Genomic maps of long noncoding RNA occupancy reveal principles of RNA-chromatin interactions

Abstract

Long noncoding RNAs (lncRNAs) are key regulators of chromatin state, yet the nature and sites of RNA-chromatin interaction are mostly unknown. Here we introduce Chromatin Isolation by RNA Purification (ChIRP), where tiling oligonucleotides retrieve specific lncRNAs with bound protein and DNA sequences, which are enumerated by deep sequencing. ChIRP-seq of three lncRNAs reveal that RNA occupancy sites in the genome are focal, sequence-specific, and numerous. Drosophila roX2 RNA occupies male X-linked gene bodies with increasing tendency toward the 3' end, peaking at CES sites. Human telomerase RNA TERC occupies telomeres and Wnt pathway genes. HOTAIR lncRNA preferentially occupies a GA-rich DNA motif to nucleate broad domains of Polycomb occupancy and histone H3 lysine 27 trimethylation. HOTAIR occupancy occurs independently of EZH2, suggesting the order of RNA guidance of Polycomb occupancy. ChIRP-seq is generally applicable to illuminate the intersection of RNA and chromatin with newfound precision genome wide.

Figure 1. Chromatin isolation by RNA purification

(A) Workflow of ChIRP. Chromatin is crosslinked to lincRNA:protein adducts in vivo. Biotinylated tiling probes are hybridized to target lncRNA, and chromatin complexes are purified using magnetic streptavidin beads, followed by stringent washes. We elute lncRNA bound DNA or proteins with a cocktail of Rnase A and H. A putative lincRNA binding sequence is schematized in orange. (B) Northern blot shows HOTAIR RNA is sheared into the size range of 100-500nt by sonication. (C) Design of antisense DNA tiling probes, grouped into “even” and “odd” sets based on their positions along the RNA. (D) Complementary DNA tiling oligonucleotides effectively retrieve ~95% of HOTAIR RNA from chromatin, as compared to ~10% by morpholino probes. Mean ± s.d. are shown.

Figure 2. ChIRP enriches for TERC RNA and detects TERC-associated telomere DNA and TCAB1 protein

(A) TERC-asDNA probes retrieve ~88% of cellular TERC RNA and undetectable GAPDH. LacZ-asDNA probes retrieve neither RNAs. Mean ± s.d. are shown. (B) Effect of different crosslinking agents on ChIRP-southern. After 1% glutaraldehyde crosslinking, TERC retrieval co-purifies telomeric repeats, but not Alu repeats. (C) TERC ChIRP retrieves TCAB1, a known telomerase holocomplex chaperone proteins. As a negative control tubulin was not detected.

Figure 3. ChIRP-seq reveals roX2 binding sites on X chromosome

(A) roX2 co-localizes with MSL complex and CES. (B) 308 roX2 binding sites are all on the X chromosome, indicating an FDR ~ 0. (C) roX2 ChIRP-seq is overall highly correlated to MSL3 ChIP-seq (R = 0.77 for log2 intensity > 10). (D) roX2 binds across X-linked gene bodies with bias towards the 3’ end, in a manner similar to MSL3 but with higher dynamic range. ChIRP-seq or ChIP-seq signal intensity for all bound genes on X or genes on chromosome 2L were averaged and to gene start and end annotations. (E) roX2 binding sites are strongly enriched for a sequence motif that is nearly identical to MSL3 motif.

Figure 4. TERC ChIRP-seq

(A) Fold enrichment of reads from TERC-ChIRP-seq and Input sample that map to telomere and Alu sequences. (B) TERC peaks are focal. (C) TERC occupancy at the MYC promoter, overlaying regions of TERT occupancy (Park et. al., 2009) and regions of dense transcription factors occupancy identified by the ENCODE project (bottom). (D) A cytosine-rich motif enriched among TERC-binding sites (e-Value = 3.7e-966).

Figure 5. HOTAIR ChIRP-seq suggests mechanisms of HOTAIR-recruitment of PRC2

(A) HOTAIR binding sites are enriched in genic regions, notably enhancers and introns. (B) Metagene analysis of genomic regions aligned by 832 HOTAIR ChIRP peaks show focal HOTAIR peaks in association with broad domains PRC2 occupancy (evidenced by subunits EZH2 and Suz12) and H3K27Me3. (C) HOTAIR nucleates broad domains of PRC2 occupancy. A HOTAIR binding site between HOXD3 and HOXD4 lies in the center of a broad domain of Suz12 and H3K27Me3 occupancy that are both lost upon HOTAIR knock down (Tsai et. al., 2010, Rinn et. al., 2007). (D) GA-rich homopurine motif enriched in HOTAIR binding sites.

Figure 6. HOTAIR binds chromatin in a PRC2-independent manner

(A) Heatmap of HOTAIR ChIRP-seq signal in peak regions. Each row is a 4 kB genomic window centered on a HOTAIR ChIRP peak in control cells; the peaks are aligned for the 832 HOTAIR bound sites (left panel). Red color intensity indicates the number of ChIRP-seq reads. The equivalent genomic windows in control and shEZH2 cells show that LacZ ChIRP retrieved no signal (right panel) while shEZH2 did not diminish or alter the profile of HOTAIR occupancy (middle panel). (B) ChIRP-qPCR validation of peaks from (A). TERC and GAPDH served as negative controls. Mean ± s.d. are shown.