Accessory corner cones (ACC) have recently been suggested to be UV-sensitive photoreceptor cells. With a view toward explaining prey detection, we examined the topography of retinal ganglion cells and ACCs in two Antarctic nototheniids occupying different ecological niches: the cryopelagic Pagothenia borchgrevinki and the benthic Trematomus bernacchii. Isodensity maps of retinal ganglion cells showed that the main visual axis, coincident with the feeding vector, was in a forward direction in both species. Visual acuity was determined as 3.64 and 4.77 cycles/degree for the respective species. In P. borchgrevinki the highest density of ACCs was associated with the eye's main visual axis. This suggested that this species uses UV-vision during forward-swims and probably in encounters with prey. On the other hand, T. bernacchii possessed two horizontal band-shaped high-density areas of ACCs, which stretched from temporal to nasal and ventral to peripheral retinal regions. Therefore, this species appears to use UV-vision to watch prey across the entire circumference of the lateral area and in the water column above its head.
The presence of UV-absorbing visual pigment in the single cone at the corner of the retinal square cone mosaic, named accessory corner cone (ACC), has been reported in many freshwater fishes and euryhaline fishes by using microspectrophotometrical and/or molecular analyses (Avery et al., 1983; Hárosi and Hashimoto, 1983; Whitmore and Bowmaker, 1989; Archer and Lythgoe, 1990; Bowmaker et al., 1991; Hawryshyn and Hárosi, 1991; Raymond et al., 1993; Hisatomi et al., 1996, 1997). Furthermore, in juveniles of some salmonids species (Salmo trutta, Salmo gairdneri (=Oncorhynchus mykiss), Salmo salar, Oncorhynchus nerka) and yellow perch Perca flavescens, the loss of UV-photosensitivity associated with the almost complete disappearance of ACCs from the retinal cone mosaic has been confirmed (Bowmaker and Kunz, 1987; Hawryshyn et al., 1989; Loew and Wahl, 1991; Kunz et al., 1994; Novales-Flamarique, 2000).
Recently, we reported the presence of ACCs histologically from the eyes of adults of seven Antarctic nototheniid species as follows: Trematomus bernacchii, Trematomus newnesi, Trematomus hansoni, Trematomus pennellii, Trematomus scotti, Pagothenia borchgrevinki, Lepidonotothen squamifrons (=kempi) (Miyazaki et al., 2001). Although several specific visual tasks for UV-vision have been proposed (foraging, navigation, and/or communication) and we have very few direct observations on visual feeding in Antarctic fishes, we suggested that Antarctic fishes could increase their feeding efficiency and stock energy by making use of UV-vision during the summer month (Miyazaki et al., 2001).
Topographical analyses of retinal ganglion cells as well as cone photoreceptors can provide us with valuable information on the visual capabilities of fishes. In adult teleosts, the position of the “area centralis (AC)” is related to both habitat and main visual vector of feeding behaviour (Collin and Pettigrew, 1988b, c; Williamson and Keast, 1988; Browman et al., 1990). Furthermore, the distribution of cone photoreceptors determines how the spectral sensitivity of a fish eye changes across its visual field, for different kinds of cones may contain visual pigments with different spectral absorption characteristics (McFarland and Munz, 1975; Beaudet et al., 1997).
In several Antarctic fish species, retinal organizations have been described for specific regions and estimates on visual resolution or photosensitivity were given (Meyer-Rochow and Klyne, 1982; Eastman, 1988; Pankhurst and Montgomery, 1989, 1990; Fanta et al., 1994, Meyer-Rochow et al., 1999). However, the topography of the ganglion cells or cone cells of the whole retina, and in particular the position of the ACs, is still poorly known. Our primary objective was to histologically examine the prey detection system in nototheniids. We also hoped to establish whether the ACCs, as the putative UV-sensitive cones, could assist the fish in prey detection and uptake. The UV-wavelength absorption of ACC's visual pigment has been shown by the previous studies mentioned above, although we have not verified about Antarctic fish yet. In this paper, we therefore examined the retinal topographies of ganglion cells and ACCs for entire retinas of two nototheniids with different habits, namely Pagothenia borchgrevinki and Trematomus bernacchii. We also determined the fish eyes' UV-visual fields. Finally, visual acuity and optical axes were estimated on the basis of retinal ganglion cell distributions, since retinal ganglion cells are providing the link between eyes and behavioural output via the optic nerve.
MATERIALS AND METHODS
P. borchgrevinki and T. bernacchii were caught by line-fishing from a depth of ca. 30 m in Lützow-Holm Bay near Syowa Station (69° S, 39° E). Total lengths of the P. borchgrevinki and T. bernacchii specimens used were 210 mm and 215 mm, respectively. According to Collin and Pettigrew (1988a), there is no remarkable difference in the topography of cells among specimens of the almost same size of the same species. Based on their observation results, we used a single specimen for each species. Following capture, the fish were anaesthetized with MS222 and then immediately fixed in 10% formalin. Thereafter the eyes were enucleated and cornea, iris, lens, and vitreous humor were removed. The remaining retinae were prepared as follows.
Examination of ganglion cells
The left retinae were radically incised and flatly mounted, and then whole-mount preparations were made following the protocol of Ito and Murakami (1984). The preparations were glued onto slides with the ganglion cell layer facing upward, and stained with crecyl violet. Ganglion cells numbers per 0.13×0.13 mm of each fish retina, about 20 mm in diameter in both fish, were counted at 2mm intervals, under a light microscope, using a calibrated eyepiece. In P. borchgrevinki and T. bernacchii, 127 and 98 sampling points were used. Values for the ganglion cells were converted into numbers of cells per mm2, and then isodensity counter maps of ganglion cells were constructed. The retinal resolving powers were calculated according to the protocol of Collin and Pettigrew (1989), the values thus obtained representing upper limits based on the following equations:
The distance from lens center to retinal focus (posterior nodal distance, PND) was calculated from lens radius (r; mm) and Matthiessen's ratio (2.55):
Examination of ACCs
Right retina of the P. borchgrevinki and T. bernacchii was divided into 24 and 30 topographical locations, respectively, and thus the position of each piece in the original retina was determined accurately. The pieces were dehydrated in alcohol and embedded in paraffin. Tangential sections were cut at 7 μm thickness and stained with haematoxylin and eosin. Cone photoreceptor layers were observed under the light microscope, and numbers of accessory corner cone cells per 0.1×0.1 mm were counted for respective retinal pieces.
Distribution of ganglion cells
All ganglion cells lying within the retinal ganglion cell layer in each fish species were counted. The averages of cell density, estimated from the counts of ganglion cells and counted number of points, were approx. 1.5×103 cells/mm2 in P. borchgrevinki and 2.5×103 cells/mm2 in T. bernacchii (Table 1). The calculated cell densities were upper estimates, which may include displaced amacrine cells.
Summery of ganglion cell counts and calculations of visual acuity in two nototheniids.
Ganglion cells were not uniformly distributed in the retinae of either species. Isodensity maps of P. borchgrevinki and T. bernacchii resembled each other in that in both species, the region of highest cell density, termed area centralis (AC), was located in the temporal retina (Figs. 1a and 1b). The position of the AC indicates that the visual axis was in a forward direction. In both species, however, a weak horizontal visual streak was also observed a little above the optic papilla (see in Figs. 1a and 1b).
Ganglion cell peak densities were 3313 and 6804 cells/mm2 in P. borchgrevinki and T. bernacchii, respectively, and the corresponding lens diameters were 5.65 mm and 5.15 mm. The PNDs, α s, and the numbers of ganglion cells sub-tended by one degree of visual arc in the respective species were calculated as shown in Table 1. Consequently, the upper limit of the theoretically derived visual acuity, based on ganglion cell counts, can be given as 3.64 and 4.77 cycles/degree in P. borchgrevinki and T. bernacchii, respectively.
Distribution of ACCs
Single and double cones were alternately and regularly aligned in the photoreceptor layer, and formed a square mosaic pattern in both species (Figs. 2a and 2b). The ACCs, which were located at the corners of the double cones, could be distinguished from the central single cones by the direction of the axes of the double cones (Figs. 2c and 2d). In both species, the ACCs were present in the entire retina, but the density varied across the retina. Figs.3a and 3b illustrate ACC density distributions in the retinae of P. borchgrevinki and T. bernacchii, respectively. The highest density of the ACCs was in the temporal retina of both species, which corresponded to the AC. Peak densities of the ACCs were 32 and 29 cells par 0.01mm2 in P. borchgrevinki and T. bernacchii, respectively.
In P. borchgrevinki, ACC density was high near the optic papilla of the temporal retina. On the other hand, in T. bernacchii, the highest density was seen in the region of the horizontal streak just dorsal of the mid-retina. An additional, high density area was seen in the peripheral part of the ventral retina.
Triple cones were observed near the optic papilla in T. bernacchii (Fig. 3c), but not in P. borchgrevinki.
Both P. borchgrevinki and T. bernacchii belong to the Nototheniidae of the suborder Notothenioidei. However, the former species is cryopelagic, whereas the latter is benthic. Furthermore, the two differ in body shape and position of the eye. The body in P. borchgrevinki is laterally compressed and the eyes are positioned laterally. This body shape is commonly associated with fast-swimming fishes and, indeed, this species is often seen as small schools swimming just beneath the sea-ice (Meyer-Rochow, 1982; Meyer-Rochow and Klyne, 1982; Eastman and DeVries, 1985; Foster et al., 1987; Gon and Heemstra, 1990; Pankhurst and Montgomery, 1990). On the other hand, the head of T. bernacchii is broader, and the eye is situated higher on the head and directed anterolaterally. This body shape is typical of benthic fishes, such as gobies and sculpins.
In the present investigation, the main visual axis, thought to provide a clue for the high resolving power of the AC (Shand et al., 2000), was determined to be forward-looking in both species, which is in agreement with the AC's location in the temporal retina. Therefore, the best vision in both species to detect prey items must be considered to lie in front of the fish. Earlier workers have also suggested that T. bernacchii has a forward-directed feeding vector, since it has a large anterior aphakic space. As for P. borchgrevinki, it lacks the anterior aphakic space and may therefore posses an extended visual field (Pankhurst and Montgomery, 1989; Macdonald and Montgomery, 1991; Eastman, 1993).
The binocular visual field, based on eye position and body morphology in T. bernacchii, is broader than that of P. borchgrevinki but less far-reaching. On the contrary, P. borchgrevinki appears to be able to detect prey further away from the body than T. bernacchii (Fig. 4). Consequently, it is likely that P. borchgrevinki can focus earlier on prey during cruising, while T. bernacchii waits and watches prey over a wide range before it initiates an attack. A similar difference in visual fields has been noted between two other nototheniids, namely Gobionotothen gibberifrons and Trematomus newnesi. The latter two species share the same environmental depth of 40–80 m bottom, but they differ in their feeding-behavioural strategies (Fanta et al., 1994). The expansion of fish visual field should also be discussed based on the eye movement. Regarding eye movement of nototheniids, the maximum eye rotation has been reported as about 15 and 20 degree for P. borchgrevinki and Dissoticus mawsonii, respectively (Montgomery and Macdonald, 1985). Nevertheless, although we don't know how nototheniids use monocular and binocular vision for their visual interest, the eye movement might allow fish to observe its surroundings without moving head or body (Fritsches and Marshall, 2002). Both of P. borchgrevinki and T. bernacchii may be able to aim prey with the expansion of visual covering range by eye movement.
Regarding visual acuity, there was no remarkable difference in the present two species. The correlation between the visual acuity level of fishes and their swimming types has not been clearly shown in Tamura (1957). Retinal resolving power in cycles per degree can be converted to that of minutes of arc (Murayama and Somiya, 1998), e.g., according to the equationP. borchgrevinki and T. bernacchii, respectively. These values are almost the same as those in other marine teleosts, which were derived from maximum cone densities (Pagrus major: 6.4′ for a fish of 200 mm in body length; Lateolabrax japonicus: 8.5′ for an 180 mm long fish; Sebastiscus marmoratus: 6.7′ also for an 180 mm long fish, Tamura, 1957).
Meyer-Rochow and Klyne (1982) have counted cone and ganglion cells in both species (P. borchgrevinki: approx. 8300 cells/mm2 for a fish of approx. 200 mm total body length; T. bernacchii: approx. 4200 cells/mm2 for a fish of approx. 240 mm total body length: converted from original data). Furthermore, there have been studies on the theoretical visual acuity calculated from retinal cone densities for both species (P. borchgrevinki: 25–50′ for fish of 63–220 mm total body length: Pankhurst and Montgomery, 1990; T. bernacchii: 10′ in the temporal retina for a fish of 192 mm total body length; Miyazaki, unpublished data). Some differences between earlier values and our present values might have been caused by differences in fish size or retinal region analyzed for each examination. Nevertheless, a close agreement between values based on cone cell and ganglion cell estimates, as well as behavioural tests clearly exists (Collin and Pettigrew, 1989; Arrese et al., 2000).
The maximum distance (L; cm), at which a fish could recognize a prey item (φ; cm), can be estimated from the fish's visual acuity according toMiyazaki and Nakamura, 1990; Miyazaki et al., 2000). On that basis, for example, the recognizable distance for a prey of 3 cm diameter would be about 12.6 m in P. borchgrevinki and 16.4 m in T. bernacchii.
While these figures do not seem to vary too greatly between the two species, the distribution of the ACCs in the retina, on the other hand, was quite dissimilar. In P. borchgrevinki, the region in which the density of the ACCs was highest coincided with the position of the AC. Therefore, UV-vision in P. borchgrevinki must be assumed to be most effective in a forward direction, and consequently P. borchgrevinki ought to principally recognize prey in the swimming direction from a larger distance than T. bernacchii (Cronin et al., 1994; Losey et al., 1999). As for T. bernacchii, the high-density area was the horizontal streak, which stretched from the temporal to the nasal retina. An additional area of high density was developed in the peripheral parts of the ventral retina. UV-vision in T. bernacchii, therefore, appears to be effective not only along the main visual axis of the forward-movement but also across a vast lateral area and into the upper water column itself. This extensive field could improve prey discovery, allowing T. bernacchii to remain relatively motionless in anticipation of prey. All of these conclusions are supported by the eye's morphology. We believe P. borchgrevinki detects pelagic prey, e.g. copepods (Hoshiai and Tanimura, 1981; Hoshiai et al., 1989) as silhouettes against downwelling light, whereas T. bernacchii relies mainly on laterally incident light to feed, since T. bernacchii in contrast to P. borchgrevinki displays an obvious corneal iridescence (Macdonald and Montgomery, 1991; Eastman, 1993), which screens out bright downwelling light.
The benefit for fish using UV-vision for feeding, lies in enhancing the contrast of prey against a UV-background and it is thought that UV-sensitivity is widespread in zooplanktivorous fishes (e.g., juvenile rainbow trout Oncorhynchus mykiss, pumpkinseed sunfish Lepomis gibbosus: Browman et al., 1994; juvenile yellow perch Perca flavescens: Loew and Wahl, 1991; Loew et al., 1993; pomacentrid fish Dascyllus trimaculatus, Pomacentrus coelestis, Chromis punctipinnis: McFarland and Loew, 1994). In Antarctic nototheniids, UV-vision may represent one strategy to increase foraging efficiency during the Antarctic summer, when light conditions are appropriate for visual feeding and a visual threshold of 2×109 photons cm−2 s−1 is reached (Morita et al., 1997). However, other and additional means of detecting predators and prey are entirely possible, especially if it is remembered that individuals with aberrant eyes can, apparently, reach full adulthood (Meyer-Rochow, 1990). To (a) determine the extent that nototheniids can make use of UV-light and (b) understand the role of triple cones, present in the retina of T. bernacchii (but not P. borchgrevinki), are thus our next goals.
We are thankful to members of 34th and 35th Japanese Antarctic Research Expedition for their kind support during respective research in Antarctica. This study was partly supported by the research fund of NIRS Brain Research Project.