Samd7 represses short-wavelength cone genes to preserve long-wavelength cone and rod photoreceptor identity

Significance

Color vision depends on the expression of spectrally distinct opsins across multiple photoreceptor subtypes. Here, we show that the transcriptional corepressor samd7 acts in both zebrafish and mice to preserve mutually exclusive patterns of opsin expression. In the absence of samd7, both species show defects in rod differentiation and aberrant expression of short-wavelength-sensitive cone opsin in long-wavelength-sensitive cones. Zebrafish additionally show loss of green cones and a doubling of the number of blue cones. Together, these studies reveal rod- and cone-subtype-specific functions of samd7 in zebrafish and mouse retina.

Abstract

The role of transcription factors in photoreceptor gene regulation is fairly well understood, but knowledge of the cell-type-specific function of transcriptional cofactors remains incomplete. Here, we show that the transcriptional corepressor samd7 promotes rod differentiation and represses short-wavelength cone genes in long-wavelength cones in zebrafish. In samd7^−/− retinas, red cones are transformed into hybrid red/ultraviolet (UV) cones, green cones are absent, the number of blue cones is approximately doubled, and the number of rods is greatly reduced. We also find that mouse Samd7 represses S-opsin expression in dorsal M-cones—analogous to its role in repressing UV cone genes in zebrafish red cones. Thus, samd7 plays a key role in ensuring appropriate patterns of gene expression in rods and cone subtypes of both zebrafish and mice.

Color vision depends on the differential activation of two or more cone types tuned to different wavelengths of light (1). Differential spectral tuning among cone types is therefore a critical feature of vertebrate retinal development and function. The retina of the common vertebrate ancestor is believed to have possessed one rod type, defined by the expression of rhodopsin (RH1), and four cone types expressing UV opsin (SWS1), blue opsin (SWS2), green opsin (RH2), and red opsin (LWS) (2, 3). Zebrafish (Danio rerio) retain a photoreceptor repertoire similar to that of the ancestral vertebrate, with each photoreceptor type expressing distinct opsins and phototransduction components (4–7). Studies in fish have identified multiple transcription factors (TFs) that regulate the specification and diversification of rods and cone types, most notably nrl, mafba, nr2e3, foxq2, tbx2a, tbx2b, thrb, gdf6a, six6a, six6b, and six7 (4, 8–20). Studies in mice have also defined a range of TFs controlling photoreceptor gene expression and subtype specification (e.g., Otx2, Crx, Rora, Rorb, Nrl, Nr2e3, Casz1, Esrrb, Thrb, Rxrg, NeuroD1, Nr2f1/2 (CoupTFI/II), Pias3, Sall3, Bmal1, and Clock) (21–35). Mutations in vertebrate photoreceptor TFs typically cause one of three phenotypes: 1) gene expression dysregulation, often resulting in cell death; 2) partial cell fate transformation, resulting in formation of a “hybrid” cell type which coexpresses genes of two photoreceptor types; or 3) complete transfating of one photoreceptor subtype into another cell type. For example, mutations in mouse Nr2e3 transform rods into hybrid cells that coexpress rod and a subset of S-cone genes, while mutations in Nrl cause complete conversion of rods into S-cones (23–25). Suppression of alternative photoreceptor cell fates (or gene expression programs) is a common function of cell-type-enriched TFs and may have been a key driver of photoreceptor subtype diversification in evolution (12–14, 24, 25).

In the zebrafish retina, suppression of alternative photoreceptor gene expression programs is necessary to maintain the separate identities of five distinct photoreceptor types (6). In contrast to fish, placental mammals possess a reduced number of photoreceptor types due to evolutionary loss of ancestral blue (SWS2) and green (RH2) cones (7). Most present-day placental mammals therefore possess only three photoreceptor types: rods, UV (SWS1) cones, and red (LWS) cones (7). While the vast majority of mammalian species show mutually exclusive expression of UV and red opsin in the two cone types, the situation in mouse is more complex. Mouse “primordial S-cones” (which exclusively express SWS1 opsin) are present throughout the retina but are concentrated in ventral regions (36, 37). Mouse “M-cones,” in contrast, exclusively express LWS opsin (“M-opsin”) in the dorsal third of the retina but show coexpression of LWS (“M-opsin”) and SWS1 opsin (“S-opsin”) in ventral regions (36). Previous studies showed that S-opsin expression is up-regulated in the dorsal M-cones of multiple mouse mutants (28–30, 34, 35), demonstrating that the pattern of cone opsin coexpression is under tight genetic control.

A recent study in mice suggested that Samd7 might play a key role in these processes. In that paper, the authors showed that mutations in Samd7 cause upregulation of a subset of cone genes in rods, including S-opsin (38). Samd7 lacks a DNA-binding domain and is thought to act as a transcriptional corepressor by binding the rod-specific TF Nr2e3 and recruiting the polycomb repressive complex 1 (PRC1) to cis-regulatory regions of S-cone genes via its sterile alpha motif (SAM) domain (38). Whether Samd7 plays a comparable role in maintaining cell identity in cones is currently unknown.

To better understand the role of Samd7 in regulating cone photoreceptor cell fate specification and diversification, we created samd7^−/− zebrafish and evaluated effects on photoreceptor subtype composition and gene expression. We found that samd7 plays a critical role in the specification of red cones, green cones, and rods by regulating distinct and overlapping patterns of gene expression in these cell types. In the samd7^−/− retina, red cones are transformed into hybrid red/UV cones, green cones are absent, blue cones are approximately doubled, and most rods fail to differentiate. We also generated Samd7^−/− mice and analyzed patterns of cone opsin expression in the retina. We found that dorsal M-opsin⁺ cones are transformed into hybrid M/S-cones, resembling the transformation of red to red/UV cones we observe in samd7^−/− zebrafish. Overall, Samd7 acts to maintain rod photoreceptor identity and suppress short-wavelength-sensitive cone gene expression in long-wavelength-sensitive cones, thereby preserving normal photoreceptor function.

Results

Samd7 Is Expressed in Immature Photoreceptors and Mature Red Cones, Green Cones, and Rods.

Previous studies showed samd7 to be enriched in red cones, green cones, and rods in adult zebrafish retina (4, 5). To determine the expression pattern of samd7 in larval retina, we used fluorescence-activated cell sorting (FACS) to isolate GFP-expressing cells from the retinas of 4-days-post-fertilization (dpf) crx:GFP transgenic zebrafish, which express GFP in developing (immature) and mature photoreceptors and bipolar cells. We then subjected these cells to single-cell RNA sequencing (scRNA-seq). Next, we identified a cluster of immature photoreceptors based on expression of prdm1a/b, a marker of early differentiating photoreceptors, and classified photoreceptor subtypes based on cone opsin (opn1sw1, opn1sw2, opn1mw1, and opn1lw2) and rod opsin (rho) expression (SI Appendix, Fig. S1) (39, 40). We found that samd7 is expressed in immature photoreceptors and mature rods, green cones, and red cones but not in UV or blue cones (Fig. 1 A and B).

Fig. 1.

Samd7 is expressed in immature photoreceptors, red cones, green cones, and rods in the larval zebrafish retina. (A) UMAP (Uniform Manifold Approximation and Projection) of scRNA-seq data from 4-dpf crx:GFP⁺ photoreceptors. (B) scRNA-seq results show that samd7 is restricted to immature photoreceptors, red cones, green cones, and rods. See SI Appendix, Fig. S1 for detailed expression patterns. (C–K) Analysis of samd7 expression in the 3-dpf larval eye by fluorescent in situ hybridization. (C) Optical cross-section of the peripheral larval retina stained with probes against samd7 (green) and red/green-cone-specific arr3a (red) shows that samd7 is expressed in immature arr3a⁻ photoreceptors (“immature photoreceptors”). (D) Whole-mount view of the eye in C. (E) Close-up view of the dotted square in D shows samd7⁺ puncta in arr3a⁺ red and green cone cell bodies. (F) Optical cross-section of the peripheral larval retina stained with probes against samd7 (green) and blue/UV-cone-specific arr3b (red). (G) Whole-mount view of the eye in F. (H) Close-up view of the dotted square in G shows that samd7 is not expressed in arr3b⁺ blue and UV cone cell bodies. (I) Optical cross-section of the peripheral larval retina stained with probes against samd7 (green) and rod-specific gnat1 (red). (J) Whole-mount view of the same eye as in I. (K) Close-up view of the dotted square in J shows samd7⁺ puncta in gnat1⁺ rod cell bodies. n = 3 eyes analyzed per in situ. [Scale bar in (C, D, F, G, I, and J), 50 µm; Scale bar in (E, H, and K), 5 µm.] D, V, N, T: Dorsal, Ventral, Nasal, Temporal. ONH; optic nerve head. ONL; outer nuclear layer.

To confirm this pattern of samd7 expression, we performed fluorescent in situ hybridization at 3-dpf, a time point at which immature and mature photoreceptors can be readily observed in whole-mount retinal preparations (Fig. 1 C–K). We found that samd7 expression is restricted to the photoreceptor layer (outer nuclear layer; ONL), confirming that samd7 is expressed exclusively in photoreceptors (Fig. 1 C, F, and I). We also observed samd7 expression in the dorsal retina, where photoreceptors have yet to initiate arr3a and arr3b expression, indicating that samd7 is expressed in immature photoreceptors (Fig. 1 C and F). To define the mature photoreceptor subtypes in which samd7 is expressed, we evaluated samd7 coexpression in arr3a⁺ green and red cones, arr3b⁺ UV and blue cones, and gnat1⁺ rods. We found samd7 to be expressed in arr3a⁺ green and red cones (Fig. 1 D, Inset; Fig. 1E) and excluded from arr3b⁺ UV and blue cones (Fig. 1 G, Inset; Fig. 1H). In addition, we found samd7 to be expressed in gnat1⁺ rods (Fig. 1 J, Inset; Fig. 1K). These results demonstrate that samd7 is expressed in immature photoreceptors, red cones, green cones, and rods at larval stages.

Loss of samd7 Results in Subtype-Specific Changes in Photoreceptor Gene Expression.

To elucidate the role of samd7 in zebrafish photoreceptor development, we engineered a mutation in the gene. Using CRISPR-Cas9 technology, we created a 10 bp deletion in the fourth exon of samd7, which is predicted to result in a frameshift and premature stop codon upstream of the SAM domain (SI Appendix, Fig. S2). To evaluate the mutation’s effects on gene expression, we performed RNA-seq on the eyes of clutch-matched 5-dpf samd7^−/− and wild-type (WT) control larvae. We found that only 90 genes were significantly dysregulated in the samd7^−/− eye compared to controls (p-adj < 0.05; Dataset S1). We first focused on changes in red cone, green cone, and rod gene expression since these are the cell types in which samd7 is expressed. Strikingly, we observed a nearly complete loss of green opsin (both larva-expressed green opsin paralogs, opn1mw1 and opn1mw2, are absent) (41) as well as loss of the red/green-cone-specific cone arrestin, arr3a (5) (Fig. 2 A and C and Dataset S1). These findings suggest that green cones are absent from the eye of samd7^−/− larvae. Aside from loss of arr3a expression, we observed only subtle changes in red-cone-specific gene expression. Specifically, we saw an ~twofold reduction in red-cone-specific si:busm1-57f23.1 (5, 14) transcripts and a sixfold increase in opn1lw1, one of the two tandemly arrayed zebrafish red opsin paralogs (6) (Fig. 2 B and C). In addition to these changes, we observed a striking reduction in rod-specific gene expression (Fig. 2B). Among the 66 significantly down-regulated genes in the mutant eye, 28 were rod-enriched. On average, the expression of individual rod genes was reduced by ~70% in the mutant eyes (Dataset S1).

Fig. 2.

Rod- and green cone-specific gene expression is reduced, and blue and UV-cone-specific gene expression is increased in samd7^−/− eyes. (A) Volcano plot of bulk RNA-seq data from samd7^−/− vs. WT 5-dpf eyes demonstrates a marked reduction in the expression of red/green-cone-specific arr3a and green-cone-specific opsins opn1mw1 and opn1mw2. (B) A magnified view of the dotted portion of panel A demonstrates a marked reduction in the expression of multiple rod-specific genes, a subset of which are annotated. The expression of multiple blue and UV cone-specific genes is increased in the mutant, including blue/UV-cone-specific arr3b and blue opsin (opn1sw2). UV opsin (opn1sw1) appeared somewhat increased, but the change is not statistically significant (p-adj = 0.07, Dataset S1). Among red-cone-specific genes, si:busm1-57f23.1 is decreased, and opn1lw1 is increased. Exorh, which is normally restricted to a subset of pineal photoreceptors, is also increased (42). (C) Bar plots of opsin expression (DESeq2-normalized counts) from RNA-seq data in panels A and B (mean ± SD; n = 3 per group) *p-adj < 0.05. ns, not significant. SD, standard deviation.

We also observed an increase in blue- and UV-cone-specific gene expression in the samd7^−/− eye. Among blue cone-specific genes, we detected an increase in blue cone opsin (opn1sw2) expression, as well as a corresponding increase in foxq2, a TF required for blue cone fate specification (11) (Fig. 2 B and C). Among UV cone-specific gene expression changes, we observed an increase in tbx2a, a TF that regulates the expression of UV cone opsin (opn1sw1) (4). There was also a suggestion of increased opn1sw1 expression in the mutant, but this change was not statistically significant (p-adjusted = 0.07) (Fig. 2 B and C and Dataset S1). We also observed an increase in arr3b and skor1a, which are coexpressed in both blue and UV cones (4, 5) (Fig. 2B). Overall, the increases in blue and UV-cone-specific gene expression ranged from 1.5 to 2-fold (Dataset S1), suggesting that the total number of blue and UV cones may be increased by a corresponding amount. We also observed an increase in the expression of exorh, a homolog of rho normally restricted to a subset of photoreceptors in the zebrafish pineal gland (42) (Fig. 2 B and C). We found that samd7 transcript levels are also increased in the samd7^−/− eye, raising the possibility that Samd7 represses its own expression (Fig. 2B).

Loss of samd7 Causes Changes in Photoreceptor Cell Fate.

To determine whether the changes in gene expression in the samd7^−/− retina are due to changes in the proportions of photoreceptor types, we used a combination of antibody staining and transgenic reporter expression to quantify rods and cone subtypes in 5-dpf clutch-matched samd7^−/− and WT zebrafish. We found that there were approximately twice as many UV opsin⁺ outer segments in the central retina of samd7^−/− compared to WT (Fig. 3 A–E). Similarly, we found about twice as many blue opsin⁺ outer segments in the central retina of samd7^−/− compared to WT (Fig. 3 F–J). The increase in UV and blue opsin⁺ cells is commensurate with the ~twofold increase in UV and blue cone-specific gene expression in the samd7^−/− eye (Fig. 2 B and C and Dataset S1). We were unable to detect any green opsin⁺ outer segments in the samd7^−/− retina (Fig. 3 K–O), consistent with the lack of green opsin expression we observe in the mutant eye by RNA-seq (Fig. 2A and Dataset S1). The number of red cones (thrb:tdTomato⁺) was unchanged in the mutant (Fig. 3 P–T).

Fig. 3.

The numbers of blue and UV cones are approximately doubled, whereas green cones and rods are severely reduced in samd7^−/− larvae. Confocal whole-mount images and high magnification insets of 5-dpf WT and samd7^−/− eyes. (A–E) The number of UV cones stained with the UV opsin antibody is approximately doubled. (F–J) The number of blue cones stained with the blue opsin antibody is approximately doubled. (K–O) Green cones, as stained with the green opsin antibody, are absent from the samd7^−/− eye. (P-T) The number of red cones, as identified with the thrb:tdTomato transgene, is unchanged, although the percentage of red cones that are directly contiguous appears to be increased in the mutant. (U and V) There is a severe reduction in the number of rods stained with the Gnat1 antibody, although a small population of rods with reduced Gnat1 signal can be observed in the ventral samd7^−/− retina. (mean ± SD; n = 3 per group). All fields of view quantified in (E, J, O, and T) were 40 × 40 µm². Scale bars in all close-up views (B, D, G, I, L, N, Q, and S) = 10 µm. [Scale bar in (U and V), 100 µm.] D, V, N, T: Dorsal, Ventral, Nasal, Temporal. Statistical comparisons were performed using the two-tailed, unpaired t test assuming unequal variance. *P < 0.05, **P < 0.005. ns, not significant. SD, SD.

To evaluate changes in the number of rods in samd7^−/− larvae, we characterized the distribution of Gnat1⁺ cells in the retina. In WT retina, Gnat1⁺ rods are highly concentrated in the ventral-most region and more sparsely distributed throughout other regions (Fig. 3U). In the samd7^−/− retina, Gnat1⁺ cells were absent from all regions except the ventral-most retina, in which the intensity of Gnat1 labeling was reduced relative to WT (Fig. 3V). This severe reduction in the number of Gnat1⁺ rods correlates with the ~80% reduction in gnat1 expression we observe in the samd7^−/− eye (Fig. 2B and Dataset S1). To determine whether cell death might account for the reduction in rods at 5-dpf, we characterized rho:GFP expression in clutch-matched samd7^−/− and WT zebrafish at 3-dpf, the earliest time point at which rods can be observed throughout the larval retina (43). We found that at 3-dpf the majority of rods (rho:GFP⁺) are already absent from the central and dorsal samd7^−/− retina, although the ventral patch of rods remains (SI Appendix, Fig. S3 A and B). These results suggest that rods fail to be born in the samd7^−/− retina. To determine whether the remaining population of ventral rods might be transformed into hybrid rod/UV cones, as occurs in the Samd7^−/− mouse retina, we stained 5-dpf WT;rho:GFP⁺ and samd7^−/−;rho:GFP⁺ eyes for UV opsin (38). We found that neither WT nor samd7^−/− rods express UV opsin (SI Appendix, Fig. S3 C–J). Taken together, our results suggest that the vast majority of rods fail to differentiate at 3-dpf and that the remaining ventral population does not adopt a rod/UV-hybrid phenotype, in contrast to the hybrid phenotype observed in Samd7^−/− mouse rods (38). However, it remains possible that some rod precursors are transfated to UV cones in the central/dorsal samd7^−/− retina, but are undetectable with available rod-specific markers.

In summary, we find that both the number of UV opsin⁺ and blue opsin⁺ cones are approximately doubled, green opsin⁺ cones are absent, the number of red cones is unchanged, and the number of rods is severely reduced in the samd7^−/− retina. Next, we sought to determine the identity of the supernumerary UV opsin⁺ cells in the samd7^−/− retina. Based on the expression of samd7 in red cones, green cones, and rods, we hypothesized that loss of samd7 could result in the upregulation of UV opsin in one of these cell types. To test this hypothesis, we first characterized the expression of UV opsin in red cones (thrb:tdTomato⁺ cells) in clutch-matched samd7^−/− and WT 5-dpf larvae. As expected in the WT retina, we detected no UV opsin in red cones (Fig. 4A). In contrast, we observed strong UV opsin staining in samd7^−/− red cones (Fig. 4B). Similarly, opn1sw1:GFP⁺, a UV-cone-specific transgene, was up-regulated in nearly all red cones in the samd7^−/− retina (Fig. 4 C and D). These results suggest that red cones are transformed into hybrid red/UV cones in the samd7^−/− retina. When red-cone-derived UV cones (thrb:tdTomato⁺;opn1sw1:GFP⁺) are excluded from our analysis, we still observe a statistically significant increase in the number of UV cones in the samd7^−/− retina (Fig. 4E), suggesting that a small subset of supernumerary UV cones in the samd7^−/− retina derive from another unknown cell type.

Fig. 4.

Red cones are transformed into hybrid red/UV cones in the larval samd7^−/− retina. (A–D) Confocal images of 5-dpf WT and samd7^−/− eyes show that UV opsin and UV-cone-specific opn1sw1:GFP is up-regulated in samd7^−/− red cones (thrb:tdTomato⁺). (E) UV cones (opn1sw1:GFP⁺ cells per 1,600 µm²) are increased in the samd7^−/− retina. The subset of UV cones that are not derived from red cones (opn1sw1:GFP⁺;thrb:tdTomato⁻) are also increased, suggesting that a small subset of supernumerary UV cones in the samd7^−/− retina derive from another unknown cell type or precursor (mean ± SD; n = 4 WT, n = 6 samd7^−/−). (F) Volcano plot from bulk RNA-seq comparing gene expression in isolated 5-dpf WT thrb:tdTomato⁺ cells (i.e., red cones) and samd7^−/− thrb:tdTomato⁺ cells (i.e., transformed hybrid red/UV cones). UV-cone genes are up-regulated in samd7^−/− hybrid red/UV cones including tbx2a, tbx2b, opn1sw1, arr3b, mir729, tgfa, and grk7b (5). Red-cone-specific changes in gene expression were also observed, including reduced expression of the thyroid-hormone-inactivating enzyme, dio3b; increased expression of the red-shifted red opsin paralog, opn1lw1 (λ_max = 558 nm); and decreased expression of the blue-shifted red opsin paralog, opn1lw2 (λ_max = 548 nm). Similar changes in red opsin paralog expression were previously observed in larval zebrafish treated with thyroid hormone (44). Red/green-cone-specific arr3a expression was absent, and exorh and samd7 expression were increased in samd7^−/− hybrid red/UV cones. (Scale bar, 10 µm.) Statistical comparisons in E were performed using the two-tailed, unpaired t test assuming unequal variance. *P < 0.05 and ***P < 0.0005. SD, standard deviation.

To evaluate the possibility that the supernumerary blue opsin⁺ cells observed in samd7^−/− retina might represent hybrid blue/red cones, we characterized the expression of blue opsin and blue-cone-specific opn1sw2:GFP in red cones (thrb:tdTomato⁺) of WT and samd7^−/− larvae (45). As expected, we found a near doubling of the number of opn1sw2:GFP⁺ cells in the samd7^−/− retina (SI Appendix, Fig. S4E), consistent with the results of blue opsin antibody staining (Fig. 3 F–J). However, we found that neither blue opsin nor the opn1sw2:GFP transgene are expressed in red cones in the samd7^−/− retina (SI Appendix, Fig. S4 A–D), suggesting that supernumerary blue cones are not derived from red cones but may instead arise from another population of samd7⁺ photoreceptors, possibly transfated rods or green cones.

RNA-seq Reveals Hybrid Red/UV Cones in the samd7^−/− Retina.

To further evaluate gene expression changes in samd7^−/− red cones, we used FACS to isolate red cones (thrb:tdTomato⁺ cells) from the eyes of 5-dpf samd7^−/− and WT larvae and compared their transcriptomes by RNA-seq. As expected, we observed an upregulation of UV opsin (opn1sw1) in samd7^−/− red cones relative to WT (Fig. 4F). In addition, we saw upregulation of mir729, tgfa, tbx2a, and grk7b, which were previously shown to be UV-cone-enriched in adult zebrafish photoreceptors (4, 5) (Fig. 4F). We also observed an increase in tbx2b, a UV/blue-cone-enriched transcription factor (4, 5). Both tbx2a and tbx2b had previously been shown to regulate opn1sw1 expression in UV cones (4, 11, 12). We did not detect upregulation of the blue-cone-specific genes opn1sw2 and foxq2 in samd7^−/− red cones (Dataset S2). These results support the conclusion that red cones are transformed into hybrid red/UV cones in the samd7^−/− retina.

We also observed dysregulation of several red-cone-specific genes in samd7^−/− red cones. Red/green-cone-specific arr3a transcripts are essentially absent. In addition, the expression of opn1lw2 [λ_max = 548 nm (6)], the predominant red cone opsin transcript in WT larvae, is reduced; whereas the expression of opn1lw1 [λ_max = 558 nm (6) is correspondingly increased (Fig. 4F and Dataset S2). A similar decrease in opn1lw2 and increase in opn1lw1 was previously observed in zebrafish larvae treated with thyroid hormone (44). Therefore, we hypothesized that thyroid hormone signaling was increased in samd7^−/− red cones. Consistent with this hypothesis, we observed a decrease in the expression of the thyroid-hormone-inactivating enzyme dio3b in samd7^−/− red cones (46) (Fig. 4F and Dataset S2). This decrease in dio3b may result in elevated levels of active thyroid hormone, thereby mimicking changes in opn1lw1/2 paralog expression observed in larvae treated with thyroid hormone. Previous reports also showed that application of thyroid hormone reduced si:busm1-57f23.1 expression, consistent with the reduction in this gene’s expression which we observed in samd7^−/− whole eye by RNA-seq (Fig. 2B and Dataset S1) (47).

We also detect a reduction in green opsin (opn1mw1/2) expression in samd7^−/− red cones (Dataset S2), which we attribute to trace (~4%) contamination of green cones in the WT sorted cells, as detailed in the Methods section. In summary, we find that UV-cone-specific gene expression is up-regulated in samd7^−/− red cones, while blue cone-specific-gene expression (i.e., foxq2 and opn1sw2) is unchanged (Fig. 4F and Dataset S2). In addition, we find that thyroid hormone signaling may be increased, resulting in increased opn1lw1 and reduced opn1lw2 expression (Fig. 4F).

The Adult samd7^−/− Retina Recapitulates Aspects of the Larval Phenotype and Demonstrates a Reduced Electrical Response to Light.

To determine whether the function of samd7 in the adult retina is similar to that in the larval retina, we isolated retinas from 3-mo-old samd7^−/− and WT zebrafish and compared their transcriptomes by RNA-seq. We find changes in gene expression similar to those we observed in the samd7^−/− larvae, with some notable exceptions. Consistent with the larval phenotype, we observe a near absence of all green opsin transcripts (opn1mw1/2/3/4) as well as loss of red/green-cone-specific arr3a from the adult samd7^−/− retina, suggesting that green cones are absent (SI Appendix, Fig. S5A and Dataset S3). We also observe an increase in blue and UV-cone-specific gene expression. Among blue-cone-specific genes, we observe a 1.7-fold increase in foxq2 and a 1.9-fold increase in blue opsin (opn1sw2). Similarly, we find that UV-cone-specific gene expression is increased in the adult mutant retina; opn1sw1, mir729, tbx2a, and tgfa transcript levels are all increased between twofold and fourfold (SI Appendix, Fig. S5A and Dataset S3). We also see an increase in exorh expression, as seen in the larval samd7^−/− eye (Fig. 2B, SI Appendix, Fig. S5A, and Datasets S1 and S3). As in larvae, these changes are consistent with an absence of green cones and increased numbers of blue cones in the mutant, as well as upregulation of UV-cone-specific genes in red cones.

Among red-cone-enriched genes, we find that one red cone opsin paralog (opn1lw2) is reduced by ~97% while the second (opn1lw1) is unchanged in the adult samd7^−/− retina compared to WT (SI Appendix, Fig. S5A and Dataset S3). Transcripts of red-cone-specific thrb, si:busm1-57f23.1, and mir726 are also reduced (SI Appendix, Fig. S5A and Dataset S3). Although red-cone-specific gene expression is generally reduced, the sustained expression of opn1lw1 and the concurrent increase in UV-cone-specific gene expression suggests that red cones assume a hybrid red/UV-cone phenotype, as we observe in the larval samd7^−/− retina (Fig. 4 A–D). In contrast to the larval phenotype, we do not observe downregulation of rod-specific-phototransduction components in the adult samd7^−/− retina, with the exception of saga (SI Appendix, Fig. S5A and Dataset S3). However, we do observe a reduction in nrl expression by ~60%. We also find an adult rod-enriched gene, ppdpfa, a putative signaling molecule of unknown function (5, 48), to be reduced by ~60% (SI Appendix, Fig. S5A and Dataset S3). Conversely, we find that mafba, a TF recently shown to regulate rod specification in conjunction with nrl in the adult zebrafish retina (9), to be increased by ~twofold (SI Appendix, Fig. S5A and Dataset S3). Overall, we find that the changes in rod gene expression in the mutant adult retina and larval eye are distinct; although rod genes represent about half of the dysregulated genes in the larval mutant dataset (Dataset S1), very few of the dysregulated genes we observe in the adult dataset are rod-specific (SI Appendix, Fig. S5A and Dataset S3). These differences may reflect the distinct developmental origin of the embryonic vs. adult zebrafish retina and the apparent differences in the requirement of nrl in the specification of larval and adult rods (8, 9, 49). Thus, samd7 appears to play distinct gene regulatory roles in mediating rod gene expression in the larval vs. adult retina.

Strikingly, we also observed upregulation of a subset of genes associated with Müller-glia-mediated regeneration in the adult samd7^−/− retina. In response to some types of retinal injury, Müller glia reenter the cell cycle, proliferate, and give rise to new neurons, including both rods and cones (50). In the adult samd7^−/− retina, we see an increase in the expression of TFs ascl1a and insm1a (SI Appendix, Fig. S5A and Dataset S3), which were previously shown to be up-regulated in Müller glia during regeneration and required for the regenerative response (51, 52). We also observed upregulation of genes involved in the cell cycle, including cyclin D1 (ccnd1), cdk1 (cyclin-dependent kinase 1), and mki67 (a marker of proliferation). Finally, we observe an increase in prdm1a, a TF expressed transiently in immature photoreceptors (39, 40) (SI Appendix, Fig. S5A and Dataset S3). These data suggest that in the adult samd7^−/− retina, Müller glia up-regulate genes associated with regeneration, proliferation, and the birth of new photoreceptors, perhaps in response to ongoing photoreceptor loss.

In the adult zebrafish retina, cone photoreceptors are arranged in a highly regular “row mosaic” in which a single row of UV and blue cones alternates with double rows of red and green cones (49). To evaluate changes in the mosaic in samd7^−/− retina, we compared retinal flat-mount preparations from adult samd7^−/− and WT fish which were stained with antibodies against Arr3a [a marker of red and green cones (53)] and Arr3b [a marker of UV and blue cones (54)]. In WT retina, we observed the regular mosaic arrangement of photoreceptors: double rows of Arr3a⁺ red and green cones alternating with a single row of Arr3b⁺ blue and UV cones (SI Appendix, Fig. S5B). In contrast, samd7^−/− retinas show a severe disruption of the photoreceptor mosaic, including a disordered arrangement of Arr3b⁺ cells and a complete absence of Arr3a⁺ cells (SI Appendix, Fig. S5C). The latter finding is consistent with the loss of arr3a expression we observed by RNA-seq analysis (SI Appendix, Fig. S5A and Dataset S3).

To determine whether rod and cone photoreceptor function is affected in samd7^−/− fish, we performed ex vivo electroretinogram (ERG) recordings of adult samd7^−/− and WT retinas. First, we measured the response of rods to 530 nm flashes of increasing intensity under scotopic conditions. Next, we measured the response of cones to 448, 530, and 655 nm flashes of increasing intensity. These wavelengths are expected to stimulate distinct subsets of cone opsins (SI Appendix, Fig. S6A and Table S1). We find that the maximal response amplitude (R_max) is significantly reduced in rods, and in cones stimulated at 530 and 655 nm (SI Appendix, Fig. S6 B, D, and E and Table S1). We also detected a possible reduction of R_max in cones stimulated at 448 nm, but the effect was not statistically significant (P = 0.09; SI Appendix, Fig. S6C and Table S1). Response kinetics do not appear to be significantly affected in any of the rod or cone stimulating conditions (SI Appendix, Fig. S6 F–I).

The rod ERG response is reduced despite the fact that we observed very few changes in rod-specific gene expression by RNA-seq in the adult retina (SI Appendix, Figs. S5A and S6B and Dataset S3). The reduction of the ERG response to 448 and 530 nm stimuli (SI Appendix, Fig. S6 C and D) is consistent with the absence of green opsin expression in the adult samd7^−/− retina (SI Appendix, Fig. S5A and Dataset S3), because both 448 and 530 nm stimuli are within the range of green opsin absorbances (SI Appendix, Fig. S6A). However, we do observe a trend in the increase of the normalized 448 nm response (absolute amplitude normalized to maximum response amplitude) in the samd7^−/− retina (SI Appendix, Fig. S6 C, Inset), suggesting there may be a relative gain in the contribution of blue cones in the mutant retina, consistent with increased blue opsin expression (opn1sw2) (SI Appendix, Fig. S5A and Dataset S3). We also observe a reduction in the red-cone ERG response to 655 nm stimuli (SI Appendix, Fig. S6E), consistent with the reduction in the expression of red opsin (opn1lw2) in the samd7^−/− retina (SI Appendix, Fig. S5A and Dataset S3). Taken together, these experiments reveal an across-the-board reduction in the absolute rod and cone electrical responses and a potential increase in the relative responses of blue cones in the adult samd7^−/− retina, consistent with the observed changes in expression of opsins and other phototransduction components.

Samd7 Represses S-Opsin Expression in Dorsal M-Cones of the Mouse Retina.

In the samd7^−/− zebrafish retina, red cones are transformed into hybrid red/UV cones in which both red (opn1lw1) and UV opsin (opn1sw1) are coexpressed (Fig. 4 B and F). In the WT mouse retina, M-cones in the ventral zone show a strikingly similar phenotype: coexpression of M- and S-opsin (36). Since mouse M- and S-opsins are orthologous to zebrafish red and UV cone opsins, respectively, we hypothesized that the upregulation of S-opsin in these cells could be the result of reduced Samd7 expression (7). Indeed, a recent study found that Samd7 expression was reduced in ventral mouse cones compared to dorsal (55). To further evaluate our hypothesis, we engineered Samd7^−/− mice (Materials and Methods and SI Appendix, Fig. S2) and performed antibody staining for M- and S-opsin on flat-mount preparations of adult WT and Samd7^−/− retina. We found that in the WT retina, M-opsin is enriched in the dorsal retina and S-opsin is enriched in the central/ventral retina, as previously reported (36) (Fig. 5 A–C). In the Samd7^−/− retina, we found that S-opsin expression was markedly up-regulated throughout the dorsal retina (Fig. 5D). Closer inspection of individual outer segments in the Samd7^−/− retina revealed that all dorsal M-opsin⁺ cones coexpress S-opsin (Fig. 5 M–O), in contrast to the WT retina in which dorsal M-opsin⁺ cones largely lack S-opsin (Fig. 5 G–I). Quantitative analysis revealed a severe reduction in the number of M-opsin-only cones in the dorsal Samd7^−/− retina and a corresponding increase in the number of M- and S-opsin coexpressing cones (Fig. 5S). We conclude that Samd7 is required for the repression of S-opsin in dorsal M-cones.

Fig. 5.

Dorsal M-cones are transformed into hybrid M/S-cones in the mouse Samd7^−/− retina. (A–F) Flat mounts of adult WT and Samd7^−/− mouse retina stained with S- and M-opsin antibodies demonstrate that S-opsin signal is increased throughout the dorsal region of Samd7^−/− retina. (G–R) Close-up views of boxed insets from panels C and F reveal that S-opsin is up-regulated specifically in dorsal Samd7^−/− M-opsin⁺ cones (i.e., M-cones are transformed into hybrid M/S-opsin⁺ cones). (S) The number of exclusively M-opsin⁺ cones in the dorsal Samd7^−/− retina is reduced to a similar degree as the increase in the number of mixed M/S-opsin⁺ cones, supporting the conclusion that M-cones are transformed into hybrid M/S-opsin⁺ cones (mean ± SD; n = 3/genotype). (T) RNA-seq analysis of WT and Samd7^−/− mouse retina demonstrates a reduction in rod-specific gene expression and an increase in cone-specific gene expression, with the exception of M-opsin (Opn1mw), which is somewhat reduced. Labeled genes were previously found to be dysregulated by qPCR analysis in another Samd7 mutant (38). Genes touching the top edge of the Volcano plot (e.g., Opn1sw) have p-adjusted values equal to zero (n = 3 retinas/genotype). Samd7^minus19A mutants were used for flat-mount imaging and quantification in panels A–S. Samd7^minus10A mutants were used for RNA-seq in panel T. [Scale bar in (A–F), 500 µm; Scale bar in (G–R), 20 µm.] D, V: Dorsal, Ventral. Statistical comparisons in S were performed using two-tailed, unpaired t test assuming unequal variance. **P < 0.005. SD, standard deviation.

To further analyze the Samd7 mutant phenotype, we performed RNA-seq analysis on WT and Samd7^−/− retina of adult mice in three biological replicates. We found that 195 genes were significantly dysregulated by >twofold in the Samd7^−/− retina (Dataset S4). A much larger number of genes were more subtly dysregulated (i.e., <twofold; Dataset S4). As previously reported, we saw upregulation of a subset of cone genes (including Opn1sw) and downregulation of a subset of rod genes (including Rho and Gnat1) (38). Our findings are fully concordant with the expression changes verified by qPCR in a prior study of Samd7^−/− mouse retina, with the exception of M-opsin (Opn1mw), which we find to be reduced by ~60% in our mutant (Fig. 5T and Dataset S4) (38). We also found that Samd7 transcript levels are increased in the Samd7^−/− retina, raising the possibility that Samd7 represses its own expression (Fig. 5T and Dataset S4), similar to our observations in zebrafish (Fig. 2B).

Discussion

In the present study, we identify key roles for samd7 in preserving normal rod and cone photoreceptor identity in zebrafish and mice. In the zebrafish samd7^−/− retina, we found that red cones are transformed into hybrid red/UV cones, green cones are absent, the number of blue cones is approximately doubled, and the number of rods is greatly reduced (Fig. 6). These diverse phenotypes suggest that samd7 mediates distinct functions in individual photoreceptor types. Loss of green cones and a concomitant increase in the number of blue cones might be explained by direct transfating of green cones into blue cones in the samd7^−/− background. Alternatively, supernumerary blue cones could originate from another population of samd7⁺ precursors, such as those that normally give rise to rods. Future lineage-tracing experiments will be required to test these ideas. We also found a role for samd7 in zebrafish red cones. In the samd7^−/− retina, red cones up-regulate multiple UV-cone genes while maintaining expression of red-cone genes. In contrast, in the thrb^−/− retina, red cones are completely transfated into UV cones (13, 14). We therefore propose that thrb regulates red cone identity via two distinct mechanisms: 1) direct activation of red-cone genes and 2) repression of UV-cone genes via induction of samd7 expression (Fig. 6).

Fig. 6.

The role of samd7 in zebrafish photoreceptor specification. Diagram of cone subtypes in the WT and samd7^−/− zebrafish retina. In the samd7^−/− retina, red cones are transformed into hybrid red/UV-sensitive cones, green cones are absent, blue cones are approximately doubled, and the number of rods is greatly reduced. We propose that Samd7 represses UV-cone-specific gene expression by repressing tbx2a and tbx2b, two key regulators of opn1sw1 expression (4, 11, 12) Activation is indicated with an arrowhead, and repression with a flat bar. Black coloring indicates that the factor is expressed, and gray indicates that it is not. A dashed line represents a proposed regulatory relationship and a solid line represents a previously verified regulatory function.

In the Samd7^−/− mouse retina, we observed S-opsin upregulation in dorsal M-cones, analogous to the upregulation of UV opsin in zebrafish samd7^−/− red cones. A similar phenotype has also been reported in other mouse mutants (28–30, 34, 35), suggesting that Samd7 may be part of a regulatory network required to repress S-opsin dorsally. SAMD7 is also enriched in long-wavelength-sensitive cones in chickens and humans (56–58), and human patients with SAMD7 mutations display selective degeneration of the cone-rich macula (59). Thus, SAMD7 may play a role in the cones of other species, but the extent to which its function is conserved across vertebrates remains to be determined.

Materials and Methods

Animal Husbandry.

All zebrafish and mouse studies were conducted in accordance with the Guide for the Care and Use of Laboratory Animals and the Animal Welfare Act and were approved by the Washington University in St. Louis Institutional Animal Care and Use Committee and by the corresponding body at the University of California, Irvine. Additional details are described in SI Appendix, SI Materials and Methods.

Hybridization Chain Reaction (HCR).

Fluorescent in situ hybridization was performed on whole larvae according to the manufacturer’s protocol (HCR v.3.0) (60). Additional details are described in SI Appendix, SI Materials and Methods.

Generation of samd7 Mutant Zebrafish.

Mutant zebrafish were generated using a gRNA (GGGAAGCAGAGTTGCGGCAG) targeting exon 4 of samd7. Additional details are described in SI Appendix, SI Materials and Methods.

Zebrafish Genotyping.

Mutant zebrafish were genotyped by PCR using primers: 5′-GAGCTGCTTCTATTACTGACACCT-3′ and 5′-CTGCCCCTGAGCCAGGAT-3′ to detect the difference between the WT and samd7^stl888 mutant allele (10 bp deletion). Additional details are described in SI Appendix, SI Materials and Methods.

Cell Dissociation and FACS.

5-dpf WT;thrb:tdTomato^+/− and samd7^−/−;thrb:tdTomato^+/− larval eyes were dissociated into single cells using papain (Worthington Biochem), and tdTomato⁺ cells were sorted by FACS. Additional details are described in SI Appendix, SI Materials and Methods.

RNA-seq.

RNA was isolated, cDNA was generated, sequencing adapters were incorporated, and samples were sequenced using 150 bp paired-end reads. Additional details are described in SI Appendix, SI Materials and Methods.

Zebrafish Retina Immunohistochemistry and Imaging.

To facilitate visualization of immunostaining patterns, larvae were treated with 200 μM N-phenylthiourea (PTU; P7629, Sigma) from 24 hpf, then killed at 5 dpf by tricaine overdose, fixed in 4% PFA with 5% sucrose overnight at 4 °C, and stored in PBS at 4 °C. Primary antibodies used were as follows: Rabbit anti-Gnat1 (1:100, SC-389, Santa Cruz), rabbit anti-green opsin [1:300, gift from David Hyde, University of Notre Dame, Notre Dame, IN (61)], rabbit anti-blue opsin (1:300, EJH012, Kerafast), rabbit anti-UV opsin (1:300, EJH013, Kerafast), mouse anti-Arr3a [1:100; mouse monoclonal zpr1 (53, 62), Zebrafish International Resource Center], Arr3b [1:100, produced at Pacific Immunology (PAC 16825-16826) by immunizing chickens with Arr3b peptide ANEEDNIDEKVEKKDTC, as previously described (54)]. Secondary antibodies used were as follows: donkey polyclonal anti-rabbit IgG conjugated with Alexa-555 (1:200; A31572, Invitrogen), donkey polyclonal anti-rabbit IgG (H⁺L) conjugated with Alexa Fluor™ Plus 488 (1:200, A32790TR, Invitrogen), and goat polyclonal anti-chicken IgY conjugated with Alexa-488 (1:200; A11039, Invitrogen). Additional details are described in SI Appendix, SI Materials and Methods.

Photoreceptor Quantification in Zebrafish Larvae.

For quantification of photoreceptors in the retina of 5-dpf larvae, the number of cells within a 40 × 40 µm square immediately dorsal to the optic nerve head was counted using Fiji software (63).

Transretinal ERG Recording and Analysis.

Details are described in SI Appendix, SI Materials and Methods.

Generation of Tg(crx:GFP) Zebrafish, Isolation of 4-dpf crx:GFP⁺ Cells, and Single-Cell RNA-seq Analysis.

Details are described in SI Appendix, SI Materials and Methods.

Samd7 Knockout Mouse Generation.

Samd7 knockout mice were generated by nucleofection of C57BL/6J mouse embryos with Cas9 RNP with guide RNA targeting exon 5 of Samd7 (CATAGAAGAGTGGTATTCCC) by the Genome Engineering and Stem Cell Center at Washington University School of Medicine (GESC). Additional details are described in SI Appendix, SI Materials and Methods.

Opsin Immunostaining of Retinal Flat Mounts and Quantitation.

Five-week-old WT and Samd7^{minus19A/minus19A} male mice were killed, and the sclera was lightly branded dorsally for orientation. Retinas were fixed in 4% PFA for 30 min and then permeabilized in PBS containing 0.5% Triton X-100 for 30 min. Primary antibodies used were as follows: Goat anti-S-opsin (Santa Cruz sc-14363, 1:1,000) and rabbit anti-M-opsin (Sigma AB5405, 1:1,000). Secondary antibodies used were as follows: donkey anti-goat Alexa 555 (Invitrogen A21432, 1:1,000) and donkey anti-rabbit Alexa 488 (Invitrogen A21206, 1:1,000). Additional details are described in SI Appendix, SI Materials and Methods.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Data, Materials, and Software Availability

RNA-seq data are available under GEO accession numbers GSE249756 (64) and GSE269667 (65). Analysis of differentially expressed genes in WT and samd7^−/− zebrafish 5-dpf whole eyes, FAC-sorted thrb:tdTomato⁺ cells, and adult retina is available in Datasets S1–S3, respectively. Analysis of differentially expressed genes in WT and samd7^{minus10A/minus10A} mouse retina is in Dataset S4.