Eukaryotes possess a characteristic abundance of short, repetitive sequences scattered throughout their genomes 
. As a consequence, catastrophic levels of DNA damage, such as those created by acute exposure to IR, or DNA damaging chemicals have the potential to generate genome rearrangements by interactions between homologous repetitive sequences on the same or different chromosomes 
. For instance, exposing diploid budding yeast cells to levels of IR sufficient to cause hundreds of DSBs per genome, with several in or near the hundreds of short, delta repeats strewn throughout the genome results in the formation of an abundance of chromosomal translocations by HR between unlinked repeats 
. This vigorous biological response to acute levels of DNA damage suggests that non-conservative HR between repetitive genomic sequences can be a potent mechanism for genome rearrangement with important implications regarding the advent of cancer and eukaryotic genome evolution 
While HR can be an important mechanism of genomic change in response to DNA damage, it is also an important mechanism for damage tolerance. For instance, in budding yeast resistance to IR is primarily dependent on HR 
. However, this resistance is most likely obtained through conservative HR between homologous sequences on sister chromatids or allelic sequences on homologous chromosomes, since the majority of survivors of even the most acute exposure display normal karyotypes 
. Further, the genetic control of radiation survivorship resembles that of conservative HR events, such as gene conversion more than it does non-conservative HR, such as SSA. For example, the repair of a HO catalyzed DSB by gene conversion and IR resistance are both greatly dependent on Rad51, while DSB-stimulated deletion and translocation formation by SSA are not ()
. Thus, DSBs can be repaired by separate conservative and non-conservative mechanisms of HR, suggesting that the maintenance of genome stability following DNA damage may require that the cell promotes one while it inhibits the other.
The data presented in this paper suggest that the interaction between Rad51 and Rad52 that facilitates Rad51 nucleoprotein filament formation may be crucial to the maintenance of genome stability by facilitating conservative HR while opposing non-conservative HR. Strikingly, whether translocation formation by HR was initiated spontaneously, or by one, or two DSBs, either the rad51Δ allele that results in the total loss of Rad51, or the rad52-329 allele that results in the loss of the interaction between Rad51 and Rad52, stimulates translocation formation ( and ). This suggests that filament formation can inhibit a mechanistically diverse set of non-conservative HR events. Further, it suggests that releasing the attenuating effect of filament formation would have a broadly destabilizing effect on the genome.
Previously, the attenuating effect of RAD51
on SSA, such as its effect on T2 was interpreted as reflecting a competition between separate apparatus for strand invasion-dependent and -independent events ( and )
. While competition between these processes may contribute to the balance between conservative and non-conservative HR, the genetic data described in this paper clearly suggested that it is not simply the presence or absence of filaments at DSBs that steers them toward conservative and away from non-conservative mechanisms of repair. Epistasis analysis clearly suggested that Rad51 filament formation specifies the necessity for the canonical SSA machinery encoded by SRS2
as these genes were largely dispensable for T2 in rad51
Δ and rad52-329/rad52-329
homozygotes ( and ). This suggests that in wild type cells, filaments are present at DSBs regardless of whether they are engaged in strand invasion-dependent or –independent repair. Further, this suggests the existence of a novel, and highly efficient alternative mechanism of non-conservative HR that resembles SSA but displays distinct genetic control.
Previous investigation into the interaction between Srs2 and Rad51 during the formation of chromosomal deletions by SSA suggested that loss of Srs2 does not inhibit SSA, but, instead inhibits recovery from a Rad51-dependent DNA damage checkpoint 
. Therefore, the reduced frequencies of translocation formation observed in srs2
Δ homozygotes (), as well as the epistasis interactions between srs2
Δ, and rad1
Δ () could be due to effects on checkpoint recovery. This possibility was explored by determining the plating efficiency before and after DSB formation in srs2
Δ homozygotes, and srs2
Δ and srs2
Δ double homozygotes (Table S3
). Because plating efficiencies in the mutants were not reduced from wild type levels, changes in checkpoint recovery were not indicated. This suggests that the effects of srs2
Δ on translocation frequency were unlikely to be due to changes in checkpoint recovery. Failure to observe altered checkpoint recovery in our experiments may be due to our use of diploid strains, where broken chromosomes have intact homologs with which to pair and attenuate the checkpoint response. Alternatively, they may be due to the fact that the DSBs occur very close to the translocation substrates, obviating the necessity to create extensive lengths of single-stranded DNA before complementary sequences are revealed.
The T2 that proceeds in the absence of Rad51, while independent of the bulk of the canonical SSA machinery, is markedly more dependent on Rad52, as translocation frequencies are 21- and 108-fold lower with the 60 bp and 300 bp substrates in the rad51
Δ double homozygotes than in the rad52
Δ homozygotes ( and ). Perhaps, the much less dramatic effect of the loss of Rad52 in the presence of Rad51 reflects the ability of Rad59, which also possesses single-stranded DNA annealing activity to act like Rad52, a notion supported by the synergistically reduced frequencies of T2 observed previously in rad52
Δ double homozygotes 
. However, experiments suggest that Rad51 can only form nucleoprotein filaments when Rad52 is present to displace RPA from single-stranded DNA, and Rad59 can neither displace RPA from single-stranded DNA, nor anneal single-stranded DNA molecules bound by RPA 
. This suggests that the contributions of Rad51 and Rad59 to T2 in the rad52
Δ homozygotes may not involve their functions in filament formation or annealing, While the role of Rad59 could possibly be related to its participation in the removal of non-homologous tails, the fact that loss of filament formation conferred by rad51
Δ and rad52-329
both suppress the necessity for Rad59 counters that suggestion as the complete absence of Rad52 conferred by rad52
Δ should also block filament formation ( and ). The role of Rad51 in translocation formation by SSA clearly merits further investigation.
While the role of Rad51 filaments in T2 is unknown, the genetic data described here, and previously support the following speculative model (, –)
. Following the creation of Rad51 filaments at DSBs, Srs2 removes Rad51, permitting the complementary sequences on 3′ single strands to be annealed by Rad52 and Rad59. Rad59 then collaborates with Msh2-Msh3 and Rad1-Rad10 to coordinate the removal of the non-homologous tails created by annealing the complementary segments of the 3′ single strands. Epistasis analysis suggests that Srs2 may also play a role in removing Rad51 that remains associated with non-homologous tails, or has reassociated, which may inhibit their removal by Rad1-Rad10 (). Removal of the tails facilitates repair synthesis and ligation, forming a covalent joint and a translocation chromosome. In contrast, when Rad51 filaments do not form at DSBs, Rad52 anneals complementary single strands without the assistance of Srs2 and Rad59. Without Rad59 specifying the use of Msh2-Msh3 and Rad1-Rad10 for the removal of non-homologous tails, other factors execute this task so that repair synthesis and ligation can complete the recombination event 
. All aspects of this model are currently under investigation at the genetic and molecular levels.
Because all eukaryotes likely share with budding yeast a genomic structure that features an abundance of short repetitive sequences, non-conservative recombination between unlinked repeats must be minimized in favor of conservative recombination so that genome stability can be maintained. Also conserved from yeast to man is the central strand exchange protein, Rad51, which facilitates conservative HR, even between non-allelic sequences (, )
. While Rad52 is also conserved in most eukaryotes, its function in facilitating nucleoprotein filament formation by Rad51 appears to have been replaced in many species by Brca2 
. Fascinatingly, like the rad51
Δ and rad52-329
mutations in yeast, mutations that disable Rad51, or disrupt the interaction between Brca2 and Rad51 in worms, flies and mammals block conservative GC and stimulate non-conservative SSA 
. Given the parallels with yeast, it is tempting to speculate that defects in filament formation in higher eukaryotes release the restriction against a mechanism of SSA that is analogous to the one described here. Given the probable link between SSA and the non-allelic HR observed in tumors and during genome evolution, revealing the molecular basis of this mechanism could have far reaching implications