Template switching during break-induced replication is promoted by the Mph1 helicase in Saccharomyces cerevisiae.

Chromosomal double-strand breaks (DSBs) that have only one end with homology to a donor duplex undergo repair by strand invasion followed by replication to the chromosome terminus (break-induced replication, BIR). Using a transformation-based assay system, it was previously shown that BIR could occur by several rounds of strand invasion, DNA synthesis, and dissociation. Here we describe a modification of the transformation-based assay to facilitate detection of switching between donor templates during BIR by genetic selection in diploid yeast. In addition to the expected recovery of template switch products, we found a high frequency of recombination between chromosome homologs during BIR, suggesting transfer of the DSB from the transforming linear DNA to the donor chromosome, initiating secondary recombination events. The frequency of BIR increased in the mph1Δ mutant, but the percentage of template switch events was significantly decreased, revealing an important role for Mph1 in promoting BIR-associated template switching. In addition, we show that the Mus81, Rad1, and Yen1 structure-selective nucleases act redundantly to facilitate BIR.

CHROMOSOMAL double-strand breaks (DSBs) are cytotoxic DNA lesions and if left unrepaired, or repaired incorrectly, can result in a loss of genetic information, genome instability, or even cell death. There are two main pathways to repair DSBs: homologous recombination (HR) and nonhomologous end joining (NHEJ). HR uses a homologous duplex to template repair of the break and is generally considered to be an error-free process; however, repair from a nonsister chromatid template can lead to loss of heterozygosity (LOH) or genome rearrangements (Symington 2002).

Initiation of HR requires degradation of the 5′-terminated strands to produce 3′-single-stranded DNA (ssDNA) tails that are first bound by replication protein A (RPA) and then replaced by Rad51 (Symington 2002). The Rad51 nucleoprotein filament searches for homologous sequences and promotes strand invasion to form a displacement loop (D-loop) (Figure 1). The 3′ end of the invading strand is used to prime DNA synthesis, templated by the donor duplex. The D-loop intermediate can be further channeled into one of several subpathways: synthesis-dependent strand annealing (SDSA), canonical double-strand break repair (DSBR), or break-induced replication (BIR). During SDSA, the invading strand that has been extended by DNA synthesis is displaced and anneals to complementary sequences exposed by 5′–3′ resection of the other side of the break. The remaining gaps can be filled by DNA synthesis and the nicks ligated to yield exclusively noncrossover products (Nassif ; Ferguson and Holloman 1996). If the second end of the break is captured by the D-loop, a double Holliday junction (dHJ) can be generated after DNA repair synthesis and ligation (Szostak ). dHJs can either be dissolved by the Sgs1–Top3–Rmi1 complex to form noncrossover products (Ira ; Wu and Hickson 2003) or resolved by structure-selective nucleases to generate crossovers or noncrossovers. Mus81–Mms4 is the main nuclease responsible for mitotic crossovers in Saccharomyces cerevisiae with Yen1 serving as a back-up function (Blanco ; Ho ; Agmon ; Muñoz-Galván ). Mus81–Mms4 is also proposed to act on the captured D-loop intermediate, prior to maturation to a dHJ, to form exclusively crossover products (Osman ; Mazón and Symington 2013).

Figure 1

Models for homology-directed DSB repair. DNA end resection produces 3′ ssDNA tails that are utilized by Rad51 for the homology search and strand invasion to form a D-loop intermediate. After extension of the 3′ end in the context of the D-loop, the invading strand is thought to be displaced by a DNA helicase, such as Mph1, and to anneal to sequences on the other side of the DSB (SDSA). If the displaced strand of the D-loop pairs with the other side of the DSB, a double Holliday junction (dHJ) intermediate is formed after gap filling and ligation (DSBR). The dHJ intermediate can be dissolved by the Sgs1–Top3–Rmi1 complex or resolved by structure-selective nucleases. If homology is restricted to one DSB end, DNA synthesis can occur in the context of the D-loop to the chromosome terminus, referred to as BIR.

BIR is a one-ended double-strand break repair mechanism in which DNA synthesis primed within the D-loop proceeds to the chromosome end, copying tens of kilobases (Morrow ; Malkova ). BIR is far more mutagenic than normal S-phase DNA synthesis and recent studies have shown the nascent strands are synthesized by a conservative mechanism, similar to other gene conversion reactions (Arcangioli 2000; Ira ; Deem ; Donnianni and Symington 2013; Saini ). BIR is the mechanism proposed for alternative lengthening of telomeres in the absence of telomerase and plays a minor role in repair of internal chromosome DSBs that have two ends with homology to a donor duplex (McEachern and Haber 2006; Ho ). However, it is a preferential pathway to generate LOH in old yeast cells (McMurray and Gottschling 2003), and microhomology-mediated BIR (MMBIR) has been implicated in genomic rearrangements (Hastings ).

BIR is most easily studied by creating a DSB where just one of the two ends can undergo homology-dependent strand invasion (Bosco and Haber 1998; Davis and Symington 2004; Malkova ; Lydeard ). In previous work from this laboratory, we described a plasmid-based assay to detect BIR in yeast. The chromosome fragmentation vector (CFV) contains the selectable marker, , , a tract of (G1-3T)n to provide a site for telomere addition and a unique DNA segment to target invasion of chromosomal sequence (Morrow ; Davis and Symington 2004). The CFV is linearized between the chromosome homology and telomere seeding sequence in vitro and used to transform yeast, selecting for Ura+ colonies. Most transformants arise by de novo telomere addition to heal one end of the CFV and strand invasion at the other end into the endogenous yeast locus to copy the entire chromosome arm, yielding a stable chromosome fragment (CF). When performed in diploid cells with polymorphic copies of chromosome III, ∼15% of the CFs recovered contained sequences from both chromosome homologs, suggesting the replication intermediate formed during BIR is unstable and the invading end can switch to a different template—a process referred to here as “template switching” (Smith ). Template switching was observed not only between chromosome homologs but also between ectopic homologies on different chromosomes, revealing the potential for this mechanism to generate genome rearrangements. Template switching has also been reported during BIR initiated at a chromosomal DSB, indicating that it is a general phenomenon and not restricted to the plasmid system (Ruiz ; Pardo and Aguilera 2012).

Template switching could be facilitated by helicases that displace the invading 3′ strand, as proposed for SDSA, or by cleavage of the D-loop intermediate by structure-selective nucleases (Pardo and Aguilera 2012). The obvious candidate for displacement of the 3′ strand during template switching is the Mph1 3′–5′ helicase, an ortholog of FANCM that displaces D-loops and extended D-loops following Rad51-mediated strand invasion (Sun ; Prakash ; Sebesta ). Interestingly, when is overexpressed in yeast, it inhibits BIR at a chromosomal DSB but not at telomeres (Luke-Glaser and Luke 2012). The Pif1 5′–3′ helicase is required for mitochondrial DNA maintenance and functions in nuclear genome stability by unwinding G quadruplexes and preventing de novo telomere addition at DSBs (Schulz and Zakian 1994; Ribeyre ; Bochman ; Paeschke ). Recent studies identified a role for Pif1 during BIR (Hu ; Saini ; Wilson ).

Here we describe a modification of the transformation assay to facilitate analysis of template switching associated with BIR in yeast. The assay is based on transformation of diploid yeast strains heteroallelic for or with CFVs that invade just upstream of the or locus. In addition to the expected recovery of CFs with a functional copy of or by template switching, recombination between the chromosome homologs targeted by the CFV occurred at a high frequency during BIR. The percentage of Ade+ transformants was significantly decreased in the Δ and mutants, and most of the Ade+ events recovered from the Δ mutant were due to chromosomal recombination, revealing an important role for Mph1 in template switching associated with BIR.

Materials and Methods

Media, growth conditions, and genetic methods

Rich medium (1% yeast extract, 2% peptone, 2% dextrose) (YPD), synthetic complete medium (SC) lacking the appropriate amino acids or nucleic acid bases, sporulation medium, and genetic methods were as described previously (Sherman ). Rich medium supplemented with 2% galactose instead of dextrose was used for induction of endonuclease. Standard procedures were used for genetic crosses and transformation (Ito ; Sherman ).

Yeast strains and plasmids

S. cerevisiae strains used in this study are derivatives of W303 (Table 1). Strains containing ade2-I, ade2-n, his3Δ::hphMX4, his3Δ::natMX4, leu2ΔB, ΔR, Δ::kanMX6, Δ::kanMX6, rad1::LEU2, Δ::HIS3, and -m2 alleles were described previously (Mortensen ; Wagner ; Ho ). All deletion alleles are denoted by ∆ in the text. The locus on the left arm of chromosome XV (Chr XV) was replaced by in the -n strains. Diploids were made by crossing appropriate haploids to generate strains that were /met22::TRP1 -I/-n Δ::hphMX4/Δ::natMX4 or ΔR/ΔB.

Table 1

Yeast strains

Strain	Genotype	Source
LSY2165-11C	MATα ade2-I his3Δ::hphMX4 mus81Δ::kanMX6	Ho et al. (2010)
LSY2165-30B	MATa ade2-I his3Δ::hphMX4 mus81Δ::kanMX6 yen1Δ::HIS3	Ho et al. (2010)
LSY2240	MATa leu2::ΔEcoRI lys2	Mortensen et al. (2002)
LSY2241	MATα leu2::ΔBsteII trp1-1	Mortensen et al. (2002)
LSY2332-2B	MATα ade2-I his3Δ::hphMX4 mph1Δ::kanMX6	Mazón and Symington (2013)
LSY2543	MATa met22::KlTRP1 ade2-n his3Δ::natMX4	This study
LSY2633	MATa met22::KlTRP1 ade2-n his3Δ::natMX4 mus81Δ::kanMX6	This study
LSY2638	MATα ade2-I his3Δ::hphMX4	This study
LSY2650	MATα ade2-I his3Δ::hphMX4 mph1Δ::kanMX6	This study
LSY2653	MATa met22::KlTRP1 ade2-n his3Δ::natMX4 mph1Δ::kanMX6	This study
LSY2708-1D	MATa ade2-I his3Δ::hphMX4 pif1-m2	This study
LSY2709-4C	MATα met22::KlTRP1 ade2-n his3Δ::natMX4 pif1-m2	This study
LSY2738-15B, 16D	MATα ade2-I his3Δ::hphMX4 mlh1Δ::kanMX6	This study
LSY2739-8B	MATα ade2-I his3Δ::hphMX4 mph1Δ::kanMX6 pif1-m2	This study
LSY2739-16A	MATa ade2-I his3Δ::hphMX4 mph1Δ::kanMX6 pif1-m2	This study
LSY2740-8B, 19B	MATα met22::KlTRP1 ade2-n his3Δ::natMX4 mus81Δ::kanMX6 yen1Δ::HIS3	This study
LSY2802-5B	MATa met22::KlTRP1 ade2-n his3Δ::natMX4 mph1Δ::kanMX6	This study
LSY2803-7C	MATa met22::KlTRP1 ade2-n his3Δ::natMX4 mlh1Δ::LEU2	This study
LSY2804-5D	MATα met22::KlTRP1 ade2-n his3Δ::natMX4 mph1Δ::kanMX6 pif1-m2	This study
LSY2804-17A	MATa met22::KlTRP1 ade2-n his3Δ::natMX4 mph1Δ::kanMX6 pif1-m2	This study
LSY2950	MATa/MATα leu2::ΔEcoRI/leu2::ΔBsteII lys2/LYS2 trp1-1/TRP1	This study
LSY2951	MATa/MATα ade2-I/ade2-n his3Δ::hphMX4/his3Δ::natMX4 MET22/ met22::KlTRP1	This study
LSY2332-2B × LSY2802-5B, LSY2650 × LSY2653	Identical to strain LSY2951 except mph1Δ::kanMX6/mph1Δ::kanMX6	This study
LSY2708-1D × LSY2709-4C	Identical to strain LSY2951 except pif1-m2/pif1-m2	This study
LSY2739-8B × LSY2804-17A, LSY2739-16A × LSY2804-5D	Identical to strain LSY2951 except mph1Δ::kanMX6/mph1Δ::kanMX6 pif1-m2/pif1-m2	This study
LSY2165-11C × LSY2633	Identical to strain LSY2951 except mus81Δ::kanMX6/mus81Δ::kanMX6	This study
LSY2165-30B × LSY2740-8B, LSY2165-30B × LSY2740-19B	Identical to strain LSY2951 except mus81Δ::kanMX6/mus81Δ::kanMX6 yen1Δ::HIS3/ yen1Δ::HIS3	This study
LSY2738-15B × LSY2803-7C, LSY2738-16D × LSY2803-7C	Identical to strain LSY2951 except mlh1Δ::LEU2/mlh1Δ::LEU2	This study
LSY2767-72D	MATa-inc lys2::natMX4 AVT2::lysΔ3′-HOcs::kanMX6 SYN8::TRP1-5′Δys2 (Chr. I 128 kb) ade3::PGAL-HO bar1::LEU2	Donnianni and Symington (2013)
LSY2884-53D	Identical to strain LSY2767-72D except mph1Δ::kanMX6	This study
LSY2884-188C	Identical to strain LSY2767-72D except pif1-m2	This study
LSY2885-297B	Identical to strain LSY2767-72D except URA3::GPD-TK7	This study
LSY2885-160B, 248A	Identical to strain LSY2767-72D except URA3::GPD-TK7 mus81Δ::kanMX6	This study
LSY2885-121B, 127C	Identical to strain LSY2767-72D except URA3::GPD-TK7 yen1Δ::HIS3	This study
LSY2885-294A	Identical to strain LSY2767-72D except mus81Δ::kanMX6 yen1Δ::HIS3	This study
LSY2885-236A	Identical to strain LSY2767-72D except mus81Δ::kanMX6 rad1::LEU2	This study
LSY2886-1C, 51B	Identical to strain LSY2767-72D except URA3::GPD-TK7 rad1::LEU2	This study
LSY2921-10B, 42B	Identical to strain LSY2767-72D except mus81Δ::kanMX6 rad1::LEU2 yen1Δ::HIS3	This study
LSY2921-93D	Identical to strain LSY2767-72D except rad1::LEU2 yen1Δ::HIS3	This study

All strains are derived from the W303 background (ade2-1 can1-100 his3-11, 15 leu2-3, 112 trp1-1 ura3-1); only mating type and differences from this genotype are shown.

CFVs used in this study are ARS derivatives of CFV/D8B-tg (Davis and Symington 2004; Marrero and Symington 2010). Vectors pLAG6 and pLAG10 contain inserts from the right arm of Chr XV (Saccharomyces Genome Database, SGD coordinates 558476–563456 and 551748–556242, respectively) while vectors pLS192 and pLAG8 contain sequences from the left arm of Chr III (SGD coordinates 96821–102096 and 93450–98535, respectively).

Transformation-based template switching assays

ade2 assay:

A total of 500 ng of BglII-digested pLAG6 or pLAG10 was used to transform competent yeast cells, selecting for Ura+ transformants. The percentage of Ura+ Ade+ (white) was determined from the total number of Ura+ transformants. All Ura+ Ade+ transformants were further analyzed by patching onto YPD medium to allow growth in nonselective conditions and then replica plating to 5-fluoroorotic acid (5-FOA)-containing medium to select for cells that became Ura− due to loss of the CF (Boeke ). Cells that lost the CF were scored again for their color, allowing distinction between 5-FOA red (Ade−; W/R; template switching) and 5-FOA white (Ade+; W/W; BIR-associated chromosomal recombination) transformants, and scored on SC −Met, SC −Trp, YPD +nourseothricin (Nat), and YPD +hygromycin B (Hyg) for marker loss. Ura+ Ade− transformants derived from pLAG6 in wild type, Δ, and Δ strains were also scored for the heterozygous markers on the left and right arms of Chr XV after CF loss. Pulsed-field gel electrophoresis of intact chromosomal DNA prepared from Ura+ transformants was used to confirm formation of full-length CF products (Schwartz and Cantor 1984; Davis and Symington 2004).

leu2 assay:

A total of 500 ng of linearized CFVs pLS192 and pLAG8, digested with SnaBI and BglII, respectively, was used to transform competent yeast cells, selecting for Ura+ transformants. Transformants were replica plated on SC −Leu medium to determine the percentage that were Leu+. Ura+ Leu+ transformants were further analyzed by patching on YPD, to allow growth in nonselective conditions and replica plated on 5-FOA to select for Ura− cells that lost the CF, followed by replica plating to SC −Leu to distinguish between Leu− (template switching) and Leu+ (BIR-associated chromosomal recombination) transformants.

Statistical significance was determined after totaling Ade+ and Leu+ transformants from several independent transformations using two-tailed Pearson’s chi-squared test (χ2; http://in-silico.net/). For the assay, the numbers of Ade+ Ura+ transformants were compared to Ade− Ura+ transformants (Figure 2D, Figure 4A, Figure 6A, and Supporting Information, Figure S3), while for the assay Leu+ Ura+ transformants were compared to the number of Leu− Ura+ transformants (Figure 2D). To determine statistical significance of template switching vs. chromosomal recombination, the number of template switching transformants (W/R on 5-FOA) was compared to the number of BIR-associated chromosomal recombinants (W/W on 5-FOA; Figure 3A, Figure 4B, Figure 6B, and Figure S3). To determine the statistical significance of LOH for a specific transforming DNA and strain after growth of transformants on 5-FOA-containing medium, the number of transformants that lost Hyg or Nat resistance (LOH transformants) was compared to those with no LOH (Figure 3C and Figure S1).

Figure 2

Template-switching assays. (A) Schematic showing the invasion sites for pLAG6 and pLAG10 on Chr XV and pLAG8 and pLS192 on Chr III. (B) BIR efficiencies for the indicated vectors in wild type. Transformations were performed five to nine times and mean values are presented with standard deviation (SD). (C) The simplest explanation for generation of Ura+ Ade− transformants is by invasion of one of the two Chr XV homologs by the CFV, and the generation of Ura+ Ade+ transformants could result from invasion of ade2-I, displacement, and invasion of the ade2-n allele. (D) Percentage of Ade+ or Leu+ transformants in wild type for both assays. pLS192 serves as control in the ade2 assay (invasion of a non-ade2-containing chromosome), and pLAG6 is a control in the leu2 assay (invasion of a non-leu2-containing chromosome).

Figure 4

Mph1 promotes template switching during BIR. (A) Percentage of Ura+ Ade+ transformants and (B) template switching in the wild type (WT), mph1Δ, pif1-m2, and mph1Δ pif1-m2 mutants transformed with linearized pLAG6. Total number of Ade+ Ura+ events analyzed for each strain is shown in parentheses. (C) Chromosomal BIR assay. A DSB 34 kb from the left telomere of Chr V is induced by plating on medium containing galactose. After DSB induction, the kanMX6 gene and the chromosomal sequences distal to the break are lost, while Chr V invades the left arm of Chr I, and DNA synthesis restores a functional LYS2 gene. The orange boxes indicate the extent of homology shared by the lysΔ3′ and 5′Δys2 cassettes. (D) BIR frequency of the indicated mutants. Assays were performed five to six times and mean values with SD are presented.

Figure 6

Role of structure-selective nucleases in BIR. Percentage of Ura+ Ade+ transformants (A) and template switching (B) in the mus81Δ mutant transformed with linearized pLAG6. Total number of Ade+ Ura+ events analyzed for each strain is shown in parentheses. (C) BIR frequencies of the indicated strains using the chromosomal assay. Assays were performed five to six times for each strain and mean values with SD are presented.

Figure 3

Chromosomal recombination associated with BIR. (A) Analysis of Ade+ transformants on 5-FOA plates to distinguish between template switching and BIR-associated chromosomal recombination. (B) Percent template switching for both assays in the wild type. Total number of Ade+ Ura+ and Leu+ Ura+ events analyzed for each vector is shown in parentheses. (C) LOH for the natMX4 and hphMX4 markers for wild type transformed with pLAG6 or pRS416. For pLAG6, Ura+ transformants were analyzed following loss of the CF by plating on 5-FOA-containing medium. R, Ade− transformants; W/R, Ade+ transformants that revert to Ade− after CF loss; and W/W, Ade+ transformants that remain Ade+ after CF loss.

BIR efficiency was determined using 500 ng of the specified linearized CVF or 100 ng of circular replicative plasmid pRS416 for each transformation, selecting for Ura+ transformants. BIR efficiency is expressed as the ratio of Ura+ transformants obtained per nanogram of CFV and per nanogram of pRS416. The experiments were repeated at least three times for each CFV and strain. Mean BIR efficiencies were compared using Welch two-sample t-test from R Statistical software.

Chromosomal assay for BIR

Two-milliliter YPD cultures were grown overnight at 30°. Cells were diluted and regrown to OD600 0.3–0.8 in YP with 2% raffinose and adjusted to an OD600 of 0.3. Aliquots of serial dilutions were plated on YPD, and SC −Lys and, for galactose induction, on YPGal plates and grown for 2–3 days. Colonies that grew on YPGal plates were replica plated to SC −Lys plates to ensure reconstitution of the functional gene by BIR. The percent BIR rate was calculated as the ratio of colony-forming units (CFUs) on SC −Lys after induction and CFUs on YPD (Donnianni and Symington 2013). Significance was determined using the Student’s t-test.

Results

A genetic assay to detect template switching during BIR

We developed a simple genetic assay to screen candidate genes for their roles in template switching during BIR. Diploid yeast strains heteroallelic for or mutations were transformed using linearized CFVs that invade upstream of the or loci and the resulting transformants were tested for adenine or leucine prototrophy, respectively. The strains for the assay contain and alleles on Chr XV homologs (Mozlin ); the mutations are 0.95 kb apart and both inactivate , resulting in accumulation of a red pigment. CFVs pLAG6 and pLAG10 contain sequences homologous to Chr XV to promote strand invasion 1 kb and 8 kb upstream of the locus, respectively (Figure 2A). To confirm the generality of results from the assay, heteroalleles were used. The and mutations are 0.6 kb apart and inactivate on Chr III (Mortensen ). CFVs pLAG8 and pLS192, which invade 1 and 4 kb upstream of the region, respectively, were used for transformation (Figure 2A).

We first analyzed the BIR efficiency for both assays in cells with normal recombination functions (“wild type”); pLAG8 and pLAG192 have 30% higher BIR efficiency than pLAG6 and pLAG10 (P = 0.013 and P = 0.008, respectively). This difference could be explained by the shorter region of DNA that needs to be copied in the assay as compared to the assay (100 kb vs. 560 kb, respectively; Figure 2B) (Donnianni and Symington 2013).

Transformation of an heteroallelic diploid with pLAG6 or pLAG10 results in generation of mainly red (Ade−) Ura+ transformants. The simplest mechanism to generate white Ade+ Ura+ transformants would be by invasion of Chr XV containing the allele by the CFV and copying the 3′ part of the gene followed by displacement of the invading strand and invasion of the allele copying the 5′ part of the gene, resulting in formation of a functional gene carried on the CF (Figure 2C), though other mechanisms are possible. pLAG6, which invades closer to the locus than pLAG10, resulted in 2.6-fold higher frequency of Ade+ Ura+ transformants compared to pLAG10 (P < 0.0001), in accordance with the previously observed 10-kb window from the invasion site in which template switching occurs (Smith ). In the assay, pLAG8 yielded 1.8-fold more Ura+ Leu+ transformants than pLS192 (P < 0.0001). The frequency of Ade+ Ura+ transformants using pLAG6 was 22-fold higher as compared to pLS192 (P < 0.0001) while the frequency of Leu+ Ura+ transformants with pLAG8 was 28-fold higher than with pLAG6 (P < 0.0001), confirming that Ade+ Ura+ or Leu+ Ura+ transformants are a consequence of BIR upstream of the test locus and not caused by a genome-wide increase in spontaneous mitotic recombination during BIR (Figure 2D). Similarly, Ade+ Ura+ or Leu+ Ura+ transformants were only rarely recovered from transformation with the replicating plasmid pRS416. We acknowledge that only a small fraction of invasion, displacement, and reinvasion events are likely to generate Ade+ or Leu+ prototrophs and many template switch events will be genetically invisible.

BIR-induced chromosomal recombination

The Ade+ or Leu+ events due to template switching should have the or locus on the CF and after loss of the CF the cells would be expected to revert to an Ade− or Leu− phenotype. To test this prediction, Ade+ Ura+ or Leu+ Ura+ transformants were grown under nonselective conditions, replica plated to medium containing 5-FOA to select for Ura− cells that had lost the CF, and then analyzed for adenine or leucine prototrophy (Figure 3A). Surprisingly, only 47% or 19% of Ade+ transformants with pLAG6 or pLAG10, respectively, contained a CF-linked gene indicative of template switching. For the assay, 70% and 49% of Leu+ transformants with pLAG8 and pLS192, respectively, were Leu− after CF loss (Figure 3B). Thus, a significant fraction of Ade+ and Leu+ transformants are due to BIR-induced recombination between chromosome homologs.

Induction of recombination between homologs was previously reported during plasmid gap repair and suggested to result from mismatch repair of a heteroduplex DNA intermediate (Silberman and Kupiec 1994). However, another possibility is that the D-loop intermediate is cleaved by a structure-selective nuclease effectively transferring the DSB from the linear fragment to the donor chromosome, initiating recombination with the homolog. The diploids have additional heterozygous markers on Chr XV to evaluate chromosome loss or LOH for markers on the left or right arm (Ho ). The natMX4 and hphMX4 markers are inserted 150 kb centromere proximal to the locus. We expected heterozygosity for the flanking markers to be retained during BIR. However, if BIR-induced a chromosomal DSB, which initiated recombination between homologs, we would expect a significant fraction of the Ade+ chromosomal recombinants to exhibit LOH for the natMX4 and hphMX4 markers (Nickoloff ; Barbera and Petes 2006; Ho ). Seven of 48 (14.6%) white 5-FOA resistant colonies (W/W) derived from pLAG6 exhibited LOH for Nat or Hyg resistance (Figure 3C). Only 2 of 40 red 5-FOA resistant colonies (W/R) exhibited LOH for natMX4 or hphMX4, but due to the low sample size this is not significantly less than observed for the chromosomal Ade+ events (W/W). Because the number of Ade+ transformants was low, we also analyzed Ura+ Ade− transformants to determine whether there was a general increase in LOH involving Chr XV homologs during BIR, regardless of whether template switching or chromosomal recombination occurred within the locus. More than 500 Ade− Ura+ transformants derived from pLAG6 (R) were passaged on 5-FOA to lose the CF and then tested for LOH of the Chr XV markers. Of these, 31 (6.1%) exhibited LOH for Nat or Hyg resistance, one colony was Trp− Nats, indicative of chromosome loss, and one colony was Met−. The frequency of natMX4 or hphMX4 marker loss was significantly higher for the Ade+ colonies due to chromosomal recombination than for Ade− BIR events (R) (P = 0.036). Ura+ colonies obtained from transformation with pRS416 were also analyzed for marker loss and only 2/1185 tested showed LOH. The significant increase in LOH for the natMX4 or hphMX4 markers following transformation with pLAG6 compared with pRS416 (P < 0.0001) is consistent with induction of recombination between homologs during BIR.

Mph1 contributes to template switching during BIR

During SDSA, the extended invading 3′ end is displaced by helicases and, as proposed previously, displacement of the invading strand could facilitate template switching during BIR (Smith ; Llorente ). Therefore, we expected the inactivation of helicases that dissociate D-loop intermediates to increase the frequency of BIR, but to decrease the frequency of Ade+ Ura+ transformants. Mph1 has emerged as the main helicase responsible for SDSA in yeast since it is able to unwind Rad51 generated D-loops and DNA polymerase extended D-loops in vitro, and the Δ mutant displays increased crossovers during mitotic recombination (Sun ; Prakash ; Tay ; Sebesta ; Mazón and Symington 2013; Mitchel ). The percentage of Ade+ Ura+ transformants derived from pLAG6 was reduced by 3.9-fold in the Δ mutant as compared to wild type (0.68%; P < 0.0001; Figure 4A). Furthermore, analysis of CF-cured Ade+ Ura+ transformants from the Δ mutant revealed that only 23.8% were due to template switching, 2-fold less than wild type (P = 0.0029; Figure 4B). These results are in agreement with the role of Mph1 in the displacement of D-loops (Figure 5A). To determine whether maintenance of heterozygosity during BIR was affected by the Δ mutation, 542 Ade− Ura+ transformants derived from pLAG6 were grown on 5-FOA-containing medium to lose the CF and then tested for loss of Nat or Hyg resistance. The frequency of LOH for the natMX4 and hphMX4 markers was 7.4%, not significantly different from wild type (Figure S1).

Figure 5

Models for template switching and chromosomal recombination associated with BIR. (A) The 3′ end of the CFV invades one homolog, and, after DNA synthesis to copy part of the ade2 locus, is displaced by Mph1. In the absence of homologous sequences to pair with, the extended 3′ end invades the ade2 locus of either homolog; if a different homolog is used, DNA synthesis can result in generation of a functional ADE2 sequence. Continued replication to the chromosome terminus generates a CF. In the mph1Δ mutant, dissociation of the D-loop intermediate is decreased and replication from the initial strand invasion event proceeds to the telomere, resulting in fewer Ade+ products. (B) If the initial strand invasion intermediate is cut by structure-selective nucleases, the CFV can become covalently linked to chromosomal sequences downstream of the site of strand invasion (half crossover). The resulting truncated chromosome can then invade the other homolog, generating a functional copy of ADE2 and resulting in homozygosis of the downstream markers, or invade the newly formed CF (not shown) retaining heterozygosity for natMX4 (n) and hphMX4 (h) resistance.

Pif1 promotes BIR but does not affect template switching

We measured template switching in the mutant since previous studies have shown a decreased frequency of BIR in mutants (Chung ; Hu ; Wilson ). The allele expresses a form of Pif1 that retains mitochondrial activity, but is largely defective for nuclear function (Schulz and Zakian 1994). The percentage of Ade+ Ura+ transformants derived from pLAG6 was reduced by eightfold in the strain (0.31%; P < 0.0001), and even further decreased in the mutant (0.16%, P = 0.031 and P < 0.001 compared to and Δ strains, respectively; Figure 4A), suggesting an additive effect of and . However, the proportion of Ade+ events due to template switching was the same for and wild-type strains. The number of Ade+ events recovered from the Δ pif1-m2 strain was very low and the percent due to template switching was not significantly different from the single mutants or wild type (Figure 4B).

Pif1 is a negative regulator of telomerase and the frequency of de novo telomere addition increases in the mutant (Schulz and Zakian 1994; Boule ). To eliminate the possibility that Ura+ transformants were generated by telomere addition at both ends of the linear vector, independent Ura+ transformants were analyzed by pulsed-field gel electrophoresis and all were confirmed to have full-length CF products (Figure S2). The heterozygous markers on the left and right arms of Chr XV can be used to monitor half-crossover events as an alternative product to BIR. Surprisingly, of 23 Ade+ events analyzed from the mutant, only one was consistent with formation by a half crossover; thus, most of the CFs recovered result from BIR and not from half crossovers.

We could not accurately assess the BIR frequency of the mutant using the plasmid-based assay due to more efficient telomere healing at the telomere-seed sequence of the linear CFV leading to an artificially high transformation frequency (data not shown) (Schulz and Zakian 1994). To more accurately determine the BIR frequency of the Δ and mutants, a chromosomal assay was used. The chromosomal assay is based on reconstitution of a functional gene by BIR (Donnianni and Symington 2013). A recipient cassette with a 3′-truncated gene (designated lysΔ3′), a 36-bp HO endonuclease cut site (HO cs), and kanMX6 to confer resistance to geneticin (G418) was inserted 34 kb from the left telomere of Chr V. There are no essential genes distal to the site of the insertion. A donor cassette composed of and a 5′ truncation of (denoted by 5′Δys2) was inserted 128 kb from the left telomere of Chr I (Figure 4C). The truncated Δ3′ and 5′Δys2 genes share 2.1-kb homology. These donor and recipient cassettes were incorporated into a haploid strain with a deletion of the native locus, with the MATa-inc allele to prevent cleavage at the endogenous HO cs and which is capable of expressing from a galactose-regulated promoter. After DSB formation, the lysΔ3′ sequence invades the donor cassette and copies to the end of Chr I, generating a Lys+ nonreciprocal translocation product (Figure 4C). The other side of the Chr V DSB lacks homology to sequences in the yeast genome and is degraded, resulting in loss of the kanMX6 gene and sensitivity to G418. The frequency of BIR was determined by the plating efficiency of cells on galactose-containing medium ( constitutively expressed) that are Lys G418s relative to medium with glucose ( off). The BIR frequency was reduced by 1.9-fold in the pif1-m2 mutant (P = 0.0076), whereas the BIR frequency increased by 2.4-fold in the Δ background (P = 0.0002) (Figure 4D). These results confirm previous studies showing positive and negative roles of and , respectively, during BIR (Luke-Glaser and Luke 2012; Hu ; Saini ; Wilson ).

MUS81 is not required for BIR-associated template switching or chromosomal LOH

The increased LOH of the natMX4 and hphMX4 markers during BIR is consistent with transfer of the DSB from the CFV to the chromosome to initiate interhomolog recombination. Structure-selective nucleases could cleave the strand invasion intermediate resulting in several possible outcomes. Mus81–Mms4 cleavage of the D-loop, followed by gap filling and ligation, would be predicted to covalently link the CFV to the donor chromosome (Figure 5B). Subsequent cleavage of the front end of the D-loop would result in a half crossover, creating a CF and truncated donor chromosome. The broken chromosome could then engage in BIR with the chromosome homolog, resulting in LOH for natMX4 or hphMX4 and the potential to form an Ade+ recombinant (Figure 5B). Alternatively, the truncated chromosome could invade the newly formed CF, retaining heterozygosity for natMX4 and hphMX4 after CF loss. The frequency of Ade+ recombinants and the percent of Ade+ events due to chromosomal recombination in the Δ mutant were not significantly different from wild type (Figure 6). We also analyzed Ade− Ura+ transformants from the Δ mutant after growth on 5-FOA-containing medium to lose the CF; the frequency of LOH for Nat or Hyg resistance was 4.1%, not significantly different from wild type (Figure S1).

The failure to detect a decrease in chromosomal Ade+ recombinants in the absence of Mus81 could be due to Yen1 cleavage of the BIR intermediate (Blanco ; Ho ; Agmon ; Muñoz-Galván ). Because the Δ Δ diploid grows very slowly and has low transformation efficiency we were unable to generate sufficient Ura+ Ade+ transformants for meaningful analysis. Furthermore, the spontaneous chromosome instability of the Δ Δ resulted in a high frequency of LOH (∼2%) even in cells transformed with the replicative plasmid, pRS416, making it difficult to identify events associated with BIR (data not shown). Since the growth defect conferred by Δ and Δ is less in haploids than in diploids (Ho ), we measured the BIR frequency of Δ and Δ mutants using the haploid chromosomal system. The Δ and Δ mutations did not affect the BIR frequency and no significant decrease was observed for the Δ Δ double mutant. We also tested the effect of deleting because the BIR assay involves recombination between limited homology substrates, and Rad1–Rad10 can be important to cleave intermediates with a branched structure adjacent to a heterology (Lyndaker and Alani 2009; Mazón ). The BIR frequency was lower in the Δ Δ double mutant than in wild type (P = 0.0003) and reduced further in the Δ Δ Δ triple mutant as compared to the Δ Δ mutant (P = 0.0004) (Figure 6C), indicating that the nucleases have overlapping roles in promoting BIR.

Mismatch repair does not influence template switching or Ade+ chromosomal recombination

Ade+ chromosomal recombinants could potentially occur through two rounds of mismatch repair of a heteroduplex DNA intermediate formed by pairing of an extended single strand from the CFV and one of the chromosome homologs (Ray ; Silberman and Kupiec 1994). Therefore, we tested if eliminating the Mlh1 protein, an essential component of the mismatch repair system, which plays no role in processing branched DNA intermediates (Harfe and Jinks-Robertson 2000), would change either the percentage of Ade+ transformants or the template switching/BIR-associated chromosomal recombination ratio. In both cases, data from the mutant were the same as with wild type, with 1.78% Ade+ transformants and 47.02% due to template switching (Figure S3), indicating that mismatch repair has no apparent role in template switching or chromosomal recombination during BIR.

Discussion

It has been previously shown that switching between DNA templates can occur during BIR and if it takes place between ectopic homologous sequences can result in genome rearrangements (Smith ; Ruiz ). The transformation-based assay using heteroallelic diploids described here allows simplified detection of template switching events by a change of phenotype. Still, only those events that occur by switching between the and mutations (located 0.95 kb apart) are detected as Ade+ recombinants and many template switch events are expected to be genetically invisible. In addition to template switching, we found a high frequency of events in which one of the chromosomal loci became Ade+, reminiscent of earlier studies reporting “tripartite” recombination induced by a DSB on an ectopic template (Ray ; Silberman and Kupiec 1994). Here, we investigated the roles of the helicases encoded by and in template switching associated with BIR.

The Mph1 helicase is known to catalyze dissociation of the invading strand of Rad51-generated D-loops or extended D-loops in vitro and consequently would be predicted to facilitate template switching during BIR (Sun ; Prakash ; Sebesta ). In agreement with a previous study (Luke-Glaser and Luke 2012) we found the frequency of BIR to be increased in the Δ mutant, which we interpret to be due to Mph1-catalyzed dissociation of the invading 3′ end and the need for additional strand invasion events to complete BIR. The percentage of Ura+ transformants that were Ade+ was reduced in the Δ mutant and template switching among the Ade+ events was also decreased, consistent with the need for Mph1 to dismantle the D-loop for template switching. In the Δ mutant, the more stable D-loop intermediate might more readily transition to a migrating D-loop after recruitment of Pif1 (Wilson ) or be cut by structure-selective nucleases leading to secondary recombination events. The Ade+ transformants recovered from the Δ mutant were mainly due to chromosomal recombination, consistent with the idea that the stabilized D-loop is cleaved by structure-selective nucleases. If the D-loop were cut by nucleases to create a half crossover, linking the CFV to sequences downstream of the site of strand invasion, the fragmented chromosome would likely engage with the homolog to repair the one-ended DSB by BIR, resulting in homozygosis of the downstream marker. Indeed, we found that 6.1 and 7.4% of all Ura+ transformants derived from pLAG6 in the wild-type and Δ strains, respectively, exhibited LOH for the natMX4 or hphMX4 markers after loss of the CF, suggesting secondary BIR events between a fragmented Chr XV and its homolog.

The frequency of BIR using a Chr III disome assay is decreased in the Δ mutant and the products recovered are mainly due to half crossovers, similar to the phenotype observed for the Δ mutant ( encodes a nonessential subunit of DNA polymerase δ) (Deem ; Smith ; Chung ; Saini ; Wilson ). Loss of Pif1 also prevents Rad51-dependent telomere elongation in the absence of telomerase (Hu ). In a reconstituted in vitro assay, Pif1 was shown to function with the DNA polymerase δ complex to extend the 3′ end of a Rad51-generated D-loop (Wilson ). Pif1 facilitated extensive DNA synthesis by liberating the newly synthesized ssDNA to establish a migrating D-loop. We could not accurately assess the BIR frequency in the mutant using the CFV system because the increased efficiency of telomere elongation from the minimal telomere seed sequence of the CFV counteracted the decrease in BIR (data not shown). A more accurate assessment of the BIR frequency was obtained using the chromosomal BIR assay, revealing a twofold decrease in the mutant. The pif1-m2 mutant is reported to retain some nuclear function and this might account for the higher BIR frequency than reported using the Δ mutant (Schulz and Zakian 1994; Hu ; Wilson ). The fraction of Ade+ events among Ura+ transformants was significantly reduced in the pif1-m2 strain, but we did not observe a change in the ratio of template switching and chromosomal recombination associated with BIR, suggesting Pif1 plays no role in D-loop dissociation.

In agreement with a previous study (Silberman and Kupiec 1994), we observed a high frequency of recombination between chromosome homologs induced by the linear transforming DNA. Ray also reported stimulation of triparental recombination by a chromosomal DSB. In the previous studies, chromosomal conversion was suggested to result from mismatch repair of an hDNA intermediate; however, we observed no effect on the frequency of this class of events in an Δ mutant. Instead, the high frequency of LOH for the natMX4 and hphMX4 markers is more consistent with transfer of the DSB from the CFV to the chromosome triggering secondary recombination events. A high frequency of LOH for Nat or Hyg resistance was observed for all Ura+ transformants derived from pLAG6, regardless of whether they were Ade+ or Ade−, suggesting a significant number of BIR intermediates are cleaved by structure-selective nucleases. Mus81–Mms4 cleaves model D-loop structures in vitro and could create the first cleavage necessary to form a half crossover with another nuclease cutting the 5′ flap represented by the leading edge of the D-loop to complete the event (Osman ). Slx1–Slx4 is able to cleave a variety of branched structures, including 5′ flaps, and Yen1/GEN1 also cuts 5′ flaps in addition to HJs (Fricke and Brill 2003; Rass ). The frequency of LOH for total Ura+ transformants (after selection for loss of the CF) from the Δ mutant was not significantly less than observed for wild type, and chromosomal recombinants were recovered at a similar frequency to wild type. Thus, Mus81–Mms4 either does not play a major role in formation of half crossovers or Yen1, Rad1–Rad10, and Slx1–Slx4 act redundantly with it. Pardo and Aguilera (2012) reported a decreased frequency of half crossovers to circularize Chr III in the absence of Mus81, Yen1, and Slx1, suggesting these nucleases function redundantly to process branched DNA structures.

Recent studies show BIR occurs by a migrating D-loop mechanism resulting in conservative inheritance of the nascent strands (Donnianni and Symington 2013; Saini ; Wilson ). Structure-selective nucleases are not predicted to function in BIR by this mechanism, but could be required if progression of the D-loop stalls and requires restart by cleavage and reinvasion. Using the chromosomal assay, the frequency of BIR was unaffected in any of the nuclease single mutants, but we did find a significant decrease in the Δ Δ double mutant and a further decrease in the Δ Δ Δ triple mutant, suggesting most of the BIR events detected in this system require cleavage of a branched DNA intermediate. The requirement for Rad1–Rad10 could be to remove the heterology presented by the HO cs sequence that would prevent pairing between the 3′ end and donor template or to cleave at the heterology boundary if the D-loop intermediate branch migrated to the end of the homology shared by the sequences (Figure 7) (Fishman-Lobell and Haber 1992). Mus81 is in the same family of nucleases as Rad1 and there is some evidence for overlapping functions to process recombination intermediates between limited homology substrates (Schwartz and Heyer 2011; Mazón ). Yen1 could potentially cut the front end of the D-loop or the crossed strands resulting from strand invasion. The products of these cleavage events would necessitate a second cycle of BIR to form viable products.

Figure 7

Model for how structure-selective nucleases promote BIR. After strand invasion, the heterology represented by the HO cs needs to be removed to allow extension of the 3′ end by DNA synthesis. Cleavage of the trailing end of the D-loop intermediate coupled with cleavage of the 5′ flap represented at the leading end of the D-loop would collapse the recombination intermediate, and the truncated chromosomes would need to engage in a second round of BIR to form viable products.

In summary, our data support the hypothesis that Mph1 is the main helicase responsible for template switching during BIR. Unexpectedly, we observed a high frequency of LOH for Chr XV markers located 150 kb downstream of the site of strand invasion during BIR. We suggest these events are due to cleavage of the strand invasion intermediate, transferring the DSB from the linear CFV to the chromosome to initiate secondary BIR events, further increasing the genomic instability associated with this mutagenic DNA repair mechanism.