Report

Airn Transcriptional Overlap, But Not Its lncRNA Products, Induces Imprinted Igf2r Silencing

See allHide authors and affiliations

Science  14 Dec 2012:
Vol. 338, Issue 6113, pp. 1469-1472
DOI: 10.1126/science.1228110

Abstract

Mammalian imprinted genes often cluster with long noncoding (lnc) RNAs. Three lncRNAs that induce parental-specific silencing show hallmarks indicating that their transcription is more important than their product. To test whether Airn transcription or product silences the Igf2r gene, we shortened the endogenous lncRNA to different lengths. The results excluded a role for spliced and unspliced Airn lncRNA products and for Airn nuclear size and location in silencing Igf2r. Instead, silencing only required Airn transcriptional overlap of the Igf2r promoter, which interferes with RNA polymerase II recruitment in the absence of repressive chromatin. Such a repressor function for lncRNA transcriptional overlap reveals a gene silencing mechanism that may be widespread in the mammalian genome, given the abundance of lncRNA transcripts.

Macro long noncoding (lnc) RNAs such as Airn (1), Kcnq1ot1 (2), or Nespas (3) that silence imprinted gene clusters offer important epigenetic models for the numerous lncRNAs mapped in the mammalian genome (46). In the Igf2r imprinted cluster, the paternally expressed Airn (antisense Igf2r RNA noncoding) macro lncRNA silences in cis the paternal alleles of Igf2r, Slc22a3, and Slc22a2 (1). Airn may use different silencing mechanisms, because Igf2r is silenced in all embryonic, extraembryonic, and adult tissues that express Airn, whereas Slc22a2 and Slc22a3 are only silenced in some extraembryonic lineages (7, 8). In support of this, Slc22a3 silencing in the placenta depends on the Airn lncRNA product recruiting EHMT2 histone methyltransferase, whereas Igf2r silencing does not (9). Igf2r silencing is also not dependent on Polycomb-group proteins or DNA methylation (10, 11). Thus, the mechanism by which Airn silences Igf2r, the only gene in this cluster with an essential embryonic function (12), remains unknown. Airn transcription overlaps the Igf2r promoter but not the Slc22a3 or Slc22a2 promoters (fig. S1A), indicating that silencing could depend on Airn transcriptional overlap independent of the Airn lncRNA product.

To test the role of Airn transcription versus product in Igf2r silencing, we used homologous recombination in embryonic stem (ES) cells to insert polyadenylation (polyA) cassettes on the paternal chromosome that truncate Airn to different lengths (figs. S1 to S5), (13). ES cell differentiation was used to recapitulate the developmental onset of Airn and Igf2r imprinted expression (14) (fig. S1B). PolyA cassettes inserted before (T3, T16) or after (T31, T51) the Igf2r promoter truncated the 118-kb Airn to 3, 16, 31, and 51 kb, respectively (Fig. 1). RNA tiling array hybridization (Fig. 1A) demonstrated Airn truncation and the absence of novel spliced variants in all alleles. Although Airn was lost downstream, normal levels of unspliced Airn were maintained upstream of each truncation site (Fig. 1, B to D, and fig. S3B). Wild-type Airn is mostly unspliced, but 5% of nascent transcripts are spliced to four variants (fig. S1A) that constitute ~30% of steady-state Airn because of their high stability (15). All four truncation alleles showed ~40% loss of total Airn; this reflects a loss of spliced products, as splice acceptors lie downstream of each truncation (Fig. 1D). Together, the truncations of Airn at 3, 16, 31, and 51 kb removed 97.5, 86.5, 73.8, and 56.8% of the 118-kb Airn product, respectively, including all spliced variants. Furthermore, the truncations did not change Airn expression kinetics or its characteristic inefficient splicing, nor did they interfere with the methylation-free state of its paternal promoter (Fig. 1D and fig. S3C).

Fig. 1

Airn is shortened to different lengths by a targeted polyA cassette. (A) Tiling arrays confirm Airn loss upstream of Igf2r in truncated alleles and absence of novel spliced products. Data are relative hybridization intensity plots (13); error bars are means ± SD of 21 windows of the averaged signal from nine tiled oligos. (B) Loss of Airn 53.2 kb (Airn-mid) and 98.7 kb (Airn-end) upon polyA insertion 3, 16, 31, and 51 kb downstream from the Airn TSS in day 14 differentiated (d14) cells. Airn is not expressed in undifferentiated (d0) ES cells. (C) RNase protection shows normal Airn levels 7.4 kb from the TSS in wild type (WT), T16, T31, and T51 and its loss in T3. Actin was used as a loading control. y–, probe lacking RNase; y+, probe plus RNase; *T3, MEFs with a previously generated 3-kb truncation (1). (D) Reduced spliced + unspliced Airn (Airn-total), but normal levels of unspliced Airn (Airn-uns), upstream of the inserted polyA cassette in four truncated alleles. In (B) and (D), data are means ± SD of three technical replicates.

We next tested whether Airn truncation alleles silence the paternal Igf2r promoter. Igf2r allelic expression was analyzed using two polymerase chain reaction (PCR) assays for steady-state expression. Undifferentiated ES cells showed the expected biallelic Igf2r expression in the absence of Airn expression (14) (Fig. 2A and fig. S6A). Upon differentiation, T3 and T16 cells maintained biallelic Igf2r expression, whereas T31 and T51 cells showed a gain of Igf2r imprinted expression similar to wild-type cells. This resulted in a factor of ~2 Igf2r increase in T3 and T16 cells relative to wild-type, T31, and T51 cells (Fig. 2B). The T3 truncation has been examined in a mouse model (1), which validates the ES cell model used here. The data also show that the T3 and T16 truncations do not interfere with Igf2r expression, as the derepressed paternal and wild-type maternal alleles expressed similar Igf2r levels. Similarly, the T31 truncation cassette inserted on the maternal chromosome allows wild-type Igf2r expression (fig. S3A). Lastly, RNA fluorescence in situ hybridization (FISH) demonstrated loss of Igf2r imprinted expression at a transcriptional level in T3 and T16 but not in T51 cells (Fig. 2C and fig. S6C). Repression of the paternal Igf2r allele is accompanied by gain of DNA methylation on its promoter CpG island (CGI) (14). This methylation is absent in T3 and T16 but is present in T31 and T51 differentiated cells (Fig. 2D and fig. S6B). The T31 allele represses Igf2r and truncates before the first Airn splice acceptor at 37 kb, showing that all spliced Airn variants are unnecessary for Igf2r silencing. The T3 and T16 truncations show that the first 16 kb of Airn are insufficient to silence Igf2r, and the T31 and T51 truncations show that the last 87 kb of Airn are unnecessary. Together, they localize Airn repressor activity to the remaining 12.7% between the T16 and T31 truncations (fig. S6D). Because this region contains the Igf2r promoter, the data support the hypothesis that repressor activity results from Airn transcription.

Fig. 2

The greater part of the Airn lncRNA product is not required for repression of Igf2r. (A) Allele-specific reverse transcription quantitative PCR (RTqPCR) showing Igf2r imprinted expression in WT, T31, and T51, but not in T3 or T16, indicated by increased maternal/paternal Igf2r ratio in differentiated/d14 ES cells compared to a ratio of ~1 in undifferentiated (d0) cells. Bars show means ± SD of one to three biological replicates, each with three technical replicates (13). (B) RTqPCR showing total steady-state Igf2r in T3, T16, T31, and T51 relative to WT cells; means ± SD as in (A). (C) Igf2r RNA FISH in differentiated (d5) cells shows loss of imprinted Igf2r expression, indicated by increased numbers of cells with double signals in T3 and T16 compared to T51 or WT (fig. S6C shows representative images). Discontinuous transcription of active genes (20) results in many nuclei with no signal using intronic probes. Single spots, imprinted or stochastic biallelic expression; double spots, biallelic expression; n, nuclei counted; bars, total counts of two biological replicates and two technical replicates (13). (D) The paternal Igf2r promoter is methylated in differentiated WT, T31, and T51 but not in T3 or T16 cells or in undifferentiated (d0) ES cells, as shown by the 5-kb EcoRI fragment resistant to NotI digestion. *T3, E14 ES cells with a 3-kb Airn truncation (1); NIH3T3, MEFs with both parental alleles; Thp/+, unipaternal MEFs. T31 cells were assayed on a separate gel.

The nuclear size of the Airn lncRNA product correlates with silencing Slc22a3 in the placenta (9). We used RNA FISH to test whether Airn nuclear size or its subnuclear localization correlates with Igf2r silencing in embryonic cells. Both parameters showed no difference between T51, which silences Igf2r, and T16, which does not (Fig. 3A and fig. S7). The larger size of wild-type Airn is therefore unrelated to Igf2r silencing. The majority of FISH signals lay in the mid-nuclear plane, with both repressing and nonrepressing Airn alleles showing a similar localization (Fig. 3B) and a similar relative position to the nucleolus (fig. S7). Together, these data indicate no organizational role for the Airn product in Igf2r silencing, thereby supporting claims (9) that Airn silences Igf2r and Slc22a3 by different mechanisms.

Fig. 3

Airn macro lncRNA size and location do not determine Igf2r repression. (A) The size of the Airn RNA FISH signal (white arrow) relative to the nucleus [ring identified by 4′,6-diamidino-2-phenylindole (DAPI)] is similar for T51 and T16 but different from WT. Horizontal line denotes median; P values are results of t tests using two biological replicates performed in two technical replicates (13); n, number of nuclei. (B) Similar subnuclear localization for WT, T51, and T16 alleles. The nuclear area was binned into pseudo-colored inner, middle, and outer circles with equal spacing, and percentages of Airn RNA FISH signals (red arrow) in different distance bins were scored using the same data set as in (A).

A prediction of a transcriptional overlap model is that the interfering promoter should impose repressor activity. To test this, we moved the Airn promoter ~700 base pairs before the Igf2r transcription start site (TSS) in ES cells that lack an endogenous paternal Airn promoter (16) (figs. S8 and S9). FAP (forward Airn promoter) cells contain the repositioned Airn promoter and the first 1.8 kb of the Airn lncRNA product (also present in the T3 and T16 alleles that do not silence Igf2r) in wild-type orientation, and express normal levels of Airn that overlap the paternal Igf2r promoter (Fig. 4A and fig. S9B). RAP (reverse Airn promoter) cells contain an inverted repositioned Airn promoter and do not transcribe Airn over the paternal Igf2r promoter. Undifferentiated ES cells showed biallelic Igf2r expression in FAP or RAP cells (Fig. 4B and fig. S10A) similar to that seen in wild-type cells (Fig. 2A). Upon differentiation, RAP cells maintained biallelic Igf2r expression but FAP cells showed paternal-specific Igf2r silencing with a maternal/paternal Igf2r ratio similar to that in wild-type and T31 and T51 truncated cells. The FAP allele shows that the Airn promoter imposes repressor activity and also excludes the 11-kb Airn region spanning the T16 to FAP insertion sites. Together with the truncation alleles, this excludes 96.7% of the Airn lncRNA product as necessary for silencing Igf2r. The FAP and T31 alleles that both silence Igf2r have in common a 4-kb Airn product that overlaps the Igf2r promoter (Fig. 4A). Another prediction of the transcriptional overlap model is that repressor activity is maintained if the Igf2r promoter is substituted. We previously replaced this 4-kb region in vivo with a Tk-neo reporter gene that preserves imprinted expression and methylation (17). To exclude the possibility that the Tk-neo reporter fortuitously reconstituted endogenous elements, we demonstrated that it lacks any nucleotide or structural similarity to this region (fig. S11). Thus, this 4-kb endogenous Airn product is unnecessary to silence the Igf2r promoter.

Fig. 4

Airn transcriptional overlap is sufficient for Igf2r repression. (A) Left: Map showing parts of the Airn product excluded by T3, T16, T31, and T51 truncation alleles and the region tested by the FAP allele. Right: Wild-type levels of Airn are expressed from the FAP but not the RAP allele (means ± SD of three technical replicates). (B) FAP but not RAP cells maintain imprinted Igf2r expression. cDNA sequence single-nucleotide polymorphism (SNP) quantitation (fig. S10C) shows increased maternal/paternal ratio in FAP but not RAP cells (means ± SD of four biological replicates). (C) The repressed FAP Igf2r promoter has reduced levels of capped mRNA. cDNA sequence SNP quantitation shows the ratio of maternal/paternal capped Igf2r mRNA in day 5 differentiated (d5) FAP and RAP cells (mean of two sequence reads per bar; *P = 0.0001, t test). (D) Chromatin accessibility in DNase I–treated FAP1 and RAP1 nuclei; blot was hybridized with probe-NE4 to identify the Igf2r promoter (image levels are nonlinearly adjusted to improve visualization). The repressed FAP Igf2r promoter (6-kb FAP-DNase I/BglII fragment) has open chromatin similar to the active RAP promoter (5-kb RAP-DNase I/BglII fragment) (fig. S12A). (E) DNA methylation in genomic DNA digested with BglII and methyl-sensitive NotI, hybridized with probe-NE4. Absence of a 12-kb BglII fragment in FAP cells indicates an unmethylated (UMe) silent paternal Igf2r promoter (fig. S6B). In control R2Δ/+ cells, the 9.5-kb fragment indicates normal methylation (Me) of the silent paternal Igf2r promoter in differentiated ES cells. (F) H3K9me3 chromatin immunoprecipitation qPCR on Airn and Igf2r promoters (P) and Igf2r intron1 (B) in FAP (silenced paternal Igf2r promoter) and RAP (active paternal Igf2r promoter) cells that also contain a silenced maternal Airn promoter. Data are means ± SD of three technical replicates.

These data are consistent with Airn silencing the Igf2r promoter by transcriptional interference, which reduces recruitment of functional RNA polymerase II (RNAPII) to the Igf2r promoter, independent of Airn lncRNA products. In FAP cells, the proximity of the repositioned Airn promoter prevents a direct analysis of RNAPII on the repressed Igf2r promoter. To circumvent this, we assayed for S5P-RNAPII–dependent capped Igf2r mRNA and found that it was reduced in FAP but not in RAP cells (Fig. 4C and fig. S10B). Transcriptional interference models (18) predict suppression of the “sensitive” promoter by an “interfering” promoter, initially in the absence of repressive chromatin. The repressed FAP Igf2r promoter maintained features associated with active chromatin, such as a strong DNase I–hypersensitive site (Fig. 4D and fig. S12, A and B) and H3K4me3 (fig. S12C), similar to the active RAP Igf2r allele. The wild-type paternal Igf2r promoter is modified late in development by DNA methylation that is unnecessary for Igf2r repression in embryo or placenta (10) and by H3K9me3 (15, 19), which is unnecessary for Igf2r silencing in the placenta (9). The repressed FAP Igf2r promoter remained free of DNA methylation (Fig. 4E), possibly due to the proximity of the repositioned Airn promoter CGI. Low-level H3K9me3 was less than on the silent Airn promoter by a factor of 10, similar to ratios in mouse embryonic fibroblasts (MEFs) (19), (Fig. 4F). Together, these data show that Airn transcriptional overlap interferes with functional RNAPII recruitment to the Igf2r promoter in the presence of active chromatin, supporting a model whereby Airn induces silencing by transcriptional interference (fig. S12D).

Collectively, our data demonstrate a role for Airn transcription, but not its spliced or unspliced lncRNA products, in silencing the Igf2r promoter. The demonstration that Igf2r silencing depends on Airn transcription reflects hallmark features of macro lncRNAs, such as inefficient splicing, extreme length, high repeat content, lack of conservation, and short half-life (15), which all indicate that transcription is more important than product. It is not yet known how many of the growing number of mammalian lncRNAs share these hallmarks. If they do, the range of lncRNA functions in the mammalian genome could be substantially enlarged.

Supplementary Materials

www.sciencemag.org/cgi/content/full/338/6113/1469/DC1

Materials and Methods

Figs. S1 to S12

References (21, 22)

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: We thank E. F. Wagner and A. Wutz for the R2Δ and D3 ES cell lines, and G. Superti-Furga, C. Bock, and A. P. Bird for comments on the manuscript. Array data (GSE41444) are available at www.ncbi.nlm.nih.gov/geo. Supported by grants from Austrian Science Fund FWF (F4302-B09 and W1207-B09) and Genome Research in Austria (GEN-AU 820980).
View Abstract

Navigate This Article