Placing very long branch taxa in the plant tree of life: a case study with Dioscoreales mycoheterotrophs

Premise: The inclusion of heterotrophic plants in broadly focused systematic studies is challenging, reflecting substantial morphological modification, gene loss, and accelerated nucleotide substitution following photosynthesis loss. In particular, inclusion of very long branch (VLB) taxa with highly elevated rates may lead to phylogenomic misinference. Methods: We explore how rapidly evolving plastomes behave in model-based phylogenetic inferences of the yam order Dioscoreales, which experienced convergent photosynthesis losses. The taxon sampling includes all photosynthetic genera, and mycoheterotrophic lineages with a broad range of substitution rates. Results: When only moderately rapidly evolving heterotrophs are included, relationships within Dioscoreales are congruent with recent mitochondrial analysis, and generally strongly supported. Photosynthetic Stenomeris (represented by only three genes) is moderately supported as sister to Dioscoreaceae. Notably, Burmanniaceae and Thismiaceae, two mycoheterotrophic families including highly reduced plastomes, are strongly supported as distantly related within Dioscoreales. Constrained analysis rejects combining Thismiaceae with Burmanniaceae, conflicting with current angiosperm classification. Including VLB mycoheterotrophs in phylogenetic inference can lead to localized to broad reduction in branch support, substantially so for Afrothismiaceae, a family placing in disparate locations in Dioscoreales (and beyond) in variant analyses, with no ability to differentiate among its alternative placements. Conclusions: Inclusion of highly rate-elevated taxa in analyses of Dioscoreales phylogeny can depress branch support. Inferred relationships in the order are otherwise congruent with studies based on mitochondrial data; the family-level classification needs updating. The extraordinarily rapidly evolving family Afrothismiaceae places inconsistently in diverse plastid analyses, pointing to probable analytical limits for plastid-based phylogenomic analyses of VLB heterotrophs.