A Comprehensive and Dated Phylogenomic Analysis of Butterflies

Abstract

Butterflies (Papilionoidea), with over 18,000 described species [1], have captivated naturalists and scientists for centuries. They play a central role in the study of speciation, community ecology, biogeography, climate change, and plant-insect interactions and include many model organisms and pest species [2, 3]. However, a robust higher-level phylogenetic framework is lacking. To fill this gap, we inferred a dated phylogeny by analyzing the first phylogenomic dataset, including 352 loci (> 150,000 bp) from 207 species representing 98% of tribes, a 35-fold increase in gene sampling and 3-fold increase in taxon sampling over previous studies [4]. Most data were generated with a new anchored hybrid enrichment (AHE) [5] gene kit (BUTTERFLY1.0) that includes both new and frequently used (e.g., [6]) informative loci, enabling direct comparison and future dataset merging with previous studies. Butterflies originated around 119 million years ago (mya) in the late Cretaceous, but most extant lineages diverged after the Cretaceous-Paleogene (K-Pg) mass-extinction 65 mya. Our analyses support swallowtails (Papilionidae) as sister to all other butterflies, followed by skippers (Hesperiidae) + the nocturnal butterflies (Hedylidae) as sister to the remainder, indicating a secondary reversal from diurnality to nocturnality. The whites (Pieridae) were strongly supported as sister to brush-footed butterflies (Nymphalidae) and blues + metalmarks (Lycaenidae and Riodinidae). Ant association independently evolved once in Lycaenidae and twice in Riodinidae. This study overturns prior notions of the taxon's evolutionary history, as many long-recognized subfamilies and tribes are para- or polyphyletic. It also provides a much-needed backbone for a revised classification of butterflies and for future comparative studies including genome evolution and ecology.

Out of the Orient: Post‐Tethyan transoceanic and trans‐Arabian routes fostered the spread of Baorini skippers in the Afrotropics

“…Two hesperiid fossils, Pamphilites abdita Scudder and Protocoeliades kristenseni de Jong, were candidates to be included as fossil calibrations for this study (de Jong, ). However, preliminary testing in beast revealed that the age estimates obtained with these two fossils were largely underestimated compared with the ages from Espeland et al () or Chazot et al (). This discrepancy is partly due to the difficulty of placing the oldest described skipper fossil into an extant taxon: Protocoeliades kristenseni from Denmark, dated at c .…”

“…Gene trees were inferred using fasttree 2.1.5 (Price et al , ) and inspected for contamination and sequence quality. Trees were rooted with Macrosoma hyacinthina Warren (Hedylidae), as the Hedylidae are the sister group to the Hesperiidae (Espeland et al , ; Toussaint et al , ). The final matrix included 89 species, including 70 Baorini, and 16 concatenated loci, totalling up to 12 928 aligned nucleotides.…”

“…However, because relationships among Coeliadinae genera and their monophyly are uncertain, any placement of the fossil elsewhere than on the stem of Coeliadinae remains contentious (de Jong, , ). Therefore, because Coeliadinae is sister to the remainder of Hesperiidae (Espeland et al , ; Toussaint et al , ), the fossil would only calibrate a minimum age of 55 Ma for crown Hesperiidae, while Espeland et al () estimated this node to be c . 76 Ma and the split between Hesperiidae and Hedylidae to be c .…”

“…99 Ma. Because Chazot et al () has a more extensive taxon sampling, including Baorini genera, we chose this study for secondary calibrations over the more robust phylogenomic tree of Espeland et al (). We constrained the node corresponding to the split between Hesperiidae and Hedylidae with a uniform prior encompassing the 95% credibility interval (lower = 80.73, upper = 119.22) estimated for this node in Chazot et al ().…”

“…We use a hypothetico‐deductive framework to test whether the phylogenomic pattern, divergence time estimates and ancestral range estimation are consistent with one or more of the hypotheses. We do not test the Indian biotic ferry hypothesis, because divergence time evidence suggests that Baorini are much younger than the invoked geological events (Espeland et al , ).…”