sspn, Tmod1, and Trp53 that are predicted to be regulated by miR-770. If these genes are bona fide targets, loss of miR-770 that occurs in the associated pUPD syndromes could contribute to the observed phenotypes by increasing the levels of its targeted genes, many of which are implicated in placental, bone, and muscle biology.
Unlike other genes in this region where oligo(dT) and random primed reverse transcription reactions were used as a substrate for PCR, we also performed strand-specific RT reactions to ascertain if the observed product corresponded to Peg11 or its antisense transcript. As Figure 5A–C demonstrates, Anti-Peg11 was expressed from the maternal allele in both PMEFs and adult brain. In contrast, Peg11 was undetectable in PMEFs and was expressed at lower levels in the brain in comparison to its antisense transcript. Our results showed that Peg11 was paternally expressed in brain. Northern analyses were performed with double stranded DNA probes and then sequentially with two single-stranded probes. Figure 5D shows that Peg11 is broadly expressed but at low levels in many tissues. Anti-Peg11 on the other hand is far more restricted in its expression with a prominent signal only detected in the brain.
To date, there are no reports of a cloned cDNA corresponding to either Peg11 or Anti-Peg11 in any species. In mice, a computationally predicted open reading frame (ORF) termed Retrotransposon-like 1 (Rtl1) has been deposited in GenBank (NM_184109) that maps to Peg11. Hence, an arrayed day 19 embryonic cDNA library (Origene MEA-1001) was screened by PCR with Anti-Peg11 477Up and 1052Dn primer pair. This primer set does not distinguish between strands and would yield a product for both Peg11 and Anti-Peg11. Ultimately, two Peg11 cDNAs were isolated and sequenced. Both clones (Accession numbers: EU434914 and EU434918) exceeded 6 kb in length, revealing that Peg11 has two exons (see Figure 2). Given the length of the Peg11 as determined by Northern, our cDNA clones are nearly full length. Likewise, a 2.3 kb Anti-Peg11 cDNA was isolated and sequenced (Accession number EU434921). As Figure 2 shows, a large number of microRNAs map to the Anti-Peg11/Peg11 region and all are in the same transcriptional orientation as Anti-Peg11. Our Anti-Peg11 cDNA clone most likely represent a cleavage product of a longer Anti-Peg11 message that resulted as the primary miRNA transcript is processed into pre-miRNAs. This conclusion is based on the observation that our cDNA begins at the end of miR-431 precursor and ends at the beginning of miR-127 precursor. Multiple experiments were unsuccessful in isolating cDNA clones that were upstream or downstream of our cloned sequence. Since Anti-Peg11 serves as a primary microRNA transcript, it would be nearly impossible to detect a full length Anti-Peg11 under conditions where the Microprocessor complex is functioning.
A large number of microRNAs map to the Anti-Peg11/Peg11 region (see Figure 2 and Table 2). Three microRNAs in this region, miR-431, miR-127, and miR-136, were shown previously to regulate Peg11 through a siRNA-like mechanism . Also, miR-127 was shown to regulate Bcl6 .Since individual microRNAs regulate large sets of genes often numbering in the hundreds , , we investigated the predicted targets for all microRNAs in this region. As Table 2 shows, several predicted targets for microRNAs in this region have been implicated in processes that when perturbed could result in phenotypes associated with the human and mouse UPD syndromes. These genes include Bmp2, Casq2, E2f1, E2f3, Grb10, Igf2, Jak2, Junb, Ldb3, Peg10, Plagl2, Magel2, Nnat, Pax3, Ryr1, Sgce, Sntb2, Sspn, Tcf21, Tln1, Tpm2, Trp53, Ttn, and Xpo5.
As Figure 6A demonstrates, Meg8 is a maternally expressed gene in mice as deduced from digests of RT-PCR products using an informative HgaI restriction site polymorphism between 129S1 and Czech II/Ei mice. To date, no reports have documented the cloning of Meg8 in any mammalian species. As such, the Origene arrayed day 19 cDNA library (MEA-1001) was screened for a full length Meg8 cDNA. The sequence of the isolated Meg8 cDNA has been deposited in GenBank (Accession number: EU434919). Comparison of Meg8 cDNA with the mouse genome revealed that this transcript has 26 exons (Figure 2). Conceptual translation of Meg8 revealed that it is a noncoding RNA that lacks significant ORFs (data not shown). Importantly, we found that Meg8 and Irm share ten common exons. Northern analysis for the region using a probe that is unique to Meg8 cDNA showed that it is expressed highly in the adult mouse brain and skin (Figure 6B). Significantly lower levels of Meg8 were detected in the heart. In contrast, Irm is barely detectable in the skin (Figure 7B), demonstrating that Meg8 and Irm differ in their expression patterns. Moreover, only Meg8 transcripts have the intron-encoded microRNAs miR-341, miR-1188, and miR-370 that lie upstream of Irm (see Figure 2 and Table 2). Of note, miR-370 is also conserved in humans. Several genes that are predicted to be regulated by these microRNAs are shown in Table 2 and include Grb10, Lmna, Peg10, and Trp53.
From the original large scale transcription profiling of AG and PG PMEFs, EST W89392 was an attractive candidate for a novel imprinted transcript. Since no polymorphisms were detected within this EST, a cDNA library derived from mouse day 19 embryos (Origene) was screened to obtain a full-length clone. The sequence of the longest of the three isolated cDNA clones (2361 nt) has been deposited in Genbank (Accession number AF498294). This cDNA was named Irm, Imprinted RNA near Meg3/Gtl2. Initial blast searches with Irm cDNA of the nonredundant Genbank database yielded only a single cDNA match to AK017440, an uncharacterized RNA that differs from Irm by alternate 5′ splice site choice in exon 8. Subsequent searches yielded partial matches to two additional RNAs, the noncoding RNA Rian  and the long form of MBII-343 snoRNA . In addition, four snoRNAs are present in introns of Irm . Conceptual translation of Irm revealed that it is a noncoding RNA as it only has several short ORFs that lack significant homology to any known protein as assayed by blast searches (data not shown).
Corresponding regions in Irm cDNA from 129S1 and CzechII/Ei mice were analyzed, leading to the identification of a single nucleotide polymorphism at position 1094A→T creating a NlaIII restriction site in the CzechII/Ei cDNA (Accession numbers: AF498295 and AF498296). RT-PCR was performed on RNA derived from individual PMEF lines with the following genotypes, 129S1, CzechII/Ei, 129S1×CzechII/Ei, or CzechII/Eix129S1. RT-PCR products were purified and digested with NlaIII restriction enzyme. The resulting restriction digestion pattern shown in Figure 7A demonstrates that this cDNA is indeed maternally expressed.
Since Meg8 and Irm cDNAs overlap substantially and are both maternally expressed, one possibility is that Irm represents a 5′ truncated transcript derived from the Meg8 promoter. An alternate hypothesis is that Meg8 and Irm have distinct promoters. Even though our Northern results support the latter hypothesis, RLM-RACE was performed to identify the transcriptional start site of Irm . In this procedure, RNA is incubated with calf intestinal phosphatase (CIP), leading to the 5′ dephosphorylation of tRNA, rRNA, and fragmented transcripts. After CIP inactivation, samples are treated with tobacco acid pyrophosphatase that catalyzes the removal the 7mG cap with the liberation of a free 5′ monophosphate. An RNA adapter is subsequently ligated and RT-PCR performed with adapter- and gene-specific primers. Sequencing of the resulting RT-PCR products revealed that the transcriptional start site of Irm is located twelve nucleotides upstream of the original Irm cDNA sequence (5′gccaatgatgac with Irm 1-120, data not shown). This site is identical to the Irm promoter and corresponding transcriptional start site predicted by a neural network based algorithm (http://www.fruitfly.org/seq-tools/promoter.html). Moreover, the 620 bp length of the first Irm exon supports that the fact that it is the first transcribed exon, as less than 1% of internal exons are longer than 400 bp .
Over 60 mouse ESTs corresponding to Irm RNA are present in the Genbank database, providing considerable insight into the Irm expression pattern. In adult mice, Irm ESTs are found in the pancreas, eye, testis, and different areas of the brain, including the hippocampus, hypothalamus, amygdala, and striatum. In mouse embryos, Irm ESTs are found in the developing pancreas, brain, heart, testis, optic vesicle, and the neural retina. Given the profile of Irm ESTs, Irm expression was evaluated using Northern and whole mount in situ hybridization techniques using a probe common to exon 1 of both Irm and Rian RNAs. Northern analysis on adult tissues (Figure 7B) revealed high levels and multiple transcripts for Irm RNA in the brain. Significantly lower Irm expression was detected after longer exposure times in adult mouse testis, stomach, skin, heart, and muscle. Northern analysis of embryos at different developmental stages revealed that Irm expression is temporally regulated during development (Figure 7B).
As mentioned previously, Irm RNA partially overlaps a cloned noncoding RNA termed Rian for RNA Imprinted and Accumulated in the Nucleus ; however, Rian Northerns have not been reported to verify its existence and size. Rian is characterized by several peculiar features. Most notably, it contains three retained introns, lacks major polyadenylation signals at its 3′ end, and terminates in a region of genomic DNA that has 15 adenosine residues. Therefore, the identified Rian cDNA most likely represents an internally primed Irm RNA that is reversed transcribed from the polyA stretch and does not exist in the cell as a mature species. Since our probe for the Irm Northern would detect Rian equally well, our results demonstrate that Irm is the major transcript derived from the Irm promoter (~2.5 kb) and raises doubts about the existence of Rian.
The spatio-temporal distribution of Irm transcripts within the embryo was determined by whole mount in situ hybridization at different stages of development. Prominent and persistent expression of Irm RNA was detected in the somites from embryonic day 9.5 (E9.5) onwards (Figures 8A–C). Other tissues expressing Irm include the first and second branchial arches and the limb cartilage (Figures 8A–C). In extraembryonic tissues, Irm expression exhibited a weak, but scattered expression in the yolk sac and to the maternal side of placenta (Figures 8A and 8C). Irm transcripts are also found in the brain, most prominently in the forebrain (Figure 8A) and later at E10.5 in the layers enriched with neural precursors in the midbrain (Figure 8B). Other regions within the nervous system that were positive for Irm expression were the cranial ganglia (predominantly the trigeminal ganglion), telencephalon, and midline of the caudal neural tube (Figures 8D and 8E). In summary, Northern, whole mount embryonic in situ hybridization, and EST analyses demonstrate that Irm RNAs are predominantly expressed in the brain but is not strictly brain-specific. RT-PCR experiments on seven adult mouse tissues for the long form of MBII-343 yielded a similar conclusion  and is noteworthy since this RNA overlaps Irm exon 6.
Multiple cDNA clones derived from Riken libraries map to the region between Irm and Meg9. Their accession numbers are from centromere to telomere as follows: AK050713, AK141557, AK163826, AK053394 and AK046809. For AK141557 and AK163826, RT-PCR with four sets of primer pairs did not detect a transcript in PMEFs or four adult tissues: brain, heart, muscle, and white adipose. These two putative cDNAs are unspliced, lack major polyadenylation signals, and end in a stretch of polyA mouse genomic sequence, suggesting that they may represent genomic DNA contaminants. For AK050713, we found that this transcript is maternally expressed in PMEFs (Figure 9A). Likewise, AK053394 is expressed exclusively from the maternal chromosome in PMEFs (Figure 9B). AK053394 is a spliced transcript with two exons and contains the entire sequence of AK046809.
Several spliced mRNAs and ESTs such as AK013406, CB193969, and BQ175773 are located approximately 73–88 kb downstream of the 3′ end of Irm. Sequences corresponding to AK013406 were determined for 129S1 and CzechII/Ei mice (Accession numbers: AF498297 and AF498298). The two alleles differ by a 14 bp insertion that abolishes an MboII restriction site in CzechII/Ei mice. Utilization of this informative restriction site on RT-PCR products revealed that this gene is maternally expressed. We thus named it Meg9 for Maternally expressed gene 9 (Figure 10A). A day 19 mouse embryo cDNA library was screened, resulting in the identification of three transcripts that are all longer than 3 kb (Accession numbers: EU434920, EU616812, and EU616813). These transcripts overlap the recently identified Mirg (MicroRNA containing gene) . In comparison to Mirg, the longest Meg9 clone has nine additional upstream exons and contains four more microRNAs: miR-382, miR-134, miR-668, and miR-485. The location of the 5′ end of Mirg indicated that this truncated transcript results as result of processing of the exonic miR-485 precursor. As Figure 10 shows, Northerns using a probe that would detect both Meg9 and Mirg demonstrate that Meg9/Mirg is widely expressed in many adult mouse tissues with the highest levels being in the brain, skin, and testis. Meg9/Mirg expression is also regulated during embryogenesis.
In total, Meg9 has thirteen microRNAs as shown in Figure 2 and Table 2. Since several hundred genes are now known that contain at least one microRNA (see http://microrna.sanger.ac.uk/sequences/), we propose that this transcript for the sake of clarity be referred to as Meg9/Mirg to avoid unnecessary confusion with the vast number of other microRNA containing genes. In this imprinted region alone, Meg3/Gtl2, Anti-Peg11, and Meg8 also contain microRNAs as already presented. The name Meg9/Mirg has already been used in four previous publications , , , . For the thirteen microRNAs that are contained in the Meg9/Mirg transcription unit, examples of genes that are predicted to be targeted by individual microRNAs as determined by miRanda are shown in Table 2. Some noteworthy mRNAs are Atp2a2, Bmp1, Calm2, Camk2g, Ctcf, Dicer1, Dlk1, Fgf10, H19, Kras, Mycn, Myf5, Myog, Myot, Mypn, Obscn, Peg10, Plagl1, Pdlim5, Sgce, Trp53, and Vim. Moreover, miR-134 was recently shown to regulate negatively Limk1 , Nanog, and LRH1 .
Previously, Dio3 has been shown to be preferentially expressed from the paternal allele with the maternal allele expressed at roughly one-fourth the level of the paternal allele , , . Our results confirm this finding (Figure 11A). In addition, the imprinting status of Dio3-as was investigated. Dio3-as was biallelically expressed without a preference for alleles (Figure 11B). This result is striking in that Dio3 and Dio3-as appear to use the same promoter but are bidirectionally expressed.
Imprinted genes are characterized by allele specific expression and to a much lesser extent differential pattern of methylation. Several methods have exploited these characteristics to identify novel imprinted genes. Imprinted genes such as U2af-rs1 , , Grf1 , Gnas , and Nesp  were identified by differential methylation analyses. Most screens, though, exploit the differential expression of imprinted genes per se. Subtractive cDNA hybridization  and fluorescent differential display ,  are also effective at identifying imprinted genes. In general, these techniques are quite labor intensive and require sufficient quantities of androgenetic and parthenogenetic embryonic material for analysis that is somewhat difficult to obtain. We utilized DNA microarray analysis on androgenetic and parthenogenetic PMEFs as an alternative approach to overcome these limitations. Six imprinted genes were among the forty most differentially expressed transcripts as determined by microarray analysis, revealing the relative efficiency of such an approach in identifying novel imprinted genes.
One maternally expressed transcript designated Irm (Imprinted RNA near Meg3/Gtl2) was a novel RNA that was cloned in its entirety. The Irm transcriptional start site as determined by RLM-RACE lies approximately 75 kb downstream from the end of Meg3/Gtl2 on distal mouse chromosome 12. Irm noncoding RNA partially overlaps Rian, whose in vivo existence is questionable given its cDNA sequence and our Northern results. Rian cDNA contains at least three retained introns, lacks major polyadenylation signals at its 3′ end, and terminates at a stretch of 15 adenosines in the genomic sequence. Since Irm maps to a region with significant parent of origin effects in three species, an exhaustive screen for imprinted genes was undertaken for a 1.96 Mb region centered on Irm. We demonstrated that ten genes are imprinted. For six genes (Dlk1, Peg11, Anti-Peg11, Meg8, Irm, and Meg9/Mirg), we isolated and sequenced mouse cDNA clones. As Table 3 illustrates, the majority of these genes have not been identified as a cloned cDNA in any mammalian species. For mouse Dlk1 and Meg9/Mirg, cDNA clones were described previously; however, our results increase our understanding of these genes. For Dlk1, we show by Northern, RT-PCR, and cDNA cloning that this gene has an abundant, alternatively polyadenylated transcript in the brain. Cross species comparisons reveal that this alternative polyadenylation site choice is a common characteristic of mammalian Dlk1. This finding clarifies the relationship between Dlk1 and “DAT” by demonstrating that mouse “DAT” is not an independent gene but an alternate Dlk1 transcript. For Meg9/Mirg, our cDNA contains nine upstream exons in relation to the original Mirg transcript. Mirg apparently is a truncated RNA generated when the exonic miR-485 precursor is removed from the Meg9 transcript. Of note, it has been previously suggested that all the maternally expressed transcripts might generated by alternative splicing from the same gene ; however, our results exclude this possibility.
The most significant implication of our results relates to how the identified imprinted genes may contribute to the parent of origin specific phenotypes that are observed in three mammalian species. Mice with either paternal or maternal uniparental disomy (UPDs) for this chromosomal segment exhibit non-viability, skeletal abnormalities, and embryonic growth defects . Specifically, paternal UPD leads to placentomegaly, costal cartilage defects and late gestational lethality, while maternal UPD causes retarded growth and perinatal death. A second imprinting effect was revealed by a gene trap-LacZ insertion upstream of Meg3/Gtl2 that causes dwarfism in mice inheriting the Gtl2LacZ mutant allele from their father , . These mouse mutants phenocopy many features associated with human individuals that harbor UPD for the syntenic region of chromosome 14. Patients with UPD14 have numerous pathologies including cardiomyopathy, dwarfism, short limbs, mental retardation, scoliosis, musculoskeletal abnormalities, and intrauterine growth defects , , . In addition, human 14q32 was recently identified as an imprinted modifier for bipolar affected disorder , delineating another potential role for this chromosomal segment in disease. In sheep, this region has been implicated as the locus responsible for the callipyge (“beautiful buttocks”) mutation that causes skeletal muscle hypertrophy when a single mutant allele of paternal origin is present . This collection of phenotypes clearly illustrates the importance of developing a better understanding of this imprinted cluster.
Each of the three paternally expressed, protein coding genes in this imprinted region has been knocked out in mice. As previously discussed, Dio3 is an atypical imprinted gene in that the paternal allele is preferentially but not exclusively expressed. A homozygous knockout of Dio3 has a pronounced phenotype characterized by partial lethality (~25%) and a 65% reduction in size at weaning age that persists throughout life . Heterozygous mice where the mutated Dio3 allele is inherited from the father are viable and are subtly growth retarded . Dlk1 knockout mice are characterized by 50% lethality by two days of age . These mice have both prenatal and postnatal growth retardation with survivors exhibiting accelerated adiposity. Finally, mouse deletions of Peg11 and several of its associated antisense microRNAs have recently been described . Upon either maternal or paternal inheritance of the knockout, mice are characterized by a lethal phenotype that is mouse strain-specific. Upon paternal inheritance of the knockout, Peg11 expression is lost while the maternally expressed microRNAs were unaffected in their expression. Maternal inheritance of the knockout results in loss of microRNAs that are antisense to Peg11, leading to elevated Peg11 expression. This finding suggested that loss or overexpression of Peg11 protein is important; however, an alternate hypothesis that can not be excluded is that Peg11 mRNA is important instead as it serves as a functional sink to limit the action of several microRNAs. In summary, one can envision that the observed lethality for mouse mUPD12 is caused by loss of at least two if not all three protein coding genes in this region.
For mice with pUPD12, the identification of the responsible for the pathology is far more elusive. Maternal inheritance of the IG-DMR knockout upstream of Meg3/Gtl2 leads to an absence of all investigated maternally expressed noncoding RNAs and is a fully penetrant perinatal lethal . The significance of noncoding RNAs in imprinting, as well as other cellular processes such as transvection and RNA interference is well documented , , . Many possible roles could be envisaged for Meg3/Gtl2, anti-Peg11, Meg8, Irm, and Meg9/Mirg genes. The significance of some of these genes as microRNA hosts will be discussed extensively later. Another role that these genes may have is in transcriptional regulation as these genes may compete for shared regulatory elements such as enhancers. The developmental expression pattern of Irm is strikingly similar to that found for Meg3/Gtl2 , suggesting that these genes may be coordinately regulated during embryogenesis to at least day E12.5. We showed by Northern analysis that Dlk1, Meg3/Gtl2, Meg8, Irm, and Meg9/Mirg are all highly expressed in the adult mouse brain, although they differ somewhat in their relative expression levels in other tissues. These results suggest that these genes may be coordinately regulated. Validating this hypothesis, Su and colleagues performed large scale transcription profiling using microarrays on 61 mouse tissues, showing that Dlk1, Meg3/Gtl2, Anti-Peg11, Irm, and Meg9/Mirg are in a genomic region of coordinate gene expression . Further experimental evidence supports this conclusion as the sheep callipyge (“beautiful buttocks”) mutation leads to altered expression of the five genes investigated in this imprinted cluster , .
The expression pattern of Irm and its coordinately regulated noncoding RNAs in this region is consistent with these genes contributing to the mutant phenotypes mentioned above. Irm expression, as determined by EST analysis, occurs in both the developing and adult retina, while uniparental disomy of syntenic human chromosome 14 in some individuals is associated with complete congenital achromatopsia . Whole mount in situ data revealed that Irm RNA is present in the developing brain, branchial arches, somites, and limb cartilage. In adults, Northern analysis revealed that Irm levels are high in the adult brain and present, albeit to a much lesser extent in the heart, skin, muscle, and testis. The expression in the first branchial arch is particularly intriguing, as maternal UPDs of distal mouse chromosome 12 are associated with failures in ossification of the maleus and incus inner ear bones, derived from this embryonic anlage . The Irm expression profile occurs almost exclusively in tissues that are adversely affected by documented parent of origin effects of this chromosomal region in sheep, mice, and humans.
The most obvious candidate for an essential noncoding RNA gene responsible for the pUPD phenotypes are microRNAs. In total, fifty-two microRNAs maps to this region and more than 80% have been identified in the orthologous human region. As previously reported, microRNA host genes and their associated microRNAs were shown to be in general coexpressed . Given the coordinate nature of gene regulation for this region, the expression patterns of these microRNAs should be similar to each other and to their respective host genes. One of the most important observations about microRNAs is that each one apparently regulates on average ~100 genes at the mRNA level , . Therefore, it maybe somewhat naive to attribute a phenotype related to specific microRNA loss with overexpression of a single mRNA target. As Table 2 illustrates, the microRNAs in the Meg3/Gtl2-Meg9/Mirg region are predicted to target genes involved in placental, bone, and muscle biology. For example, six microRNAs are predicted to regulate Dlk1 (underlined and written in bold in Table 2) suggesting another mechanism for interactions for allelic gene products besides Peg11 and its antisense microRNAs. This observation is intriguing as it may account, in part, for the polar overdominance of the callipyge mutation. In addition, several well known genes involved in cancer initiation and/or progression are predicted to be microRNA targets. For example, loss of specific microRNAs may lead to overexpression of well known oncogenes such as Bcl2, Fos, Igf2, Mycn, and Wnt1. Likewise, microRNA overexpression could effectively downregulate tumor suppressors including Brca1, Fhit, Rb1, Trp53, and Wwox. This finding is particularly important given that a common Adeno-Associated Virus (AAV) insertion site in mice has recently been identified that maps between miR-341 and miR-370 in Meg8 that causes hepatocellular carcinomas, suggesting that perturbed expression of microRNAs may be responsible. Ultimately, site directed mutagenesis in mice will help to resolve the relative importance of specific regions and genes within the Dlk1-Dio3 region in the establishment of imprinting, in regulation of transcription, and in contributions to linked disorders. Of special interest, the role of microRNAs will need to be addressed in detail.
Recently, two additional partial mRNA sequences have been deposited in GenBank for the putative transcripts, Mico1 (EF648170) and Mico1-os (EF648171) that map between mouse Dlk1 and Meg3/Gtl2 . These cDNAs are 2012 bp in length and are exact reverse complements of each other.
Sequences of Oligonucleotide Primers Used in this Study.
(0.10 MB DOC)
Restriction site polymorphisms in cDNAs between 129S1 and CzechII/Ei
(0.07 MB DOC)
We thank Drs. Dedra Cutler and Jr-Gang Cheng, Lidia Hernandez, and Lori Sewell for excellent technical assistance and fruitful discussions.
Competing Interests: The authors have declared that no competing interests exist.
Funding: This research was supported in part by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research; and with Federal funds from the National Cancer Institute, National Institutes of Health. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.