All procedures involving frogs and tadpole embryos were approved by the Animal Care and Use Committee at the University of Calgary. X. laevis egg production was induced by chorionic gonadotrophin injection (Intervet Canada Ltd.) and in vitro fertilization in accordance with standard procedures (see protocols at www.xenbase.org/ RRID:SCR_003280). Embryos were stored at 16°C or 22°C and were kept on a 12:12 h cycle of light and dark. For light exposure experiments, embryos were kept in the dark from stage 24 until they reached stage 43 and were exposed to 30 min of white light (approximately 800 lux) immediately before fixing. Embryo stages were determined using Nieuwkoop and Faber’s developmental data for X. laevis stages on Xenbase. ¹

Embryos at desired stages were collected and fixed in MEMFA [(0.1 M 3-N-morpholino propanesulfonic acid (MOPS), 2 mM ethylene glycol tetra-acetic acid (EGTA), 1 mM MgSO₄, and 4% formaldehyde in diethylpyrocarbonate (DEPC)-treated H₂O)] at 4°C. After 24 h, embryos were placed in 30% sucrose in DEPC water until sunken and then frozen in optimal cutting temperature (OCT) compound (Tissue Tek, Sakura Finetek Inc., United States) and stored at −80°C. OCT-embedded embryos underwent cryostat sectioning (12 μm thick sections) for slide in situ hybridization (ISH). Sense and antisense riboprobes were synthesized from linearized plasmids containing the partial sequences of X. laevis opn4a, opn5, opn6a, opn6b, opn8, and c-fos (GenBank access numbers: MN820837, MN820850, MN820855, MN820856, MN820855 and BC079689.1, respectively). SP6 and T7 RNA polymerases (Roche, Quebec, Canada) with digoxigenin (DIG)-labeled nucleotides (Roche) were used to generate the riboprobes. Riboprobes were hybridized at 65°C and detected with anti-DIG antibodies (1:2500, Roche) conjugated to alkaline phosphatase, and the signal developed with BCIP (Roche; 5-bromo-4-chloro-3-indolyl phosphate) and NBT (nitro blue tetrazolium chloride; Roche) in a NTMT solution (100 mM NaCl, 100 mM Tris-Cl pH 9.5, 50 mM MgCl₂, and 1% Tween 20).

DIG- and fluorescein isothiocyanate (FITC)-labeled probes were synthesized and detected by using anti-DIG (1:500 dilution) and anti-FITC (1:500 dilution) specific antibodies coupled to horse radish peroxidase. The signal was developed with tyramide signal amplification (TSA) cyanine 3 or TSA fluorescein system kits, respectively (both Perkin Elmer, United States), following the manufacturer’s protocol.

DIG-labeled probes developed with TSA cyanine 3 were used for single FISH of retina sections. Following single FISH, immunohistochemistry was performed by incubating the slides with antibodies for Otx2 (rabbit polyclonal 1:200; ab21990, Abcam), Pax6 (rabbit polyclonal 1:100, Cedarlane), Rhodopsin (clone 4D2; mouse 1:200; Millipore) or Calbindin (rabbit polyclonal 1:200; Swant) for 1 hour. Alexa Fluor-tagged secondary antibodies (1:1000, green, 488 nm excitation, anti-mouse or anti-rabbit, Molecular Probes) were used to identify the primary antibody. DAPI staining after ISH and immunohistochemistry labeled cell nuclei and guided cell counting.

For single RNA ISH, bright field optics on a Zeiss compound microscope were used to examine slides through a 20× or 40× objective. Digital images were obtained with Axiovision software. A Zeiss LSM 900 confocal laser scanning microscope was used to image both FISH followed by immunohistochemistry, and double FISH. Z-stacks of digital images (1 μm thick) were generated by using Zen Blue software (Zeiss) on the confocal microscope. A minimum of 8 embryos (N) were analyzed for single ISH, FISH, double FISH, or FISH followed by immunohistochemistry.

Publicly available single cell RNA sequencing (scRNA-seq) whole eye datasets for the adult zebrafish (Danio rerio), chicken (P10) (Gallus gallus) and mouse (P60) (Mus musculus) were downloaded from GitHub at https://github.com/jiewwwang/Single-cell-retinal-regeneration (Hoang et al., 2020). For each species, RStudio (R version 4.3.0) and Seurat V3 were used to generate a Seurat object. All metadata, including cell annotations and sample identification, were preserved from the original study. The Seurat objects were subset to include only control (uninjured) retinae. Quality control was performed to include only cells with between 500 and 2,500 gene features (standard values) and less than 15% mitochondrial genes. A standard Seurat integration pipeline with the “vst” selection method and 2000 variable features was performed for each species. Integrated data were scaled, and an unbiased principal component analysis was performed. RunUMAP was performed with 19 dimensions determined by JackStraw statistics. Original annotation was re-evaluated and confirmed by examining the expression of common cell type markers (Supplementary Table S1) in UMAP clusters indicated by the “FindAllMarkers” function in Seurat.

Neuropsin expression in distinct clusters was evaluated using a custom R script which parses through a gene count matrix to find the number of cells of a particular cell type that express a gene at or above an expression level of one UMI. A list of genes was provided to the script which were then iteratively analyzed, and a data-frame was generated for all results. Github repository including all information pertaining to the downloading and instructions for how to use the function of the “blackshaw” dataset is linked to the following website https://github.com/s-storey/opn_expression/tree/main.

Digital images from confocal microscopy that captured expression in single transverse central retinal sections containing the optic nerve were collected for analysis of neuropsin-positive cells in the retina. If the optic nerve section was not preserved, an adjacent anterior or posterior section was used to count cells The number of retinal slices analyzed (n) and the mean with 95% confidence interval are indicated in the corresponding figures. Immunohistochemical analysis were performed at less two times (N ≥ 2) containing a minimum of 4 independent tadpoles in each experiment. The number of central retina sections analyzed (n) is indicated in figure legends. GraphPad Prism software (v.9.1.2, GraphPad Software, San Diego, CA, United States) was used for statistical analysis.

The onset of neuropsin expression in the retina is unknown. We analyzed the developmental expression of retinal neuropsins found in X. laevis: opn5, opn6a, opn6b and opn8 (Bertolesi et al., 2020, 2021). Neuronal differentiation initiates at stage 24 (1 day of development), classical opsins are detected a day later at stage 35, and the RGC axons target the optic tectum by stage 39/40 (3 days of development) (Table 1). Two opn5 genes exist in X. laevis. These genes likely originated from the X. laevis specific tetraploid in that both localize to homologous chromosome 5; opn5.L (gene number 100500792) and opn5.S (gene number 100500792) located on homologous chromosomes “long” and “short,” respectively, which correspond to the sub-genomes (2n each) provided by the distinct progenitors that gave rise to the tetraploid. ISH on 12 μm retinal sections using an opn5.L riboprobe (1,222 bp length; GenBank number: MN820850) revealed a specific, yet sparse, expression pattern first detected at stage 37/38 (Figure 1A), when the retinal circuits initially become light-activated (Bertolesi et al., 2014). At stage 41, when the retinal-brain circuits are functional, opn5 expression was strongest in cells located in the outer region of the INL (Figure 1A). By stage 43, opn5 was expressed sparsely in the INL, and several cells (~20% of total opn5⁺ cells) were present in the GCL (Figure 2). Of note, the opn5.L probe possessed 91% identity to opn5.S (GenBank number XM_018264817.1), therefore expression of both opn5 genes was likely detected. opn5 expression in the INL and GCL in X. laevis resembles the pattern reported previously in post-hatching chick retina (Yamashita et al., 2010), but differs from mammals, where opn5 expression appears restricted to the GCL (Calligaro et al., 2022; D’Souza et al., 2022).

We previously identified three opn6 genes in X. laevis that are phylogenetically related, and that we named opn6a.L, opn6a.S, and opn6b.L (Bertolesi et al., 2020). Two homologous genes exist for opn6a, located on the homologous chromosomes 5 L and 5 S (opn6a.L, gene number 108716317 and opn6a.S, gene number 108717820), while opn6b.L (gene number 108698982) is located on chromosome 8 L as a single copy. opn6a and opn6b are not homologous genes resulting from X. laevis tetraploidy. Instead opn6b.L was potentially generated by inter-chromosomal transposition. The probe generated to detect opn6a (972 bp) was produced from the opn6a.L template (GenBank number MN820855) and contained 94% identity with the predicted opn6a.S transcript sequence. Therefore, the opn6a probe likely recognized both opn6a.L and opn6a.S homologous transcripts. In contrast, no significant similarity was found between the recognition sequence of the opn6a.L probe and that of opn6b.L (gene number 108698982). Furthermore, the sequence recognized by the opn6b probe (1,047 bp length) showed no significant similarity to opn6a.L or opn6a.S. Of note, at the time of this publication, opn6b.L is classified as an opn5 gene in the NCBI database and should be renamed.

opn6a expression was first detected at stage 35 as diffuse staining, mainly in cells of the dorsal-most outer layer of the retina, with a few positive cells located in the inner portion of the retina (Figure 1B). By stage 37/38, opn6a was expressed strongly by cells of the ONL (Figure 1B). opn6a was present in the ONL by stage 43, but apparently in a smaller number of cells than at stage 37/38, with the strongest expression observed in cells that border the CMZ (Figures 1B, 2B). The CMZ-associated expression at stage 43/44 is consistent with our previous data suggesting that opn6a is robustly expressed in a transient manner in newborn photoreceptors (Bertolesi et al., 2021). Interestingly, the opn6b expression was similar to opn6a, re-enforcing the idea that opn6b is related to opn6a and not to opn5. opn6b expression was first detected at stage 35, and most strongly expressed at stage 37/38 in the ONL. Fewer cells expressed opn6b mRNA in the ONL by stage 43, with the strongest label located at the CMZ border (Figure 1C). Interestingly, the developmental time-course of opn6a/b expression at stage 35/36 and 37/38 matches the characterized processes of specification, migration, and differentiation of X. laevis cone and rod photoreceptors (Table 1) (Chang and Harris, 1998). The decrease in expression seen in maturing photoreceptors integrated into functional retinal circuits (stages 41 and 43/44), however, is different from what is observed for the classical opsins that continue to be highly expressed with ongoing larval development (Chang and Harris, 1998).

We detected three related opn7 genes in Xenopus laevis: opn7a.L and opn7b.L and opn7b.S (gene numbers 108718575, 108700890 and 108702595, respectively). opn7b.L and opn7b.S are homologous genes located on chromosomes 9–10 L and S, respectively, while opn7a.L is located on chromosome 6 L (Bertolesi et al., 2020). mRNA for these genes is almost undetectable in the eye at stage 43/44 (Bertolesi et al., 2021), and therefore we did not analyze them here.

Two opn8 genes (opn8.L; gene number 108717224 and opn8.S; gene number 108718281) are located on chromosome 5 L and 5 S, respectively. Expression of the opn8 genes was analyzed by ISH using a probe generated against opn8.L (904 bp length; GenBank number MN820854; with 91% identity to opn8.S). opn8 expression was first detected at stage 37/38 in a few cells located in the INL. opn8 remained sparsely expressed in the INL, GCL and in cells bordering the CMZ with ongoing development (stage 41 and 43) (Figure 1D).

Together, our results show a specific pattern of neuropsin expression in the retina. opn6a and opn6b are expressed by photoreceptors in the ONL, with expression beginning about the same time as the classical opsins at ~stage 35. With further development, however, opn6a and opn6b expression diminishes in fully differentiated photoreceptors, but is present likely in newborn photoreceptors that arise from the CMZ. The expression of opn5 and opn8 turns on slightly later at stage 37/38. In the mature larval retina (stage 43), opn5 and opn8 are both expressed mainly in cells located in the INL, with the remainder of cells in the GCL.

Next, we identified the specific cell types that express the different neuropsins. We started with the analysis of which photoreceptors in the ONL express opn6. Our ISH data show strong opn6a and opn6b expression at stage 37/38. Thus, we chose this stage to determine whether these two opsins are co-expressed and in which photoreceptor classes. Co-expression was determined by double fluorescent ISH (dFISH). At stage 37/38, most photoreceptors appeared to express both opn6a and opn6b (Figure 2A). Cell types were assigned through FISH, followed by immunohistochemistry either against the known rod marker, rhodopsin, or calbindin, a calcium binding protein expressed by cones (Chang and Harris, 1998). Initially, opn6a and opn6b were expressed by both rhodopsin⁺ and calbindin⁺ cells, with the fluorescent signal commonly localizing to the inner segments of photoreceptors as a cluster of fluorescent dots (Figures 2B,C). However, by stage 43/44, once retinal circuits are functional and connected to the brain, opn6a mRNA was found preferentially in differentiated rods (rhodopsin⁺), and in only a few differentiated cones (calbindin⁺) (Figures 2B,D). In the opposite manner, opn6b mRNA was found preferentially in differentiated calbindin⁺ cones and only a few differentiated rhodopsin⁺ rods (Figure 2D). Interestingly, for the newly born photoreceptors located proximal to the CMZ opn6 opsins were expressed in both rhodopsin⁺ and calbindin⁺ cells, similar to what was observed in the central retina at a younger developmental stage (stage 37/38) (Figure 2D).

We find opn6a and opn6b are co-expressed by all newly generated photoreceptors shortly after their birth, either in the central embryonic retina (stage 37/38) or those generated subsequently by the CMZ (stage 43). With differentiation in the central retina, however, opn6a and opn6b become selectively expressed in rod and cone photoreceptors, respectively.

Next, we identified the cell types that express opn5 and opn8 in the INL and GCL. The ISH data indicated that by stage 41, when retinal differentiation and circuity are defined (Table 1), opn5 and opn8 are mainly expressed in the INL and a subset of cells in the GCL. To determine the cell types located in the INL, we performed FISH followed by immunohistochemistry against the transcription factors Otx2 and Pax6 that in the INL label BCs (Viczian et al., 2003) and ACs (Hirsch and Harris, 1997), respectively.

opn5 mRNA was seen by FISH as clusters of subcellular red fluorescent dots that were either nuclear or associated with the nuclear membrane (Figure 3A). Most of the opn5⁺cells were in the outer portion of the INL (Figures 1, 3A). Approximately 70%–75% of the opn5⁺ cells were BCs, as they colocalized with Otx2 (75.25%, 149 opn5⁺ Otx2⁺ cells/198 opn5⁺ cells; n = 6 retinal sections) (Figures 3A,B). The remaining populations (24.74%, 49 opn5⁺ Otx2⁻ cells/198 opn5⁺ cells) were distributed either in the inner part of the INL or the GCL (Figures 3A,B). We found that these cells were Pax6⁺ (Figure 3C); the opn5⁺ Pax6⁺ cells in the INL were ACs (10.8%, 21 cells opn5⁺ Pax6⁺ in the INL/194 opn5⁺cells; Figure 3D), and the remaining were located in the GCL (21.13%, 41 cells opn5⁺ Pax6⁺ in the GCL/194 opn5⁺cells, Figure 3E).

For opn8-expressing cells, we used immunohistochemistry against Pax6 after FISH at stage 41, when opn8 is expressed in both the INL and the GCL. Like opn5, opn8 expression was seen by FISH as subcellular red fluorescent dots that were either nuclear localized or associated with a nuclear membrane (Figure 4A). opn8 was expressed by only a few retinal cells per section. These cells were all Pax6⁺ and were found in either the AC-populated inner portion of the INL or the GCL (Figures 4B,C). Thus, for the opn8⁺ cells, 70% were INL-located cells, likely ACs, and 30% were GCL-located cells (Figure 4B).

The expression of opn5 and opn8 in Pax6⁺cells in the INL and the GCL led us investigate possible neuropsin co-expression by performing double ISH against opn5 and opn8 on stage 43/44 retinal sections. Interestingly, we found that opn5 and opn8 mRNAs did not colocalize in the INL (2% co-localization; 5 cells opn5⁺ opn8⁺/248 opn5⁺ cells; 52 opn8⁺ cells counted) and only a few cells expressed both genes in the GCL (17% co-localization; 8 cells opn5⁺ opn8⁺/47 opn5⁺ cells; n = 8 retinal sections) (Figures 5A,B).

We also analyzed if neuropsins were co-expressed with melanopsin. In the Xenopus retina, opn4 and opn4a are co-expressed (91% colocalization) in a subset of HCs and RGCs (Bertolesi et al., 2014). Because of the high level of co-expression, we could use the opn4a antisense probe alone in order to determine both opn4 and opn4a colocalization with opn5 and opn8. Double FISH at stage 43/44 revealed that only 24% of opn5⁺ cells also expressed opn4a in the INL (57 opn5⁺ opn4a⁺cells/240 opn5⁺ cells), while 41% of cells expressed opn4a and opn5 in the GCL (15 opn5⁺ opn4a⁺ cells/37 opn5⁺cells; n = 8 retina central sections) (Figures 5C,D). In contrast, opn8 was not expressed by melanopsin⁺ cells of the INL or GCL (4% in the INL; 2 cells opn8⁺ opn4a⁺/45 opn8⁺ cells; n = 148 opn4a⁺cells counted; n = 8 retinal sections) (5% in the GCL; 1 cell opn8⁺ opn4a⁺/19 opn8⁺ cells; n = 19 opn4a⁺ cells counted in eight retinal sections) (Figures 5E,F).

Together, these results suggest that neuropsins are differentially expressed in distinct cells, with opn5 in BCs, ACs, and RGCs, and opn8 in ACs and in a different subset of RGCs. Additionally, mRGCs do not express opn8 and only a few express opn5. Thus, our results suggest that the neuropsins opn5 and opn8 allow for a greater total number of potentially photosensitive RGCs.

Next, we investigated whether the neuropsin-expressing cells participate in light-stimulated retinal circuits. To identify light-activated cells, we used c-fos, an immediate early gene marker expressed in neurons activated in response to synaptic stimuli (Yoshida et al., 1996; Huerta et al., 1997). We showed previously that several cells express c-fos after 30 min of white light illumination, particularly in the retina of dark-adapted X. laevis embryos (Bertolesi et al., 2014). Specifically, light-induced activation occurs in at least two cells of the INL (ACs and BCs) for each activated RGC. No c-fos expression is induced in photoreceptors (Bertolesi et al., 2014). Thus, we used double FISH to analyze at stage 43/44 the co-expression of c-fos with opn5 and opn8 upon light exposure.

We found that the majority of opn5-expressing cells in the INL (64%) expressed c-fos (n = 363 cells counted) after 30 min of light stimulation, as did almost all cells (87%) in the GCL (n = 71 cells counted) (Figures 6A,B). Similarly, 70% and 77% of opn8-expressing cells also expressed c-fos in the INL (n = 77 cells counted) and the GCL (n = 35 cells counted), respectively (Figures 6C,D). It is worth noting that without light stimulation, almost no c-fos expression was detected in the retina of dark-adapted embryos (Supplementary Figure S1 and Bertolesi et al., 2014).

Together, our data show that the majority of opn5⁺ and opn8⁺ cells in the INL and GCL express c-fos after light stimulation. We also define a second group of opsin-expressing RGCs, in addition to the larger number of melanopsin-expressing cells in the GCL (Bertolesi et al., 2014), which are in light-activated circuits.

We next compared our data obtained in X. laevis to publicly available scRNA-seq datasets from the adult zebrafish, mouse, and chicken retina (Hoang et al., 2020). The datasets were generated recently to analyze gene expression of normal and injured retinas (Hoang et al., 2020). We isolated the scRNA-seq data of the control (uninjured) retina and analyzed the expression of neuropsins in different retinal cell-type clusters. opn5 was the only neuropsin gene detected in the chicken and mouse dataset, however, almost no cells showed expression (5 opn5⁺ AC and 1 opn5⁺ RGC of 8,895 chicken retinal cells; no opn5⁺ neuropsin gene of 14,891 mouse retinal cells). These results can be explained by a lack of evolutionary conservation of neuropsins, the small number of retinal cells in the chicken and mouse datasets, and low opn5 mRNA levels. In contrast, the zebrafish dataset contained more retinal cells (18,057 cells), and several neuropsins were detected (Figures 7A,B). Of note, some of the retinal cluster annotations present in the original dataset, such as GABAergic and glycinergic ACs as well as activated and resting MG cells, were grouped as ACs and MG cells, respectively, in our analysis.

opn5 was detected in only a handful of cells in the adult zebrafish retina (11 opn5⁺ cells/18,057 total cells), spread between different cell types (rods, cones, BC, HC, AC and MG cells) (Figure 7B). In contrast, other neuropsins were expressed in more cells and the data was supportive of our results in X. laevis: (i) opn6a and opn6b were detected mainly in cones and rods [opn6a: 175/1427 cones (12.26%), 79/3,594 rods (2.2%); opn6b: 251/1,427 cones (17.6%); 91/3,594 rods (2.5%)]. Of note, zebrafish opn6b is homologous to opn6a generated during the teleost whole genome duplication and differs from the opn6b in X. laevis; (ii) the three opn7 genes were barely expressed in the retina; and (iii) opn8c was expressed mainly in MG cells (116 cells/5,097 MG, 2.27%) and ACs (13/2,170 cells, 0.6%). Additionally, opn9 that is present in teleosts but lost in tetrapods, was expressed in HCs (388/2037 cells, 19%), which agrees with findings from ISH performed in adult zebrafish (Davies et al., 2015).