Role of homology in site-specific recombination of bacteriophage lambda: evidence against joining of cohesive ends.

Bacteriophage lambda integration and excision take place at specific loci called attachment sites. Earlier work has shown that efficient recombination requires the identical sequence to be present in both attachment sites throughout the seven-base-pair region between the points of strand exchange. A plausible model for the role of homology postulates that Int, the site-specific recombinase, makes double-strand breaks at attachment sites such that each broken end has a short single-strand protrusion. Recombination would then depend upon the capacity of these protrusions to form Watson-Crick helices--i.e., to anneal--a process that might require perfect complementarity between the cohesive ends. To test this model, we have studied Int-promoted crosses in which one attachment site is a heteroduplex. Specifically, we constructed sites in which the seven-base-pair region between the points of strand exchange contains one or more noncomplementary pairs. The double-strand break and annealing mechanism predicts that crosses with these heteroduplex sites should yield one completed recombinant and one broken site. We find that such nonreciprocal recombination is uncommon and that the typical outcome of crosses involving a heteroduplex site is a reciprocal recombinant in which both products are resealed. Moreover, the occasional appearance of nonreciprocal products can be explained by our finding that Int can cleave heteroduplex attachment sites after recombination is completed. Taken together, our data strongly indicate that bacteriophage lambda recombination does not proceed by the homology-dependent annealing of cohesive ends; acceptable alternatives for the role of homology are discussed.