Cyto-nuclear discordance and ancient reticulation
The conflict between plastid trees and inferences from the nuclear genome goes back to the earliest studies on plastid phylogenetics (Palmer et al., 1982 ). Such conflict subsequently appeared in many early plastid restriction site analyses, in which one or more individuals of one species were nested within plastid-based clades of another species (Rieseberg et al., 1991 ). These results demonstrate the high frequency of cytoplasmic gene flow in angiosperms, as well as its extent within certain lineages (Rieseberg & Soltis 1991 ). Accordingly, the deep discordance between plastid and nuclear trees might be interpreted as evidence of ancient hybridization, given the propensity for interspecific hybridization among extant species (Gitzendanner et al., 2018 ), including more than hundred records of interspecific hybridization among rosid taxa alone (Rieseberg & Soltis 1991; Rieseberg et al., 1996 ).
Although conflict between plastid and nuclear trees is typically attributed to hybridization, other processes such as incomplete lineage sorting (ILS) may also cause phylogenetic incongruence between nuclear and plastid DNA (Soltis & Kuzoff, 1995; Wendel & Doyle, 1998 ). While hybridization and ILS were historically difficult to distinguish due to their similar phylogenetic signatures (Wendel & Doyle, 1998 ), the multispecies coalescence (Mirarab et al., 2014; Mirarab & Warnow, 2015 ) offers a clear path for testing the relative roles of ancient hybridization and ILS in explaining gene tree congruence (Folk et al. 2018 ). Sun et al. (2015)proposed that the incongruence in the positions of the COM orders between studies based on plastid, mitochondrial, and nuclear genes was possibly the result of ancient hybridization and introgression events. This hypothesis requires further study with probabilistic methods and larger samples of the three genomic compartments of plants.
Here, we retrieved different relationships of superrosids based on 122 single-copy nuclear genes compared to plastid genes, especially for the placement of Vitales and Saxifragales with respect to the rosids, non-monophyly of the COM clade, and the re-circumscription of rosids and fabids seen here. As reviewed above and in Discussion S1, while methodological decisions and data properties influence recovered topologies, the primary cause for differing deep relationships among superrosids appears to rely on different studies of either cytoplasmic or nuclear markers. Although cytonuclear discordance is often attributed to hybridization, our simulation results suggest a role for both ILS and hybridization. The pattern of gene tree discord between the three major superrosid lineages (Vitales, Saxifragales, and rosids) was within ILS expectations and consistent with this branch of the tree being in the anomaly zone. However, the degree of gene tree heterogeneity related to backbone relationships within the core rosids was unexpected based on ILS alone, and is therefore likely due to ancient hybridization. These results contribute to a growing awareness that complex evolutionary processes should be considered, even for deep-level plant phylogenetics (Folk et al. 2018 ; Stull et al. 2022 ).