UC San Diego

Abstract Background The S-RNases of the Solanaceae are highly polymorphic self-incompatibility (S-) alleles subject to strong balancing selection. Relatively recent diversification of S-alleles has occurred in the genus Physalis following a historical restriction of S-allele diversity. In contrast, the genus Solanum did not undergo a restriction of S-locus diversity and its S-alleles are generally much older. Because recovery from reduced S-locus diversity should involve increased selection, we employ a statistical framework to ask whether S-locus selection intensities are higher in Physalis than Solanum. Because different S-RNase lineages diversify in Physalis and Solanum, we also ask whether different sites are under selection in different lineages. Results Maximum-likelihood and Bayesian coalescent methods found higher intensities of selection and more sites under significant positive selection in the 48 Physalis S-RNase alleles than the 49 from Solanum. Highest posterior densities of dN/dS (ω) estimates show that the strength of selection is greater for Physalis at 36 codons. A nested maximum likelihood method was more conservative, but still found 16 sites with greater selection in Physalis. Neither method found any codons under significantly greater selection in Solanum. A random effects likelihood method that examines data from both taxa jointly confirmed higher selection intensities in Physalis, but did not find different proportions of sites under selection in the two datasets. The greatest differences in strengths of selection were found in the most variable regions of the S-RNases, as expected if these regions encode self-recognition specificities. Clade-specific likelihood models indicated some codons were under greater selection in background Solanum lineages than in specific lineages of Physalis implying that selection on sites may differ among lineages. Conclusions Likelihood and Bayesian methods provide a statistical approach to testing differential selection across populations or species. These tests appear robust to the levels of polymorphism found in diverse S-allele collections subject to strong balancing selection. As predicted, the intensity of selection at the S-locus was higher in the taxon with more recent S-locus diversification. This is the first confirmation by statistical test of differing selection intensities among self-incompatibility alleles from different populations or species.