Light environment drives evolution of color vision genes in butterflies and moths

Abstract

Opsins, combined with a chromophore, are the primary light-sensing molecules in animals and are crucial for color vision. Throughout animal evolution, duplications and losses of opsin proteins are common, but it is unclear what is driving these gains and losses. Light availability is implicated, and dim environments are often associated with low opsin diversity and loss. Correlations between high opsin diversity and bright environments, however, are tenuous. To test if increased light availability is associated with opsin diversification, we examined diel niche and identified opsins using transcriptomes and genomes of 175 butterflies and moths (Lepidoptera). We found 14 independent opsin duplications associated with bright environments. Estimating their rates of evolution revealed that opsins from diurnal taxa evolve faster-at least 13 amino acids were identified with higher dN/dS rates, with a subset close enough to the chromophore to tune the opsin. These results demonstrate that high light availability increases opsin diversity and evolution rate in Lepidoptera.

Fig. 1. Opsin annotation and diel activity are mapped onto a Lepidoptera species tree. Duplications are associated with taxa active in bright light.

Taxa are color coded by diel-niche and RH4/UV, RH5/Blue, RH6/LW, and RH7 opsin recovery is marked. Red dots at nodes indicate duplication events identified by the tree reconciliation analysis for that lineage and darkened colors indicate a duplication in a particular opsin. Duplications (red dots) are more commonly associated with diurnal (yellow) and and partially diurnal or”both” taxa (green) which are more active in bright light environments. The tree is a pruned cladogram of tree of the most recent Lepidoptera phylogeny. 98/104 taxa are included in the figure, taxa for which the identification were only till genus level are excluded, but all the species with duplications are shown. Asterisk (*) indicates that these superfamilies are not monophyletic. Family names and some superfamily names are abbreviated, see Supplementary Data 1 for expanded names and complete classification, raw data data used to create these figures can be found at ref. .

Fig. 2. Opsin selection (dN/dS) rates between nocturnal and diurnal Lepidoptera species. Different models show that rates for visual opsins are higher in diurnal taxa.

A Blue, LW (n = 17 taxa) and UV (n = 14 taxa) opsin dataset were run using opsin gene trees. B Expanded dataset (n = 24 taxa) for UV, Blue, and LW opsins run using gene trees. RH7 was analyzed using more species (n = 33 taxa). C Expanded dataset run using a robust species trees (UV, LW n = 24 taxa, Blue = 23 taxa). *p-value < 0.05, **p-value < 0.01. The significance values indicate how well different models fit the data using likelihood ratio test (LRT). Model parameters used to run PAML and estimate the rates represented in the figures can be found at ref. . Opsin selection rates modeled for various datasets with different sample sizes and different model parameters all show that RH4/UV, RH5/Blue, and RH6/LW visual opsins have higher dN/dS rates in diurnal species than nocturnal species. RH7, a non-visual opsin, shows no significant difference. Smaller sample sizes (A) often fail to detect differences, which were more apparent when using larger data sets (B). Using a robust species tree rather than the gene tree shows more consistent differences across the visual opsins (C).

Fig. 3. Opsin structural prediction with retinal binding sites, sites with higher dN/dS (positively selection) and putative tuning sites, the subset that interacts with both.

A UV-opsin PHOBIUS transmembrane prediction. Opsins sites with higher dN/dS rates and amino acids close to (<=4 Å) retinal, which form the retinal binding region (RBR) have been marked. B UV-opsin model made using squid rhodopsin as the template. Retinal is marked, secondary structure prediction of alpha helices are shown in different colors and the expanded view shows the positively selected sites that can interact (<=2 Å) with the retinal binding sites. Protein models used to get these metrics can be found at ref. . Sites under positive selection are highlighted in red, sites in the retinal binding region in yellow and positively selected sites that are capable of interacting with the retinal binding sites in pink, identifying likely candidates for opsin tuning (A). Other positively selected are located within the transmembrane region and could be indirectly involved in opsin tuning (B).