Background

Overview

R-loops are triple-stranded nucleic acid structures that form when an RNA hybridizes with a complementary ssDNA, leading to displacement of the second DNA strand. R-loops were first described as a by-product of transcription and a source of genomic instability that needed to be resolved. However, research over the last decade has shown that R-loops can be associated with specific histone modifications and transcriptional status, induce the targeting of proteins, or act as promoters (Chedin and Benham, 2020; Niehrs and Luke, 2020). To study R-loop function, protocols have been developed to map these structures in multiple organisms. Differences have been observed between species: R-loop-forming sequences in mammals have a GC skew; those in Saccharomyces cerevisiae have an AT skew; and in plants, both GC and AT skews have been reported (Sanz et al., 2016; Wahba et al., 2016; Xu et al., 2017; Hartono et al., 2018). Differences in the properties and localization of R-loop-forming sequences between species and the association of R-loops with different diseases and gene misexpression highlight the importance of developing standard R-loop mapping protocols to perform rigorous comparisons.

R-loop maps in plants have shown that some R-loops can form differentially during development (Fang et al., 2019; Xu et al., 2020) and may potentially be involved in changes in transcriptional status. In mammals, active enhancers have been associated with R-loop formation, and these R-loops can act as promoters (Tan-Wong et al., 2019). The study of R-loops in the context of development in intact organisms with available genetic tools to evaluate their formation, resolution, and function, will allow a better understanding of R-loop biology.

Here, we provide a detailed strand-specific DRIP-seq protocol to identify R-loop-forming sequences in Drosophila melanogaster embryos and tissue culture cells (Figure 1). Similar to other DRIP protocols, our protocol relies on the S9.6 antibody to identify R-loop-forming sequences but also considers that the weak affinity of the S9.6 antibody for dsRNA can generate false positive signals (Phillips et al., 2013; Hartono et al., 2018; König et al., 2017). This single-strand DRIP-seq protocol in association with the genetic tools available for Drosophila melanogaster should be a powerful system to study R-loop function in the whole organism, during development, and in different tissues (Figure 2, Table 1). DRIP-seq can be combined with genetic manipulation to overexpress or knockdown genes with a view to identifying their effect on R-loop formation, or with cell synchronization to follow R-loop formation during the cell cycle. Finally, the use of sonication rather than restriction enzymes to fragment the genomic DNA means that this protocol could potentially be adapted to other organisms with minimal optimization.

Limitations

The protocol relies on the S9.6 antibody to immunoprecipitate RNA–DNA hybrid-containing fragments. This antibody is not a perfect tool: it shows differential binding affinity based on the sequence of the R-loop with no correlation with the GC content and a preference for longer RNA–DNA hybrids. It can bind dsRNA, albeit with an affinity 5-times lower than that for RNA–DNA hybrids (Phillips et al., 2013; König et al., 2017). The latter two limitations can be overcome during the preparation of nucleic acids. After nucleic acid extraction, ssRNA and dsRNA are removed by incubating the nucleic acids with RNase A in the presence of 0.5 M NaCl to avoid the degradation of RNA in the RNA–DNA hybrids. This first digestion step is followed by an incubation with RNase III to degrade any dsRNA that were not degraded by RNase A. By using two different RNases before immunoprecipitation, we removed all ssRNA and dsRNA but also potentially some R-loops, which can be sensitive to these RNases. Thus, these steps make the experiment more stringent and limit false positives but could be modified to balance stringency with sensitivity, since some R-loops may be sensitive to these treatments. Removal of RNA that can interact with S9.6 is most relevant when the RNA component of the R-loop is to be sequenced; although, we have observed that the RNA–DNA hybrid pulldown efficiency is increased when stringent RNase treatment is used. It is also essential to have an RNase H-treated negative control to evaluate the specificity of the pulldown (Figure 3). We found that commercially available E. coli RNase H does not consistently completely digest RNA–DNA hybrids in nucleic acid preparations. This can be solved by producing and using highly active human RNase H1 and RNase H2. This negative control is especially important during optimization of the DRIP protocol; it ensures that the technique is specific. It is also used to call peaks when DRIP-seq is performed (Figures 2, 4, and 5).

To avoid bias toward longer fragments during the immunoprecipitation step, we fragment the nucleic acids by sonication (Figure 4A). Sonication allows us to obtain a homogenous population of fragments with an average size of 300 bp. Sonication is performed after RNase A treatment to avoid the possibility of generating new RNA–DNA hybrids during the DRIP procedure. Sonication is reported to reduce the recovery of some RNA–DNA hybrids as compared with restriction digestion (Crossley et al., 2020).

Advantages

Advantages compared with other DRIP protocols

This protocol uses a gentle lysis step to extract the nucleic acids. Cell lysis can be performed on tissue culture cells, whole embryos, or dissected tissues from larval, pupal, or adult Drosophila melanogaster. Although the ~200 mg Drosophila embryos needed to perform one DRIP-seq experiment is relatively high, the possibility to freeze tissues and pool them for a single nucleic acid extraction makes it feasible. By dissecting Drosophila melanogaster and performing DRIP on discs, organs, or sorted cell populations, it should be possible to identify cell- or tissue-specific R-loops.

This DRIP protocol, contrary to several protocols developed for mammalian cells (e.g., Sanz et al., 2016), does not use a restriction enzyme cocktail to fragment the nucleic acids. Instead, sonication is used, which leads to the fragmentation of nucleic acids at an average size of 300 bp (Figure 4A). These fragments have a homogenous size and the sonication is highly reproducible. An increase in resolution by using sonication, as compared with restriction digestion, in mammalian cells has recently been demonstrated (Crossley et al., 2020). By using sonication, the protocol can easily be adapted to other organisms with little optimization; the only step that may require optimization is lysis. Sonication has another advantage: it leads to disruption of the displaced single-strand DNA of the R-loop, which makes it possible to prepare strand-specific sequencing libraries using either the DNA or the RNA moiety of the RNA–DNA hybrid (Wahba et al., 2016).

Advantages compared with other methods

Another method to identify R-loop-forming sequences relies on a catalytically inactive form of the RNase H1 enzyme (dRNase H1) (Ginno et al., 2012; Chen et al., 2017). The use of dRNase H1 in cells presents several potential problems, which may explain the differences in the identification of R-loop-forming sequences with this method versus the S9.6 antibody. Firstly, RNase H1 is not the only regulator of R-loops in cells; it has been suggested that topoisomerases are the main enzymes responsible for the resolution of R-loops that form co-transcriptionally in human cells (Manzo et al., 2018; Zhang et al., 2019), while RNase H1 and H2 target R-loops once they are formed. Secondly, a recent article by Lockhart et al., 2019 demonstrated that RNase H1 is activated upon stress, while RNase H2 displays the main activity under physiological conditions in S. cerevisiae. Thirdly, with dRNase H1, R-loop identification may be limited to R-loops that are accessible, protein-free, and normally degraded by RNase H enzymes. Fourthly, expression of dRNase H1 may stabilize R-loops. While this may allow detection of transient R-loops, it could also skew interpretation of where stable R-loops exist. Finally, RNase H1 has two RNA–DNA hybrid binding domains, both of which need to bind to induce degradation of the RNA moiety of the hybrid (Nowotny et al., 2008). This may prevent or limit the detection of smaller RNA–DNA hybrids. Thus, while dRNase H1 may be a useful tool in some contexts, the interpretation of DRIP results may be more straightforward. Comparison of results from both methods could also yield complementary information.

Similar to DRIP, dRNase H1 has been used to isolate RNA–DNA hybrids after extraction of nucleic acids from tissue culture cells; however, bias of the enzyme toward longer hybrids and its weaker affinity as compared with the S9.6 antibody make it less efficient (Ginno et al., 2012).

Finally, native bisulfite sequencing is an alternative high-resolution method for identifying R-loops. Bisulfite converts cytosine to uracil in single-stranded DNA (Yu et al., 2003); thus, this method does not detect RNA–DNA hybrids but the ssDNA strand that is displaced when they form. However, the presence of methyl-cytosine in the genome blocks modification of the ssDNA, and ssDNA can be displaced by the formation of other non-canonical DNA structures such as G-quadruplexes and I-motifs, which could lead to false negative or false positive results, respectively.

The single-strand DRIP-seq protocol presented here has been optimized to identify R-loop-forming sequences in Drosophila melanogaster embryos and tissue culture cells, and can potentially be used in other organisms with minimal optimization. This protocol could be a standardized means to evaluate R-loops across developmental stages or in different tissues or organisms.

Figure 2. Examples of shared and distinct R-loops among S2 cells and early and later stage embryos. A-B. R-loops form over the CG1513, CG3008, and CG6040 genes in S2 cells and 2-6 H and 10-14 H embryos (red arrow). No R-loop is formed over the CG12923 gene. C. At the alt gene, three R-loops are detected in 2-6 H but not in 10-14 H embryos. Two of these R-loops (red arrows) are also present in S2 cells, while one of them is replaced by an R-loop on the opposite strand (purple arrow). D. R-loop formed over Spn1 is only present in 2-6 H embryos (purple arrows). E. R-loops form over scrt on the + strand in 2-6 H embryos and the - strand in 10-14 H embryos and in S2 cells. “Unstr” corresponds to unstranded DRIP-seq data, “+” and “-” indicate the strand-specific track, and RNase H-treated sample acts as a negative control. The red arrowhead indicates the orientation of the transcript. Note that the “+” and “-” strands refer to the DNA strand such that “+” strand R-loops should arise from transcription in the leftward direction. Thus, in A, the R-loop in the CG3008 gene is in the expected orientation to have arisen from gene transcription, while that in the gene CG1513 would be derived from antisense transcription.

Table 1. Data availability