Diagnosing Systematic Errors and Incomplete Lineage Sorting in the Deep Phylogeny of Collembola

Resolving deep phylogeny is complicated by systematic errors and biological sources of gene tree discordance, a challenge that is clearly presented by Collembola, one of the earliest diverging hexapod lineages. To address the long-standing conflict among the four collembolan orders, we expanded taxon sampling (113 species) and marker representation (4,070 single-copy orthologs), conducting a systematic evaluation of potential error sources. Using both concatenation and multispecies coalescent approaches under site-homogeneous (LG) and site-heterogeneous (CAT-PMSF) models, we recovered two primary competing topologies: a Neelipleona-first and an Entomobryomorpha-first hypothesis. Pronounced branch-length heterogeneity across orders and two exceptionally short internal branches suggested susceptibility to systematic error. Analyses of empirical data, combined with extensive simulations, showed that the Neelipleona-first topology arises under conditions known to induce long-branch attraction, including branch-length imbalance, compositional heterogeneity, and model misspecification, with fast-evolving and compositionally constrained sites further amplifying these artifacts. Coalescent simulations demonstrated that incomplete lineage sorting and gene tree estimation error jointly account for much of the deep gene tree-species tree discordance. In contrast, analyses using site-heterogeneous models and multispecies coalescent approaches, both intended to reduce systematic errors, consistently supported the Entomobryomorpha-first topology, recovering Entomobryomorpha + (Symphypleona + (Poduromorpha + Neelipleona)). Our findings clarify the mechanistic origins of phylogenomic conflict in Collembola and highlight the need to jointly consider systematic error and ILS when resolving ancient radiations. We propose the name 'Brachyantennamorpha' for the clade Poduromorpha + Neelipleona.