Phylogenomic discordances reveal conflicting hybridization episodes and widespread lineage-sorting events in...

The grass subtribe Loliinae has great ecological and economic importance, as it includes community-dominant species of mountain grasslands and the most extensively cultivated pasture, fodder and turf grasses (fescues, ryegrasses). Resolving the phylogeny of recently evolved Loliinae lineages has proven challenging due to frequent introgressions and polyploidizations that occurred throughout their history. Here we present the first target-capture phylogeny of Loliinae using 270 orthologous single-copy nuclear coding loci for a large sample of 132 representative taxa, covering all its 29 evolutionary lineages. Additionally, we assembled plastome sequences to complement the inferred hybrid speciation history of the Loliinae. Concatenated maximum likelihood and multispecies coalescent trees of ortho-homeolog single-copy genes showed well-supported relationships for major lineages, which were generally consistent across analyses and genomes, and with previous taxonomic and phylogenetic findings. However, they also revealed high levels of both nuclear and cyto-nuclear discordances estimated to be caused by hybridization and incomplete lineage sorting. We complemented this with a phylogenetic network analysis with representative samples from the main clades to infer reticulation events in the evolution of these grasses. Furthermore, we performed gene tree - species tree reconciliation methods using gene duplication and loss models and multi-labeled trees for polyploidy analysis to estimate the proportion of duplicated genes and the nature of polyploidy of the major Loliinae lineages. These analyses agreed that the Fine-Leaved (FL) Loliinae clade could have evolved from hybridization between more ancestral Broad-Leaved (BL) Loliinae lineages and that both groups underwent ancestral and recent hybridizations. The time-calibrated phylogeny of for the main Loliinae clades supports an early Miocene origin for Loliinae and Mid-Late Miocene splits for its main BL and FL lineages, while current species-rich groups radiated in the Late Miocene. Hybridization tests of nuclear data and topological incongruence assays between the nuclear single-copy genes and plastome-based trees using various approaches and different sampling subsets confirmed the rampant hybridization experienced by Loliinae at deep and shallow nodes. However, hybridization rates differed from lineage to lineage within the major clades and were not correlated with time or ploidy level, but rather depended on their different propensities to hybridize with species within and/or outside their own clade. Our analyses detected high hybridization rates in four broad-leaved (Subulatae-Hawaiian, Tropical-South African, Mexico-Central-South American (MCSA I-II) and Leucopoa p. p.) and five fine-leaved Loliinae lineages (American II, Aulaxyper, Afroalpine, American-Neozeylandic, Australia-Tasmania) containing rogue species that probably originated from trans-clade crosses and are more likely to hybridize greatly. In contrast, they recovered low hybridization rates in four broad-leaved (Schedonorus-Lolium, Subbulbosae, Drymanthele-Pseudoscariosae-Lojaconoa, Leucopoa p. p.) and six fine-leaved lineages (Festuca, Psilurus-Vulpia, American I, Exaratae-Loretia, American-Vulpia-Pampas, Eskia), with species derived from single ancestors that hybridize only with close congeners. Inferences from gene duplications and allopolyploidizations, along with inheritance probabilities from the phylogenetic network, point to the BL Loliinae lineages as ancient hybrids or paleo-allopolyploids, while the FL lineages, especially those of the core FL clade, correspond to more recent meso- or neo-allopolyploids. [allopolyploidization, lineage-specific hybridization rates, nuclear and cyto-nuclear discordances, Festuca and related genera, phylogenomics, nuclear single-copy genes, plastome].