SEARCH

SEARCH BY CITATION

Keywords:

  • cone photoreceptors;
  • rod photoreceptors;
  • color vision;
  • retina

Abstract

  1. Top of page
  2. Abstract
  3. MAMMALIAN PHOTORECEPTORS: BASIC PATTERN
  4. DEVIATIONS FROM BASIC PATTERN
  5. CONCLUSION
  6. Acknowledgements
  7. LITERATURE CITED

All mammalian retinae contain rod photoreceptors for low-light vision and cone photoreceptors for daylight and color vision. Most nonprimate mammals have dichromatic color vision based on two cone types with spectrally different visual pigments: a short-wavelength-sensitive (S-)cone and a long-wavelength-sensitive (L-)cone. Superimposed on this basic similarity, there are remarkable differences between species. This article reviews some striking examples. The density ratio of cones to rods ranges from 1:200 in the most nocturnal to 20:1 in a few diurnal species. In some species, the proportion of the spectral cone types and their distribution across the retina deviate from the pattern found in most mammals, including a complete absence of S-cones. Depending on species, the spectral sensitivity of the L-cone pigment may peak in the green, yellow, or orange, and that of the S-cone pigment in the blue, violet, or near-ultraviolet. While exclusive expression of one pigment per cone is the rule, some species feature coexpression of the L- and S-pigment in a significant proportion of their cones. It is widely assumed that all these variations represent adaptations to specific visual needs associated with particular habitats and lifestyles. However, in many cases we have not yet identified the adaptive value of a given photoreceptor arrangement. Comparative anatomy is a fruitful approach to explore the range of possible arrangements within the blueprint of the mammalian retina and to identify species with particularly interesting or puzzling patterns that deserve further scrutiny with physiological and behavioral assays. © 2005 Wiley-Liss, Inc.

Mammalian retinae conform to the general blueprint of the vertebrate retina. Their basic cell types can be easily identified across species. But as common in biology, species-specific differences are superimposed on the basic similarities. We consider them as evolutionary adaptations to particular visual needs and constraints. Comparative studies exploiting these experiments of nature are a suitable strategy to elucidate the construction principles and the malleability of the mammalian retina. Mammals have occupied nearly every habitat available on earth and adopted various lifestyles associated with different light conditions and visual challenges. In some species, vision is the dominant sense, in others it is less important. While all mammals have eyes, their different size and retinal circuitry indicate different image processing capabilities. One extreme is the foveate retina of diurnal primates with its high visual acuity and refined color vision. At the other end of the spectrum, there is the subcutaneous minute eye of the blind mole-rat Spalax, which has lost all image processing capability and is thought to serve only the entrainment of the circadian clock.

Photoreceptors have been a longstanding focus of interest in comparative studies. They are the input stage to the visual system, and their properties determine which signals become available to the postreceptoral retinal and cortical neurons for further processing. The histological study of photoreceptors has been greatly advanced by the availability of visual pigment-specific antibodies. They now allow us to assess the spectral cone photoreceptor types and hence the prospects for color vision even in species that would not easily be available to physiological or behavioral studies. Over the past years, we and others have analyzed the photoreceptor types and their distributions in a wide range of mammals with different habitats and lifestyles. This has provided insights into the adaptive flexibility of mammalian photoreceptor arrangements, but it has also uncovered some unexpected deviations over which we still puzzle. The present overview highlights the diversity of mammalian photoreceptor arrangements and indicates where further study is needed to understand the underlying selective pressures.

MAMMALIAN PHOTORECEPTORS: BASIC PATTERN

  1. Top of page
  2. Abstract
  3. MAMMALIAN PHOTORECEPTORS: BASIC PATTERN
  4. DEVIATIONS FROM BASIC PATTERN
  5. CONCLUSION
  6. Acknowledgements
  7. LITERATURE CITED

Rods and Cones

Vertebrate retinal photoreceptors are divided into two basic categories: rods and cones (Fig. 1). The rods are more light-sensitive and used for vision at low light levels (scotopic vision, night vision); the cones need higher light levels and are used for daylight (photopic) vision. At intermediate (mesopic) light levels, both rods and cones contribute to vision. The different sensitivity of rods and cones originates in their different morphological and biochemical properties, e.g., different amounts of visual pigment and different amplification factors of the phototransduction cascade. The sensitivity of rod vision is further increased by a high convergence of the rods onto their postsynaptic neurons, the rod bipolar cells, which improves the signal-to-noise ratio. Rod/cone ratios vary considerably across mammals, roughly correlating with the daily activity pattern (Fig. 2). Nocturnal species have between 0.5% and 3% cones among their photoreceptors, crepuscular and arrhythmic species have between 2% and 10% cones, and diurnal mammals show a larger range of cone proportions from 8% to 95% cones (for overviews, see Ahnelt and Kolb, 2000; Peichl et al., 2000). It is notable that most diurnal mammals still have rod-dominated retinae [e.g., pig about 10–20% cones (Hendrickson and Hicks, 2002), guinea pig about 8–17% cones (Peichl and González-Soriano, 1994), degu about 30% cones (Jacobs et al., 2003]. Rod-dominated retinae may be a shared basal trait of mammals retained from some nocturnal ancestors. Only a few diurnal species are known to have more cones than rods [e.g., ground squirrel about 85% (Kryger et al., 1998), tree shrew about 95% (Müller and Peichl, 1989)]. The rodless fovea of diurnal primates is a regional patch of all-cone retina.

thumbnail image

Figure 1. Sections through the retina of the ox (Bos taurus, left) and the crab-eating raccoon (Procyon cancrivorous, right); semithin sections of Epon-embedded tissue stained with toluidine blue. Bottom panels show the retinal layers in transverse sections with the photoreceptors on top (1, outer segments; 2, inner segments; 3, somata in the outer nuclear layer), followed by the outer plexiform layer (4, first synaptic layer), the inner nuclear layer (5, somata of the interneurons, i.e., bipolar cells, horizontal cells, and amacrine cells), the inner plexiform layer (6, second synaptic layer), and the ganglion cell layer (7, somata of the retinal output neurons). In this orientation, light would reach the retina from below. Top panels show horizontal sections at the level of the photoreceptor inner segments, revealing the image sampling array of the rods and cones. In the ox, the cones are readily identified by their fatter inner segments and darker staining; the more numerous rods are slender and pale. In the raccoon, the staining shows no obvious difference between the rods and cones. Scale bars = 20 μm (top); 50 μm (bottom).

Download figure to PowerPoint

thumbnail image

Figure 2. Cone densities and proportions in species with different activity patterns. Micrographs of retinal flat mounts immunolabeled for cone opsin; the focus is on the cone outer segments. The strictly nocturnal African giant rat (Cricetomys gambianus) has a very low cone density and proportion (A). The nocturnal to crepuscular wolf (Canis lupus) has a slightly higher cone density and proportion (B), and the diurnal to crepuscular mouflon (Ovis musimon) has a significantly higher cone density and proportion (C). The strictly diurnal tree shrew (Tupaia belangeri) is exceptional in having about 95% cones and only about 5% rods (D). The spaces between the cones are filled by the unlabeled rods. All micrographs from mid-peripheral retina and shown at same magnification. Scale bar = 50 μm.

Download figure to PowerPoint

Rod densities generally are higher in nocturnal than in diurnal species. Drastically diurnal species are the tree shrew, with only 500–3,500 rods/mm2 (Müller and Peichl, 1989), and the ground squirrel, with 1,000–13,000 rods/mm2 (Kryger et al., 1998); a nocturnal extremist is the African pouched rat, with 390,000–730,000 rods/mm2 (Peichl and Moutairou, 1998). Most mammals have intermediate rod densities in the range of 200,000–400,000/mm2.

Early comparative work has claimed the existence of pure rod retinae and pure cone retinae as extreme adaptations to a strongly noctural or diurnal lifestyle, respectively (see, for instance, Kolmer, 1936; Walls, 1942; Rochon-Duvigneaud, 1943). In contrast, contemporary methodology has found rods and cones in all species examined. In some species, the cones are morphologically very distinct from the rods; in others, they are not (Fig. 1). The latter are difficult to detect by classical histology, particularly when present in low numbers. Today, immunohistochemistry with rod-specific and cone-specific antibodies allows an unambiguous identification. The current understanding is that all mammals have a duplex retina containing rods and cones.

One reason for the retention of at least a sparse cone population in each species lies in the circuitry of the mammalian retina. Rod and cone pathways are strictly separated in the outer retina; rods connect to rod bipolar cells and cones to cone bipolar cells. In the inner retina, cone bipolar cells form synapses with ganglion cells (the output cells of the retina). Rod bipolar cells do not directly contact ganglion cells but feed into cone bipolar cells via the specialized AII amacrine cell in the inner retina. If the cone circuitry were not present, rod signals could not reach the ganglion cells (for recent review, see Wässle, 2004). Thus, a pure rod retina would require major rewiring of retinal circuits. As far as we know, this has not happened. A pure cone retina would be feasible, as evidenced by the foveal circuitry in primate retina. But to date there is no evidence for a pure cone retina in any species.

Photoreceptors and Visual Acuity

Visual sensitivity and spatial resolution (acuity) are conflicting fundamental capacities (Land and Nilsson, 2002). Sensitivity requires the summation of the signals of many photoreceptors in order to optimize the signal-to-noise ratio, thus precluding spatial resolution. Such convergence is a hallmark of the rod pathway. Good spatial resolution requires low convergence and stimuli of high brightness or color contrast and reasonably high light levels. This is a task for the cones, which have the low convergence onto cone bipolar cells and ganglion cells that preserves the fine-grain information to higher processing stages. However, the spatial resolution capacity of a retina is determined by its ganglion cells, because in most mammals even the smallest ganglion cells in the Area centralis, the retinal region of highest neuron density and hence visual acuity, sum the inputs of several cones. The only known exception is the primate fovea, where the 1:1 connection between the cones and the ganglion cells conserves the pixelation of the cone mosaic down to the ganglion cells (by the so-called midget system). In other species, a higher cone density also indicates better visual acuity, but the acuity cannot be calculated directly from that density unless the convergence ratio of cones to ganglion cells is known. Commonly, both cone densities and ganglion cell densities peak in the Area centralis and decrease toward the retinal periphery (see summary schemes, Fig. 5).

Mammalian Cone Opsins and Color Vision

The cones also provide for color vision. Color discrimination requires two or more types of photoreceptor with spectrally discrete visual pigments. Furthermore, the spectral signals must be kept segregated in the postreceptoral circuitry, such that they can be compared, e.g., by color-opponent retinal ganglion cells. The cones have evolved into such spectral types containing different visual pigments and sending spectral information to the appropriate ganglion cells via different types of interneurons (for recent reviews, see Dacey and Packer, 2003; Jacobs and Rowe, 2004).

Vertebrates have one type of rod visual pigment and four types of cone visual pigment located in four spectral cone types. A visual pigment consists of a protein, the opsin, enclosing a chromophore (retinal) as the actual photon-activated molecule. The spectral sensitivity of a given pigment is determined by the amino acid sequence of the opsin. The four classes of vertebrate cone opsins are named by acronyms roughly reflecting their spectral sensitivity maxima: SWS1 (short-wavelength-sensitive 1, near-ultraviolet to violet range), SWS2 (short-wavelength-sensitive 2, violet to blue range), RH2 (middle-wavelength-sensitive, green range), and LWS (long-wavelength-sensitive, yellow to red range) (for reviews, see Yokoyama, 2000; Ebrey and Koutalos, 2001; Jacobs and Rowe, 2004). For convenience, wavelength ranges are here referred to by the color they have for a human observer. The rod opsin RH1 is most closely related to the cone opsin RH2, indicating that the rod evolved out of a cone-like photoreceptor.

Many diurnal bonefish, reptiles, and birds possess all four cone types and thus the potential for tetrachromatic color vision. Mammals have lost the opsin classes RH2 and SWS2 and only retained the classes LWS and SWS1 (for reviews, see Yokoyama, 2000; Ebray and Koutalos, 2001; Jacobs and Rowe, 2004). Consequently, the most common mammalian condition is dichromatic color vision on the basis of LWS cones and SWS1 cones, allowing discrimination of shorter wavelengths from longer wavelengths, but no discrimination between longer wavelengths (for review, see Jacobs, 1993). Depending on the population density of the cones (see above and Fig. 2), this dichromatic color vision may range from robust to feeble. Interestingly, in most mammals the SWS1 pigment has shifted its sensitivity from near ultraviolet to violet or blue, i.e., to the position of the lost SWS2 pigment, and the LWS pigment sensitivity ranges from green to red depending on species (Jacobs, 1993), suggesting that spectral tuning, effectuated by amino acid changes in the opsins, is under strong selective pressure. Only in Old World primates and man has trichromatic color vision reevolved by a duplication of the LWS opsin gene to yield two spectrally discrete LWS opsins, our green and red cone opsins (for reviews, see Jacobs, 1993; Nathans, 1999). The color vision of nonprimate mammals may be roughly comparable to that of a red-green blind human.

In the following, the two mammalian cone opsin types are termed L and S for convenience. Assessment of their spectral sensitivities and their roles in color vision requires physiological and behavioral experiments. Histologically, the two cone types can be assessed by antibodies against the L- and S-opsin, respectively (for reviews, see Szél et al., 1996, 2000; Ahnelt and Kolb, 2000). Some of these are directed against conserved epitopes and recognize the respective opsin types across species. Immunocytochemical labeling of the cone opsins has two major advantages. First, it allows a determination of the population properties of L- and S-cones, including their topographic distribution across the retina (Figs. 3 and 4). These properties have important functional consequences, and they cannot be obtained by the other methods. Second, the antibodies work in fixed tissue, hence the presence and distribution of the two cone types can be assessed in species that are not easily available for physiological or behavioral experiments. Studies using opsin antibodies have greatly increased our knowledge about the diversity of cone arrangements across mammals. They have identified a number of species with particularly interesting cone properties, which have subsequently been scrutinized by molecular and physiological approaches.

thumbnail image

Figure 3. Spectral cone types across mammals. Micrographs of retinal flat mounts, double immunofluorescence labeling for L-cone opsin (in red) and S-cone opsin (in green); the focus is on the cone outer segments. A: The macaque (Macaca fascicularis) has the most common pattern of a majority of L-cones and a minority of S-cones. Macaques are cone trichromats; the L-opsin antiserum labels the green and red cones. Cone density is low in this region of peripheral retina. B: The subterranean cururo (Spalacopus cyanus) has a similar arrangement of spectral cone types. However, cone density is unexpectedly high (mid-peripheral retina). C: The ringed seal (Phoca hispida) completely lacks S-cones. D: A large region of the retina of the guinea pig (Cavia porcellus) is dominated by cones coexpressing various levels of L- and S-opsin (dual pigment cones, showing various shades of yellow and orange by the merge of red and green fluorescence). Here only a minority of the cones shows exclusive L- or S-opsin label (red or green fluorescence). All micrographs shown at same magnification. Scale bar = 50 μm.

Download figure to PowerPoint

thumbnail image

Figure 4. Dorsoventral S-cone gradient in the retina of an insectivore, the common shrew (Sorex araneus). A: Low-power view of a nasal half-retina, double-immunofluorescence-labeled for L-cone opsin (in red) and S-cone opsin (in green); the focus is on the cone outer segments. L-cones have a high density across the retina, whereas the overall lower S-cone density sharply increases from dorsal to ventral retina. D, dorsal; V, ventral; N, nasal. A nasal cut was made to flatten the retina. A central patch of retina is covered by pigment epithelium, which obscures the photoreceptors at the red filter setting but not the green filter setting. B and C: Higher-power views of fields in dorsal and ventral retina. In dorsal retina, less than 1% of the cones are S-cones (B); in ventral retina, the proportion is raised to several percent (C). There are no dual pigment cones. Scale bars = 200 μm (A); 50 μm (B for B and C).

Download figure to PowerPoint

A limitation of opsin immunocytochemistry is that the antibodies identify the opsin type (L or S) but not the exact spectral tuning, and that they do not discriminate between the two spectral types of L-cone present in trichromatic primates (Fig. 3A). Immunolabeling will not tell whether a species is a cone dichromat (most New World primates) or a cone trichromat (Old World primates). As much of this overview is based on immunocytochemical data, the primates will be largely excluded. The photoreceptor arrangements and types of color vision present across primates have been extensively studied and competently covered in a number of reviews (see, for instance, Jacobs, 1993; Ahnelt and Kolb, 2000).

Basic Mammalian Pattern

The photoreceptor properties most commonly found in mammals can be summarized in a few statements describing the basic mammalian pattern (Jacobs, 1993; Szél et al., 1996, 2000; Ahnelt and Kolb, 2000). One, all mammals possess rods and cones (duplex retinae). In most mammals, the rods greatly outnumber the cones. Two, the rod/cone proportion loosely correlates with the activity pattern of the species. Diurnal mammals commonly have higher cone densities than nocturnal ones. Three, most nonprimate mammals possess two spectral cone types, L-cones (commonly green- or yellow-sensitive) and S-cones (commonly blue- or violet-sensitive) and thus the potential for dichromatic color vision. Four, commonly, a mammalian cone contains only one visual pigment, coexpression of different opsins rarely occurs. Five, commonly, the L-cones form a dominant majority, and the S-cones a minority of 5–15%. Six, usually, both cone types show a centroperipheral density gradient, coinciding with the ganglion cell density gradient. Highest densities are present in the Area centralis, lowest densities in the retinal periphery.

However, there are quantitative as well as qualitative deviations from the basic pattern. Some striking examples are reported in the following sections.

DEVIATIONS FROM BASIC PATTERN

  1. Top of page
  2. Abstract
  3. MAMMALIAN PHOTORECEPTORS: BASIC PATTERN
  4. DEVIATIONS FROM BASIC PATTERN
  5. CONCLUSION
  6. Acknowledgements
  7. LITERATURE CITED

S-Opsin Loss in Terrestrial Mammals

Every now and then immunocytochemical studies have come across a species that completely lacks S-opsin expression and only possesses L-cones (summary scheme Fig. 5B). With just one spectral cone type, cone-based color vision is thought to be precluded. The short list of terrestrial L-cone monochromats includes two primates (owl monkey and bushbaby) (Wikler and Rakic, 1990; Jacobs et al., 1996), three carnivores (common raccoon, crab-eating raccoon, and kinkajou) (Jacobs and Deegan, 1992; Peichl and Pohl, 2000) and a number of rodents (for summaries, see Jacobs, 1993; Szél et al., 1996; Crognale et al., 1998; Peichl and Moutairou, 1998; Ahnelt and Kolb, 2000). Most likely the list will become longer as further species are studied.

thumbnail image

Figure 5. Schematic illustrations of the diversity of cone arrangements found across mammals. AD: L-cone topographies (left) and S-cone topographies (right). The retina is stylized as a circle; the small central circle indicates the generic location of the optic nerve head. In all maps, dorsal is up and temporal to the right. A: The pattern found in most species is a majority of L-cones and a minority of S-cones, both with a centroperipheral density gradient. Peak densities are commonly located at the Area centralis in the dorsotemporal quadrant. B: Normal topography of L-cones and complete absence of S-cones, found in whales, seals, and some terrestrial nocturnal species. C: Concentration of S-cones in ventral retina, while L-cones either have a relatively normal topography (e.g., in the common shrew) or concentrate in dorsal retina (e.g., in some mouse species). The latter pattern is often associated with opsin coexpression. D: Normal topography of L-cones and reverse topography of S-cones with high densities in peripheral and low densities in central retina, e.g., in the tarsier and mouse lemur. E: Cone opsin expression patterns. Left: In most species, the cones exclusively express either the L-opsin (white) or the S-opsin (black). Right: In some species, there is a regional dominance of dual pigment cones coexpressing both opsins (hatched), while pure L- and S-cones are a minority.

Download figure to PowerPoint

A common feature of these terrestrial L-cone monochromats is that they are all nocturnal, and that they are relatively isolated in their deficit, often having dichromatic close relatives. For example, in the family Procyonidae, the nocturnal raccoons (Procyon lotor and P. cancrivorous) and the nocturnal kinkajou (Potos flavus) lack S-cones while the diurnal coati (Nasua nasua) has L- and S-cones (Jacobs and Deegan, 1992; Peichl and Pohl, 2000). Within the genus Mus, the S-cone lack in M. pahari and M. plathythrix contrasts with a standard cone complement in some other species (Szél et al., 1996). Among New World monkeys, the owl monkey (Aotus trivirgatus) is the only nocturnal species, and it is the only species known to lack S-cones (Martin et al., 2000). The conclusion was that the loss of S-opsin had occurred independently several times during mammalian evolution, perhaps as genetic accidents, and that it did not jeopardize survival because for nocturnal species mostly using rod vision, color blindness is not a severe handicap. Hence the S-cone loss was considered a curiosity without much general impact. But of course it raises the question: Why then have most nocturnal mammals retained both cone types and the option to see color? We have no answer yet.

The finding of an S-cone absence in the Syrian golden hamster had more impact (von Schantz et al., 1997; Calderone and Jacobs, 1999). Golden hamsters have been a key species for studies of the photic entrainment of circadian rhythms. It was shown that short wavelengths had the strongest influence on circadian regulation, and an involvement of the S-cones was discussed (for overview, see Brainard et al., 1994). So the news that these animals lack S-cones came as a surprise.

In this context, it is noteworthy that no mammal has yet been found to lack the L-opsin completely. L-cones commonly are the more numerous cone type and hence, in addition to their role in color vision, they also provide a larger input to the luminosity and visual acuity channels in photopic and mesopic viewing conditions. To lose this cone type would severely impede visual performance in any species that chances on a situation requiring cone-based vision. S-cones are supposed to be used only for color vision and thus more easily expendable.

Actually, limited color discrimination is also possible with just one spectral cone type. This has been shown in a human cone monochromat (Reitner et al., 1991) and may be true for other mammals. It is attributed to interactions of the remaining cone type and the rods, and it is restricted to mesopic lighting conditions where both the cones and the rods are in their working range.

S-Opsin Loss in Marine Mammals

While an S-cone loss is the exception among terrestrial species, it appears to be the rule in marine whales (Cetacea) and seals (Pinnipedia/Carnivora). This has been established by immunocytochemical, electrophysiological, and molecular genetic studies. A survey of by now 13 cetacean species (11 odontocetes, 2 mysticetes) and 10 pinniped species (6 phocids, 4 otariids) has found not a single species that possesses S-opsin immunoreactive cones (Peichl and Moutairou, 1998; Peichl et al., 2001a; Griebel and Peichl, 2003). The L-cones are present at low densities (around 2% of the photoreceptors; Fig 3C). Flicker-photometric electroretinography (ERG) in the harbor seal has confirmed that this species only has one cone type, the L-cone (Crognale et al., 1998). Molecular studies in several whale and seal species have shown the presence of S-opsin genes with deleterious mutations that preclude the expression of the protein (Fasick et al., 1998; Robinson and Newman, 2002; Levenson and Dizon, 2003).

Pinnipeds belong to the caniform carnivores. The closest relatives of the cetaceans are the terrestrial artiodactyls. Terrestrial caniforms and artiodactyls have L- and S-cones, and the presence of a nonfunctional S-opsin gene in seals and whales means that their ancestors also had S-cones. It thus appears that the S-cone loss in all marine members of these two distant orders represents convergent evolution and hence a common selective pressure. The puzzling fact is that in clear open ocean waters, blue light (450–480 nm) penetrates deepest, while longer as well as shorter wavelengths are preferentially scattered and absorbed. Probably color vision is of little use in this monochrome blue underwater world, and two spectral cone types are not really needed. But losing the S-cones seems a rather unfitting adaptation, as they would be suited best to perceive brightness and contrast information. The rod and L-cone opsins of seals and whales are tuned to shorter wavelengths than those of their terrestrial relatives, indicating an adaptation to the blue-dominated visual environment, but they are not as effective as the lost S-opsin would have been.

One hypothesis to explain this apparent paradox is that the S-cones were lost in an early coastal period of cetacean and pinniped evolution. In many coastal waters, the underwater light spectrum is red-shifted due to blue-light absorption by organic and inorganic debris. In such conditions, a loss of the S-cones could have been an economical adaptation, simplifying retinal and cortical visual information processing. Some descendant species have stayed in coastal waters and for them the loss remains useful or neutral. Other descendant species have later conquered the open ocean. They may have profited from a functional S-opsin, but they could not reverse the deleterious gene defect. All they could apparently do was shift the spectral tuning of their L-cones and rods to shorter wavelengths. A detailed account of the above data and hypotheses has been given by Griebel and Peichl (2003).

Whales and seals are arrhythmic with activity phases at both day and night. Their retinae show the low cone proportions (around 2%) that are commonly found in nocturnal mammals. This indicates that their eyes are well-adapted to the low light levels encountered during diving to sometimes considerable depths, and not so much to the daytime photopic light levels encountered when emerging for breathing or when resting on land (in the case of pinnipeds). If selection went in favor of rod-based vision, the L-cone monochromacy was no major disadvantage, as in the case of the nocturnal terrestrial species described above. There is behavioral evidence for residual color discrimination in some pinnipeds and the bottlenose dolphin (for review, see Griebel and Peichl, 2003). It may be based on rod/L-cone interactions.

Other aquatic or semiaquatic mammals have retained two spectral cone types. For the manatee, this has been shown by behavioral tests (Griebel and Schmid, 1996) and by immunocytochemistry (Ahnelt and Kolb, 2000). The pygmy hippopotamus, the river otter, and the polar bear also have L- and S-cones (Peichl et al., 2001a, 2005b). In these semiaquatic species, their terrestrial activities probably had some impact on keeping cone dichromacy. The manatee is fully aquatic, but it lives in very shallow waters with spectrally broad lighting.

Ultraviolet Sensitivity

The S-cones of most mammals have sensitivity maxima in the blue to violet part of the spectrum, and the lens and cornea absorb most of the UV. As the ancestral mammalian S-pigment was a UV pigment (Hunt et al., 2001) requiring UV-transmissive eye optics, the long-wave shift of the S-pigment may be interpreted as accompanying the evolution of UV-absorbing optics to shield the retina from potentially damaging UV light. However, within the past 15 years, a number of rodents have been shown to possess S-cones with maximal sensitivity in the near UV (around 360 nm). Currently, the list of UV-sensitive rodents includes house mouse, rat, pocket gopher, gerbil, Siberian hamster, cururo, and degu (Jacobs et al., 1991, 2003; Govardovskii et al., 1992; Calderone and Jacobs, 1999; Chávez et al., 2003; Peichl et al., 2005a; Williams et al., 2005). Some other rodents have been shown to have violet- or blue-sensitive S-cones [e.g., ground squirrel (Jacobs et al., 1985), guinea pig (Parry and Bowmaker, 2002; Parry et al., 2004)]. Interestingly, the change from UV sensitivity to violet/blue sensitivity can be effected by just a single amino acid substitution (Cowing et al., 2002; Parry et al., 2004).

Most of the rodent species that have retained UV-sensitive S-cones are nocturnal, and the scotopic light levels make UV damage less of an issue. However, the degu is strongly diurnal, and the gerbil is rather diurnal. Here one has to ask what selective pressure has favored UV sensitivity over blue sensitivity. The urine, which rodents generously use to scent-mark their territories and trails, is highly UV-reflective. We have recently suggested that rodents might profit from seeing their scent marks in addition to smelling them (Chávez et al., 2003), but this hypothesis still awaits behavioral testing.

Outside the rodents, no mammals with UV-sensitive S-cones have been identified to date, with the exception of marsupials and possibly bats. These two orders are discussed in separate sections below.

Dual Pigment Cones

Most mammals follow the rule book by expressing only one opsin type per cone. This is the canonical way of being spectrally selective and thus useful for color vision. However, in several species immunocytochemistry has revealed substantial populations of cones that label for both L- and S-opsin. After clarifying that these findings were not artifacts of antibody cross-reactivity, it is now established that opsin coexpression in cones is more widespread. Ágoston Szél, Pal Röhlich, and their colleagues, who discovered opsin coexpression, have published thorough surveys of the occurrence and distributions of mammalian dual pigment cones (Szél et al., 2000; Lukáts et al., 2005).

In their original report, Röhlich et al. (1994) described dual pigment cones in the house mouse, guinea pig, and rabbit. These species had normal cone arrangements with an L-cone majority and an S-cone minority in the dorsal retina, but parts of the ventral retina dominantly or exclusively contained S-cones. These S-fields extended over the entire ventral retina in mouse and guinea pig, but only over the ventral periphery in rabbit. In the transition zone between the normal part and the S-field, many of the cones coexpressed both opsins in various mixtures (Fig. 3D, summary scheme Fig. 5E). Since then, opsin coexpression has been most thoroughly studied in the mouse. With more sensitive opsin antibodies, it was shown that the vast majority of S-opsin expressing cones also express low levels of L-opsin (Glösmann and Ahnelt, 1998; Applebury et al., 2000). From this, Applebury et al. (2000) have postulated that there is only one cone type in the mouse, which produces a pattern of subtypes across the retina by regulating the relative expression levels of L- and S-opsin. Others have challenged this hypothesis (Neitz and Neitz, 2001). A recent study provides evidence that across the entire mouse retina there is a sparse population of genuine S-cones with the appropriate circuitry of S-cone-selective bipolar cells (Haverkamp et al., 2005). From a comparative point of view, it would be comforting if the mouse had the standard set of two cone types and if dual pigment cones would just be L-cones that fail to turn off S-opsin expression.

Physiological evidence confirms the presence of dual pigment cones in mouse and shows that the coexpressed pigments are both linked to the phototransduction cascade (Lyubarsky et al., 1999; Ekesten et al., 2002). This should turn the cone into a spectral broadband detector with little use for color vision. On the other hand, the spectral tuning seems to be determined by the relative amounts of S- and L-pigment that a dual pigment cone contains, so it could still signal spectral information. In behavioral experiments, mice are able to discriminate colors, albeit poorly (Jacobs et al., 2004). Hence, while a substantial population of dual pigment cones certainly does not help color vision, it does not completely prevent it either. It is still unclear how pigment coexpression comes about, whether it may have some functional benefit, or whether it is just an accidental mild affliction of some species (Lukáts et al., 2005).

Unusual Cone Topographies

In most mammals, areas of highest cone density correspond to areas of maximum ganglion cell density. Cone densities peak in the regions of high visual acuity (Area centralis and visual streak) and are lowest in the retinal periphery. Commonly, S-cones constitute a constantly low fraction of the cones across the entire retina. However, there are a number of species that deviate from this pattern. One example is the tree shrew with its cone-dominated retina (Fig. 2D). Here, the densities of both L- and S-cones are highest in the ventronasal mid-periphery, not in the temporally located Area centralis (Müller and Peichl, 1989; Petry et al., 1993). While this may not seriously affect visual acuity, as in the tree shrew the Area centralis cone density is still higher than the acuity-limiting ganglion cell density, it is completely open what adaptive advantage the high ventronasal cone densities might provide.

A further topographic oddity found in several mammals is a marked regional increase in S-cone proportions. All four insectivore species studied so far show several-fold higher S-cone proportions in ventral than in dorsal retina (Peichl et al., 2000). For the common shrew (Sorex araneus), this is illustrated in Figure 4. The hyena also has highest S-cone densities in the ventral retina (Calderone et al., 2003). The shift toward dominant S-opsin expression in the ventral retina is even stronger, for example, in the house mouse and in the subterranean pocket gopher. In these two species, but not in the insectivores, it is associated with the coexpression of S- and L-opsin in a large fraction of the cones. Mouse opsin coexpression has been discussed above; pocket gopher cone properties are detailed further below. A summary scheme of such bipartite S-cone topographies is given in Figure 5C.

The ventral increase of S-cone proportions has been suggested to be an adaptation to the natural light distribution. Skylight, projected onto the ventral retina, contains higher proportions of short wavelengths than light reflected from the ground, falling on the dorsal retina. Increased short-wave sensitivity in the ventral retina may thus improve contrast when viewing objects, e.g., aerial predators, against skylight (Szél et al., 1996). The hypothesis is flawed by the fact that some species facing very similar visual environments do not have similarly divided retinae. Even within the genus Mus, some species have a ventral S-cone dominance, some have a standard S-cone distribution, and some lack S-cones completely (Szél et al., 1996).

In two dichromatic nocturnal primates, the tarsier Tarsius spectrum (Hendrickson et al., 2000) and the mouse lemur Microcebus murinus (Dkhissi-Benyahya et al., 2001), the S-cone density shows a reverse gradient with highest densities in the retinal periphery and lowest densities in the central retina (summary scheme, Fig. 5D). Central S-cones are so sparse that they can hardly contribute to visual processing; presumably these animals are essentially colorblind in central retina. Some other lemurs have conventional S-cone distributions (Peichl et al., 2001b). A similar but weaker reverse S-cone gradient exists in a marsupial, the honey possum (Arrese et al., 2003). A still different S-cone topography is present in the ground squirrel and in four marsupials (tammar wallaby, fat-tailed dunnart, quokka, and quenda), where highest S-cone proportions are located in dorsal peripheral retina (Kryger et al., 1998; Hemmi and Grünert, 1999; Arrese et al., 2003, 2005b). The role of these S-cone-rich peripheral regions for vision has yet to be elucidated.

Subterranean Mammals

A surprisingly large number of about 300 mammalian species have adopted a completely or near-completely subterranean lifestyle. It is generally assumed that the visual systems of subterranean mammals have undergone extensive convergent regression in adaptation to the lightless underground habitat (for summaries, see Burda et al., 1990; Nevo, 1999). The model species to support such claims is the muroid blind mole-rat Spalax ehrenbergi, which has atrophied subcutaneous eyes and retinal deficits (Sanyal et al., 1990; Cernuda-Cernuda et al., 2002). Spalax possesses rods with a functional rod opsin (Sanyal et al., 1990; Janssen et al., 2000; Cernuda-Cernuda et al., 2002). While cones have not been conclusively identified by morphology, there is a functional L-cone opsin but no functional S-cone opsin (David-Gray et al., 2002). The only role of the Spalax eye is seen in mediating the entrainment of circadian rhythms (Sanyal et al., 1990; David-Gray et al., 2002). Where the photic cues would come from in a subterranean environment is still open.

Recent studies on other subterranean rodents have challenged the view of general convergent eye regression and unearthed some unexpected photoreceptor properties. In African mole-rats (Cryptomys spp. and Heterocephalus glaber, Bathyergidae), the eyes are relatively small but the optical apparatus and the retina are normally developed. The photoreceptor layer contains rods and an unexpectedly high proportion of about 10% cones. About 90% of the cones dominantly express the S-opsin; most of them also express low levels of L-opsin (dual pigment cones), and only about 10% appear to be pure L-cones. This is the largest S-/L-opsin ratio reported in any mammal (Peichl et al., 2004). Most other neurons of the retinal circuitry are present, but their structural organization is less regular than in surface dwellers (Mills and Catania, 2004).

The eyes of the South American cururo (Spalacopus cyanus, Octodontidae), a subterranean caviomorph rodent, are only slightly smaller than those of its surface-dwelling octodontid relatives. Here, too, the photoreceptor layer contains rods and about 10% cones. In contrast to the arrangement in the African mole-rats, a conventional majority of the cururo cones are L-cones, and S-cones constitute a regionally varying proportion of 2–20%; there are no cones coexpressing both opsins (Fig. 3B) (Peichl et al., 2005a). The L-cones are green-sensitive and the S-cones UV-sensitive. This spectral tuning is very similar to that in the diurnal surface-dwelling degu (Octodon degus) and may be an octodontid family trait (Chávez et al., 2003; Jacobs et al., 2003). The diurnal degu has about 30% cones (Jacobs et al., 2003), but the nocturnal surface-dwelling Octodon bridgesi and O. lunatus only have 2% cones (Peichl et al., 2005a).

The subterranean pocket gopher (Thomomys bottae), a geomyid rodent, has an even higher cone proportion of about 25%. A UV-sensitive S-opsin and a green-sensitive L-opsin are present, but their distribution patterns are again different. All cones contain the S-opsin, and the L-opsin is coexpressed in practically all cones of the dorsal half-retina, while it is completely absent from the ventral half-retina (Williams et al., 2005). Finally, the subterranean European mole (Talpa europaea, Insectivora) also has about 10% cones. In dorsal retina, there are about twice as many L-cones as S-cones; in ventral retina, this ratio is reversed. A subpopulation of the cones coexpresses both opsins (Glösmann et al., 1999).

Hence, a common feature in these subterranean species is that their cone proportions (10–25%) are more similar to those of diurnal than to those of nocturnal surface dwellers. The peak cone densities vary considerably across species, from about 30,000/mm2 in the pocket gopher and cururo to about 15,000/mm2 in the African mole-rats. In fact, the latter are not very different from the cone densities in mouse (about 12,500/mm2) (Jeon et al., 1998). This points to another surprising feature: the rod densities in all these species are markedly lower than in nocturnal surface dwellers, hence similar cone densities constitute larger cone proportions. Rod density is about 440,000/mm2 in the mouse, about 220,000/mm2 in the cururo, 100,000–150,000/mm2 in Cryptomys and Talpa europaea, and only 55,000–100,000/mm2 in the pocket gopher (all figures taken from the respective papers). The reason for these lower densities is that the individual rods are larger.

It thus appears that subterranean mammals have reduced their rod populations and maintained or increased their cone populations. This does not look like adaptation to a lightless habitat. The large diversity of cone opsin arrangements across subterranean species, from a normal mix of L- and S-cones, to a dominance of S-opsin, to coexpression of the opsins, also argues for species-specific adaptations to different demands and against convergent adaptation to the common subterranean darkness. Perhaps a complete absence of light does not exert any selective pressure, because ever so large increases in sensitivity would not enable vision. So even rare and short episodes of surface activity, where vision would be useful to avoid predators, could have been the relevant factor in shaping the photoreceptor patterns. The above-ground activities of most subterranean species have yet to be studied. Some are known to emerge to the surface every now and then, but it is still open how much of this happens at light conditions favoring cone-based vision.

Bats

Microchiropteran bats (microbats) are a further large group of mammals in which vision is traditionally considered of negligible importance. They are predominantly nocturnal and most famous for their echolocation. The eyes are small, commonly less than two millimeters in diameter, and obviously ill-suited for refined vision. Some older histological studies have claimed that microbats have all-rod retinae (see, for instance, Kolmer, 1936; Walls, 1942; Rochon-Duvigneaud, 1943), and for a long time, vision research has stopped there. More recent studies have established that vision contributes to orientation in some microbats, and that visual acuity is comparable to that of murid rodents (for data and references, see, for instance, Pettigrew et al., 1988; Eklöf et al., 2002; Rydell and Eklöf, 2003). Hence, there is a renewed interest to study the photoreceptor composition of bats with modern techniques.

A molecular genetic analysis has demonstrated the presence of L- and S-opsins in two frugivorous megachiropteran and one insectivorous microchiropteran species (Wang et al., 2004), and we have started a survey across microchiropteran superfamilies to assess the presence and distribution of cone types by opsin immunolabeling (Müller and Peichl, 2005). The immunocytochemical evidence so far is that all microbats possess substantial cone populations in addition to the rods. Probably because of their rod-like morphology, they have been overlooked by early researchers. Some species have L- and S-cones, others lack the S-opsin. Further work will have to elucidate the full diversity of cone properties present across microbats and to uncover potential correlations with the various feeding habits and visual needs.

While our immunocytochemical approach can identify the cone types and their population magnitudes and distributions, it cannot identify the exact spectral sensitivity of the S- and L-opsin. Molecular analysis of the opsins in two megabats and one microbat suggests that the L-opsin is rather long-wavelength-sensitive and the S-opsin is UV-sensitive (Wang et al., 2004). Together with rodents and marsupials, the bats might be the third mammalian order where UV vision occurs. A further surprise was the presence of duplicate L-opsin genes in one megabat species, the first reported case of opsin gene duplication outside the primates (Wang et al., 2004). This will be a fascinating avenue to follow.

A recent behavioral study has shown that the phyllostomid flower bat Glossophaga soricina is sensitive to ultraviolet stimuli in conditions of dark adaptation (Winter et al., 2003). A test for color vision (color discrimination) was negative. The conclusion was that the rods were responsible for the UV sensitivity. In addition to their main sensitivity peak (the α-band), all rod and cone visual pigments have a secondary sensitivity peak in the near UV (the β-band). In many mammals, this is not relevant for vision because UV light is blocked by the optics. However, species with UV-sensitive cones, such as microbats or some rodents, have UV-transmissive optics, and the β-band may contribute to retinal spectral sensitivity in rod vision as well as cone vision.

Marsupials

Strictly speaking, the mammalian reduction to two spectral cone types SWS1 and LWS may only hold for eutherian mammals. Arrese et al. (2002, 2005b) have demonstrated the presence of three cone types and hence the potential for trichromatic color vision in four marsupials (honey possum, fat-tailed dunnart, quokka, and quenda). The three cone types have sensitivity peaks in the yellow to orange, green, and ultraviolet to violet range, respectively. By molecular analysis, the first and last have now been shown to belong to the LWS and SWS1 group (Strachan et al., 2004; Arrese et al., 2005a, 2005b). The molecular identity of the green-sensitive cone opsin is still unknown. Arrese et al. (2005a) could exclude that it is a modified LWS opsin, and they did not find evidence for the presence of an RH2 opsin (which would have indicated that these marsupials have retained a preeutherian vertebrate opsin). As they found a duplicated rod opsin gene in the dunnart, it is possible that the green cone opsin derives from a rod opsin.

In contrast, the tammar wallaby seems to be a dichromat possessing only two cone opsins, as shown by behavioral, immunocytochemical, and molecular studies (Hemmi, 1999; Hemmi and Grünert, 1999; Deeb et al., 2003). Thus, present evidence suggests that there are dichromatic as well as trichromatic marsupials. At least two scenarios come to mind. Perhaps there were three cone types at the root of the marsupial ramification, and some branches have lost one. This would indicate that marsupial dichromacy has evolved independently of eutherian dichromacy. Alternatively, the last common ancestor of marsupials and eutherians already only had two cone types, and some marsupials have reinvented a third spectral cone type, as the primates have done in the eutherian branch. Further studies are necessary to give us a better view on the emergence of marsupial cone types and color vision.

CONCLUSION

  1. Top of page
  2. Abstract
  3. MAMMALIAN PHOTORECEPTORS: BASIC PATTERN
  4. DEVIATIONS FROM BASIC PATTERN
  5. CONCLUSION
  6. Acknowledgements
  7. LITERATURE CITED

The examples reviewed here highlight the impressive diversity of photoreceptor properties and arrangements in mammals. Comparative analysis has identified certain patterns that are common to many species and has deduced a set of basic rules. But practically all of these are soft rules, permitting stark deviations from the basic bauplan in one or the other taxon. Figure 5 schematically summarizes the main variants of mammalian cone arrangements. The rodents are the mammalian order with the largest diversity of photoreceptor patterns (Jacobs, 1993; Ahnelt and Kolb, 2000; examples above). This is not surprising, as they are the most species-rich order, having conquered the broadest range of habitats. The artiodactyls appear to be the least diverse order in terms of photoreceptors. As far as they have been studied, all have the standard set of rods, L-cones and S-cones (Jacobs, 1993; Ahnelt and Kolb, 2000). If one includes cetaceans among the artiodactyls, as molecular phylogenetic data suggest, the S-cone loss is the single outstanding deviation.

Comparative studies of photoreceptor diversity increase our zoological knowledge and appreciation for the evolutionary malleability of photoreceptor properties in response to specific visual challenges. They also specify to which extent findings in standard laboratory models can be generalized to all mammals. For example, cone-based performance in the mouse retina with its many dual pigment cones certainly is not representative for the mammalian retina, and the spectral sensitivity of light-driven entrainment of the circadian clock in normal dichromatic mammals may be different from that in the L-cone monochromatic golden hamster. Last but not least, comparative studies are an appropriate way to identify suitable new model species for tackling particular questions of photoreceptor and retinal organization. As the Danish physiologist and Nobel laureate August Krogh has put it, “for a large number of problems there will be some animal of choice or a few such animals on which it can be most conveniently studied” (Krogh, 1929).

This review has also indicated numerous instances where we are yet unable to interpret an observed photoreceptor specialization in adaptive terms. Some of these deviations seem to reflect strong selective pressure, but behavioral and physiological studies to prove the adaptive advantage are lacking. Within a single rodent genus, Mus, there is the whole gamut of cone arrangements from conventional S-/L-cone ratios to regional opsin coexpression to an absence of S-cones, despite the fact that some of these species have similar habitats, lifestyles, and feeding habits (Szél et al., 1996; Lukáts et al., 2005). With reference to the Krogh quotation above, we have identified some animals of choice, but we have not yet grasped the problems we can study on them. Finally, the photoreceptors are just the input stage to retinal processing. For most species mentioned here, the postreceptoral neurons, their circuits, and their adaptative specializations are terra incognita. All this is a rewarding field for further endeavours.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MAMMALIAN PHOTORECEPTORS: BASIC PATTERN
  4. DEVIATIONS FROM BASIC PATTERN
  5. CONCLUSION
  6. Acknowledgements
  7. LITERATURE CITED

The author is grateful to all collaborators in their comparative studies that form the backbone of this overview and to Martin Glösmann for valuable comments on an earlier draft of the manuscript.

LITERATURE CITED

  1. Top of page
  2. Abstract
  3. MAMMALIAN PHOTORECEPTORS: BASIC PATTERN
  4. DEVIATIONS FROM BASIC PATTERN
  5. CONCLUSION
  6. Acknowledgements
  7. LITERATURE CITED