Synthesis-dependent strand-annealing (SDSA)-mediated homologous recombination replaces the sequence around a DNA double-strand break (DSB) with a copy of a homologous DNA template, while maintaining the original configuration of the flanking regions. In somatic cells at the 4n stage, Holliday-junction-mediated homologous recombination and nonhomologous end joining (NHEJ) cause crossovers (CO) between homologous chromosomes and deletions, respectively, resulting in loss of heterozygosity (LOH) upon cell division. However, the SDSA pathway prevents DSB-induced LOH. We developed a novel yeast DSB-repair assay with two discontinuous templates, set on different chromosomes, to determine the genetic requirements for somatic SDSA and precise end joining. At first we used our in vivo assay to verify that the Srs2 helicase promotes SDSA and prevents imprecise end joining. Genetic analyses indicated that a new DNA/RNA helicase gene, IRC20, is in the SDSA pathway involving SRS2. An irc20 knockout inhibited both SDSA and CO and suppressed the srs2 knockout-induced crossover enhancement, the mre11 knockout-induced inhibition of SDSA, CO, and NHEJ, and the mre11-induced hypersensitivities to DNA scissions. We propose that Irc20 and Mre11 functionally interact in the early steps of DSB repair and that Srs2 acts on the D-loops to lead to SDSA and to prevent crossoverv.
DOUBLE-strand DNA breaks (DSBs) are generated by exposure to ionizing radiation or chemical compounds, such as topoisomerase inhibitors and apoptosis inducers, reactive oxygen species (ROS) generated as by-products of oxidative respiration, transposition of transposable elements, and by the actions of restriction enzymes or meiosis-specific endonucleases, such as the Spo11 complex and intron-homing nucleases. DSBs are repaired through various recombination-dependent pathways (Figure 1), and recombination deficiencies cause genome instability and carcinogenesis.
Holliday-junction-mediated homologous recombination (Figure 1A) is a type of homologous recombination that produces either crossover or noncrossover products. These products are generated on the basis of the resolution of the double-Holliday structure (Resnick and Martin 1976; Szostak et al. 1983). It was proposed that the noncrossover products (B5) are generated from a double-Holliday structure (A4), with reverse migration of the Holliday junctions without cleavage (Wu and Hickson 2003).
Synthesis-dependent strand-annealing (SDSA)-mediated homologous recombination (Figure 1B) is the other type of homologous recombination, which produces only noncrossover products that are not associated with a flanking crossover. In this pathway, a D-loop, formed with a single-strand tail from the terminus of a DSB (B2), migrates with associated DNA synthesis primed at the 3′ termini of the invading single-strand tail, but without forming Holliday junctions. These reactions are followed by the annealing of the newly synthesized strands dissociated from the template DNA with the other termini of the DSB (Nassif et al. 1994). This pathway was postulated on the basis of homologous recombination studies in phage T4 (Mueller et al. 1996), Saccharomyces cerevisiae (McGill et al. 1989; Nelson et al. 1996; Pâques et al. 1998), Ustilago maydis (Ferguson and Holloman 1996), and Drosophila melanogaster (Nassif et al. 1994; Adams et al. 2003).
The two homologous recombination pathways described above share common steps up to the D-loop formation of a single-strand tail, from a DSB with a homologous double-strand DNA as the template for DSB repair.
Nonhomologous end joining (NHEJ) (Figure 1C) religates DSB termini lacking extensive homologous sequences and generates imprecise or precise end-joining products, according to the features of the DSB termini. This pathway mainly depends on Ku70 (Boulton and Jackson 1996b), Ku80 (Boulton and Jackson 1996a), and Lig4 (Schär et al. 1997; Ramos et al. 1998). The alternative NHEJ pathway is independent of Ku and Lig4, but depends on Rad52 (Hegde and Klein 2000).
When a DSB is generated between the centromere and a gene in somatic cells heterozygous for the gene, the Holliday-junction-mediated homologous recombination at S/G2 (4n) and the imprecise end joining at G1/S (2n) or S/G2 (4n) frequently produce a crossover between homologous chromosomes and long deletions, respectively, and then cause loss-of-heterozygosity (LOH) of the alleles upon cell division. In contrast, the SDSA pathway leads only to noncrossover products that are free of deletions around the DSB sites and thus prevents DSB-induced LOH in somatic cells. LOH is a major cause of carcinogenesis (Bignon 2004). Thus, it is important to study the SDSA-mediated homologous recombination pathway to clarify the underlying mechanisms that maintain heterozygosity and genome stability.
Various mitotic or meiotic genetic events, which can be explained by the SDSA model, have been reported from yeast studies. For instance, most of the quite frequent expansions and contractions in a tandem array of 375-bp units, which were induced by a DSB, were not associated with crossovers (Pâques et al. 1998). In another SDSA-based mechanism, new sequences were created by template switching between Kluyveromyces lactis URA3 and S. cerevisiae URA3 and were mostly independent of mismatch repair (Hicks et al. 2010). His+ and Leu+ markers placed on opposite sites of a DSB site were converted, without the conversion of an intervening heterologous insertion surrounded by a few hundred-base duplication on the donor chromatid. This was explained by the annealing of two strands newly synthesized on the few hundred-base duplication, resulting in the loss of the intervening marker (McMahill et al. 2007). In many of the meiotic recombination events initiated at the HIS4 recombination hotspot, the heteroduplex formation occurs on only one side of the DSB site and then often produces noncrossovers (Merker et al. 2003).
These genetic observations support the presence of the SDSA mechanism in vegetative and meiotic yeast cells. The presence of SDSA in yeast was demonstrated by a triparental system, in which the gapped leu2 plasmid and two donor templates on two different chromosomes were used (Pâques et al. 1998). However, the genetic requirements for the SDSA pathway, and especially the DNA helicases that are likely to determine the SDSA pathway (Figure 1B) or the Holliday junction-mediated pathway (Figure 1A), remain to be identified. SRS2 encodes a DNA helicase (Rong and Klein 1993). Mutants with defective srs2 show a hyperrecombination phenotype (Aguilera and Klein 1988) and increased mitotic crossovers, which rarely occur in the wild-type cells (Ira et al. 2003). These results suggested that the Srs2 helicase functions negatively in somatic homologous recombination. The negative function is explained by the observation that Srs2 disrupted in vitro Rad51 filaments formed on single-strand DNA and inhibited the DNA strand exchange mediated by Rad51 (Veaute 2003). A synthetic D-loop, formed from double-strand DNA with a homologous 3′-single-strand tailed double-strand DNA was efficiently dissociated by the Srs2 DNA helicase (Dupaigne et al. 2008). This observation favors a model in which Srs2 promotes the migration of a D-loop, accompanied with DNA synthesis initiated at the 3′ terminus of the invading single-strand tail in the SDSA pathway (Dupaigne et al. 2008) (Figure 1, B3). Thus, it is critical to elucidate whether the Srs2 helicase inhibits D-loop formation or acts after the D-loop formation, to determine whether the SDSA pathway or the Holliday junction-mediated pathway is utilized in somatic cells.
In this study, we developed a new yeast genetic assay involving a triparental system, consisting of the gapped ura3 plasmid and two donor templates on two different chromosomes, to determine the genetic requirements either for SDSA-mediated noncrossover, as distinguished from those for noncrossover through a double-Holliday intermediate, or for the NHEJ pathway. Using this assay, we showed that the elimination of the Srs2 helicase decreased SDSA production, but increased imprecise end joining. We then tested Irc20, which is a new putative DNA/RNA helicase classified into the Srs2 subgroup by a phylogenetic analysis. On the basis of our studies on the effects of the inactivation of Srs2 and Irc20 within various genetic backgrounds, we discuss the roles of these two DNA helicases in SDSA-mediated homologous recombination.
Materials and Methods
Yeast strains and media
The S. cerevisiae strains used in this study are listed in Table 1. YPD (1% yeast extract, 2% peptone, and 2% glucose) was used as the complete medium, and SD (2% glucose or 2% galactose, 0.67% Bacto-yeast nitrogen base, and appropriate supplements) was used to select cells carrying selective markers. Yeast transformation was performed using a lithium acetate-based protocol (Gietz et al. 1995) for the construction of knockout strains, the SDSA/NHEJ assay, and the targeted integration (crossover) assay.
All of the knockout alleles used in this study were constructed by the one-step gene-disruption method (Rothstein 1983). For the preparation of knockout DNA fragments, we used the strategy of short flanking homology PCR (Capozzo et al. 2000). The primers used for the knockouts are listed in Supporting Information, Table S2. The knockout fragment with the kanamycin-resistant marker was created by PCR amplification of the KanMX4 module, using a primer pair containing a 5′ portion homologous to the gene to be disrupted and a 3′ portion homologous to the multiple cloning sites (MCS) of the plasmid pFA6a-KanMX4, which was introduced into competent cells. YPD plates with 200 μg/ml of G418 (Invitrogen, CA) were used to select the clones bearing geneΔ::KAN. The knockout fragment with a hygromycin-resistant marker was created by PCR amplification, using the primer pair containing the 5′ portion homologous to the gene to be disrupted and the 3′ portion homologous to the MCS of the plasmid pAG34-HYG, which was introduced into competent cells. YPD plates with 200 μg/ml of hygromycin B (Invitrogen, CA) were used to select the clones bearing geneΔ::HYG. The knockout fragment with the TRP1 marker was created by PCR amplification, using the primer pair containing the 5′ portion homologous to the gene to be disrupted and the 3′ portion homologous to the TRP1 gene derived from the plasmid pAS2-1, and was used for transformation into Trp+ clones, which carried geneΔ::TRP1.
5′Δ-ura3 (one template allele, residing at the ura3 locus on chromosome V; Figure 2A): At first, the ura3-1 allele of the W303 strain was reverted to Ura+, by transformation with the URA3 fragment encompassing base pairs 2734 to 3841 in the sequence of the pYES2 plasmid, to construct the strain KTYY84. To construct the strain with a deletion at the 5′ end of this URA3 locus, two primers, PRI245 (with the homology arm for the upstream region of the URA3 locus) and PRI249 (with the homology arm for the internal region of the URA3 gene), were designed to create a deletion from the promoter to the 39th codon and to introduce the HIS3 fragment derived from the T334 strain. The fragment of 5′Δ-ura3 with the HIS3 marker was used to transform the URA3 locus of KTYY84 into 5′Δ(::HIS3)-ura3, to construct the strain KTYY86. This structure was confirmed by PCR with the PRI240 (downstream of the URA3 locus) and PRI245 (upstream of the URA3 locus) primers, which produced the 1848-bp fragment.
ura3-3′Δ (the other template allele, integrated into the AUR1 locus on chromosome XI; Figure 2A): To construct the strain with the ectopic ura3 allele in which the 3′ end was deleted, the region from the promoter to the 139th codon of the URA3 gene, derived from the pYES2 plasmid, was amplified with two primers, PRI231 (with a KpnI site tag) and PRI232 (with an XbaI site tag). The 684-bp fragment of ura3-3′Δ was ligated into the KpnI and XbaI sites of the MCS of the pAUR101 plasmid, which was used for targeted integration. The StuI-cleaved pAUR101::ura3-3′Δ plasmid was introduced into the AUR1 locus to construct the strain KTYY90, with AUR1-ura3-3′Δ-AmpR-pUCOri-AUR1C. This structure was confirmed by PCR with the primer pair PRI231 and PRI232, which produced the 684-bp fragment of ura3-3′Δ, and with another primer pair, PRI766 (at the 5′ end of AUR1 gene) and PRI769 (at the multiple cloning site of pAUR101), which produced the 3613-bp fragment including the AUR1-C segment.
ura3-intΔisceI (the recipient allele with the internal deletion of the URA3 gene, sealed with an I-SceI site, residing within the plasmid with the LEU2 marker; Figure 2A): The pAUR101 plasmid was first cleaved with the HindIII restriction enzyme and was self-ligated, to create the 3162-bp plasmid pUCOri-AmpR-MCS, lacking the AUR1-C gene. The 1173-bp KpnI/XbaI fragment containing the URA3 gene derived from the pYES2 plasmid, obtained from the amplification by PRI231 (with a KpnI site tag) and PRI273 (with an XbaI site tag), was inserted into the KpnI and XbaI sites of this plasmid, to create the plasmid pUCOri-AmpR-URA3. To introduce the CEN6/ARSH4 sequence into this plasmid, the 554-bp SphI/XbaI fragment, including CEN6/ARSH4 derived from the plasmid pYC6/CT, was amplified with the PRI274 (with an SphI site tag) and PRI275 (with an XbaI site tag) primers and was inserted into the SphI and XbaI sites of the plasmid pUCOri-AmpR-URA3, to create pUCOri-AmpR-URA3-CEN6/ARSH4. To add the LEU2 transformation marker to this plasmid, the 1951-bp SphI fragment, including the LEU2 gene derived from the plasmid pRS315, was amplified with the PRI271 and PRI278 primers (both with an SphI site tag) and was inserted into the SphI site of this plasmid, to create pUCOri-AmpR-URA3-CEN6/ARSH4-LEU2. To enlarge the 3′-end homology of the recipient ura3 allele, the 1580-bp fragment, from the StuI site within URA3 to the downstream region of URA3, was amplified by PRI276 and PRI277 (with an XbaI site tag), and the 1501-bp StuI/XbaI fragment was replaced with the 473-bp StuI/XbaI region of the plasmid pUCOri-AmpR-URA3-CEN6/ARSH4-LEU2. To create the gap of the 458-bp PstI/StuI region in the recipient ura3 allele, this plasmid was cleaved by the PstI and StuI restriction enzymes and treated with T4 DNA polymerase, and then the NruI linker (CTCG|CGAG) was inserted. Finally, the sequence of the I-SceI site (5′-TAGGGATAA|CAGGGTAAT/3′-ATCCC|TATTGTCCCATTA) was introduced into the NruI site, to create the pSDSA/NHEJ plasmid. This assay plasmid has 1492-bp of 3′-end homology and 3365-bp of 5′-end homology.
To prepare competent cells of KTYY90 and its mutant derivatives (Table 1) which have 5′Δ-ura3 and ura3-3′Δ, a culture was grown in liquid SD media containing 1 μg/ml 5-fluoroorotic acid monohydrate (5-FOA, Wako, Osaka, Japan), which prevented the generation of Ura+ cells by homologous recombination between two ura3 alleles. A portion of the preculture was added to new liquid medium, to an initial OD of 0.05. The liquid culture was incubated at 30° to mid-log phase (OD, 0.5∼0.6), and then the cells were collected and suspended in a lithium acetate solution. Aliquots (0.1 ml) of the cell suspension were placed into 1.5-ml microtubes and were frozen slowly at −80°. This slow-freezing procedure generated cells that were mostly viable after mild thawing on ice for 60 min or less, while in contrast, the use of either solid carbon dioxide or liquid nitrogen caused severe cell death. The I-SceI cut plasmid DNA (1 μg) was mixed with 0.1 ml of the thawed competent cell suspension, and then appropriate treatments in accordance with the lithium acetate method were performed. After an incubation in 1 ml of nonselective SD medium at 30° for 2 hr, the cell suspension was washed and plated onto selective media. Cells transformed with the SDSA products were selected for Ura+ Leu+, and cells transformed with the NHEJ products were selected for 5FOAR Leu+. After an incubation at 30° for 3 days, the transformed colonies were scored, and the number of transformant colonies per microgram was determined as the transformation efficiency. The same amount (1 μg) of the uncut plasmid was concurrently transformed in the same manner as the I-SceI cut plasmid. The Leu+ transformant colonies were selected and scored, and the number of transformant colonies per microgram was determined as the transformation competency. The SDSA frequency was calculated by dividing the Ura+ Leu+ transformation efficiency with the I-SceI cut plasmid by the transformation competency with the uncut plasmid. The NHEJ frequency was calculated by dividing the 5FOAR Leu+ transformation efficiency with the I-SceI cut plasmid by the transformation competency with the uncut plasmid. For the structural analysis of the NHEJ products, whole DNA was extracted from each of the 5-FOAR Leu+ colonies. Each ura3 plasmid DNA was amplified by the PRI231 (at the 5′ end of URA3 gene) and PRI273 (at the 3′ end of URA3 gene) primers with the whole DNA samples and was digested with I-SceI nuclease. The I-SceI-sensitive clones were judged as precise end-joining products, and the I-SceI-resistant clones were sequenced.
Targeted integration (crossover) assay
This assay was performed as previously reported (Yamana et al. 2005). pAUR101 (6687 bp; Takara, Kyoto, Japan), which bears the dominant AUR1-C allele conferring resistance to Aureobasidin A (AurR), does not possess either a centromere (CEN) or an autonomous replication site (ARS) (Hashida-Okado et al. 1998). The base substitution producing the dominance is at 4522 within the pAUR101 sequence (AB012282; Hashida-Okado et al. 1998). The blunt-ended cleavages by the MscI and StuI restriction enzymes occur at 3231 and 4143 of pAUR101, respectively, and each recognition site resides within the AUR1-C allele. This plasmid DNA was cleaved by the MscI enzyme or the StuI enzyme and was introduced into W303α and its mutant derivatives. The resultant AurR transformants with AUR1-(pUCOri-AmpR)-AUR1-C were scored as targeted integrants. The pRS315-AurR plasmid (8599 bp; Yamana et al. 2005), an AUR1-C derivative of the pRS315 plasmid (Sikorski and Hieter 1989), was used to determine the transformation competency and to normalize the differences in DNA uptake capacity among preparations of competent cells. Competent cells of W303α and its mutant derivatives were prepared and transformed in the same manner as for the SDSA/NHEJ assay. The StuI cut pAUR101 DNA (1 μg) and 1.28 μg of uncut pRS315-AurR DNA (the same molar amount as the StuI cut DNA) were used for a set of transformations. SD plates with 0.4 μg/ml of Aureobasidin A (Takara, Kyoto, Japan) were used for the selection. The targeted integration frequency (percentage) by crossover events with associated gene conversion was calculated by dividing the AurR transformation efficiency with 1 μg of the MscI cut or StuI cut pAUR101 DNA by the AurR transformation efficiency (competency) with 1.28 μg of the uncut pRS315-AurR plasmid.
To detect the targeted integration (crossover) events associated with longer gene conversion, we prepared the plasmid pAUR101-I-SceI, constructed by inserting an 18-mer bearing the I-SceI recognition sequence into the StuI site. This 18-mer insertion inhibits the function of the AUR1-C gene products. The plasmid DNA linearized by MscI digestion results in products conferring the AurR phenotype when gene conversion occurs beyond the 18-mer insertion, which is 912 bp away from the DSB end, to regain the StuI sequence.
Measurement of sensitivity to DNA scissions
Sensitivity to EcoRI expression:
The strains to be tested, carrying the plasmid pGAL1::EcoRI (URA3), which bears the EcoRI restriction gene expressed by the GAL1 promoter (Lewis et al. 1998), were grown to mid-log phase (OD, 0.2∼0.4) in minimal medium containing 2% glucose. The cells were washed twice to remove the glucose, diluted, and spread onto minimal plates with 2% glucose or 2% galactose. The plates were incubated for 4 days at 30°, and then the number of viable colonies was counted. The surviving fraction was calculated by dividing the colony-forming units on the plates with 2% galactose by those on plates with 2% glucose. For the spot assay, 10 μl portions of a suspension of serial 10-fold dilutions from a mid-log phase culture with the indicated genotype were spotted onto minimal plates containing 2% glucose or 2% galactose.
Sensitivity to bleomycin:
Bleomycin (Zeocin; InvivoGen, Carlsbad, CA) was supplemented within solid media. The strains to be tested were grown to mid-log phase (OD, 0.2∼0.4) in YPD medium. The cells were diluted and spread onto YPD plates with or without bleomycin. The plates were incubated for 3 days at 30°, and then the number of viable colonies was counted. The surviving fraction was calculated by dividing the colony-forming units on the plates with the reagent by those on plates without the reagent. For the spot assay, 10 μl portions of a suspension of serial 10-fold dilutions from a mid-log phase culture with the indicated genotype were spotted onto YPD plates containing the indicated concentration of a DNA scission reagent, such as bleomycin, methyl methanesulfonic acid, (MMS; ICN Biomedicals, OH), camptothecin (CPT; Wako, Osaka, Japan), and hydroxyurea (HU; Wako, Osaka, Japan).
SDSA/NHEJ assay system
SDSA-mediated recombination, which generates noncrossovers, progresses via the following steps: D-loop formation between a single-strand tail from a processed DSB terminus and the homologous double-strand DNA (Figure 1, A1B1–A2B2), priming of DNA repair synthesis at the 3′ end in the D-loop (A2B2-B3), dissociation of the newly synthesized strand from the double-strand DNA-template (B3–B4), and annealing of the new strand with the complementary strand of the other processed DSB end (B4–B5). To detect the noncrossover products formed by SDSA and to distinguish them from those formed by other pathways, we developed a yeast SDSA assay, consisting of two discontinuous templates on different chromosomes and a linearized recipient plasmid DNA containing a gene disrupted by a DSB. The disrupted gene will be repaired for activation by copying both templates, followed by annealing of the complementary single-strand copies (Figure 2A). In this two-separate-template system, no double-Holliday structure is formed from the two discontinuous templates on different chromosomes, and the noncrossover products are generated solely by SDSA (Figures 1 and 2). This assay consists of three different ura3 alleles: (i) 5′Δ-ura3, one template allele with a 5′-end deletion of URA3, set at its locus on chromosome V; (ii) ura3-3′Δ, the other template allele with a 3′-end deletion of URA3, integrated into the aur1 locus on chromosome XI; and (iii) ura3-intΔisceI, the recipient allele with the gap sealed with an I-SceI site, carried on the plasmid with the LEU2 marker. Transformation of the I-SceI cut recipient plasmid would generate Ura+ Leu+ colonies as the products of SDSA (Figure 2B), and 5FOAR Leu+ colonies as the DSB repair products by precise or imprecise NHEJ (Figure 2C). This assay has advantages over the existing assays consisting of either a template allele with direct repeats or a single-mutant template allele, since these do not distinguish the SDSA-mediated noncrossover products from those formed through double-Holliday intermediates.
Control experiments without any plasmid showed that Ura+ clones in competent cell suspensions of the double-template strain were spontaneously generated at 6.9E-08 ± 3.0E-9 (n = 3), indicating that the spontaneous frequency of Ura+ clones in competent cells is sufficiently low to contain them in a population of Ura+ transformants possessing SDSA products (Figure 3A).
A quantitative analysis with the SDSA/NHEJ plasmid in the double-template strain revealed that SDSA and NHEJ occurred at 4.7 and 0.99%, respectively (Figure 3, A and B, and Table S1). In a similar triparental system, in which a plasmid with a gapped leu2 allele and two donor templates on two different chromosomes were used, the frequency of SDSA was reportedly ∼0.3% (Pâques et al. 1998), suggesting that the difference in the frequency from their system would be due to the homology length at the DSB ends. A molecular analysis of the SDSA products from the wild-type strain, by a PCR analysis of each of the Ura+ Leu+ colonies and treatments with PstI and StuI (Figure 2B and Figure S1A), revealed that all of them (24/24) had the 458-bp PstI–StuI fragment, indicating that they were SDSA products (Figure S1A). A molecular analysis of each NHEJ product (5FOAR Leu+ colony) from the wild-type strain, by PCR and a treatment with I-SceI (Figure 2C and Figure S1B), showed that all of them (144/144) were sensitive to the I-SceI endonuclease, indicating that they were precise end-joining products (Figure S1B and Table 2).
Rad52 promotes the SDSA pathway and the Lig4-independent NHEJ pathway
Rad52 plays a central role in homologous recombination and mediates D-loop formation by Rad51 nucleoprotein filaments between a single-strand tail from a processed DSB end and homologous double-strand DNA (Sung 1997). The RAD52 gene knockout caused a 2500-fold reduction in SDSA (from 4.7 to 0.0019%; Figure 3, A and B, and Table S1), indicating that SDSA-mediated noncrossover formation requires Rad52 function. On the other hand, the rad52 knockout did not affect NHEJ (Figure 3, A and B, and Table S1), indicating that Rad52 is not required for the major NHEJ pathway.
DNA ligase 4 (Lig4, LIG4 gene product) is known to promote the Ku-dependent NHEJ (Schär et al. 1997; Ramos et al. 1998). The LIG4 gene knockout caused a fivefold reduction in NHEJ (from 0.99 to 0.18%), but did not significantly affect SDSA (Figure 3, A and B, and Table S1). These results indicate that the NHEJ detected by this assay is mostly dependent on the main pathway promoted by Ku and Lig4, consistent with previous reports (Boulton and Jackson 1996b; Schär et al. 1997; Ramos et al. 1998). Molecular analyses of the NHEJ products from the lig4 knockout strain revealed that 6% of the NHEJ products (9/144) were resistant to the I-SceI endonuclease, indicating that some products are generated by imprecise NHEJ (Table 2). They appear to be joined with the associated deletion of a few bases (from 1 to 8 bases) within the I-SceI recognition sequence (Table 2). Next, we tested whether the remaining NHEJ capacity in the lig4 mutant is dependent on Rad52. The addition of the rad52 knockout to the lig4 defective strain caused a 22-fold further reduction in NHEJ (from 0.18 to 0.0083%; Figure 3, A and B, and Table S1). As the reduction of NHEJ by the rad52 knockout was observed only in the lig4-deficient background, the cells may have an alternative NHEJ pathway requiring Rad52 function, in addition to the Lig4-dependent NHEJ pathway.
The Srs2 DNA helicase and the Irc20 putative helicase promote the SDSA pathway
The Srs2 DNA helicase recognizes synthetic DNA structures that mimic D-loops as substrates and translocates them on RPA-coated single-strand DNA (Dupaigne et al. 2008). We tested the possible functions of Srs2 in SDSA, using the in vivo SDSA/NHEJ assay system. The knockout of SRS2 caused a fourfold reduction in SDSA (from 4.7 to 1.1%) (Figure 3, A and B, and Table S1), indicating that Srs2 actively promotes SDSA-mediated noncrossover production, probably by its D-loop processing activity.
The negative effect of the srs2 knockout on SDSA production was much smaller than that of the rad52 knockout. We interpreted this result as suggesting the existence of another DNA helicase that can replace Srs2 and looked at the putative DNA/RNA helicase Irc20, which was classified into the Srs2 subgroup by a phylogenetic analysis for the helicase-related proteins in S. cerevisiae (Shiratori et al. 1999), as a candidate for an Srs2 functional analog. Irc20 protein is conserved in Ascomycota; however, its function is unknown. We tested the effects of its knockout on DSB repair. The IRC20 knockout caused a fivefold reduction in SDSA (from 4.7 to 0.98%), similar to the srs2 knockout (Figure 3, A and B, and Table S1), indicating the function of Irc20 in SDSA. The double knockout of SRS2 and IRC20 decreased SDSA as much as each single knockout (Figure 3, A and B, and Table S1), indicating that Irc20 works in the SDSA pathway that involves the Srs2 helicase.
Maintenance of precise NHEJ by DNA helicases
The efficiencies of NHEJ by the knockout strains of srs2 (1.9%), irc20 (1.2%), and srs2 irc20 (1.4%) were slightly higher than that of the wild-type strain (0.99%; Figure 3, A and B, and Table S1). We analyzed the structures of the NHEJ products. Unlike the wild-type strain (144/144), in which all of the NHEJ products were I-SceI sensitive, in the srs2 strain, 18% of the NHEJ products (21/120) were resistant to I-SceI (Table 2). In the irc20 strain, 8% of the NHEJ products (11/132) were resistant to I-SceI (Table 2). These results indicate that imprecise NHEJ is enhanced in these DNA helicase mutants and that the NHEJ products lost a few bases (from 1 to 11 bases) within the linker sequence (Table 2). These results suggest that Srs2 and Irc20 contribute to the prevention of imprecise NHEJ.
Irc20 works before D-loop processing by Srs2
To investigate the step at which Irc20 functions, we utilized a targeted integration (crossover) assay, which produces the duplication of AUR1 and AUR1-C as the crossover form, thus conferring AurR (Materials and Methods; Yamana et al. 2005; Figure 4A). The targeted integration occurred with a 4.1% frequency in the wild-type strain and a 9.3% frequency in the srs2 knockout strain (Figure 4, B and C), thus revealing a 2.3-fold enhancement by the srs2 knockout, as observed in a previous report (Ira et al. 2003). The single knockout of RAD52 caused a 147-fold reduction in the targeted integration (from 4.1 to 0.028%) (Figure 4, B and C), indicating the strong dependency of the integration on Rad52. The rad51 knockout decreased the targeted integration to a similar extent as the rad52 knockout (Yamana et al. 2005). We tested whether the irc20 knockout enhanced the targeted integration, as in the case of the srs2 knockout. However, the single knockout of IRC20 caused a threefold reduction in the targeted integration (from 4.1 to 1.3%) (Figure 4, B and C). Since the irc20 knockout reduced both SDSA and crossover, it is likely that Irc20 functions in the steps before D-loop formation (Figure 1, A2B2). We further tested whether the irc20 knockout suppressed the enhancement of the crossover by the srs2 knockout. The irc20 srs2 double knockout generated a fourfold reduction in crossover (from 4.1 to 1.1%), as compared to the wild type, which is in accordance with the level observed in the single irc20 knockout cells (Figure 4, B and C), indicating the complete suppression by irc20 of srs2-induced crossover enhancement. These results suggest that, contrary to our expectations, the Irc20 putative DNA helicase has a different function from the Srs2 DNA helicase in the SDSA pathway. Instead, the results support a model in which Irc20 works before D-loop processing by Srs2.
Irc20 is a novel factor for Mre11-directed DSB-end processing
The Mre11 protein forms the MRX complex with Rad50 and Xrs2 and processes DSB ends to initiate DSB repair by homologous recombination (Usui et al. 1998; Moreau et al. 1999). To investigate the genetic relationship between MRE11 and IRC20, we tested the effect of the irc20 knockout on SDSA and NHEJ production in the mre11 knockout strain. The mre11 single knockout caused a 52-fold reduction in SDSA (from 4.5 to 0.086%) (Figure 5, A and B), indicating that Mre11 promotes SDSA. The irc20 mre11 double knockout caused a 10-fold enhancement in SDSA (from 0.086 to 0.85%), as compared with the mre11 single knockout (Figure 5, A and B), indicating that the irc20 defect partially suppressed the mre11-induced SDSA inhibition. The single mre11 knockout also caused an 18-fold reduction in NHEJ (from 1.1 to 0.059%) (Figure 5, A and B), consistent with previous reports (Moore and Haber 1996; Lewis et al. 1999). The irc20 mre11 double knockout caused a sixfold enhancement in NHEJ (from 0.059 to 0.33%), as compared with the mre11 single knockout (Figure 5, A and B), indicating that the irc20 defect partially suppressed the mre11-induced NHEJ inhibition.
We next examined whether the irc20 single knockout causes hypersensitivity to DNA scission and whether it can suppress the mre11-induced hypersensitivity to DNA scission. As shown in Figure 6A, the irc20 single knockout did not show any sensitivity to EcoRI expression; however, the knockout completely suppressed the hypersensitivity of the mre11 knockout strain. As shown in Figure 6B, the irc20 single knockout did not show any sensitivity to bleomycin or other DNA scission reagents, such as MMS, HU, and CPT; however, the knockout suppressed the hypersensitivity of the mre11 knockout strain to these reagents.
We quantitatively examined the effects of the irc20 exo1 double knockout on the resistance to EcoRI expression and bleomycin treatment. The double mutant did not show strong sensitivity to either DNA scission treatment (Figure 7). Foster et al. (2011) reported that the elimination of Ku suppressed mre11 defects, in terms of the sensitivity to DNA scission reagents, and that this suppression was totally dependent on Exo1 function. To test whether the irc20 knockout-mediated suppression of mre11 defects requires Exo1 function, we quantitatively measured the effects of the exo1 knockout on the irc20-mediated suppression of EcoRI resistance and bleomycin resistance (Figure 7). In the presence of Exo1, the addition of the irc20 knockout to the mre11 mutant background showed a 4.7-fold increase in EcoRI resistance, and even in the absence of Exo1, it showed a 2.0-fold increase (Figure 7A). Such weak effects of the exo1 knockout were observed only with a low dose of bleomycin, and not with a higher dose (Figure 7B). These results show that the irc20-mediated suppression of mre11 defects weakly depends on Exo1 function, suggesting that this suppression is different from the ku-mediated suppression, in terms of Exo1 dependency, and involves other exonucleases or helicases.
We also tested the effect of the irc20 knockout on the crossover in the mre11 knockout strain. When the AUR1-C-bearing plasmid DNA, linearized by MscI digestion, was used to measure crossover (Figure 8A), the single knockout for MRE11 caused a 153-fold reduction in the crossover (from 4.4 to 0.029%) (Figure 8D). The irc20 mre11 double knockout showed a 23-fold enhancement in the crossover (from 0.029 to 0.67%), as compared with the mre11 single knockout (Figure 8D), indicating that the irc20 defect partially suppresses the mre11-induced crossover inhibition. In the absence of Exo1, the irc20 mre11 double knockout showed a fivefold enhancement in the crossover (from 1.2 to 5.8%), as compared with the mre11 knockout (Figure 8D), indicating that the irc20 defect-mediated suppression of the mre11-induced crossover inhibition partially depends on Exo1 function.
Finally, to evaluate whether Irc20 is involved in the early steps of the pathway where the resection of DSB ends occurs, we tested the effects of irc20 mutations on the crossover associated with longer-tract (more than 912-bp) gene conversion, using the MscI-digested pAUR101-I-SceI plasmid DNA (Figure 8B). In this crossover assay, AurR transformation requires the longer-tract crossover beyond the 18-mer insertion, 912 bp away from the DSB ends. Longer resection of DSB ends or longer branch migration of Holliday junctions is considered to lead to the longer-tract formation. The single knockout of IRC20 displayed a fourfold increase in the crossover associated with longer conversion (from 2.4 to 9.2%) (Figure 8E), in contrast to the twofold decrease in the crossover associated with non-longer conversion (from 4.4 to 2.0%) (Figure 8D). With regard to the longer-tract crossover production, a similar result was obtained using the single knockout of Exo1 (Figure 8E). These results suggest that Irc20 plays a role at an early step, where Exo1 functions. The mre11 single knockout caused a 60-fold reduction in the longer-tract crossover (from 2.4 to 0.041%) (Figure 8E). The irc20 mre11 double knockout generated a 3.6-fold enhancement in the longer tract crossover (from 0.041 to 1.5%), as compared with the mre11 single knockout (Figure 8E), indicating that the mre11-induced inhibition of longer-tract crossover is partially suppressed by the irc20 defect. In the absence of Exo1, the knockouts of irc20 and mre11 showed a 2.7-fold increase in the longer-tract crossover (from 0.23 to 0.62%), as compared with the mre11 knockout (Figure 8E), indicating that the irc20 defect-mediated suppression of the mre11-induced inhibition of longer-tract crossover was not strongly dependent on Exo1 function.
These results with respect to the two homologous recombination assays, the NHEJ assay and the DNA scission resistance, suggested the functional interaction between Irc20 and Mre11. The role of Irc20 is discussed later.
Roles of Srs2 in the SDSA pathway
We developed a novel yeast genetic assay to measure SDSA-mediated noncrossover, distinguished from noncrossover through double-Holliday intermediates and crossover. Our quantitative analysis clearly revealed that both Rad52 and Srs2 promote SDSA-mediated noncrossover. This result is consistent with the recombination mediator activity of Rad52 for the Rad51-catalyzed D-loop formation between an RPA-coated single-strand tail and a homologous double-strand DNA (Sung 1997). On the other hand, the Srs2 DNA helicase was suggested to prevent D-loop formation by disrupting the Rad51 nucleoprotein filament (Veaute et al. 2003). In addition, in a recent biochemical study using a synthetic D-loop (Figure 1, A2B2) (Dupaigne et al. 2008), the Srs2 helicase promoted the migration of a D-loop formed by Rad51 and Rad52 to stimulate SDSA. Our results showed that the Srs2 helicase actively stimulates SDSA (Figure 3) , prevents crossover (Figure 4), and thus support a model in which Srs2 acts on a D-loop formed by Rad51 and Rad52 to guide it into SDSA and to prevent the formation of a double-Holliday intermediate.
The roles of Rad52, Srs2, and Irc20 in NHEJ
The current genetic assay using the ura3 internal deletion allele allowed the quantification of NHEJ and the determination of whether precise or imprecise end joining occurred for each event. The single knockout for RAD52 did not reduce NHEJ. However, the lig4 rad52 double knockout significantly showed a 22-fold reduction in NHEJ, as compared to the lig4 single knockout. This indicates that Rad52 promotes more than 95% of NHEJ events, in the absence of the main Ku/Lig4-dependent pathway in this assay. Since imprecise end-joining events occurred at only a 6% frequency in lig4 mutants, Rad52 mostly promotes precise end joining between I-SceI-induced 3′-protruding DSB ends. Rad52 has potent single-strand DNA annealing activity (Mortensen et al. 1996), and this activity may contribute to single-strand annealing between the I-SceI-generated 3′-protruding ends. Other ligases and/or annealing proteins might be involved in imprecise end joining involving one or more base pairs (short microhomologies).
Teo and Jackson (1997) showed that EcoRI-induced DSB ends are rejoined imprecisely in about half of the end-joining products in lig4 mutants, in contrast to the results in the current assay. A major difference between these two assay systems is that the I-SceI-induced DSB ends in our assay are 3′ protruding, whereas the EcoRI-induced DSB ends in Teo and Jackson’s (1997) assay were 5′ protruding. We interpret this difference to mean that a 5′–3′ exonuclease might be involved in the processing prior to annealing, which may result in the discrepancy of the end-joining accuracy between 5′-protruding ends and 3′-protruding ends.
The single knockout for SRS2 or IRC20 did not reduce NHEJ (Figure 3), but increased the amount of imprecise products in NHEJ (Table 2). The imprecise NHEJ products contained small deletions, as in the case of the lig4 knockout. These results suggest that Srs2 and Irc20 ensure precise end joining in the Ku/Lig4-dependent pathway. In another NHEJ assay system, in which the DSB ends lack homology with any chromosomal sequences, the single mutation of rad52 or srs2 showed a significant reduction in the NHEJ frequency (Hegde and Klein 2000), suggesting that Rad52 and Srs2 promote the NHEJ pathway in parallel with the Ku/Lig4 pathway, in the absence of homology around the DSB. In the case where the DSB ends share homology with a chromosomal locus, as in our assay system, Rad52 and Srs2 predominantly act on the intermediate molecules formed along the homologous recombination pathways, rather than the NHEJ pathway.
Roles of the Irc20 protein in homologous recombination
IRC20 encodes a putative DNA/RNA helicase, and we expected it to replace the Srs2 function. The single irc20 mutation reduced SDSA-mediated noncrossover as much as the single srs2 mutation. The SDSA in the srs2 irc20 double mutant was almost the same as that in either single mutant. This indicates that Irc20 works in the SDSA pathway that involves Srs2. We used crossover assays to elucidate the step in the SDSA pathway where Irc20 functions, considering that SDSA-mediated noncrossover and double-Holliday-mediated crossover share steps up to D-loop formation in DSB repair (Figure 1, A1B1–A2B2). While the srs2 single mutation enhanced the targeted integration, the irc20 single mutation reduced the targeted integration and suppressed the srs2-induced enhancement of the targeted integration in srs2 irc20 double-knockout cells. These findings further support a model in which Irc20 works before D-loop formation, and Srs2 processes the D-loop to enhance SDSA and to prevent double-Holliday intermediate formation in the homologous recombination repair of a DSB.
The Mre11–Rad50–Xrs2 (MRX) complex and Exonuclease 1 are involved in DSB end processing, to form the 3′-single-strand DNA tails used for D-loop formation (Moreau et al. 2001; Lewis et al. 2002; Mimitou and Symington 2011). Recent studies in S. cerevisiae and in vitro have implied that the MRX complex with Sae2 removes short oligonucleotides from the 5′ end of a DSB, to initiate DSB-end resection. Subsequently, the short 3′-single-strand tail is extended in alternative manners, depending on either Exo1 or the Sgs1–Top3–Rmi1 complex with Dna2, to provide a longer 3′-single-strand DNA for D-loop formation (Gravel et al. 2008; Mimitou and Symington 2008; Zhu et al. 2008; Cejka et al. 2010; Nicolette et al. 2010; Shim et al. 2010). We observed that the irc20 single mutant showed approximately fourfold increase in longer-tract crossover (Figure 8E), as opposed to the effect on shorter-tract crossover (Figure 8D). To our surprise, we found that the knockout of IRC20 suppressed the mre11 knockout-induced hypersensitivities to DNA scission treatments (Figures 6 and 7) and the mre11 knockout-induced inhibition of SDSA, NHEJ (Figure 5), and crossovers (Figure 8). This irc20-mediated suppression of mre11 defects was partially dependent on Exo1 function (Figure 7 and Figure 8D), and thus Exo1 and other nucleases would act on DSB ends for resection in the absence of Mre11 and Irc20. Since the longer-tract crossover was increased by the loss of Irc20, and this increase was observed even in the absence of both Mre11 and Exo1 (Figure 8E), other nucleases would produce longer 3′ tails in the absence of Irc20. We interpret that Irc20 somehow causes the DSB ends to be predominantly directed to Mre11, rather than to other exonucleases, to induce the DSB-end resection. It might be important to express the checkpoint function of the MRX complex (Haber 1998). Another explanation for the increased longer-tract crossover in the absence of Irc20 is that the Irc20 helicase inhibits the branch migrations of Holliday junctions to produce shorter-tract crossovers.
The loss of Irc20 reportedly resulted in an increase in spontaneous Rad52 foci in diploid cells (Alvaro et al. 2007). Rad52 foci are formed on a 3′-single-strand tail generated by resection, to mediate D-loop formation by Rad51. We found that the loss of Irc20 causes the increase of the crossovers associated with longer-tract conversion (Figure 8E). We interpret these results to mean that in the absence of Irc20, other exonucleases would produce longer 3′-single-strand DNA at DSB ends, and then more Rad52 molecules would bind to the single-strand DNA region containing RPA proteins, resulting in more visible Rad52 foci.
The meaning of the suppression by DNA helicase defects for nuclease-deficient phenotypes
Previous Escherichia coli studies revealed that genetic defects in RecQ helicase suppressed the inhibition of DSB-induced crossover and the hypersensitivity to DNA scissions due to defects in a 5′- to 3′-single-strand DNA exonuclease, RecJ (Kusano et al. 1994; Lovett and Sutera-Jr 1995; Handa et al. 2009). In addition, a certain point mutation of the uvrD gene and the insertion mutation of the helD gene, encoding DNA helicase IV, suppressed the negative effects due to the recJ mutation (Lovett and Sutera-Jr 1995). The S. cerevisiae RECQ homolog, sgs1, is not another suppressor of the mre11 defects, because the mre11 sgs1 double mutant is more sensitive to X-ray irradiation than the single mre11 mutant (Budd and Campbell 2009). These genetic interactions revealed that a DNA helicase acts on the DNA substrate prior to the nuclease, to create a suitable substrate, and concurrently, to prevent the actions of other nucleases.
The authors thank Yu-ichi Sakoyama and Akihiro Fujita for plasmid construction and Takehiko Shibata for providing constructive criticism of this paper. This research was supported by a Grant-in-Aid from the Ministry of Education, Culture, Sports, Science and Technology of Japan to K.K.
Supporting information is available online at http://www.genetics.org/content/suppl/2012/02/23/genetics.112.139105.DC1.
Communicating editor: J. C. Schimenti
- Received December 1, 2011.
- Accepted February 1, 2012.
- Copyright © 2012 by the Genetics Society of America