Ontogeny of cone photoreceptor mosaics in zebrafish

Abstract

Cone photoreceptors in fish are typically arranged into a precise, reiterated pattern known as a "cone mosaic." Cone mosaic patterns can vary in different fish species and in response to changes in habitat, yet their function and the mechanisms of their development remain speculative. Zebrafish (Danio rerio) have four cone subtypes arranged into precise rows in the adult retina. Here we describe larval zebrafish cone patterns and investigate a previously unrecognized transition between larval and adult cone mosaic patterns. Cone positions were determined in transgenic zebrafish expressing green fluorescent protein (GFP) in their UV-sensitive cones, by the use of multiplex in situ hybridization labelling of various cone opsins. We developed a "mosaic metric" statistical tool to measure local cone order. We found that ratios of the various cone subtypes in larval and adult zebrafish were statistically different. The cone photoreceptors in larvae form a regular heterotypic mosaic array; i.e., the position of any one cone spectral subtype relative to the other cone subtypes is statistically different from random. However, the cone spectral subtypes in larval zebrafish are not arranged in continuous rows as in the adult. We used cell birth dating to show that the larval cone mosaic pattern remains as a distinct region within the adult retina and does not reorganize into the adult row pattern. In addition, the abundance of cone subtypes relative to other subtypes is different in this larval remnant compared with that of larvae or canonical adult zebrafish retina. These observations provide baseline data for understanding the development of cone mosaics via comparative analysis of larval and adult cone development in a model species.

Figure 1

Schematic of the planar mosaic arrangement of cone photoreceptors in zebrafish: a heterotypic mosaic of cone subtypes organized in a precise, reiterated row pattern. Four cone photoreceptor subtypes are present, including UV (magenta), B (blue), G (green), and R (red) in a precise ratio: twice as many R or G cones relative to UV or B cones, and equal numbers of R and G cones, and equal numbers of B and UV cones. The spatial arrangement is highly stereotyped, with alternating rows of R/G double cones and B/UV single cones. The starbursts represent proliferating cells that give rise to new cone photoreceptors throughout the life of the fish in the marginal germinal zone -- an annulus orthogonal to the cone rows and at the boundary between neural retina and ciliary epithelium.

Figure 2

Cone photoreceptor subtype positions in a retina from an adult zebrafish determined by markers for opsin expression. A. Triple-label multiplex in situ hybridization using riboprobes against blue-, green- and red-sensitive opsins on a transgenic retina expressing GFP in UV-sensitive cones. Fluorescent signals are pseudo-coloured cyan, green, red and magenta, respectively. Panel B is the same field as A with only UV and B channels; D is the same field as A with only the G and R channels. C. Excel plot of experimental data presented in panel A. Note the central retina (i.e., optic nerve head) is to the left in this image and the retinal periphery is to the right, such that the rows of cones are slightly further apart towards the periphery. Panel C and D are replicated as Supplemental Figure in magenta-green. Scale bar represents 50 μm.

Figure 3

Ratios of cone photoreceptor subtypes in various stages of zebrafish development. The canonical row mosaic of adult zebrafish predicts ratios of cones (black bars and dotted lines, e.g. see Figure 1). The ratios observed in adults are not significantly different from expected (χ², p> 0.05). Cone ratios in larvae are significantly different from expected (χ², p< 0.03), with an excess of UV cones relative to other types, and an excess of B cones relative to R or G cones compared to that expected from the canonical row mosaic. Cone ratios within the region of retina surrounding the optic nerve head that was generated during larval development (the larval remnant) are also different from expected (χ², p< 0.03). See also Table 1. Data is represented as ratios ± standard deviation.

Figure 4

Cone photoreceptor subtype positions in a retina from a larval zebrafish (four days post-fertilization) determined by markers for opsin expression. A. Double-label in situ hybridization using riboprobes against blue- and greensensitive opsins on a transgenic retina expressing GFP in UV-sensitive cones. Fluorescent signals are pseudo-coloured cyan, green, and magenta, respectively. The areas not occupied by fluorescent signal can be expected to predominantly contain red-sensitive cones and a few rods (data not shown). Panels B thru F are the same field as A, with only subsets of the channels displayed. Panel B displays B and G channels; Panel C displays UV and B channels; Panel D displays UV and G channels; Panel E displays the G channel; Panel F displays the B channel. Figure 4 is replicated as the Supplemental Figure 2 with the same data represented in magentagreen. Scale bar represents 50 μm.

Figure 5

The mosaic metric index reveals local cone patterns in adult retinas consistent with the canonical row pattern, whereas larval retinas show more variability. A. The mosaic metric catalogues the identity of the six cones closest to each B cone in the sample by first finding the distance from a B cone to every cone (lines with empty circles or arrowheads) and indentifying the six closest cones (yellow circles). This is then reiterated for every B cone and data is compiled for comparison to expected values. The samples are labelled for B, UV and G cones only, thus R cones are represented by dashed circles. For the canonical row mosaic we expect the result to be 2 UV cones, 0 B cones and 4 G cones. B. Mosaic metric analysis: Cone patterns in adult retinas are not significantly different from expected (χ², p>0.25; 17 images from n=3 fish). Cone patterns in a larval retina are significantly different from expected (χ², p<0.025, n= 6 fish; labeled on figure by ‘a’). Cone patterns in a remnant of larval retina retained in the adult (n=149 blue cones in one retina) are also significantly different from expected (χ², p<0.0001, noted on figure by ‘b’), although the nature of the difference is distinct in the larvae and larval remnant. Dotted lines note expected values, which are also indicated by the data set on the left of the graph.

Figure 6

Flat-mounted retina demonstrates that the retina that was generated in the larval zebrafish is retained as a ‘larval remnant’, i.e. a relatively disorganized region of cone mosaic in the adult fish. The pattern of UV cones is more regular in the retina that was generated later, by continuous retinal growth in the adult. This transgenic zebrafish (UV cones labeled with GFP) was treated with BrdU at 7 dpf and sacrificed when 3-months-old. The lack of a row pattern of UV cones within and immediately adjacent to the ring of BrdU indicates that this ‘larval remnant’ of retina remains less organized in the adult fish. Inset shows location of the main image (gray dashed lines) within the entire retinal flat-mount.

Figure 7

Flat-mounted retina demonstrates that the row pattern appears in the retinal region that is generated after larval stages. This transgenic zebrafish (UV cones labeled with GFP, pseudocoloured magenta) was treated with BrdU at 20 dpf, and IdU at 36 dpf, and sacrificed when 8-months-old. A. Rings of BrdU (blue) or IdU (yellowish-green) mark the extent of the retina at the time of treatment; the retina within the BrdU ring was generated when the fish was younger than 20 dpf and includes the optic nerve head (onh). B. UV cones expressing GFP (pseudocoloured magenta) are aligned in regular rows throughout most of the retina. C. Merge of UV cone and thymidine analogues reveals timing of change in cone pattern. D. Higher magnification view: UV cones begin to be patterned into rows between the BrdU and IdU rings. The pattern of UV cones was transiently disrupted by IdU treatment (arrowhead in B) however it is apparent that the cones were patterned prior to this (D). Scale bar represents 500 μm.