Selection versus demography: a multilocus investigation of the domestication process in maize.

The domestication of maize (Zea mays ssp. mays) from its wild ancestor (Zea mays ssp. parviglumis) led to a loss of genetic diversity both through a population bottleneck and through directional selection at agronomically important genes. In order to discriminate between those effects and to investigate the nature of the domestication bottleneck, we analyzed nucleotide diversity data from 12 chromosome 1 loci in parviglumis. We found an average loss of nucleotide diversity of 38% across genes, but this average was skewed downward by four putatively selected loci (tb1, d8, ts2, and zagl1). To better understand the domestication process, we used the coalescent with recombination to simulate bottlenecks under various lengths and population sizes. For each locus, we determine the likelihood of the observed data using three summary statistics: the number of segregating sites, an estimate of the population recombination parameter, and Tajima's D. Based on the eight neutrally evolving loci, a model with a bottleneck had a significantly higher likelihood than a model without one. The four putatively selected loci had significantly different likelihood optimums than the neutral loci, and this approach confirmed that ts2 and d8 were selected either during domestication or breeding. Overall, the best-fitting models had a bottleneck in which the population size and the bottleneck duration had a ratio of approximately 4- to approximately 5; for example, if the initial domestication event occurred over a 500-year period, the population size was roughly 2,000 to 2,500 individuals. However, this range did vary with the summary statistic used to assess the fit of simulations to data. In this context, Tajima's D performed poorly as a goodness-of-fit statistic, probably because Z. mays ssp. parviglumis has a frequency spectrum that is significantly skewed toward low-frequency variants. Finally, we found that demography is unlikely to account for the previously observed positive correlation between nucleotide diversity and the population-recombination parameter in maize, leaving this observation difficult to interpret.