Comprehensive Species Sampling and Sophisticated Algorithmic Approaches Refute the Monophyly of Arachnida

Abstract

Deciphering the evolutionary relationships of Chelicerata (arachnids, horseshoe crabs, and allied taxa) has proven notoriously difficult, due to their ancient rapid radiation and the incidence of elevated evolutionary rates in several lineages. Although conflicting hypotheses prevail in morphological and molecular data sets alike, the monophyly of Arachnida is nearly universally accepted, despite historical lack of support in molecular data sets. Some phylotranscriptomic analyses have recovered arachnid monophyly, but these did not sample all living orders, whereas analyses including all orders have failed to recover Arachnida. To understand this conflict, we assembled a data set of 506 high-quality genomes and transcriptomes, sampling all living orders of Chelicerata with high occupancy and rigorous approaches to orthology inference. Our analyses consistently recovered the nested placement of horseshoe crabs within a paraphyletic Arachnida. This result was insensitive to variation in evolutionary rates of genes, complexity of the substitution models, and alternative algorithmic approaches to species tree inference. Investigation of sources of systematic bias showed that genes and sites that recover arachnid monophyly are enriched in noise and exhibit low information content. To test the impact of morphological data, we generated a 514-taxon morphological data matrix of extant and fossil Chelicerata, analyzed in tandem with the molecular matrix. Combined analyses recovered the clade Merostomata (the marine orders Xiphosura, Eurypterida, and Chasmataspidida), but merostomates appeared nested within Arachnida. Our results suggest that morphological convergence resulting from adaptations to life in terrestrial habitats has driven the historical perception of arachnid monophyly, paralleling the history of numerous other invertebrate terrestrial groups.

Keywords: Chelicerata; orthologs; phylogenomics; supermatrix; total evidence.

Fig. 1.

Higher-level phylogeny of Chelicerata showing well-resolved groups (boldface text adjacent to nodes), based on Ontano et al. (2021). Circles indicate whole-genome duplication events (WGD) subtending specific taxa. Branch lengths are not to scale.

Fig. 2.

Comprehensive phylogenomic sampling of all extant chelicerate orders recovers horseshoe crabs as derived arachnids. (a) Phylogenomic relationships of 506 chelicerate data sets based on maximum likelihood analysis of slowly evolving loci (Matrix 3) and site heterogeneous evolutionary models. Colors correspond to orders; note that Acariformes and Parasitiformes are each treated as orders in this study. Dots on nodes indicate high (> 95% bootstrap; blue color), medium (90–95% bootstrap; orange color), or low (< 90% bootstrap; red color) support. (b) Summary of relationships inferred under site heterogeneous models by three matrices. Numbers on nodes correspond to bootstrap resampling frequencies below 100%; all unlabeled nodes are maximally supported. (c) Gene (gCF) and site (sCF) concordance factors exhibit higher support for the derived placement of Xiphosura under all three 506-taxon matrices. Asterisks indicate tree topologies wherein Xiphosura was recovered as sister group to Ricinulei + Solifugae. (d) Tests of monophyly consistently rejected the monophyly of Arachnida and Acari over the unconstrained topology for Matrices 1–3. Nonsignificant result for Matrix 3 results from the unconstrained recovery of Panscorpiones in this analysis.

Fig. 3.

Site heterogeneous model-based approaches using CAT+GTR+Γ and SR4 recoding refute the monophyly of Arachnida. (a) Summary tree of eight chains from PhyloBayes-mpi analysis of Matrix 4. Numbers on nodes correspond to posterior probabilities below 1.00; all unlabeled nodes are maximally supported. Lower right: Distribution of support across a posteriori trees for arachnid monophyly (yellow) versus nested placement of Xiphosura (blue). (b) Summary trees from PhyloBayes-mpi analysis separated by starting tree topology. Top: Chains started on maximum likelihood tree topology for Matrix 4 (Xiphosura nested). Middle: Chains started on maximum likelihood tree topology for Matrix 4 with a constraint for arachnid monophyly. Bottom: Chains started on random tree topologies. Nodal support values and pie charts for each summary tree reflect the conventions for (a). (c) Maximum likelihood tree topology based on SR4 recoding and multi-profile tiered site heterogeneous models (RL2 approach). Numbers on nodes correspond to bootstrap resampling frequencies below 100%; all unlabeled nodes are maximally supported.

Fig. 4.

Dissection of phylogenetic signal shows that a minority of genes supports arachnid monophyly. (a) ΔGLS distributions mapping phylogenetic support for competing hypotheses reveal that a minority of genes (39–41%) supports arachnid monophyly, regardless of orthology criterion (Matrices 1 and 2) and filtering of fast-evolving genes (Matrix 3). These proportions are similar to the proportions of genes supporting falsified groupings (supplementary fig. S3, Supplementary Material online). (b) Genes supporting the derived placement of Xiphosura exhibit comparable or better metrics of systematic bias (e.g., saturation, evolutionary rate, missing data) than genes supporting Arachnida. (c) ΔSLS distributions reveal that the majority of sites (68%) support a derived placement of Xiphosura. Whereas the two categories of sites are similar with respect to missing data, sites supporting arachnid monophyly exhibit high levels of Shannon entropy (exceeding entropy values for falsified groupings; supplementary fig. S4, Supplementary Material online).

Fig. 5.

Inclusion of morphology does not rescue arachnid monophyly in total evidence analyses. (a) Strict consensus of 1,000 equally parsimonious trees inferred for a morphological matrix of 259 morphological characters (482 extant and 32 fossil taxa). Merostomata comprises extinct groups Eurypterida (sea scorpions), Chasmataspidida, and Synziphosurina, as well as Xiphosurida, including the extant Limulidae. Numbers on nodes indicate bootstrap resampling frequencies. (b) Summary tree from Bayesian inference analysis of the morphological matrix. Numbers on nodes indicate posterior probabilities below 1.00; unlabeled nodes are maximally supported. (c) Maximum likelihood total evidence topology based on 152 slowly evolving genes and morphological characters. Numbers on nodes represent bootstrap resampling frequencies; unlabeled nodes are maximally supported. Note that the timing of WGD events in Xiphosura cannot be pinpointed on the branches subtending this group.