|Home | About | Journals | Submit | Contact Us | Français|
To distinguish among possible mechanisms of repair of a double-strand break (DSB) by gene conversion in budding yeast, Saccharomyces cerevisiae, we employed isotope density transfer to analyze budding yeast mating type (MAT) gene switching in G2/M-arrested cells. Both of the newly synthesized DNA strands created during gene conversion are found at the repaired locus, leaving the donor unchanged. These results support suggestions that mitotic DSBs are primarily repaired by a synthesis-dependent strand-annealing mechanism. We also show that the proportion of crossing-over associated with DSB-induced ectopic recombination is not affected by the presence of nonhomologous sequences at one or both ends of the DSB or the presence of additional sequences that must be copied from the donor.
Mitotic double-strand breaks (DSBs) are efficiently repaired by homologous recombination. Repair can occur by several mechanisms, including gene conversion, break-induced replication, or single-strand annealing (reviewed in references 18 and 26). In wild-type cells, where the DSB is flanked by sequences that are homologous to intact sequences located on a sister chromatid, a homologous chromosome, or an ectopic donor, repair occurs mostly by gene conversion. Several models have been advanced to explain how gene conversion occurs (Fig. (Fig.1).1). In the seminal DSB repair model of Szostak at al. (35) (Fig. (Fig.1A),1A), the ends of a DSB are resected by a 5′-to-3′ exonuclease, producing 3′-ended single-stranded DNA (ssDNA). Recombination proteins from the Rad52 epistasis group assist the ssDNA ends to search for an intact homologous sequence where the ssDNA end can invade and prime new DNA synthesis (reviewed in reference 18). The invasion of both ends produces a symmetric structure linked together by two Holliday junctions (HJs), whose cleavage by a Holliday junction resolvase can yield gene conversions either with an associated crossing-over or without crossover. Hereafter we will refer to this model as the double Holliday junction, or dHJ, model (Fig. (Fig.1A).1A). Support for this mechanism has come predominantly from studies of meiotic cells, where key molecular steps of this process have been identified by the physical monitoring of DNA undergoing recombination (12, 32).
In the past few years it has become evident that there are additional mechanisms of gene conversion besides that described by Szostak et al. (35). First, the dHJ model postulates the formation of an equal number of crossover and noncrossover outcomes, but very low proportions of mitotic recombination events show crossing-over in budding yeast, Saccharomyces cerevisiae, including mating type (MAT) switching (<1%) (17), interchromosomal recombination (generally about 5%) (6, 14, 20, 24, 36), and Rad51-dependent intraplasmid recombination (8%) (13). A similar constraint appears in interhomolog and intersister mitotic recombination in mammalian cells (<3%) (34). Second, heteroduplex-containing DNA involved in mitotic recombination between slightly divergent sequences is found only at the recipient, but not also at the donor locus as would be predicted by the dHJ model (22, 29, 37). Finally, genetic and physical analyses of both mitotic and meiotic recombination intermediates suggest that crossover and noncrossover outcomes may result from different recombination intermediates (2, 14).
Gene conversions where there is little or no accompanying crossing-over have been explained by several versions of synthesis-dependent strand-annealing (SDSA) mechanisms (5, 7, 24, 26, 30, 34) (Fig. (Fig.1B).1B). Here, resection of DSB ends occurs as in the dHJ mechanism, but after strand invasion and the initiation of new DNA synthesis, the newly synthesized DNA is displaced from the template. Alternatively, intermediates similar to those of the dHJ model are transiently formed but are unwound by helicases/topoisomerases (10, 22, 24) (Fig. (Fig.1C).1C). Support for this idea comes from the demonstration in budding yeast that the Sgs1 helicase and its associated Top3 topoisomerase suppress crossing-over (14) and that human homologs of Sgs1-Top3 (Blm/Topo3α) resolve a covalently closed synthetic dHJ into noncrossovers (38). Two basic features of SDSA are that recombination is generally not associated with crossing-over and that the donor locus remains unchanged while the recipient locus receives all newly synthesized DNA. In yeast and in Drosophila melanogaster, evidence supporting SDSA has come from so-called tripartite recombination experiments where sequences homologous to each double-strand break end are located on different chromosomes (27, 33). The completion of these events requires the invasion and unwinding of newly synthesized DNA from both template sequences and the annealing of these newly copied strands. In the dHJ model, where there is no strand displacement, repair would be impossible if the two ends invaded homologous template sequences on different molecules. Further support for the SDSA model comes from gene conversion experiments in which the donor sequence contains a series of repeated sequences; changes in the number of repeats are seen frequently in the recipient locus and rarely in the donor (1, 25, 27).
Looking for a definitive way to determine the predominant mechanism of DSB repair in mitotic cells, we examined the fate of newly synthesized DNA during gene conversion using the isotope density transfer method initially used by Meselson and Stahl (23) to demonstrate semiconservative DNA replication. We examined gene conversion in Saccharomyces cerevisiae during MAT gene switching, a well-characterized intrachromosomal mitotic gene conversion event (reviewed in reference 8). MAT switching is an intrachromosomal ectopic gene conversion event induced by the site-specific HO endonuclease, in which normally about 700 bp of either Ya or Yα sequences at MAT are replaced by a similar-size copy of the opposite mating type copied from either the HMLα or HMRa donor loci (Fig. (Fig.2).2). Previous studies have shown that the genetic requirements for MAT switching are similar to those for other ectopic and allelic recombination events (26); however, MAT switching differs from some DSB gene conversion events in that one DSB end has the Y region that is not homologous to the donor and must eventually be removed to complete gene conversion using the more-distant, homologous X region. Recombination is initiated by strand invasion of the 3′-ended ssDNA in the Z region, where the donor and the recipient are perfectly matched. The 3′ end of the invading strand primes new DNA synthesis that copies the donor Y region and continues into the X region. The joining of the X region at MAT to the newly copied DNA occurs only about 30 min after strand invasion (37). Using isotope density methods, we show here that both of the newly synthesized strands created during gene conversion are inherited at the repaired recipient molecule, further supporting SDSA models of recombination.
Strain tGI268 was derived from strain RM14-3a provided by M. K. Raghuraman (University of Washington) and has the genotype MATa hmlΔ::NAT HMRα::λ bar1::ADE3 cdc7-1 his6 leu2 trp1 ura3 carrying the centromeric YCp50 plasmid with GAL1::HO marked by LEU2 (pJH727). Plasmid pGI6 contains a modified HMRα locus marked by URA3 (plasmid pXW172 ) and by the insertion of a 2,918-bp phage λ fragment at a BamHI site introduced in Yα 96 bp 5′ from the HO cut site (Fig. (Fig.2A).2A). A linear HindIII fragment was used to replace HMRa. Cells were grown in 13C- and 15N-containing medium (heavy medium; 0.1% potassium acetate and 0.01% ammonium sulfate) for over seven generations, and then growth was arrested with nocodazole (15 μg/ml) in the G2/M phase of the cell cycle, so that there was no ongoing DNA replication. Cells whose growth was nocodazole arrested were washed and resuspended in light medium (containing 12C and 14N) plus nocodazole. Thirty minutes after the isotope shift, 2% galactose was added to induce the DSB at MATa. To repress HO expression, 2% glucose was added after 1 h, and the cells were harvested 6 h after HO induction. Purified DNA was digested with BamHI and StyI restriction enzymes and fractionated by CsCl equilibrium density sedimentation as described previously (21). The refractive index of each sample was measured, and then DNA was separated by agarose gel electrophoresis. Hybridization with a phage λ-specific probe showed sedimentation of both the donor and recipient DNA fragments (Fig. (Fig.33).
Three ectopic recombination substrates were created by a modification of our previously described assay (14) in which an HO-induced break at MAT on chromosome III is repaired by recombination with MAT sequences inserted at the arg5,6 locus on chromosome V. All three ade3::GAL::HO strains contain MATa and have HML and HMR deleted. The chromosome V donor contains either MATa-inc, differing from MATa by a single base pair that prevents HO cleavage (pGI354); MATα-inc (pGI439); or 117 bp of the BglII-HincII region covering the HO cleavage site in MATa into which a 708-bp bacterial hisG segment was introduced (pGI438).
MAT gene switching in budding yeast can be induced synchronously in a population of cells by induction of a galactose-regulated HO endonuclease gene. For strain tGI268, when an HO-induced DSB is created in MATa, the Ya sequence is replaced by the 3.6-kb Yα-λ sequence copied from HMRα-λ. Thus, this gene conversion event requires the synthesis of >3.6 kb of new DNA. The insertion of these additional sequences was needed in order to generate appropriately sized restriction fragments whose density would be well resolved by equilibrium density gradient centrifugation. The insertion of phage λ DNA did not affect the silencing of the HMR locus, as HO endonuclease did not cut its cleavage site, which is occluded by positioned nucleosomes in this locus (28; data not shown). To assess the nature of repair DNA synthesis, it was necessary to carry out the isotope incorporation under conditions in which normal DNA replication was not occurring. We studied HO-induced recombination in G2-arrested cells, after replication was complete, rather than in G1, prior to DNA replication, because Cdk1 kinase, which is required for completing HO-induced recombination, is inactive in G1 (Fig. (Fig.2B)2B) (15). Cells were grown in medium containing heavy isotopes of carbon and nitrogen (13C and 15N); after growth was arrested in G2/M with nocodazole, the cells were shifted to medium containing normal, light isotopes. Thirty minutes after the shift in the medium, the HO endonuclease gene was induced by adding galactose to a final concentration of 2%, to induce a DSB at MATa. Most of the cells (>90%) remained blocked in G2 during the 6-h duration of the experiment; therefore, incorporation of new nucleotides synthesized from light isotopes was limited to DNA synthesis during the gene conversion event. Southern blot analysis showed that the replacement of Ya by Yα-λ DNA sequences at MAT was completed within 6 h (Fig. (Fig.2B).2B). Repair efficiency is only about 50% of that observed for HO-induced cycling cells.
By using CsCl equilibrium density centrifugation, we determined whether newly incorporated light-isotope-labeled (LL) DNA was incorporated at MAT and/or at HMRα::λ. The results of Southern blot analyses of fractions taken from the CsCl gradient are shown in Fig. Fig.3.3. The position of LL DNA was determined from MATα::λ cells grown in normal 12C 14N medium, and the position of semiconservatively replicated DNA from the same cells but grown first in heavy (13C 15N) medium and then allowed to go through one round of DNA synthesis in light (12C 14N) medium (HL DNA) was determined. It appears that all the newly synthesized DNA is located in the recipient, because the MATα::λ cell DNA sediments at nearly the same position as LL DNA isolated from cells grown in normal medium (Fig. (Fig.3A).3A). The small shift of the newly created MATα-λ fragment toward the HL peak probably results from the fact that the 3.7-kb BamHI-StyI fragment that we tested is 0.7 kb longer than the newly synthesized 3.0-kb Yα-λ sequence (Fig. (Fig.2).2). The MAT Z region and more-distal sequences are partially resected and later will be filled in by light (12C 14N) nucleotides, but the strand ending 3′ at the HO cut site is not replaced (37); thus, the BamHI-StyI fragment would be expected to be about 17% heavier than a fully LL segment.
In contrast, after completion of MAT switching, the HMRα-λ fragment remained fully heavy isotope labeled (HH) and sedimented in the same fractions as it did before repair, implying that no newly synthesized DNA was inherited by the donor locus (Fig. (Fig.3B).3B). The donor fragment analyzed is 7.6 kb long, and only about 3.3 kb participates in recombination (Yα-λ plus Z), so that only about half of the region was likely to have been replicated during repair (Fig. (Fig.2).2). Therefore, about 50% of one strand of the donor molecule would be light if MAT switching occurred by the dHJ mechanism. Consequently, if the dHJ mechanism were used, about 25% of the total DNA would be LL, and we should observe a peak halfway between HL and HH for the donor. Because we do not observe this shift, we conclude that the molecular weight of the donor did not change during recombination. We note, however, that nocodazole arrest of growth is not perfect, and a certain percentage of cells escape the arrest; consequently 5% of HMRa-λ fragments sedimented as HL (lighter than we would expect from the dHJ model of recombination). To make sure that this was due to the escape from nocodazole arrest and is not relevant to the repair, we tested the sedimentation of DNA fragments with two other loci on the same chromosome (THR4 and HIS4) by hybridizing the same membrane (sampled 6 h after DSB induction) with THR4- and HIS4-specific probes. In both cases, we observed a very small fraction of HL DNA due to escape from nocodazole arrest, similar to the results with the HMRα-λ fragment (Fig. (Fig.3B3B).
Our results show clearly that both DNA strands that are newly synthesized during DSB repair are recovered in the recipient locus. This finding is not compatible with a dHJ mechanism involving an HJ resolvase in which almost all of the resolved recombinants are recovered as noncrossovers. Such a mechanism would produce HL restriction fragments for both the donor and the recipient, contrary to what was observed. These data provide strong evidence that either SDSA or a dHJ unwinding model is the predominant repair process in mitotic cells. Based on results from previous studies, we suggest that both pathways are used but that the predominant one is SDSA. In particular, SDSA best explains both the frequent expansion/contraction of repeated sequences copied into the recipient during gene conversion (27, 31) and the tripartite recombination events, where sequences homologous to each double-strand-break end are located on different chromosomes.
We note that, in order to use density transfer methods, we were obliged to study DSB repair that includes the copying of a larger nonhomologous donor sequence than the usual 700-bp segment, so that the new DNA was incorporated into a 3-kb fragment. Moreover, MAT switching is inherently asymmetric, in that one side of the DSB has nonhomologous sequences that prevent the immediate engagement of an end that can prime new DNA synthesis. It is possible that when both ends are equally capable of strand invasion and initiation of new DNA synthesis and when there is no large heterologous region to be copied, a higher proportion of cells will use the dHJ mechanism. To address whether the exceptional features of MAT switching create a special case in which SDSA predominates, we tested the ratios of crossovers and noncrossovers in three different ectopic recombination assays between chromosomes III and V, in which we altered the donor and recipient (Fig. (Fig.4).4). The first assay showed that both DSB ends are almost perfectly homologous (there is a single-base-pair difference that prevents the MATα-inc donor from being cleaved by HO). The equivalence of the two ends might promote dHJ formation and an increased level of crossovers (Fig. (Fig.4A).4A). The second assay showed that the recipient has one nonhomologous end (the 0.65-kb Ya sequence) and the donor has a heterologous segment (the 0.7-kb Yα sequence) that needs to be copied to the recipient; this scenario perfectly imitates MAT switching, except that the donor is on a different chromosome (Fig. (Fig.4B).4B). The third assay showed that the recipient has two nonhomologous tails, 75 and 46 bp long, and a 708-bp hisG insertion that needs to be copied during repair (Fig. (Fig.4C).4C). If nonhomologous segments at the ends or a large heterologous segment at the donor locus would have any impact on the engagement of the two DSB ends and the formation of a dHJ crossover intermediate, then we would expect to see significant differences in crossover frequency in these three cases. The results of Southern blot analyses following DSB repair after HO induction are shown in Fig. Fig.4D.4D. There is a delay in recombination when the two ends are nonhomologous, as we have seen before (4). However, in all cases the exchange frequencies were the same, indicating that the selection of a noncrossover SDSA pathway or a crossover-generating dHJ pathway is not affected by nonhomologous tails or heterology in the donor. These new results are in agreement with our previous studies of HO-induced recombination when both ends are homologous to those of the donor (14, 19). Thus, it seems that MAT switching is a representative function for studying DNA synthesis during DSB repair. Nevertheless, we acknowledge that it is still possible that recombination between fully homologous sequences, such as between two sister chromatids, can occur more often by the dHJ mechanism than by SDSA.
The low level of crossing-over in mitotic cells suggests that dHJs are formed only infrequently, and probably about half of them are removed by Sgs1-Top3 activity (14). In interchromosomal ectopic HO-induced recombination, only about 4% of cells repair the DSB with an accompanying crossover (14). In MAT switching, the proportion of crossovers that generate deletions between MAT and HMR is also small (9, 17). Such a small proportion would not have been detected in these experiments. Crossing-over might reflect the operation of the dHJ mechanism, or it might indicate that sometimes SDSA mechanisms can “trap” HJs, as has been suggested for some variations of the SDSA mechanism (2, 7, 26). Two enzymes that promote the noncrossover SDSA pathway are the Srs2 and Mph1 DNA helicases (14; our unpublished data). We are currently investigating the roles of these two helicases in promoting noncrossover recombination pathways.
Previously, Arcangioli (3) used density transfer to study mat gene switching in Schizosaccharomyces pombe. Approximately 25% of fission yeast cells switch mating types in any generation, and correlating with this proportion, about 25% of the newly synthesized mat DNA was LL, suggesting that an SDSA mechanism was involved. This gene conversion event is fundamentally different from that examined here, in that there is no HO-like nuclease to make the DSB; rather, a DSB is generated during the S phase itself from a preexisting nick, and the repair process is dependent on cells being in S phase (11, 16). In the case we studied here, repair is initiated outside S phase and both strands of the Y region of MAT must be copied de novo.
We are grateful to M. K. Raghumaran for his suggestions for carrying out density transfer experiments. We thank Phil Hastings and Neal Sugawara for critical comments.
This project was supported by NIH grant GM20056 to J.E.H. and by a startup grant from Baylor College of Medicine to G.I. G.I. was supported by a Charles King Trust fellowship.
Published ahead of print on 9 October 2006.