Compositional phylogenomic modelling resolves the ‘Zoraptera problem’: Zoraptera are sister to all other polyneopteran insects

Abstract

The evolution of wings propelled insects to their present mega-diversity. However, interordinal relationships of early-diverging winged insects and the timescale of their evolution are difficult to resolve, in part due to uncertainties in the placement of the enigmatic and species-poor order Zoraptera. The 'Zoraptera problem' has remained a contentious issue in insect evolution since its discovery more than a century ago. This is a key issue because different placements of Zoraptera imply dramatically different scenarios of diversification and character evolution among polyneopteran. Here, we investigate the systematic placement of Zoraptera using the largest protein-coding gene dataset available to date, deploying methods to mitigate common sources of error in phylogenomic inference, and testing historically proposed hypotheses of zorapteran evolution. We recover Zoraptera as the earliest-diverging polyneopteran order, while earwigs (Dermaptera) and stoneflies (Plecoptera) form a monophyletic clade (Dermoplectopterida) sister to the remainder of Polyneoptera. The morphology and palaeobiology of stem-zorapterans are informed by Mesozoic fossils. The gut content and mouthparts of a male specimen of Zorotypus nascimbenei from Kachin amber (Cretaceous) reveal a fungivorous diet of Mesozoic zorapterans, akin to extant species. Based on a set of 42 justified fossil and stratigraphic calibrations, we recover a Devonian origin of winged insects and Polyneoptera, suggesting that these groups coincided with the rise of arborescence during the diversification of early terrestrial plants, fungi, and animals. Our results provide a robust framework for understanding the pattern and timescale of early winged insect diversification.

Phylogenomic Analyses of the Tenthredinoidea Support the Familial Rank of Athaliidae (Insecta, Tenthredinoidea)

“…Our current knowledge of the evolutionary relationships among insect orders is extensive and supported by numerous large-scale phylogenomic analyses [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 ]. However, inconsistencies persist regarding the phylogenetic relationships between a few groups (e.g., Psocoptera, Antliophora + Siphonaptera) and have challenged the perfect picture of insect evolution [ 1 , 10 , 11 , 12 ]. It is widely accepted that further advances in insect phylogenetics cannot rely only on deeper and broader genome and taxon sequencing.…”

Systematic bias and the phylogeny of Coleoptera—A response to Cai et al. (2022) following the responses to Cai et al. (2020)

“…It is not possible to evaluate this unstated assumption of CEA22 because the authors did not provide their log files and did not clearly report the number of estimated categories from the model, despite their continued bold and vocal support for CAT (CEA20, Tihelka et al, 2020a, 2020b, 2020c, 2021). Category occupancy in CAT is known to be the variable that captures most of the data complexity under this model, and that this is driven by rare sites, which themselves may be insufficient for parameterizing the model (Li et al, 2021b).…”

“…Because we cannot evaluate model fit using the files provided by CEA22, we here consider the other CAT‐GTR centred studies published by some of the authors. In those prior studies (Tihelka et al, 2020a, 2020b, 2020c, 2021) 15 mean category count and distribution of categories are provided for some or all CAT chains, though only in the Mendeley data. In all cases, the number of categories is one or more orders of magnitude greater than the finite C10–C60 models.…”

“…Because of these issues, coupled with the inadequacy of their model testing, the results reported in the ‘phylogenomic reconstruction’ section should not be accepted, nor can the other primary results of this study more broadly, as these are dependent on the GTR‐CAT+G4 analysis. There also appears to be a more general pattern in which some of the authors of CEA22 have claimed superiority of the CAT model without adequate reporting and model testing, despite claiming to have conducted analyses under best practises and on the ‘most comprehensive dataset for [taxon] to date’ (Tihelka et al, 2020a, 2020b, 2020c, 2021). Close evaluation of the basic and major claims of CEA22 above shows that the methodological justification for their claims is unfounded and that, separate from the inadequacy of their model comparison exercises, the focal analysis upon which all of the subsequent claims were built had not achieved statistical convergence, that is, it had failed.…”

“…This is false 12 . The bpcomp ‘maxdiff’ statistic from the CAT‐GTR + G4 PhyloBayes results reported in the supplementary material is 0.39332 13 (see also misreporting of maxdiff in Tihelka et al, 2021 [Zoraptera]). This value indicates that the MCMC sampling was not at convergence.…”