The control of polycomb repressive complexes by long noncoding RNAs

Abstract

The polycomb repressive complexes 1 and 2 (PRCs; PRC1 and PRC2) are conserved histone-modifying enzymes that often function cooperatively to repress gene expression. The PRCs are regulated by long noncoding RNAs (lncRNAs) in complex ways. On the one hand, specific lncRNAs cause the PRCs to engage with chromatin and repress gene expression over genomic regions that can span megabases. On the other hand, the PRCs bind RNA with seemingly little sequence specificity, and at least in the case of PRC2, direct RNA-binding has the effect of inhibiting the enzyme. Thus, some RNAs appear to promote PRC activity, while others may inhibit it. The reasons behind this apparent dichotomy are unclear. The most potent PRC-activating lncRNAs associate with chromatin and are predominantly unspliced or harbor unusually long exons. Emerging data imply that these lncRNAs promote PRC activity through internal RNA sequence elements that arise and disappear rapidly in evolutionary time. These sequence elements may function by interacting with common subsets of RNA-binding proteins that recruit or stabilize PRCs on chromatin. This article is categorized under: RNA Interactions with Proteins and Other Molecules > Protein-RNA Recognition RNA Interactions with Proteins and Other Molecules > RNA-Protein Complexes RNA Interactions with Proteins and Other Molecules > Protein-RNA Interactions: Functional Implications.

Keywords: Airn; HOTAIR; Xist; lncRNA; polycomb.

FIGURE 1

Composition of PRC1 subcomplexes. PRC1 has two main forms, canonical PRC1 (cPRC1) and variant PRC1 (vPRC1). cPRC1 induces PRC1 oligomerization and chromatin compaction. The CBX subunit of cPRC1 recognizes and is recruited by H3K27me3. The RYBP subunit of vPRC1 stimulates vPRC1 catalytic activity and can also recognize H2AK119ub1

FIGURE 2

Composition of the PRC2 subcomplexes. PRC2 has two main forms, PRC2.1 and PRC2.2. For more information, see (Laugesen et al., 2019; Schuettengruber et al., 2017; J. R. Yu et al., 2019). The PCL1–3 subunits of PRC2.1 appear to be important for recruiting PRC2.1 to CpG Islands. The JARID2 subunit of PRC2.2 recognizes and is recruited by H2AK119ub1. EPOP may function to promote a low level of transcription from within domains repressed by PRC2. PALI appears to stimulate the enzymatic activity of PRC2 in vitro and in vivo. Both JARID2 and AEBP2 allosterically activate PRC2. AEBP2 also binds methylated DNA and may stabilize PRC2 over a subset of genomic regions harboring methylated DNA

FIGURE 3

Modes of interaction between the PRCs and RNA. (a) PRC2 binds RNA promiscuously, but this direct binding blocks the catalytic activity of PRC2, rendering it unable to place H3K27me3. (b) The RNA-binding protein HNRNPK bridges the lncRNA Xist and PRC1, bringing PRC1 to the X chromosome (Pintacuda, Wei, et al., 2017). (c) PRC2 can be recruited by RBFOX2 through pre-mRNA to dampen expression of genes that are already susceptible to PRC2-mediated silencing (Wei et al., 2016). (d) Either through direct or indirect interactions, RNA appears to be essential to stabilize PRC2 at specific sites throughout the genome (Long et al., 2020)

FIGURE 4

Recruitment of PRCs by Xist during mouse XCI. vPRC1 recruitment by HNRNPK and Repeats B/C recruits PRC2 and cPRC1 to the inactive X (Almeida et al., 2017; Bousard et al., 2019; Colognori et al., 2019; Pintacuda, Wei, et al., 2017). Repeat A is also necessary to recruit the PRCs (Colognori et al., 2020). Repeat E associates with components of the PRCs and may play some role in recruitment to the inactive X (Hendrickson et al., 2016; M. K. Ray et al., 2016)

FIGURE 5

Xist transcript structure and Repeat regions. Cartoon structures are modeled from a combination of RNA-fold predictions (Gruber et al., 2008) and the specific works referenced in each panel. (a) UCSC Genes’ human and mouse Xist isoforms, annotated with location of Xist repeats (Haeussler et al., 2019). Repeat B is split by insertion in human Xist. (b) Repeat A is present in both human and mouse and interacts with many RBPs including SPEN, HNRNPC RALY, and several SR proteins (Chu et al., 2015; Cirillo et al., 2016; Graindorge et al., 2019; Pintacuda, Wei, et al., 2017; Trotman et al., 2020). The cartoon structure represents what is in our opinion one of the more likely conformations of Repeat A—one in which the predominant base-pairing occurs between, and not necessarily within, each repeat (Duszczyk et al., 2011; Z. P. Lu et al., 2016; Maenner et al., 2010). For more information, we direct the reader to recent reviews that summarize the structural data that have been collected for Repeat A (Jones & Sattler, 2019; Pintacuda, Young, & Cerase, 2017). (c) Repeat B robustly binds HNRNPK and largely consists of repeating HNRNPK binding-motifs (Cirillo et al., 2016; Colognori et al., 2019; Nakamoto et al., 2020; Pintacuda, Wei, et al., 2017). Cartoon structure is modeled from (Nakamoto et al., 2020) and RNA-fold predictions. (d) Repeat C is rodent specific and binds HNRNPK and HNRNPU, among other proteins (Bousard et al., 2019; Cirillo et al., 2016; Graindorge et al., 2019). (e) Repeat D is the longest and most complex Xist repeat, and is present in most non-rodents (Nesterova et al., 2001; Sprague et al., 2019; Van Nostrand et al., 2016; Yen et al., 2007). (f) Repeat E is largely unstructured and is known to associate with PTBP1, MATR3, TDP-43, CELF1, and CIZ1 (Cirillo et al., 2016; Pandya-Jones et al., 2020; Ridings-Figueroa et al., 2017; Smola et al., 2016; Sunwoo et al., 2017; Van Nostrand et al., 2016). Cartoon structure is modeled from (Smola et al., 2016) and RNA-fold predictions

FIGURE 6

Imprinted lncRNAs in their imprinted domains. UCSC Genes’ noncoding RNA gene structures (left) and schematic of imprinted targeted domains (right) of the (a) Airn, (b) Kcnq1ot1, and (c) Meg3 lncRNAs in mice. In the left-hand panels, filled rectangles correspond to UCSC-annotated exons, and fishbone structures correspond to UCSC-annotated introns (Haeussler et al., 2019). In the right-hand panels, the location of each imprinted domain relative to its position on its corresponding chromosome is shown. The Airn, Kcnq1ot1, and Meg3 genes are colored purple, and the protein-coding genes that are repressed in each domain are colored red. Genes whose expression is bi-allelic within the imprinted domain are colored black. (a) The Airn locus is located on mouse chromosome 17. The Airn lncRNA is expressed predominantly as an unspliced and unstable 90–120 kb transcript from the paternally inherited allele and silences up to 13 Mb of chromatin. (b) The Kcnq1ot1 locus is located on mouse chromosome 7. The Kcnq1ot1 lncRNA is expressed predominantly as an unspliced and unstable 80–95 kb transcript from the paternally inherited allele and silences up to 3 Mb of chromatin. (c) The Meg3 locus is located on mouse chromosome 12. The Meg3 lncRNA is alternatively spliced and from the maternally inherited allele. While Meg3 is implicated to silence neighboring genes in cis (shown here), it has been suggested to regulate silencing of other genes in trans (not shown here)

FIGURE 7

A potential model for how the Airn lncRNA coordinates gene silencing and PRC deposition throughout a 13 Mb domain. (Top) Linear schematic representation of the 13-Mb Airn lncRNA-targeted domain on mouse chromosome 17. (Left) Predicted conformation of targeted domain on the maternal nonexpressing allele. Within the targeted domain, gene promoters that harbor CpG islands bound by the PRCs may form 3D contacts with each other and the CpG islands near Airn, creating a pre-existing conformational state that allows Airn to access distal targets despite remaining localized to its site of transcription. (Right) Predicted conformation of the targeted domain on the paternal lncRNA-expressing allele. Upon Airn expression, the lncRNA may preferentially contact regions of chromatin that are in close proximity to its site of transcription—in this case, we would predict that those regions of contact would harbor CpG islands that were already bound by the PRCs. HNRNPK and potentially other RBPs that interact with Airn help to create a high local concentration of PRCs around the lncRNA locus; these PRCs could then engage with nearby chromatin and spread over broad domains

FIGURE 8

Epigenetic control of the plant FLC gene by PRC2 and lncRNAs COOLAIR, COLDAIR, and COLDWRAP. Cold-induced repression of the FLC floral repressor gene occurs in three consecutive steps. First, upon exposure to cold temperatures, FLC transcription is transiently silenced in a PRC2-independent manner that involves expression of the antisense COOLAIR lncRNAs. Second, following extended time in the cold, the sense COLDAIR and COLDWRAP lncRNAs are expressed, which promote PRC2 spreading and H3K27me3 accumulation in a nucleation region at the 5′ end of FLC. Third, after a return to warmer growing temperatures, PRC2 and H3K27me3 spread outward from the nucleation region, causing FLC to be maintained in a silenced state that permits the transition to flowering