The Air Noncoding RNA Epigenetically Silences Transcription by Targeting G9a to Chromatin

See allHide authors and affiliations

Science  12 Dec 2008:
Vol. 322, Issue 5908, pp. 1717-1720
DOI: 10.1126/science.1163802


A number of large noncoding RNAs (ncRNAs) epigenetically silence genes through unknown mechanisms. The Air ncRNA is imprinted—monoallelically expressed from the paternal allele. Air is required for allele-specific silencing of the cis-linked Slc22a3, Slc22a2, and Igf2r genes in mouse placenta. We show that Air interacts with the Slc22a3 promoter chromatin and the H3K9 histone methyltransferase G9a in placenta. Air accumulates at the Slc22a3 promoter in correlation with localized H3K9 methylation and transcriptional repression. Genetic ablation of G9a results in nonimprinted, biallelic transcription of Slc22a3. Truncated Air fails to accumulate at the Slc22a3 promoter, which results in reduced G9a recruitment and biallelic transcription. Our results suggest that Air, and potentially other large ncRNAs, target repressive histone-modifying activities through molecular interaction with specific chromatin domains to epigenetically silence transcription.

Large ncRNAs such as Xist, Air, and Kcnq1ot1 are required for epigenetic silencing of multiple genes in cis (1), whereas the recently described HOTAIR ncRNA silences genes in trans (2). For Xist, evidence suggests that specific interactions between the ncRNA and chromatin are essential for gene silencing; however, molecular details of such interactions are lacking. The Kcnq1ot1 and Air ncRNAs are involved in silencing clusters of multiple imprinted genes in cis on mouse chromosomes 7 and 17, respectively. Deletion of their promoters or truncation of the ncRNA results in biallelic, nonimprinted expression of genes in their respective clusters (36). These results suggest that the ncRNA molecules are involved in allele-specific silencing; however, a role for the process of transcription itself has not been ruled out (7).

Air expression is imprinted, driven by an antisense promoter located in intron 2 of Igf2r, regulated by DNA methylation specific to the parent of origin (8). In cells of the embryo proper, Air expression results in silencing of the paternal Igf2r gene in cis. However, in placenta, two additional distant genes, Slc22a2 and Slc22a3, are silenced in cis by Air (5). We confirmed parent-of-origin–specific placental expression of Air, Igf2r, Slc22a2, and Slc22a3 by reverse transcription polymerase chain reaction (RT-PCR) in reciprocally derived heterozygous Thp mice (9, 10) in which a large, naturally occurring deletion removes the entire gene cluster. Air is expressed predominantly from the paternal allele, whereas expression of Igf2r, Slc22a2, and Slc22a3 is predominantly maternal (fig. S1). Slc22a3 reverts to biallelic expression by embryonic day 15.5 (E15.5), consistent with previous analyses in mouse (11) and human (12). We verified that this expression pattern is the result of differential transcriptional regulation by primary transcript RT-PCR in Thp placentas (fig. S1) and by RNA fluorescence in situ hybridization (FISH) on wild-type placentas with Air and Slc22a3 probes (Fig. 1A). At E11.5, Slc22a3 is transcribed primarily on the maternal allele (Fig. 1A, left), in trans to the paternal Air signal, as expected. At E15.5, Slc22a3 transcription occurs both cis and trans to Air (paternal and maternal alleles, respectively) (Fig. 1A, right, and 1B). These results demonstrate that Slc22a3 escapes imprinted silencing of transcription later in gestation.

Fig. 1.

Air ncRNA envelopes the paternal Slc22a3 locus in correlation with allelic silencing. (A) RNA FISH on mouse placental tissue sections with probes for Air (red) and Slc22a3 intron 1 (green) [4′,6′-diamidino-2-phenylindole (DAPI), blue]. Representative cell nuclei with Slc22a3 signals in trans to Air signals (left, maternal Slc22a3 transcription) and biallelic Slc22a3 signals (right, maternal and paternal Slc22a3 transcription). Scale bar, 5 μm. (B) The ratio of Slc22a3 RNA FISH signals in trans (pink, maternal) versus those in cis (blue, paternal) to the Air signal in Air-positive cells from E11.5 and E15.5 placentas. (C) Representative high-magnification RNA-DNA FISH signals on mouse placental sections. (Top) Merged images of Slc22a3 DNA FISH signals (red) and Air RNA FISH signals (green). (Middle and bottom) Separate Slc22a3 DNA and Air RNA images, respectively. Scale bar, 1 μm. (D) Distribution of the Air RNA cover index on the Slc22a3 locus in E11.5 and E15.5 placentas. (E) Distribution of the Air RNA cover index on Slc22a3 RNA FISH signals in cis with Air in E15.5 placentas. Here, the cover index quantifies the extent of overlap between Slc22a3 RNA FISH signals and Air RNA FISH signals.

Air is largely unspliced and is retained in the nucleus (13). Air RNA FISH signals are considerably larger than primary transcript signals for protein-coding genes (Fig. 1A and fig. S2). We considered the possibility that Air functions to silence genes in cis in a manner, analogous to Xist “coating” of the inactive X chromosome. To test this, we performed RNA-DNA FISH to detect Air RNA and Slc22a3 locus DNA in placentas from E11.5 and E15.5 (Fig. 1C). Although the Air and Slc22a3 transcription units are >230 kb apart and oriented in different directions, we found that the Air RNA “cloud” frequently enveloped paternal Slc22a3. However, the extent to which the Air RNA overlapped Slc22a3 varied markedly between E11.5 and E15.5. To quantify the extent to which Air “covered” the Slc22a3 locus, we captured z-stacked confocal images of the signal pair, and determined the “cover index” [see (10) for definition of cover index].

At E11.5, most paternal Slc22a3 alleles are enveloped by Air and have a high Air cover index (80% > 0.4) (Fig. 1D and table S1). This suggests the possibility that Air coats Slc22a3 chromatin in placenta in conjunction with imprinted silencing, reminiscent of descriptions of Xist-mediated X inactivation (14). At E15.5, when paternal Slc22a3 escapes Air-directed silencing, we found a significant reduction (P = 0.016; Fisher's exact test) in Slc22a3 loci enveloped by Air (Fig. 1D and table S1; 47.5% cover index < 0.4). This cannot be accounted for by reduced expression of Air or smaller Air clouds at E15.5 compared with E11.5 because Air expression is considerably higher at E15.5 (fig. S1) and because Air clouds are generally larger (see below). These results show a correlation between high Air cover index on the Slc22a3 locus and gene silencing. To confirm that dissociation from the Air cloud correlates with gene activation, we measured the Air cover index on the transcriptionally active paternal Slc22a3 alleles using RNA FISH on E15.5 placental sections. The results show that transcribed Slc22a3 alleles have an extremely low cover index (90% < 0.4) (Fig. 1E and table S1), which demonstrates that the active Slc22a3 alleles are those most dissociated from Air.

To determine whether there is a molecular interaction between Air and the imprinted cluster in placenta, we employed RNA TRAP (tagging and recovery of associated proteins). RNA TRAP is derived from RNA FISH and uses the molecular targeting power of in situ hybridization to covalently tag chromatin in the immediate vicinity of a specific RNA in the nucleus (15). As a control, we also performed Air RNA TRAP on adult heart tissue, in which Air is highly expressed (16). Air silences paternal Igf2r in adult heart, but expression of Slc22a3 is biallelic (not imprinted), and Slc22a2 is not expressed (13). We noted that Air RNA FISH signals in adult heart were smaller than Air signals in wild-type placenta (Fig. 2A), which suggested that nuclear accumulation of Air in heart is less abundant than in placenta. Mice expressing a truncated Air (Air-T) (5) show greatly reduced accumulation in placental nuclei.

Fig. 2.

Air accumulates at the Slc22a3 promoter. (A) Detection of biotin deposition in RNA TRAP experiments with identical conditions. Air signal in Air-T placenta (arrowhead). DAPI, blue. Scale bar, 5 μm. (B) Air RNA TRAP results on designated tissues. Mean fold enrichment ± SEM in pull down to input [see (10) for details] normalized to 2.5 Mb upstream, from three biological replicates.

The RNA TRAP results (Fig. 2B) show that enrichment across the imprinted domain varies considerably among the three tissues (see tables S2 and S3). The highest and most significant level of enrichment was observed over the Slc22a3 promoter in E11.5 placenta when paternal Slc22a3 is silenced (Fig. 2C, red curve). Air enrichment at the Slc22a3 promoter was significantly reduced at E15.5 [P < 0.001; two-way analysis of variance (ANOVA) followed by contrast test] when Slc22a3 escapes imprinting and is biallelically transcribed. A lower, but still significant, level of enrichment was also seen over the body of the Slc22a3 gene in E11.5 placenta (P < 0.001 versus adult heart; P = 0.024 versus 2.3 Mb downstream). These results are wholly consistent with the differential Air cover index. Rather than uniformly coating the entire imprinted domain, Air accumulates specifically at the Slc22a3 promoter at E11.5, which suggests an interaction between the ncRNA (which probably exists as a ribonucleoprotein complex) and Slc22a3 promoter chromatin.

We next assessed allele-specific histone modifications throughout the imprinted cluster to test the possibility that accumulation of Air on the paternal Slc22a3 promoter is associated with a specific chromatin signature. We used E11.5 and E15.5 placentas from mice with maternally or paternally derived Thp deletion to assess each chromosome separately. We analyzed histone H3 lysine 4 di- and trimethylation (H3K4me2 and me3) as typical modifications at active promoters, and histone H3 lysine 9 trimethylation (H3K9me3) and histone H3 lysine 27 trimethylation (H3K27me3) as typical modifications at repressed promoters (17). We found a higher peak at E11.5 of H3K9me3 enrichment at the paternal Slc22a3 promoter, three times that at the maternal promoter (Fig. 3), which correlated strongly with the presence of Air (see supporting text for results of other modifications and loci). Paternal H3K9me3 at the Slc22a3 promoter is markedly reduced at E15.5, which correlates with the disappearance of Air from the promoter and biallelic transcription. The correlation between the high level of H3K9me3 enrichment at the paternal Slc22a3 promoter at E11.5 and the accumulation of Air suggests that the two may be functionally linked.

Fig. 3.

H3K9 methylation at the Slc22a3 promoter correlates with Air accumulation. Allele-specific H3K9me3 analyzed by native ChIP using +/Thp placentas (for maternal allele) and Thp/+ placentas (for paternal allele) at E11.5 and E15.5. Mean fold enrichment ± SD in pull down to input normalized to 2.5 Mb upstream. A second biological replicate showed similar results.

One possibility is that Air silences paternal Slc22a3 transcription by recruiting histone-modifying activities to the Slc22a3 promoter to methylate chromatin on H3K9. The G9a histone methyltransferase is involved in H3K9 di- and trimethylation and allelic silencing of some genes in the Kcnq1-imprinted domain in placenta (18). We investigated a potential functional link between G9a and Air in placenta by RNA immunoprecipitation (RNA IP) (19). RNA IP with antibodies raised against two distinct G9a epitopes bring down Air RNA (Fig. 4A and fig. S6). Furthermore, allele-specific chromatin immunoprecipitation (ChIP) with these G9a antibodies shows consistent substantial enrichment only at the paternal Slc22a3 promoter (Fig. 4B and fig. S7). To examine the possibility that G9a is targeted to the Slc22a3 promoter through recruitment by Air, we assessed G9a association with Slc22a3 promoter chromatin by ChIP at E11.5 in wild-type and +/Air-T placentas. Air-T mice express truncated Air, which results in biallelic expression of Slc22a3, Slc22a2, and Igf2r in placenta (5). Air-T RNA FISH signals are weak (Fig. 2A), and Air RNATRAP in E11.5 Air-T placenta does not bring down any detectable DNA from the cluster, which suggests that accumulation of truncated Air at the Slc22a3 promoter is greatly reduced. The ChIP results (Fig. 4C) show a significant reduction in G9a enrichment over the Slc22a3 promoter in +/Air-T placentas compared with wild type. Comparison of this finding with the parental allele-specific data from Fig. 4B obtained with the same antibody shows the expected trend in reduction of G9a occupancy at the Slc22a3 promoter from the respective genotypes. These data show that full-length Air is necessary for normal G9a targeting to the paternal Slc22a3 promoter.

Fig. 4.

Air recruitment of G9a to the Slc22a3 promoter is required for gene silencing. (A) RNA IP with and without G9a-specific antibody or immunoglobulin G (IgG) in E11.5 placenta. Air RNA detected by RT-PCR. Arbitrary units on the y axis, mean ± SEM derived from several measurements. (B) Allele-specific ChIP with G9a from Thp placentas at E11.5. Mean fold enrichment ± SEM of G9a pull down over IgG pull down, normalized to the pull down at 2.5 Mb upstream is shown. (C) G9a ChIP with wild-type and +/Air-T E11.5 placentas as described in (B), *P < 0.05; **P < 0.01; one-way ANOVA and contrast test. (D) Slc22a3 RNA FISH signals in trans to Air signal (red bars, maternal), biallelic (pink/blue tapering segments, maternal and paternal), or cis to Air (blue bars, paternal) in three E9.5 wild-type (WT) and three G9a/ gene knockout (KO) placentas. The ratio of paternal Slc22a3 signal is shown on the right. (E) Igf2r RNA FISH signals analyzed as in (D).

We assessed allelic transcription in wild-type and G9a–/– placentas by RNA FISH at E9.5 [one day before death of G9a–/– embryos (18)]. We observed significant shifts toward 50% paternal to maternal Slc22a3 transcription in G9a–/– placentas compared with wild type (P = 0.012; t test), which showed that G9a is required for silencing paternal Slc22a3 (Fig. 4D). Monoallelic transcription of Igf2r was maintained in G9a–/– placentas (P = 0.679; t test), which demonstrated that G9a is not required for imprinted silencing of paternal Igf2r (Fig. 4D). We excluded the possibility that these results were caused by a developmental delay by showing that Slc22a3 and Igf2r are imprinted in wild-type placenta a day earlier at E8.5 (fig. S8). Slc22a2 is not expressed on E8.5 or E9.5 (fig. S8), which precludes its analysis in G9a–/– placenta. These results are consistent with our RNA TRAP and ChIP data showing significant enrichment of Air and G9a at the Slc22a3 promoter, but not at the Igf2r promoter (see Figs. 2B and 4B). These data show that Slc22a3 and Igf2r are silenced by different mechanisms in the placenta.

We conclude that Air silences the imprinted Slc22a3 and Igf2r genes by different mechanisms in placenta. We propose that Air silences transcription of the distal paternal Slc22a3 gene via a specific interaction between the ncRNA and chromatin at the Slc22a3 promoter. This interaction may be mediated by factor(s) associated with the ncRNA and/or the Slc22a3 promoter. Accumulated Air at the promoter recruits G9a and leads to targeted H3K9 methylation and allelic silencing. Although paternal Igf2r silencing is also controlled by Air, our RNA TRAP data show that this is not through a ncRNA interaction with the Igf2r promoter, and analysis of G9a–/– placentas shows that it does not require G9a. An alternative, although not mutually exclusive, explanation for our data is that Air creates a repressive nuclear compartment, similar to that described for Xist (20) and that the observed interaction between Slc22a3 and Air is the result of the recruitment of Slc22a3 promoter (and G9a) to the Air compartment. In summary, our results show that Air (and potentially other large ncRNAs such as Xist) can function through specific interaction with chromatin and mediates targeted recruitment of repressive histone-modifying activities to epigenetically silence transcription.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S8

Tables S1 to S5


References and Notes

View Abstract

Navigate This Article