Target capture data resolve recalcitrant relationships in the coffee family (Rubioideae, Rubiaceae)

Subfamily Rubioideae is the largest of the main lineages in the coffee
family (Rubiaceae), with over 8,000 species and 29 tribes. Phylogenetic
relationships among tribes and other major clades within this group of
plants are still only partly resolved despite considerable efforts. While
previous studies have mainly utilized data from the organellar genomes and
nuclear ribosomal DNA, we here use a large number of low-copy nuclear
genes obtained via a target capture approach to infer phylogenetic
relationships within Rubioideae. We included 101 Rubioideae species
representing all but two (the monogeneric tribes Foonchewieae and
Aitchinsonieae) of the currently recognized tribes, and all but one
non-monogeneric tribe were represented by more than one genus. Using data
from the 353 genes targeted with the universal Angiosperms353 probe set we
investigated the impact of data type, analytical approach, and potential
paralogs on phylogenetic reconstruction. We inferred a robust phylogenetic
hypothesis of Rubioideae with the vast majority (or all) nodes being
highly supported across all analyses and datasets and few incongruences
between the inferred topologies. The results were similar to those of
previous studies but novel relationships were also identified. We found
that supercontigs (coding sequence [CDS] + noncoding sequence) clearly
outperformed CDS data in levels of support and gene tree congruence. The
full datasets (353 genes) outperformed the datasets with potential
paralogous genes removed (186 genes) in levels of support but increased
gene tree incongruence slightly. The pattern of gene tree conflict at
short internal branches was often consistent with high levels of
incomplete lineage sorting (ILS) due to rapid speciation in the group.
While concatenation- and coalescence-based trees mainly agreed, the
observed phylogenetic discordance between the two approaches may be best
explained by their differences in accounting for ILS. The use of target
capture data greatly improved our confidence and understanding of the
Rubioideae phylogeny, highlighted by the increased support for previously
uncertain relationships and the increased possibility to explore sources
of underlying phylogenetic discordance.