Mapping R-Loops and RNA:DNA Hybrids with S9.6-Based Immunoprecipitation Methods

R-loops are three-stranded nucleic acid structures that form primarily during transcription upon hybridization of the nascent RNA transcript to the template DNA strand. This results in the formation of an RNA:DNA hybrid and causes the displacement of the non-template DNA strand in a single-stranded looped state. Biochemical reconstitution^1,2,3,4 and mathematical modeling⁵, in combination with other biophysical measurements^6,7, have established that R-loops are more likely to occur over regions that exhibit specific favorable characteristics. For instance, regions that display strand asymmetry in the distribution of guanines (G) and cytosines (C) such that the RNA is G-rich, a property called positive GC skew, are favored to form R-loops when transcribed owing to the higher thermodynamic stability of the DNA:RNA hybrid compared to the DNA duplex⁸. Regions that have evolved positive GC skew, such as the early portions of many eukaryotic genes^4,9,10,11, are prone to forming R-loops in vitro and in vivo^3,4,12. Negative DNA superhelical stress also greatly favors structure formation^13,14 because R-loops efficiently absorb such topological stresses and return the surrounding DNA fiber to a favorable relaxed state^5,15.

Historically, R-loop structures were considered to result from rare, spontaneous, entanglements of RNA with DNA during transcription. However, the development of DNA:RNA immunoprecipitation (DRIP) coupled with high-throughput DNA sequencing (DRIP-seq) allowed the first genome-wide mapping of R-loops and revealed that those structures are far more prevalent than expected in human cells^4,16. R-loops occur over tens of thousands of conserved, transcribed, genic hotspots in mammalian genomes, with a predilection for GC-skewed CpG islands overlapping the first intron of genes and the terminal regions of numerous genes¹⁷. Overall, R-loops collectively occupy 3%-5% of the genome in human cells, consistent with measurements in other organisms, including yeasts, plants, flies, and mice^{18,19,20,21,22}.

Analysis of R-loop forming hotspots in human cells revealed that such regions associate with specific chromatin signatures²³. R-loops, in general, are found over regions with lower nucleosome occupancy and higher RNA polymerase density. At promoters, R-loops associate with increased recruitment of two co-transcriptionally deposited histone modifications, H3K4me1 and H3K36me3¹⁷. At gene termini, R-loops associate with closely arranged genes that undergo efficient transcription termination¹⁷, consistent with prior observations²⁴. R-loops were also shown to participate in the initiation of DNA replication at the replication origins of bacteriophage, plasmid, mitochondrial, and the yeast genomes^{25,26,27,28,29,30,31}. In addition, 76% of R-loop-prone human CpG island promoters function as early, constitutive replication origins^32,33,34,35, further reinforcing the connections between R-loops and replication origins. Collectively, these studies suggest that R-loops represent a novel type of biological signal that can trigger specific biological outputs in a context-dependent manner²³.

Early on, R-loops were shown to form at class switch sequences during the process of immunoglobulin class switch recombination^3,36,37. Such programmed R-loops are thought to initiate class switch recombination through the introduction of double-stranded DNA breaks³⁸. Since then, harmful R-loop formation, generally understood to result from excessive R-loop formation, has been linked to genomic instability and processes such as hyper recombination, transcription-replication collisions, replication, and transcriptional stress (for review^{39,40,41,42,43}). As a consequence, improved mapping of R-loop structures represents an exciting and essential challenge to better decipher the distribution and function of these structures in health and disease.

DNA:RNA immunoprecipitation (DRIP) relies on high affinity of the S9.6 monoclonal antibody for DNA:RNA hybrids⁴⁴. DRIP-seq permits robust genome-wide profiling of R-loop formation^4,45. While useful, this technique suffers from limited resolution due to the fact that restriction enzymes are used to achieve gentle DNA fragmentation. In addition, DRIP-seq does not provide information on the directionality of R-loop formation. Here, we report a variant of DRIP-seq that permits the mapping of R-loops at high resolution in a strand-specific manner. This method relies on sonication to fragment the genome prior to immunoprecipitation and the method is thus called sDRIP-seq (sonication DNA:RNA immunoprecipitation coupled with high throughput sequencing) (Figure 1). The use of sonication permits an increased resolution and limits restriction enzyme-linked fragmentation biases observed in DRIP-seq approaches⁴⁶. sDRIP-seq produces R-loop maps that are in strong agreement with the results from both DRIP-seq and the previously described high-resolution DRIPc-seq method in which sequencing libraries are built from the RNA strands of immunoprecipitated R-loop structures⁴⁵.

Faced with a plethora of methods to choose from, users may wonder which particular DRIP-based approach is preferable for their needs. We offer the following advice. DRIP-seq, despite its limitations, is technically easiest and is the most robust (highest yields) of all three methods discussed here; it thus remains broadly useful. Numerous DRIP-seq datasets have been published, which provide a useful comparison point for new datasets. Finally, the bioinformatic analysis pipeline is simpler as the data is not stranded. It is recommended that new users begin honing their R-loop mapping skills with DRIP followed by quantitative polymerase chain reaction (qPCR) and DRIP-seq. sDRIP-seq represents a slightly higher degree of technical difficulty: the yields are slightly reduced due to sonication (discussed below) and the sequencing library process is slightly more complex. Yet, the gain of strandedness and higher resolution is invaluable. It is noted that sDRIP-seq will capture both two-stranded RNA:DNA hybrids and three-stranded R-loops. Due to the library construction steps, DRIP-seq will not capture two-stranded RNA:DNA hybrids. DRIPc-seq is the most technically demanding and requires higher amount of starting materials. In return, it offers the highest resolution and strandedness. Because sequencing libraries are built from the RNA moiety of R-loops or hybrids, DRIPc-seq may suffer from possible RNA contamination, especially since S9.6 possesses residual affinity for dsRNA^19,47,48. sDRIP-seq permits strand-specific, high resolution mapping without worries about RNA contamination since sequencing libraries are derived from DNA strands. Overall, these three methods remain useful and present differing degrees of complexity and slightly different caveats. All three, however, produce highly congruent datasets⁴⁸ and are highly sensitive to RNase H pre-treatment, which represents an essential control to ensure signal specificity^45,49. It is noted that given the size selection imposed on sequencing libraries, small hybrids (estimated <75 bp), such as those forming transiently around lagging strand DNA replication priming sites (Okazaki primers) will be excluded. Similarly, since all DRIP methods involve DNA fragmentation, unstable R-loops that require negative DNA supercoiling for their stability will be lost⁵. Thus, DRIP approaches may underestimate R-loop loads, especially for short, unstable R-loops that may be best captured using in vivo approaches^45,48. It is noted that R-loops can also be profiled in an S9.6-independent manner at deep coverage, high-resolution, and in a strand-specific manner on single DNA molecules after sodium bisulfite treatment¹². Additionally, strategies using a catalytically inactive RNase H1 enzyme have been employed to map native R-loops in vivo, highlighting short, unstable R-loops that form primarily at paused promoters^50,51,52.