Does crossover interference count in Saccharomyces cerevisiae?

We previously proposed a "counting model" for meiotic crossover interference, in which double-strand breaks occur independently and a fixed number of noncrossovers occur between neighboring crossovers. Whereas in some organisms (group I) this simple model alone describes the crossover distribution, in other organisms (group II) an additional assumption--that some crossovers lack interference--improves the fit. Other differences exist between the groups: Group II needs double-strand breaks and some repair functions to achieve synapsis, while repair in group I generally occurs after synapsis is achieved; group II, but not group I, has recombination proteins Dmc1, Mnd1, and Hop2. Here we report experiments in msh4 mutants that are designed to test predictions of the revised model in a group II organism. Further, we interpret these experiments, the above-mentioned differences between group I and II meiosis, and other data to yield the following proposal: Group II organisms use the repair of leptotene breaks to promote synapsis by generating double-Holliday-junction intermediates that lock homologs together (pairing pathway). The possible crossover or noncrossover resolution products of these structures lack interference. In contrast, for both group I and group II, repair during pachytene (disjunction pathway) is associated with interference and generates only two resolution types, whose structures suggest that the Holliday junctions of the repair intermediates are unligated. A crossover arises when such an intermediate is stabilized by a protein that prevents its default resolution to a noncrossover. The protein-binding pattern required for interference depends on clustering of sites that have received, or are normally about to receive, meiotic double-strand breaks.