Evidence for nonindependent evolution of adjacent microsatellites in the human genome.

Microsatellites are short tandem repeats that evolve predominantly through a stepwise mutation model. Despite intensive study, many aspects of their evolution remain unresolved, particularly the question of how compound microsatellites containing two different motifs evolve. Previous work described profound asymmetries in the probability that any given second motif lies either 3' or 5' of an AC repeat tract. Here we confirm and extend this analysis to examine the length dependence of these asymmetries. We then use the differences in length between homologous human and chimpanzee microsatellites as a surrogate measure of the slippage-based mutation rate to explore factors that influence this process. We find that the dominant predictor of mutation rate is the length of the tract being considered, which is a stronger predictor than the length of the two tracts combined, but other factors also have a significant impact, including the length of the second tract and which of the two tracts lies upstream. We conclude that compound microsatellites rarely arise through random point mutations generating a second motif within a previously pure tract. Instead, our analyses point toward a model in which poorly understood mutation biases, probably affecting both slippage and point mutations and often showing 3'-5' polarity, promote the formation of compound microsatellites. The result is convergent evolution. We suggest that, although their exact nature remains unclear, these biases are likely attributable to structural features, such as the propensity of AC tracts to form Z-DNA.