Activation of paternally expressed genes and perinatal death caused by deletion of the Gtl2 gene

The imprinted Dlk1-Gtl2 (delta like 1 homolog – gene trap locus 2) locus is approximately 1 Mb long and located on mouse distal chromosome 12 and on human chromosome 14q32 (da Rocha et al., 2008). This locus consists of multiple genes, which are active either on the chromosome inherited from the mother (maternally expressed) or the one inherited from the father (paternally expressed). The maternally expressed genes (MEGs) in this locus are all non-coding genes, including Gtl2, also known as maternally expressed gene 3 (Meg3) (Miyoshi et al., 2000), RNA imprinted and accumulated in the nucleus (Rian) (Hatada et al., 2001), miRNA containing gene (Mirg) (Seitz et al., 2003), anti-retrotransposon-like 1 (anti-Rtl1) and numerous C/D-box-containing small nucleolar RNAs (snoRNAs) (Cavaille et al., 2002) and micro RNAs (Seitz et al., 2004). Only three paternally expressed genes (PEGs), all protein coding, are found in this locus, including Dlk1 (Schmidt et al., 2000; Takada et al., 2000), Rtl1 (Byrne et al., 2010) and deiodinase, iodothyronine type III (Dio3) (Hernandez et al., 2002). The imprinting of these genes plays an important role in prenatal development of multiple tissues, including muscle, bone and brain, and controls postnatal neurological and metabolic functions (Charalambous et al., 2007; Wilkinson et al., 2007). Duplication of maternal distal chromosome 12 along with a deficiency of the corresponding paternal distal portion of the chromosome, MatDp(dist12), causes prenatal growth retardation, hypotrophic immature skeletal muscle and perinatal death (Tevendale et al., 2006); whereas duplication of paternal distal chromosome 12 with deletion of the corresponding maternal region, PatDp(dist12), causes a much more severe phenotype, i.e. death before 16.5 days of gestation (Tevendale et al., 2006). In humans, uniparental disomy of maternal chromosome 14 (mUPD14) or paternal chromosome 14 (pUPD14) is compatible with life. However, mUPD14 patients manifest short stature, small hands, scoliosis, mild developmental delay and early puberty (Kotzot, 2004). By contrast, pUPD14 patients display characteristic facial anomalies and skeletal defects (Cotter et al., 1997; Kagami et al., 2005). Recently, Stadtfeld et al. (Stadtfeld et al., 2010) reported that mouse induced pluripotent stem cells (iPSCs) with repressed expression of MEGs, including Gtl2, Rian, Mirg and numerous mirRNAs, contributed poorly to chimeras and failed to generate all-iPSC mice, whereas iPSCs with normal MEG expressions contributed to high-grade chimeras and produced viable all-iPSC mice. These data indicate that overexpression or lack of expression of imprinted genes in the Dlk1-Gtl2 locus has detrimental consequences to prenatal development and postnatal growth.

Imprinting of the Dlk1-Gtl2 locus is regulated by at least one cis-element called the intergenic differentially methylated region (IG-DMR) located between Dlk1 and Gtl2 (Lin et al., 2003). The IG-DMR on the maternally inherited chromosome is unmethylated, whereas the paternally inherited chromosome is methylated (Takada et al., 2002). Deletion of the IG-DMR on the maternal chromosome silences expression of MEGs and activated maternal alleles of the PEGs. However, a deletion on the paternal chromosome does not have this effect (Lin et al., 2003). The mechanism by which the IG-DMR regulates Dlk1-Gtl2 imprinting is unclear. It is likely that the IG-DMR acts as a positive regulator activating Gtl2 transcription and its downstream non-coding RNAs. Transcription of these genes in turn suppresses expression of PEGs on the same chromosome (da Rocha et al., 2008). Gtl2 is the first maternally expressed gene downstream of the IG-DMR. The Gtl2 gene contains a well-defined TATA-containing promoter. By contrast, no apparent promoters have been found for any other maternally expressed genes downstream of Gtl2 (Tierling et al., 2006). In addition, all these gene transcripts have the same orientation as Gtl2. Therefore, it has been suggested that all maternally expressed genes are transcribed from the Gtl2 promoter as one long polycistronic RNA, which is processed post-transcriptionally to generate individual intergenic transcripts (da Rocha et al., 2008). The Gtl2 gene was discovered in dwarf mice carrying paternal transmission of a lacZ insertion on distal chromosome 12 (Schuster-Gossler et al., 1998). The insertion location is approximately 1.7 kb upstream of the Gtl2 gene (Steshina et al., 2006). Paternal transmission of the lacZ insertion reduces expression of PEGs including Dlk1, Dio3 and Rtl1, which was believed to cause dwarfism (Sekita et al., 2006; Steshina et al., 2006). Interestingly, maternal transmission of this insertion reduces Gtl2 expression and causes an increase in Dlk1 expression on a C57BL/6 background (Steshina et al., 2006). The role of the Gtl2 gene in regulating imprinting remains elusive.

To investigate the in vivo function of Gtl2 and its role in regulating Dlk1-Gtl2 imprinting, we created a mouse model carrying a 5.9 kb deletion of genomic DNA including ∼300 bp of the Gtl2 promoter and exon 1 through exon 5 of the Gtl2 gene. We found that the maternally inherited Gtl2 deletion in embryos resulted in methylation of the IG-DMR, completely abolished expression of downstream MEGs, activated expression of PEGs and caused perinatal death. By contrast, the paternally inherited deletion did not have this effect. Our data indicate that activation of Gtl2 and its downstream maternal genes plays a crucial role in regulating Dlk1-Gtl2 imprinting, possibly by maintaining active status of the IG-DMR.

Gtl2 knockout mice were created using inGenious Targeting Laboratories (iTL, Stony Brook, NY). A 15.9 kb DNA fragment containing partial Gtl2 gene sequences from a mouse 129SvJ BAC clone was used to construct the targeting vector (Fig. 1A). An 8.0 kb region upstream of exon 1 and a 2.0 kb region after exon 5 were cloned into a vector to flank the neo cassette; designated long arm and short arm, respectively (Fig. 1B). After linearization, the target vector was transfected into iTL 129 Sv/Ev embryonic stem (ES) cells by electroporation. Seven positive clones were injected into mouse blastocysts and implanted into pseudopregnant C57BL/6J females. Male chimeras from two ES clones were mated to C57BL/6J females to achieve germline transmission. Two F1 male Gtl2 heterozygous mice were backcrossed to C57BL/6J females and bred up to ten generations. Phenotypic analysis was performed using generation F7 mice or later. Similar numbers of mice derived from each of the two F1 male Gtl2 heterozygous mice were used in the study. Animals were housed in the Center for Comparative Medicine at Massachusetts General Hospital and the protocol was approved by the Subcommittee on Research Animal Care.

The targeted allele was identified by PCR using tail DNA with primers GtlN2F (5′-AAAAGTCGCCTCCCCTACCCGGTAGAATTGAC-3′) and GtlwmR1 (5′-TAGAAGGGAACCAGAGCCGTGAAACCAATGGAGC-3′). The wild-type allele was identified with primers GtlwtF1 (5′-AGTGGACATGAACCTGTTTCCAGCCTGCATCC-3′) and GtlwmR1. Genotyping was confirmed by Southern blotting. Genomic DNA was digested with BamHI, and probed P_l and P_s (Fig. 1).

Embryonic tissues were fixed in Bouin's solution, dehydrated and embedded in paraffin. Six-micrometer sections were prepared and stained with Hematoxylin and Eosin (H&E). For glycogen examination, embryo forelimbs were dissected, snap-frozen (cooled isopentane) in liquid nitrogen and embedded in OCT. Ten-micrometer sections were prepared and stained using PAS (Periodic acid-Schiff).

Whole embryo forelimbs without skin were fixed overnight at 4°C with a solution of 2.5% formaldehyde and 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer, pH 7.4. Fixed muscles were washed with cacodylate buffer, post-fixed in osmium tetroxide, dehydrated in graded ethanol and Epon embedded. Semithin sections were stained with Toluidine Blue alkaline. Ultrathin sections were stained with uranyl acetate, lead citrate and viewed with a JEOL 1200EX transmission electron microscope.

Total embryo and placental RNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, CA). Gene expression was determined by northern blotting. Ten micrograms of total RNA was resolved on a 1% denaturing agarose gel. Blots were pre-hybridized at 65°C in a pre-hybridization solution (7% SDS and 0.5 M sodium phosphate, pH 7.5) and hybridized with radiolabeled Gtl2-specific cDNA probes (accession number: Y13832) (Schuster-Gossler et al., 1998), Dlk1 (NM_010052), anti-Rtl1 and Rtl1 (NM_184109), Rian (AK017440) (Cavaille et al., 2002), Mirg (AJ517767) (Seitz et al., 2003) or Dio3 (NM_172119) (Hernandez et al., 1999). The blots were stripped and re-probed with radiolabeled Gapdh probe as internal controls. All cDNA probes were labeled with [α-³²P]dCTP using Ready-To-Go DNA Labeling Beads (GE Healthcare, Piscataway, NJ, USA). Expression was quantified using a Storm 860 Phosphorimager and associated ImageQuant program (GE Healthcare). Total RNA from oocytes (approximately 100) was isolated and treated with DNase using a NucleoSpin RNA II Kit following the manufacturer's instructions (Macherey-Nagel, Bethlehem, PA). The first strand cDNA was synthesized using the ProtoScript AMV First Strand cDNA Synthesis Kit using the manufacturer's protocol (New England BioLabs, Ipswich, MA, USA). Expression of neo was detected by RT-PCR with the primer set 5′-ACAGTCGATGAATCCAGAAAAGCGG-3′ (forward) and 5′-GGTAGCCAACGCTATGTCCTGATAGC-3′ (reverse).

Ten nanograms of RNA were used to synthesize cDNAs for micro RNA using a TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA). Quantitative PCR was performed in quadruplicate using mir-127 or mir-410 specific TaqMan probes following the manufacturer's instructions on an Applied Biosystems 7500 Fast Real-Time PCR system. SnoRNA202 was used as the endogenous control. Relative quantities for individual microRNA were calculated using the associated SDS software. Values from wild-type embryos were designated as 100%. Values from Gtl2 KO/+ and Gtl2 +/KO embryos were normalized against those from wild-type embryos.

Bialleic expression of the Dlk1 gene in embryos was done as previously described (Lin et al., 2003). Total RNA was isolated from 16.5 days post coitum (dpc) embryos. After reverse transcription, a fragment flanking the polymorphism was amplified by PCR using primers: 5′-TCCTGAAGGTGTCCATGAAAGAGC-3′ (forward) and 5′-AAGCATAGCGTTCACTCGATTCCAC-3′ (reverse). Multiple PCR products were generated from several embryos obtained from at least three litters for each genotype and their respective wild-type littermates. The biallelic expression of Dlk1 was detected by sequencing the PCR products.

Fifteen micrograms of genomic DNA was used for restriction digestion, as described below. After being resolved on a 2% agarose gel, DNA was transferred onto Nytran SPC membranes using a Whatman TurboBlotter (GE Healthcare). The membrane was pre-hybridized with hybridization buffer (1% bovine serum albumin in 7% SDS and 0.5 M sodium phosphate, pH 7.5) at 65°C and hybridized with ³²P-labelled DNA probes as described below. The image was acquired using a Storm 860 phosphorimager. For the IG-DMR, DNA was digested with StuI plus methylation-insensitive MspI or methylation-sensitive HpaII or HhaI. The blot was hybridized with probe M4 as previously described (Takada et al., 2002). For Gtl2-DMR, DNA was digested with BsrI and enzymes indicated, and hybridized with a 1 kb DNA probe approximately 350 bp upstream the first Gtl2 exon. For the Dlk1-DMR, DNA was digested with NheI and enzymes indicated, and hybridized with D2 probe as previously described (Takada et al., 2000). Bisulfite treatment was performed using an EpiTect Bisulfite Kit from Qiagen following the manufacturer's instructions. After purification of the treated DNA, two round PCRs were used to amplify the IG-DMR region using Takara Premix EX Taq. Primers for the first round PCR were 5′-TAAGTGTTGTGGTTTGTTATGGGTA-3′ (forward) and 5′-CCATCCCCTATACTCAAAACATTCT-3′ (reverse). Primers for the second round PCR were 5′-TGGTTTGTTATGGGTAAGTTTTATG-3′ (forward) and 5′-CTTCCCTCACTCCAAAAATTAAAA-3′ (reverse). PCR products were cloned using the TOPO TA cloning kit (Invitrogen, Carlsbad, CA) and sequenced by the Massachusetts General Hospital DNA Core facility. For bisulfite sequencing of gamete DNA, approximately 120 oocytes, collected as previously described (Morita et al., 2000), were treated with an Easy DNA Methylation-Direct Kit using the manufacturer's protocol (Zymo Research, Orange, CA, USA).

Fisher's exact test was used to compare survival rates between embryos and neonates (Table 1). Student's t-test was used to analyze data of embryo body weight, placenta weight, myofiber count (Fig. 2G) and placental expression of Dlk1 and Rtl1 (Fig. 6).

Analysis of genomic sequences surrounding the transcription initiation site of the Gtl2 gene revealed that the Gtl2 core promoter contains a TATA box and a CAAT box. The Gtl2 gene consists of ten exons (Miyoshi et al., 2000; Schuster-Gossler et al., 1998). To inactivate the Gtl2 gene, we deleted the first five exons of the gene and the adjacent upstream promoter sequences of ∼300 bp by gene targeting (Fig. 1). The targeted region encodes approximately half of the mature Gtl2 transcript – approximately 1900 nucleotides long (Schuster-Gossler et al., 1998). Two lines of Gtl2 knockout KO mice obtained from two independent ES clones were used. Because it is a maternally expressed gene, we maintained the generated mice by crossing heterozygous males with wild-type C57BL/6J females.

Newborn pups carrying maternal or paternal Gtl2 gene deletions (Gtl2 KO/+ and Gtl2 +/KO, respectively) were obtained and examined postnatally (Table 1). All pups carrying the paternal deletion appeared healthy and lived well beyond 1 year. By contrast, none of the pups carrying the maternal Gtl2 deletion survived. They had similar body weight to their wild-type littermates (data not shown). No milk was visible in the dead pups' stomachs and their lungs sank in water, consistent with pre- or perinatal death. We also examined mice with deletions of both Gtl2 alleles and found that half of the heterozygous pups were dead postnatally, while all Gtl2 KO/KO pups were dead (Table 1). We next examined embryos before birth at 18.5 dpc to determine whether pups carrying the maternal deletion died prenatally. As expected, all Gtl2 +/KO embryos were alive before birth (Table 1) and weighed approximately 13.3% less than their wild-type littermates (Gtl2 +/KO 1.02±0.10 g versus Gtl2^+/+ 1.16±0.11 g; P=2×10e-7). Gtl2 KO/+ embryos were mostly alive at 18.5 dpc (Table 1). In contrast to the Gtl2 +/KO embryos, Gtl2 KO/+ embryos did not significantly differ in weight from their wild-type littermates (Gtl2 KO/+ 1.09±0.16 g versus Gtl2^+/+ 1.14±0.10 g; P=0.227). We examined 23 Gtl2 KO/KO embryos; 21 of them were alive. They also had similar body weight to their wild-type littermates (Gtl2 KO/KO 1.23±0.13 g versus Gtl2^+/+ 1.21±0.11 g; P=0.638). No apparent physical abnormalities were observed in these embryos. These data indicate that paternal transmission of the Gtl2 deletion did not cause perinatal death but resulted in embryonic growth retardation. By contrast, maternal deletion of the gene resulted in perinatal death, but did not affect embryonic growth.

Gtl2 KO/+ and Gtl2 KO/KO embryos of 18.5 dpc were dissected and examined histologically for abnormalities. No apparent anatomic defects were observed in the brain, heart, lung, kidney, liver or spleen. However, they contained obvious defects in all skeletal muscles, including diaphragm, intercostal, paraspinal and limb muscles. As shown in Fig. 2, for example, paraspinal muscles from wild-type embryos as well as Gtl2 +/KO embryos consist of mainly small, rounded myofibers with peripherally placed nuclei and eosinophilic contractile apparatus throughout the myocyte cytoplasm. An extremely low number of myofibers contain small areas of clearing (Fig. 2A,B). By contrast, the muscles from Gtl2 KO/+ and Gtl2 KO/KO embryos consisted of mainly myofibers with centrally localized nuclei or areas of central clearing (Fig. 2C,D). Staining of longitudinally sectioned muscles demonstrated that the area of central clearing is the space between centrally located nuclei along myofibers (Fig. 2D, inset). Central nucleation and central clearing are characteristics of immature muscle fibers called myotubes (Platzer, 1978; Stromer et al., 1974) or can be found postnatally in a severe congenital myopathy called X-linked myotubular myopathy (Pierson et al., 2005). Indeed, H&E skeletal muscle staining from wild-type 15.5 dpc embryos showed that virtually all myotubes contain either centrally localized nuclei or areas of central clearing (Fig. 2E). Similar phenotypes were observed in muscles from Gtl2 KO/+ embryos at the same age, although the myotubes showed larger areas of clearing (Fig. 2F). The number of myofibers with central nucleation or central clearing decreases with age in wild-type embryos as skeletal muscles mature (Fig. 2G). However, the number remained high in Gtl2 KO/+ embryos compared with age-matched wild-type littermates (Fig. 2G). Ultrastructural examination revealed appropriate organization of organelles and contractile apparatus in skeletal muscles from 18.5 dpc wild-type embryos (Fig. 3A). By contrast, myofibers from Gtl2 KO/+ embryos at the same age displayed internally placed nuclei and areas of central clearing, consistent with deposition of a substance that was dissolved during processing for electron microscopy (Fig. 3B,C). The most common muscle constituent with these staining characteristics is glycogen. We stained longitudinal section of forelimb muscle from 18.5 dpc Gtl2 KO/+ embryos by periodic acid Schiff (PAS) (Fig. 3D). The PAS-positive areas resided in the middle of the myofibers. The staining pattern was very similar to that of the central clearing found in the H&E-stained longitudinal section of Gtl2 KO/+ paraspinal muscle (Fig. 2D, inset) suggesting that the areas of clearing seen on H&E staining were filled with glycogen (Fig. 3D). These findings are consistent with reported ultrastructural studies of myotubes at gestational day 14-15 in mice (Platzer, 1978). These data suggest a delay in muscle maturation beyond the myotube stage in embryos with maternal deletion of the Gtl2 gene.

Two cDNA probes were used to determine Gtl2 transcript levels: one overlaps the exon 3 region (P_ex3), which had been deleted in the targeted allele; the other spans exons 8, 9 and 10 (P_ex10), which had been retained in the targeted allele. As expected, Gtl2 +/KO embryos expressed Gtl2 at a level comparable with the wild-type embryos (Fig. 4A,C), whereas no Gtl2 expression was detected in Gtl2 KO/+ embryos with both probes (Fig. 4A), suggesting that Gtl2 gene transcription was completely abolished in embryos with the maternal transmission of the deletion. To investigate how this deletion affects other downstream MEGs, we examined expression of several maternally expressed non-coding RNAs in 16.5 dpc embryos. By northern blotting, RNA levels of anti-Rtl1, Rian and Mirg in Gtl2 KO/+ embryos were reduced to 9.0, 15.0 and 5.8%, respectively, of those in the wild-type littermates (Fig. 4B,C). By quantitative RT-PCR, levels of miR-127 and miR-410 in Gtl2 KO/+ embryos were reduced to 3.5% and 4.1%, respectively, of those in wild-type littermates (Fig. 4D). By contrast, their expression levels did not change in Gtl2 +/KO embryos compared with wild-type littermates (Fig. 4D). Paternal deletion of the Gtl2 gene did not affect expression of PEGs, including Dlk1, Rtl1 and Dio3 (Fig. 4B,C). However, the transcript levels of Dlk1, Dio3 and Rtl1 were 210, 180 and 490% in Gtl2 KO/+ embryos, respectively, compared with wild-type littermates (Fig. 4C). This phenomenon is similar to that found in embryos carrying maternal transmission of the IG-DMR deletion, where elevated levels of Dlk1, Rtl1 and Dio3 in IG-DMR KO/+ embryos are due to activation of their maternal alleles (Lin et al., 2003). Increased Rtl1 expression is also partly due to lack of expression of anti-Rtl1, which is believed to contain micro RNAs inhibitory to Rtl1 (Lin et al., 2003; Seitz et al., 2003). To investigate whether deletion of the Gtl2 gene activates the maternal allele of PEGs, we crossed DBA/2 mice with heterozygous Gtl2 mice to generate embryos with paternal or maternal deletion of the Gtl2 gene. Dlk1 contains a polymorphism in the coding region (Lin et al., 2003). Transcripts from both alleles of Dlk1 were detected in embryos carrying maternal transmission of the Gtl2 deletion, indicating that maternal deletion of the Gtl2 gene activates the maternal allele of the paternal gene (Fig. 5).

Placentas of 18.5 dpc were stained with H&E. No apparent defects were observed in placentas carrying paternal, maternal or homozygous deletion of the Gtl2 gene (data not shown). In addition, maternal deletion and homozygous deletion of the Gtl2 gene did not significantly affect placental growth. The average weight of the Gtl2 KO/KO placentas was 0.104±0.018 g, while that of the litter-matched Gtl2^+/+ placentas was 0.105±0.015 (P=0.8814). Similarly, the Gtl2 KO/+ placental weight was 0.121±0.023 g versus the litter-matched Gtl2^+/+ placentas, at 0.113±0.016 g (P=0.2982). By contrast, paternal gene deletion resulted in a significant decrease in placental weight (Gtl2 +/KO 0.086±0.015 g versus Gtl2^+/+ 0.101±0.015 g, P=0.0241). The smaller placenta may be responsible for the smaller size of the Gtl2 +/KO embryos.

To determine the effect of Gtl2 deletion on placental gene expression, we investigated expression of maternal and paternal genes in 18.5 dpc placentas. Expression of MEGs was completely abolished in placentas carrying maternal deletion of the Gtl2 gene (Fig. 6), similar to the expression patterns in embryos (Fig. 4). In addition, Rtl1 expression was increased to approximately 400% of that seen in wild type (Fig. 6). Surprisingly, maternal Gtl2 deletion did not affect placental Dlk1 expression, which is in contrast to the increased expression seen in Gtl2 KO/+ embryos (Figs 4 and 6). We failed to detect any Dio3 expression in 18.5 dpc placenta. Therefore, we are unable to determine whether Dio3 expression was changed in Gtl2 KO/+ placenta. Paternal Gtl2 deletion did not affect expression of any placental MEGs (Fig. 6). By contrast, the paternal deletion significantly reduced expression of Dlk1 and Rtl1 to 50 and 60% of their respective wild types (Fig. 6). Taken together, these data raise the possibility that maternal and paternal alleles are regulated individually by separate mechanisms in placenta. We also investigated the methylation status at the IG-DMR in Gtl2 KO/+ placentas. We found that the IG-DMR was fully methylated (data not shown), which is consistent with IG-DMR methylation in embryos carrying a maternal Gtl2 deletion.

Deletion of the maternal IG-DMR, which is unmethylated, silences maternally expressed genes and activates maternal alleles of the paternally expressed genes in embryos (Lin et al., 2003). By contrast, deletion of the paternal IG-DMR, which is methylated, does not (Lin et al., 2003), suggesting that the function of IG-DMR is inactivated by methylation. Using methylation-sensitive Southern blotting, we detected both methylated and unmethylated IG-DMR bands in Gtl2^+/+ and Gtl2 +/KO embryos (Fig. 7A). However, only methylated IG-DMR was detected in Gtl2 KO/+ embryos (Fig. 7A). This finding was confirmed by bisulfite sequencing of the IG-DMR core region (Fig. 7D), indicating that the maternal deletion of Gtl2 results in IG-DMR methylation. The Gtl2 DMR covers approximately 1.5 kb upstream of the Gtl2 promoter, the first exon and part of the first intron (Paulsen et al., 2001; Takada et al., 2002). In our Gtl2 knockout model, the 3′ end of Gtl2-DMR is deleted. The remaining Gtl2-DMR is approximately 1 kb long and we examined its methylation. As shown in Fig. 7B, digestion of this region with HpaII revealed three unmethylated bands in Gtl2^+/+ embryos. However, two of them were not detected in Gtl2 KO/+ embryos, indicating that these sites were methylated in embryos carrying a maternal deletion of Gtl2. Similarly, HhaI digestion demonstrated increased methylation in the Gtl2-DMR in the Gtl2 KO/+ embryos compared with that in the Gtl2^+/+ embryos (Fig. 7B). This increase is not observed in Gtl2 +/KO embryos. The Dlk1 gene also contains a DMR located near its exon 5 (Takada et al., 2002). Neither paternal nor maternal Gtl2 gene deletion altered Dlk1-DMR methylation (Fig. 7C), indicating that maternal deletion of the Gtl2 gene leads to a partial epigenotype switch from maternal allele to paternal allele.

The neo cassette from the targeting vector was inserted approximately 12.7 kb downstream of the IG-DMR and is transcribed toward the IG-DMR (Fig. 1). Chotalia et al. (Chotalia et al., 2009) reported that transcription from the Nesp gene establishes methylation of DMRs in the Gnas locus in maternal germ cells. This raises the question as to whether gain of methylation at the IG-DMR in Gtl2 KO mice is due to transcription from the neo cassette. We examined the methylation status of the IG-DMR in oocytes obtained from mice carrying the Gtl2 deletion. We found that the IG-DMR remained unmethylated in oocytes carrying the Gtl2 deletion (Fig. 8A). It is important to mention that oocytes isolated from Gtl2 KO mice consist of two populations: those carrying the wild-type allele of Gtl2 and those carrying the targeted allele. Because the IG-DMR is unmethylated in wild-type oocytes, the fact that we did not detect any fully methylated DNA indicates that oocytes carrying the targeted allele do not contain methylated IG-DMR. We also detected neo cDNA in the oocytes obtained from Gtl2 KO mice, indicating that the neo is transcriptionally active in germ cells (Fig. 8B). These data suggest that existence of the neo does not influence IG-DMR methylation in germ cells.

We demonstrated that deletion of the maternal allele of the Gtl2 gene caused perinatal death (Table 1) and embryonic skeletal muscle developmental defects (Figs 2 and 3). Skeletal muscles from late gestation embryos carrying the maternal deletion of Gtl2 display characteristics of immature myotubes found in much younger embryos (Figs 2 and 3). The concept that developmental arrest at the myotube stage could cause severe muscle disease was first proposed by Spiro et al. (Spiro et al., 1966), where patients with severe congenital myopathy and skeletal muscle biopsies displaying small myofibers and markedly increased central nucleation were diagnosed with `myotubular myopathy'. Subsequent work has proven this term to be a misnomer; a murine model of myotubular myopathy displays normal histological development at birth, with central nucleation developing later in the course of the disease (Buj-Bello et al., 2002). Our histological findings in mice with maternal transmission of the Gtl2 deletion suggest arrest or delay in progression from the myotube stage in our mice, which may contribute to their lack of survival through the perinatal period.

Takahashi et al. (Takahashi et al., 2009) reported that deletion of both Gtl2 and its DMR abolished Gtl2 expression, but retained various levels of expression of other MEGs and did not significantly affect expression of PEGs. Interestingly, mice carrying this maternal deletion have a much milder phenotype; they were born alive and lived up to 4 weeks after birth (Takahashi et al., 2009). These data suggest that lack of Gtl2 expression may not directly cause the perinatal death and skeletal muscle developmental delays found in our Gtl2 KO/+ embryos. Dlk1 has been implicated in skeletal muscle development. It is involved in differentiation of skeletal stem cells (Abdallah et al., 2004) and skeletal muscle remodeling (Andersen et al., 2009). Davis et al. (Davis et al., 2004) reported that overexpression of Dlk1 resulted in skeletal muscle hypertrophy. However, a recent study by the Ferguson-Smith group (da Rocha et al., 2009) showed that high levels (two- to threefold of wild type) of Dlk1 expression enhances embryo growth, leading to skeletal but not skeletal muscle defects. Therefore, Dlk1 may not be the gene responsible for the delay in skeletal muscle development in Gtl2 KO mice. The gene expression pattern of the Dlk1-Gtl2 locus in Gtl2 KO/+ mice resembles that found in embryos carrying maternal deletion of the IG-DMR and with paternal uniparental disomy for chromosome 12 (da Rocha et al., 2008; Georgiades et al., 2000; Lin et al., 2007). These mice all had skeletal muscle developmental defects and died at late gestation or perinatally (da Rocha et al., 2008; Georgiades et al., 2000; Lin et al., 2007). Therefore, it is likely that the severe phenotype found in Gtl2 KO/+ mice is caused by contributions from more than one gene.

Maternal deletion of Gtl2 abolished expression of all MEGs examined (Figs 4 and 6). Although not yet proven, it has been suggested that all maternally expressed genes are transcribed into one giant polycistronic RNA, from which individual genes are derived by post-transcriptional processing (Tierling et al., 2006). Because a small portion of the Gtl2 promoter was deleted in the targeted allele, it is possible that lack of downstream MEG expression is due to promoter truncation. It is important to point out that it is not possible to discern whether the effect of the Gtl2 deletion in our animal model is due to lack of transcription from the Gtl2 promoter, which leads to silencing of all MEGs, or due to lack of Gtl2 expression, which in turn causes silencing of the downstream MEGs.

The gene expression pattern, i.e. silencing of MEGs and activation of PEGs, and the skeletal muscle defects in Gtl2 KO/+ embryos were strikingly similar to those seen in IG-DMR KO/+ embryos (Lin et al., 2003). Interestingly, the maternal allele of the IG-DMR was fully methylated in Gtl2 KO/+ embryos (Fig. 7). This suggests that IG-DMR methylation may be another mechanism by which Gtl2 deletion causes silencing of MEGs and activation of PEGs, resulting in the severe skeletal muscle phenotype. The mouse model by Takahashi et al. (Takahashi et al., 2009) carried a 10 kb DNA deletion including the Gtl2-DMR and the first five exons of Gtl2 (to distinguish this animal model from ours, we designated it as Gtl2-10kb KO). Gtl2 expression was abolished completely in embryos carrying the maternal deletion of this region (Gtl2-10kb MatKO). However, expression of other MEGs was not completely diminished, but rather modestly expressed in Gtl2-10kb MatKO embryos. In addition, the PEGs were not activated in the Gtl2-10kb MatKO embryos, but rather activated in embryos carrying a paternal deletion or a homozygous deletion (Takahashi et al., 2009). This contrasts significantly with our finding, as well as reports of the maternal IG-DMR deletion (Lin et al., 2003). One of the main differences in the targeted deletion between our Gtl2 KO mice and Gtl2-10kb KO mice is the orientation of the neo cassette in the targeted allele. It appears that transcription of the neo in the Gtl2-10kb KO mice is transcribed away from the IG-DMR and has the same orientation as the MEGs (Takahashi et al., 2009). By contrast, the neo cassette in our Gtl2 KO mice is transcribed towards the IG-DMR and has the opposite orientation to the MEGs. It is very interesting to note that Rian and Mirg levels in Gtl2-10kb KO embryos carrying a paternal or homozygous deletion are significantly higher than those in wild-type embryos. For example, levels of Rian in lung are 1.00, 0.66, 2.44 and 1.88 that of wild-type embryos, embryos with maternal, paternal deletion and homozygous deletion of Gtl2-10kb, respectively (Takahashi et al., 2009). One very likely interpretation is that the neo promoter, PGK, is responsible for the modest expression of MEGs in Gtl2-10kb KO embryos with a maternal deletion. Similarly, the neo promoter would also be responsible for additional expression of MEGs in embryos with a paternal or homozygous deletion, because the neo promoter renders MEGs bi-allelically active in these Gtl2-10kb KO mice. Therefore, one possible mechanism responsible for the discrepancy in imprinted gene expression between Gtl2-10kb KO mice and our Gtl2 KO mice is the transcription of the neo cassette.

Another major difference between our data and Takahashi's data is that maternal deletion of the Gtl2 gene in our animal model resulted in methylation of the IG-DMR (Fig. 7), whereas maternal deletion of the Gtl2 gene and its DMR did not affect IG-DMR methylation in Takahashi's model (Takahashi et al., 2009). The failure of IG-DMR methylation in Gtl2-10kb MatKO embryos may be attributed to the additional deletion upstream of the Gtl2 promoter, which may contain a cis-element necessary for IG-DMR methylation. Chotalia et al. (Chotalia et al., 2009) reported that transcription from an upstream gene results in methylation of germline DMRs in the Gnas locus. Therefore, another possibility is that the transcription from the neo cassette causes IG-DMR methylation in Gtl2 KO embryos. However, in our study, the IG-DMR was unmethylated in Gtl2 KO oocytes even though transcription of the neo was active in the oocytes (Fig. 8). Therefore, the gain of methylation at the IG-DMR in Gtl2 KO/+ mice occurs during embryogenesis. Sekita et al. (Sekita et al., 2006) reported that the IG-DMR remained unmethylated in Gtl2lacZMat mice, in which two to three copies of transgenes carrying neo cassettes, whose transcription is towards the IG-DMR, were inserted at 1.7 kb upstream of the Gtl2 gene. This suggests that the gain of methylation at the IG-DMR in our Gtl2 KO/+ embryos is not likely to be due to the existence of a neo cassette. Mice carrying a maternal transmission of a transgene insertion, such as the neo (Steshina et al., 2006) and neo/lacZ (Sekita et al., 2006), between the IG-DMR and the Gtl2 gene have reduced expression of MEGs, but the maternal IG-DMR remained unmethylated. This phenomenon is similar to that observed in Takahashi's Gtl2-10kb MatKO mice, in which the expression of MEGs, except Gtl2, was modest and the maternal IG-DMR was unmethylated (Takahashi et al., 2009). By contrast, mice carrying a maternal Gtl2 deletion did not express any MEGs and the IG-DMR was fully methylated. Taken together, we propose that a certain level of MEG expression is required to maintain an unmethylated status at the IG-DMR. MEG non-coding RNAs may act as inhibitors of IG-DMR methylation. In our Gtl2 KO mouse, Gtl2 deletion may reduce expression of downstream MEGs to a level unable to prevent IG-DMR methylation. The methylated IG-DMR in turn causes methylation at the rest of the Gtl2-DMR and further represses MEG expression.

We thank Dr Hang Lee for his assistance with statistical analysis. This work was supported by grants from the National Institutes of Health (A. Klibanski, R01DK40947), the Guthart Family Foundation and the Jarislowsky Foundation. Deposited in PMC for release after 12 months.