BURLEIGH, J. GORDON and SARAH MATHEWS.* Division of Biological Sciences, University of Missouri, Columbia, 65211. - Resolving ancient divergences in molecular phylogenetic analyses: sources of error and bias in a 9-locus data set from seed plants.
Relationships among the extant seed plants, angiosperms, conifers,
cycads, ginkgos, and Gnetales, remain controversial. Recent molecular
analyses refute the anthophyte hypothesis, which unites Gnetales and
angiosperms, but disagree with respect to an alternative hypothesis.
Different molecular trees suggest that Gnetales are sister to
Pinaceae, to conifers, or to all other seed plants. These analyses are
complicated by ancient divergence, variation in substitution rates
among lineages, and extinctions. We analyzed a nine-locus dataset from
25 seed plants to infer relationships and to test for bias. The
dataset included published sequences from the nuclear, chloroplast,
and mitochondrial genomes and some unpublished phytochrome B
sequences. Bayesian and parsimony analyses strongly support the
gnepine hyothesis, in which Gnetales and Pinaceae are sister taxa.
Moreover, Templeton tests reject the anthophyte hypothesis and the
gnetifer hypothesis, in which Pinaceae are sister to conifers.
However, results from parametric bootstrapping indicate bias in the
data toward the gnepine tree. When mitochondrial atpA and plastid psaA
are simulated on a gnetifer tree, analyses of the data infer gnepine
trees. When these data are excluded from analyses, a gnetifer tree is
inferred. Triplet likelihood ratio tests provide evidence that
organellar sequences from Gnetales are evolving significantly faster
than those from other seed plants, suggesting that bias toward the
gnepine hypothesis might result from the very long branches to
Gnetales. To test this, we used a maximum likelihood approach to
identify the most rapidly evolving nucleotide sites and then excluded
or down-weighted them in subsequent analyses. These analyses support
the gnetifer hypothesis. Our results suggest that resolving deep
divergences throughout the tree of life will require more than
additional data from the rapidly expanding databases. Rather, they
emphasize the need for careful analyses of the data to identify
potential sources of conflict and bias.
Key words: Bayesian analyses, multigene dataset, parsimony, relative evolutionary rates, seed plant phylogeny, tests for bias