Resolving an ancient, rapid radiation in Saxifragales

Area of Science:

• Phylogenomics
• Plant evolutionary biology
• Angiosperm radiation

Background:

• Deep-level relationships in Saxifragales remain unclear despite extensive prior genetic analysis.
• An ancient and rapid radiation event complicates phylogenetic resolution within this angiosperm clade.

Purpose of the Study:

• To resolve deep phylogenetic relationships within the Saxifragales clade.
• To investigate the impact of data size and type on phylogenetic accuracy.
• To accurately estimate the timing of the early diversification of Saxifragales.

Main Methods:

• Construction of large multi-gene datasets (up to 50,845 bp) from nuclear, plastid, and mitochondrial compartments.
• Phylogenetic analyses using Bayesian inference (BI) and maximum likelihood (ML) methods.
• Age estimation using penalized likelihood, assessing the impact of taxon sampling and data size.

Main Results:

• BI and ML analyses consistently resolved deep relationships with high support, indicating that sufficient data can overcome rapid radiations.
• Parsimony analyses produced conflicting topologies, particularly for long-branched taxa like Paeoniaceae, with increased data sometimes reinforcing incorrect placements.
• Slowly evolving plastid genes provided better resolution for deep divergences than rapidly evolving genes, while rapidly evolving genes supported relationships closer to the tips.
• Accurate age estimations required sufficient taxon density, not just large datasets, with early Saxifragales diversification estimated to have occurred rapidly over ~6 million years.
• Between 25,000 and 50,000 bp were necessary to achieve high support for resolving this radiation.

Conclusions:

• Large-scale phylogenomic data, particularly from slowly evolving plastid regions, are effective for resolving deep angiosperm divergences.
• Taxon sampling density is critical for accurate molecular dating of plant radiations.
• Resolving ancient, rapid radiations in angiosperms necessitates substantial genomic datasets (tens of thousands of base pairs).