We report a novel genome-wide system in budding yeast to study non-allelic homologous recombination (NAHR) between natural repeats. While previous assays isolate aspects of competitive repair addressed here, our system gauges the competition between all parameters concurrently, as what naturally transpires in a cell. The value of this new approach is evidenced by the surprising features of NAHR our system reveals. Remarkably, in purebred diploids, DSBs within a long stretch of unique sequences are not always repaired by allelic homologous recombination (AHR) as previously assumed. Rather, 17% of these DSBs repair by NAHR. This NAHR arises because the DSB activates Ty recipients 12 to 48 kb distal from the break site to recombine with non-allelic Ty donor sequences. Robust NAHR through break-distal recombination (BDR) is supported by a previous study of bridge-breakage-fusion in diploid budding yeast by Malkova and colleagues 
In this and the previous study, competition between BDR-dependent NAHR and AHR occurs after an endonuclease-induced DSB. In diploids, endonucleases can cleave one homolog prior to DNA replication and both its sister chromatids after DNA replication, thereby eliminating the sister chromatid as a donor for AHR. Therefore, the only AHR donor is the uncut homolog. However, a homolog is also the only AHR donor for repair of spontaneous DSBs that occur on unreplicated DNA in G1 or S. Indeed, recent evidence suggests that spontaneous DSBs occur on unreplicated DNA 
. We suggest that spontaneous DSBs in unique unreplicated DNA are also likely to induce robust BDR-dependent NAHR.
The fact that break-distal Ty sequences undergoes frequent NAHR reveals two surprising features of recombination that have important mechanistic implications for current models of recipient activation and choice. The first surprise is that distal Ty repeats are activated as recipients at all (presumably by becoming single-stranded) when break-proximal ssDNA can undergo AHR. Indeed, a recent study in diploid yeast suggests that ssDNA is generated at least 10 kb from a DSB before its repair is complete 
. To explain this extensive break-distal resection, we suggest that a step after resection must be slow, such as the homology search for donor sequences. A slow homology search would provide time for break-distal sequences to be resected and compete with previously resected break-proximal sequences. Such a slow homology search is consistent with studies suggesting the slow diffusion of chromosomal sequences 
The second surprise is the disproportionate use of very small break-distal Ty sequences as recipients for NAHR. They would represent only a very small proportion of the entire block of resected DNA, which can all act as a recipient for AHR. We suggest that the smaller Ty recipients encounter their potential Ty donors first because chromosome territories 
generate a high local concentration of potential intrachromosomal Ty donors. In contrast, the larger allelic recipients must travel further to partner with allelic donors on the homolog. Consistent with this model, almost all NAHR rearrangements through break-distal Ty recipients result from pairing with intrachromosomal Ty donors.
Along with recipient usage, our genome-wide system reveals the role sequence homology and genomic position play in NAHR donor choice. We find that the Ty donors chosen by a recipient are not among the most homologous in the genome by the criteria of either percent identity or longest block of uninterrupted identity. Rather the primary determinant of NAHR donor choice is local proximity. We observe a ~50-fold preference for Ty repeat donors on the same chromosome over different chromosomes. This intrachromosomal NAHR preference is consistent with previous studies 
, although the magnitude of this preference differs, possibly due to specific configurations of repeats relative to a break site, as observed in our studies. However, in contrast to previous work, our study shows this intrachromosomal bias occurs under conditions that allow unrestricted choice of repair pathways and partners amongst a natural repetitive family. Interestingly, Ty1/Ty2 elements are preferentially inserted within 750 bp upstream of tRNA genes 
, and dispersed tRNA genes cluster together 
. Our results suggest that possible Ty interchromosomal contacts mediated by tRNA clustering is not sufficient to overcome an intrachromosomal bias. It will be interesting to see whether higher-order chromosome organization may influence donor repair choice of natural repeats when only interchromosomal donors are available for NAHR.
Our system also provides insights into the preferred repair pathways that act on a family of natural repeats. We show that NAHR occurs mostly by the SSA pathway whether DSBs occur in unique sequences or a Ty repeat. The robustness of SSA is consistent with previous studies using model repeats 
. Since repair of a single DSB by SSA will occur through an intrachromosomal donor, the predominance of SSA helps explain the preferential usage of intrachromosomal donors and the resulting preference for intrachromosomal NAHR.
Importantly, our pathway analysis of NAHR also helps explain one of the most surprising and striking observations of this study: DSBs that occur outside
repeat clusters are more mutagenic than DSBs that occur inside
repeat clusters. This seemingly counterintuitive observation arises because DSBs that occur inside a Ty have better options for repair, both in efficiency of pathways and favorably positioned donors. DSBs within the Ty predominately repair through two highly efficient pathways, SSA within the Ty locus or GC with preferred intrachromosomal Ty donors 
. These types of repair preserve gene copy number since neighboring unique genes are unaffected. Since SSA and GC are compensatory pathways 
, it is possible that DSBs inside repetitive elements that cannot undergo SSA (i.e. solo insertion of LINE-1) efficiently repair through GC events 
. A recombination execution checkpoint has been suggested to maintain genome integrity by ensuring the coordination of two-ended strand invasion events during GC for conservative repair 
. Consistent with this, our results suggest that NAHR through GC between natural repeats is a major mechanism that limits changes in genome structure.
In contrast, DSBs in unique sequences that repair predominately through GC with the homolog is not as effective in limiting detrimental rearrangements. As the search for the interchromosomal homolog allows for more time to activate a break-distal Ty as a recipient, BDR occurs more frequently through SSA between distinct Ty loci or one-ended events through the BIR pathway. In this situation, SSA always, and BIR often times, change the copy number of neighboring unique genes. Hence, this opens up the possibility that DSBs in unique sequences, rather than repeats, may generate spontaneous or irradiation-induced NAHR-dependent rearrangements observed in yeast 
. Similarly, NAHR-dependent rearrangements in the human genome may also occur by a DSB in the surrounding unique DNA followed by BDR-dependent NAHR. If so, then the recombinant junction would not coincide with the site of the initiating lesion. Therefore, analysis of NAHR junctions alone may miss underlying mechanisms for genome rearrangements. Examining broad regions around NAHR junctions could potentially identify fragile sites that predispose a locus to recurrent instability, contributing to genetic diversity and disease.