Evolution of colour JAMES K. BOWMAKER vision in vertebrates Abstract It is now possible, more than 50 years later, to suggest that in one respect Walls was wrong, The expression of five major families of visual but in another, partly correct. Our underlying pigments occurred early in vertebrae mammalian dichromacy owes everything to our evolution, probably about 350-400 million teleost (more strictly, fish) and reptilian years ago, before the separation of the major ancestors (as ancient as 300-400 million years vertebrate classes. Phylogenetic analysis of ago, MYA), but our recent acquisition of opsin gene sequences suggests that the trichromacy (about 35 MYA) is a feature not ancestral pigments were cone pigments, with only of anthropOids but of Old World monkeys rod pigments evolving last. Modern teleosts, in general and even some New World monkeys. reptiles and birds have genera that possess The evidence to support these assumptions is rods and four spectral classes of cone each derived from three comparative sources: (i) representing one of the five visual pigment present-day 'representatives' of ancient families. The complement of four spectrally ancestral vertebrates, (ii) modern distinct cone classes endows these species representatives of the major vertebrate classes: with the potential for tetra chromatic colour teleosts, reptiles, birds and mammals, and (iii) vision. In contrast, probably because of their the molecular genetics of visual pigment protein nocturnal ancestry, mammals have rod­ genes. dominated retinas with colour vision reduced In this brief review it is not possible to give a to a basic dichromatic system subserved by complete survey of all the relevant published only two spectral classes of cone. It is only data, but only to select certain aspects that within primates, about 35 millions years ago, underlie and demonstrate the basic that mammals 're-evolved' a higher level of assumptions surrounding the evolution of colour vision: trichromacy. This was achieved visual pigments and colour vision. The by a gene duplication within the longer-wave consequence of such an eclectic selection is that cone class to produce two spectrally distinct much detail must be passed over and that members of the same visual pigment family generalisations will inevitably introduce which, in conjunction with a short-wavelength inaccuracies. pigment, provide the three spectral classes of cone necessary to subserve trichromacy. Early vertebrates Key words Colour vision, Cone, Evolution, Three questions are often asked with respect to Retina, Rhodopsin the evolution of the vertebrate retinal photoreceptor system. First, which are the ancestral (more primitive?) photoreceptors: The evolution of vertebrate colour vision, or rods or cones? Secondly, what was the ancestral more fundamentally, the evolution of visual visual pigment? And thirdly, when did colour pigments and photoreceptors, has long held the vision evolve? There are no simple answers to interest of zoologists and visual scientists. With these questions, though an answer to the first the great expansion of molecular genetics and might suggest an answer to the other two. the ability to identify and sequence opsin genes, Rods are more conserved throughout the there has been a reawakening of this interest. vertebrates, in terms of both structure and One of the great comparative visual visual pigments, whereas cones are highly scientists, Gordon Walls, wrote in 1942, in his J.K. Bowmaker � diverse in structure and can be divided into Department of Visual tome The Vertebrate Eye and its Adaptive many spectrally distinct classes (possibly as Science Radiation: many as five) within a given species. Rods Institute of Ophthalmology 'It seems necessary to believe that human could then be considered the simpler, and University College London colour vision owes nothing whatever to the therefore ancestral, photoreceptor. However, in Bath Street London EC 1V 9EL product of the teleost and the reptile. But other respects rods could be considered to be UK "human" colour vision is already present far more specialised than cones: they have a more Tel: +44 below man in the anthropoid stock. It is not complex morphology with isolated discs and (0)171 6086832 Fax: +44 (0)171 6086850 necessary to suppose that human colour vision are more sensitive, being capable of signalling e-mail: has evolved wholly within the genus Homo.' the detection of a single photon. A possible way [email protected] Eye (1998) 12, 541-547 © 1998 Royal College of Ophthalmologists 541 to approach the question of the ancestral vertebrate extracellular loop of opsin. The chloride binding site is photoreceptor is to look at modem species that are absent from some mammalian long-wave cone pigments thought to be direct descendants of the earliest of that have closer to nm.14 Amax 500 vertebrates. A conclusion that can be drawn from these studies of The present-day cyclostomes, i.e. lampreys and lamprey photoreceptors is that the ancestral vertebrate considered to have arisen directly from visual system was based on relatively unspecialised hagfish, are ancient ostracoderms, extinct jawless, fish-like primitive photoreceptors (perhaps more cone-like) and thatat least vertebrates of the Silurian and Devonian periods. There two spectral classes of photoreceptor were present at a has been much debate as to whether the photoreceptors very early stage. Thepresence of the two classes of visual of lampreys are rods or cones Crescitelli1), though it pigment suggests a gene duplication early in vertebrate (see is clear that they have two classes of photoreceptor, one evolution (about to give a middle-wave­ 400 MYA) with long outer segments and the other with short outer sensitive (MWS), but non-chloride-sensitive visual segments. Both types of outer segment are pigment and a longer-wave-sensitive (LWS), chloride­ morphologically cone-like in having many discs that are sensitive pigment, the chloride binding being necessary infoldings of the outer plasma membrane, but recent to spectrally tune a visual pigment to longer studies2--4 clearly demonstrate that the two classes are wavelengths. There is no reason why an ancestral functionally different. vertebrate should not have been carrying more than a Microspectrophotometrr has revealed that the single opsin gene and colour vision (requiring two shorter outer segments have a wavelength of maximum spectral classes of photoreceptor), as distinct from absorbance close to whereas the longer photopic vision, may have been present in the earliest P�max) 517 nm, contain a pigment with at about nm. Further, verterates. Amax 555 the spectral sensitivity of the dark-adapted eye is A second group of modem representatives of maximal at about 510-520 nm, but with short-wave vertebrate ancestral formsis thesturgeons (Chondrostei), background illumination the maximum sensitivity is which are considered to be degenerate ray-finned fish displaced to about 555 nm. These data strongly imply arising from ancestral Devonian forms (about 350 MYA). that the lamprey has a duplex retina with both cone-like These fish have a clearly duplex retina with rods and at and rod-like photoreceptors and exhibits a Purkinje shift least three spectrally distinct classes of cone (with oil between scotopic and photopic vision similar to thatseen droplets). The cones (containing porphyropsins in have at about and in most vertebrates. However, the rod-like Acipenser transmontanus) Amax 610, 540 465 nm, whereas the rods have at about nm. photoreceptors also have cone-like features: the pigment Amax 540 reacts with hydroxylamine in a similar manner to cone Sturgeons have colour-opponent horizontal cells (with possible up to six classes in somewhat pigments,S the cells apparently do not saturate at high Acipenser baieri), light intensities and are also involved in photopic vision, similar to teleosts.15-18 These data suggest that at a very perhaps even subserving a dichromatic colour vision early stage in vertebrate evolution there were at least system.2 Immunocytochemical reactions also support the four opsin families (a rod and three cones) with the idea that the shorter of the two outer segment types is neural mechanisms to give at least trichromatic colour rod-like and the longer cone-like.3,4 vision. Rod opsin-like genes have been isolated and sequenced for the river lamprey, and Lampetra japonica,6 from the marine lamprey, The Petromyzon marin us? Teleosts deduced amino acid sequences show about 92% similarity and have about 80% identity with rod opsins Teleost fish (having evolved in the last show a 150 MY) from higher vertebrates, but only about a 45% identity great diversity of visual pigments and colour vision with cone opsins. These rod opsins are presumably depending on their environment and life styles (for a expressed within 'short', more rod-like the review, see Bowmaker19), but in diurnal species living in photoreceptor. The gene sequence for the longer-wave relatively shallow water where a broad spectrum of light visual pigment has not so far been published, but in P. is available there are, in addition to rods, at least four the visual pigment is sensitive to the marinus spectral classes of cone including a violet- or ultraviolet­ concentration of CI- ions.8 Under normal saline sensitive (V /UVS) class. This arrangement appears conditions the pigment has close to nm (a Amax 600 common amongst cyprinids, such as the goldfish, which porphyropsin, visual pigments based on vitamin A ), but has cone