Amphibians are ideal for studying visual system evolution because their biphasic (aquatic and terrestrial) life history and ecological diversity expose them to a broad range of visual conditions. Here we evaluate signatures of selection on visual opsin genes across Neotropical anurans and focus on three diurnal clades that are well-known for the concurrence of conspicuous colors and chemical defense (i.e., aposematism): poison frogs (Dendrobatidae), Harlequin toads (Bufonidae: Atelopus), and pumpkin toadlets (Brachycephalidae: Brachycephalus). We found evidence of positive selection on 44 amino acid sites in LWS, SWS1, SWS2, and RH1 opsin genes, of which one in LWS and two in RH1 have been previously identified as spectral tuning sites in other vertebrates. Given that anurans have mostly nocturnal habits, the patterns of selection revealed new sites that might be important in spectral tuning for frogs, potentially for adaptation to diurnal habits and for color-based intraspecific communication. Furthermore, we provide evidence that SWS2, normally expressed in rod cells in frogs and some salamanders, has likely been lost in the ancestor of Dendrobatidae, suggesting that under low-light levels, dendrobatids have inferior wavelength discrimination compared to other frogs. This loss might follow the origin of diurnal activity in dendrobatids and could have implications for their chemical ecology, biodiversity, and behavior. Our analyses show that assessments of opsin diversification in understudied groups could expand our understanding of the role of sensory system evolution in ecological adaptation.

Please see the accompanying paper (Wan et al. 2023) for descriptions of the methods and analyses used to produce these data. Briefly, we obtained sequences of opsin genes from amphibians through RNAseq, public databases, and target-enrichment sequencing. We verified the gene identity of each sequence by estimating gene trees (Data S3). These data suggested that the SWS2 gene was missing from dendrobatid poison frogs, so we wrote a set of scripts to verify that blast fails to identify an SWS2 ortholog in Oophaga pumilio (Data S4). Then, we demonstrate that a short region upstream of LWS is present in Oophaga pumilio, even though the SWS2 gene appears to be absent (Data S2). Finally, we used CODEML and four types of site-based selection analyses from HyPhy version 2.5.14, on our frog opsin sequences: FEL, which estimates the rate of synonymous (dS) and non-synonymous (dN) substitutions per site with maximum likelihood and compares them using likelihood ratio tests; MEME, a mixed-effects maximum likelihood approach, to test for episodic or diversifying selection; contrast-FEL, tests for differences in selective pressures at individual sites among clades and groups of branches; and FUBAR, tests for nonsynonymous (dN) and synonymous (dS) substitution rates on a per-site basis using a Bayesian approach (Data S1).