Testing the covarion hypothesis of molecular evolution

The covarion hypothesis of molecular evolution states that the fixation ofmutations may alter the probability that any given position will fix thenext change. Tests of this hypothesis using the divergence of realsequences are compromised because models of rate variation among sites(e.g., the gamma version of the one-parameter equation) predict sequencedivergence values similar to those for the covarion process. This studytherefore focuses on the extent to which the varied and unvaried codons oftwo well-diverged taxa are the same, because fewer are expected by thecovarion hypothesis than by the gamma model. The data for these tests arethe protein sequences of Cu, Zn superoxide dismutase (SOD) for mammals andplants. Simulation analyses show that the covarion hypothesis makes betterpredictions about the frequencies of varied and unhit positions in commonbetween these two taxa than does the gamma version of the one-parametermodel. Furthermore, the analysis of SOD tertiary structure demonstratesthat mammal and plant variabilities are distributed differently on theprotein. These results support the conclusions that the variable andinvariable codons of mammal and plant SODs are different and that thecovarion model explains the evolution of this protein better than the gammaversion of the one-parameter process. Unlike other models, the covarionhypothesis accounts for rate fluctuations among positions over time, whichis an important parameter of molecular evolution.