Cone Arrangements in Teleost Retinae

Most teleost retinae contain single and double cones and these are frequently arranged to form a regular cone mosaic. Teleost cone patterns have been observed by a number of investigators, including Hannover (1840), Beer (1894), Eigenmann and Shafer (1900), Fürst (1904), McEwan (1938), Müller (1951), Ryder (1895), and Shafer (1900), but very few of them have studied the pattern in detail. The most comprehensive study is that made by Eigenmann and Shafer who enumerated seven different patterns and also claimed that the pattern was constant for a particular species. My observations on the growth of the trout [Salmo trutta) retina have shown that in this species the pattern changes as the eye grows (Lyall, 1957). Since the teleost retina grows from the edge, the differences between the central and peripheral cone arrangements, which have sometimes been observed, may also be growth changes. The results of an examination of the retinae of various teleosts, with particular reference to the types of cones and their arrangements, are presented in this paper. The origin and possible significance of cone patterns and double cones, and the evolution of double cones, are discussed in relation to these observations.

The retinae of various teleost species have been examined. Bouin and formaldehyde have been used as fixatives, and serial sections have been prepared in the manner described in an earlier paper (Lyall, 1957).

Single cones are the basic type of photopic visual cell found throughout the vertebrates. A teleost single cone consists of a conical outer segment and a cylindrical inner segment containing an ellipsoid. The nucleus is at the base of the inner segment and the cell terminates in a tapering foot-piece which passes through the layer of rod nuclei. In trout retinae I have occasionally found an unusual type of large single cones which, from their size and position in the cone pattern, appear to each represent half a double cone. Double cones are found in many vertebrate retinae and generally consist of two dissimilar halves fused together, one half resembling a single cone whereas the other half is larger and non-migratory. Teleost double cones differ from those of other vertebrates in that both halves undergo photomechanical movements and are usually of the same shape and size. These typical teleost double cones are generally called twin cones which by definition consist of two identical halves fused along their inner segments. The two halves of a trout double cone are of equal size and these elements are usually referred to as twin cones, but I have found that the two halves often stain differently with haematoxylin or Mallory and are therefore not identical. Similar staining differences were observed by Müller (1951) in Lebistes with azan stain, and Schultze (1867) observed that in some teleosts there was a difference in the appearance of the two halves of a double cone in a fresh retina, the cytoplasm of one half being more homogeneous than that of the other. According to Müller (1954) the staining properties of the cones (of Lebistes) change during dark adaptation. The majority of my sections have been taken from fully light-adapted eyes, but the differential staining of the two halves of the double cones is still evident in dark-adapted trout retinae. Several types of double elements in which one half differs morphologically from the other, to some extent, have also been observed in teleosts (e.g. by Butcher, 1938; Verrier, 1928; Walls, 1942), and these have been termed conjugate elements or unequal twin cones. The differential staining of teleost double cones with two halves of equal size indicates that the physico-chemical make-up of the two halves is not identical, and it is therefore debatable whether such structures should be referred to as twin cones or conjugate elements. It is simpler to refer to all these double elements as double cones, irrespective of the extent by which the two halves differ. Double cones can be subdivided, on the criterion of size only, into equal and unequal double cones, to distinguish the unique type of teleost double element which has two halves of equal size from the unequal double cones of some teleosts and other vertebrates.

Triple cones have previously been observed as rare and anomalous structures ; Saxen (1953) observed a number of triple cones in tadpole [Rana temporaria) retinae and Underwood (1951) found a few in gecko (Aristelliger praesensis) retinae. I have observed more than 150 triple cones, in which the three parts are of equal size and arranged linearly, in tangential sections of trout retinae (fig. 1, A). The position of the triple cones in relation to the general cone pattern indicates that these are abnormal double cones. In minnow (Phoxinus laevis) retinae triple cones are numerous and must be a fundamental type of visual element in this species. The three parts of a minnow triple cone are arranged linearly, as in trout, but the central cone is larger than the cones on each side of it. In longitudinal section the outer segments of the lateral cones are level with the ellipsoid of the large central cone. Quadruple cones consisting of three smaller cones arranged symmetrically around a large central cone are also present in minnow retinae. These quadruple cones are fairly numerous and in a few sections they are almost as abundant as triple cones. Each part of a multiple cone has the same structure as a single cone. Fig. 1, B shows double, triple, and quadruple cones in a tangential section of a minnow retina.

The cone patterns are seen most clearly in tangential sections of the retina cut through the cone inner segments, but they can usually still be traced at the level of the cone nuclei. One of the most common patterns is that found in an adult trout retina (pattern I) (figs. 1, c; 2). Each pattern unit consists of four double cones surrounding a single cone ; the double cones are arranged in two pairs so that the cones of each pair are parallel to each other and at right angles to the other pair. I have found a slightly different pattern (pattern II) in young trout retinae in which there is an additional single cone at each corner of a pattern I unit. Pattern II is shown in figs. 1, A and 3. The additional single cones are slightly shorter (33% or less) than the central single cones. The double cones are the longest and, contrary to Müller’s (1951) observations in Lebistes, each half of a double cone has a greater diameter than either type of single cone. A regular alternation of double and single cones is also evident in longitudinal sections when the plane of sectioning coincides with a row of cones, and the single cones can be identified as central single cones or additional single cones according to the plane in which the double cones are sectioned (see fig. 1, D, E). The pattern in trout (which is the one studied in greatest detail) is not regular over the whole retina, for the direction of the lines of cones sometimes changes and there are also irregularities caused by the addition and termination of cone rows. Occasionally a double cone is represented by a large single cone or by a triple cone (fig. 4), but the central single cones are always present.

Some teleost genera in which cone patterns have been observed are shown in table 1 with their respective patterns. In perch (Perea fluviatilis) and miller’s thumb (Cottus gobio) the single and double cones are arranged as in adult trout (fig. 2), but in the former species I have observed some variation in the size of the cones in certain regions. I have observed both patterns I and II in salmon (Salmo salar) retinae. The cone pattern in grayling (Thymallus vulgaris) is similar to that in adult trout. The cone arrangements in char (Salve-linus willughbii) and gwyniad (Coregonus pennantii) are rather irregular ; the cone elements are chiefly double cones which are arranged nearly parallel to each other over much of the retina. In some regions pattern I is visible, but the single cones are small and cannot always be distinguished in sections cut through the inner segments of the double cones. The pattern in pike (Esox lucius) is unusual in that the pattern units are triangular, instead of the more common rectangular units of patterns I and II. There is some difference of opinion on the patterns found in certain species, e.g. that figured by Eigenmann and Shafer (1900) for Scorpaena porous differs from that given by Beer (1894). Müller (1952) describes a cone pattern in Phoxinus laevis whereas I have found an irregular arrangement of single, double, triple, and quadruple cones in this species. In central regions of minnow retinae I have only observed single cones in tangential sections cut near the nuclear layer ; these cones are shorter than the other cones and also have a greater diameter than the constituent elements of the multiple cones. I have always found that cone patterns are associated with double cones in which the two halves are of equal size, although the retina of carp (Cyprinus carpió) has no pattern despite the presence of equal double cones. There is no pattern in the retinae of roach [Rutilus rutilus) and rudd [Scardinius erythrophthalmus) which have unequal double cones.

There is a regular arrangement of the differently stained halves of trout double cones, so that within the anatomical mosaic of single and double cones there is a staining pattern (figs, i, c; 2). With haematoxylin, one half of a double cone stains darkly while the other half is eosinophil ; a similar staining pattern is observed with Mallory. The double cones are arranged so that two light and two dark halves face each other alternately along each row and where the rows intersect two dark halves traverse two light ones. This regular staining pattern is not usually visible in young trout retinae and not in all the retinae of adult fish. The staining pattern in trout is the same as that found by Müller (1951) in Lebistes.

(a) Changes in the central region of the retina. Eigenmann and Shafer (1900) claimed that the cone arrangement in the retinae of teleosts was constant for a particular species and Müller (1951) confirmed this when he found the same pattern in the retinae of Lebistes of different ages. In trout and salmon, however, the pattern differs in young and adult fish, and many of the single cones present in the young trout retinae are absent from the cone pattern in adult trout. The loss of the additional single cones in trout, which occurs during the change-over from pattern II to pattern I, takes place gradually, so that in some regions a few pattern II units are scattered among those of pattern I. The change to pattern I involves the transformation of a retina in which the single cones and double cones were approximately equal in number to one in which the ratio of single to double cones is approximately 1:2. The loss of the single cones begins near the pole of the eye and progresses towards the periphery, and thus follows the course of earlier differentiation and development. The change-over occurs in fish 1-2 years old, so the loss of the additional single cones cannot be compared with the waves of degeneration which Glucksmann (1940) found in the retinae of frogs during differentiation. There is no evidence of the degeneration of the additional single cones and it seems likely that they are transmuted into rods (see p. 197). Fürst (1904), who observed the change of pattern in salmon, suggested that the missing cones might be found among the rods in the adult retinae.

(b) Changes between the periphery and the centre of the retina. I have examined the peripheral cone arrangements in trout, minnow, and pike retinae and, et) although they have very different cone arrangements in the centre of the retina, there is a similar basic arrangement at the periphery. The retina grows from the periphery and the peripheral region at any particular time becomes more central as the eye grows, so that the peripheral cone pattern must change into the characteristic central cone arrangement. The derivation of the more central arrangement from that found at the edge can be traced in tangential sections of the retina. The basic cone pattern at the periphery consists of parallel rows of double cones arranged with their long diameters parallel to the edge. One consequence of the parallel arrangement of double cones at the edge of the retina is that in longitudinal sections they are always cu at right angles to their long diameter and thus appear single, which gives rise to the view that double cones are absent at the edge of the retina (Verrier, 1928). A parallel arrangement of double cones was observed by Müller (1952) at the nasal edge of the retina in Lebistes, and Shafer (1900) describes a nearly parallel arrangement of double cones at the edge of the retina of Micropterus, but these were arranged at right angles to the edge.

The parallel arrangement of double cones found at the periphery of trout retinae gives rise to the square pattern of more central regions by a series of positional changes which are shown diagrammatically in fig. 5, A-E. At the extreme edge, the double cones are arranged so that in each row parallel to the edge the orientation of the two halves of each double cone is the same, but is opposite to that in the rows on each side (figs. 5, A and 6, A). The single cones are arranged in rows between the double cones and are not always visible in the outer-most sections. Every alternate single cone is smaller than the others and these represent the additional single cones of pattern II. If pattern I were formed directly from the parallel arrangement of double cones, the single cones would only be present at alternate intervals between the double cones. I have found this arrangement at the edge of a perch retina. The typical central trout cone pattern is derived from the parallel double cone arrangement by movement of the double cones so that each row becomes zigzag (fig. 5, c, D) till finally two adjacent double cones in the original rows are at right angles to each other (fig. 5, D; pattern II). The additional single cones are lost during later growth to give pattern I (fig. 5, E) which is characteristic of an adult trout retina.

The minnow retina is characterized by the numerous triple cones which it contains in addition to quadruple, double, and single cones. There is no regular cone pattern except at the edges, where only double and single cones are present and the double cones are arranged in parallel rows. It seems reasonable to assume that the minnow retina grows from the edge, like that of trout, and therefore the regular pattern of double and single cones will develop into the irregular arrangement of triple and quadruple cones characteristic of more central regions. A plan whereby this transformation may take place is shown in fig. 7. The arrangement of the double cones in rows parallel to the edge of the retina appears similar to that found at the edge of trout retinae, but the orientation of the two halves of the double cones is different. In each row parallel to the edge, the two halves of succeeding double cones are alternately arranged so that two similar halves (two light or two dark) of adjacent cones face each other (fig. 7, A). Adjacent rows have the opposite arrangement so that two dark halves face each other next to two light halves in adjoining rows. Considering the rows of double cones running at right angles to the edge, the orientation of the two halves alternates in succeeding double cones, as is also found in trout (compare figs. 5, A and 7, A). ROWS of darkly stained single cones are present between the rows of double cones (figs. 6, B; 7, B). Alternate single cones are larger than the others, as in trout, and in some sections only these larger cones are visible.

The triple cones near the edge of the retina are usually still arranged in rows and often alternate with double cones. It appears from this arrangement of cones in minnow retinae that triple cones are formed by the fusion of a double cone with a single cone. If the large single cones were the first to fuse with the double cones this would give the alternate triple and double cone arrangement which is sometimes observed (fig. 7, c). As the other single cones grow larger they will also fuse with the remaining double cones so that at this stage the original pattern is lost, and the triple cones become irregularly arranged (fig. 7, D). A quadruple cone is probably formed by the fusion of another later developing single cone with a triple cone (fig. 7, E) ; some may also be formed by irregularities in the fusion with double cones so that some single cones fuse with the triple cones instead of with double cones.

The typical pattern in pike is formed of triangular units, but it is also derived from parallel rows of cones at the edge of the retina. The position of the single cones in relation to the double cones differs from that in trout and minnow, the most common arrangement being that shown in fig. 8, A. The change of pattern can be traced and is represented in fig. 8, A-D. The single cones are usually situated in the double-cone rows at regular intervals and the double cones lying between two single cones of adjacent rows are reorientated to lie at right angles to the doublecone rows (fig. 8, B). The positions of the other double cones change, adjacent cones turning in opposite directions to form the characteristic pattern (fig. 8, D). NO difference in staining properties between the two halves of a double cone was observed in pike.

I have always found the parallel arrangement of double cones at the edges of minnow retinae but not always in trout; this may be due to the fusion of cones in minnow in addition to reorientation and the different times taken in the two species to complete the changes. If the change of pattern in trout is rapid, the change-over zone will be narrow and more difficult to locate. It is also possible that reorientation in trout may sometimes occur at the nuclear stage, before differentiation.

The change of pattern found in trout retinae involves the loss of many single cones; and the transmutation of these cones into rods seems the most probable explanation of their disappearance, for there is no evidence of degeneration, which appears to be the only alternative solution. There is evidence of the transmutation of one type of visual cell into the other in several vertebrates, e.g. the rods of certain geckos resemble cones in structure but contain rhodopsin (Underwood, 1954). Various criteria have been used to distinguish between rods and cones, but, although the rods of one species may resemble the cones of another, within one species there is usually a clear distinction between rods and cones. Apart from differences in the size and shape of the inner segments, the rods and cones in a trout retina have clearly distinguishable nuclei which differ in shape, size, and staining properties. The cone nuclei are large and rather elongated and do not stain readily, whereas the rod nuclei are smaller, more spherical, and stain darkly. They also differ in their position in relation to the external limiting membrane, the rod nuclei all being situated wholly on the inner side of the membrane whereas the cone nuclei protrude through it so that generally two-thirds of each nucleus is on the pigment epithelium side of the membrane. I have found a few visual elements in trout retinae which may be intermediate stages in the transition from cone to rod (fig. 6, c). The nuclei of these elements stain like rod nuclei with haematoxylin or Mallory, and are similar in shape. The position of these nuclei in relation to the external limiting membrane is also rod-like, since they are generally situated entirely within the membrane and always with at least two-thirds of the nucleus on the inner side. A small, lightly stained myoid separates the nucleus from the cylindrical inner segment, which is narrower and shorter than that of a typical cone. A significant feature about these elements is that they are seen in longitudinal sections between two double cones cut through their long diameters, which is the same position as that of the additional single cones in younger retinae (compare figs. 6, c; 1, D). The retinae in which these elements were found belong to fish within the change-over size range, and it seems reasonable to assume that they are intermediate stages in the transmutation of the additional single cones to rods.

The teleost retina grows from a peripheral growth zone and it is in this region that the formation of the pattern is most likely to be seen. Müller (1952) devised a scheme for the development of the cone pattern in Lebistes (which is the same as that in young trout) based on the planes of mitoses in the growth zone. Müller examined the mitoses in surface sections and assumed that all the mitoses were concerned in the formation of cone nuclei, but, as mitoses are almost entirely confined to the outermost layer of nuclei in the growth zone, some of the products of division must develop into the other types of retinal cells which will eventually lie internal to the cones. I have examined the planes of mitoses in longitudinal sections of trout retinae and my results are in general agreement with those obtained by Glucksmann (1940) from tadpole retinae, in which there is no pattern. I found 76% of the mitoses parallel to the edge, 6% at right angles to it, and 18% intermediate at approximately 45⁰ to the edge. Another difficulty with Müller’s theory is that the pattern at the extreme edge of the retina may differ from that in more central regions, in which case the mitotic planes of the dividing cells in the growth zone cannot determine the final arrangement, for the peripheral region at any particular time becomes more central as the eye grows. It thus seems doubtful if any interpretation of the mitotic axes can explain the formation of the characteristic patterns. A parallel arrangement of double cones is probably the basic pattern from which all others can be derived by positional changes and the addition of single cones, but the origin of this basic cone arrangement remains obscure.

No functional significance has been attributed to cone patterns in teleost retinae, but their frequent occurrence suggests that they may affect some aspect of visual perception. It is possible that cone patterns improve the perception of movement, since they are generally found in species which feed on fastmoving objects. Bateson (1889) has shown that the majority of teleosts feed by sight, but some are more sensitive to movement than others. Movement perception is generally of greater importance to predatory fish than high visual acuity. A cone pattern provides a uniform distribution of both types of cone cells and this may be important if the single and double cones have different functions.

The relative distribution of single and double cones in teleost species suggests that double cones are associated with vision in deep water, although Wunder (1925), who examined 24 fresh-water species, found they were most numerous in surface fish. In the Salmonidae, double cones are relatively more numerous in deep-water forms than in species living in shallower water, e.g. char and gwyniad, which live in deeper water than trout and grayling, have fewer single cones than the latter two species, and the loss of the additional single cones in trout and salmon may also be associated with their migration into deeper water. Walls (1942) states that double cones alone occur in some Gadus species, and I have found that there are almost exclusively double cones in the retinae of cod (Gadus morhua) and whiting (G. merlangus), both species living at considerably depths. The association of double cones with vision in deep water may be due to greater sensitivity of the double cones so that they are intermediate in sensitivity between single cones and rods, as Willmer (1953) has suggested, or to differences in the spectral sensitivities of single and double cones, since different wavelengths of light penetrate water to different depths.

There are two theories on the origin of double cones in vertebrates. The most widely held view (Bernard, 1900; Cameron, 1905; Detwiler and Laurens, 1921; Eigenmann and Shafer, 1900; Müller, 1952; Saxen, 1954) asserts that a double cone is formed by the fusion of two adjacent cells. The alternative theory, held by Dobrowolsky (1871), Howard (1908), and Franz (1913), maintains that a double cone is the result of incomplete division. It seems reasonable to assume that all multiple cones are formed in the same way, thus triple and quadruple cones will be formed in the same manner as double cones, and the abundance of triple and quadruple cones in minnow retinae supports the theory that they are formed by fusion. As Saxen (1954) points out, two successive incomplete divisions would be required to form a triple cone, therefore in minnow this would have to be a normal method of division. It is more probable that triple and quadruple cones are formed by the fusion of cone elements, and the changes in the cone arrangement observed between the peripheral and central regions of the minnow retina seem to support this view.

On the assumption that the peripheral arrangement of cones in minnow retinae develops into that of more central regions, there is evidence that double cones of equal size develop into unequal double cones. The double cones found at the edge have two halves of equal size with different staining properties, but during the formation of triple and quadruple cones one half of the original double cone must increase in size disproportionately to form the central cone. Most of the double cones present in the central regions of the retina also have one half larger than the other. Walls (1942) believes that as teleosts are a terminal group in evolution, non-teleost double cones cannot have evolved from teleost (equal) double cones, and the presence in teleosts of certain double elements with dissimilar halves may indicate incomplete equalization of the two halves. If, however, equal double cones had evolved by equalization from unequal double cones, it is unlikely that in the teleosts the latter would initially appear as equal double cones and only become unequal during later growth. The occurrence of the majority of unequal double cones among some of the most primitive families has been noted by Walls in discussing the evolution of teleost double cones, but it seems injudicious to consider the distribution of these elements in relation to the evolution of double cones when related genera may have different types of double cones. The distribution of equal and unequal double cones in teleosts cannot be correlated with the phylogeny of the fish : among the cyprinodonts Fundulus has unequal double cones (Butcher, 1938) but Lebistes has equal double cones (Müller, 1951), and similar differences are found in the Cyprinidae. The simplest theory of the evolution of double cones is that unequal double cones have evolved from equal double cones by the two halves becoming increasingly dissimilar, and the differential staining of some teleost double cones may represent the first stage in this evolution.

I wish to express my sincere thanks to Professor R. J. Pumphrey, F.R.S., for his helpful advice and criticism. I am also indebted to Dr. J. W. Jones for providing the fish and to Mr. W. Irvine for taking the photographs. This work was carried out during the tenure of a D.S.I.R. maintenance grant.