Main

To investigate the capacity of transcript RNA to recombine with genomic DNA, we sought to discover whether a chromosomal DSB could be repaired directly by endogenous RNA in yeast S. cerevisiae cells. We designed a strategy by which we could induce a DSB in the HIS3 marker gene and monitor precise repair of the DSB by a homologous transcript messenger RNA by restoration of HIS3 function resulting in histidine prototrophic (His+) cells (see Methods). We developed two experimental yeast cell systems, trans and cis, in strains YS-289, 290 and YS-291, 292, respectively (Extended Data Table 1). The trans system is designed to test the ability of a spliced (intron-less) antisense his3 transcript from chromosome III to repair a DSB in a different his3 allele on chromosome XV, which contains an engineered homothallic switching endonuclease cutting site (Fig. 1a and Extended Data Fig. 1a, b). The cis system is designed to test the capacity of the spliced antisense his3 transcript from chromosome III to repair a homothallic-switching-endonuclease-induced DSB located inside the intron of the same his3 locus (Fig. 1b and Extended Data Fig. 1c). In both the trans and cis cell systems, the spliced antisense his3 transcript RNA can serve as a homologous template to repair the broken his3 DNA and restore its function. However, given the abundance of Ty retrotransposons in yeast cells, the spliced antisense his3 RNA could potentially be reverse transcribed by the Ty reverse transcriptase in the cytoplasm to cDNA that could then recombine with the homologous broken his3 sequence or be captured by non-homologous end joining at the homothallic switching endonuclease break site to produce His+ cells6,7,8. To distinguish DSB repair mediated by the transcript RNA template from repair mediated by the cDNA template, we performed the trans and cis assays in two yeast strains that contained either a wild-type SPT3 gene or its null allele, which prevents Ty transcription and strongly reduces Ty transposition and transpositional recombination3,8,9. In both assays, cells containing wild-type SPT3 produced numerous His+ colonies after DSB induction (Fig. 1c and Table 1a). As expected, the frequency of His+ colonies in the trans system was significantly higher than that in the cis system because the his3 transcript is continuously generated in the presence of galactose. In contrast, production of the full his3 transcript is immediately terminated upon DSB formation in the cis system. This frequency difference is not specific to the particular genomic loci in which the DSBs are induced, as transformation by DNA oligonucleotides (HIS3.F and HIS3.R) designed to repair the broken his3 gene produced the same frequency of His+ colonies in the two systems (Extended Data Tables 2a and 3), demonstrating that the homothallic switching endonuclease DSB stimulates homologous recombination in the trans and cis systems equally well. Notably, almost all the His+ colonies are dependent on SPT3 function, indicating that the DSB in his3 is repaired exclusively via the cDNA pathway (Fig. 1c and Table 1a). This finding demonstrates that if an actively transcribed gene is broken, it can be repaired using a cDNA template derived from its intact transcript. Moreover, these data also support the model in which reverse-transcribed products from any sort of RNA can be a significant source of genome modification at DSB sites10.

Figure 1: Repair of a chromosomal DSB by transcript RNA.
figure 1

a, b, Scheme of the trans (a) and cis (b) cell systems used to detect DSB repair by transcript RNA. AI, artificial intron; HO, homothallic switching endonuclease; pGAL1, galactose-inducible promoter; RT, reverse transcriptase. Yellow triangles, cleavage activity by HO homothallic switching endonuclease; red question marks, hypothesis for transcript-RNA-templated DSB repair mechanism. c–e, Examples of replica-plating results (n = 6) from galactose medium to histidine dropout medium demonstrating the ability of various yeast strains (relevant genotypes shown) of the trans and cis systems to generate histidine prototrophic colonies in the absence of SPT3, or DBR1 function, or with phosphonoformic acid (PFA) (c), in the presence of the plasmid carrying the pGAL1-mhis3-AI cassette (BDG606) or the control (BDG283) (d), or when the artificial intron has a 23-base-pair deletion (AIΔ23) (e). WT, wild type.

PowerPoint slide

Full size image
Table 1 Frequencies of cDNA and transcript-RNA-templated DSB repair in trans and cis systems
Full size table

For RNA to recombine with DNA, an intermediate step that is probably required is the formation of an RNA–DNA heteroduplex. We therefore deleted the genes coding for ribonuclease (RNase) H1 (RNH1) and/or the catalytic subunit of RNase H2 (RNH201), which both cleave the RNA strand of RNA–DNA hybrids11. Remarkably, while deletion of RNH1 slightly increased the frequency of His+ colonies in the trans system, deletion of RNH201 increased the frequency of His+ colonies in both the trans and cis systems, and combined deletion of RNH1 and RNH201 resulted in an even stronger increase of His+ colonies in both systems. Moreover, we detected His+ colonies in rnh1 rnh201 cells in the absence of SPT3 (Fig. 1c and Table 1a). Notably, there were more His+ colonies in cis-system rnh1 rnh201 spt3 than in trans-system, and the frequency of His+ colonies observed in the rnh1 rnh201 spt3 relative to spt3 cells was much higher in cis (>69,000) than in trans (>6,400) (Fig. 1c and Table 1a). If DSB repair in rnh1 rnh201 spt3 cells were due to cDNA, we would expect a higher His+ frequency in the trans than in the cis system, as observed in wild-type cells. The fact that the His+ frequency is higher in the cis system suggests that DSB repair is not mediated by cDNA but instead by RNA or predominantly RNA. To further examine the possibility that residual cDNA rather than transcript RNA is responsible for his3 correction in cis-system rnh1 rnh201 spt3 cells, we introduced a trans system directly into these cells and into the control cis wild-type cells. When wild-type cells of the cis system were transformed with a low-copy-number plasmid carrying the pGAL1-mhis3-AI cassette, where AI represents an artificial intron (BDG606; see Methods), they displayed a large (a factor of 4,000) increase in the His+ frequency following DSB induction in his3 compared to the same cells transformed with the control empty vector (BDG283). In contrast, BDG606 in cis-system rnh1 rnh201 spt3 cells did not significantly increase the His+ frequency (Fig. 1d and Extended Data Table 4). These results argue against the role of residual cDNA in template-dependent DSB repair in cis-system rnh rnh201 spt3 cells and support a predominant, direct template function of the cis-system his3 transcript RNA in these cells. Overall, these data support the conclusion that a transcript RNA can directly repair a DSB in cis-system rnh1 rnh201 and rnh1 rnh201 spt3 cells. The physical proximity of the his3 transcript to its own his3 DNA during transcription could facilitate annealing of the broken DNA ends to the transcript. This possibility is consistent with the fact that closer donor sequences repair DSBs more efficiently12,13 and that mature transcript RNAs are exported rapidly to the cytoplasm or degraded after completion of transcription14.

To confirm that inactivation of RNases H1 and H2 allows for direct transcript RNA repair of a DSB in homologous DNA, we conducted a complementation test in the cis system using a vector expressing either a catalytically inactive mutant of RNH201, rnh201(D39A)15, or wild-type RNH201. Results showed that when wild-type RNH201 was expressed from the plasmid in rnh1 rnh201 spt3 cells, there were no His+ colonies following DSB induction (Extended Data Fig. 2a). Deletion of SPT3 is a well-established and robust method to suppress reverse transcription and formation of cDNA in yeast3,8,9. However, to prove that the increased frequency of His+ detected in the cis- relative to the trans-system rnh1 rnh201 spt3 background was not solely linked to SPT3 deletion, we impaired cDNA formation by deleting the DBR1 gene, which codes for the RNA debranching enzyme Dbr1 (refs 16, 17), or by using the reverse transcriptase inhibitor foscarnet (phosphonoformic acid)18. Results shown in Fig. 1c and Extended Data Table 5a support our conclusion that RNA transcripts can directly repair a DSB in chromosomal DNA without being first reverse transcribed into cDNA in rnh1 rnh201 cells.

Efficient generation of His+ colonies in cis wild-type, rnh1 rnh201, or rnh1 rnh201 spt3 cells requires transcription and splicing of the antisense his3 and DSB formation in the his3 gene. Deletion of pGAL1 (the galactose-inducible promoter) upstream of his3 on chromosome III, deletion of the homothallic switching endonuclease gene, or growing cells in glucose medium, in which homothallic switching endonuclease is repressed, drastically decreased His+ frequency (Extended Data Fig. 2b, c and Extended Data Table 5b, c). Similarly, yeast wild-type, rnh1 rnh201 and rnh1 rnh201 spt3 cells of the cis system containing a 23-base-pair truncation of the artificial intron in his3 lacking the 5′ splice site (Extended Data Table 1 and Extended Data Fig. 1c) produced no His+ colonies following DSB induction (Fig. 1e and Extended Data Table 5d), yet these cells were efficiently repaired by HIS3.F and HIS3.R synthetic oligonucleotides indicating that the DSB occurred in these cells (Extended Data Table 3).

Next, to examine whether DSB repair frequencies at the his3 locus in the trans and cis systems correlate with the expression level of antisense his3 transcript, we performed quantitative real-time PCR (qPCR). The qPCR data showed that with increased time of incubation in galactose medium (from 0.25 to 8 h) the trans strains had significantly more his3 RNA than the cis strains in all backgrounds, including the rnh1 rnh201 spt3 strain. Furthermore, the levels of his3 transcript dropped significantly from 0.25 to 8 h in galactose in cis but not in trans strains, except for the cis strain in which the homothallic switching endonuclease gene was deleted (Extended Data Fig. 2d). These results are expected in the cis strains because as soon as the homothallic switching endonuclease DSB is made, a full his3 transcript cannot be generated. Therefore, these data corroborate the conclusion that the higher frequency of His+ colonies obtained in cis- than in trans-system rnh1 rnh201 spt3 cells (Fig. 1c and Table 1a) is not due to more abundant and/or more stable transcript but rather to the proximity of the transcript to the target DNA.

PCR analysis of ten random His+ colonies from each of the trans- and the cis-system rnh1 rnh201 spt3 backgrounds, and Southern blot analysis of three samples from each background showed that the his3 locus that was originally disrupted by the homothallic switching endonuclease site (trans background), or by the intron with the homothallic switching endonuclease site (cis background), was indeed corrected to an intact HIS3 sequence. No integration of the HIS3 gene at the homothallic switching endonuclease site or elsewhere in the genome was detected in tested clones (20 of 20), excluding possible mechanisms of repair via capture of cDNA by end joining or via transposition (Fig. 2a and Extended Data Figs 3 and 4a–c). We also excluded the possibility that double deletion of RNH1 and RNH201 resulted in increased level of Ty transposition. In fact, results presented in Extended Data Table 6 show transposition rates a factor of 3–14 lower in null rnh1 rnh201 than in wild-type cells. This could be due to an increase of non-productive Ty RNA–DNA substrates for the Ty integrase, resulting in abortive integrations and/or titration of the enzyme. Sequence analysis of 24 random His+ colonies from the cis-system rnh1 rnh201 spt3 background revealed that all 24 clones had the same precise sequence as the spliced antisense his3 transcript and did not present a typical end joining pattern with small insertion, deletion or substitution mutations (Extended Data Fig. 1c and Extended Data Table 2b). These results, together with our observation of no His+ colony formation in cells unable to splice the intron in his3 (Fig. 1e and Extended Data Table 5d), strongly support a homologous recombination mechanism of DSB repair by transcript RNA in cis-system rnh1 rnh201 spt3 cells.

Figure 2: Transcript-templated DSB repair follows a homologous recombination mechanism.
figure 2

a, Southern blot analysis of yeast genomic DNA derived from trans wild-type His− (lane 2) or His+ (lane 3), rnh1 rnh201 spt3 His− (lane 4) or His+ (lanes 5–7) cells, digested with BamHI restriction enzyme and hybridized with the HIS3 probe, or derived from cis wild-type His− (lane 8) or His+ (lane 9), rnh1 rnh201 spt3 His− (lane 10) or His+ (lanes 11–13) cells, digested with NarI restriction enzyme and hybridized with the HIS3 probe (Extended Data Fig. 4a, c). Lanes 1 and 14, 1-kilobase DNA ladder visible in the ethidium-bromide-stained gel (Extended Data Fig. 4b). Size of digested DNA bands is indicated by red arrows. bp, base pairs. b, Experimental scheme of Rad52-promoted annealing between RNA and DNA in vitro. Asterisk denotes 32P label. ssDNA (named no. 211) or ssRNA (no. 501) oligonucleotides are in black; DNA oligonucleotides no. 508 and no. 509, forming double-stranded DNA (dsDNA), are in blue and green, respectively. Sequences of oligonucleotides no. 201, no. 501, no. 508 and no. 509 are shown in Extended Data Table 2a. c, d, The kinetics of annealing promoted by yeast Rad52 (c) and human RAD52 (d). Nucleoprotein complexes were assembled between dsDNA (no. 508 and no. 509) with an ssDNA protruding tail (0.4 nM) and either yeast or human Rad52 (1.35 nM) in the presence (dashed lines) or absence (solid lines) of yeast or human RPA (2 nM). Annealing was initiated by addition of 32P-labelled ssRNA or ssDNA (0.3 nM). The kinetics of protein-free annealing reactions are indicated by open squares and circles. The error bars represent the standard error of the mean, n = 4. For the significance of comparisons between the last two time points we used the two-tailed Mann–Whitney U-test. P values are given in Supplementary Table 1c.

PowerPoint slide

Full size image

Previous studies showed the ability of Escherichia coli RecA to promote pairing between duplex DNA and single-strand RNA in vitro19,20. Recent work suggests that Rad51 (the homologous protein to bacterial RecA) can promote formation of RNA–DNA hybrids in yeast21. Here we show that transcript-RNA-directed chromosomal DNA repair is stimulated by the function of Rad52 but not Rad51 recombination protein22. Rad52 is important for homologous recombination both via single-strand annealing and via strand invasion1,22. DSB repair by transcript RNA was reduced over 14-fold in cis-system rnh1 rnh201 spt3 rad52 but was increased by a factor of 4 in cis-system rnh1 rnh201 spt3 rad51 compared to rnh1 rnh201 spt3 cells (Table 1b). Notably, our in vitro experiments demonstrate that both yeast and human Rad52 efficiently promote annealing of RNA to a DSB-like DNA end (Fig. 2b–d and Extended Data Fig. 4d–h). Importantly, Rad52 catalyses the reaction with RNA at nearly the same rate as the reaction with single-stranded DNA (ssDNA) of the same sequence. Moreover, in our experiments replication protein A (RPA), a ubiquitous ssDNA binding protein1, caused a moderate inhibition of Rad52-promoted annealing between complementary ssDNA molecules, but not between ssRNA and ssDNA molecules. Thus, in the presence of RPA, the annealing between ssRNA and ssDNA proceeded with higher efficiency than the reaction between ssDNA molecules (Fig. 2b–d and Extended Data Fig. 4d–g).

In vivo, cDNA and/or RNA-dependent DSB repair may be especially important in the absence of functional Rad51 that prevents repair by the uncut sister chromatid via strand invasion23. Indeed, our results show that deletion of RAD51 increases the frequency of repair by cDNA and/or RNA (Table 1b). Hence, considering the bias observed for DSB repair in cis versus trans systems when Ty reverse transcription was impaired, we propose a model that in the absence of H-type RNase function, transcript RNA mediates DSB repair preferentially in cis systems via a Rad52-facilitated annealing mechanism. In this mechanism, the transcript may provide a template that either bridges broken DNA ends to facilitate precise re-ligation or initiate single-strand annealing via a reverse-transcriptase-dependent extension of the broken DNA ends (Fig. 3). The reverse transcriptase activity could be provided by a replicative DNA polymerase3, minimal Ty reverse transcriptase, or both. The current view in the field is that RNA–DNA hybrids formed by the annealing of transcript RNA with complementary chromosomal DNA either in cis or in trans systems are mainly a cause of DNA breaks, DNA damage and genome instability24. Here we demonstrate that under genotoxic stress, transcript RNA is recombinogenic and can efficiently and precisely template DNA repair in the absence of H-type RNase function in yeast. In the central dogma of molecular biology, the transfer of genetic information from RNA to DNA is considered to be a special condition, which has been restricted to retro-elements25 and telomeres26. Our data show that the transfer of genetic information from RNA to DNA occurs with an endogenous generic transcript (his3 antisense), and is thus a more general phenomenon than previously anticipated. In addition, in vitro RNA–DNA annealing was markedly promoted not only by yeast but also human RAD52, suggesting that transcript-RNA-templated DNA repair could occur in human cells. RNA transcripts could template DNA damage repair at highly transcribed loci, in cells that do not divide (lack sister chromatids), or have more stable RNA–DNA heteroduplexes, like those defective in RNASEH2 in patients with Aicardi–Goutières syndrome27. Our findings lay the groundwork for future exploration of RNA-driven DNA recombination and repair in different cell types.

Figure 3: Models of transcript-RNA-templated DSB repair in cis.
figure 3

An actively transcribed DNA region experiencing a DSB uses its own transcript RNA as a bridging (a) or an extension (b) template for repair. The small black lines indicate initial annealing between the transcript RNA and the DSB end(s), and between the two DSB ends. Orange circles, Rad52; green triangles, RNase H1 and H2 (H1/2).

PowerPoint slide

Full size image

Methods

Experimental design to explore transcript-RNA-templated chromosomal DSB repair in yeast

In the experimental design to explore transcript-RNA-templated chromosomal DSB repair it is critical to discriminate repair of the DSB by transcript RNA from repair by the DNA region that generates the transcript. Also, translation of the repairing transcript mRNA should not produce the functional His3 protein. Moreover, it is essential that DSB repair would not restore the HIS3 marker sequence by simple end ligation via non-homologous end-joining (NHEJ). To satisfy these requirements, the DNA region that generates the transcript was constructed to contain a his3 allele on chromosome III consisting of a yeast HIS3 gene interrupted by an artificial intron in the antisense orientation (mhis3-AI cassette), which was previously used to study reverse transcription in yeast28,29. The antisense his3 RNA is not translated into the functional His3 protein. Moreover, after intron splicing, the transcript RNA sequence has no intron, while the DNA region that generates the transcript retains the intron; thus they are distinguishable. We developed two experimental yeast cell systems, trans and cis (Fig. 1a, b and Extended Data Fig. 1) in strains YS-289, 290 and YS-291, 292, respectively (Extended Data Table 1). In both systems, transcription of the antisense his3 RNA and expression of the homothallic switching endonuclease are regulated by the galactose-inducible promoter (pGAL1). In addition, these yeast cell systems are auxotrophic for histidine (His−) and thus do not grow on media without histidine. Upon induction of the homothallic switching endonuclease DSB, the broken his3 allele of the trans and cis cell systems can, in principle, only be repaired to a functional HIS3 allele by recombination with a homologous template. Alternative mechanism of HIS3 repair by ligation of the broken ends via NHEJ is inefficient in this system (<0.1 out of 107 viable cells) (data not shown), as the HIS3 gene is disrupted by a long sequence with the homothallic switching endonuclease site (trans system) or an intron and the homothallic switching endonuclease site (cis system) (Extended Data Fig. 1b, c).

To impair DSB repair by cDNA deriving from the his3 antisense, we deleted the SPT3 or the DBR1 gene. SPT3 encodes for a subunit of the SAGA and SAGA-like transcriptional regulatory complexes and its null allele reduces Ty reverse transcriptase function over 100-fold3,8,9. DBR1 encodes for the RNA debranching enzyme Dbr1 and its null allele in yeast cells impairs cDNA formation and diminishes Ty transposition up to a factor of tenfold16,17. As further proof that we can detect DSB repair by transcript RNA independently of cDNA, we performed the trans and cis assays with and without RNase H functions in the presence of foscarnet (phosphonoformic acid, PFA), an inhibitor of the HIV reverse transcriptase, which blocks Ty reverse transcription in yeast18 (and data not shown).

Yeast strains

The yeast strains used in this work are listed in Extended Data Table 1 and derive from the FRO-767 strain3, which contains the site for the site-specific homothallic switching endonuclease in the middle of the LEU2 gene on chromosome III. A gene cassette carried on plasmid pSM50 (refs 28, 29) containing the his3 gene disrupted by an artificial intron and regulated in the antisense orientation by the galactose inducible promoter pGAL1 and containing the URA3 marker gene (pGAL1-mhis3-AI-URA3) was integrated into the leu2 locus of strain FRO-767 after DSB induction at the homothallic switching endonuclease site by the gene collage technique with no PCR amplification30. The URA3 gene was then replaced with the ADE3 gene generating strain FRO-1073. To build the strains of the trans system, an homothallic switching endonuclease site was integrated into the endogenous HIS3 locus on chromosome XV of FRO-1073 exactly in the same position in which the artificial intron was inserted in the pGAL1-mhis3-AI cassette using the delitto perfetto method, as described previously30,31, to generate FRO-1075, 1080. The correct sequence and insertion position of the homothallic switching endonuclease site was confirmed by sequence analysis. For constructing strains of the cis system, first the his3 gene disrupted by the homothallic switching endonuclease site of FRO-1075 and 1080 was replaced with a TRP1 gene to generate YS-164, 165, and then an homothallic switching endonuclease cutting site was integrated into the artificial intron in the his3 cassette on chromosome III to generate strain YS-172,174. To be cautious to avoid any possibility of transcription from Ty into the pGAL1-mhis3-AI cassette in both the trans and cis systems, the Ty2 element located upstream of the leu2 locus on chromosome III, YCLWTy2-1, was deleted following the delitto perfetto method to generate YS-289, 290 (trans system) and YS-291, 292 (cis system). These new strain constructs were verified by PCR and sequence analysis to confirm correct constructions. However, no difference in the frequency of His+ cells was observed between the strains with the YCLWTy2-1 and those without it for the strains of both the trans and cis systems (data not shown). Deletion mutants for the trans YS-289, 290, and the cis 291, 292 strains contain either the kanMX4, hygMX4, natMX4 and/or the Kluyveromyces lactis URA3 (KlURA3) marker gene in place of the open reading frame or the promoter of the gene(s) of choice. All gene disruptions were confirmed by colony PCR. Strains HK-396, 400 and HK-391, 394 were constructed using the delitto perfetto method by deleting the first 23 base pairs on the 5′ end of the artificial intron via insertion of the CORE cassette, and then by popping out the CORE cassette with a pair of oligonucleotides. These constructs were confirmed by sequence analysis. Strain HK-404, 407 was obtained by deleting the SPT3 gene with kanMX4 from HK-391, 394. The FRO-1092, 1093 strain is rad52Δ and has only one his3 allele, the endogenous allele on chromosome XV that has been inactivated by the homothallic switching endonuclease site.

Standard genetic, molecular biology techniques and plasmids

Yeast genetic methods and molecular biology analyses were done as described3,30,31. The BDG606 vector32 and the BDG283 control vector (a gift from D. Garfinkel), used to verify a direct role of transcript RNA in DSB repair (Extended Data Table 4), are centromeric plasmids with the URA3 marker. BDG606 contains the pGAL1-mhis3-AI cassette fused to Ty (pGTy1-H3his3-AI/Cen-URA3) and BDG283 contains only pGAL1. The plasmids used for the complementation assay with RNase H2 are YEp195SpGAL, which is a high-copy expression plasmid containing the URA3 selectable marker33, YEp195SpGAL containing the wild-type RNH201 gene (YEp195SpGAL-RNH201) inserted by gap repair, and YEp195SpGAL-rnh201-D39A constructed by in vitro mutagenesis (Quick Change Mutagenesis Kit, Stratagene, La Jolla, CA) of YEp195SpGAL-RNH201 and confirmed by sequence analysis. To confirm occurrence of the homothallic switching endonuclease DSB following incubation in the 2% galactose medium, the percentage of G2 arrested cells was determined right before adding galactose and after 8-h incubation in galactose as previously described34 (Extended Data Fig. 2c). All primers used for strain and plasmid constructions, PCR verifications and sequence analyses are available upon request. Samples for sequencing were submitted to Eurofins MWG Operon. The Southern blot experiment was done as follows. Cells from colonies growing on rich medium containing yeast extract, peptone and 2% (w/v) dextrose (YPD) or His− media were grown on YPD overnight (O/N). Genomic DNA was extracted as described35 and digested with either BamHI or NarI restriction enzyme. After digestion, column purification was applied by using QIAquick PCR Purification Kit (Qiagen). DNA was run in a 0.8% agarose gel. Following electrophoresis and Southern blotting chromosomal regions containing the HIS3 gene were detected using a [α-32P]ATP (PerkinElmer)-labelled (Prime-It RmT Random Primer Labelling Kit, Agilent Technologies) 250-base-pair HIS3-specific probe. Membrane was exposed to a phosphor screen for 3 days. Images were taken with Typhoon Trio+ (GE Healthcare) and obtained with ImageQuant (GE Healthcare).

Trans and cis assays using patches or liquid cultures

Yeast cells of the chosen strains were patched on YPD and grown at 30 °C for 1 day. The cells were then replica-plated on medium containing yeast extract, peptone and 2% (w/v) galactose (YPGal) or YPGal containing phosphonoformic acid (PFA, 2.5 mg ml−1) to turn on transcription of the his3 antisense on chromosome III and expression of the homothallic switching endonuclease. As a control, cells were also replica-plated from the YPD medium on synthetic complete medium plates lacking histidine (SC-His−) and grown for 3 days at 30 °C. We never detected a single His+ colony from any of the trans and cis strains used in this study following replica-plating from the YPD medium on SC-His− (not shown). After 2 days’ incubation on YPGal medium, these cells were replica-plated onto SC-His− and grown for 3 days at 30 °C to form visible colonies. At this stage, plates were photographed and photo files stored. For experiments using the BDG606 and BDG283 plasmids, cells were replica-plated from SC-Ura− onto SC-Ura−Gal medium, and were then replica-plated onto SC-Ura−His−. As a control, cells were also replica-plated from the SC-Ura− medium onto SC-Ura−His− and grown for 3 days at 30 °C.

For the experiments in liquid culture, flasks with 50 ml of liquid medium containing yeast extract, peptone and 2.7% (v/v) lactic acid (YPLac) were inoculated with yeast cells of the chosen strains and incubated in a 30 °C shaker for 24 h. The density of the cultures was determined by counting cells using a hemocytometer and counting under a microscope. Generally, 107 or, in rare cases, 108 cells (we note that survival is very low on galactose medium) were then plated on YPGal medium, or YPGal medium containing PFA (2.5 mg ml−1) for experiments using PFA to obtain from 1 to ∼500 His+ colonies per plate after the replica-plating on His− medium, and grown for 2 days at 30 °C. Two aliquots of 104 cells were plated, each on one YPGal medium plate, or YPGal medium containing PFA (2.5 mg ml−1) for experiments using PFA plate, to measure the cell survival after galactose treatment. After 2 days’ incubation on YPGal medium, cells were replica-plated on His− plates and grown for 3 days at 30 °C. The frequency of DSB repair was calculated by dividing the number of His+ colonies on SC-His− medium by the number of colonies on YPGal medium. The survival was calculated by dividing the number of colonies on YPGal medium by the number of cells plated on the same medium. For experiments using the BDG606 and BDG283 plasmids, cells were treated as described above except that they were plated from YPLac on SC-Ura−Gal medium in different dilutions, and were then replica-plated on SC-Ura−His−. The frequency of His+ colonies was calculated by dividing the number of His+ colonies on SC-Ura−His− medium by the number of colonies on SC-Ura−Gal medium. The survival was calculated by dividing the number of colonies on SC-Ura−Gal medium by the number of cells plated on the same medium.

Oligonucleotide transformation

Transformation by oligonucleotides (1 nmol) was performed as described3. Induction of the homothallic switching endonuclease DSB was done by incubating cells in 2% galactose medium for 3 h.

Transposition assay

Yeast cells of the chosen strains transformed with BDG102 (empty plasmid) or BDG598 (pGTy-H3mhis3-AI) plasmid36 (containing a Ty transposon fused to the his3 gene, which is in the antisense orientation and disrupted by an artificial intron; both Ty and the his3 antisense are regulated by the galactose-inducible promoter) were patched on SC-Ura− and grown overnight at 30 °C. Cells were then replica-plated on synthetic medium lacking uracil with 2% (w/v) galactose (SC-Ura−Gal) and grown for 48 or 96 h at 30 °C or 22 °C, respectively. As control, cells were also replica-plated on SC-His− to determine the background of His+ clones. After the incubation in galactose, cells were replica-plated on SC-His− and grown for 3 days at 30 °C to form visible colonies. At this stage, plates were photographed and photo files stored. For the experiments in liquid culture, strains with BDG102 or BDG598 were grown in 5 ml SC-Ura− liquid medium or in 10 ml of YPLac liquid medium in a 30 °C shaker for 24 h. Then, 1 × 106 cells were transferred from the SC-Ura− liquid medium into 5 ml SC-Ura− or 5 ml SC-Ura−Gal liquid medium and incubated for 48 or 96 h at 30 °C or 22 °C, respectively. After 24 h, YPLac cultures were split in half. One-half was kept growing for additional 48 h at 30 °C, while galactose was directly added to the other half to reach 2% and cells were then incubated for 48 h at 30 °C. From glucose and YPLac cultures grown at 22 °C or 30 °C, 107 or 108 cells were plated on SC-His−Ura− medium, respectively, and were grown for 2 days at 30 °C. From galactose cultures grown at 22 °C or 30 °C, 105 or 106 cells were plated on SC-His−Ura− medium, respectively, and were grown for 2 days at 30 °C. Two aliquots of 5 × 102 cells were plated each on one SC-Ura− medium plate, to measure the cell survival after glucose, YPLac or galactose treatment. The rate of formation of His+ cells was calculated using the maximum-likelihood method described in ref. 37.

Quantitative real-time PCR

RNA was isolated from the chosen yeast strains of the trans and cis systems using a protocol adapted from a method described previously38. RNA was converted in to cDNA using QuantiTect Reverse Transcription Kit (Qiagen). SYBR Green qPCR Mix (BioRad) was used for analysing RNA expression in 96-well plates (Applied Biosystems). The total volume in each well was 20 µl, which consisted of 10 µl of SYBR Green qPCR Mix, 4 µl of nuclease-free water, 2 µl of primers and 4 µl of cDNA. The cDNA levels were determined using an ABI Prism 7000 RT–PCR machine (Applied Biosystems). ACT1.F and ACT1.R, HIS3.F2 and HIS3.R2 primers were used in this study (Extended Data Table 2a). ACT1 primers were used for normalization. Values for each sample were normalized with ACT1, and then a second normalization was performed by subtracting normalized values of each time point from the control normalized value per each gene39. As a negative control, CEN16.F and CEN16.R primers were used to show that there is minimal or no qPCR product derived from a chromosomal region that is not transcribed (A. El Hage, personal communication) (data not shown).

Rad52 in vitro annealing assay

In vitro assays using yeast or human Rad52 were performed as described40,41 (and references therein), with all DNA and RNA concentrations expressed in moles of molecules. All oligonucleotide sequences (no. 211, no. 501, no. 508 and no. 509) are shown in Extended Data Table 2a. A single nucleotide mismatch was incorporated into the dsDNA (relative to ssDNA or RNA) to reduce the spontaneous Rad52-independent annealing. Tailed dsDNA (no. 508 and no. 509) (0.4 nM) was incubated in the absence or presence of yeast or human RPA (2 nM) in a buffer containing 25 mM Tris acetate, pH 7.5, 100 μg ml−1 BSA, and 1 mM DTT (dithiothreitol) for 5 min at 37°C. Then yeast or human Rad52 (1.35 nM) was added to the mixture containing either yeast or human RPA, respectively, and incubation continued for 10 min. Annealing reactions were initiated by adding 32P-labelled ssRNA (no. 501) or ssDNA (no. 211) (0.3 nM). Aliquots were withdrawn at indicated time points and deproteinized by incubating samples in stop solution containing 1.5% SDS, 1.4 mg ml−1 proteinase K, 7% glycerol and 0.1% bromophenol blue for 15 min at 37 °C. Samples were analysed by electrophoresis in 10% (17:1 acrylamide:bisacrylamide) polyacrylamide gels in 1X TBE (90 mM Tris-borate, pH 8.0, 2 mM EDTA) at 150 V for 1 h and were quantified using a Storm 840 Phosphorimager and ImageQuant 5.2 software (GE Healthcare).

Data presentation and statistics

Graphs were made using GraphPad Prism 5 (Graphpad Software). The results are each expressed as a median and 95% confidence interval (in brackets), or alternatively the range when number of repeated experiments was <6. Statistically significant differences between the His+ frequencies were calculated using the nonparametric two-tailed Mann–Whitney U-test42. All P values obtained using the Mann–Whitney U-test were then adjusted by applying the false discovery rate method to correct for multiple hypothesis testing43 (Supplementary Table 1).