gin of the crop. First, we found that genetic variation in the crop is almost entirely a subset of that found in flabellifolia. Flabellifolia contains 24 haplotypes, of which 6 are found in cassava; cassava 's haplotype diversity, therefore, represents 25% of that found in M. esculenta overall. Thus, the crop is most likely derived directly from flabellifolia, rather than from several hybridizing progenitor species as traditionally thought. In addition, we found that the cassava haplotypes occur in flabellifolia populations along the southern border of the Amazon basin and not along the eastern border. This finding points to the southern Amazonian region as the likely site of domestication of cassava. Interestingly, paleobotanical and other anthropological data indicate this region as a probable zone of domestication shared with peanut, two species of chili pepper, and jack bean (Piperno and Pearsall, 1998). Finally, we found that none of the cassava haplotypes occur in M. pruinosa, suggesting that this species is not a progenitor of the crop. All of these conclusions are corroborated by an analysis of this same study system with microsatellite markers (K.M.O. and B.A.S., unpublished work).
The phylogeographic aspect of the study has focused on historical patterns of population divergence in flabellifolia and between flabellifolia and pruinosa. The distribution of the rainforest–cerrado ecotone where these species occur is likely to have shifted during the climatic changes of the Pleistocene (Behling, 1998; Burnham and Graham, 1999). Although there is not yet a consensus on the pattern or extent of habitat shifts, cooler/drier periods (associated with glaciations in temperate latitudes) are expected to have favored the expansion of cerrado and transitional forest; during warmer, humid periods (including the present), these habitats would be expected to be more restricted and fragmented as rainforest expanded. The repeated climate fluctuations of the Pleistocene are therefore predicted to have led to cycles of population fragmentation followed by range expansions and secondary contact in populations of flabellifolia and pruinosa. If these events have occurred, they should be reflected in the present phylogeographic structure of these taxa.
These hypotheses are being tested currently through a nested cladistic analysis (Templeton et al., 1995) of the G3pdh data set (K.M.O., unpublished data). Although the statistical analyses are not complete, some preliminary insights are possible by visual inspection of the G3pdh gene tree. One interesting finding is that three haplotypes are shared between flabellifolia and pruinosa, suggesting interspecific introgression and/or shared ancestral polymorphisms that predate the divergence of these species. Two of the shared haplotypes (E and J) are common in eastern flabellifolia populations, and each is found in a single pruinosa individual from a population in close proximity to flabellifolia populations. This pattern suggests introgression from flabellifolia into M. pruinosa. The position