A fish eye out of water: epithelial surface projections on aerial and aquatic corneas of the ‘four-eyed fish’Anableps anableps


Professor Shaun P Collin, School of Animal Biology, The University of Western Australia, Crawley WA 6009, AUSTRALIA, E-mail: shaun.collin@uwa.edu.au


Background:  Vertebrate corneas feature a variety of microprojections, to which a tear film adheres. These microprojections are formed by folds in epithelial cell membranes, which increase surface area, stabilise the tear film and enhance movement of nutritional and waste products across cell membranes. Differences in corneal microprojections among vertebrates have been correlated with habitat and differ markedly between terrestrial and aquatic species.

Methods:  This study investigated epithelial microprojections of both the aerial (dorsal) and aquatic (ventral) corneal surfaces of the ‘four-eyed fish’Anableps anableps using scanning electron microscopy.

Results:  The central region of the dorsal cornea, which projects above the water, had a density of 16,387 ± 3,995 cells per mm2, while the central region of the ventral cornea (underwater) had a density of 22,428 ± 6,387 cells per mm2, a difference that suggests an environmental adaptation along the two visual axes. Both corneal surfaces were found to possess microridges rather than microvilli or microplicae characteristic of terrestrial/aerial vertebrates. Microridges were 142 ± 9 nm wide and did not differ (p = 0.757) between dorsal and ventral corneas. Microridges were consistently separated by a distance of 369 ± 9 nm across both corneas.

Conclusion:  Dorsal-ventral differences in corneal epithelial cell density in Anableps anableps suggest a difference in osmotic pressure of the two corneas. The modest differences in the microprojections indicate that the need to secure the tear film underlying each optical axis is of prime importance, due to the likelihood that a persistent layer of water normally covers both dorsal and ventral corneal surfaces and that maintaining a transparent optical pathway for vision is critical for a species prone to predation from both above and below the water's surface.

The corneal surface in the vertebrate eye features microprojections formed by folds of the epithelial cell membranes. These microprojections increase the surface area available for the transport of nutritional and waste products across the cell surface.1 Although the microprojections form a seemingly rough optical surface, this is neutralised by the smooth surface of the tear film or mucus that coats the cornea, thereby ensuring maintenance of a transparent optical interface.2

Aquatic and terrestrial environments typically present very different optical, osmotic and physico-mechanical problems to the cornea. In aquatic environments, the cornea contributes little to the refractive power of the eye due to the similarity of the refractive indices of water (refractive index of 1.33) and the cornea (refractive index of 1.37).3,4 In terrestrial environments of course, the refractive index of air is 1.00 at standard temperature and pressure5 and consequently the refractive power of the cornea is markedly greater in air. In terrestrial environments, the cornea would be damaged by desiccation and abrasion by particulate matter carried in the wind, if it were not for the moist tear film that envelops it. In aquatic environments, the tear film might protect the cornea from abrasion and also from infection as mucus does for the rest of the body.6 This tear film is thought to be stabilised by the presence of corneal microprojections;1 however, the tear film can evaporate when exposed to air, whereas in water the tear film is protected from evaporation but can be washed away.

It is likely that the evolutionary pressures of these different environmental conditions have led to the variety of corneal surface projections observed among vertebrates. Aquatic organisms, especially marine teleosts, feature microridges as the dominant structure of the corneal epithelial surface,2,7 which provide greater stabilisation of the tear film in aquatic environments. Teleosts also secrete a viscous mucous coating, which further stabilises the tear film. Terrestrial vertebrates, with a less stringent requirement for a physically stable tear film feature microplicae and/or microvilli in place of microridges.2,7 Microvilli increase the surface area of the corneal epithelium and might also provide less resistance to the movement of a nictitating membrane than microridges.2,8 Further, corneas of aquatic vertebrates have more densely packed corneal epithelial cells,2,7 possibly to maintain the required level of dehydration in the presence of dissolved salts and to provide strength. The large corneal epithelial cells in terrestrial vertebrates help maintain the appropriate levels of dehydration to ensure optical transparency1 and a larger surface area for the beds of microvilli; however, the specific functions of the variety of structural adaptations in the vertebrate cornea have not yet been fully elucidated.7

Interestingly, one species of teleost, the Australian lungfish (Neoceratodus forsteri) possesses microvilli7 rather than the microridges typical of aquatic vertebrates. This might be an adaptation for occasional forays into a terrestrial environment,2 suggesting that similar adaptations of the corneal microstructure might exist in other aquatic species exhibiting specialised aerial or amphibious vision.

The genus Anableps (Anablepidae, Cyprinodontiformes) is native to southern Central America and northern South America and all three species of Anableps (A. anableps, A. microlepis and A. dowei) have similar ophthalmic adaptations to their epipelagic lifestyle.9Anableps species hunt for insects and small crustaceans primarily at the waterline of brackish rivers and estuaries.10–12 Although these fish only have two eyes, each eye is capable of simultaneously seeing both above and below the waterline, allowing the fish to feed at the surface or in shallow tidal flats while ‘keeping an eye out’ (or part thereof) for predators that might approach from above or below the water's surface. The eye of Anableps sp. is divided between two optical systems, with two pupillary apertures, one retina with asymmetric hemi-retinas showing the presence of two horizontal streaks in neuron distribution and a cornea divided into two distinct surfaces by a horizontal septum11,12 (Figure 1).

Figure 1.

Close up of Anableps anableps head and eyes showing the position of the waterline and division of dorsal and ventral corneas. The thin white line indicates approximate position of the waterline when the fish is at the surface. Note that a meniscus forms over the eye such that it is always covered by a thin layer of water.

To better understand the function of microstructural adaptations of the vertebrate corneal epithelium, we compared the dorsal (aerial) and ventral (aquatic) corneal surfaces of the eye of A. anableps. Given the extraordinary level of visual adaptation of Anableps eyes to simultaneous aerial and aquatic vision, we predicted that the corneal surface structures also reflect these different visual environments.


Four male and female adult individuals of the species Anableps anableps Linnaeus 1758 (122 to 195 mm in total length) were used in the present study. The specimens were collected at Mosqueiro Island, Belém, Pará, Brazil, and following euthanasia, the head of each individual was preserved in Karnovsky's fixative (2.5 % glutaraldehyde, 2 % paraformaldehyde in 0.1 M cacodylate buffer) and donated to this study by João Paulo Coimbra. Although the osmolarity of the modified Karnovsky's fixative was not measured directly, it is thought to be approximately 500 mOsm, given the fixative is a 1 : 4 dilution of the original Karnovsky recipe.13 The fixative used is also consistent with that used previously to examine the corneal microprojections in a large range of vertebrate corneas, including teleosts.2

Whole corneas from each of the four individuals (two left eyes, two right eyes) were removed and prepared for scanning electron microscopy (SEM). Using a Biowave scientific microwave (PELCO International, Redding, CA, USA), the corneas were post-fixed in 1 % osmium tetroxide in 0.1 M cacodylate buffer (pH 7.2), followed by a water rinse and dehydration in a graded series of alcohols. Specimens were dried with hexamethyldisilazane (Proscitech, Townsville, QLD, Australia) using a graded series with hexamethyldisilazane : ethanol at ratios of 1 : 3, then 1 : 1 and 3 : 1, and then finally three rinses in full strength hexamethyldisilazane. The samples were left immersed in hexamethyldisilazane and allowed to evaporate slowly overnight. When completely dry, samples were mounted on 12.5 mm aluminium stubs using double-sided carbon tape. Corneas were cut into quarters and oriented to ensure both the endothelial and epithelial surfaces of the dorsal and ventral corneas were displayed. The mounted specimens were covered with 10 to 15 nm of platinum using an ion coater (Eiko IB-5; Eiko Engineering Co, Ibaragi, Japan). The specimens were examined using a field emission scanning electron microscope (JEOL JSM 6400 FESEM; Japan Electron Optics Ltd, Tokyo, Japan) under an accelerating voltage of either 3 or 5 kV. Images were recorded digitally and features were measured using UTHSCSA ImageTool V3.00 (http://ddsdx.uthscsa.edu/dig/itdesc.html). Measures of cell area were performed at a magnification of greater than 10,000 times. In total, the area (in mm2) of 640 epithelial cells from a total of four eyes from four different individuals were measured and converted into cell density. Higher magnification images (viewed at 100,000 times) were also taken of selected corneal regions (central-dorsal, peripheral-dorsal, central-ventral and peripheral-ventral) for each of the four eyes to determine the mean width and separation of microridges, with 30 measures of ridge width and ridge separation measured per region per fish.

Mean cell density, ridge width and ridge separation were based on the mean of all four fish. The values for each fish were based on the mean of multiple samples (see above). Differences between corneal regions, based on a sample size of four, were analysed with the statistical package R (version 2.8.1; http://www.r-project.org) using a three factor linear mixed effects model, with each individual animal used as a random effect and dorsal/ventral and central/peripheral positions used as fixed effects and compared using ANOVA. Cell density measurements were normalised by log-transformation. Results are reported as the mean and standard deviation. The value for each fish was the result of numerous samples (sample numbers below) from each region of the cornea under investigation.


The corneal epithelium in both the dorsal and ventral corneas featured cells that were predominantly shaped as irregular pentagons or hexagons (Figure 2), appearing to be more elongated in the periphery of both the dorsal and ventral corneas than in the central regions. The dorsal (aerial) corneal surface had a mean cell density of 16,387 ± 3,995 cells per mm2 in the central region of the cornea and 19,425 ± 1,890 cells per mm2 in the peripheral regions, whereas the ventral (aquatic) corneal surface had a mean cell density of 22,428 ± 6,387 cells per mm2 centrally and 26,491 ± 5,622 cells per mm2 peripherally (Table 1). Although these dorsal-ventral differences in epithelial cell densities were large, they were not significant (p = 0.160 centrally and p = 0.055 peripherally) as a result of high variance between individuals (Table 2, Figure 3). There was no significant difference (p = 0.218 dorsally and p = 0.376 ventrally) in cell densities between the central and peripheral regions of the cornea (Table 1, Figure 3).

Figure 2.

Mean cell density (cells per mm2) from central and peripheral regions of the dorsal and ventral cornea for Anableps anableps. Mean of 40 cells per region per fish. Error bars represent one standard deviation of the mean; letters over error bars show significant differences at α= 0.05.

Table 1. Mean epithelial cell density, microridge width and microridge separation for central and peripheral regions of the dorsal and ventral corneas of Anableps anableps.
 Cell densityMicroridge
mean ± SD (cells mm-2)Width mean ± SD (nm)Separation mean ± SD (nm)
Central16,387 ± 3,995147 ± 22361 ± 56
Peripheral19,425 ± 1,890130 ± 21371 ± 72
Central22,428 ± 6,387144 ± 22369 ± 62
Peripheral26,491 ± 5,622144 ± 28374 ± 66
Table 2. Mean cell density (cells per mm2) from central and peripheral regions of the dorsal and ventral corneas for individual Anableps anableps. Mean of 40 cells per region per fish.
mean ± SDmean ± SDmean ± SDmean ± SD
1 (left eye)14,885 ± 3,87319,430 ± 6,50628,146 ± 6,40527,977 ± 7,419
2 (left eye)22,191 ± 9,87416,944 ± 7,39224,176 ± 11,90730,235 ± 21,106
3 (right eye)13,076 ± 2,83521,528 ± 17,39324,113 ± 6,55329,573 ± 7,998
4 (right eye)15,396 ± 5,92519,799 ± 6,02113,275 ± 4,34718,179 ± 6,599
Figure 3.

Scanning electron micrograph showing cell borders of the dorsal corneal surface of Anableps anableps. The borders of one cell have been traced, demonstrating how the cell area and thus cell density were estimated.

Cells of both the dorsal and ventral corneas featured microprojections that fit previous descriptions of microridges,14 with long concentric projections, far longer than they were wide, forming intricate patterns within the cell border of each epithelial cell (Figure 4). Microridges were separated by a mean distance of 369 ± 9 nm, with no significant differences (p = 0.633 centrally and p = 0.912 peripherally) between the dorsal and ventral corneas and no significant differences (p = 0.689 dorsally and p = 0.758 ventrally) between the central and peripheral corneal regions (Figure 4, Table 1).

Figure 4.

Scanning electron micrographs showing the range of corneal microridge width and microridge separation in Anableps anableps. Upper images (A and B) are dorsal, lower images (C and D) are ventral, left column (A and C) is peripheral and right column (B and D) is central.

The microridges had a mean width of 142 ± 9 nm, with no significant differences (p = 0.747 centrally and p = 0.192 peripherally) between the ventral and dorsal corneas. However, while there was no significant difference (p = 0.907) between the central and peripheral regions in the ventral cornea, there was a significant difference (p = 0.025) in mean microridge width between the central and peripheral regions in the dorsal cornea (Figure 4, Table 1).


The corneal epithelia of terrestrial vertebrates generally have, as the dominant surface feature, microplicae and/or microvilli, while the corneal epithelia of aquatic vertebrates generally have microridges.2,7 Therefore, it was anticipated that there might be dramatic differences between the corneal surface projections covering the dorsal (aerial) and ventral (aquatic) hemispheres of the four-eyed fish Anableps anableps; however, we found that both the dorsal and ventral corneas of A. anableps featured microridges (Figure 3) and there were only minor differences in cell densities and dimensions of the microridges across the cornea (Table 1).

Our observations show that the microstructure of the corneal epithelium of A. anableps possesses only minor adaptations to simultaneous aerial and aquatic vision. In particular, there was a trend towards more densely packed cells on the ventral compared with the dorsal corneal surface. In general, corneas of aquatic vertebrates have more densely packed epithelial cells than terrestrial vertebrates,2,7 which is thought to be an adaptation to differences in osmotic stresses between aerial and aquatic environments.1 The difference in cell density between dorsal (17,882 ± 8,338 cells per mm2) and ventral (24,967 ± 12,457 cells per mm2) corneas of A. anableps might reflect differences in osmolarity and rigidity/strength encountered by the aerial and aquatic corneas of a single eye, a finding that warrants further investigation.

The mean density of epithelial cells across the whole corneal surface of A. anableps was 21,182 ± 3,061 cells per mm2, which was greater than the mean cell density of 11,108 ± 4,756 cells per mm2 reported for other euryhaline species.2 In A. anableps, cell density varied greatly among individuals, possibly indicating polymegathism.2 Although others have found increased cell densities in the central cornea compared with the periphery,15,16 we found no significant differences (p = 0.218 dorsally and p = 0.376 ventrally) with corneal eccentricity in this species.

We found microridges of A. anableps were invariably separated by a distance of 369 ± 9 nm throughout all regions of both corneas, possibly indicating that the distance is constrained and has some functional role that requires precise spacing. Collin and Collin2 also reported a constant separation in the blowfish, Torquigener pleurogramma, of 170 nm, suggesting similar constraints operate on other vertebrates. Due to inevitable shrinkage during processing for SEM, the exact separation of microridges in vivo might be different and difficult to estimate in the present study. Doughty17 reported a tissue contraction of 35.8 ± 1.2 per cent as a result of fixation and critical point drying of rabbit cornea. Therefore, a correction factor should be applied to the data presented to give an estimate of the in vivo size of these microprojections and the degree of shrinkage. Doughty, Bergmanson and Blocker18 also found that epithelial cell density in the freshwater rainbow trout, Oncorhynchus mykiss, was different when the tissue was fixed in a low or high osmolality solution. As the fixation procedures used on all specimens of Anableps sp. in the present study were identical, we would predict the level of shrinkage to be consistent, but some variations in ridge width and spacing, in addition to cell density, might be predicted (when compared with similar studies in marine teleosts), based on the fact that these species live in a freshwater habitat.

A. anableps, a teleost fish with a need for aerial vision, has microridges and little to no difference in cell density or microridge structure between dorsal and ventral corneas. The general lack of adaptation of the dorsal cornea, which is often out of water, might imply that the phylogenetic constraints on corneal epithelial cell structure and density are greater than the selective pressure(s) on adaptation for aerial environments in this species occupying such a unique environment. Individuals of the closely related species A. microlepis have been observed to dip their head underwater two to four times each minute, probably to keep their eyes moist and maintain a clear optical path, and they completely submerge to escape from predators approaching from above the surface.12 As the whole eye would be always covered by at least a thin layer of water, in addition to the underlying tear film, there might not be a great difference in the physiological challenges faced by the two halves of the eye and cornea, and the selective pressure for altered corneal surface microprojections is low.

Differences in the structural, optical and physiological function of corneal microprojections have not yet been fully elucidated and only a handful of vertebrate corneas have been studied among the vast diversity of extant vertebrate species. As reviewed by Collin and Collin,2 the type of microprojections found over the corneal epithelium varies across vertebrate taxa and appears to be species specific. Whether these extensions of the epithelial cells reflect differences in the affinity of the tear film to adhere to the epithelium in different environments is yet to be established for the observed range of microprojections. Based on the present study, the osmotic pressure placed on each of the two corneas in Anableps sp. appears to be different, while the transmission of light through the cornea (as indicated by the ridge spacing) might not vary. More studies need to focus on species specialised for simultaneous aquatic and aerial vision and/or species that make transitions between the two environments before we can make conclusions regarding the evolutionary plasticity of corneal microstructures.


We are indebted to João Paulo Coimbra, Dr Luciano Montag and Valéria Oliveira for the collection and taxonomic identification of specimens. We also thank the staff at the Centre for Microscopy and Microanalaysis at The University of Queensland for their technical assistance.


Dr Shelby Temple was supported by post-doctoral fellowships from The University of Queensland and the Natural Sciences and Engineering Research Council of Canada. This research was supported by an Australian Research Council grant to Professor Shaun Collin.