Antisense non-coding RNAs and regulation of gene transcription

ABSTRACT

Transcriptome analyses have revealed the existence of a large variety of non-coding RNAs (ncRNAs) that, although specifically expressed, are largely unknown in function. The best-studied role of ncRNAs is the regulation of gene expression, mostly participating in transcription. We report here the role of an antisense ncRNA that represses transcription of LEF1 by recruiting Polycomb repressive complex 2 (PRC2), which trimethylates H3K27. This antisense LEF1 transcript undergoes splicing losing its inhibitory function. We also discuss the possible general relevance of this finding.

The development of technologies for extensive RNA sequencing has enabled the analysis of the global transcriptomes of specific cells and tissues. These analyses have shown that 70–90% of the genome is actually transcribed, yielding RNAs lacking protein-coding capacity.^1,2 With the exception of miRNA transcripts, most of these non-coding RNAs have not been characterized and do not have any known function. During recent years several long non-coding RNAs (lncRNAs) have been described as crucial regulators of gene expression, mainly acting at transcription, although some lncRNAs also control RNA stability, splicing or translation.³ Therefore, lncRNAs can increase stability of the target RNA by direct interaction and acting as a miRNA sponge⁴; favor translation by recruiting the sense RNA to the polysomes⁵ and affect RNA splicing, either by preventing the association of the spliceosome⁶ or by creating a chromatin environment that prevents the interaction of a chromatin-splicing adaptor.⁷ Moreover, other lncRNAs, such as ncNRA-a7, work as specific enhancers, through their interaction with Mediator complex; these RNAs are required for enhancer looping and function.⁸ It remains to be established how specific this binding to Mediator is and which subunits of this large protein complex provide specificity for the interaction with the enhancer-specific lncRNA.

However, most lncRNAs act on their target genes by inhibiting their transcription. This was initially described for the first lncRNAs to be characterized, such as Xist/Rep1 or HOTAIR, which are capable of repression in cis or trans by interacting with the Polycomb repressive complex 2 (PRC2),^9,10 a complex that induces trimethylation of histone H3 lysine27 (H3K27me3).¹¹ Although PRC2 binding to different promoters has been extensively described, the signals guiding its recruitment are still controversial. While PRC2 binds to its own product (H3K27me)¹² and to the H2AK119ub mark deposited by PRC1,^13,14 it has gradually become accepted that binding of PRC2 to target promoters depends on RNA, either HOTAIR or another specific for the target promoter. The few cases where DNA binding is triggered by transcriptional repressors might depend on the association to an intermediate RNA. For instance, we have described that PRC2 is required for the repression of CDH1 expression by Snail1 and is recruited to the CDH1 promoter by Snail1.¹⁵ Since Snail1 binds to LSD1¹⁶ and this protein interacts with HOTAIR,¹⁷ it is possible that this lncRNA is required for PRC2 association to this and perhaps other Snail1-repressed promoters.

The specificity and extent of the binding of PRC2 to RNA is still a matter of discussion, as is the identity of the RNA-binding subunit of this complex. The core of PRC2 is composed of 4 subunits, Suz12, Eed, Rbbp4 and Ezh2, where Ezh2 contains the methylase catalytic activity.¹¹ Moreover, other proteins can also associate to the core PRC2; for instance, Jarid2 binds non-stoichiometrically to the complex and modulates its enzymatic activity. It has been described that Suz12, Jarid2 and Ezh2 bind RNA, with Ezh2 suggested to have the highest affinity and the lowest specificity.¹⁸ Another subunit of the complex, Eed, enhances the binding specificity of the complex to chromatin, whereas Jarid2 relieves the inhibition of Ezh2 catalytic activity caused by RNA binding.¹⁸ These results were contrasted by data from Cech and coworkers who reported that binding of RNA to the PRC2 complex is promiscuous and mainly depends on RNA size.¹⁹ However, not all RNAs are bound with the same affinity, although the differences in K_d are, in general, not greater than 10-fold for size-matched RNAs.²⁰ According to these authors, longer RNAs interact better, although factors other than length, such as the capability of the RNA to form specific secondary structures, might be relevant since these hairpins would be more abundant in long RNAs. In this regard, an enrichment of PRC2 in GC elements has been detected by ChIP-Seq analysis.²¹ Moreover, genome wide screens for RNAs bound by PRC2 in cells using CLIP technologies have identified nascent RNAs as targets for PRC2 complex at the same level as lncRNAs.²² Accordingly, PRC2 would be bound to many transcriptionally active, not-silenced (poised) promoters through these RNAs^23,24; the capability to block transcription would rest on its transfer to the chromatin depending on RNA length, stability and other transcription-related parameters of each specific locus.

We have recently described the interaction of PRC2 to an antisense lncRNA encoded by the LEF1 locus.²⁵ LEF1 is a gene induced during the process of epithelial-to-mesenchymal transition (EMT), a cellular conversion that provides the cells with invasive and migratory characteristics.²⁶ LEF1 expression is upregulated during EMT and contributes to the fixation of the mesenchymal phenotype. EMT happens through several intermediate states; these intermediate states are particularly relevant because they have been associated with the acquisition by epithelial cells of stem cell-like characteristics.²⁷

LEF1 antisense transcript (AT) is mostly in the nucleus, as expected for a lncRNA controlling transcription, and is synthetized from a promoter present in the DNA corresponding to the first intron of LEF1 mRNA. It acts in cis: a 1.5 kb fragment of this AT is sufficient to ectopically inhibit LEF1 gene expression, since it associates both with LEF1 promoter and PRC2 (Fig 1A). Curiously, its stability seems to be controlled by PRC2, since its levels are affected by Ezh2 downregulation, a characteristic that might be shared by other lncRNAs. It is expressed at levels far below those of LEF1 mRNA, suggesting that it does not act by destabilizing this RNA by direct interaction or physically preventing its export to the cytosol.

Figure 1.

(A) in the LEF1gene an antisense transcript (AT) (in black and gray) is synthetized from a promoter present in the first intron of LEF1 RNA (in dark blue; the rest of the RNA, in light blue) and overlaps the LEF1 promoter (in white). This AT contains 2 domains for binding to the LEF1 promoter (in black) and for the interaction with PRC2 (in gray); consequently, it facilitates the access of PRC2 to LEF1 promoter and limiting the transcription of the LEF1 RNA. (B) in some conditions (see text) this AT is spliced in a much shorter form unable to associate with PRC2 but still interacting with the promoter. This spliced RNA prevents the interaction of the remaining unspliced AT to the promoter and the inhibitory effect of the longer form of this transcript. (C) this cis-mediated action of the AT in the recruitment of PRC2 might be complemented by another interaction dependent on transcriptional factors and also mediated by lncRNAs, in this case acting in trans. This trans-acting lncRNAs (in dark red) will present a domain with higher affinity for PRC2 than most cis-acting lncRNAs and will also be capable to associate with co-repressor (in green, CoR) and through them, with transcriptional repressor (TR), binding to specific sequences in LEF1 promoter; therefore facilitating the recruitment of PRC2 and the inhibition of promoter activity.

Interestingly, this LEF1 AT undergoes alternative splicing, which generates a much shorter form that does not associate to PRC2 (Fig 1B). In fact, the spliced form precludes the inhibitory action of the unspliced form, since it prevents the interaction of this longer form with the LEF1 promoter and the recruitment of PRC2. Expression and splicing of this AT is associated with LEF1 mRNA expression in several experimental conditions, such as during EMT transition. In epithelial cells, where LEF1 is completely inhibited, the activity of both sense and antisense promoters is off and LEF1 AT is not transcribed. In certain conditions this locus becomes activated and LEF1 mRNA and AT start to be expressed. In these cells, corresponding to an intermediate epithelial-mesenchymal phenotype, the AT is not spliced and retains the elements required for binding to LEF1 promoter and PRC2, thus limiting LEF1 mRNA expression. However, in epithelial cells, LEF1 AT becomes spliced and is unable to downregulate LEF1 mRNA, despite being expressed at higher levels than in cells with the intermediate phenotype. In addition, as indicated above, the spliced AT behaves as a dominant negative inhibitor of the unspliced form, preventing its inhibitory activity.

This study presents several interesting implications. First, it is possible that this mechanism is more general and PRC2 binding is driven by other antisense lncRNAs acting in cis, thus overlapping and interacting with the promoter of the sense gene. These ATs would be expressed when the locus is activated and would associate with PRC2, taking advantage of their long size and maybe also their ability to form secondary structures. These lowly-expressed lncRNA would stabilize PRC2 interaction with the promoter, thereby maintaining its inhibition. PRC2 binding would act as a safety control, preventing the inadequate expression of LEF1 and other genes or limiting their expression in inappropriate conditions. Positive signals might decrease PRC2 binding to the promoter, either by the splicing of the antisense transcript or by promoting the association of transcriptional activators, activators that would prevent the interaction of the transcript with the promoter. Moreover, binding to PRC2 might also be controlled by the interaction of the lncRNA with other molecules that hide this RNA and preclude the association of PRC2. It is also possible that unspliced sense transcripts, particularly the intronic elements, might also act in a similar way, facilitating the interaction of PRC2 to the antisense promoter. As for LEF1, these antisense promoters might be located in the DNA corresponding to the first intron. The recruitment of PRC2 to the antisense promoter by the sense strand might be specific and not affecting the sense promoter; this might depend on the accessibility of PRC2 to specific regions of the sense strand. In this respect, Drosophila PRC2 interacts differently in vivo with the sense and antisense strands of the Polycomb-binding element in the vestigial gene.²⁸ It is possible that, as these authors suggest, the highly promiscuous binding of PRC2 to RNA in vitro might be corrected in vivo by RNA-binding proteins, providing some level of specificity.

It is also worth mentioning that, besides the possible action of RNA-binding proteins, LEF1 antisense-dependent inhibition relies on the lack of splicing of this antisense transcript. Some studies have shown a relationship between chromatin silencing and splicing^7,29; however, in these reports altered splicing seems to be the consequence, not the cause of chromatin modifications, contrarily what happens with LEF1 AT. Intron elimination would abolish PRC2 binding, precluding the negative effects of these transcripts. Therefore, a deficient recruitment of the splicing machinery to the antisense lncRNA would be a method to limit the activity of open, poised, promoters. It is possible that this general and promiscuous binding of PRC2 to promoters guided by cis-acting antisense lncRNAs is complemented by a more specific interaction relying on the binding of transcriptional repressors to specific target promoters. As discussed before, these repressors might recruit PRC2 through an interaction with co-repressors such as LSD1 that can also associate with HOTAIR (Fig 1C). Unlike cis-acting lncRNAs, these trans-lncRNAs should present a higher affinity than most RNAs for PRC2 and, together with the better association to DNA than transcription factors exhibit with respect to RNA, would create stronger binding sites for PRC2 in specific target promoters. Thus, the two systems might be in operation, being the antisense-dependent mechanism a general system to limit gene expression, whereas the transcription-repressor-mediated would be more specific for genes totally repressed.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Funding

Work in Antonio Garcia de Herreros' lab on lnRNAs was supported by a grant from Worldwide Cancer Research and Ministerio de Economía y Competitividad (SAF 48849-C2-1R). Manuel Beltran is currently supported by an EMBO long-term fellowship.