Harmful DNA:RNA hybrids are formed in cis and in a Rad51-independent manner

DNA:RNA hybrids constitute a well-known source of recombinogenic DNA damage. The current literature is in agreement with DNA:RNA hybrids being produced co-transcriptionally by the invasion of the nascent RNA molecule produced in cis with its DNA template. However, it has also been suggested that recombinogenic DNA:RNA hybrids could be facilitated by the invasion of RNA molecules produced in trans in a Rad51-mediated reaction. Here, we tested the possibility that such DNA:RNA hybrids constitute a source of recombinogenic DNA damage taking advantage of Rad51-independent single-strand annealing (SSA) assays in the yeast Saccharomyces cerevisiae. For this, we used new constructs designed to induce expression of mRNA transcripts in trans with respect to the SSA system. We show that unscheduled and recombinogenic DNA:RNA hybrids that trigger the SSA event are formed in cis during transcription and in a Rad51-independent manner. We found no evidence that such hybrids form in trans and in a Rad51-dependent manner.


INTRODUCTION
R loops are structures formed by a DNA:RNA hybrid and the complementary displaced single stranded DNA (ssDNA). They were observed naturally as programmed events in specific genomic sites such as the S regions of Immunoglobulin genes in mammals or mitochondrial DNA (Chang, Hauswirth, & Clayton, 1985;Garcia-Muse & Aguilera, 2019;Yu, Chedin, Hsieh, Wilson, & Lieber, 2003), where they play specific functions by promoting class switch recombination or DNA replication, respectively; but also as unscheduled nonprogrammed structures upon dysfunction of RNA binding proteins involved in the assembly or processing and export of the protein-mRNA particle (mRNP) such as the THO complex or the SRSF1 splicing factor (Huertas & Aguilera, 2003;X. Li & Manley, 2005). Also, they have been inferred in the rDNA regions of the bacterial chromosome upon Topo1 inactivation (Drolet et al., 1995).
Accumulated evidence indicates that R loops are detected from yeast to humans in many transcribed regions of the eukaryotic genome in wild-type cells as well as in cells defective in several metabolic processes covering from RNA processing to DNA replication and repair or cells deficient in specific chromatin factors (Bhatia et al., 2014;Garcia-Muse & Aguilera, 2019;Garcia-Rubio et al., 2015;Herrera-Moyano, Mergui, Garcia-Rubio, Barroso, & Aguilera, 2014;Mischo et al., 2011;Paulsen et al., 2009;Schwab et al., 2015). The biological consequences of such R loop structures are diverse and include replication stress, DNA breaks and genome instability that can be detected as hyperrecombination, plasmid loss or gross chromosomal rearrangements  Aguilera, 2019). This is consistent with the idea that it is the RNA produced in cis who invades the duplex DNA, a reaction that can be facilitated by increasing the negative supercoiling of DNA as well as by nicking the DNA template (Roy, Zhang, Lu, Hsieh, & Lieber, 2010). The evidence of DNA:RNA hybrid formation at breaks has matured in the last years (Cohen et al., 2018;D'Alessandro et al., 2018;L. Li et al., 2016;Ohle et al., 2016;Teng et al., 2018;Yasuhara et al., 2018) although the source and role of such hybrids remains still controversial Puget, Miller, & Legube, 2019). Of note, genome-wide mapping results have been interpreted in diverse manners by different labs.
Whereas some claim that DNA:RNA hybrids detected around DNA breaks mostly accumulate at transcribing sites (Cohen et al., 2018), in agreement with their co-transcriptional formation, others suggest that there is no preference for DNA:RNA hybrids to form at transcribed loci (D'Alessandro et al., 2018), implying an scenario in which DNA:RNA hybrids at breaks sites would form either de novo or with RNAs produced in trans.
DNA:RNA hybrids may be formed in trans as intermediates in the course of ribonucleoprotein-mediated reactions such as telomerase and CRISPR-Cas9 ribonucleoprotein involved in specific reactions (Collins, 2000;Jinek et al., 2012). DNA:RNA hybrids can also be formed in vitro with the aid of the bacterial strand exchange protein RecA (Kasahara, Clikeman, Bates, & Kogoma, 2000;Zaitsev & Kowalczykowski, 2000). They have also been reported to have regulatory roles in gene expression when formed by long non-coding RNAs (lncRNAs) at in trans loci such as the cases of the GAL lncRNA in yeast (Cloutier et al., 2016) or the APOLO lncRNA in plants (Ariel et al., 2020). In summary, despite the accumulating evidence that in vivo DNA:RNA hybrids formed in cis constitute a threat for genome stability, an open question is whether DNA:RNA hybrids also form in trans as a potential source of recombinogenic DNA damage. To our knowledge, this has only been addressed in the yeast Saccharomyces cerevisiae (Wahba, Gore, & Koshland, 2013). By S9.6 immunofluorescence (IF) and a yeast artificial chromosomebased genetic assay that measures gross chromosomal rearrangements, it was inferred that DNA:RNA hybrids could be formed in trans by a reaction catalyzed by the eukaryotic strand-exchange protein Rad51 (Wahba et al., 2013).
Nevertheless, the fact that the detected gross chromosomal rearrangements could depend on Rad51 and that the S9.6 antibody can also recognize dsRNAs (Hartono et al., 2018;Konig, Schubert, & Langst, 2017;Silva, Camino, & Aguilera, 2018), prompted us to address this question using a different approach. Using Rad51-independent recombination assays in which the initiation region could be unambiguously delimited, we provide genetic evidence that DNA:RNA hybrids compromising genome integrity are formed in cis and in a Rad51-independent manner.

A new genetic assay to detect recombinogenic DNA:RNA hybrids in trans
We developed a new genetic assay to infer the formation of recombinogenic DNA:RNA hybrids in trans. It is based on two plasmids, one containing the recombination system and the LacZ gene in cis (GL-LacZ recombination system), and another one providing the in trans LacZ transcripts (tet p :LacZ) ( Figure 1). The bacterial LacZ gene consists of a 3-Kb sequence with high G+C content previously reported to be hyper-recombinant and difficult to transcribe in DNA:RNA hybrid-accumulating strains, such as tho mutants (Chavez, Garcia-Rubio, Prado, & Aguilera, 2001).
The GL-LacZ recombination system is a leu2 direct-repeat construct carrying the LacZ gene in between and under the GAL1 inducible promoter so that this construct is transcribed as a single RNA unit driven from the GAL1 promoter (Piruat & Aguilera, 1998). Single-Strand Annealing (SSA) events cause the deletion of the LacZ sequence and one of the leu2 repeats leading to Leu+ recombinants in a Rad51-independent manner ( Figure 1A). To provide LacZ transcripts in trans, we did a fusion construct containing the complete bacterial LacZ gene sequence under the doxycycline-inducible tet promoter (tet p :LacZ). As a control of no expression in trans, we used transformants with an empty plasmid to avoid any possible effect from leaky transcription from the tet promoter in the presence of doxycycline.
Yeast strains carrying both GL-LacZ recombination system and the tet p :LacZ construct were used to assay SSA annealing events in the four different possible conditions: i) no transcription, with GL-LacZ construct turned transcriptionally off (2% glucose) and an empty plasmid; ii) transcription in trans, with GL-LacZ construct turned transcriptionally off (2% glucose) and the tet p :LacZ construct; iii) transcription in cis, with GL-LacZ construct turned transcriptionally on (2% galactose) and an empty plasmid; and iv) transcription in cis and in trans, with GL-LacZ construct turned transcriptionally on (2% galactose) and the tet p :LacZ construct ( Figure 1B).

RNA is not a spontaneous source of recombinogenic DNA damage in trans
The analysis of recombination in wild-type cells revealed that whereas transcription in cis elevated the frequency of recombination threefold, transcription in trans driven from the tet p :LacZ construct had no effect on recombination ( Figure 2A). These results already suggest that homologous transcripts coming from a different locus do not represent a detectable source of genetic instability in wild-type conditions and thus argue against the hypothesis that spontaneous DNA:RNA hybrids could be formed with mRNAs transcribed in trans. However, it is known that mRNA coating protects DNA from cotranscriptional RNA hybridization. Thus, we wondered if in trans transcripts could induce recombination in mRNP-defective mutants such as those of the THO complex. Hence, we performed our experiments in mft1∆ and hpr1∆ mutant strains. mft1∆ and hpr1∆ enhanced recombination slightly when transcription in cis was switched off, likely as a consequence of leaky transcription form the GAL1 promoter in glucose. More significantly and in agreement with previous reports (Chavez et al., 2000), recombination frequencies rocketed when transcription was stimulated in cis. However, transcription activation in trans did not enhance recombination, as it would be expected if additional DNA:RNA hybrids could form with RNA produced in trans.
Instead, under conditions of high transcription in cis, transcription in trans led to a partial suppression of the hyper-recombination observed by only in cis transcription. The reason for such a suppression might involve the potential ability of the mRNA produced in trans to interfere with transcription occurring in cis at the GL-LacZ construct. Given that a DNA:RNA hybrid produced in the template DNA strand can impair transcription elongation (Tous & Aguilera, 2007), one possibility was that this interference is mediated by DNA:RNA hybrids formed in trans with the transcribed strand of the GL-LacZ construct.
However, this possibility was ruled out by the fact that transcription in trans also led to a reduction of the hyper-recombination when we used an alternative recombination system (GL-LacZi), in which the LacZ sequence was inverted so that the LacZ transcript produced in trans would not be able to anneal with the transcribed strand of the GL-LacZi system ( Figure 2B). Furthermore, in this case the suppression was stronger and was also observed in glucose, when transcription in cis was off. This could be explained because, in this scenario, the RNA produced in trans is complementary to the mRNA produced in cis.
Consequently, they can hybridize together forming a dsRNA that would preclude the possibility to form DNA:RNA hybrids at the GL-LacZi construct.
RNAses H efficiently degrade the RNA moiety of DNA:RNA hybrids.
Thus, to favor the potential DNA:RNA hybrid accumulation in trans we used cells lacking also both RNAses H1 and H2 and we determined the impact on recombination. Figure 2A shows that rnh1∆ rnh201∆ cells elevated the recombination frequency when transcription was stimulated in cis. Importantly, the recombination frequencies were not altered by producing transcripts in trans, arguing again against the recombinogenic potential of putative DNA:RNA hybrids formed in trans.
Since transcription from the long LacZ gene is inefficient and leads to unstable RNA products, particularly in tho mutants (Chavez et al., 2001), we made a new construct with only the last 400 bp of LacZ (tet p :LacZ400) ( Figure   2C). In this case, recombination frequencies were not significantly affected by in trans transcription in any of the strains or conditions tested further arguing against mRNA produced in trans as a possible source of recombinogenic DNA:RNA hybrids.
Finally, to confirm that the same results were obtained in chromosome loci and not only in plasmid-born DNA sequences, we integrated the GL-LacZ system in the chromosome. As shown in Figure 3, we observed again that mRNA production in trans had no effect on recombination, neither in wild-type cells nor in the tho mutant hpr1∆. Furthermore, the results were similar after RNase H overexpression. Hence, altogether, these results argue that, in contrast to mRNA produced in cis, RNA produced at a particular locus does not lead to recombinogenic DNA damage at regions located in trans.

Rad51 is not required for DNA:RNA hybridization
We next wondered about the possible role of the recombination protein Rad51 in DNA:RNA hybridization. To examine this, in particular in relation to hybrids potentially formed in trans, we analyzed recombination frequencies in our direct-repeat assays in hpr1∆ cells in which in cis transcription was switched off, under conditions in which an homologous RNA with the potential to hybridize with the intervening sequence of repeat construct was or was not generated in trans ( Figure 4A). It is important to remark that the recombination events detected in our assays are deletions occurring by SSA between direct repeats, which do not require Rad51 (Pardo, Gomez-Gonzalez, & Aguilera, 2009).
Indeed, in agreement with SSA annealing being Rad51-independent, RAD51 deletion caused no significant changes in the recombination frequencies in our assay. Thus any conclusion about Rad51-dependency or independency of the hybridization inferred from our assay is not contaminated by a possible direct role of Rad51 in the event we are studying. Importantly, we observed no differences when RAD51 was deleted in hpr1∆ cells even when the LacZ sequence was expressed in trans. This result argues against Rad51 facilitating or impeding the formation of DNA:RNA hybrids in trans.
We then wondered whether the formation of known recombinogenic DNA:RNA hybrids formed in cis, such as those reported in the hpr1∆ mutant, requires Rad51. For this purpose, we deleted RAD51 and studied the effect in the strong hyper-recombination phenotype of hpr1∆, which has been observed in multiple direct-repeat systems in which recombinants also arise by Rad51independent SSA, such as the LY∆NS system consisting in two truncated leu2 direct-repeats under the LEU2 promoter with a 3.7-Kb sequence in between (Prado, Piruat, & Aguilera, 1997). As shown in Figure S1, hpr1∆ led to a 33-fold increase in recombination in the absence of Rad51, similarly to what was previously reported in the presence of Rad51 (Gomez-Gonzalez, Felipe-Abrio, Prado et al., 1997). This result clearly indicates that the in cis DNA-RNA hybrid-mediated hyper-recombination phenotype is actually independent on Rad51.
In parallel, we studied the formation of Rad52 foci, a marker of recombinogenic DNA breaks (Lisby, Rothstein, & Mortensen, 2001) as well as the effect of AID overexpression to enhance the recombinogenic potential of R loops (Gomez-Gonzalez & Aguilera, 2007) and RNase H overexpression to remove DNA:RNA hybrids ( Figure 5A). In agreement with the role of the THO complex in R loop prevention, hpr1∆ caused an increase in Rad52 foci that was enhanced by AID overexpression and suppressed by RNase H overexpression, as previously reported (Alvaro, Lisby, & Rothstein, 2007;Garcia-Pichardo et al., 2017;Wellinger, Prado, & Aguilera, 2006). By contrast, the accumulation of Rad52 foci observed in rad51∆ cells was not affected by either AID or RNase H overexpression. This result argues that R loops are not the cause for the genetic instability observed in the absence of Rad51. Given the role of Rad51 in recombination upstream of Rad52, the accumulation of Rad52 foci in rad51∆ cells is rather likely due to the accumulation of unrepaired recombination intermediates, as previously suggested (Alvaro et al., 2007). Importantly, hpr1∆ rad51∆ cells showed a similar result, further supporting that the accumulation of recombinogenic damage in hpr1∆ is independent on Rad51. We next directly measured DNA:RNA hybrid accumulation by immunodetection with the S9.6 antibody on metaphase spreads. Figure 5B illustrates that the number of cells with S9.6 positive signal was similar in hpr1∆ and in hpr1∆ rad51∆ cells.
Altogether, these results demonstrate that the Rad51 protein is not required for the DNA:RNA hybrid formation previously reported in THO mutants.

DISCUSSION
We have devised a new genetic assay to infer whether the source of DNA:RNA hybrids compromising genome integrity could potentially come from RNAs produced in trans. To reach this conclusion, we used an SSA assay. It is well established that SSA events are Rad51-independent; they do not require strand-exchange, but just annealing between resected single-stranded DNA (ssDNA) for which the action of Rad52 is sufficient ( Figure 1A) (Pardo et al., 2009). Our constructs show that, in contrast to the RNA produced in cis at the site where SSA occurs, an RNA produced in trans does not induce an increase in recombination. Importantly, recombination is not induced by in trans RNA production even when the major DNA:RNA removal machinery is absent (rnh1∆ rnh201∆ mutant) or when the RNA coating functions preventing DNA:RNA hybrid formation are impaired (tho mutants), arguing against the idea that harmful DNA:RNA hybrids could spontaneously form in trans. Putative DNA:RNA hybrids formed in trans would be expected to further increase recombination levels. Instead, the simultaneous induction of transcription in cis and in trans (Figure 2A) reduced the strong hyper-recombinogenic effect of tho mutants. The fact that this suppressor effect was augmented when one of the LacZ sequences was inverted ( Figure 2B) and prevented by a shorter LacZ construct ( Figure 2C), which was reported to be more stable in tho mutant backgrounds (Chavez et al., 2001), suggests that the free RNA itself, and not in the form of DNA:RNA hybrids formed at the template DNA strand, play some role in preventing the hyper-recombination, likely because stable RNAs can interfere with transcription at an homologous locus.
DNA:RNA hybrids formed by an RNA produced in trans were previously suggested to threaten genome integrity based on the results obtained with a yeast artificial chromosome and an homologous region placed at chromosome III, which transcription was inducible (Wahba et al., 2013). Recombination involving multiple substrates was first reported in S. cerevisiae, in which an induced-DSB triggered recombination between two other homologous fragments at different chromosomes (Ray, Machin, & Stahl, 1989). Tri-parental recombination assays have been successfully used since then to define specific features of the HR reaction as well as for studies of Break-Induced Recombination (BIR) or translocations and chromosomal rearrangements occurring between ectopic regions (Pardo & Aguilera, 2012;Piazza, Wright, & Heyer, 2017;Ruiz, Gomez-Gonzalez, & Aguilera, 2009). However, such events are not the most adequate to infer recombination initiation unless this has been artificially designed (as is the case of an HO-induced DSB). Hence, the assay used to infer the potential of trans DNA:RNA hybrids to induce genetic instability (Wahba et al., 2013) relied on an RNA fragment produced at a (first) DNA region that could form a DNA:RNA hybrid with a (second) ectopic homologous DNA region that would promote its deletion or loss, leading to a genetically detectable phenotype. Thus, this assay does not exclude the possibility that the RNA forms the hybrid in cis inducing subsequently a DSB that would stimulate the recombination events studied ( Figure 6). Indeed, this event would demand the action of Rad51 for DNA strand invasion, consistent with the results obtained (Wahba et al., 2013). Therefore, rather than implying that Rad51 is required for the RNA to invade in trans the second DNA sequence, the increased genetic instability obtained suggests that the 3' end of the DNA break induced by the DNA:RNA hybrid formed at the first site who makes the invasion ( Figure 6).
Our assays involve two leu2 homologous repeats that recombine by Rad51-independent SSA. Indeed, as expected, RAD51 deletion caused no decrease in the observed recombination frequencies in our assay (Figure 4).
Recombination between the leu2 repeats could be originated by either a DNA:RNA hybrid in cis or by a DSB occurring in between the repeats, or as suggested previously for tho mutants, by a bypass mechanism involving template switching ). In our case, however, we show that the hyper-recombinogenic potential of DNA:RNA hybrids is Rad51independent ( Figure 4).
Similarly, a DSB occurring at the locus where the RNA in trans was generated could give rise to Leu+ recombinants in our assay. However, such recombination events would be Rad51-dependent, as they will require a Rad51dependent invasion into the GL-LacZ construct ( Figure 6). Hence, the Leu+ recombinants obtained in rad51∆ mutant cells (Figure 4) can only be explained by Rad51-independent events occurring in cis, at the GL-LacZ construct.
Strikingly, the fact that we detected no significant increase in Leu+ recombinants by inducing transcription in trans, either in RAD51 or rad51∆ backgrounds rules out the possibility that recombinogenic DNA:RNA hybrids form in trans in our assay. It was previously shown that S9.6 signal detected by IF was reduced by rad51∆ in metaphase spreads (Wahba et al., 2013). By contrast, we detected S9.6 signal in metaphase spreads of the hpr1∆ mutant of the THO complex in both RAD51 and rad51∆ backgrounds ( Figure 5). The uncertainty about the identity of the structures detected by IF using the S9.6 antibody, which also recognizes dsRNA (Hartono et al., 2018;Konig et al., 2017;Silva et al., 2018), and the possibility that chromosomal spreads could preferentially visualize the rDNA regions, in which high levels of dsRNA structures formed by the rRNAs, makes difficult to make conclusions on S9.6 IFs in this case.
Thus, we have found no evidence for a Rad51-facilitated strand invasion from RNAs produced in trans. Further arguing against any major role of this recombinase in R loop metabolism or function, none of the so far reported DNA:RNA hybrid interactomes has identified RAD51 (Cristini, Groh, Kristiansen, & Gromak, 2018;Nadel et al., 2015;Wang et al., 2018). The fact that, in vitro, RecA can catalyze an inverse strand exchange reaction with DNA or RNA thus promoting the assimilation of a transcript into duplex DNA (Kasahara et al., 2000;Zaitsev & Kowalczykowski, 2000) does not argue that this is the case for unscheduled recombinogenic R loops in vivo. More likely, the biological significance of this process relies on its use for replication initiation of prokaryotic cells as originally proposed (Zaitsev & Kowalczykowski, 2000), for replication-dependent recombination to restart stalled forks (Pomerantz & O'Donnell, 2008) or even for transcription-induced origin-independent replication (Stuckey, Garcia-Rodriguez, Aguilera, & Wellinger, 2015). Hence, DNA:RNA hybridization could occur in trans under regulated conditions but not spontaneously as unscheduled and harmful structures that would put genome integrity into risk. Thus, the assimilation of a transcript into a duplex DNA in trans would be tightly regulated and limited to specific reactions such as the case of telomerase or CRISPR and possibly other proteins yet to be discovered. For other cases, such as that of the GADP45 factor that binds to promoters harboring hybrids formed by lncRNAs (Arab et al., 2019), it is unclear whether such hybrids are formed in trans and in a GADP45-dependent manner.
Altogether, our results suggest that RNAs do not form hybrids in trans, so that the previously reported induction of Rad51-dependent ectopic genetic instability would be explained by R loop-mediated DNA breaks in cis.

Recombination assays
Recombination frequencies were calculated as previously described (Gomez-Gonzalez, Ruiz, & Aguilera, 2011) as means of at least 3 median frequencies obtained each from 6 independent colonies isolated in the appropriate SC medium for the selection of the required plasmids. Recombinants were obtained by platting appropriate dilutions in selective medium. To calculate total number of cells, plates with the same requirements as for the original transformation were used. All plates were grown for 3-4 days at 30ºC.

Detection of Rad52-YFP foci
Spontaneous Rad52-YFP foci from mid-log growing cells carrying plasmid pWJ1344 were visualized and counted by fluorescence microscopy in a Leica DC 350F microscope, as previously described (Lisby et al., 2001). More than 200 S/G2 cells where inspected for each experimental replica.

S9.6 immunofluorescence of yeast chromosome spreads
The procedure performed is similar to (Chan et al., 2014)  rpm, lysed with 1% vol/vol Lipsol and fixed on slides using Fixative solution (4% paraformaldehide/3.4% sucrose). The spreading was carried out using a glass rod and the slides were dried from 2 hours to overnight in the extraction hood.
For the immuno-staining, the slides were first washed in PBS 1X in coplin jars and then blocked in blocking buffer (5% BSA, 0.2% milk in PBS 1X) over 10 minutes in humid chambers. Afterwards, slides were incubated with the primary monoclonal antibody S9.6 (1 mg/ml) in a humid chamber 1 hour at 23ºC. After washing the slides with PBS 1X for 10 minutes, the slides were incubated 1 hour at 23ºC in the dark with the secondary antibody Cy3 conjugated goat antimouse (Jackson laboratories, #115-165-003) diluted 1:1000 in blocking buffer.
Finally, the slides were mounted with 50 µl of Vectashield (Vector laboratories, CA) with 1X DAPI and sealed with nail polish. More than 300 nuclei were visualized and counted to obtain the fraction of nuclei with DNA:RNA hybrids.