Under one type of SI (gametophytic SI), a model to describe the evolutionary processes of the S-locus can be constructed using only two parameters: the effective population size and the rate at which mutation changes the mating specificity. Due to this tractability, we can construct explicit statistical tests to examine and compare the evolutionary history of SI plants.
For instance, why the S-locus genealogies of two genera in the Solanaceae (Physalis and Solanum) exhibit such dramatically different tree shapes? The differences between the two genera seem to be caused by a higher bifurcation rate within the Physalis tree relative to that of Solanum. Our model derives a maximum likelihood estimate of relative coalescence rates in the two groups, and shows a 9.9-fold difference between groups (Uyenoyama and Takebayashi 2003). We developed a statistical test based on coalescence theory to distinguish between two possible explanations for the differing coalescence rates: (1) effective population sizes differed between genera, or (2) some other aspect of the S-locus diversification process differs between genera. Our test clearly rejects the first hypothesis, and suggests that more complex models are needed; models that consider heterogeneous evolutionary processes between the two genera. The diversification processes of S-alleles may differ, for example, if the two genera experienced different historical population growth rates, or if there are differences in the deleterious mutations around the S-locus region where recombination is suppressed (see Uyenoyama 2003, TPB, 63: 281-293). These complexities are now being incorporated into the model.
To aid in the search for the pollen-S, we have developed a statistical analysis of DNA sequence to identify characteristics expected for a pollen-S gene (Takebayashi et al. 2003). A pollen-S candidate gene, 48A, was recently isolated from Nicotiana alata, and we used our predictions to determine if 48A might actually be a pollen-S gene. The answer is no. Within pistil-S and pollen-S genes, the amino acid residues determining the mating specificity should experience strong diversifying selection. First, we applied an empirical Bayesian approach originally developed by Yang and his colleagues for identification of positively selected codon sites. Positively selected sites were identified in the known pistil-S, but not in 48A. Another expected characteristic of pollen-S is complete linkage to the pistil-S. Under complete linkage, neutral regions of S-RNase and pollen-S should show similar genealogies. We developed and applied several statistics to test this null hypothesis of complete linkage, and found indications of historical recombination between the pistil-S and 48A.
Although our work shows that 48A is not the pollen-S gene, the approach we developed here will be useful in screening other candidate loci, and eventually identifying the real pollen-S. There are many open reading frames (ORFs) around the S-locus region, and our sequence analysis is a much more efficient way to screen pollen-S candidate ORFs than the tedious and ineffectual crosses previously required to observe recombinants.