The imprinted Airn macro long non-coding (lnc) RNA is an established example of a cis-silencing lncRNA. Airn expression is necessary to initiate paternal-specific silencing of the Igf2r gene, which is followed by gain of a somatic DNA methylation imprint on the silent Igf2r promoter. However, the developmental requirements for Airn initiation of Igf2r silencing and the role of Airn or DNA methylation in maintaining stable Igf2r repression have not been investigated. Here, we use inducible systems to control Airn expression during mouse embryonic stem cell (ESC) differentiation. By turning Airn expression off during ESC differentiation, we show that continuous Airn expression is needed to maintain Igf2r silencing, but only until the paternal Igf2r promoter is methylated. By conditionally turning Airn expression on, we show that Airn initiation of Igf2r silencing is not limited to one developmental ‘window of opportunity’ and can be maintained in the absence of DNA methylation. Together, this study shows that Airn expression is both necessary and sufficient to silence Igf2r throughout ESC differentiation and that the somatic methylation imprint, although not required to initiate or maintain silencing, adds a secondary layer of repressive epigenetic information.
Genomic imprinting is an epigenetic process that causes parental-specific expression of a subset of mammalian genes (Ferguson-Smith, 2011). The two parental alleles of an imprinted gene co-exist in the same nuclear environment, but silencing is restricted to one allele; thus, genomic imprinting is a cis-acting silencing mechanism (Barlow, 2011). To date, 150 mouse imprinted genes have been identified (Williamson et al., 2012), with the majority occurring in clusters. In eight clusters, imprinted expression is controlled by a cis-regulatory DNA sequence - the imprint control element or ICE that acquires a DNA methylation imprint on one parental chromosome during gamete formation (Bartolomei and Ferguson-Smith, 2011). Imprinted protein-coding genes are silenced on the parental chromosome carrying the unmethylated ICE. In six clusters, the unmethylated ICE activates a lncRNA (Koerner et al., 2009) that, in three cases, controls silencing of the clustered protein-coding genes (Mancini-Dinardo et al., 2006; Sleutels et al., 2002; Williamson et al., 2011). These functional imprinted lncRNAs, Airn, Kcnq1ot1 and Nespas, represent invaluable epigenetic models for understanding how lncRNAs repress genes in cis. Global transcriptome analyses show that lncRNAs are found throughout the mammalian genome (Derrien et al., 2011). LncRNA abundance, tissue-specific and developmental regulation indicate functional cellular roles that may depend on recruiting chromatin modifiers for trans-regulation (Guttman and Rinn, 2012). Imprinted lncRNAs that silence in cis possess hallmarks - inefficient splicing, high repeat content, low conservation and short half-life - that indicate their transcription is more important than their lncRNA product. This lncRNA class has been termed ‘macro’ and may exert a silencing function on promoters and enhancers by transcriptional overlap (Guenzl and Barlow, 2012; Pauler et al., 2012).
In this study, we use the mouse Igf2r imprinted cluster as a model to investigate developmental regulation of the repressive action of the Airn macro lncRNA. Airn is paternally expressed and silences three protein-coding genes in cis: Igf2r, Slc22a2 and Slc22a3 (Sleutels et al., 2002). Of these, only Igf2r is essential for development (Wang et al., 1994) and shows imprinted expression in all embryonic, extra-embryonic and adult tissues that co-express Airn (Yamasaki et al., 2005). Imprinted expression of Slc22a2 and Slc22a3 is restricted to extra-embryonic lineages such as placenta and visceral yolk-sac endoderm (Hudson et al., 2011; Zwart et al., 2001). The Airn lncRNA promoter lies in Igf2r intron 2 within a 3.7 kb region genetically defined as the ICE (Lyle et al., 2000; Wutz et al., 1997). On the maternal chromosome, an ICE methylation imprint silences the Airn promoter, allowing expression of the three protein-coding genes (Wutz et al., 1997; Zwart et al., 2001). On the paternal chromosome, the unmethylated ICE drives expression of the 118 kb Airn transcript, a nuclear-localized, mostly unspliced and unstable lncRNA that overlaps the Igf2r promoter in antisense orientation (Seidl et al., 2006). Upon truncation of the Airn lncRNA to 3 kb, all three protein-coding genes are expressed biallelically, showing that Airn is required to initiate silencing (Sleutels et al., 2002). In placenta, the Airn lncRNA product has been shown to maintain Slc22a3 silencing by recruiting EHMT2 histone methyltransferase (Nagano et al., 2008). However, Igf2r silencing is independent of both EHMT2 and the Airn lncRNA product, but requires Airn transcriptional overlap that interferes with RNAPII recruitment to the Igf2r promoter (Latos et al., 2012).
An unresolved issue is whether Airn transcription is sufficient or whether it requires additional factors to initiate Igf2r silencing. Igf2r imprinted expression is developmentally regulated and established after embryonic implantation (Lerchner and Barlow, 1997; Szabo and Mann, 1995). This developmental regulation is reproduced in differentiating mouse embryonic stem cells (ESCs) (Latos et al., 2009), where Igf2r expression switches from biallelic to monoallelic after the onset of Airn expression (Fig. 1A). If Airn requires additional factors, their expression may be restricted to the same developmental window during which Airn establishes Igf2r silencing. Testing whether Airn-mediated silencing is limited to a developmental window is the first step towards identifying such factors. Another unresolved issue concerns the maintenance of imprinted silencing. Once its expression is turned on, Airn is transcribed continuously where Igf2r shows imprinted expression. However, it is unknown whether continuous expression is needed to maintain silencing. Among the three genes silenced by Airn, Igf2r is the only one to gain DNA methylation on the silenced promoter (Zwart et al., 2001). This somatic imprint, gained late in development, is not required for initiation (Li et al., 1993; Seidl et al., 2006) but could play a maintenance role.
Here, we investigate developmental control of Igf2r silencing by altering the timing of Airn expression, using inducible systems with general applicability to lncRNA genetic studies. We find that eliminating Airn transcription in differentiated ESCs reverses Igf2r silencing, unless the paternal Igf2r promoter is methylated. This shows that Airn is continuously required to maintain Igf2r silencing, but only in the absence of DNA methylation. This methylation mark is maintained independently of Airn, indicating no role for Airn in its propagation. Furthermore, Airn can initiate Igf2r silencing in early and late differentiated ESCs, although with decreasing efficiency, indicating a ‘window of opportunity’ does not limit its repressive effects. Finally, we show that Igf2r repression is maintained in the absence of DNA methylation. Together, our results indicate that Airn acts alone to silence Igf2r and that the somatic methylation imprint, although dispensable for silencing initiation and maintenance, may play a reinforcing role.
MATERIALS AND METHODS
Targeted ESC generation
The R26CreERT2 targeting vector was a gift from Austin Smith (CSCR, Cambridge, UK). CKO and CRes targeting vectors were built using a plasmid with a 7.3 kb 129Sv homology region (chr17:12,738,432-12,745,760, UCSC build GRCm38/mm10). In the CKO construct, a 1.9 kb PacI-NsiI region (chr17:12,740,792-12,742,677) was flanked by loxP sites. First, a loxP-flanked PGK-Neo-pA sequence was subcloned into the NsiI site and the resulting plasmid transformed into EL350 E. coli, expressing arabinose-inducible Cre recombinase (a gift from Alexander Stark, IMP, Vienna, Austria). Cre recombination was induced by 0.1% L-(+)-arabinose resulting in Neo excision and generation of a single loxP site at the NsiI position. The second loxP site, together with an FRT-flanked PGK-Neo-pA selection cassette, was subcloned from plasmid pK-II (a gift from Maria Sibilia, ICR, Vienna, Austria) into the PacI site. For the CRes construct, a 1.2 kb rabbit β-globin polyA cassette (Sleutels et al., 2002) and loxP site, plus the same FRT-Neo-FRT+loxP cassette used above, were subcloned into the BamHI site at chr17:12,744,359. Electroporation and neomycin selection were performed under standard conditions. S12/+ cells [a feeder-dependent D3 ESC line carrying a SNP in Igf2r exon12 (Latos et al., 2009)] were used to obtain R26CreER ESCs (S12RC/+), which were used to obtain CKO and CRes ESCs. The selection cassette was removed by electroporating the pMC-Cre plasmid in R26CreER cells or the pCAGGS-FLPe plasmid in CKO and CRes cells.
ESCs were grown on irradiated primary mouse embryo fibroblasts. Differentiation was induced by feeder-cell depletion, LIF withdrawal and 0.27 μM all-trans RA. Embryoid body formation was induced by ESC aggregation in AggreWell plates (Stemcell Technologies) for 8 hours and culture on ultra-low adherence flasks. The tetracycline-inducible promoter in APD-TET-Rolo cells was induced with 1 μg/ml doxycycline hyclate. Cre recombination in CKO and CRes cells was induced with 1 μM 4-hydroxytamoxifen, unless otherwise stated.
DNA and RNA analysis
Genomic DNA isolation and Southern blots used standard protocols and signal intensities were quantified with ImageQuant. qPCR and RNA FISH were as described previously (Latos et al., 2012). Table S1 in the supplementary material lists primers and probes.
Western blot analysis was performed as described previously (Gratz et al., 2011), using a 1:1000 dilution of the Covance rabbit anti-Cre antibody (a gift from Juergen Knoblich, IMBA, Vienna, Austria).
Bisulfite conversion, cloning and sequence analysis were as described previously (Koerner et al., 2012). PCR used primers in supplementary material Table S1 and conditions were 1 minute at 94°C, 30 seconds at 58°C and 1 minute at 72°C for 40 cycles.
Chromatin immunoprecipitation (ChIP) was carried out as described previously (Regha et al., 2007).
Two inducible systems to control the Airn lncRNA
We previously reported a tetracycline (Tet)-inducible Airn allele (Stricker et al., 2008). However, owing to gain of Tet-Airn DNA methylation, these cells were not suitable for further experiments (supplementary material Fig. S1). We developed an alternative genetic system to control Airn expression during ESC differentiation using a D3 ESC line named S12/+ (the maternal allele is written on the left side throughout the text), which carries an Igf2r exon 12 SNP to discriminate maternal and paternal expression, and reproduces the developmental onset of Igf2r imprinted expression during differentiation (Latos et al., 2009) (Fig. 1A). The CreERT2 gene was inserted into the ROSA26 locus (Zambrowicz et al., 1997) to ensure expression throughout ESC differentiation (Fig. 1B, top; supplementary material Fig. S2A-C). CreER expression was verified at mRNA and protein levels (supplementary material Fig. S2D,E), and the cells were designated S12RC/+. The expressed CreER product remains inactive in the cytoplasm until 4-hydroxytamoxifen (TAM) treatment (Feil et al., 1997). S12RC/+ that carry no additional modification in the Airn/Igf2r locus compared with parental S12/+ cells are referred to as wild type. Using S12RC/+ ESCs, the Airn locus was modified to generate Airn promoter conditional knockout (CKO) and Airn expression conditional rescue (CRes) cell lines (Fig. 1B).
Airn CKO ESCs
Airn CKO ESCs were generated by introducing loxP sites flanking 1.9 kb containing the Airn promoter and CGI (Fig. 2A). The 5′ boundary was a PacI site 580 bp upstream of the Airn TSS and 385bp from Igf2r exon 3. The 3′ boundary was an NsiI site 1.3 kb downstream of the Airn TSS. Two independent clones (S12RC/CKOFl+cas1,2; Fig. 2B) were targeted on the paternal allele that carries the unmethylated ICE and expresses the Airn lncRNA (Fig. 2C). A targeting vector containing the selection cassette in opposite orientation generated no homologously targeted clones (supplementary material Table S2). Selection cassette removal generated clones S12RC/CKOFl1,2 (Fig. 2D). CKOFl cells were TAM treated to delete the loxP-flanked Airn promoter, thus generating the CKOΔ allele (supplementary material Fig. S3A). CreER-mediated excision efficiency was tested in undifferentiated ESCs (supplementary material Fig. S3A). Independent of TAM dose, >80% of CKOFl alleles undergo recombination by 24 hours, with complete excision by 48 hours.
Conditional deletion of the Airn promoter
Imprinted Igf2r expression arises between days 2 and 3 of ESC differentiation (Fig. 1A). To test whether Airn expression is needed to maintain Igf2r silencing after it is initiated, CKO ESCs were differentiated using retinoic acid (RA), then the Airn promoter was deleted at day 5, 9 or 13 by TAM addition, and cells were harvested 4 days later (Fig. 3A). Airn has a half-life of less than 2 hours and transcripts are absent ∼10 hours after promoter deletion (Seidl et al., 2006). Cre-mediated excision of CKOFl was quantified by Southern blot (Fig. 3B, top; supplementary material Fig. S3B). In contrast to undifferentiated ESCs (supplementary material Fig. S3A), the Airn promoter showed 88% recombination at day 5 of differentiation, which was reduced to 58-72% by day 9 or day 13 (Fig. 3B, top). qPCR quantification shows 83% recombination at day 5 and 59-63% at day 9 or day 13 (Fig. 3C, left). To test whether Cre recombination improves in a different lineage, we performed the same experiment on CKOFl cells differentiated by embryoid body (EB) formation. As shown by Southern blot (Fig. 3B, bottom; supplementary material Fig. S3B) and qPCR quantification (Fig. 3C, right), the Airn promoter is deleted more efficiently in EB differentiated ESCs, with only 19-26% residual unrecombined alleles.
The effect of the conditional promoter deletion on Airn steady-state levels was assessed by RT-qPCR (Fig. 3D). As expected, Airn is upregulated in differentiated CKO cells carrying an intact promoter (Fig. 3D, bars 2-5), showing that loxP sites in the CKOFl allele do not interfere with promoter activity. However, Airn is not expressed if its promoter is deleted before differentiation (Fig. 3D, bar 6), confirming that the deletion removes all sequences required for Airn transcription. When the promoter is deleted during differentiation, Airn steady-state levels are reduced to ∼15% of controls in EB differentiated cells (Fig. 3D, right, bars 7-9). Higher residual levels of Airn, seen when the deletion is induced during late RA differentiation (Fig. 3D, left, bars 8-9), are explained by inefficient recombination of the CKOFl allele. The data show that promoter deletion during ESC differentiation can eliminate Airn expression.
Igf2r silencing requires continuous Airn expression until DNA methylation is acquired
To determine the effect of Airn removal after Igf2r silencing is initiated, we examined Igf2r imprinted expression in differentiated CKO cells. Allele-specific Igf2r expression was assayed non-quantitatively using the maternal-specific SNP in exon 12 that destroys a PstI site (Fig. 4A; supplementary material Fig. S4). PstI digestion of amplified cDNA from undifferentiated ESCs, which express Igf2r biallelically, yields an undigested maternal band and two paternal PstI-cut fragments (Fig. 4A, sample 1; supplementary material Fig. S4). Reduced paternal Igf2r fragments relative to the maternal fragment in differentiated cells that express Airn indicate maternal-specific Igf2r upregulation (Fig. 4A, samples 2-5; supplementary material Fig. S4). When the Airn promoter is deleted from CKO cells at day 0, Igf2r expression remains biallelic with visible paternal-specific bands throughout differentiation (Fig. 4A, sample 6; supplementary material Fig. S4), in agreement with previous Airn promoter deletion alleles that fail to establish Igf2r imprinted expression (Stricker et al., 2008; Wutz et al., 2001). To determine whether Airn is required to maintain Igf2r silencing, we turned Airn expression off at day 5, day 9 or day 13 of differentiation, after Igf2r silencing has occurred (seen in the untreated ‘no TAM’ day 5-17 controls). Four days after TAM treatment, re-expression of paternal Igf2r occurs at all tested times (Fig. 4A, samples 7-9; supplementary material Fig. S4), indicating that Igf2r silencing is not maintained in the absence of Airn.
We quantified Igf2r allele-specific expression by RT-qPCR using forward primers specific for the wild-type paternal or the SNP-modified maternal Igf2r allele and a common reverse primer (Koerner et al., 2012). Control differentiated cells that lack the Airn promoter and express Igf2r biallelically were used to set the maternal:paternal ratio to 50:50 (Fig. 4B, bar 6). Untreated (no TAM) control cells expressing wild-type levels of Airn show maternal-specific Igf2r expression, with low-level paternal expression (4-24% of total Igf2r levels; Fig. 4B, bars 2-5). Confirming results from Fig. 4A, the qPCR assay shows that paternal Igf2r silencing is relieved to different extents when Airn is turned off during differentiation. In RA differentiated cells, paternal Igf2r expression is 38% of total levels when the Airn promoter is deleted at day 5 (Fig. 4B, left, bar 7, blue bar), but is reduced to ∼30% when Airn is removed at day 9 or day 13 (Fig. 4B, left, bars 8 and 9, blue bars). Correcting for recombination efficiency in RA day 9/day 13 differentiated cells, to consider only the subpopulation of cells with no Airn promoter, shows that paternal Igf2r is re-expressed to ∼40% of total levels when the Airn promoter is deleted during late differentiation (Fig. 4B, left, black bars). Quantification of allele-specific Igf2r expression in EB differentiated cells in which the Airn promoter is deleted with higher efficiency shows that when Airn is removed at day 5 paternal Igf2r is re-expressed to ∼45% of total levels (Fig. 4B, right, bar 7). However, when Airn is turned off at day 9 or day 13, paternal Igf2r re-expression is 21-23% of total levels (Fig. 4B, right, bars 8 and 9). Together, the analysis in RA or EB differentiated cells shows that Airn is continuously required to maintain paternal Igf2r silencing, but additional factors influence silencing in late differentiated cells.
Igf2r silencing by Airn during embryonic development and ESC differentiation is marked by a late gain of DNA methylation on the paternal Igf2r promoter CGI (Latos et al., 2009; Stöger et al., 1993). This methylation mark, although not needed to silence Igf2r up to 8.5 dpc of embryonic development (Li et al., 1993), could play a later maintenance role. We tested Igf2r promoter methylation in differentiated CKO cells by Southern blot analysis of a methyl-sensitive NotI site diagnostic of the methylation status of the Igf2r CGI (Stöger et al., 1993) (Fig. 4C; supplementary material Fig. S5). In differentiated control cells lacking the Airn promoter, the paternal Igf2r promoter is expressed and lacks DNA methylation, as shown by the presence of the single NotI-digested 1 kb band (Fig. 4C, lane 5; supplementary material Fig. S5A). In control-differentiated cells that express Airn and establish Igf2r imprinted expression, the paternal Igf2r promoter is progressively methylated during differentiation, as shown by gain of a methylated, NotI-undigested 5 kb band (Fig. 4C, lanes 1-4; supplementary material Fig. S5A). Maximum methylation levels of ∼20% were seen in RA differentiation (Fig. 4C, left, lane 4; supplementary material Fig. S5A) and of ∼40% in EB differentiation (Fig. 4C, right, lane 4; supplementary material Fig. S5A). Notably, after Airn removal and re-expression of the paternal Igf2r promoter, the DNA methylation that was gained was maintained despite the absence of Airn (Fig. 4C right, compare lanes 6-8 with lanes 1-4; supplementary material Fig. S5A). This shows that DNA methylation on the paternal Igf2r promoter is maintained independently of the Airn lncRNA.
Airn CRes ESCs
To test whether Airn can silence Igf2r at any differentiation stage, we established CRes ESCs, in which the silencing function of Airn can be switched on during differentiation. We introduced a loxP-flanked polyA signal into S12RC/+ cells, at a BamHI site 3 kb after the Airn TSS (Fig. 5A), to create a conditional version of an Airn 3 kb truncation allele that cannot silence Igf2r (Sleutels et al., 2002). Paternal targeting of two independently targeted clones (S12RC/CResFl+cas1,2; Fig. 5B) was confirmed (Fig. 5C) and the selection cassette removed to generate clones S12RC/CResFl1,2 (Fig. 5D). Deletion of the loxP-flanked polyA signal in the CResFl allele generated the CResΔ allele (supplementary material Fig. S6A). Compared with CKOFl cells (supplementary material Fig. S3A), recombination is faster in undifferentiated CRes cells, which have loxP sites further downstream of the Airn promoter [over 80% recombination 12 hours after TAM treatment and complete excision by 24 hours (supplementary material Fig. S6A)].
Conditional deletion of the truncation signal rescues full-length Airn transcription
To test whether removing the polyA signal restores full-length Airn transcription to wild-type levels, RA differentiated CRes cells were induced to delete the polyA signal daily between day 1 and day 10 (Fig. 6A), and harvested after 3-4 days (Fig. 6A). CResΔ cells (TAM treated at day 0) were co-differentiated for 4-14 days as a control for wild-type Airn levels. Cre-mediated excision monitored by Southern blot showed the CResFl allele is recombined efficiently (over 85%) throughout RA differentiation (Fig. 6B; supplementary material Fig. S6B). Full-length Airn is not detected in differentiated cells carrying the unrecombined CResFl allele (Fig. 6C, bar 8), confirming the polyA signal truncates Airn. Airn is strongly upregulated during differentiation in control CResΔ cells, showing that truncation of Airn is reversible (Fig. 6C, bars 2-6). Importantly, when the polyA signal is removed during differentiation, full-length Airn expression is restored to levels comparable with wild-type controls (Fig. 6C). Overall, the data show that the CRes system efficiently rescues full-length Airn transcription during ESC differentiation, allowing a switch from a short, non-functional Airn to its longer, functional form at any time.
Airn expression can silence Igf2r at any time during ESC differentiation
To test whether Airn can silence Igf2r at any time or whether its effects are restricted to a developmental window, we examined Igf2r imprinted expression in CRes cells using the PstI assay (Fig. 7A; supplementary material Fig. S7A). In agreement with mouse studies (Sleutels et al., 2002), differentiated CResFl cells carrying the truncated Airn allele fail to establish Igf2r imprinted expression and display paternal-specific bands throughout differentiation (Fig. 7A, no TAM day 8, day 14; supplementary material Fig. S7A). By contrast, control CResΔ display wild-type gain of Igf2r imprinted expression during differentiation (Fig. 7A, left; supplementary material Fig. S7A). We next restored full-length Airn expression at 24-hour intervals, testing early (Fig. 7A, top right; supplementary material Fig. S7A) and late (Fig. 7A, bottom right; supplementary material Fig. S7A) differentiation time points. Compared with the truncated Airn control that does not silence Igf2r, we observed Igf2r repression at all time points (Fig. 7A, compare samples 9-13 and sample 8 in each row; supplementary material Fig. S7A). However, paternal-specific bands were more visible compared with wild-type controls, especially at late differentiation time points (Fig. 7A, compare samples 9-13 and samples 2-6 in each row; supplementary material Fig. S7A).
We quantified allele-specific Igf2r expression (Fig. 7B) setting to 1 the ratio between maternal and paternal expression in undifferentiated control cells that carry the Airn truncation and express Igf2r biallelically (Fig. 7B, day 0 control BAE Igf2r). During differentiation, these cells show no gain of Igf2r imprinted expression and the maternal/paternal Igf2r ratio remains ∼1 at day 8 and day 14. Control CResΔ express full-length Airn and gain wild-type levels of Igf2r imprinted expression during differentiation, with maternal:paternal ratios of 6-18 for early and late differentiation (Fig. 7B, control imprinted Igf2r). When Airn is turned on between days 1-10 of differentiation, we observe a gain of Igf2r imprinted expression at all time points, with maternal:paternal ratios between 4 and 11 (Fig. 7B, CRes experiment). This ratio is similar to control cells when the polyA signal is removed at day 1 or day 2 (Fig. 7B, left, compare CRes experiment and control imprinted Igf2r). When full-length Airn is restored after day 3, the maternal:paternal Igf2r ratio remains at ∼4-5 for all time points (Fig. 7B, compare CRes experiment and control imprinted Igf2r). Together, this shows that Airn silencing of Igf2r is not restricted to one developmental window but silencing is less efficient when functional Airn is expressed after day 3.
We next analysed DNA methylation of the Igf2r promoter CGI by Southern blot (Fig. 7C; supplementary material Fig. S7B,C). Undifferentiated ESCs or differentiated control cells that express truncated Airn and show biallelic Igf2r lack DNA methylation, as shown by the single 1 kb band (Fig. 7C, lanes 1, 7 and 8; supplementary material Fig. S7B). Differentiated control cells expressing full-length Airn gradually gain Igf2r promoter methylation on the repressed paternal allele, as shown by increased intensity of the methylated 5 kb band (Fig. 7C, lanes 2-6; supplementary material Fig. S7B). Unexpectedly, when Airn function is rescued during differentiation, we observed little or no DNA methylation on the Igf2r promoter (Fig. 7C, lanes 9-13; supplementary material Fig. S7B). Methylation levels comparable with wild-type controls are observed only when the polyA signal is removed at day 1 (Fig. 7C, top, compare lane 9 and lane 2; supplementary material Fig. S7B). When Airn length is functionally restored between days 2 and 4, low methylation is detected; rescuing at day 6 or later results in no detectable (nd) DNA methylation on the Igf2r promoter (Fig. 7C, bottom, compare lanes 9-13 and lanes 2-6; supplementary material Fig. S7B). Bisulfite sequencing of the Igf2r CGI supports these observations (Fig. 7D; supplementary material Fig. S8A,B). The inability of the repressed Igf2r allele to gain DNA methylation when Airn function is restored in late differentiation correlates with Dnmt3b and Dnmt3l downregulation (supplementary material Fig. S8C). However, low levels of repressive H3K9me3 modification are gained at the Igf2r promoter when Airn function is restored at day 10 (supplementary material Fig. S9). Together, the data show that Igf2r silencing by Airn during late differentiation is accompanied by low-level H3K9me3, but not DNA methylation, on the Igf2r promoter.
We describe here inducible ESC systems that control endogenous Airn lncRNA expression to investigate the developmental regulation of imprinted Igf2r silencing. Airn is a well-established example of a cis-repressing lncRNA that silences the paternal Igf2r allele, which becomes methylated in all embryonic, extra-embryonic and adult tissues where they are co-expressed (Sleutels et al., 2002; Yamasaki et al., 2005). Although Airn expression is also necessary to silence the paternal Slc22a2 and Slc22a3 alleles in extra-embryonic tissues, ESCs cannot yet be differentiated into these tissues and these genes show low-level non-imprinted expression in differentiated ESCs, typical of embryonic tissues (Hudson et al., 2010; Latos et al., 2009; Zwart et al., 2001). Using two inducible systems, we tested whether Airn expression is continuously needed to maintain Igf2r silencing and whether Airn silencing is restricted to a ‘window of opportunity’ during ESC differentiation. The data show that although Airn expression is necessary and sufficient to initiate and maintain Igf2r silencing at any stage during ESC differentiation, DNA methylation adds an extra layer of epigenetic information that may act to safeguard the silent state.
Inducible ESC systems to control endogenous gene expression
We have previously characterized an Igf2r imprinting model using the S12/+ ESC line, modified here, which faithfully recapitulates the developmental onset of Igf2r imprinted expression (Latos et al., 2009). ESCs are frequently used as models for X-chromosome inactivation (XCI) (Navarro and Avner, 2010) and are becoming more appreciated for genomic imprinting studies (Kohama et al., 2012). An ESC study of the Kcnq1 imprinted cluster demonstrated that Cdkn1c was silenced during RA differentiation without acquiring the DNA methylation seen in mouse embryos (Wood et al., 2010). However, we show that the Cdkn1c promoter acquires ∼20% methylation after EB differentiation (supplementary material Fig. S10A). Our results confirm the utility of ESC models for studying some aspects of epigenetic silencing of imprinted genes, but demonstrate that differentiation protocols need consideration.
We initially attempted to control endogenous Airn expression using a TetOn system (Stricker et al., 2008). However, the Tet-driven Airn promoter was modified by DNA methylation and the effects of inducing Airn expression could be assayed only in a subset of cells. Therefore, we switched strategies and created two inducible Cre-loxP systems, with general applicability for lncRNA genetic studies, to control Airn expression during ESC differentiation. The CKO system used loxP sites flanking the Airn promoter to delete it during ESC differentiation, whereas the CRes system used loxP sites flanking a polyA signal to functionally elongate Airn during ESC differentiation. Both genetically modified ESC lines differentiated normally, as shown by downregulation of pluripotency markers and upregulation of differentiation markers (supplementary material Fig. S10B). The effect of deleting or inducing functional Airn was tested 3-4 days after TAM treatment to allow time for chromatin state to change and existing Igf2r mRNA to decay. In the CKO system, where loxP sites span the expressed Airn promoter, we observed reduced recombination efficiency in RA compared with EB differentiation and therefore based conclusions on experiments with the latter. This difference may be related to promoter activity, as Airn was more highly expressed in RA than in EB differentiated cells (supplementary material Fig. S10C). Overall, the inducible Cre-loxP strategy proved a valid alternative to the Tet-inducible system.
Continuous Airn expression is necessary for Igf2r silencing
By deleting the Airn promoter during ESC differentiation, we show that continuous Airn expression is needed to maintain Igf2r silencing but only in the absence of DNA methylation at the Igf2r promoter. Removing Airn transcription at day 5 of ESC differentiation, when fewer than 10% of cells have gained Igf2r promoter methylation, results in almost complete loss of Igf2r silencing. A similar effect is observed when Airn is removed at later stages in RA differentiated cells, which gain only ∼20% Igf2r methylation. However, removing Airn in late-differentiated EBs, which gain ∼2 fold more Igf2r methylation, causes incomplete loss of silencing. Continuous Airn expression is therefore necessary for Igf2r silencing, but only until DNA methylation is established, determining a switch from Airn-dependent to Airn-independent Igf2r silencing during development. Importantly, the data also show that continuous Airn expression is not necessary for DNA methylation to be propagated, as removing Airn at any time point during ESC differentiation did not cause loss of the DNA methylation already established on the Igf2r promoter. This was not due to cell cycle arrest, as both RA and EB differentiated cells continued to proliferate throughout the observation period (supplementary material Fig. S10D). In a recent mouse study, maintenance of imprinted silencing at the Kcnq1 cluster was analysed by conditionally deleting the promoter for the Kcnq1ot1 macro lncRNA that controls this cluster (Mohammad et al., 2012). Similar to observations of Airn during ESC differentiation, continuous Kcnq1ot1 expression is necessary to maintain imprinted silencing of genes in embryos. However, in contrast to Airn, DNA methylation at the promoters of two silenced genes is lost in the absence of the Kcnq1ot1 lncRNA (Mohammad et al., 2012). The results here show that the Igf2r somatic imprint is maintained in a lncRNA-independent fashion, most likely through the hemi-methyltransferase activity of DNMT1 (Ooi et al., 2009).
Our results raise questions concerning the developmental regulation of Igf2r silencing by Airn transcription (Latos et al., 2012). First, if Airn is dispensable to maintain Igf2r silencing once DNA methylation is established, as our results in early development show, it is unclear why the lncRNA is continuously expressed. Similar to Airn, the Xist lncRNA responsible for XCI is also continuously expressed in mouse tissues, although XCI is maintained independently of Xist in both differentiated ESCs and somatic cells (Csankovszki et al., 1999; Wutz and Jaenisch, 2000). In general, somatic imprints modify the repressed alleles of very few imprinted protein-coding genes and for some of these, methylation is not conserved in humans (John and Lefebvre, 2011). Thus, the role of DNA methylation in maintaining imprinted gene silencing is unclear. In the mouse, many imprinted genes show imprinted expression for only a limited time and switch to biallelic expression during development (Santoro and Barlow, 2011). It is tempting to speculate that the absence of DNA methylation from most silent imprinted gene promoters is due to the need to re-express the silent allele during development. Conversely, DNA methylation could represent a means to ensure stable epigenetic repression of essential imprinted genes throughout life (John and Lefebvre, 2011).
Airn expression can silence Igf2r at any time during ESC differentiation
The Airn lncRNA is among the few lncRNAs for which a precise function has been described (Guttman and Rinn, 2012; Pauli et al., 2011). It has been recently shown that Airn transcription, but not the lncRNA transcript, is responsible for Igf2r silencing (Latos et al., 2012). One way to investigate lncRNA mechanism of action is to ask whether its activity is restricted to a permissive developmental context or time frame that contains essential co-factors or chromatin environments. For example, a ‘window of opportunity’ has been described for the Xist lncRNA, which can only initiate XCI within 48 hours of ESC differentiation (Wutz and Jaenisch, 2000). In adult mice, most cells are resistant to Xist but permissiveness for XCI is transiently re-established in hematopoietic precursor cells (Savarese et al., 2006). In contrast to Xist, Airn can initiate Igf2r silencing throughout ESC differentiation. Airn is normally upregulated between days 2 and 3 of ESC differentiation (Latos et al., 2009) and activating functional Airn after day 3 induces paternal Igf2r repression at all time points, showing that silencing activity is not restricted to a window and is unlikely to depend on developmentally regulated factors. Although Igf2r silencing is usually followed by gain of DNA methylation (Latos et al., 2009; Stöger et al., 1993), Igf2r repression after day 5 is not. This correlates with decreased levels of the de novo methyltransferase DNMT3B and of the DNMT3L co-factor during ESC differentiation. Importantly, Igf2r silencing can be maintained up to 8.5 dpc in the absence of DNA methylation, as shown by Dnmt1 knockout mice that silence Igf2r biallelically and upregulate Airn twofold (Li et al., 1993; Seidl et al., 2006). The data here show that DNA methylation, although able to maintain the silent state, is not necessary for its maintenance and can only be established within an early developmental window.
Although Airn-mediated silencing is observed throughout ESC differentiation, the data show that Igf2r repression after day 3 is less efficient than in the continuous presence of Airn. It is noteworthy that Airn and Igf2r show similar expression kinetics in mouse tissues and differentiated ESCs (Latos et al., 2009; Pauler et al., 2005). This could indicate that Airn repressor activity is limited by higher Igf2r promoter activity. Transcriptional interference, whereby one transcriptional process suppresses another one in cis (Palmer et al., 2011) has been shown to act at the Igf2r locus (Latos et al., 2012). The data presented here, that Airn represses Igf2r most efficiently when the latter is weakly expressed and that silencing efficiency decreases when the Igf2r promoter is expressed strongly, are in agreement with a transcriptional interference model.
Understanding the order of events that lead to stable silencing of imprinted protein-coding genes by macro lncRNAs is not only relevant for other imprinted clusters, but may be informative for the growing number of lncRNAs identified in the mammalian genome, particularly those associated with abnormal gene silencing in human disease (Wang and Chang, 2011). Human imprinting syndromes arising from aberrant expression of imprinted genes or loss of the parental allele expressing the protein-coding gene can benefit from therapeutic strategies that relieve the dormant alleles. One example is the Angelman syndrome, where topoisomerase inhibitors have recently been used to reactivate the silent Ube3a gene, which correlated with downregulation of the antisense Ube3a-as lncRNA (Huang et al., 2012). The data here, which show Airn expression is continuously required for Igf2r silencing until DNA methylation is acquired, underline the importance of understanding how epigenetic silencing is maintained, before strategies to reactivate epigenetically silenced genes can be designed, as removing only DNA methylation or only the lncRNA product would not relieve silencing from similar loci.
We thank Nina Gratz for help with western blots; Martin Leeb for the CreER lysate; Tomasz Kulinski for help with EB differentiation; Meinrad Busslinger, Michael Jantsch and the Barlow lab for discussions; and Giulio Superti-Furga and Quanah Hudson for reading the manuscript.
This project was supported by Austrian Science Fund [FWF F4302-B09 and W1207-B09] and by Genome Research in Austria [GEN-AU 820980].
Competing interests statement
The authors declare no competing financial interests.
Supplementary material available online at http://dev.biologists.org/lookup/suppl/doi:10.1242/dev.088849/-/DC1
- Accepted January 1, 2013.
- © 2013. Published by The Company of Biologists Ltd