Nonhomologous tails direct heteroduplex rejection and mismatch correction during single-strand annealing in Saccharomyces cerevisiae.

Single-strand annealing (SSA) is initiated when a double strand break (DSB) occurs between two flanking repeated sequences, resulting in a deletion that leaves a single copy of the repeat. We studied budding yeast strains carrying two 200-bp URA3 sequences separated by 2.6 kb of spacer DNA (phage lambda) in which a site-specific DSB can be created by HO or Cas9 endonucleases. Repeat-mediated deletion requires removal of long 3'-ended single-stranded tails (flaps) by Rad1-Rad10 with the assistance of Msh2-Msh3, Saw1 and Slx4. A natural 3% divergence of unequally spaced heterologies between these repeats (designated F and A) causes a significant reduction in the frequency of SSA repair. This decrease is caused by heteroduplex rejection in which mismatches (MMs) in the annealed intermediate are recognized by the MutS (Msh2 and Msh6) components of the MM repair (MMR) pathway coupled to unwinding of the duplex by the Sgs1-Rmi1-Top3 helicase. MutL homologs, Mlh1-Pms1 (MutL), are not required for rejection but play their expected role in mismatch correction. Remarkably, heteroduplex rejection is very low in strains where the divergent repeats were immediately adjacent (Tailless strains) and the DSB was induced by Cas9. These results suggest that the presence of nonhomologous tails strongly stimulates heteroduplex rejection in SSA. DNA sequencing analysis of SSA products from the FA Tailed strain showed a gradient of correction favoring the sequence opposite each 3' end of the annealed strand. Mismatches located in the center of the repair intermediate were corrected by Msh2-Msh6 mediated mismatch correction, while correction of MMs at the extremity of the SSA intermediate often appears to use a different mechanism, possibly by 3' nonhomologous tail removal that includes part of the homologous sequence. In contrast, in FA Tailless strains there was a uniform repair of the MMs across the repeat. A distinctive pattern of correction was found in the absence of MSH2, in both Tailed and Tailless strains, different from the spectrum seen in a msh3Δ msh6Δ double mutant. Previous work has shown that SSA is Rad51-independent but dependent on the strand annealing activity of Rad52. However Rad52 becomes dispensable in a Tailless construct where the DSB is induced by Cas9 or in transformation of a plasmid where SSA occurs in the absence of nonhomologous tails. Author summary: Deletions between dispersed, divergent, repeated sequences are frequently found among cancer mutations. We have studied deletion formation by single-strand annealing (SSA) between naturally divergent 200 bp repeats in budding yeast. SSA was initiated by creating a site-specific double-strand break (DSB) in the region between the two repeats, either using HO endonuclease or guide RNA-directed Cas9. In this process, the ends are resected 5' to 3', leaving long 3'-ended single-stranded sequences that can anneal. After annealing any 3'-ended nonhomologous tails are clipped off by the Rad1-Rad10 endonuclease. The presence of nonhomologous tails triggers a Msh2-Msh3-dependent and Sgs1-dependent heteroduplex rejection when there are 7 heterologies within the 200-bp repeats being annealed, whereas a "tailless" construct is not rejected. These results suggest that the presence of nonhomologous tails strongly stimulates heteroduplex rejection in SSA. Mismatch repair within the heteroduplex is Msh2-Msh6-dependent; however, a msh2 deletion proved have a spectrum of mismatch repair different from a msh3 msh6 double mutant, implying that Msh2 may have some additional role when the intermediates have nonhomologous 3' tails. "Tailed" repeats yielded a gradient of repair that favored sequences in the "left" repeat, but "Tailless" SSA yielded a uniform correction of all 7 heterologies in the 200-bp region. A cluster of repeats near the left end of the sequence provokes correction of these markers in a Msh6-independent fashion; this correction is not attributable to 3' to 5' exonuclease activity of DNA polymerase δ and may reflect instances where part of the homologous sequence itself is clipped off by Rad1-Rad10. The presence or absence of nonhomologous tails also affected the requirement for the Rad52 strand annealing protein. Both in the chromosomal SSA assay and in a plasmid assay, Rad52 was largely dispensable when there were no nonhomologous tails after strand annealing. [ABSTRACT FROM AUTHOR]