A new method is described for determination of the origin of polyploid lineages. It tests the hypothesis that a tetraploid lineage originated via autopolyploidization vs. allopolyploidization. The method is based on the hypothesis that, in the case of autopolyploidy, any genetic marker in the first tetraploid ancestor is represented by two copies (one for each homoeologous chromosome of the haploid complement), whereas in allopolyploidy some markers absent from one of the hybridizing species will display one copy at most. The model requires knowledge of the phylogeny (topology and branch lengths) of a sample of species descending from the same tetraploidization event, together with the number of homoeologous copies present in each species for a set of neutral markers. The likelihood of a given proportion of the markers being present in both homoeologous chromosome pairs of the ancestral tetraploid is expressed as a function of the deletion rate of a marker. In the case of an autopolyploid origin, this proportion equals one. A likelihood-ratio test was carried out to test this hypothesis. The method was used to examine five microsatellite loci in eight species of Barbus (sensu lato). Assuming the validity of the hypotheses on phylogenetic relationships and evolutionary rates, the test rejects the possibility that European tetraploid barbs originated through autopolyploidy. This is the first test that can reject autopolyploidy, and it would appear particularly useful for phylogenetic studies in taxa where hybridization is known and where, consequently, undetected reticulate evolution may impair phylogenetic reconstruction.
