Data from: Allele phasing is critical to revealing a shared allopolyploid origin of Medicago arborea and M. strasseri (Fabaceae)

Background: Whole genome duplication plays a central role in plant
evolution. There are two main classes of polyploid formation:
autopolyploids which arise within one species by doubling of similar
homologous genomes; in contrast, allopolyploidy (hybrid polyploidy) arise
via hybridization and subsequent doubling of nonhomologous (homoeologous)
genomes. The distinction between polyploid origins can be made using gene
phylogenies, if alleles from each genome can be correctly retrieved. We
examined whether two closely related tetraploid Mediterranean shrubs
(Medicago arborea and M. strasseri) have an allopolyploid origin - a
question that has remained unsolved despite substantial previous research.
We sequenced and analyzed ten low-copy nuclear genes from these and
related species, phasing all alleles. To test the efficacy of allele
phasing on the ability to recover the evolutionary origin of polyploids,
we compared these results to analyses using unphased sequences. Results:
In eight of the gene trees the alleles inferred from the tetraploids
formed two clades, in a non-sister relationship. Each of these clades was
more closely related to alleles sampled from other species of Medicago, a
pattern typical of allopolyploids. However, we also observed that alleles
from one of the remaining genes formed two clades that were sister to one
another, as is expected for autopolyploids. Trees inferred from unphased
sequences were very different, with the tetraploids often placed in poorly
supported and different positions compared to results obtained using
phased alleles. Conclusions: The complex phylogenetic history of M.
arborea and M. strasseri is explained predominantly by shared
allotetraploidy. We also observed that an increase in woodiness is
correlated with polyploidy in this group of species and present a new
possibility that woodiness could be a transgressive phenotype. Correctly
phased homoeologues are likely to be critical for inferring the hybrid
origin of allopolyploid species, when most genes retain more than one
homoeologue. Ignoring homoeologous variation by merging the homoeologues
can obscure the signal of hybrid polyploid origins and produce inaccurate
results.