Mapping Native R-Loops Genome-wide Using a Targeted Nuclease Approach

Abstract

R-loops are three-stranded DNA:RNA hybrids that are implicated in many nuclear processes. While R-loops may have physiological roles, the formation of stable, aberrant R-loops has been observed in neurological disorders and cancers. Current methods to assess their genome-wide distribution rely on affinity purification, which is plagued by large input requirements, high noise, and poor sensitivity for dynamic R-loops. Here, we present MapR, a method that utilizes RNase H to guide micrococcal nuclease to R-loops, which are subsequently cleaved, released, and identified by sequencing. MapR detects R-loops formed at promoters and active enhancers that are likely to form transient R-loops due to the low transcriptional output of these regulatory elements and the short-lived nature of enhancer RNAs. MapR is as specific as existing techniques and more sensitive, allowing for genome-wide coverage with low input material in a fraction of the time.

Keywords: DNA:RNA hybrids; R-loops; chromatin; gene expression; transcription.

Figure 1.. MapR, a Native and Antibody-Independent R-Loop Detection Strategy

R-loop recognition and recovery by MapR. Step 1: cells are immobilized on concanavalin A beads and permeabilized. Step 2: equimolar amounts of a catalytic deficient mutant of RNase H fused to micrococcal nuclease (GST-RHΔ-MNase) or GST-MNase is added to immobilized cells. Step 3: the RHΔ module recognizes and binds R-loops on chromatin. Step 4: controlled activation of the MNase moiety by addition of calcium results in cleavage of DNA fragments in proximity to R-loops. Step 5: Released R-loops diffuse out of the cell; the DNA is recovered and sequenced.

Figure 2.. MapR and RHΔ CUT&RUN Signals Are Enriched at Similar Regions Genome-wide

(A) Schematic of RHΔC&R using FLAG M2 antibody (left) and MapR using GST-RHΔ-MNase (right) in HEK293. (B) Enriched regions identified by RHΔC&R and R-ChIP in HEK293. GRO-seq and H3K4me3 tracks indicate active gene transcription. (C) Venn diagram of gene-level overlap between RHΔC&R and R-ChIP. Total number of unique genes with an R-loop at the promoter region (−2kb/+2kb from the TSS) and their overlap are shown. p < 10⁻¹⁵, hypergeometric distribution. (D) Peak distribution of MapR and RHΔC&R showing percent of peaks mapping to promoter regions (−2kb/+2kb from the TSS), gene bodies (entirety of gene including introns, excluding promoter region), or intergenic regions. Total peak numbers are shown in parentheses. Background genomic distribution is shown for comparison. (E) MapR and RHΔC&R signals at the XIST and TSIX genes. GST-MNase and IgG controls are shown for MapR and RHΔC&R, respectively. H3K4me3 (Thurman et al., 2012) and H3K27Ac (Frietze et al., 2012) chromatin immunoprecipitation sequencing (ChIP-seq) and GRO-seq (Chen et al., 2017) tracks are shown as proxies for transcriptional activity. (F) Venn diagram of gene-level overlap between RHΔC&R and MapR. Total number of unique genes with an R-loop at the promoter region (−2kb/+2kb from the TSS) and their overlap are shown. p < 10⁻¹⁵, hypergeometric distribution. (G) Correlation scatterplot showing read densities for the union of peaks from MapR and RHΔC&R (log2 scale). r = 0.76, Spearman correlation coefficient. (H) Heatmaps of H3K27Ac, H3K27me3, MapR, and RHΔC&R signal intensity across all TSS sorted by MapR signal. GRO-seq signals were summed and collapsed into a box per gene. (I) Metagene plots of MapR (left) and RHΔC&R (right) signals at all TSSs (black) and TESs (red).

Figure 3.. Characterization of R-Loops Obtained by MapR

(A) Schematic of actinomycin D (ActD) treatment followed by MapR identification of R-loops. (B) Genome browser views of GBAP1 and IPP genes showing MapR signals with and without ActD treatment. GRO-seq tracks show active transcription. (C) Metagene plots of MapR signals at TSS of all genes with and without ActD treatment. (D) Schematic of RNase H treatment followed by MapR identification of R-loops. (E) Genome browser views RWDD1 and ANP32E genes showing MapR signals with and without RNase H treatment. GRO-seq tracks show transcription at these genes. (F) Metagene plots of MapR signals at TSS of all genes with and without RNase H treatment. (G) Heatmaps of MapR signals across all TSS in control, RNase-H-, and ActD-treated HEK293 cells, sorted by MapR signal. GRO-seq signals from untreatedHEK293 were summed and collapsed into a box per gene. (H) Genome browser views of RCHY1 gene showing overlapping MapR and DNase I hypersensitivity signals. GRO-seq tracks show active transcription. (I) Genome browser views of NIT2 gene showing DNase I hypersensitivity signals without appreciable MapR signal. GRO-seq tracks show active transcription. (J) Base-level overlap between DNase I signal and MapR, RHΔC&R, RChIP, RDIP, and DRIP signal.

Figure 4.. Similarities and Differences between MapR and Other R-Loop Detection Methods

(A) Genome browser view of the USP24 gene showing MapR, RHΔC&R, R-ChIP, DRIP, and RDIP signals. The scale for the y axis is in RPM. (B) Gene-level overlap between MapR and R-ChIP datasets (left) and MapR and DRIP datasets (right). Total number of unique genes with an R-loop at the promoter region (−2kb/+2kb from the TSS) and their overlap are shown. p < 10⁻¹⁵, hypergeometric distribution. (C) Genome browser view of PTAR1 that shows MapR but no R-ChIP, DRIP, or RDIP signals. GRO-seq tracks indicate active transcriptional status. (D) Ratio in the −2kb/+2 kb window around the TSS in MapR signal in untreated cells and cells treated with actinomycin D to inhibit transcription. MapR signal from genes identified by each R-loop detection method and genes that did not contain R-loops are shown. (E) Genes called from MapR, RHΔC&R, R-ChIP, DRIP, and RDIP in 293 cells show similar frequency of predicted G quadruplex structures at their promoter, indicative of R-loop presence, whereas non-R-loop genes have a lower frequency of G quadruplexes. (F) Distance between MapR peaks and peaks detected by RHΔC&R, RChIP, RDIP, and DRIP at genes with R-loop detected at promoter. (G) Genome browser view of VAX2, that shows MapR signals that do not overlap with RDIP peaks in close proximity. Simple tandem repeat (STR) and GRO-seq tracks are shown. The scale for the y axis is in RPM. (H) Percent of peaks that contain STRs in MapR, RHΔC&R, RDIP, and R-ChIP experiments in HEK293. Results from published RDIP (IMR90)(Nadel et al., 2015) and DRIP (293 and K562) (Manzo et al., 2018; Sanz et al., 2016) datasets are also shown. (I) Distance (kb) from peaks detected by each technology to the nearest STR. (J) Heatmaps of GRO-seq signal intensity across STRs identified by each method in HEK293. Each heatmap is sorted by intensity.

Figure 5.. MapR Identifies R-Loops Formed at Enhancers with Higher Sensitivity

(A) Genome browser view of an intergenic enhancer with an R-loop detected by MapR, RHΔC&R, DRIP, RDIP, and R-ChIP. Chromosome and genome coordinates are shown above the panels. H3K27ac, H3K4me1, and H3K4me3 tracks are also shown to verify that the region has a signature characteristic of an active enhancer. (B) Genome browser view of an intergenic enhancer with an R-loop detected by MapR and RHΔC&R but not by DRIP, RDIP, or R-ChIP. Chromosome and genome coordinates are shown above the panels. H3K27ac, H3K4me1, and H3K4me3 tracks are also shown to verify that the region has a signature characteristic of an active enhancer. (C) Percentage of intergenic peaks that overlap with an enhancer for each technology (left), and the percentage of intergenic sequence that is characterized as an enhancer by chromHMM (right). (D) Percentage of active and poised enhancers that contain a R-loop detected by MapR or RHΔC&R.