Quantitative Studies of Retinal Ganglion Cells in a Turtle, Pseudemys scripta elegans I. NUMBER AND DISTRIBUTION OF GANGLION CELLS ELLENGENE H. PETERSON AND P. S. ULINSKI Department of Anatomy, University of Chicago, 1025 East 57th Street, Chicago, I1linois 60637
ABSTRACT Multiple pathways for the transmission of visual information from retina to brain have been described in reptiles, but little is known about their functional organization. These parallel channels begin a t the retina, and we have therefore begun to study the functional organization of retinal ganglion cells in the turtle, Pseudemys scripta elegans. This paper describes the numbers and distribution of cells in the ganglion cell layer. To develop criteria for the identification of ganglion cells, we labelled them retrogradely by applying horseradish peroxidase (HRP) to the optic nerve. Ganglion cells were found to vary substantially in size and cytology. In low density areas of the retina, ganglion cells typically have cytoplasm with well developed Nissl substance, a distinct, pale nucleus, and a large nucleolus. In high density areas of retina, ganglion cells are small, densely staining, and gliaform. The average minimum proportion of ganglion cells in the ganglion cell layer i s 7580%of total profiles. No more than five or six percent of profiles in the ganglion cell layer are neurons which do not send an axon into the optic nerve (displaced amacrine cells or intraretinal association cells). The ganglion cell layer of P. s. elegans can be divided into a number of regions on the basis of cell density. Isodensity maps constructed from Nisslstained, wholemounted retinas indicate that there is an elongated region of high ganglion cell density, the visual streak, which extends from nasal t o temporal retina and is oriented such that its long axis follows the horizontal axis of the eye. The streak is aligned with the externally visible iris line. Seen in crosssection, the ganglion cell layer in the streak is three to four cells thick; in nonstreak retina, ganglion cells form only a monolayer of somas. Ganglion cell density drops off more rapidly above the streak than below it. The temporal arm of the streak is both shorter and broader than the nasal arm. There is a peak in ganglion cell density at the midpoint of the streak, in the approximate center of the retina. Here, ganglion cell densities exceed 20,000 cells mm-2. The total number of ganglion cells in the retina is 350,000-390,000. The retina in reptiles projects to at least six targets within the central nervous system: to one or more loci within the hypothalamus, dorsal thalamus, ventral thalamus, pretectum, tectum, and midbrain tegmentum (Butler, '79; Cruce and Cruce, '78; Ebbesson, '72). Each of these retinorecipient zones makes subsequent efferent connections (Northcutt, '78; Ulinski, '77, '79). Thus the visual system of reptiles, as of other vertebrates, consists of multiple, parallel pathways for the transmisJ. COMP. NEUR. (1979) 186: 17-42.
sion of visual information. We would like to know whether these parallel channels process different categories of visual information and, if so, how this is effected. Accordingly, we have begun a series of experiments designed to explore the functional architecture of the visual brain in a n emydid turtle, Pseudemys scripta elegans. We have chosen this species because more information is available on its retinal physiology (Cervetto, '76; Fuortes, '76; Granda and Dvorak, '771, central visual path-
17
18
ELLENGENE H. PETERSON AND P. S. ULINSKI
ways (Butler, '79; Northcutt, '78; Ulinski, '791, visual discrimination (Burghardt, '77; Peterson, '79), and psychophysical performance (Granda and Dvorak, '77) than for any other reptile. The first experiments have characterized the intrinsic organization and connections of each structure in the tectofugal pathway, both qualitatively and quantitatively (Balaban, '78a,b; Peterson, '77, '78a,b; Rainey, '78, '79; Schechter and Ulinski, in preparation; Ulinski, '78, '79). In this paper we begin an analysis of the neural retina in Pseudemys with an account of ganglion cell numbers and distribution in Nissl-stained, wholemounted retinas. A problem with such studies has been the difficulty in distinguishing between neurons and neuroglia in the ganglion cell layer (Hughes, '75a; Stone, '65). Therefore, we first developed criteria for the identification of ganglion cells in Pseudemys by labelling them retrogradely with horseradish peroxidase (HRP) applied to the optic nerve. We then used these criteria to map ganglion cell distribution. The principal finding of this study is that the ganglion cell layer in Pseudemys exhibits a number of regional specializations, including a horizontall y aligned visual streak and an area of peak density in the center of the retina. MATERIALS AND METHODS
Eye dimensions and retinal preparations Turtles were sacrificed by transcardial perfusion of 0.9%saline followed by 10%formolsaline. The eyes were removed and stored in formol-saline for four to five days. Eyes from five turtles were measured under a dissecting microscope with a calibrated eyepiece reticle. Horizontal, vertical, and axial bulb diameters and pupil diameter were recorded. To study retinas, eyes were opened under saline by an incision just distal to the ora serrata and the anterior portion of the bulb was discarded. Retinas were then teased free from the posterior eyecup and cleaned of any remaining pigment epithelium. A series of radial cuts was made in each retina so that it could be flattened and mounted, fiber layer up, on a gelatinized slide. The retinas were air dried for 24 hours, and then defatted by passing them through a graded series of alcohols and xylene. A final overnight wash in xylene markedly improved the subsequent staining. After defatting, the retinas were rehydrated and then stained with 0.1%unbuffered cresyl violet a t 40°C for 30 minutes, differentiated in
acid alcohol plus one wash of a 1:6:1 mixture of ether: chloroform: 100%ethyl alcohol, dehydrated, cleared, and coverslipped. Two sets of measurements were taken on these retinal wholemounts. To measure shrinkage, two fixed retinas were mounted on a slide and measured before air drying. Five measurements were taken between pairs of identifiable landmarks on each retina. Each measurement was repeated after wholemount processing, and the average percent shrinkage was calculated. To compute retinal areas, processed retinas were placed in a microprojector and the enlarged image was traced onto graph paper, cut out and weighed. In addition, a known area was comparably enlarged onto the same graph paper, weighed, and used to calculate retinal areas. Two eyes were embedded in paraffin, and two were embedded in celloidin. Serial sections were cut a t 10 p and 30 p , respectively, and stained with cresyl violet. Two of these eyes were cut in a plane parallel to the pigmented iris line (fig. 1).The two remaining eyes were cut orthogonal to the iris line and parallel t o a plane dividing the eye into nasal and temporal halves. Horseradish peroxidase preparations
To distinguish between neuroglia and retinal ganglion cells, horseradish peroxidase (HRP) was applied t o the optic nerve in four turtles, and retinal wholemounts from these animals were examined. In this procedure, the olfactory bulbs were removed to expose the underlying optic nerves of anesthetised turtles (Brevital Sodium, 9-15 mgkg; Wang et al., '77). A transverse incision was made in one optic nerve of each animal, and the resulting "pocket" was filled with crystals of HRP (Sigma, Type VI). After three to six days, both eyes were excised under deep anesthesia. The fresh retinas were removed in a bath of 0.15 M phosphate buffer, fixed in 1%glutaraldehyde in phosphate buffer for ten minutes, rinsed in fresh buffer, and reacted with diaminobenzidine. The retinas were then wholemounted and processed as described above. The contralateral retina served as a control for endogenous peroxidase. The retinas were examined with oil immersion objectives at 1,250 x. Labelling was maximal in two turtles with survival times of five and six days, and only these two cases were analyzed. Since the optic nerve was not uniformly saturated with HRP, labelled ganglion cells were distributed in
19
TURTLE GANGLION CELLS
Fig. 1 Pseudemys scriptu eleguns. The photograph shows a freely swimming turtle. Note the black iris line. The turtle's head is extended and raised, but compensatory eye rotation maintains the black iris line approximately at horizontal (water line).
patches over t h e wholemounted retinas. Heavily labelled areas were sampled as follows. All of the profiles within a field were drawn with a camera lucida and marked as labelled or unlabelled. Unlabelled cells were designated as either presumed neurons or neuroglia according to criteria outlined below (RESULTS). Four samples of approximately 100 cells each were taken; two were taken within the streak, and two within the peripheral retina. For comparison, four samples were taken from normal wholemounted retinas, a t approximately comparable loci, and the proportion of neuroglia in each sample was estimated.
Distribution and absolute number of ganglion cells Four retinal wholemounts were examined a t 1,000 x . A 100 p x 100 p grid was advanced systematically over the surface of the retina in equal increments of either 200 p (high resolution sample; 1 retina) or 500 p (low resolution sample; 3 retinas). Profiles falling within the grid were counted and identified as either (presumed) neurons or neuro-
glia. Of cells which intersected the borders of the grid, only those touching the upper and right edges were included. In addition, a high resolution sample (i.e., one 100 p x 100 p grid per 200 p x 200 p area) was taken over approximately the central 2 mm X 2 mm of one low resolution retina (fig. 7) since the density gradients are relatively steep in this area. Because ganglion cell distribution in all four retinas appeared virtually identical, i t was not necessary to taken high resolution samples from the two remaining retinas. No correction was made for shrinkage in calculating ganglion cell density. To estimate the total number of ganglion cells in each retina, the number of ganglion cells counted was totalled and multiplied by either 4 (high resolution retina) or 25 (low resolution retinas). RESULTS
Pseudemys scripta,' the pond slider, is a I On the basis of cranial characteristics (McDowell, '64), osteology (Weaver and Rose, '671, and soft anatomy (Zug, '66),several authors have suggested that the genus Pseudemys be reclassified and subsumed under the Chrysemys group. Thue P. 8. elegans would become Chrysemys scripto elegons. This usage has been adopted in Some recent communications (e.g.,Ernst and Barbour, '72).
20
ELLENGENE H. PETERSON AND P. S. ULINSKI
cryptodiran turtle of the family Emydidae (Testudinidae). The subspecies P.s. elegans is distinguished by a red postorbital patch from which it derives its common name, the redeared slider. P. scrzpta is a New World, freshwater turtle which is omnivorous and primarily diurnal. It is markedly opportunistic, both as to habitat and to food (Boyer, '65; Cagle, '50; Cahn, '37; Moll and Legler, '71). Although this paper focuses on the ganglion cell layer of the retina, it is useful first t o outline some general features of the eye and retina. Morphology of the eye and macroscopic features of the retina The dimensions of the eyeball are given for ten eyes in table 1.The horizontal diameter is slightly greater than the vertical diameter. Both are greater than the axial diameter, indicating that the bulb is flattened from front to back. Similar eye dimensions are reported by Bowling ('76). The pupil is circular (fig. 1) and, in our specimens, approximately 1.74 mm in diameter (table 1). Granda and Dvorak ('77) and Bowling ('76) report roughly comparable diameters. The iris is green except for a black iris line which runs across it horizontally, or approximately so (fig. 1). Brown ('69) has reported that this iris line is maintained a t horizontal by non-visual reflexes of the head and neck (DISCUSSION). Retinal dimensions are given in table 1. The mean horizontal diameter is somewhat greater than the vertical diameter. The average area of these retinas is 87.5 mm2.The four wholemounted retinas, taken from somewhat larger turtles, average 116.5 mm2 (table 3). Granda and Haden ('70) report a retinal area of 138 mm2for P. scripta, but they did not report the sizes of the turtles from which the retinas were taken so that i t is difficult to compare their data with ours. Shrinkage always occurred during processing, but it averaged no more than two percent. As noted by other authors (e.g., Hughes, '75a), shrinkage was greatest at the edges of the retina and along radial cuts or tears. At such points, ganglion cell counts were often elevated, and such samples were discarded in constructing maps of ganglion cell density. To calculate visual angle, we used dimensions for the schematic eye of P. s. elegans given by Granda and Dvorak ('77). The distance between their average nodal point and the junction between retina and pigment epithelium is 4.92 mm. From this we estimate a
retinal magnification factor, near the center of the retina, of approximately 98 p per degree of visual field, or 11.5'-12.0" of visual field per millimeter of retina. This latter figure is reasonably close to a figure of approximately 13.0' of visual field per millimeter of retina derived for the high resolution retina (M28-R; figs. 6 , 8 , 9 ) by assuming that the entire retina (14.0 mm vertical diameter) sees 180' of visual space. To the unaided eye, the most striking feature of the vitreal surface is a prominent line which extends horizontally across the retina. It appears white in fresh preparations and as a dark blue streak in stained wholemounts (fig. 2A). Brown ('69) reported that a similar streak is visible a t the photoreceptor level as an increased density of receptors with colored oil droplets. One goal of this paper was to determine the relationship between ganglion cell distribution and this streak. However, it was first necessary to develop criteria by which ganglion cells can be distinguished from neuroglia and from other neurons lying in the ganglion cell layer. Identification of ganglion cells To determine the cytological characteristics of ganglion cells in Pseudemys retina, we used two cases in which HRP was applied to the optic nerve. Following five or six days survival, these retinas were wholemounted and the ganglion cell layer examined for the presence of HRP positive profiles. No evidence of peroxidase granules was ever observed in the two retinas contralateral to the HRP application. The ipsilateral retinas, in contrast, were heavily labelled in some areas, and the fiber layer was so deeply stained that it obscured much of the ganglion cell layer. These HRP cases revealed t h a t ganglion cell size and cytology varies with ganglion cell density. In peripheral retina where cell density is relatively low (figs. 3A,B), labelled cells resembled ganFig. 2 Visual streak. A Retina M28-L. The wholemounted retina has been stained with cresyl violet, and the visual streak appears as a dark blue line extending horizontally from nasal to temporal periphery. Note the expansion of the streak at the center of the retina which corresponds to the region of peak density. The optic nerve exits from the temporal-ventral quadrant (lower right). B High power photograph of the dorsal border of the visual streak, taken from a point approximately midway between the peak density area and the nasal edge of the retina. Note the sharp increase in cell density in the streak and that cells in the streak are relatively small, densely staining, and gliaform.
TURTLE GANGLION CELLS
21
N
22
ELLENGENE H. PETERSON AND P. S. ULINSKI TABLE 1
Dimensions of
eye
and retina in Pseudemys scripta elegans
Animal Eye number
Weight (g)
110 R L 112 R
Carapace length (cm)
280
13.5
250
13.0
L 116 R L 119 R L 120 R L
470
15.5
1,050
21.5
1,000
19.5
X SD
610 388.52
16.6 3.75
-
Bulb diameters (mm) Horizontal
7.2 7.4 7.0 6.9 8.4 8.5 8.7 8.6 9.4 8.9 8.1 0.89
Vertical
7.2 7.2 6.6 6.6 8.2 8.1 8.2 8.3 8.6 8.4 7.7 0.76
Axial
Retinal diameters (mm)
'
5.6 6.0 5.4 5.6 7.0 6.6 6.8 6.9 7.2 7.0 6.4 0.69
Pupil diameter
1.6 1.6 1.6 1.5 2.0 2.0 1.6 1.7 1.9 1.9 1.74 0.19
Retinal area (mm ?)
Horizontal
Vertical
12.5
12.0
77.3
11.5
11.2
67.3
12.5
12.5
92.1
13.0
12.8
90.6
14.0
13.8
110.4
12.7 1,908
12.46 0.963
87.53 16.35
' Axial diameter measured for posterior pole of bulb to plane of limbus.
glion cells as described in detail by several authors (e.g., Stone, '65, '78; Hughes, '75a; Tiao and Blakemore, '76). They have a pale nucleus and a large, darkly staining nucleolus (figs. 3, 4). The amount of cytoplasm and the development of Nissl substance is variable, being relatively greater in the largest cells. In smaller neurons, the cytoplasm is often rimmed with Nissl substance. Such ganglion cells are characteristically irregular in outline and assume a violet color in Nissl preparations. In contrast, within high density areas, labelled ganglion cells are very small and so densely stained that often no nucleus is visible. These cells tended to resemble glia, and thus our HRP cases were particularly useful in helping us determine the subtle cytological differences between neurons and glia within the high density retina. Here, as in the retinal periphery, ganglion cells can be distinguished primarily by their relatively irregular outline and violet stain. Our HRP material also provides a lower limit for the proportion of ganglion cells in the ganglion cell layer. Seventy-five to eighty percent of profiles in the ganglion cell layer were HRP positive and thus are conventional ganglion cells in the sense that their axons exit the retina in the optic nerve (table 2). Conversely, 20-25% of total profiles were unlabelled. Most of these cells (70430% of unlabelled profiles) were judged to be neuroglia because they resembled published accounts of glia in reptiles (Stensaas and Stensaas, '68;
Kruger and Maxwell, '67) or profiles seen in the fiber layer near the optic disk (fig. 5 ) . Such presumptive glia are generally smaller than neurons, a t least in low density areas. Their nuclei may be dark and uniformly stained or pale, often with clumped chromatin. Cytoplasm, when visible, is pale and filmy and may be eccentric. Neuroglia tend t o have a smooth outline in Nissl preparations and a relatively blue color. An average of 16%of total profiles in both labelled and normal wholemounts were judged to be neuroglia according to these criteria (table 2). We found no consistent evidence that the proportion of neuroglia increases with eccentricity. However, our observations were limited to areas of retina within two millimeters of the streak, and the proportion of neuroglia may be relatively greater in the far periphery. Possible identification of Fig. 3 Retinal ganglion cells labelled with horseradish peroxidase (HRP) granules following application of HRP to the ipsilateral optic nerve. Five days survival. A Heavily labelled area approximately two millimeters above the visual streak. Three profiles in this field are unlabelled. One, incompletely shown in upper right comer (A), is clearly a neuron because of its size. A second unlabelled cell (arrow) is probably a neuron because it is very similar to a labelled profile nearby (dark asterisk). The third unlabelled cell, slightly out of focus, ia probably a glial cell (white asterisk; cf. fig. 5F). B A less heavily labelled retinal area. T w o neurons are unlabelled and one probable glial cell (asterisk). The glial cell resembles that shown in figure 51. Scale as in A. C A heavily labelled area of the dorsal periphery of the streak. One neuron (arrow) and one probable glial cell (asterisk; cf. fig. 5H) are unlabelled.. Scale as in A.
TURTLE GANGLION CELLS
Figure 3
23
24
ELLENGENE H. PETERSON AND P. S. ULINSKI
Fig. 4 Varieties of ganglion cells seen in Pseudemys scripta elegans. A,B,G These large cells occur infrequently. They seldom reach 20 I./ in (average long and short) diameter, and are almost never seen in the streak (cf. fig. 2B). Next to the large, multipolar ganglion cell in B, is a probable glial cell (asterisk; cf. fig. 5H). C Two medium ganglion cells seen frequently in Pseudemys retina. D-F,H Medium ganglion cells typical of non-streak retina, but also seen occasionally in the streak. I Ganglion cells in the visual streak. All profiles in this picture are probably ganglion cells.
25
TURTLE GANGLION CELLS TABLE 2
Composition of the ganglion cell layer Visual streak
'
-
X
SD
91 80 20 16 4
99.50 78.00 22.00 16.00 6.25
11.03 8.29 8.29 4.55 3.86
86 19
98.30 15.50
11.32 4.36
Periphery
HRP cases Total profiles % labelled % unlabelled % glia % unidentified
110 66 34 22 12
89 81 19 15 5
Total profiles % glia
111 10
104 14
108 85 15 11 4
Normal cases 92 19
' T w o samples, taken from dorsal border of streak (8,000-10,000 cells -mm2). 'Two samples, taken approximately 2 mm above streak samples (4,000 cells -mm') 3All percent figures refer to percent of total profiles in sample.
the remaining unlabelled profiles in the ganglion cell layer is considered in the DISCUSSION.
Ganglion cell distribution Having formulated criteria for the identification of ganglion cells and neuroglia, we next assessed ganglion cell distribution in four retinas by systematically sampling ganglion cell density over the retinal surface in increments of either 200 p (1high-resolution retina) or 500 p (3 low-resolution retinas). From these data we constructed isodensity maps for the four retinas (figs. 7,8; table 3). In addition, we plotted ganglion cell density in the high resolution case as a series of equally spaced transects (both horizontal and vertical) through the retina (figs. 8,9).From these data it appears that the retina of Pseudemys can be divided into a number of regions on the basis of ganglion cell density. The remainder of this section deals with these areal variations. The most striking feature of ganglion cell distribution is the presence of a horizontally aligned area of high ganglion cell density (figs. 2, 3 0 . We call this region the visual streak by analogy to a similar ganglion cell configuration in a number of other vertebrate species (DISCUSSION). The streak lies slightly less than two millimeters above the optic disk, just above a line dividing the retina into dorsal and ventral halves (figs. 2A, 6, 7). The shape and disposition of the streak are virtually identical in all four retinas. The streak is vertically asymmetric in that the ganglion cell density gradient is more pronounced above the streak than below it. This can be seen in the isodensity maps (figs. 6, 7)
and in figure 8 which shows ganglion cell density in the high-resolution retina (M28-R) plotted as a function of vertical distance from the streak. Above the streak, ganglion cell counts drop from peak values of more than 20,000 cells mm-2 t o less than 10,000 cells mm-2within a linear distance of 400 @ or 4.7" of visual space (figs. 6, 8). Below the streak, a comparable drop in ganglion cell density occurs 800 p or 9.4" from the streak center. In a second retina (HRPS-R; fig. 71, ganglion cell density drops from peak values of more than 18,000 cells mm-2to approximately 7,000 cells mm-2 a t 400 p (4.7') above the streak center and 1,400 p (16.45") below it. Thus the streak is sharply demarcated from the dorsal periphery of the retina throughout its naso-temporal extent, but below the streak the density gradients are less precipitous and the transition between the streak and the ventral periphery of the retina is not sharply defined. The streak is not precisely linear because the isodensity lines in the middle third of the retina expand inferiorly forming an obtuse triangle with its apex toward, and just nasal to, the optic disk. Because of this expansion, ganglion cell counts are higher in the ventral than in the dorsal hemiretina. However, the proportion of retina with ganglion cell density of less than 4,000 cells mm-2is approximately equal above and below the streak. An examination of the isodensity maps (figs. 6, 7) and of horizontal transects through the streak center (fig. 9, transect F) reveal that ganglion cell density is not homogeneous within the streak. Instead, densities increase systematically from nasal and temporal periphery toward the midpoint of the streak. Correspondingly, the elliptical isodensity lines
26
ELLENGENE H. PETERSON AND P. S. ULINSKI
Fig. 5 Varieties of neuroglia seen in the retina of Pseudernys scripfa elegans. A-D Glia in the fiber layer over the optic disk. E Small glial cell with distinct, eccentric cytoplasm and patches of clumped chromatin in its pale nucleus. The larger profile is a ganglion cell. F Two ganglion cells and a glial cell of type most commonly observed in Pseudemys retina. G,H Neuroglia with pale nucleus, clumped chromatin and no visible cytoplasm. I Ambiguous profile. This is probably a glial cell because of its filmy cytoplasm which show little basophilia and has no Nissl bodies.
27
TURTLE GANGLION CELLS A-18,000 C E L L S I ~ ~ ' B - 16,000 C - 14.000 D - 12 ,000 E - 10,000 F- 8,000 G - 6.000 H - 4.000 I - 2,000
T E M POR A L
VENTRAL
Fig. 6 Isodensity map of Pseudemys retina, animal M28-R (right eye). Cell density (presumed neurons within 100-p grid) waa sampled every 200 CL over the retinal surface. Isodensity lines depict points at which neuron density decreased from highest (A) to progressively lower densities. For example, in the area between isocount lines I and H, there are more than 2,000and less than 4,000 presumed neurons per square millimeter. The isodensity lines form a series of concentric ellipses. A t higher densities, these ellipses become increasingly eccentric in shape to form a visual streak. The streak appears slightly curved with the concavity toward the inferior retina. This is probably due to mechanical deformation of the retina during mounting (see text). The isodensity lines converge upon a peak density area which is nearly circular and is located at approximately the center of the retina, less than two millimeters above the optic disk (shaded). The retina is depicted as seen in the microscope, i.e., reversed left to right. For this and the following retina (fig. 71,approximate retinal areas and total neurons in the ganglion cell layer are given in table 3. Immediately above the optic disk, an island of cell density greater than 16,000 is unlabelled.
become truncated nasotemporally until the isocount lines are nearly circular a t the point of highest ganglion cell density. Here, ganglion cell counts are distinctly higher than in the
surrounding streak. This is illustrated in the F transects of figures 8 and 9, both of which pass through the region of peak density (cf. fig. 6). Similar plots through the peak density area in
A = 18,000 mm-'
lines are unlabelled.
Fig. 7 Isdensity map of Pseudemys retina, animal HRP-3-R(right eye). Cell density was sampled every 500 w over the retinal surface. For the inset, cell density was sampled every 200 p . In the inset, one 18,000 and two 16,000 isocount
N+T V
D
G = 6,000 H = 4,000 I = 2,000
D = 12,000 E = 10,000 F = 8,000
C = 14,000
6 = 16,000
TURTLE GANGLION CELLS TABLE 3
Areas, total ganglion cells, and peak neuron densities offour Pseudemys retinas Retinal area (mm2)
Total ganglion cells '
Peak neuron density (cells/mm2)
M28-R M28-L HRP3-R 14-L
112.8 105.9 122.1 125.0
362,000 355,000 350,000 391,000
21,000 20,500 19,000 20,000
X
116.5
364,500 18,339
20,125 853.91
Eye number
-
SD
8.77
' Isodensity maps for two of these four retinas (M28-R: HRP3-R) are illustrated in figures 6 and 7, respectively. Rounded to three significant figures.
a second retina (HRP3-R; fig. 7 ) show that, in this retina also, density peaks sharply a t a point midway along the nasotemporal axis. In three retinas, the point of peak density was located just nasal to a line passing through the optic disk and perpendicular to the streak, i.e., at the approximate center of the retina. In the fourth retina, the highest density area lay slightly temporal to this, just above the disk. Thus, there is a region of peak ganglion cell density which is consistently located a t or near the center of the retina. The streak is horizontally asymmetric about this peak density area. Immediately adjacent to the region of highest density, ganglion cell counts decrease more rapidly nasally than temporally (fig. 9, transect F). Thus the highest density of ganglion cells lies, first, a t the center of the retina and, second, just temporal to this. In the periphery, however, this difference in nasal and temporal density gradients is reversed, and the isodensity lines are somewhat more narrow and elongated nasally than temporally. For example, the 6,000 isodensity line extends to the nasal periphery but it lies approximately two millimeters from the temporal edge of the retina (figs. 6, 7 ) . Similarly, in horizontal transects through the streak center and one millimeter below it, ganglion cell density can be seen to drop to less than 1,000 cells mm-*in the far temporal retina; in contrast, there are approximately 5,000 cells mm-2 in the far nasal periphery (fig. 9, transects F, G). Finally, there is a tendency for the isodensity lines to be somewhat expanded toward the temporal-ventral quadrant of the retina. Thus the temporal arm of
29
the streak is both shorter and broader than the nasal arm. Given the calculated retinal magnification factor of 98 g per degree of visual field, we can approximate the visual space subtended by the visual streak. For this purpose, we have arbitrarily chosen the 6,000 isodensity line to define the border of the streak because it is the lowest isocount line which has the streaklike configuration. In retina M28-R (fig. 6), the maximum length of the 6,000 isocount line is 12.2 mm, and its maximum width is 2.0 mm. Thus the streak encompasses 143.4' of visual field from nasal to temporal periphery, and 23.5' from its superior to its inferior border. Similarly, if the peak density area is taken as the 18,000 isocount line, then i t subtends a sector of visual space approximately 2.4" by 4.7" in diameter. To determine the orientation of the streak relative to the externally visible iris line, the eyes were removed from one formol-saline perfused turtle and the position of the iris line, as it crosses the ora serrata, was marked by radial cuts. The two retinas were then processed in the usual manner. Ganglion cell counts made a t the nasal and temporal periphery of each retina indicate that the isodensity lines a t the periphery converge upon the radial cuts. Thus the visual streak is aligned with the pigmented iris line in Pseudemys. This experiment suggests that the apparent curvature of the streak in figures 6 and 7 is due largely or entirely to mechanical deformation of the retina during mounting. To summarize briefly, our observations on retinal wholemounts indicate that ganglion cells in P. scripta are distributed so as to form a horizontally aligned visual streak and that the point of highest ganglion cell density lies within the streak, a t the approximate center of the retina. These observations were confirmed and extended in embedded and sectioned retinas. In retinas sectioned perpendicular t o the iris line, and thus perpendicular to the streak, a local increase in ganglion cell density was observed midway between the dorsal and ventral borders of the retina (fig. 10). A sharp transition is visible between this point of increased density and the retina superior to it. The transition is defined not only by an increase in ganglion cell density, but also by a marked decrease in cell size and an increase in the intensity of staining. Ganglion cells in the dorsal periphery form a continuous
30
ELLENGENE H. PETERSON AND P. S. ULINSKI
monolayer of somas, whereas cells a t the point of high density are three or four layers thick. Below the point of highest density, the thickness of the ganglion cell layer decreases gradually until it is no more than one to two cells thick. There appears to be a local increase in the number of photoreceptors and in the thickness of the inner nuclear layer associated with the point of highest ganglion cell density. No attempt was made to quantify this observation; however, it is consistent with reports that photoreceptor oil droplets in Pseudemys reach a maximum density along the approximate equator of the retina (Brown, '69; Granda and Haden, '70). We believe that the high density area seen in the ganglion cell layer of sectioned retinas corresponds to the visual streak of wholemounted retinas because i t is located appropriately and because the size and staining properties of the cells are similar. Absolute number of ganglion cells
The average number of neurons in the ganglion cell layer of four retinas was estimated t o be 364,000(table 3). Our HRP experiments indicate that no more than five or six percent of total profiles (7% of presumed neurons) are nonganglion cell neurons (table 2). Thus our estimates of total neuronal profiles reflects primarily the population of ganglion cells which send an axon into the optic nerve. DISCUSSION
This paper has examined the ganglion cell layer in the turtle, Pseudemys scripta elegans. There are three principal findings. The first is that ganglion cells comprise 75-80% of total profiles and 90-95%of the neurons in the ganglion cell layer. The second is that ganglion cells are distributed inhomogeneously within this layer, forming a horizontally aligned visual streak which contains an area of peak density a t its midpoint. The third finding is that the total number of ganglion cells is approximately 364,000.The remainder of this section deals with each of the findings in turn. Composition of the ganglion cell layer
To develop criteria for distinguishing between ganglion cells and neuroglia, we attempted to label ganglion cells retrogradely by applying HRP to the optic nerve. In these experiments, an average of 78% of profiles in any field were labelled. In general, the cytology of labelled cells was similar to that of ganglion cells already described (e.g., Hughes,
'75a; Stone, '65, '78; Wassle et al., '75). As in other vertebrates, retinal ganglion cells in Pseudemys show considerable variation both in cytology and in size. An average of 22% of the profiles in the ganglion cell layer were unlabelled, and there are a t least three interpretations of these cells. Unlabelled ganglion cells The vagaries of HRP labelling are now well documented (e.g., LaVail, '75; Vanegas et al., '78; Winer, '771, and it is possible that one or more categories of ganglion cells did not transport sufficient HRP t o be detected (cf. Bunt et al., '74). Alternatively, in some cells the enzyme may have been inactivated by the time we examined the tissue (LaVail, '751, although this seems unlikely since survival times shorter than five or six days resulted in poorly labelled retinas. Furthermore, neither optic nerve was saturated with HRP so that large patches of the two retinas were labelled erratically or not a t all. To maximize the probability that any ganglion cells in our samples would be labelled, we analyzed only those regions of the retinas which showed the heaviest deposition of grains. Nevertheless, some unlabelled profiles were probably ganglion cells which were never exposed to the enzyme in a concentration sufficient to cause labelling. Neuroglia Seventy t o eighty percent of unlabelled cells in the ganglion cell layer were probably neuroglia. Outside the streak, the distinction between neurons and glia proved relatively easy to make, both in our HRP cases and in the wholemounts from which our isodensity maps were constructed. Glia were typically the smallest profiles in a given field. They had little visible cytoplasm and were either heavily basophilic or pale with patches of clumped Fig. 8 Density of neurons in the ganglion cell layer is plotted as a series of equally spaced vertical transects through the visual streak. Data are from retina depicted in figure 6. Distance between the transects is one millimeter. Their approximate location is indicated on the inset schematic retina. Arrow a t F indicates transect passing through the area of peak density. Within each transect, the distance between points is 200 p . Note that neuron density falls off more sharply above the streak than below it and that the shape of the gradient is the same across the retina from nasal to temporal periphery. Breaks in the plots occur at the optic disk (shaded) and where a transect intersects one of the radial cuts made to flatten the retina. VS, center of visual streak, i.e., imaginary line running through the peak density area; N, nasal; T, temporal.
31
TURTLE GANGLION CELLS T
ABCDEFGHIJK
N A
-
‘ 5I!
+6mm
+4mm+2mm
+6mm
+4mm+
VS
T
-2mm
-4mm
-6mm
N
E E
n z
a cn
I
3
I
2mm
VS
-2mm
-4mm
-6mm
+2mm
VS
-2mm
-4mm
-6mm
0
I
I-
10
5
w
u
I
& +6mm
10-
5-
+4mm
3, ---&
VERT I CAL
D I STANCE
FROM V ISUAL
Figure 8
STREAK
(vs)
32
ELLENGENE H. PETERSON AND P. S. ULINSKI
15-
E 10-
5-
-
N
E
I
f6mm
E
I
I
+4mm
f2mm
VS
I
-2mm
-4mm
-6mm
\
cn 1 1 W
0
n
20
3 0
15
F
z Q cn
I
I-
>-
10
I-
v)
z
w n
5
1 1 W
0
15
-
10
-
5-
I
V E R T I C A L D I S T A N C E FROM V I S U A L STREAK (VS) Figure 8 fcontinuedl
33
TURTLE GANGLION CELLS 15-
H
10-
5-
1
N
E
+6mm
+4mm
+2mm
VS
-2mm
-4mm
-6mm
+6mm
+4mm
+2mm
VS
-2mm
-4mm
-6mm
+6km
+4km
+
ABSTRACT Multiple pathways for the transmission of visual information from retina to brain have been described in reptiles, but little is known about their functional organization. These parallel channels begin a t the retina, and we have therefore begun to study the functional organization of retinal ganglion cells in the turtle, Pseudemys scripta elegans. This paper describes the numbers and distribution of cells in the ganglion cell layer. To develop criteria for the identification of ganglion cells, we labelled them retrogradely by applying horseradish peroxidase (HRP) to the optic nerve. Ganglion cells were found to vary substantially in size and cytology. In low density areas of the retina, ganglion cells typically have cytoplasm with well developed Nissl substance, a distinct, pale nucleus, and a large nucleolus. In high density areas of retina, ganglion cells are small, densely staining, and gliaform. The average minimum proportion of ganglion cells in the ganglion cell layer i s 7580%of total profiles. No more than five or six percent of profiles in the ganglion cell layer are neurons which do not send an axon into the optic nerve (displaced amacrine cells or intraretinal association cells). The ganglion cell layer of P. s. elegans can be divided into a number of regions on the basis of cell density. Isodensity maps constructed from Nisslstained, wholemounted retinas indicate that there is an elongated region of high ganglion cell density, the visual streak, which extends from nasal t o temporal retina and is oriented such that its long axis follows the horizontal axis of the eye. The streak is aligned with the externally visible iris line. Seen in crosssection, the ganglion cell layer in the streak is three to four cells thick; in nonstreak retina, ganglion cells form only a monolayer of somas. Ganglion cell density drops off more rapidly above the streak than below it. The temporal arm of the streak is both shorter and broader than the nasal arm. There is a peak in ganglion cell density at the midpoint of the streak, in the approximate center of the retina. Here, ganglion cell densities exceed 20,000 cells mm-2. The total number of ganglion cells in the retina is 350,000-390,000. The retina in reptiles projects to at least six targets within the central nervous system: to one or more loci within the hypothalamus, dorsal thalamus, ventral thalamus, pretectum, tectum, and midbrain tegmentum (Butler, '79; Cruce and Cruce, '78; Ebbesson, '72). Each of these retinorecipient zones makes subsequent efferent connections (Northcutt, '78; Ulinski, '77, '79). Thus the visual system of reptiles, as of other vertebrates, consists of multiple, parallel pathways for the transmisJ. COMP. NEUR. (1979) 186: 17-42.
sion of visual information. We would like to know whether these parallel channels process different categories of visual information and, if so, how this is effected. Accordingly, we have begun a series of experiments designed to explore the functional architecture of the visual brain in a n emydid turtle, Pseudemys scripta elegans. We have chosen this species because more information is available on its retinal physiology (Cervetto, '76; Fuortes, '76; Granda and Dvorak, '771, central visual path-
17
18
ELLENGENE H. PETERSON AND P. S. ULINSKI
ways (Butler, '79; Northcutt, '78; Ulinski, '791, visual discrimination (Burghardt, '77; Peterson, '79), and psychophysical performance (Granda and Dvorak, '77) than for any other reptile. The first experiments have characterized the intrinsic organization and connections of each structure in the tectofugal pathway, both qualitatively and quantitatively (Balaban, '78a,b; Peterson, '77, '78a,b; Rainey, '78, '79; Schechter and Ulinski, in preparation; Ulinski, '78, '79). In this paper we begin an analysis of the neural retina in Pseudemys with an account of ganglion cell numbers and distribution in Nissl-stained, wholemounted retinas. A problem with such studies has been the difficulty in distinguishing between neurons and neuroglia in the ganglion cell layer (Hughes, '75a; Stone, '65). Therefore, we first developed criteria for the identification of ganglion cells in Pseudemys by labelling them retrogradely with horseradish peroxidase (HRP) applied to the optic nerve. We then used these criteria to map ganglion cell distribution. The principal finding of this study is that the ganglion cell layer in Pseudemys exhibits a number of regional specializations, including a horizontall y aligned visual streak and an area of peak density in the center of the retina. MATERIALS AND METHODS
Eye dimensions and retinal preparations Turtles were sacrificed by transcardial perfusion of 0.9%saline followed by 10%formolsaline. The eyes were removed and stored in formol-saline for four to five days. Eyes from five turtles were measured under a dissecting microscope with a calibrated eyepiece reticle. Horizontal, vertical, and axial bulb diameters and pupil diameter were recorded. To study retinas, eyes were opened under saline by an incision just distal to the ora serrata and the anterior portion of the bulb was discarded. Retinas were then teased free from the posterior eyecup and cleaned of any remaining pigment epithelium. A series of radial cuts was made in each retina so that it could be flattened and mounted, fiber layer up, on a gelatinized slide. The retinas were air dried for 24 hours, and then defatted by passing them through a graded series of alcohols and xylene. A final overnight wash in xylene markedly improved the subsequent staining. After defatting, the retinas were rehydrated and then stained with 0.1%unbuffered cresyl violet a t 40°C for 30 minutes, differentiated in
acid alcohol plus one wash of a 1:6:1 mixture of ether: chloroform: 100%ethyl alcohol, dehydrated, cleared, and coverslipped. Two sets of measurements were taken on these retinal wholemounts. To measure shrinkage, two fixed retinas were mounted on a slide and measured before air drying. Five measurements were taken between pairs of identifiable landmarks on each retina. Each measurement was repeated after wholemount processing, and the average percent shrinkage was calculated. To compute retinal areas, processed retinas were placed in a microprojector and the enlarged image was traced onto graph paper, cut out and weighed. In addition, a known area was comparably enlarged onto the same graph paper, weighed, and used to calculate retinal areas. Two eyes were embedded in paraffin, and two were embedded in celloidin. Serial sections were cut a t 10 p and 30 p , respectively, and stained with cresyl violet. Two of these eyes were cut in a plane parallel to the pigmented iris line (fig. 1).The two remaining eyes were cut orthogonal to the iris line and parallel t o a plane dividing the eye into nasal and temporal halves. Horseradish peroxidase preparations
To distinguish between neuroglia and retinal ganglion cells, horseradish peroxidase (HRP) was applied t o the optic nerve in four turtles, and retinal wholemounts from these animals were examined. In this procedure, the olfactory bulbs were removed to expose the underlying optic nerves of anesthetised turtles (Brevital Sodium, 9-15 mgkg; Wang et al., '77). A transverse incision was made in one optic nerve of each animal, and the resulting "pocket" was filled with crystals of HRP (Sigma, Type VI). After three to six days, both eyes were excised under deep anesthesia. The fresh retinas were removed in a bath of 0.15 M phosphate buffer, fixed in 1%glutaraldehyde in phosphate buffer for ten minutes, rinsed in fresh buffer, and reacted with diaminobenzidine. The retinas were then wholemounted and processed as described above. The contralateral retina served as a control for endogenous peroxidase. The retinas were examined with oil immersion objectives at 1,250 x. Labelling was maximal in two turtles with survival times of five and six days, and only these two cases were analyzed. Since the optic nerve was not uniformly saturated with HRP, labelled ganglion cells were distributed in
19
TURTLE GANGLION CELLS
Fig. 1 Pseudemys scriptu eleguns. The photograph shows a freely swimming turtle. Note the black iris line. The turtle's head is extended and raised, but compensatory eye rotation maintains the black iris line approximately at horizontal (water line).
patches over t h e wholemounted retinas. Heavily labelled areas were sampled as follows. All of the profiles within a field were drawn with a camera lucida and marked as labelled or unlabelled. Unlabelled cells were designated as either presumed neurons or neuroglia according to criteria outlined below (RESULTS). Four samples of approximately 100 cells each were taken; two were taken within the streak, and two within the peripheral retina. For comparison, four samples were taken from normal wholemounted retinas, a t approximately comparable loci, and the proportion of neuroglia in each sample was estimated.
Distribution and absolute number of ganglion cells Four retinal wholemounts were examined a t 1,000 x . A 100 p x 100 p grid was advanced systematically over the surface of the retina in equal increments of either 200 p (high resolution sample; 1 retina) or 500 p (low resolution sample; 3 retinas). Profiles falling within the grid were counted and identified as either (presumed) neurons or neuro-
glia. Of cells which intersected the borders of the grid, only those touching the upper and right edges were included. In addition, a high resolution sample (i.e., one 100 p x 100 p grid per 200 p x 200 p area) was taken over approximately the central 2 mm X 2 mm of one low resolution retina (fig. 7) since the density gradients are relatively steep in this area. Because ganglion cell distribution in all four retinas appeared virtually identical, i t was not necessary to taken high resolution samples from the two remaining retinas. No correction was made for shrinkage in calculating ganglion cell density. To estimate the total number of ganglion cells in each retina, the number of ganglion cells counted was totalled and multiplied by either 4 (high resolution retina) or 25 (low resolution retinas). RESULTS
Pseudemys scripta,' the pond slider, is a I On the basis of cranial characteristics (McDowell, '64), osteology (Weaver and Rose, '671, and soft anatomy (Zug, '66),several authors have suggested that the genus Pseudemys be reclassified and subsumed under the Chrysemys group. Thue P. 8. elegans would become Chrysemys scripto elegons. This usage has been adopted in Some recent communications (e.g.,Ernst and Barbour, '72).
20
ELLENGENE H. PETERSON AND P. S. ULINSKI
cryptodiran turtle of the family Emydidae (Testudinidae). The subspecies P.s. elegans is distinguished by a red postorbital patch from which it derives its common name, the redeared slider. P. scrzpta is a New World, freshwater turtle which is omnivorous and primarily diurnal. It is markedly opportunistic, both as to habitat and to food (Boyer, '65; Cagle, '50; Cahn, '37; Moll and Legler, '71). Although this paper focuses on the ganglion cell layer of the retina, it is useful first t o outline some general features of the eye and retina. Morphology of the eye and macroscopic features of the retina The dimensions of the eyeball are given for ten eyes in table 1.The horizontal diameter is slightly greater than the vertical diameter. Both are greater than the axial diameter, indicating that the bulb is flattened from front to back. Similar eye dimensions are reported by Bowling ('76). The pupil is circular (fig. 1) and, in our specimens, approximately 1.74 mm in diameter (table 1). Granda and Dvorak ('77) and Bowling ('76) report roughly comparable diameters. The iris is green except for a black iris line which runs across it horizontally, or approximately so (fig. 1). Brown ('69) has reported that this iris line is maintained a t horizontal by non-visual reflexes of the head and neck (DISCUSSION). Retinal dimensions are given in table 1. The mean horizontal diameter is somewhat greater than the vertical diameter. The average area of these retinas is 87.5 mm2.The four wholemounted retinas, taken from somewhat larger turtles, average 116.5 mm2 (table 3). Granda and Haden ('70) report a retinal area of 138 mm2for P. scripta, but they did not report the sizes of the turtles from which the retinas were taken so that i t is difficult to compare their data with ours. Shrinkage always occurred during processing, but it averaged no more than two percent. As noted by other authors (e.g., Hughes, '75a), shrinkage was greatest at the edges of the retina and along radial cuts or tears. At such points, ganglion cell counts were often elevated, and such samples were discarded in constructing maps of ganglion cell density. To calculate visual angle, we used dimensions for the schematic eye of P. s. elegans given by Granda and Dvorak ('77). The distance between their average nodal point and the junction between retina and pigment epithelium is 4.92 mm. From this we estimate a
retinal magnification factor, near the center of the retina, of approximately 98 p per degree of visual field, or 11.5'-12.0" of visual field per millimeter of retina. This latter figure is reasonably close to a figure of approximately 13.0' of visual field per millimeter of retina derived for the high resolution retina (M28-R; figs. 6 , 8 , 9 ) by assuming that the entire retina (14.0 mm vertical diameter) sees 180' of visual space. To the unaided eye, the most striking feature of the vitreal surface is a prominent line which extends horizontally across the retina. It appears white in fresh preparations and as a dark blue streak in stained wholemounts (fig. 2A). Brown ('69) reported that a similar streak is visible a t the photoreceptor level as an increased density of receptors with colored oil droplets. One goal of this paper was to determine the relationship between ganglion cell distribution and this streak. However, it was first necessary to develop criteria by which ganglion cells can be distinguished from neuroglia and from other neurons lying in the ganglion cell layer. Identification of ganglion cells To determine the cytological characteristics of ganglion cells in Pseudemys retina, we used two cases in which HRP was applied to the optic nerve. Following five or six days survival, these retinas were wholemounted and the ganglion cell layer examined for the presence of HRP positive profiles. No evidence of peroxidase granules was ever observed in the two retinas contralateral to the HRP application. The ipsilateral retinas, in contrast, were heavily labelled in some areas, and the fiber layer was so deeply stained that it obscured much of the ganglion cell layer. These HRP cases revealed t h a t ganglion cell size and cytology varies with ganglion cell density. In peripheral retina where cell density is relatively low (figs. 3A,B), labelled cells resembled ganFig. 2 Visual streak. A Retina M28-L. The wholemounted retina has been stained with cresyl violet, and the visual streak appears as a dark blue line extending horizontally from nasal to temporal periphery. Note the expansion of the streak at the center of the retina which corresponds to the region of peak density. The optic nerve exits from the temporal-ventral quadrant (lower right). B High power photograph of the dorsal border of the visual streak, taken from a point approximately midway between the peak density area and the nasal edge of the retina. Note the sharp increase in cell density in the streak and that cells in the streak are relatively small, densely staining, and gliaform.
TURTLE GANGLION CELLS
21
N
22
ELLENGENE H. PETERSON AND P. S. ULINSKI TABLE 1
Dimensions of
eye
and retina in Pseudemys scripta elegans
Animal Eye number
Weight (g)
110 R L 112 R
Carapace length (cm)
280
13.5
250
13.0
L 116 R L 119 R L 120 R L
470
15.5
1,050
21.5
1,000
19.5
X SD
610 388.52
16.6 3.75
-
Bulb diameters (mm) Horizontal
7.2 7.4 7.0 6.9 8.4 8.5 8.7 8.6 9.4 8.9 8.1 0.89
Vertical
7.2 7.2 6.6 6.6 8.2 8.1 8.2 8.3 8.6 8.4 7.7 0.76
Axial
Retinal diameters (mm)
'
5.6 6.0 5.4 5.6 7.0 6.6 6.8 6.9 7.2 7.0 6.4 0.69
Pupil diameter
1.6 1.6 1.6 1.5 2.0 2.0 1.6 1.7 1.9 1.9 1.74 0.19
Retinal area (mm ?)
Horizontal
Vertical
12.5
12.0
77.3
11.5
11.2
67.3
12.5
12.5
92.1
13.0
12.8
90.6
14.0
13.8
110.4
12.7 1,908
12.46 0.963
87.53 16.35
' Axial diameter measured for posterior pole of bulb to plane of limbus.
glion cells as described in detail by several authors (e.g., Stone, '65, '78; Hughes, '75a; Tiao and Blakemore, '76). They have a pale nucleus and a large, darkly staining nucleolus (figs. 3, 4). The amount of cytoplasm and the development of Nissl substance is variable, being relatively greater in the largest cells. In smaller neurons, the cytoplasm is often rimmed with Nissl substance. Such ganglion cells are characteristically irregular in outline and assume a violet color in Nissl preparations. In contrast, within high density areas, labelled ganglion cells are very small and so densely stained that often no nucleus is visible. These cells tended to resemble glia, and thus our HRP cases were particularly useful in helping us determine the subtle cytological differences between neurons and glia within the high density retina. Here, as in the retinal periphery, ganglion cells can be distinguished primarily by their relatively irregular outline and violet stain. Our HRP material also provides a lower limit for the proportion of ganglion cells in the ganglion cell layer. Seventy-five to eighty percent of profiles in the ganglion cell layer were HRP positive and thus are conventional ganglion cells in the sense that their axons exit the retina in the optic nerve (table 2). Conversely, 20-25% of total profiles were unlabelled. Most of these cells (70430% of unlabelled profiles) were judged to be neuroglia because they resembled published accounts of glia in reptiles (Stensaas and Stensaas, '68;
Kruger and Maxwell, '67) or profiles seen in the fiber layer near the optic disk (fig. 5 ) . Such presumptive glia are generally smaller than neurons, a t least in low density areas. Their nuclei may be dark and uniformly stained or pale, often with clumped chromatin. Cytoplasm, when visible, is pale and filmy and may be eccentric. Neuroglia tend t o have a smooth outline in Nissl preparations and a relatively blue color. An average of 16%of total profiles in both labelled and normal wholemounts were judged to be neuroglia according to these criteria (table 2). We found no consistent evidence that the proportion of neuroglia increases with eccentricity. However, our observations were limited to areas of retina within two millimeters of the streak, and the proportion of neuroglia may be relatively greater in the far periphery. Possible identification of Fig. 3 Retinal ganglion cells labelled with horseradish peroxidase (HRP) granules following application of HRP to the ipsilateral optic nerve. Five days survival. A Heavily labelled area approximately two millimeters above the visual streak. Three profiles in this field are unlabelled. One, incompletely shown in upper right comer (A), is clearly a neuron because of its size. A second unlabelled cell (arrow) is probably a neuron because it is very similar to a labelled profile nearby (dark asterisk). The third unlabelled cell, slightly out of focus, ia probably a glial cell (white asterisk; cf. fig. 5F). B A less heavily labelled retinal area. T w o neurons are unlabelled and one probable glial cell (asterisk). The glial cell resembles that shown in figure 51. Scale as in A. C A heavily labelled area of the dorsal periphery of the streak. One neuron (arrow) and one probable glial cell (asterisk; cf. fig. 5H) are unlabelled.. Scale as in A.
TURTLE GANGLION CELLS
Figure 3
23
24
ELLENGENE H. PETERSON AND P. S. ULINSKI
Fig. 4 Varieties of ganglion cells seen in Pseudemys scripta elegans. A,B,G These large cells occur infrequently. They seldom reach 20 I./ in (average long and short) diameter, and are almost never seen in the streak (cf. fig. 2B). Next to the large, multipolar ganglion cell in B, is a probable glial cell (asterisk; cf. fig. 5H). C Two medium ganglion cells seen frequently in Pseudemys retina. D-F,H Medium ganglion cells typical of non-streak retina, but also seen occasionally in the streak. I Ganglion cells in the visual streak. All profiles in this picture are probably ganglion cells.
25
TURTLE GANGLION CELLS TABLE 2
Composition of the ganglion cell layer Visual streak
'
-
X
SD
91 80 20 16 4
99.50 78.00 22.00 16.00 6.25
11.03 8.29 8.29 4.55 3.86
86 19
98.30 15.50
11.32 4.36
Periphery
HRP cases Total profiles % labelled % unlabelled % glia % unidentified
110 66 34 22 12
89 81 19 15 5
Total profiles % glia
111 10
104 14
108 85 15 11 4
Normal cases 92 19
' T w o samples, taken from dorsal border of streak (8,000-10,000 cells -mm2). 'Two samples, taken approximately 2 mm above streak samples (4,000 cells -mm') 3All percent figures refer to percent of total profiles in sample.
the remaining unlabelled profiles in the ganglion cell layer is considered in the DISCUSSION.
Ganglion cell distribution Having formulated criteria for the identification of ganglion cells and neuroglia, we next assessed ganglion cell distribution in four retinas by systematically sampling ganglion cell density over the retinal surface in increments of either 200 p (1high-resolution retina) or 500 p (3 low-resolution retinas). From these data we constructed isodensity maps for the four retinas (figs. 7,8; table 3). In addition, we plotted ganglion cell density in the high resolution case as a series of equally spaced transects (both horizontal and vertical) through the retina (figs. 8,9).From these data it appears that the retina of Pseudemys can be divided into a number of regions on the basis of ganglion cell density. The remainder of this section deals with these areal variations. The most striking feature of ganglion cell distribution is the presence of a horizontally aligned area of high ganglion cell density (figs. 2, 3 0 . We call this region the visual streak by analogy to a similar ganglion cell configuration in a number of other vertebrate species (DISCUSSION). The streak lies slightly less than two millimeters above the optic disk, just above a line dividing the retina into dorsal and ventral halves (figs. 2A, 6, 7). The shape and disposition of the streak are virtually identical in all four retinas. The streak is vertically asymmetric in that the ganglion cell density gradient is more pronounced above the streak than below it. This can be seen in the isodensity maps (figs. 6, 7)
and in figure 8 which shows ganglion cell density in the high-resolution retina (M28-R) plotted as a function of vertical distance from the streak. Above the streak, ganglion cell counts drop from peak values of more than 20,000 cells mm-2 t o less than 10,000 cells mm-2within a linear distance of 400 @ or 4.7" of visual space (figs. 6, 8). Below the streak, a comparable drop in ganglion cell density occurs 800 p or 9.4" from the streak center. In a second retina (HRPS-R; fig. 71, ganglion cell density drops from peak values of more than 18,000 cells mm-2to approximately 7,000 cells mm-2 a t 400 p (4.7') above the streak center and 1,400 p (16.45") below it. Thus the streak is sharply demarcated from the dorsal periphery of the retina throughout its naso-temporal extent, but below the streak the density gradients are less precipitous and the transition between the streak and the ventral periphery of the retina is not sharply defined. The streak is not precisely linear because the isodensity lines in the middle third of the retina expand inferiorly forming an obtuse triangle with its apex toward, and just nasal to, the optic disk. Because of this expansion, ganglion cell counts are higher in the ventral than in the dorsal hemiretina. However, the proportion of retina with ganglion cell density of less than 4,000 cells mm-2is approximately equal above and below the streak. An examination of the isodensity maps (figs. 6, 7) and of horizontal transects through the streak center (fig. 9, transect F) reveal that ganglion cell density is not homogeneous within the streak. Instead, densities increase systematically from nasal and temporal periphery toward the midpoint of the streak. Correspondingly, the elliptical isodensity lines
26
ELLENGENE H. PETERSON AND P. S. ULINSKI
Fig. 5 Varieties of neuroglia seen in the retina of Pseudernys scripfa elegans. A-D Glia in the fiber layer over the optic disk. E Small glial cell with distinct, eccentric cytoplasm and patches of clumped chromatin in its pale nucleus. The larger profile is a ganglion cell. F Two ganglion cells and a glial cell of type most commonly observed in Pseudemys retina. G,H Neuroglia with pale nucleus, clumped chromatin and no visible cytoplasm. I Ambiguous profile. This is probably a glial cell because of its filmy cytoplasm which show little basophilia and has no Nissl bodies.
27
TURTLE GANGLION CELLS A-18,000 C E L L S I ~ ~ ' B - 16,000 C - 14.000 D - 12 ,000 E - 10,000 F- 8,000 G - 6.000 H - 4.000 I - 2,000
T E M POR A L
VENTRAL
Fig. 6 Isodensity map of Pseudemys retina, animal M28-R (right eye). Cell density (presumed neurons within 100-p grid) waa sampled every 200 CL over the retinal surface. Isodensity lines depict points at which neuron density decreased from highest (A) to progressively lower densities. For example, in the area between isocount lines I and H, there are more than 2,000and less than 4,000 presumed neurons per square millimeter. The isodensity lines form a series of concentric ellipses. A t higher densities, these ellipses become increasingly eccentric in shape to form a visual streak. The streak appears slightly curved with the concavity toward the inferior retina. This is probably due to mechanical deformation of the retina during mounting (see text). The isodensity lines converge upon a peak density area which is nearly circular and is located at approximately the center of the retina, less than two millimeters above the optic disk (shaded). The retina is depicted as seen in the microscope, i.e., reversed left to right. For this and the following retina (fig. 71,approximate retinal areas and total neurons in the ganglion cell layer are given in table 3. Immediately above the optic disk, an island of cell density greater than 16,000 is unlabelled.
become truncated nasotemporally until the isocount lines are nearly circular a t the point of highest ganglion cell density. Here, ganglion cell counts are distinctly higher than in the
surrounding streak. This is illustrated in the F transects of figures 8 and 9, both of which pass through the region of peak density (cf. fig. 6). Similar plots through the peak density area in
A = 18,000 mm-'
lines are unlabelled.
Fig. 7 Isdensity map of Pseudemys retina, animal HRP-3-R(right eye). Cell density was sampled every 500 w over the retinal surface. For the inset, cell density was sampled every 200 p . In the inset, one 18,000 and two 16,000 isocount
N+T V
D
G = 6,000 H = 4,000 I = 2,000
D = 12,000 E = 10,000 F = 8,000
C = 14,000
6 = 16,000
TURTLE GANGLION CELLS TABLE 3
Areas, total ganglion cells, and peak neuron densities offour Pseudemys retinas Retinal area (mm2)
Total ganglion cells '
Peak neuron density (cells/mm2)
M28-R M28-L HRP3-R 14-L
112.8 105.9 122.1 125.0
362,000 355,000 350,000 391,000
21,000 20,500 19,000 20,000
X
116.5
364,500 18,339
20,125 853.91
Eye number
-
SD
8.77
' Isodensity maps for two of these four retinas (M28-R: HRP3-R) are illustrated in figures 6 and 7, respectively. Rounded to three significant figures.
a second retina (HRP3-R; fig. 7 ) show that, in this retina also, density peaks sharply a t a point midway along the nasotemporal axis. In three retinas, the point of peak density was located just nasal to a line passing through the optic disk and perpendicular to the streak, i.e., at the approximate center of the retina. In the fourth retina, the highest density area lay slightly temporal to this, just above the disk. Thus, there is a region of peak ganglion cell density which is consistently located a t or near the center of the retina. The streak is horizontally asymmetric about this peak density area. Immediately adjacent to the region of highest density, ganglion cell counts decrease more rapidly nasally than temporally (fig. 9, transect F). Thus the highest density of ganglion cells lies, first, a t the center of the retina and, second, just temporal to this. In the periphery, however, this difference in nasal and temporal density gradients is reversed, and the isodensity lines are somewhat more narrow and elongated nasally than temporally. For example, the 6,000 isodensity line extends to the nasal periphery but it lies approximately two millimeters from the temporal edge of the retina (figs. 6, 7 ) . Similarly, in horizontal transects through the streak center and one millimeter below it, ganglion cell density can be seen to drop to less than 1,000 cells mm-*in the far temporal retina; in contrast, there are approximately 5,000 cells mm-2 in the far nasal periphery (fig. 9, transects F, G). Finally, there is a tendency for the isodensity lines to be somewhat expanded toward the temporal-ventral quadrant of the retina. Thus the temporal arm of
29
the streak is both shorter and broader than the nasal arm. Given the calculated retinal magnification factor of 98 g per degree of visual field, we can approximate the visual space subtended by the visual streak. For this purpose, we have arbitrarily chosen the 6,000 isodensity line to define the border of the streak because it is the lowest isocount line which has the streaklike configuration. In retina M28-R (fig. 6), the maximum length of the 6,000 isocount line is 12.2 mm, and its maximum width is 2.0 mm. Thus the streak encompasses 143.4' of visual field from nasal to temporal periphery, and 23.5' from its superior to its inferior border. Similarly, if the peak density area is taken as the 18,000 isocount line, then i t subtends a sector of visual space approximately 2.4" by 4.7" in diameter. To determine the orientation of the streak relative to the externally visible iris line, the eyes were removed from one formol-saline perfused turtle and the position of the iris line, as it crosses the ora serrata, was marked by radial cuts. The two retinas were then processed in the usual manner. Ganglion cell counts made a t the nasal and temporal periphery of each retina indicate that the isodensity lines a t the periphery converge upon the radial cuts. Thus the visual streak is aligned with the pigmented iris line in Pseudemys. This experiment suggests that the apparent curvature of the streak in figures 6 and 7 is due largely or entirely to mechanical deformation of the retina during mounting. To summarize briefly, our observations on retinal wholemounts indicate that ganglion cells in P. scripta are distributed so as to form a horizontally aligned visual streak and that the point of highest ganglion cell density lies within the streak, a t the approximate center of the retina. These observations were confirmed and extended in embedded and sectioned retinas. In retinas sectioned perpendicular t o the iris line, and thus perpendicular to the streak, a local increase in ganglion cell density was observed midway between the dorsal and ventral borders of the retina (fig. 10). A sharp transition is visible between this point of increased density and the retina superior to it. The transition is defined not only by an increase in ganglion cell density, but also by a marked decrease in cell size and an increase in the intensity of staining. Ganglion cells in the dorsal periphery form a continuous
30
ELLENGENE H. PETERSON AND P. S. ULINSKI
monolayer of somas, whereas cells a t the point of high density are three or four layers thick. Below the point of highest density, the thickness of the ganglion cell layer decreases gradually until it is no more than one to two cells thick. There appears to be a local increase in the number of photoreceptors and in the thickness of the inner nuclear layer associated with the point of highest ganglion cell density. No attempt was made to quantify this observation; however, it is consistent with reports that photoreceptor oil droplets in Pseudemys reach a maximum density along the approximate equator of the retina (Brown, '69; Granda and Haden, '70). We believe that the high density area seen in the ganglion cell layer of sectioned retinas corresponds to the visual streak of wholemounted retinas because i t is located appropriately and because the size and staining properties of the cells are similar. Absolute number of ganglion cells
The average number of neurons in the ganglion cell layer of four retinas was estimated t o be 364,000(table 3). Our HRP experiments indicate that no more than five or six percent of total profiles (7% of presumed neurons) are nonganglion cell neurons (table 2). Thus our estimates of total neuronal profiles reflects primarily the population of ganglion cells which send an axon into the optic nerve. DISCUSSION
This paper has examined the ganglion cell layer in the turtle, Pseudemys scripta elegans. There are three principal findings. The first is that ganglion cells comprise 75-80% of total profiles and 90-95%of the neurons in the ganglion cell layer. The second is that ganglion cells are distributed inhomogeneously within this layer, forming a horizontally aligned visual streak which contains an area of peak density a t its midpoint. The third finding is that the total number of ganglion cells is approximately 364,000.The remainder of this section deals with each of the findings in turn. Composition of the ganglion cell layer
To develop criteria for distinguishing between ganglion cells and neuroglia, we attempted to label ganglion cells retrogradely by applying HRP to the optic nerve. In these experiments, an average of 78% of profiles in any field were labelled. In general, the cytology of labelled cells was similar to that of ganglion cells already described (e.g., Hughes,
'75a; Stone, '65, '78; Wassle et al., '75). As in other vertebrates, retinal ganglion cells in Pseudemys show considerable variation both in cytology and in size. An average of 22% of the profiles in the ganglion cell layer were unlabelled, and there are a t least three interpretations of these cells. Unlabelled ganglion cells The vagaries of HRP labelling are now well documented (e.g., LaVail, '75; Vanegas et al., '78; Winer, '771, and it is possible that one or more categories of ganglion cells did not transport sufficient HRP t o be detected (cf. Bunt et al., '74). Alternatively, in some cells the enzyme may have been inactivated by the time we examined the tissue (LaVail, '751, although this seems unlikely since survival times shorter than five or six days resulted in poorly labelled retinas. Furthermore, neither optic nerve was saturated with HRP so that large patches of the two retinas were labelled erratically or not a t all. To maximize the probability that any ganglion cells in our samples would be labelled, we analyzed only those regions of the retinas which showed the heaviest deposition of grains. Nevertheless, some unlabelled profiles were probably ganglion cells which were never exposed to the enzyme in a concentration sufficient to cause labelling. Neuroglia Seventy t o eighty percent of unlabelled cells in the ganglion cell layer were probably neuroglia. Outside the streak, the distinction between neurons and glia proved relatively easy to make, both in our HRP cases and in the wholemounts from which our isodensity maps were constructed. Glia were typically the smallest profiles in a given field. They had little visible cytoplasm and were either heavily basophilic or pale with patches of clumped Fig. 8 Density of neurons in the ganglion cell layer is plotted as a series of equally spaced vertical transects through the visual streak. Data are from retina depicted in figure 6. Distance between the transects is one millimeter. Their approximate location is indicated on the inset schematic retina. Arrow a t F indicates transect passing through the area of peak density. Within each transect, the distance between points is 200 p . Note that neuron density falls off more sharply above the streak than below it and that the shape of the gradient is the same across the retina from nasal to temporal periphery. Breaks in the plots occur at the optic disk (shaded) and where a transect intersects one of the radial cuts made to flatten the retina. VS, center of visual streak, i.e., imaginary line running through the peak density area; N, nasal; T, temporal.
31
TURTLE GANGLION CELLS T
ABCDEFGHIJK
N A
-
‘ 5I!
+6mm
+4mm+2mm
+6mm
+4mm+
VS
T
-2mm
-4mm
-6mm
N
E E
n z
a cn
I
3
I
2mm
VS
-2mm
-4mm
-6mm
+2mm
VS
-2mm
-4mm
-6mm
0
I
I-
10
5
w
u
I
& +6mm
10-
5-
+4mm
3, ---&
VERT I CAL
D I STANCE
FROM V ISUAL
Figure 8
STREAK
(vs)
32
ELLENGENE H. PETERSON AND P. S. ULINSKI
15-
E 10-
5-
-
N
E
I
f6mm
E
I
I
+4mm
f2mm
VS
I
-2mm
-4mm
-6mm
\
cn 1 1 W
0
n
20
3 0
15
F
z Q cn
I
I-
>-
10
I-
v)
z
w n
5
1 1 W
0
15
-
10
-
5-
I
V E R T I C A L D I S T A N C E FROM V I S U A L STREAK (VS) Figure 8 fcontinuedl
33
TURTLE GANGLION CELLS 15-
H
10-
5-
1
N
E
+6mm
+4mm
+2mm
VS
-2mm
-4mm
-6mm
+6mm
+4mm
+2mm
VS
-2mm
-4mm
-6mm
+6km
+4km
+
