Displaced Amacrine Cells Disappear from the Ganglion Cell Layer in the Central Retina of Adult Fish during Growth

Abstract

purpose. Fish grow throughout life, including enlargement of eye and retina. Retinal growth involves several mechanisms of adjustment, such as cell addition and dendritic growth. To discover possible other means with which the animals adjust to changing eye size, the distribution of displaced amacrine cells (DACs) and ganglion cells (GCs) was analyzed in the retina of three sizes of a South American cichlid, the blue arcara Aequidens pulcher.

methods. DACs were identified by staining with antibodies specific for the calcium-binding protein parvalbumin. They were also weakly positive for staining against choline acetyl transferase (ChaT). GCs were labeled retrogradely with rhodamine dextran. Densities for both DACs and GCs were lower in the retinas of large fish. To distinguish changes due to eye size from specific adjustments, the proportions of DACs to GCs were examined, rather than the absolute cell densities, in various retinal regions in cryostat sections and wholemount preparations from fish of the three sizes.

results. The analyses suggest that, in small and large fish, DACs and GCs were produced in similar proportions (ratio of DACs to GCs, ∼0.62) in the retinal periphery where new retinal tissue was added by the germinal zone. However, in the central retina of large fish, this proportion was shifted toward GCs (DAC-GC ratio as low as 0.25).

conclusions. During growth of the eye, the proportion of DACs in the ganglion cell layer decreases, indicating that these cells are eliminated from the ganglion cell layer by some unknown mechanism.

Two salient features distinguish the retinas of fish from those of mammals: In fish and many nonmammalian vertebrates, the retina is often arranged in a highly regular pattern, ^{1 2} and retinal growth continues throughout life. Retinal growth in fish is achieved by two mechanisms: expansion of existing tissue and addition by cell proliferation. ^{3 4} Tissue expansion results in increasing distances between cell bodies, which translates into lower cell densities in a growing eye. ⁵ Tissue addition occurs in a peripheral growth zone located circumferentially at the edge of the eye, where all retinal cell types are continually generated, with the exception of rod photoreceptors, for which a specific stem cell population exists. ⁶ As a consequence of peripheral tissue addition, older retinal areas become relatively shifted toward the center of the optical axis as the eye grows. A given retinal tissue unit therefore progressively sends information of a more central visual angle to central brain areas through ganglion cell axons. 

Several adjustment mechanisms have been identified on the synaptic, cellular, and tissue levels that indicate how the continuously growing eye copes with the growth-related changes taking place. Scotopic sensitivity is maintained because rod photoreceptors are the only cell type to maintain its density during growth. ⁷ Rods are added throughout the retina in the mature outer nuclear layer (ONL) from a proliferating population of precursor cells. ^{6 8} As a consequence of the insertion of new rods an increasing number of synapses is received by the existing differentiated bipolar cells. ⁹ At least one type of ganglion cell in the goldfish maintains a complete dendritic coverage of the retina during growth and forms new synapses as an adjustment to the decrease of cell density due to stretching. ^{10 11} A region of higher cell density is maintained presumably by differential tissue expansion ¹² or asymmetric cell addition. ¹³ It has been suggested ¹⁴ that the shifting of axon terminals in the optic tectum adjusts for the peripheral annular tissue addition to maintain retinotopic organization. Little is known about adjustments of other cells in the inner retina, specifically horizontal and amacrine cells. Moreover, it is not clear whether modifications occur to compensate for the change in relative location (peripheral versus central) in retinal cells. 

In this study, we examined a specific type of amacrine cell, the interneuron, that strongly influences the output signals from retinal ganglion cells (GCs). Most of the amacrine cells in fish contain inhibitory transmitters, either γ-amino-butyric acid (GABA) or glycine, ^{15 16} often colocalized with various neuropeptides. The retina of cyprinid fish contains an abundance of amacrine cell types. ¹⁷ Most are located on the inner side of the inner nuclear layer (INL); however, one or several populations of amacrine cells lie among GC bodies in the ganglion cell layer (GCL) and are therefore referred to as displaced amacrine cells (DACs). ^{18 19 20 21} Because of their proximity to GCs, DACs are likely to exert strong influence on the output activity of GCs. 

To gain more insight into how neuronal organization adjusts to continued growth, we analyzed the proportion of GCs and non-GCs in the GCL of a growing teleost. We showed a change in the relative amounts of DACs and GCs during retinal growth. 

Materials and Methods

We used the blue arcara (Aequidens pulcher), a South American cichlid fish that has been established as a model to study morphologic changes in the retina. ²² Moreover, retinal growth in other cichlid species has been investigated extensively (e.g., Ref. ¹² ). Fish were bred in our colony in a 12-hour light–dark cycle. To characterize GCs and distinguish them from DACs in the GCL, we labeled both cell types in the adult fish retina by immunocytochemistry and retrograde tracing, respectively. 

Retrograde Labeling of GCs

Fish were killed by cervical section and their eyes removed. Animal maintenance and experimental procedures were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The sclera was removed from the back of the eye, but the lens and cornea were left in place. Retinal GCs were labeled by placing crystals of tetramethylrhodamine dextran (MW 3000, cat. no. 3308; Molecular Probes, Leiden, The Netherlands) on the cut end of the optic nerve outside the eye. After several minutes in a moist Petri dish to allow uptake of the tracer, the eyes were placed in minimal essential medium in a 5% O₂ and 95% CO₂ gas atmosphere at 20°C. After a culture period of 24 hours, the retinas were isolated (the cornea, lens, and the rest of the sclera were removed) and fixed in 4% paraformaldehyde for 2 hours. Subsequently, they were further processed for cryostat sections or as wholemount preparations. After infiltrating the tissue with 30% sucrose for at least 12 hours, we prepared cryostat sections at a thickness of 20 μm and mounted them on gelatin-coated slides. 

Immunocytochemistry

We used antibodies against parvalbumin (product no. P3088, mouse monoclonal, diluted 1:1000; Sigma-Aldrich, Deisenhofen, Germany) and choline acetyl transferase (ChaT; cat. no. AB144P, goat polyclonal, diluted 1:100; Chemicon, Hofheim, Germany) on wholemounts or cryostat sections. All antibody solutions were prepared in 0.3% Triton X-100 and 1% dimethylsulfoxide (DMSO). Nonspecific staining was blocked with normal goat serum or, in the case of anti-ChaT staining, with donkey serum. Primary antibodies were applied to cryostat sections overnight or to free-floating retinas for 48 to 72 hours at 4°C. After rinses in PBS, sections or isolated retinas were incubated with rabbit anti-mouse and/or donkey anti-goat secondary antibodies conjugated to Alexa488 or Alexa660 (Molecular Probes). The anti-parvalbumin stain in some cases was combined with nuclear counterstain (Sytox Green; Molecular Probes) diluted 1:50,000 and incubated for 30 minutes before coverslipping. When anti-parvalbumin and -ChaT stains were combined, primary and secondary antibodies were applied as cocktails. Wholemounts were carefully mounted with the pigmented epithelium side down and the GC side facing the coverslip. Sections and wholemounts were viewed on a confocal microscope (LSM 410; Carl Zeiss Meditec, Oberkochen, Germany) equipped with three lasers with excitation wavelengths at 488, 543, and 633 nm, and appropriate long-pass and band-pass filter sets, depending on whether multiple lasers were used simultaneously or in separate scans. Laser attenuation and photomultiplier gain control were carefully used to avoid fluorescence emission cross talk. Control experiments for antibody stains were performed by omitting the primary antibodies. 

Quantification of whole flatmounted preparations was performed by scanning the entire retina in square images (320 × 320 μm) with a 40× objective (numeric aperture [NA] 1.2 W). Within each scanned image, labeled GCs and DACs were identified and recorded in a representative area of 100 × 100 μm. Thus, the DAC-to-GC ratio was determined directly, rather than from calculated cell densities. Six retinas were analyzed as wholemounted retinas; however, not all could be evaluated completely because of insufficient retrograde tracing in larger fish or problems with antibody penetration and collection of data from the periphery in small fish. We therefore performed a quantitative comparison on cryostat sections to be sure that we could account for every cell in the GCL. 

For a quantitative analysis of different aged animals, we used three groups of fish (two animals each) with an average standard length of 2, 5, and 9 cm, corresponding to approximate ages of 3, 9, and 18 months. Cryostat sections in the approximate equatorial plane of retinas labeled retrogradely with rhodamine dextran were collected and stained against parvalbumin, as described earlier. Only sections containing both temporal and nasal peripheral regions and central retina were analyzed, by counting cells in the GCL and recording the total distance between the peripheral germinal zones along the sections. This total distance in each section was then divided into five equal sectors or bins to which the determined cell counts were allotted. Cells were counted along a 300-μm linear distance of the sections within the sectors. The number of parvalbumin-positive cells in the GCL and rhodamine-dextran–labeled cell bodies were always counted in the same scanned image to determine their ratio. A minimum of 10 counts per sector and size group were recorded. Before the ratio was calculated, we corrected for section thickness and different soma sizes according to Abercombie ²³ and determined cell densities per unit retinal area on the level of the GCL (Tables 1 2) . With the 40× W objective (NA 1.2) and the pinhole in the confocal light path set at 1 Airy unit, the thickness of confocal optical sections was approximately 1 μm. Soma sizes for each cell type and animal size were determined in stacks of optical images through the entire extent of cryostat sections in which the center of the cell body, (i.e., largest diameter) was identified and measured. We determined an average soma size of 4.51 ± 0.54, 4.44 ± 0.32, and 4.01 ± 0.36 μm for DACs in small, medium, and large fish, respectively. For GCs, the corresponding sizes were 5.2 ± 0.94, 4.9 ± 0.53, and 4.55 ± 0.32 μm, respectively. We performed analysis of variance (ANOVA) to test for differences in the DAC-to-GC ratios of retinal sectors within each size group and across size groups. Student’s t-test was used for statistical comparison of two sectors of density ratios within one fish group. 

Results

We were interested in cells in the GCL in cichlids, in particular the proportion and distribution of non-GCs in this retinal cell layer. The combination of two cell-specific labels together with a nuclear counterstain turned out to be useful in this study: retrograde labeling from the cut end of the optic nerve and immunohistochemical staining for parvalbumin. In retrogradely labeled retinas, many but not all cells in the GCL and some cells in the INL were positively stained with the rhodamine dextran dye. Based on backlabeling from the optic nerve through their axons, they were identified as GCs or displaced GCs, respectively (Fig. 1) . The dendritic processes of these cells were stained as well but were distinctly identifiable only in larger GCs. Dendritic trees of smaller GCs were either weakly filled with the dye, or the branching could not be traced because of overlap with processes of neighboring cells. 

Staining retinal sections with antibodies against parvalbumin showed positively labeled cells of various intensities. Strong antibody binding was present in a single row of cells on the inside of the INL, which were identified as a population of amacrine cells. Several cell populations were identified with weaker staining intensities. One population of parvalbumin-positive cells was located in the outer INL, one in the inner INL, and one cell population in the GCL (Fig. 1) . In the inner plexiform layer (IPL) there were two major parvalbumin-positive bands, one strongly labeled at approximately 20% of the IPL width and a less intensely labeled band at 75% (0% = inner edge of the INL, 100% = outer edge of the GCL; Fig. 1 ). Sometimes a weaker band could be detected at 50% of the IPL width (see Figs. 1 2 3 ). 

Combining the parvalbumin antibody stain with the retrograde rhodamine dextran stain revealed that the two labels did not overlap. Yet, almost all cells in the GCL were stained by either one of the two labels. This was confirmed with a nuclear counterstain (Sytox Green; Molecular Probes), visualizing all cell nuclei in the GCL. Rarely, a nucleus was found in the GCL that could not be associated with either the parvalbumin or the rhodamine dextran stain. Thus, parvalbumin-positive cells in the GCL were not GCs but could be characterized as DACs, because parvalbumin-positive cells were also found among amacrine cells in the INL. 

To place this staining pattern in context with other known cell types in the inner retina, we used antibodies against ChaT to label cholinergic amacrine cells. Several cell populations were ChaT positive. Strongly immunoreactive cells with large cell bodies were found only in the INL; weakly stained cells with smaller cell bodies were present in the INL and GCL (Fig. 2) . The weaker ChaT signal in the smaller cells may be due to their scant cytoplasm. In the IPL, two broad ChaT-positive stripes at 25% to 40% and 60% to 70% of the IPL width were clearly visible in retinal cross sections, separated by a ChaT-negative band between, at 40% to 60% IPL width. In addition, at least two more weakly stained layers were present. Triple labeling for GCs filled retrogradely and with antibodies against ChaT and parvalbumin revealed that many parvalbumin-positive DACs were faintly positive for ChaT, but none of the GCs showed ChaT immunoreactivity. Most of the amacrine cells in the INL were not double labeled, with fewer cells labeled for ChaT than for parvalbumin. A few weakly ChaT-positive cells in the INL were also positive for parvalbumin (Fig. 2 , top right, small arrows). Within the IPL, the only overlap between the parvalbumin and ChaT immunostains was found on the inner side (at approximately 75%) with weak ChaT reactivity (Fig. 2 , bottom right, arrowheads). 

When we compared different retinal areas for the presence of DACs, we observed that in the retinal periphery a higher fraction of cells were parvalbumin positive than in the central retina. To corroborate this observation we measured the density of GCs and DACs across the retina in retinal wholemount preparations (Fig. 3) . We were primarily not concerned about the absolute densities, because absolute densities in the fish retina are dependent on the eye size and the previously occurring tissue expansion. Rather, we calculated the ratio of DACs to GCs in double-labeled preparations. Investigating six retinas from four fish, we found a difference in the ratio of DACs to GCs, with a lower ratio in central retinal areas (Fig. 4) . Similar results were observed in stained retinal sections (Fig. 3 , top). 

This result suggested the possibility that the rate of cell production changes in one of these cell types as the animal grows. We therefore performed the experiments and stains on three different sizes of fish (standard lengths ∼2, ∼5, and ∼9 cm, respectively), to test for variations in the DAC-GC ratio. The rationale for this investigation was based on the fact that new tissue is added to the retina in the peripheral germinal zone and therefore the peripheral retina is younger than the central retina. Thus, the peripheral retina of a small fish becomes more centrally located as the retina grows. If the proportions change in which GCs and DACs are newly generated, then the ratio in the peripheral retina of a small and a large fish should be different. Alternatively, if the proportions of these cells change after being generated, then the DAC-GC ratio should be the same in the periphery of all retinas but should change with increasing age of the tissue. Given the staining problems described in the Materials and Methods section, we analyzed the retinal sections rather than wholemount preparations from six different fish. The results showed that, as expected, the absolute densities for both GCs and DACs decreased with increasing eye size, because of the expansion or stretching of retinal tissue (Tables 1 2) . However, the DAC-GC ratio did not change with eye size but with retinal location. We found that the ratio, on average, was highest (approximately 0.62) in the retinal periphery of all fish and lowest (as low as 0.25) in the central retina of large fish (Table 3 , Fig. 5 ). 

For statistical comparison, we divided the retinal sections into five sectors, as described earlier. We used ANOVA to test whether the determined ratios were significantly different from one another and compared the five sectors within each fish size group and each sector across the size groups. The difference between the average ratios within each group was high in the large and medium fish (P < 0.0001) but not significant in the small fish (P > 0.05). Comparing the ratios determined for each retinal sector across size groups revealed a significant difference in the central sector (P = 0.0017). However, in the flanking two sectors, the probabilities were higher (P = 0.125 for the nasocentral and P = 0.101 for the temporocentral sector) and even higher in the peripheral retina (P = 0.143 for the nasal periphery and P = 0.276 for the temporal periphery, indicating no significant difference in these areas). A pair-wise comparison between the peripheral and central areas in each size group revealed indeed a significant difference in the large- and medium-sized fish (t-test, P < 0.01), but not in the small fish. This suggests that relative to the existing GCs the number of DACs decreased over time as retinal areas became located more centrally. Similar results were observed in wholemount preparations (Fig. 4) , although a quantitative comparison between different sized animals was not performed due to insufficient retrograde tracing or antibody penetration. 

Discussion

In this report we focused on cells in the GCL of the cichlid fish retina. As has been shown for other cell types ^{3 5 24} and is expected as a result of tissue expansion, the densities of DACs and GCs decreased in the growing eye. Tissue stretching or expansion, however, did not change the proportion of cell types. To have a measure of the abundance of DACs independent of eye size, we compared ratios of DACs to GCs, rather than absolute cell densities. We found that this ratio did indeed change in existing retinal tissue in a centroperipheral gradient. It has been reported in several teleost species that the percentage of true GCs in the total number of cells in the GCL changes. ^{25 26} Although one of the studies ²⁶ agrees well with our data, the other study does not. However, that report, ²⁵ mainly concerned with variations in circumferential sectors, showed, in agreement with our data, an overall decrease of GC density, but in disagreement, a decrease of GCs to as low as 35% of cells in the GCL. Both of these studies were based on backfilled GCs and Nissl counterstains, and depended heavily on the reliability of the backlabeling technique. They identified non-GCs as DACs on the basis of the absence of tracer. Combining retrograde labeling and a marker for non-GCs with a nuclear counterstain, we showed that such a changing proportion of GCs to DACs exists. With this technique we could account for virtually all cells in the GCL as labeled either retrogradely or by the amacrine cell marker parvalbumin. Moreover, we compared the ratio of GCs to DACs in the retinas of three different sized groups of fish. The fact that this ratio was not significantly different in the peripheral sectors of all retinal sizes investigated indicates that new retinal cells are produced in constant proportions in the peripheral germinal zone. Therefore, a decreasing gradient in the DAC-GC ratio toward the center of the retina implies that DACs are eliminated from the GCL by some unknown mechanism. 

Amacrine cells displaced into the GCL have been described in many vertebrate species, ^{27 28 29 30} including teleosts. ^{20 31} In most vertebrate species, some subpopulations of inhibitory interneurons in the INL (horizontal and amacrine cells) stain positively for antibodies against parvalbumin, a calcium-binding protein. ³² This has been reported also in fish. ³³ Goldfish DACs are positive for GABA and are considered cholinergic starburst homologues. ²⁰ This was confirmed by our double staining for ChaT and parvalbumin. These cells have conventionally placed homologues that were also weakly positive for both parvalbumin and ChaT (Fig. 2) . The cholinergic system in the fish retina, however, is more complex and includes five cell classes in the goldfish, including a bistratified type. Some have large cell bodies and strong ChaT immunoreactivity. ^{20 34}  

Regarding the system described herein, we can only speculate on the mechanism of the changing cell ratio. Three theoretical possibilities appear to be plausible explanations: transdifferentiation of DACs to other cell types, selective cell death, or migration of DACs out of the GCL. The first possibility, although discussed previously, ³⁵ seems unlikely in the light of recent findings suggesting that the timing of cell cycle exit controls the expression of genes and the activity of factors determining cell fates. ^{36 37} Transdifferentiation would require that these early events be reinitiated to form other cell types. Cell death cannot be excluded as a mechanism; however, the only reports on apoptosis show relatively infrequent labeling for cell death in the retinas of adult fish. Cell death was detected during embryonic development but was absent from the retinal periphery in an adult cichlid ³⁸ and at least in the central retina of adult zebrafish. ³⁹ Nevertheless, apoptosis cannot be ruled out completely, since a relatively rare event in the adult retina would be difficult to detect. The observation of migrating amacrine cells from the GCL late in embryonic development of the rat retina ⁴⁰ supports the third possibility that DACs move through the IPL to the INL. More recently, it has been shown that substantial retinal remodeling occurs in the course of various retinal degenerations including cellular migration of amacrine cells. ⁴¹ However, although this migration was described as bidirectional, it mostly involved amacrine cells repositioned from the INL into the GCL (i.e., in the opposite direction, as described in this report). We did occasionally observe parvalbumin-positive cells in the IPL. Whether these belong to the regularly occurring interstitial amacrine cells in the IPL ⁴² or are possibly migrating cells cannot be determined at present. This intriguing possibility also raises new questions—for example, whether such cells would retain their synaptic connections and/or form new ones in a different environment. 

There have been few reports concerning the specific functions of DACs. Cholinergic DACs in the mammalian retina (starburst amacrine cells) have been shown to participate in the spontaneous rhythmic development occurring in the developing retina. ⁴³ Recently, the starburst amacrine cells have been shown to exhibit directional selectivity computed locally in their dendritic field, which is thought to be the basis for the direction-selective response found in some GCs. ⁴⁴ Thus, a high proportion of direction-selective cells in the retinal periphery would improve the detection of objects moving into the visual field of the eye. Our data suggest that this detection is well maintained in the retinal periphery but is less pronounced and probably less essential in the center of the eye. 

In a set of experiments, Nirenberg and Meister ⁴⁵ ablated DACs in the mouse retina and recorded from GCs. This resulted in a prolonged discharge by the GCs and suggested that the output of GCs is shaped by the action of DACs. In this context, a lower density of DACs relative to the density of GCs in the central retina would mean that the response from GCs is more pronounced (less inhibited) in the center of the retina than in the peripheral areas. Although the mechanism of DAC elimination remains unknown, the results presented herein suggest an adaptation process to the relative shifting of retinal peripheral tissue toward more central areas due to continued growth. 

 Submitted for publication February 23, 2004; revised June 15, 2004; accepted June 24, 2004.

 Disclosure: A.F. Mack, None; C. Süssmann, None; B. Hirt, None; H.-J. Wagner, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Ganglion Cell Density

	Temporal	Temporocentral	Central	Nasocentral	Nasal
Small	15.0 (1.79)	17.6 (5.73)	21.2 (5.36)	29.4 (5.81)	25.2 (5.72)
Medium	10.9 (2.23)	14.2 (3.74)	14.0 (2.98)	13.3 (3.73)	11.7 (1.88)
Large	7.9 (2.05)	7.0 (1.77)	9.83 (2.52)	12.2 (3.24)	11.6 (1.57)

 Absolute, corrected densities of retinal ganglion cells labeled retrogradely and measured on cryostat sections for three sizes of fish. The entire retinal section from nasal to temporal periphery was measured and divided into five equal segments. Within each segment, cells were counted in portions of 300 μm, corrected for optical section thickness and soma sizes and calculated for a retinal area of 1000 μm². Data are expressed as the mean ± SD.

Table 2.

 

Displaced Amacrine Cell Density

	Temporal	Temporocentral	Central	Nasocentral	Nasal
Small	9.1 (1.67)	10.4 (4.74)	10.9 (3.59)	13.6 (5.28)	14.7 (5.22)
Medium	8.5 (3.19)	8.2 (2.95)	5.4 (2.46)	4.7 (1.26)	7.9 (1.44)
Large	5.3 (1.87)	3.1 (0.75)	3.2 (1.3)	5.3 (2.6)	8.0 (1.8)

 Absolute, corrected densities of displaced amacrine cells identified by parvalbumin-immunoreactivity, and measured on cryostat sections for three sizes of fish. The entire retinal section from nasal to temporal periphery was measured and divided into five equal segments which were the same as used for the ganglion cell counts. Cells were counted and data expressed as in Table 1 .

Figure 1.

 

Radial retinal cryostat section stained with antibodies against parvalbumin (green) and nuclear counterstain (in the blue channel; Sytox Green; Molecular Probes). Retinal GCs were retrogradely labeled before fixation with rhodamine dextran (red). In these confocal images, all cells in the GCL were labeled by either rhodamine dextran (GCs) or parvalbumin (DACs).

Figure 2.

 

Confocal scans of a cryostat section from a small fish retina with retrogradely labeled GCs (red) stained with antibodies against parvalbumin (green) and choline acetyl transferase (ChaT, white). Some of the parvalbumin-positive DACs in the GCL were weakly positive for ChaT (top right, large arrows). Dendritic processes positive for both antibody labels were distinctly layered in the IPL and did not overlap except for a small band (bottom right, arrowheads). Except for a few cells weakly positive for ChaT (small arrows), amacrine cells in the INL were not double labeled.

Figure 3.

 

Confocal images of cryostat sections (top) and wholemount preparations (bottom) of retrogradely labeled retinas stained with antibodies against parvalbumin (green). The optical sections in the wholemounts through the GCL and the radial sections showed more DACs (arrows) in the peripheral retina than in the central retinal.

Figure 4.

 

Relationship of DACs to GCs in wholemounted retina from a medium-sized fish (standard length, 5 cm). Left: a montage of confocal scans of GCs taken with a 20× objective. For quantification, DAC and GCs were counted in 100-μm² areas in confocal images scanned with a 40× objective. The size of the columns on the right corresponding to the montage indicates the DAC/GC ratio. D, dorsal; N, nasal; T, temporal; V, ventral.

Table 3.

 

Ratio of Displaced Amacrine Cells to Ganglion Cells

	Temporal	Temporocentral	Central	Nasocentral	Nasal
Small	0.61 (0.10)	0.56 (0.15)	0.53 (0.16)	0.46 (0.17)	0.59 (0.17)
Medium	0.77 (0.23)	0.57 (0.11)	0.38 (0.15)	0.36 (0.09)	0.69 (0.12)
Large	0.70 (0.28)	0.46 (0.14)	0.33 (0.11)	0.41 (0.11)	0.69 (0.16)

 Ratios of displaced amacrine cells to ganglion cells. Data are averages of ratios for corrected cell densities determined for each cell count and are plotted in Figure 5 . Data are expressed as the mean ± SD.

Figure 5.

 

Left: schematic diagrams indicating the size increase of the fish retina; new tissue is added substantially to existing retina in the peripheral growth zone (PGZ, top). The lines indicate that tissue generated by the peripheral growth zone of a small fish becomes located more centrally in larger fish. The actual increase of retinal circumference for the three fish groups used in this study is shown on the bottom left diagram as the average distances from nasal to temporal PGZ in the three groups of fish. The increase is due to tissue stretching and cell proliferation to about equal amounts. ⁴⁶ The hatched areas indicate the estimated addition of retinal area due to tissue stretching. The area within the connecting lines represents expansion of the existing retina, the area outside the lines increase by proliferative tissue addition. Right: DAC-to-GC ratio for the three sizes of fish and five retinal regions, determined on cryostat sections (the counts of at least 160 GCs and corresponding numbers of DACs were included in any of the histograms). ANOVA revealed significant differences among the retinal regions within the medium and large fish (P < 0.0001) but not in the small fish (P > 0.05). ANOVA across fish groups and within retinal regions showed a significant difference only in the central region (rectangle; P < 0.005). Pair-wise comparison of central to peripheral areas in large and medium fish showed a significantly lower ratio in the central than in peripheral, young retinal tissue (*P < 0.01, Student’s t-test). Error bars, SD.

The authors thank Barbara Wallenfels-Thilo for technical assistance and the staff of the medical biometry institute for statistical advice.