A genomic evaluation of taxonomic trends through time in coast horned lizards (genus Phrynosoma)

Abstract

Determining the boundaries between species and deciding when to describe new species are challenging practices that are particularly difficult in groups with high levels of geographic variation. The coast horned lizards (Phrynosoma blainvillii, Phrynosoma cerroense and P. coronatum) have an extensive geographic distribution spanning many distinctive ecological regions ranging from northern California to the Cape Region of Baja California, Mexico, and populations differ substantially with respect to external morphology across much of this range. The number of taxa recognized in the group has been reevaluated by herpetologists over 20 times during the last 180 years, and typically without the aid of explicit species delimitation methods, resulting in a turbulent taxonomy containing anywhere from one to seven taxa. In this study, we evaluate taxonomic trends through time by ranking 15 of these species delimitation models (SDMs) using coalescent analyses of nuclear loci and SNPs in a Bayesian model comparison framework. Species delimitation models containing more species were generally favoured by Bayesian model selection; however, several three-species models outperformed some four- and five-species SDMs, and the top-ranked model, which contained five species, outperformed all SDMs containing six species. Model performance peaked in the 1950s based on marginal likelihoods estimated from nuclear loci and SNPs. Not surprisingly, SDMs based on genetic data outperformed morphological taxonomies when using genetic data alone to evaluate models. The de novo estimation of population structure favours a three-population model that matches the currently recognized integrative taxonomy containing three species. We discuss why Bayesian model selection might favour models containing more species, and why recognizing more than three species might be warranted.

Genomic data reveal deep genetic structure but no support for current taxonomic designation in a grasshopper species complex

“…As in many other species complexes (e.g., Leaché et al, ) the taxonomy of O. antigai and O. navasi has changed through time in conjunction with the discussion about the validity of the different criteria considered to establish the boundaries between the two putative taxa (Clemente et al, ; Ragge & Reynolds, ). After the two species were described by Bolívar in the transition between the XIX and XX centuries (Bolívar, , ) and Clemente, García, and Presa () confirmed their taxonomic distinction, Reynolds () and Ragge and Reynolds () attributed their subtle morphological diversity to the effect of population geographic isolation, denied differences in acoustic communication (i.e., courtship songs), and recommended the synonymy of O. navasi to O. antigai .…”

“…The Bayes factor (BF) test statistic (2 × ln BF) was calculated, where BF is the difference in MLE between two competing models (base scenario‐alternative scenario). These analyses were performed using a matrix with no missing data (393 SNPs; see Leaché, McElroy, & Trinh, for a similar approach) and including two individuals from each sampled population plus two individuals from the outgroup O. femoralis , to be able to test for the single‐species model.…”

“…The Bayes factor (BF) test statistic (2 × ln BF) was calculated, where BF is the difference in MLE between two competing models (base scenario-alternative scenario). These analyses were performed using a matrix with no missing data (393 SNPs; see Leaché, McElroy, & Trinh, 2018 for a similar approach) and including two individuals from each sampled population plus two individuals from the outgroup O. femoralis, to be able to test for the singlespecies model. bpp version 3.3 can perform species delimitation analyses using a fixed input guide tree specified by the user (option A10, guided analysis) or estimating a species tree (option A11, unguided analysis; Yang, 2015;Yang & Rannala, 2014).…”

“…As in many other species complexes (e.g., Leaché et al, 2018) the taxonomy of O. antigai and O. navasi has changed through time in conjunction with the discussion about the validity of the different criteria considered to establish the boundaries between the two putative taxa (Clemente et al, 1999;Ragge & Reynolds, 1998).…”

“…It is also remarkable that many nearby populations (BOR-MON, NER-SAR, BOI-LLE and SET-BAT) assigned to the same genetic groups by PCA and Bayesian clustering analyses (structure, faststructure and construct) were consistently delimited as distinct species (Figure 4). Accordingly, the species delimitation model most supported by bfd* analyses was the one considering the highest number of species, one per genetic cluster inferred by structure (see also figure 2b in Leaché et al, 2018). Recent empirical and theoretical studies have shown the limitations of Bayesian species delimitation approaches based on the multi-species coalescent (MSC) model (Rannala & Yang, 2013;Yang & Rannala, 2010), pointing out that such methods are not able to statistically distinguish genetic structure due to population isolation from true species boundaries and tend to over-split genetically differentiated populations rather than capturing species divergence (Huang, 2018;Leaché et al, 2018;Leaché, Zhu, Rannala, & Yang, 2019;Sukumaran & Knowles, 2017).…”