The evolution of color vision in nocturnal mammals

Abstract

Nonfunctional visual genes are usually associated with species that inhabit poor light environments (aquatic/subterranean/nocturnal), and these genes are believed to have lost function through relaxed selection acting on the visual system. Indeed, the visual system is so adaptive that the reconstruction of intact ancestral opsin genes has been used to reject nocturnality in ancestral primates. To test these assertions, we examined the functionality of the short and medium- to long-wavelength opsin genes in a group of mammals that are supremely adapted to a nocturnal niche: the bats. We sequenced the visual cone opsin genes in 33 species of bat with diverse sensory ecologies and reconstructed their evolutionary history spanning 65 million years. We found that, whereas the long-wave opsin gene was conserved in all species, the short-wave opsin gene has undergone dramatic divergence among lineages. The occurrence of gene defects in the short-wave opsin gene leading to loss of function was found to directly coincide with the origin of high-duty-cycle echolocation and changes in roosting ecology in some lineages. Our findings indicate that both opsin genes have been under purifying selection in the majority bats despite a long history of nocturnality. However, when spectacular losses do occur, these result from an evolutionary sensory modality tradeoff, most likely driven by subtle shifts in ecological specialization rather than a nocturnal lifestyle. Our results suggest that UV color vision plays a considerably more important role in nocturnal mammalian sensory ecology than previously appreciated and highlight the caveat of inferring light environments from visual opsins and vice versa.

Fig. 1.

SWS1 species tree showing substitution rates, indels and stop codons based on the ORF. The tree topology and divergence dates follow consensus published phylogenies (SI Text). Numbers of insertions and deletions in the ORF are shown on the branches by downward (▼) and upward oriented (▲) triangles, respectively. Ancestral stop codons were inferred by probabilistic reconstruction and are denoted by squares (■), and critical sites implicated in spectral tuning were obtained using the same method for all functional ancestral nodes predating stops. Stops found in the tip sequences are shown in Fig. 3. Key changes in the critical spectral tuning amino acid sites are shown by sequence logos in which the height of amino acids at a given position are proportional to their posterior probability. Species with sequences containing stop codons are shown in red font, whereas the loss of functionality inferred from indels or stops is depicted by red branches. Branch length represents millions of years and numbers at the nodes represent divergence time in millions of years (SI Text). Numbers along the terminal branches are the d_N/d_S ratios after removing indels and stops, calculated by PAML (40) (see also SI Text).

Fig. 2.

The TBR maximum likelihood tree (-ln likelihood = 2,103.5) for the M/LWS dataset under the GTR+Γ+Ι model of sequence evolution. Numbers above the branches are the ML bootstrap values/Bayesian posterior probabilities as percentages, 100* = clades that received 100% ML bootstrap support and had posterior probabilities of 1; numbers below the branches are the d_N/d_S values estimated by PAML (40) (see also SI Text).

Fig. 3.

Schematic to show SWS1 bat sequences with insertions (▼), deletions (▲) and stops (■) in the ORF. Sizes of indels are given in base pairs.