Single Strand Annealing

Evidence for SSA in yeast was observed when it was found that the HO endonuclease could stimulate deletions events between ura3 sequences in vivo. Reciprocal crossover products could not be identified by Southern hybridization analysis, suggesting a non-conservative process such as SSA. Subsequently these deletion events were characterized and it was observed that as repeated sequences were moved farther apart, deletions occurred less frequently and arose at a later times. This is consistent with the proposal that it takes more time for a nuclease to reach repeats located further from a DSB.

Also consistent with SSA model was the observation that if several repeats are present, deletions preferentially occur between repeats closest to the break. In this experiment a DSB preferentially caused a deletion between B and C rather than A and C.

Single strand annealing was also characterized in several other respects. We have measured its homology dependence and showed that the product formation increases with increasing amounts of homology. Product formation plateaus at about 400 bp and the smallest amount of homology that we have tested was 29 bp which recombined at a level of 0.1%. Interestingly, if one repeat is immediately adjacent to the DSB, as little as 10 bp can be used at a very efficient frequency. We have also separated the repeats by as much as 25 kb and still obtain efficient repair by SSA deletion events. This process takes 6 hours during which time the cells arrest their cell cycle until the DSB can be repaired. This experiment suggests that the tails remain relatively stable during this time.

Genetic Requirements.

SSA requires several gene products that are also required for other types of recombination assays. Rad52p, for example, is required for most if not all types of recombination processes including SSA. Rad52 was shown to be a DNA binding protein that can anneal ssDNA in vitro and in its absence Rad51 fails to bind to the single-stranded DNA tails. Rad52 isolated from humans possesses the ability to bind to the ends of DNA. While RAD52 is clearly required for SSA using ura3 and leu2 substrates it has been reported that RAD52 is not required for SSA events located within the ribosomal DNA array (100 repeats of a 9 kb sequence) and is partially required for SSA within a CUP1 array of repeats. We suggest that large amounts of homology can compensate for deficiencies created by rad52 mutations.

RAD59.A homolog of RAD52, RAD59, is also required for SSA. Rad59 mutants were isolated in a rad51 background and based on their relationship in some epistasis assays, may define two recombination pathways. In view of its homology with RAD52, it was found that over-expression of RAD52 could compensate for certain deficiencies of rad59. Purified Rad59 possesses DNA binding properties and strand annealing activity.

RPA. Another DNA binding factor, the RPA complex, is required for SSA. An allele of RPA1, rpa1-t11, is partially defective for SSA but is still capable of carrying out its role in DNA replication as cells appear to grow normally. The defect is more apparent when the repeats in the SSA substrate are small.

RAD50, MRE11, XRS2, RAD51, RAD54, RAD55 and RAD57. Interestingly Rad51 is not required for SSA. Rad51 is the homolog of the bacterial recA gene and is required for DSB-induced gene conversion. Rad51 may be involved in strand invasion or strand exchange, and hence might not be expected to play a role in SSA, where two strands interact by intertwining around each other. Rad54, Rad55 and Rad57 are also not required for SSA, although they are required for DSB-induced gene conversion. Another RAD gene, RAD50, is partially required for SSA. Rad50 forms a complex with Mre11 and Xrs2. Deletions of rad50 or xrs2 slow down the formation of single stranded DNA and reduce the amount of product formed.

Interestingly Rad51 is not required for SSA. Rad51 is the homolog of the bacterial recA gene and is required for DSB-induced gene conversion. Rad51 may be involved in strand invasion or strand exchange, and hence might not be expected to play a role in SSA, where two strands interact by intertwining around each other. Rad54, Rad55 and Rad57 are also not required for SSA, although they are required for DSB-induced gene conversion. Another RAD gene, RAD50, is partially required for SSA. Rad50 forms a complex with Mre11 and Xrs2. Deletions of rad50 or xrs2slow down the formation of single stranded DNA and reduce the amount of product formed

Mismatch Repair Proteins and RAD1-RAD10. The mismatch repair proteins Msh2 and Msh3 as well as Rad1 and Rad10 are required for efficient SSA and appear to be needed to remove the non-homologous 3' tails from the annealed intermediate. A gene conversion assay devised to test the ability of various mutants to remove non-homologous sequences adjacent to the DSB implicates these four genes as being necessary for non-homologous tail removal. Several research groups have shown that Msh2 and Msh3 form a complex in vivo and have a strong preference for recognizing "loop-out" structures such as those formed by frame shift replication errors. Msh2 has been implicated in binding to palindromic structures formed during meiotic recombination in vivo and has also been shown to bind to Holiday junctions in vitro. Based on these observations we proposed that Msh2 and Msh3 bind to the branched junction between the single and double stranded DNA. The complex stabilizes the annealed intermediate and/or signals the Rad1-Rad10 endonuclease to cleave the single stranded tail. Msh2 and Msh3 are dispensable if the length of homology is large (1 kb), suggesting that they may act to stabilize junctions where the repeats are small (0.2 kb).

We have also used SSA as a tool to study mismatch repair. To do this, we have introduced mismatches between the repeated sequences such that the mismatch repair proteins will repair the heteroduplex to one or the other sequence. The mismatch repair proteins will also lead to a reduction in the amount of product formed in a process termed heteroduplex rejection. Many models can be envisioned; perhaps the easiest is to imagine that a nuclease excises the mismatches but proceeds too far and destroys the intermediate.

We have analyzed which mismatch repair proteins are important for these mismatch repair and heteroduplex rejection, and have found that Msh6 and Msh2 are important for both processes. The mutL homologs, PMS1, MLH1, MLH2 and MLH3 are also important for both processes but the situation appears to be more complex due to some apparent redundancy among the components. 

Heteroduplex Rejection via Strand Unwinding. One can imagine several ways that heteroduplex rejection may occur. For example one mechanism might be similar to mismatch repair and involve the excision of mismatched base pairs. If the excision is extensive, the SSA intermediate would be destroyed. The SSA intermediate would also be destroyed, if two excisions started on different strands. An alternative model proposes that the strands are unwound so that they can search for a better match.

To explore these possibilities we made two strains containing three repeats that can result in two SSA products. When the repeats are identical both products are made at about equal levels. When heterologies are introduced into the middle repeat, recombination with that repeat goes down and the other goes up.

Overall viabilities of the two strains are about equal. This is important since the "destruction" models predict a decrease in viability when heterologies are present. On the other hand, the "unwinding" model predicts equal viabilities, if the unwound strands are not destroyed and are able to search for a more perfect match.

Supporting this model, was the finding that deletion of the Sgs1 helicase gene caused a loss of heteroduplex rejection. Product levels using mismatched repeats approached that of substrates with identical repeats. Interestingly, while mutants with sgs1 are defective in heteroduplex rejection they are not defective in mismatch repair indicating that they are separable pathways. Loss of the Srs2 helicase did not have an effect on heteroduplex rejection. Similarly loss of the nuclease ExoI did not have an effect.

Biological Role. A question that is often raised is why would organisms evolve a repair mechanism that deletes genetic material that might be essential for survival? One possibility is that SSA is well suited to repair DSBs that occur within tandem arrays of sequences such as those found within the rDNA locus. A DSB would initiate SSA resulting in a deletion event of redundant genes which can be restored by mechanisms that normally maintain the copy number of the arrays. The experiments above suggest that heteroduplex rejection provides another means to maintain homogeneity of tandem arrays in addition to mismatch repair.

Another consideration is that SSA is only one of several mechanisms that can repair a DSB. SSA offers a way for a cell to possibly survive, if gene conversion is too slow or is not an option.