RADICL-seq identifies general and cell type-specific principles of genome-wide RNA-chromatin interactions

Abstract

Mammalian genomes encode tens of thousands of noncoding RNAs. Most noncoding transcripts exhibit nuclear localization and several have been shown to play a role in the regulation of gene expression and chromatin remodeling. To investigate the function of such RNAs, methods to massively map the genomic interacting sites of multiple transcripts have been developed; however, these methods have some limitations. Here, we introduce RNA And DNA Interacting Complexes Ligated and sequenced (RADICL-seq), a technology that maps genome-wide RNA-chromatin interactions in intact nuclei. RADICL-seq is a proximity ligation-based methodology that reduces the bias for nascent transcription, while increasing genomic coverage and unique mapping rate efficiency compared with existing methods. RADICL-seq identifies distinct patterns of genome occupancy for different classes of transcripts as well as cell type-specific RNA-chromatin interactions, and highlights the role of transcription in the establishment of chromatin structure.

Fig. 1. RADICL-seq method for the identification of RNA–chromatin interactions.

a Schematic representation of the RADICL-seq protocol. Top: sequence of enzymatic reactions occurring in fixed nuclei after partial lysis of the nuclear membrane. The adduct formed by genomic DNA (black), RNA (red), and proteins (blue circles) is subjected to controlled DNase I digestion and chromatin preparation. After RNase H digestion, an adapter (dark blue) containing an internally biotinylated residue (black dot) bridges the RNA and DNA molecules that lie in close proximity. Bottom: sequence of enzymatic reactions performed in solution. After reversal of crosslinks, the RNA–DNA chimera is converted into a fully double-stranded DNA (dsDNA) molecule and digested by the EcoP15I enzyme to a designated length (adpr, adapter). After ligation of the sequencing linkers (yellow) and biotin pull-down, the library is PCR amplified and high-throughput sequenced. b Reproducibility of the RNA–DNA interaction frequencies across replicates, assessed by counting the occurrences of genic transcripts and 25-kb genomic bin pairs. c RNA and d DNA tag origins. The inner pie charts represent a broader classification into intergenic and genic (annotated genes), while the outer circles show a finer classification of the genic portion. e, f Comparison between (e) nuclear and (f) cytosolic RNA-seq tag counts and RADICL-seq genic transcript counts. The former are normalized to tags per million (TPM), while the latter are normalized to reads per kilobase (RPK). The linear regression lines are shown in red. g Density of the normalized counts of DNA reads detected by RADICL-seq around ATAC-seq (blue), DHS-seq (green), and H3K9me3 ChIP-seq (red) peaks; dashed lines represent the density profiles of aggregated signal from random genomic reads equal in number and size to the real peaks.

Fig. 2. Comparison of RADICL-seq with similar methods.

a Summary of the features that distinguish RADICL-seq from GRID-seq. b Analysis of the read length and mapping outcome. Unique (dark gray) and multi-mapping (blue) reads are reported as percentage of the total number of reads pool. RADICL-seq reads were artificially trimmed down to 20 nt for direct comparison with the GRID-seq dataset. c Assessment of the genomic coverage as a function of the sequencing depth for RADICL-seq (blue) and GRID-seq (yellow). The coverage was calculated for both datasets by sub-sampling with a step of 1,000,000 reads up to the maximum available number of reads. d Distribution of the linear genomic distance between RNA and DNA tags derived from the same read for the GRID-seq (yellow) and RADICL-seq (blue) datasets. Data are presented as mean ± s.d.; statistical significance was calculated with one-sided two-proportions z test; *P ≤ 0.05. e Comparison of Malat1 transcript target DNA loci in mESCs identified by RAP-DNA (yellow), RADICL-seq (grey), and GRID-seq (blue) methods. f Comparison of Rn7sk transcript target DNA loci in mESCs identified by ChIRP-seq (yellow), RADICL-seq (grey), and GRID-seq (blue) methods. All panels were generated using RADICL-seq total dataset. Source data are available in the Source Data File.

Fig. 3. RADICL-seq identifies genome-wide RNA–chromatin interactions.

a, b RNA–DNA interactions in mESCs shown as a single point per 25-kb bin and colored (see key in c) by the most-represented RNA region or RNA class, respectively, in that bin. c RNA–DNA interactions in mESCs, quantified according to genomic distance between RNA and DNA tags. d, e RNA–DNA interaction matrix for mOPCs, similar to that shown in a, b. f RNA–DNA interactions in mOPCs, quantified according to genomic distance between RNA and DNA tags. All panels were generated using significant total datasets.

Fig. 4. RNA interactions and genomic structural features.

a–d Metadata profiles showing the average coverage of a, b DNA tags or c, d RNA tags at the boundaries of TADs in mESCs (a, c) and mOPCs (b, d) for the indicated conditions. e, f Average RNA-binding signal at TAD boundaries in total mESCs and mOPCs, respectively. The signal was split based on whether the RNA originated within or outside of the TAD. Coverage distance from the TAD is relative to the corresponding TAD width. g, h Percentage of RNA–DNA interactions in total RADICL-seq libraries from mESCs and mOPCs, respectively, divided into incremental RNA–DNA distances and with RNA tags grouped by the identity of intersected repeat elements. NR, RNA tags not mapped on any repeat. Statistical significance was calculated with the two-tailed Student’s t test. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001. All panels were generated using significant datasets. Source data are available in the Source Data File.

Fig. 5. Cell-type-specific RNA–chromatin interaction patterns.

a For each gene, the Jaccard distance between the genome-wide RNA–DNA binding profiles in mESCs and mOPCs is compared with the difference in capture rate (proxy for gene expression) between the two cell types. b Distribution of the log₂ ratios of RADICL-seq mESC to mOPC DNA normalized tag counts for targeted gene promoter regions, which are defined as ±2 kb around the TSS. Cell-type-specific marker gene positions are highlighted for both mESCs (red) and mOPCs (blue). Counts were normalized by library size. c RADICL-seq counts of unique genomic targets per interacting RNA in mESCs vs. mOPCs. d, e Circos plots depicting Neat1 genomic interactions in mESCs and mOPCs, respectively. f, g Circos plots depicting Fgfr2 genomic interactions in mESCs and mOPCs, respectively. Each line represents the interaction between the genic transcript and the contacted genomic bin, while its color indicates log₂ of the RADICL-seq count. The chromosome of origin of the RNA under investigation is shown enlarged (gray shading) on the left portion of each circos plot. All panels were generated using significant total datasets.