(A) The gene structure of zebrafish nrl and the CRISPR-Cas9 target site used for knocking out nrl are shown. Orange boxes, the exons encoding the MTD and bZIP domains; red arrow, the CRISPR-Cas9 target site; black arrows, the primers used for mutation detection. (B) Sequencing validation of the homozygous nrl del8 mutation (c.230_237del8). (C) The nrl mRNA levels in the nrl-KO zebrafish were measured using qPCR. The data are shown as mean with SD (n = 3). *, p < 0.05. (D) Immunostaining of retinal sections from WT and nrl-KO zebrafish from 1 mpf to 18 mpf with the anti-Rho antibody for the rod outer segments. Scale bars: 25 μm. ROS, rod outer segment; CN, cone nuclear layer; RN, rod nuclear layer; INL, inner nuclear layer; GCL: ganglion cell layer.

(A) The rod outer segments were visualized via immunostaining on the flattened whole-mount retinas from WT and nrl-KO zebrafish at 2 mpf using the anti-Rho antibody. The overall views of WT and nrl-KO retinas are shown in the low-magnification images. Scale bar: 200 μm. (B) High-magnification images from the dorsal and ventral retinal regions are shown. The density of rods is reduced in the nrl knockout retinas, especially in the ventral retinal region. Scale bar: 50 μm.

Rods were labeled with EGFP by crossing the WT or nrl-KO zebrafish with the Tg(rho:EGFP) transgenic line. Fluorescence was observed every 2 days from 3 dpf. Representative images of WT and nrl-KO retinas are shown in (A) for 3–9 dpf and (B) for 11–17 dpf. The dotted circles indicate the boundaries of the retinas. D, dorsal; V, ventral; N, nasal; T, temporal. Scale bars: 100 μm. (C) Immunostaining of retinal sections for Gnat1 (the rod transducin alpha-subunit) showed exact co-localization with EGFP in rods at 7 dpf. Rods were barely observed in the nrl-KO retinas. Scale bar: 50 μm.

(A) The Tg(rho:EGFP) transgenic line was used to label rods with EGFP. Representative images of flattened whole-mount retinas from WT and nrl-KO zebrafish at 1 mpf, 2 mpf, and 3 mpf are shown. The dashed lines indicate the edges of the retinas. Scale bars: 200 μm. (B) EGFP-labeled rods were observed on the retinal sections from WT and nrl-KO Tg(rho:EGFP) transgenic zebrafish at 1 mpf. The overall views are shown in the upper panel. The regions nearby the ciliary marginal zone (labeled with boxes) showed no or little fluorescence signal of rods. Enlarged images of the dorsal retinal regions are shown in the lower panel. ROS, rod outer segment; CN, cone nuclear layer; RN, rod nuclear layer; INL, inner nuclear layer; GCL: ganglion cell layer. Scale bars: 50 μm.

(A) The volcano plot shows the 386 differentially expressed genes (196 up-regulated and 190 down-regulated) between WT and nrl-KO retinas at 2 mpf identified via RNA-seq. The red and blue points indicate the up-regulated and down-regulated genes, respectively. (B) The functional categories enriched among the top 100 differentially expressed genes. (C) The expression patterns of rod- and cone-specific phototransduction genes in nrl-KO retinas shown as a heatmap. Most of the rod genes were down-regulated in the nrl-KO retinas.

(A) The protein levels of rod-specific (rho, gnat1, and gnb1) and cone-specific (gnat2 and gnb3) genes in WT and nrl-KO retinas from 14 dpf to 3 mpf were evaluated using western blotting. Tubulin was used as a loading control. The asterisk indicates a non-specific band. (B) Quantitative analysis of the protein levels of rod- and cone-specific genes based on at least three independent experiments. WT samples, blue color; nrl-KO samples, red color. Data from different ages are arranged from left to right (from 14 dpf to 3 mpf) and indicated by different fill patterns. The data are shown as mean with SD (n = 3). *, p < 0.05. **, p < 0.01. (C) The mRNA levels of the rod and cone opsins in the WT and nrl-KO retinas at 2 mpf and 5 mpf were measured using qPCR. The data are shown as mean with SD (n = 3). ns, non-significant; *, p < 0.05; **, p < 0.01. (D) Detection of green-cones on retinal sections of WT and nrl-KO zebrafish at 2 mpf, 3 mpf, and 13 mpf by immunostaining using the anti-Opn1mw antibody. The dorsal retinal regions are shown. White arrows indicate the mislocalized green-cone outer segments. GCOS, green-cone outer segments; CN, cone nuclear layer; RN, rod nuclear layer; INL, inner nuclear layer. Scale bars: 50 μm.

(A) tSNE visualization of the unsupervised cell clusters from 5-month-old WT and nrl-KO zebrafish. Left, the distribution of cell clusters between the WT and nrl-KO groups. Right, the retinal cell types identified via scRNA-seq (see also S6 Fig). BC, bipolar cells; Cone_R/G, red and green cones; Cone_B/UV, UV and blue cones; RGC, retinal ganglion cells; HC, horizontal cells; AC, amacrine cells; RPE, retinal pigment epithelium. (B) Proportions of each retinal cell type in the WT and nrl-KO zebrafish. (C) The cell compositions and proportions of photoreceptor sub-clusters in WT and nrl-KO retinas. (D) The heatmap shows the clustering pattern of the top 20 down-regulated and top 20 up-regulated genes in the WT and nrl-KO rods. Yellow, high expression; purple, low expression. (E) Mis-expression of green-cone opsin in a proportion of rods in the nrl-KO retinas. The dorsal retinal regions of WT and nrl-KO zebrafish at 3 mpf are shown. Green, EGFP-labeled rods. Red, immunofluorescence signals of the anti-Opn1mw antibody. The green fluorescent signal in the nrl-KO group was artificially enhanced to discern the morphology of rods. ROS, outer segments of rods; GCOS, outer segments of green cones; RN, rod nuclear layer. Scale bar: 20 μm.

(A) Clustering analysis of the large MAF genes and the genes involved in photoreceptor development by using the expression data from scRNA-seq. The cell-type-specific expression pattern of mafba was highly similar to those of nrl and nr2e3. (B) The distribution of rods (labeled with EGFP) in the nrl-KO zebrafish carrying the WT, heterozygous, and homozygous mafba alleles at 9 dpf (left panel) and 15 dpf (right panel). Knocking out mafba further reduced the genesis of rods in the nrl-KO zebrafish. The dotted circles indicate the boundaries of the retinas. Scale bars: 100 μm. (C) The expression levels of nrl, nr2e3, and mafba in the nrl-KO retinas at 2 mpf and 5 mpf were measured via RNA-seq and qPCR. The data are shown as mean with SD (n = 3). *, p < 0.05; **, p < 0.01.

The genes nearby mafba and their relative orders were extracted from the genomes of a wide range of vertebrates, including ancient fishes (sea lamprey, elephant shark, spotted gar, and coelacanth), modern fishes (European eel, zebrafish, Japanese medaka, and torafugu), amphibians (tropical clawed frog), reptilians (American alligator), birds (chicken), and mammals (house mouse and human). The mafba and mafbb genes appear to have originated from the ancestral gene mafb through genome duplication specifically in teleost fishes.

(A) The TUNEL assay results revealed that multiple types of retinal cells, including rods, cones, RPE, and inner retinal cells, undergo apoptosis in the nrl-KO retinas. Representative images are shown. Scale bar: 20 μm. (B) Quantitation of the apoptotic cells per section from 5 mpf to 13 mpf. The results are shown as mean with SD (n = 3). *, p < 0.05; **, p < 0.01. (C) RPE morphology in the WT and nrl-KO zebrafish at 20 mpf are shown, as assessed by immunostaining the retinal whole-mounts for ZO-1 Scale bar: 20 μm. (D) The up-regulation of GFAP in nrl-KO retinas as detected by immunostaining. CN, cone nuclear layer; RN, rod nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar: 25 μm. (E) Regeneration of photoreceptors was more active in the nrl-KO retinas at 9 mpf and 18 mpf, as reflected by the increase in the number of Pcna-positive cells (proliferating cells) in the ONL, compared with the WT levels. Scale bars: 50 μm. (F) Quantitation of the Pcna⁺ cells located in the ONL per section. The results are shown as mean with SD (n = 4) from 7 dpf to 29 mpf. *, p < 0.05; **, p < 0.01.

There may be two types of rod precursors defined as nrl-dependent and nrl-independent in the zebrafish retinas. The former is responsible for the production of rods at embryonic stage, and knocking out nrl abolishes the differentiation into rods (left panel). The latter is responsible for the production of rods at the juvenile and adult stages. In the presence of nrl and mafba, these cells differentiate into rods. Knocking out nrl does not severely affect the production of rods from these cells but causes abnormal expression of rod- and cone-specific genes and a progressive increase of green cones (M-cones) with age. Knocking out both nrl and mafba eliminates all rods, suggesting that mafba also plays an important role in the fate determination of this type of rods (right panel).

Unillustrated author statements