One of the least understood aspects of homologous recombination is the process by which the ends of a double-strand break (DSB) search the entire genome for homologous templates that can be used to repair the break. We took advantage of the natural competition between the alternative donors HML and HMR employed during HO endonuclease-induced switching of the budding yeast MAT locus. The strong mating-type-dependent bias in the choice of the donors is enforced by the recombination enhancer (RE), which lies 17 kb proximal to HML. We investigated factors that improve the use of the disfavored donor. We show that the normal heterochromatic state of the donors does not impair donor usage, as donor choice is not affected by removing this epigenetic silencing. In contrast, increasing the length of homology shared by the disfavored donor increases its use. This result shows that donor choice is not irrevocable and implies that there are several encounters between the DSB ends and even the favored donor before recombination is accomplished. The increase by adding more homology is not linear; these results can be explained by a thermodynamic model that determines the energy cost of using one donor over the other. An important inference from this analysis is that when HML is favored as the donor, RE causes a reduction in its effective genomic distance from MAT from 200 kb to ∼20 kb, which we hypothesize occurs after the DSB is created, by epigenetic chromatin modifications around MAT.

ONE of the least-understood aspects of homologous recombination is how a Rad51-coated nucleoprotein filament of single-stranded DNA can search the entire genome for homologous sequences with which to effect repair of a double-strand break (DSB). A well-studied model system of DSB repair is mating-type gene (MAT) switching in budding yeast (Figure 1A) (reviewed by Haber 1998; Haber 2007). This homothallic process allows haploid cells to grow into isogenic diploid colonies. Switching is initiated by the site-specific HO endonuclease, which creates a single DSB within the MAT locus adjacent to Ya or Yα sequences that encode cell-type-specific regulators. The HO-cut MAT sequences engage in homologous recombination (a gene conversion event without crossing over) in which either the HML or HMR sequences, located 200 and 100 kb from MAT, respectively, are used as templates to repair the DSB (Figure 1A). Although HML and HMR harbor equivalent HO cleavage sites to that at MAT, they are not cleaved, because of the heterochromatin structure of these sites. In most strains HML carries Yα and HMR carries Ya. Following HO cleavage, the ends of the DSB are resected by 5′–3′ exonucleases to produce long, 3′-end single-stranded DNA (ssDNA), which is coated first by the ssDNA-binding protein complex RPA and then by a filament of the Rad51 recombinase (Figure 1B). Collisions between the nucleoprotein filament and other DNA sequences results eventually in the recognition of homologous sequences, which can be opened up within the filament to allow the ssDNA to base pair with its complementary strand, creating a D-loop strand-invasion intermediate. These steps have been well studied in vitro (Sung et al. 2003; Krogh and Symington 2004). In vivo, chromatin immunoprecipitation techniques can be used to demonstrate the association of Rad51 first with the ssDNA tail at MAT and later with donor sequences (Sugawara et al. 2003; Wolner et al. 2003; Wang and Haber 2004; Hicks et al. 2011).

Figure 1 

MAT switching in Saccharomyces cerevisiae. (A) Chromosomal arrangement of MAT and its donors on chromosome III. An HO endonculease cleavage at the boundary between Ya and Z1 sequences at MAT initiates replacement of the ∼650-bp Ya sequences by ∼700 bp Yα sequences copied from the heterochromatic HMLα locus. HO cleavage sites in HML or HMR are not cleaved because these regions are untranscribed and heterochromatic (indicated by hatching); silencing depends on cis-acting E and I sites adjacent to these loci. Regions of homology W, X, Z1, and Z2 shared by HML and MAT are indicated, along with the X and Z1 regions shared by MAT and HMR. In MATa cells, the recombination enhancer (RE) (shown as a green square) increases to 90% the choice of HML; when RE is deleted or in MATα cells, where RE is inactivated, HML usage drops to 10%. (B) Molecular events during MAT switching. The ends of a DSB are resected, allowing Rad51 filament formation. Homology searching by this filament engages the donor (here, HMLα). On the right side (region Z) of the DSB, strand invasion can form a fully interwound plectonemic structure, which can recruit PCNA and other DNA replication proteins. However, on the left side, the paired W and X regions may form only a weak, side-by-side paranemic joint, because the first 650 nt of the 3′ end of the left end of the DSB is not homologous to the donor. Conversion of this joint to an interwound plectonemic joint requires either the action of a topoisomerase or the removal of the nonhomologous tail by the Rad1–Rad10 flap endonuclease. Initiation of new DNA synthesis from the 3′ Z end produces a single strand of newly copied DNA that is displaced and can anneal with the second end, allowing that end of copy the new MAT sequences.

Strand invasion is followed by the initiation of new DNA synthesis using the 3′ end of the invading strand as a primer (White and Haber 1990; Hicks et al. 2011). Because switching to the opposite mating type replaces Y sequences to the left of the DSB, strand invasion and primer extension initially occur in the Z region, where the end of the DSB is perfectly matched to the donor. In contrast, primer extension from the WX region can occur only after a longer resection has taken place and the nonhomologous 700-bp Y region is clipped off to expose a functional primer end (White and Haber 1990). Eventually the first strand copied is displaced from the template and is itself used as the template for the rightward copying of the second strand, leaving all the newly synthesized DNA at the MAT locus and the donor unchanged (Ira et al. 2006).

It is important to note that the amount of homology shared by HML or HMR with MAT is sufficient to ensure highly efficient recombination. When both donors contain the same Y region, more than 90% of cells complete MAT switching, and this number rises to nearly 100% when the competing nonhomologous end ligation of the HO-cut ends is eliminated (Valencia et al. 2001). As noted above, recombination is initiated with the MAT-Z end, which shares only 230 bp homology with the distal side of HMR or 320 bp at HML. Replacement of HML with HMR does not alter the left donor usage in MATa cells (Wu and Haber 1995), so that 230 bp to the right of the DSB is sufficient at either donor to effect highly efficient recombination.

A remarkable aspect of MAT switching is its donor preference, such that MATa usually recombines with HMLα (i.e., MATa is replaced by MATα) and, similarly, MATα is usually replaced by MATa, by preferentially copying HMRa. Preference is not dictated by the Y sequences at the donors nor differences in the sequences in and around the donors, because the choice remains the same if the Y regions are replaced or if HMLα is replaced by HMRα; choice is dictated by chromosome location. Donor preference is enforced by a 275-bp cis-acting recombination enhancer (RE) sequence, located 17 kb from HMLα (Wu and Haber 1996). In MATa cells, RE binds several proteins including Mcm1, Fkh1, and Swi4/Swi6, whereas in MATα cells the Matα2Mcm1 repressor binds to RE and establishes a region of highly positioned nucleosomes that prevent binding of other proteins (Wu et al. 1998; Coïc et al. 2006a,b). How RE works has not yet been precisely elucidated, but it clearly influences how HML is used as a donor; when RE is deleted in MATa cells, they behave like MATα cells, using HML less than 10% of the time. We have proposed that the chromosome III architecture is different in MATa and MATα cells in a way that favors HML or HMR accessibility (Bressan et al. 2004; Coïc et al. 2006a).

The presence of two donors allows us to assay competition between these donors, to examine key steps leading up to recombination. We have previously used competition assays to demonstrate that an intrachromosomal donor outcompetes an interchromosomal donor, a result that is supported by observing the kinetics of DSB repair (Wu et al. 1997; Ira et al. 2003; Keogh et al. 2005). These observations and others (Agmon et al. 2009) suggest that the search for homologous sequences by the Rad51 filament is more efficient intrachromosomally. Here we wished to determine if the use of HML and HMR is affected by changing the length of homology shared between one donor and MAT. We note again that, in this situation, the natural length of homology at each donor is sufficient to accomplish nearly 100% gene conversion, so that we are not looking at a net increase in recombination efficiency as one might do if each donor were inherently inefficient. Such increases in recombination efficiency (in the absence of competition) can be seen if the initial lengths of homology are close to the minimum efficient processing segment (MEPS) and allow only inefficient repair of a DSB (Shen and Huang 1989). In the case of budding yeast, MEPS appears to be about 70 bp (Ira et al. 2003).

In the case we study here, the bias in usage of one donor is enforced by RE and we are interested in whether donor preference is so strongly dictated that the ratio of the preferred to the excluded donor cannot be changed. The strong exclusion of the “wrong” donor can be partially relieved by increasing the homology shared with MAT on the distal side of the DSB from ∼150 bp to ≥ 2 kb. A series of donors with increasing amounts of homology distal to the DSB reveals that use of the unfavored donor increases, but not linearly, to about one-third of all switches. These data argue that the choice of the preferred donor is not irrevocable. We conclude that successful recombination usually results only after multiple encounters between the donor and the Rad51 filament covering DSB ends at MAT, so that the rare encounters of the “wrong” donor with the DSB ends become more likely to be successful when the size of homology increases. Greater homology between the disfavored donor and MAT leads to a higher chance that each encounter will progress from an initial, transient strand association to a successful recombination event.

These data, combined with data from the kinetic analysis of the early steps in homologous recombination, allow us to construct a thermodynamic model for donor preference that accounts quantitatively for the effect of changing the homology length. The free energy cost for pairing each of the two donors with the Rad51-coated MAT sequences can be calculated and used to infer how RE acts to facilitate the use of the more distant HML donor. We conclude that RE markedly lowers the free energy of association with MAT equivalent to moving HML from its normal genomic distance of 200 kb to a proximity that is equivalent to having a donor only 23 kb away. This prediction of the thermodynamic model suggests that there is a dramatic change in the architecture of chromosome III after induction of the DSB, which could be explained by RE associating with the DSB region.

Materials and Methods


Donor preference was measured in derivatives of DBY745 (ho MATa ade1-100 ura3-52 leu2-3,112). The replacement of HML and its associated silencers sequence E and I have already been described (Wu and Haber 1995). A 3.4-kb region including HMR and its adjacent silencers was replaced with a 4.8-kb EcoRI–BamHI MATa-inc-URA3 (uncuttable by the HO endonuclease) fragment that lacks the silencers E and I. To avoid the expression of mating-type genes from the additional MAT copies introduced, the Y sequences were deleted at these loci. Yα was replaced by a NAT cassette obtained from pAG25 (Goldstein and McCusker 1999) by transformation with a PCR fragment containing the NAT cassette flanked by the 50 first and last sequence of Yα with the primers: YαNATU (5′-tat gtc tag tat gct gga ttt aaa ctc atc tgt gat ttg tgg att taa aag cat agg cca cta gtg gat ctg-3′) and YαNATL (5′-cga agt agt ccc ata ttc cgt gct gca ttt tgt ccg cgt gcc att ctt cac agc tga agc ttc gta cgc-3′). Equally, Ya was replace by a KanMX4 cassette by transformation with a PCR fragment obtained from pFA6-KanMX4 and primers MATKANp1 (5′-tag gta aat tac agc aaa tag aaa aga gct ttt tat tta tgt cta gta cag ctg aag ctt cgt acg c-3′) and MATKANp2 (5′-caa cca ctc tac aaa acc aaa acc agg gtt tat aaa att ata ctg ttg cgc gaa gta gtc cca gca tag gcc act agt gga tct g-3′). These alleles are referred to as matYΔn or matYΔk, respectively.

At the HML loci, the extra homology centromere-distal to the matYΔn extra copy was truncated in a series of strains, in each case by replacement with a KAN cassette. In strain ECY446, replacement was accomplished by transformation with a PCR fragment containing the KAN cassette flanking by sequences surrounding the distal sequences, leaving the same 325 bp of homology that exists between MAT and HML. The PCR fragment was obtained with primers MATdistKANU (5′-tat gta ttt gta taa aat atg ata tta ctc aga ctc aag caa aca atc aac agc tga agc ttc gta cgc-3′) and HMLMATdistKANL (5′-att tgc tca aaa tta tcc atg aaa gaa gta cca atg aag cac tag cct gtg cat agg cca cta gtg gat ctg-3′). A similar strategy was then used to truncate the homology between the matYΔn sequences at hmlΔ and those at MAT to lengths of 148 (ECY487), 653 (ECY474), 754 (ECY488), 994 (ECY489), 1320 (ECY475), 2216 (ECY490), 2717 (ECY491). The specific primers used are available upon request.

The same approach was employed to delete the distal homology of matYΔk at the HMR locus by replacement with the NAT cassette. The PCR fragment suitable for transformation was obtained with primers MATdistKANU (see above) and HMRMATdistNATL (5′-tgt ggg gtt gca ttg tag ata aaa gta ata ata tta ggt ata tag aat atg cat agg cca cta gtg gat ctg-3′).

RE was deleted by transformation of a DNA fragment containing an URA3 cassette surrounded by RE flanking sequences isolated from the plasmid pJH1561 (Wu and Haber 1995).

The MATa reΔ strains carrying HMRα-B and hmlΔ::matYΔn, with various NAT-marked truncations of the homology (described above) were then transformed with a hmrΔ::LEU2 fragment from plasmid JKM106 to delete HMRα-B, yielding a set of strains in which a DSB at MAT can be repaired only by recombination with the modified HML sequence. Because the HML construct lacks an HO cleavage site, cells that switch will not suffer further HO cleavage; hence the strains can be directly plated on rich media containing galactose (YEP-Gal plates). Standard yeast genetic and molecular techniques were used (Guthrie and Fink 1991).

Analysis of donor preference

Quantification of donor preference on Southern blot was performed as previously described (Wu et al. 1998). Signals were quantified using ImageQuant V1.2 (Molecular Dynamics).


Cells were grown in YEP-lactate overnight to mid-log phase and then appropriate dilutions were plated on rich media containing dextrose (YEPD) and rich media containing galactose (YEP-Gal plates, where HO is induced).


Extra homology adjacent to the Z region, but not its heterochromatic state, increases competitiveness of the less frequently used donor

In a MATa cell in which the two donors carry differently marked Yα sequences, HMR is used only ∼10% of the time to repair the DSB (Figure 2A; Wu and Haber 1995). We asked if extra homology introduced at the HMR locus would increase its use. In these experiments we replaced HMR by MATa-inc sequences, eliminating the cis-acting silencers adjacent to HMR, but preventing HO from cleaving the unsilenced locus, because the HO cut site is mutated by a single base pair substitution in this construct (Weiffenbach et al. 1983) (Figure 2B). In addition, Ya was replaced by the G418-resistance KAN gene, so that the unsilenced donor would not express mating-type information that would alter the cell’s mating type. The replacement introduced an additional 1155 bp of homology shared with MAT to the left of the 703-bp X region (the amount of homology shared between MAT and the HMR locus is now WX = 1858); 2159 bp of homology was also added to the right of the 238-bp Z1 region (Z = 2397). Because the modification that is most decisive involves the length of the Z sequences (see below), we designate this locus as hmrΔ::matYΔk-Z2397. Following expression of a galactose-inducible HO gene for 90 min, the DSB could be repaired by recombining either with the modified, unsilenced HMR locus or with the normal, heterochromatic HMLα locus. The ratio between these two recombination outcomes was determined by probing a southern blot with MAT-proximal sequences (Figure 2B).

Figure 2 

Alterations of the donor loci to examine competition between HML and HMR. (A) Quantification of donor preference in MATa strains bearing heterochromatic donors (indicated by hatching) HMLα and HMRα-B (from Wu and Haber 1995 and Wu et al. 1998). The proportion of the two expected products MATα and MATα-B, resulting from an HO-induced DSB at MAT (arrow), are determined by probing a southern blot of HindIII (H)–BamHI (Ba) enzyme-digested DNA with Yα sequences (horizontal bar). The size of the restriction fragment detected on the Southern blot is indicated. The results of the quantification is given in strains carrying or deleted for the RE. (B) The HMR locus was replaced by MATa-inc sequences, deleting the E and I silencers leading to hmr∆::matY∆k-Z2397. Ya was replaced by the G418-resistance KAN gene (matY∆k, green), so that the unsilenced donor would not express mating-type information (strain ECY246). The replacement introduced additional sequences homologous to the MAT locus: 1155 bp was added upstream of the 703 bp of X sequences, increasing the homology to 1858 bp (WX1858). On the right side, the homology increased from 238 bp (Z1) to 2397 bp (Z2397). The additional homology at Z was subsequently removed by inserting the NAT locus, yielding to hmr∆::matY∆k-Z325 (strain ECY447). Following expression of a galactose-inducible HO gene for 90 min, the DSB could be repaired by recombining either with the modified, unsilenced hmr∆::matY∆k locus or with the normal, heterochromatic HMLα locus. The ratio between these two recombination outcomes was determined by probing a Southern blot of HindIII (H)–BlpI (Bl) enzyme-digested DNA with MAT proximal sequences upstream of the region of extended homology (horizontal bar). In addition to the bands corresponding to the parental and products fragments, a signal corresponding to the proximal sequences of MAT is also revealed (MAT prox.) (C) An analogous replacement of HML and adjacent E and I sites was carried out, adding 433 bp homology to the left (WX1858) and 3504 bp homology to the right (Z3829) of MAT leading to hml∆::matY∆n-Z3829. Yα was replaced by NAT (matY∆n, yellow) in ECY375. In strain ECY446, the additional homology was removed by KAN insertion leading to Z325. RE was deleted in strains ECY377 and ECY446. Following induction of the HO endonuclease, the proportion of the alternative recombination outcomes was determined as in B. (D) Competition between hmr∆::matY∆k-Z2397 and hml∆::matY∆n-Z3829 (ECY271 RE+, ECY292 reΔ). Note that the size of Z at the HML locus is nearly twice the size as at the HMR locus. Deleting the additional homology at both loci (ECY459) restores the rule of donor preference when the HML and HMR donors are unmodified. Donor usage was determined as in B for ECY459; the genomic DNA of ECY271 and ECY292 was digested with HgaI and probed with the same MAT proximal sequences. The left donor usage is the mean of at least three independent determinations in all experiments.

With the additional homology, HMR usage in this MATa strain increased from 10% in the unaltered strain to 39% (compare Figure 2, A and B). This increase in donor usage could have come either from unsilencing the donor locus or from the creation of greater shared homology. To distinguish between these alternative explanations, we deleted the additional MAT-distal sequences at HMR with a NAT (nourseothricin resistance) gene (see Materials and Methods) leaving only 325 bp of homology to the right of the cleavage site, the same as that found at HML, creating hmrΔ::matYΔk-Z325. In this construct, where the KAN gene replaced Ya, HMR usage returned to its normal level, 12%. Therefore, the improved usage of unsilenced MAT replacing HMR is not explained by the lack of silencing, but by the increased length of homology shared between the donor and the recipient on the centromere-distal side of MAT.

It should be noted that the increased amount of homology on the left side of the donor (increased from 703 bp in the X region of HMR to 1858 bp) did not alter HMR usage when the right side was reduced to 325 bp. This result is consistent with the idea that the initial steps of recombination are dictated by the ssDNA of the Z region, which matches the donor exactly (except for a single-base-pair substitution), whereas the left side of the DSB has a large segment of heterologous DNA at its end (in this case the KAN sequences and part of Ya) that prevent stable synapsis and primer extension (White and Haber 1990).

By a similar approach we unsilenced the HML donor by deleting the adjacent E and I silencer sequences and inserting an equivalent segment from a MATα locus (Figure 2C). Nearly all of the Yα region and part of the HO cleavage site was then replaced by a NAT gene. The initial segment of MAT that we inserted extended the amount of homology from 1425 bp in the W and X regions to 1858 bp and from 325 bp in the Z1/Z2 regions to 3829 bp. The resulting construct was designated as hmlΔ::matYΔn-Z3829. Following induction of the HO endonuclease, the ratio between the alternative recombination outcomes could be determined by probing a Southern blot with MAT proximal sequences upstream of the region of extended homology. When RE is active, the use of the modified HML was significantly higher (nearly 100%, Figure 2C) compared to a normal HML (90%, Figure 2A). When RE was deleted, the use of HML also increased, from the normal level of about 10 to 34% when the Z sequences share 3829 bp with MAT. As with the alterations at HMR discussed above, the improved use of the “wrong” donor is attributable to the length of the shared homology and not to the unsilenced donor, because when the hmlΔ::matYΔn-Z3829 sequences were truncated back to 325 bp to the right of the cut site, using a KAN-marked modification, i.e., hmlΔ::matYΔn-Z325, the use of this HML locus was again ∼10% in a RE deleted strain.

Finally, we examined left donor usage in strains where both HML and HMR loci had been replaced by extended, unsilenced MAT sequences (Figure 2D). We observed a wild-type donor preference: 92% of the cells used the left donor when RE is active but only 30% when RE is deleted. This result again confirms than the E and I silencer sequences are not involved in donor preference. The greater use of hmlΔ::matYΔn-Z3829 in the MATa reΔ strain (30%) than that seen in normal strains (10%, Figure 2A) may reflect the fact that this donor shares more homology with the right side of the HO-cut MAT (3829 bp vs. 2397 bp). When the extra homologies introduced at both donor loci were deleted in a MATa reΔ strain, so that each donor had 325-bp homology in the Z region, the left donor was largely excluded. Left donor usage (12%) was not significantly different from strains carrying HMLα and HMRα–BamHI (10%, Figure 2A). These data also show that the NAT or the KAN cassettes used both to delete the extra homology and to replace the Y sequences at both donors do not interfere with their usage.

Length dependence of the increased use of HML

To better understand the relationship between the length of homology shared by HML and MAT in competition with a normal, silenced HMR in a RE deleted strain, we constructed a series of eight truncations of the right-hand homology, leaving as little as 148 bp and ranging up to 2717 bp, in each case with the shared homology bounded by the KAN cassette. These strains are designated by the total homology shared between MAT and the modified donor (hmlΔ::matYΔn-Z148, hmlΔ::matYΔn-Z326 to hmlΔ::matYΔn-Z2717). The results of this analysis are shown in Figure 3. When matYΔn shares only 148 bp with its donor, the efficiency of repair is considerably reduced compared to the normal 325 bp. Increasing homology leads to increased use of the HML donor, although the increase is not strictly proportional to distance. Thus, additional homology at the “wrong” donor can partially overcome the donor preference mechanism. A best-fit line through these data is based on a “toy” model based on a model in which (1) an encounter between either the preferred or disfavored donors of normal size will result only in the completion of recombination in ∼25% of attempts, and (2) the success in completing DSB repair increases as the size of the Z region increases (see Discussion).

Figure 3 

Homology length dependence on the use of HML in a MATa reΔ HMRα-B strain. (A) Rationale. The length of homology shared by hml∆::matY∆n-Z and MAT is modulated at the donor by the introduction of a KAN cassette containing variable amounts of Z sequences (from 148 to 2717 bp). (B) Donor usage in the resulting nine strains (see Materials and Methods) was measured on Southern blots as in Figure 2C. (C) The percentage use of the “wrong” HML locus, hml∆::matY∆n-Z, measured as the average of at least three independent quantifications, is plotted as a function of the length of homology shared with MAT. A ball-tossing model of donor preference that reflects the consequences of increasing the length of HML homology with MAT is shown as a red curve. This line is a best fit through the data, representing the case where the search is directed toward HMR nine times as often as HML, and where SR/SL max= 0.23, SL max = 1, Lopt = 1800 bp, and the minimum length of homology used by Rad51 is 100 nt. A nearly identical line is obtained if the minimum homology length is 70 pb. (D) Donors with limited homology on the Z side are efficient. MATa reΔ strains with varying extents of homology on the Z site of hml∆::matY∆n-Z were further modified by the LEU2-marked deletion of HMRα-B. Strains with Z148, Z325, and Z653 were grown in YEP-lactate and dilutions were plated on YEPD to determine cell number, and on YEP-Gal to measure viability after HO induction. The percentage viability plotted is the results of three independent experiments.

Efficiency of the “wrong” donor in the absence of competition

If there is no competition, will donors with limited homology on the Z side prove to be adequate donors? To address this question we deleted HMRα-B in the set of strains described above where RE is deleted. Because the modified HML donor in each of these strains lacks an HO cleavage site, the efficiency of switching could be assessed by plating the cells directly on a rich medium containing galactose (YEP-Gal) and comparing the viability to cells plated on rich medium containing dextrose (YEPD), Southern blot analysis being uninformative here. The results, shown in Figure 3D, demonstrate that in the absence of competition, even a donor with only 148 bp of homology is quite efficient (60% viability) compared to only ∼1% usage when it is in competition with the preferred HMRα-B donor (Figure 3, B and C). At 325 bp (the normal size of the Z1 and Z2 regions in HML) or 653 bp, the length of homology is sufficient to repair the DSB in nearly 90% of the cells. Therefore, the low usage of the modified HML donor can be unambiguously attributed to the competition with HMRα-B and not to a poor efficiency in recombination because of the smaller amount of homology with MAT.

A very important additional finding is that the primary factor regulating the use and efficiency of HML is the homology to the right of the HO cleavage site, where the DSB end is perfectly homologous to the donor, as we noted when extra homology was added to HMR (see previous section). In most of the constructs there is considerably more homology to the left of the DSB (i.e., 1858 bp including the W and X regions) than to the right. As noted above, the lack of influence of the donor sequences to the left of the DSB most likely reflects the fact that these homologous sequences are separated from the DSB end by ∼1 kb of nonhomologous sequences, so that they are unable to form a stable association with the donor (see Discussion).


Donor efficiency is unaffected by its heterochromatic state

An important finding in this study is that the normally heterochromatic HML and HMR loci do not become more efficient donors if they are unsilenced. A notable feature of the silent donors is that HO endonuclease cannot cleave the same sequence that is readily cut at MATa or MATα, apparently because of highly positioned nucleosomes that occlude the cleavage site (Weiss and Simpson 1998; Ravindra et al. 1999). Yet the cleaved MAT-Z end, as part of a Rad51 nucleoprotein filament, is capable of invading precisely the same sequences, promoting strand invasion and the initiation of new DNA sequences. Rad51 and its associated proteins must therefore be able to overcome any chromatin-based barrier to DSB repair. These conclusions appear to be different from those reached in the study of gene conversion promoting immunoglobulin diversity in vertebrate DT40 cells (Cummings et al. 2007).

Establishment of strand invasion must involve multiple attempts

The competition between HMR and a series of HML donors with different lengths of homology provides an opportunity to consider the dynamics of the search for a donor by the Rad51 filament associated with the single-stranded 3′ ends of the DSB at MAT. In a MATa reΔ strain, HMR is normally used 90% of the time. If this choice was dictated prior to switching and the choice was irrevocable (e.g., if the preferred donor were already held close to MAT), then the modifications we made at HML should have no effect on the preferred usage of HMR. But this is not the case. These results are consistent with the fact that a close association between MAT and its donor does not occur until after HO creates a DSB (Bressan et al. 2004; Houston and Broach 2006). It is also unlikely that the relatively small sequence that we introduced into the HML region changes the probability of the initial encounter between MAT and the donors, 100 or 200 kb distant; rather they are influenced by global constraints on chromosome movement or architecture. The fact that the excluded donor usage can be improved by increasing the size of homology with MAT therefore suggests that encounters between MAT and its favored donor are not always productive. Synapsis may result only in a paranemic joint that is not converted into an interwound, more stable plectonemic joint that can recruit PCNA and DNA replication factors to launch primer extension (Sugawara et al. 2003). Thus, even though the preferred donor will make contact with the Rad51 filament nine times as often as the wrong donor, successful recombination is not guaranteed at any given encounter.

We postulate that the Rad51::ssDNA filament is likely to encounter the preferred donor several times before creating a stable recombination intermediate that will then lead to a completed switching event. Thus, although RE will ensure that the DSB in a MATa strain will encounter HML nine times more often than it will collide with HMR, when both donors are of equivalent size, it is likely that many more than 10% of the cells will encounter HMR at least once before HML is used as the donor. Similarly, when RE is deleted, the Rad51 filament will more frequently encounter the now-disfavored HML donor than is evident in its 10% usage.

We suggest that when the homology length of the wrong donor is increased, the frequency of productive collisions between MAT and the normally disfavored donor increases. This will lead to an overall increase in stable strand invasion and the initiation of new DNA synthesis with this donor, thus compensating for its infrequent collisions with the Rad51::ssDNA filament. A toy model that illustrates the changes in donor preference imagines a pail of ping-pong balls that are thrown nine times as often at a box in the right corner of a room than at an equivalently sized box in the left corner. The ball is hard to throw accurately, such that one gets the ball in a box only a small fraction of the time. Here we use the example that a throw succeeds 25% of the time. Therefore, on average, 75% of the balls will be on the floor and there will be nine times more balls in the right box. If one increases the width of the disfavored box by a factor of 2, now the tosses toward this box will have a twofold higher probability of landing in the bucket, while the number of throws to the right still exceeds the number of throws toward the left box by a factor of 9.From this model the success of using the “wrong” box can be calculated in the following manner: FL and FR are the fraction of times the ball is tossed toward left and right boxes (FL + FR = 1), while SL and SR are the probabilities that a toss in the left or right direction is successful (SL = SR = 0.25 in the above example). The probability of a ball landing in the box on the left (pL) is given bypL=FLSL/[FLSL+FRSR]=1/[1+FR/FL(SR/SL)].(1)Thus when FR/FL = 9 and SR = SL the use of the left donor will be pL =1/[1 + 9 × 1] = 0.1, as is observed in the unperturbed case. If, on the other hand, SL is increased to 1.0, then the left box usage rises to pL =1/[1 + 9 × 0.25] = 0.31, which is approximately the maximum value seen in our data when Z at HML increases >2700 bp (Figure 3C).

The model can be modified to reflect some of the realities of homologous recombination in budding yeast. As homology increases, SL will increase linearly until a point at which additional homology will not be as efficiently used is reached. First, the probability of success cannot exceed 1. A decline in further efficiency with homology length may happen because the amount of Rad51 in the cell is limited to about 3500 molecules, so that only ∼5000 bp on either side of the DSB can be covered if the filament is continuous (Sugawara et al. 2003). Moreover, if strand pairing is initiated far from the 3′ end of the filament, the probability that the remaining filament will “zipper up” as a paranemic joint to engage the 3′ end in the D loop, and thus have a chance to become plectonemic, is likely to diminish with distance. We can describe this mathematically by stating that SL will vary with homology size (HL) according to the equationSL=SLmax(1exp((HLm)/Lopt)),(2)where SL max is the maximum value of SL, which is reached at the largest homology lengths, m is the minimum size of a Rad51 filament that can be used in repair (70–100 nt) (Ira and Haber 2002), and Lopt is the point at which the efficiency of the filament’s size becomes nonlinear (here we estimate that it is 1800 bp). This equation has the property that it increases linearly with HL/Lopt when HL is much smaller than Lopt, and goes smoothly to SL max when HL is much larger than Lopt.

The data in Figure 3C are well described by this model, as shown for the best fit, where SR/SL max = 0.23. From these values we also can estimate that HMR will, on average, encounter MAT four times before completing recombination, assuming SL max= 1. If, at large values of homology at HML, the probability of a MAT-HML collision leading to a successful recombination event is less than one (SL max < 1), then the corresponding value for productive collisions with HMR will be equally reduced (SR = SLmax × 0.23); i.e., the number of unsuccessful trials will go up. We estimate that in about 40% of cells, HML will have been encountered at least once.

Given all the previous experimental evidence, we believe it likely that HML accessibility in cells where RE is active is dependent on a spatial reorganization of chromosome III induced by the formation of the DSB at MAT. If we further assume that the chromatin strand can be treated as a simple Gaussian polymer chain, then the ratio of probabilities that one or the other donor will make contact with MAT is equal to the ratio of their genomic distances to MAT raised to the −3/2 power (Rippe 2001). Simply based on their distances from MAT with no other constraints, HMR usage should be favored over HML by a factor of 3: [G(RM)/G(LM)] ^ (−3/2) = [100 kb/200 kb] ^ (−1.5) = 3; however, in MATa cells, HML is used 9 out of 10 times. This preference could be achieved if HML were brought close to MAT so that its effective distance were only 23 kb. This then leads to a preference, based on polymer theory, of [100/23] ^ (−1.5) = 1/9. Although our data and that of others show that HML is not constitutively much closer to MATa than is HMR in the nucleus (i.e., in the absence of HO cleavage) (Dekker et al. 2002; Bressan et al. 2004; Miele et al. 2009), we suggest that such a reorganization will occur after DNA damage is created. If RE is itself tethered to the DSB, then HML will be about 17 kb away, consistent with the estimate of 23 kb. This proximity could be accomplished if RE promoted a chromosome remodeling in response to the DSB. For example, RE binds multiple copies of the Fkh1 protein that contains a FHA domain that can bind phosphopeptides (Sun et al. 1989; Coic et al. 2006b). If Fkh1 binds to the checkpoint kinase-modified chromatin around the DSB (Shroff et al. 2004; Kim et al. 2007), it would recruit RE—and also HML—very close to the DSB at MAT. This possibility is being investigated.

Nonhomologous sequences prevent adjacent homology from dictating donor usage

A very important additional finding is that the primary factor regulating the use of HML is the homology to the right of the HO cleavage site, where the DSB end is perfectly homologous to the donor. In hmlΔ::matYΔn-Z325 and hmrΔ::matYΔk-Z325, where Z is truncated to its normal size at HML (325 bp), the large amount of homology to the left of the DSB (i.e., 1858 bp, including the 1425-bp W and X regions) is unable to increase the use of these donors. We believe the lack of influence of the donor sequences to the left of the DSB reflects the fact that these homologous sequences are separated from the DSB end by approximately 1 kb of nonhomologous sequences, so that they are unable to form a stable association with the donor (Figure 1). We believe that this side of the DSB does make a less stable, paranemic joint with the donor, although it cannot form a fully interwound, plectonemic structure until the nonhomologous region is removed. These sequences must be clipped off by the Rad1–Rad10 nuclease (White and Haber 1990; Fishman-Lobell and Haber 1992; Colaiacovo et al. 1999). Our recent studies have shown that there is Rad51-mediated synapsis on this side of the DSB before the nonhomologous sequences are removed (Hicks et al. 2011). Moreover, MAT switching remains efficient even if the W/X region is truncated to the same 327 bp as found on the right side (Hicks et al. 2011). So, although sequences distal to the left end of the DSB do make contact with the donor, the probability that Rad51-mediated homology searching will lead to repair is dictated by homology to the right of the cleavage site. We propose that the size of homology to the right of the HO cut site was selected in S. cerevisiae consistent with the architecture of chromosome III, to assure a strong exclusion of the wrong donor and the easy repair of the break with the preferred one.


Michael Lichten and Paul Ginsparg have provided invaluable comments and assistance. E.C. was supported in part by grants from l’Association pour la Recherche sur le Cancer and the Philippe Foundation. Research was supported by National Institutes of Health grant GM20056 (J.H.) and National Science Foundation grants DMR-0706458 (J.K.) and Materials Research/Science and Engineering Center grant 0820492 at Brandeis University.


  • 1 Present address: Department of Biology, CB #3280, Coker Hall, University of North Carolina, Chapel Hill, NC, 27599-3280.

  • 2 Present address: Neopeutics Sdn. Bhd. Halaman Bukit Gambir, 11700 Gelugor, Penang, Malaysia.

  • Received July 14, 2011.
  • Accepted September 9, 2011.

Literature Cited

View Abstract