Molecular Vision 2018; 24:587-602
Received 07 April 2018 | Accepted 24 August 2018 | Published 27 August 2018
1Department of Biological Science, Florida State University, Tallahassee, FL; 2Program in Neuroscience, Florida State University, Tallahassee, FL
Correspondence to: James M. Fadool, Department of Biological Science and Program in Neuroscience, Florida State University, 319 Stadium Drive, Tallahassee, FL, 32306; Phone: (850) 644-3550; FAX: (850) 645-8447; email: firstname.lastname@example.org
Purpose: Retinitis pigmentosa (RP) is a collection of genetic disorders that results in the degeneration of light-sensitive photoreceptor cells, leading to blindness. RP is associated with more than 70 loci that may display dominant or recessive modes of inheritance, but mutations in the gene encoding the visual pigment rhodopsin (RHO) are the most frequent cause. In an effort to develop precise mutations in zebrafish as novel models of photoreceptor degeneration, we describe the generation and germline transmission of a series of novel clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9-induced insertion and deletion (indel) mutations in the major zebrafish rho locus, rh1–1.
Methods: One- or two-cell staged zebrafish embryos were microinjected with in vitro transcribed mRNA encoding Cas9 and a single guide RNA (gRNA). Mutations were detected by restriction fragment length polymorphism (RFLP) and DNA sequence analyses in injected embryos and offspring. Immunolabeling with rod- and cone-specific antibodies was used to test for histological and cellular changes.
Results: Using gRNAs that targeted highly conserved regions of rh1–1, a series of dominant and recessive alleles were recovered that resulted in the rapid degeneration of rod photoreceptors. No effect on cones was observed. Targeting the 5′-coding sequence of rh1–1 led to the recovery of several indels similar to disease-associated alleles. A frame shift mutation leading to a premature stop codon (T17*) resulted in rod degeneration when brought to homozygosity. Immunoblot and fluorescence labeling with a Rho-specific antibody suggest that this is indeed a null allele, illustrating that the Rho expression is essential for rod survival. Two in-frame mutations were recovered that disrupted the highly conserved N-linked glycosylation consensus sequence at N15. Larvae heterozygous for either of the alleles demonstrated rapid rod degeneration. Targeting of the 3′-coding region of rh1–1 resulted in the recovery of an allele encoding a premature stop codon (S347*) upstream of the conserved VSPA sorting sequence and a second in-frame allele that disrupted the putative phosphorylation site at S339. Both alleles resulted in rod death in a dominant inheritance pattern. Following the loss of the targeting sequence, immunolabeling for Rho was no longer restricted to the rod outer segment, but it was also localized to the plasma membrane.
Conclusions: The efficiency of CRISPR/Cas9 for gene targeting, coupled with the large number of mutations associated with RP, provided a backdrop for the rapid isolation of novel alleles in zebrafish that phenocopy disease. These novel lines will provide much needed in-vivo models for high throughput screens of compounds or genes that protect from photoreceptor degeneration.
Retinitis pigmentosa (RP) represents a collection of heritable retinopathies characterized by the progressive degeneration of rod photoreceptors, followed by the secondary loss of cones and circuitry remodeling. RP is associated with mutations at over 70 loci, disrupting not only phototransduction and the visual cycle, but also nearly every aspect of rod cell biology, including development, metabolism, transport, and structure (RetNet). Mutations in rhodopsin (RHO; OMIM 180380) are the most frequent causes of autosomal dominant (ad) RP, and they account for a small fraction of autosomal recessive (ar) RP . More than 150 unique mutations spanning the entire RHO coding sequence have been identified (Human Gene Mutation Database). These mutations disrupt various molecular processes, including phosphorylation, glycosylation, chromophore binding, G-protein activation, arrestin-mediated endocytosis, and targeting of RHO to the rod outer segment (ROS).
RHO mutations have been categorized according to biochemical properties or clinical standards [2-7]. In vitro, class I mutants were defined as showing levels of expression similar to the wild-type (WT) RHO, reconstitution with chromophore, and proper folding; however, in vivo, the protein products mislocalized to the plasma membrane of the cell body [2,3]. These mutations include several at the C-terminus, which disrupt a VXPX consensus sequence necessary for post-Golgi trafficking and the targeting of RHO to the ROS . C-terminal mutations also affect conserved phosphorylation sites essential for protein–protein interactions and the deactivation of RHO . Class II RHO mutations exhibit reduced expression compared to WT, show poor reconstitution with chromophore, and are retained in the trans-Golgi network, suggesting misfolded or unstable products. These mutations largely alter the 5′ and membrane spanning domains, N-linked glycosylation, or cysteine residues . For example, T17M and RHO P23H, the most common RP allele in the United States [10-14], display retention in the trans-Golgi network [2,3,15] and mutations T4K, T17M, and P23H in or near consensus glycosylation sequences [16-18] alter glycosylation profiles in vitro and similarly affect trafficking. Knowledge of the molecular pathology underlying rod death is incomplete, but these data and mounting evidence suggest that diverse mechanisms are responsible.
Animal models recapitulate many of the histopathological features of RP, and they have been invaluable for investigating the cellular and physiologic consequences of disease-causing mutations. Several of the earliest and frequently exploited rodent models, such as the rd1, rd2/rds, and rd10 mice [5,19-22] and the Royal College of Surgeons (RCS) rat [23-25], harbor spontaneous mutations in gene orthologs that are associated with human disease. The characterization of transgenic rodent, pig, dog, and frog models overexpressing mutant forms of RHO display reduced or aberrant opsin localization, thinning of the retinal outer nuclear layer (ONL), shortened or dysmorphic ROSs, rod death, and eventually cone death [26-37]. Large animal models, such as canine, with naturally occurring mutations, share common histological features with RP, and they have been incredibly useful for pre-clinical safety testing and recognizing the long-term outcomes of novel therapies [38-42]. In animal models, consistent with the in vitro phenotype of class I mutations, opsin mislocalization precedes progressive photoreceptor death [8,43-46]. Models generated through the knock-in of precise mutations into the endogenous Rho locus allow for the probing of highly specific mechanistic hypotheses leading to RP [47-50]. The relative levels of the mRNA expression and protein of the mutant alleles, to those of the WT allele influence the stage of onset and severity of the degeneration phenotype . Other studies demonstrate that a low expression level of P23H RHO was not associated with protein accumulation in the endoplasmic reticulum, but it was appropriately trafficked to the outer segment and altered disc morphogenesis preceded degeneration [48,49]. Advances in gene editing in non-mammalian species open the possibility for expanding the number of animal models harboring precise lesions of RHO to further the understanding of the rod degeneration phenotype. A recent report took advantage of clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 to target multiple rho loci in the tetraploid model Xenopus laevis. The data showed that the mutation of a single locus out of three was sufficient to alter rod survival .
Zebrafish have proven a powerful forward genetic model for studying retinal development and disease [52-56]. The majority of existing mutations were the product of numerous forward genetic screens to identify genes essential to embryonic development, photoreceptor function, and survival [52-54,57-62]. Those same features of high fecundity, external fertilization, relatively short generation time, and ease of the microinjection of one-cell stage embryos have facilitated the rapid adoption of gene-targeting technologies to test specific hypotheses of gene function in eye development [63-65], to generate reporter knock-ins [66,67], to induce precise modifications through homology-directed repair [68-70], and to model human disease [71,72].
The purpose of this study was to generate zebrafish lines harboring mutations of the major endogenous rho locus, rh1–1, as models of rod degeneration. Using CRISPR/Cas9, we targeted conserved regions of zebrafish rh1–1 that are associated with photoreceptor disease in humans. Given the large number of mutations in RHO and the ease and efficiency of gene editing using CRISPR/Cas9, we screened for novel germline mutations of the endogenous locus and then tested for those that phenocopied the rod defects associated with similar mutations in the human locus. This approach offers a strategy for targeting zebrafish genes for functional analyses in vivo, and it provides zebrafish models for investigating the initial cellular changes leading to RP with the potential to develop therapeutic treatments.
AB strain zebrafish (Danio rerio) were reared, bred, and staged according to standard methods . Experiments and procedures were approved by the Florida State University Animal Care and Use Committee. The transgenic (Tg) line Tg(Xops:EGFP) expresses enhanced green fluorescent protein (EGFP) in rods under control of the Xenopusrho promoter . Animals were anesthetized with MS222 and euthanized in ice water.
The gRNA plasmids were constructed by the Mutation Generation and Detection Core, University of Utah, to target the 5′ or 3′ coding region of zebrafish rh1-1. The gRNA plasmids were PCR-amplified (forward primer: 5′ CAC CGC TAG CTA ATA CGA CTC 3′; reverse primer: 5′ GAT CCG CAC
CGA CTC GGT GCC AC 3′) to generate 130 bp in vitro transcription templates including a T7 binding site, 20 nucleotides identical
to the 5′ or 3′ rh1-1, and a gRNA scaffold sequence. gRNA was synthesized using the T7 MEGA shortscript kit (Thermo Fisher, Waltham, MA) followed by ethanol precipitation. The 5′ rh1-1 target sequence 5′ GCC TAT GTC CAA TGC CAC CG
DNA was extracted from 20 to 25 injected G0 embryos or uninjected sibling controls, pooled into groups of five embryos, and used as a template for the PCR amplification of rh1-1. For the restriction fragment length polymorphism (RFLP) analysis of 5′ rh1-1, a 253-bp product was amplified (forward primer: 5′ ACA GTC CTG CCC AGA CAT CTA 3′; reverse primer: 5′ ATG GTG ACG TAC AGC GTG AG 3′) and digested with NciI (New England Biolabs, Ipswich, MA). Indels were detected by retention of the 253-bp band. For the RFLP analysis of 3′ rh1-1, a 325-bp product was amplified (forward primer: 5′ GCG TGG CCT GGT ACA TCT TC 3′; reverse primer: 5′ GGT CTC TGT GTG GTT TGC CG 3′) and digested with BbsI (New England Biolabs). Indels were detected with the gain of a 284-bp band. For the Sanger sequencing analysis, the entire rh1-1 coding sequence was amplified (1,198 bp, forward primer: 5′ ACA GTC CTG CCC AGA CAT CTA 3′; reverse primer: 5′ GGT CTC TGT GTG GTT TGC CG 3′), purified using the EZNA Cycle Pure kit (Omega Bio-Tek, Norcross, GA), and sequenced with nested primers for 5′ rh1-1 (5′ ATG GTG ACG TAC AGC GTG AG 3′) or 3′ rh1-1 (5′ GCG TGG CCT GGT ACA TCT TC 3′).
The remaining injected G0 embryos were reared to adulthood and outcrossed with WT animals or inbred to generate F1 fish. DNA was extracted either from F1 embryos (typically 10–15) or from F1 adult fin clip tissue and screened for mutations, as described previously.
Immunolabeling and fluorescence microscopy of retinal cryosections (10 µm) were performed, as described previously [59,74]. Adults and larvae were genotyped by DNA sequence analysis. The following primary antibodies were used: 1D1 (mouse monoclonal antibody (mab), 1:20) against Rho , 4C12 (mouse mab, 1:20) to label the rod plasma membrane , and Zpr-1/FRet 43 (mouse mab, 1:20) against Arr3a to label double cone cells [75,76]. Species-specific Alexa-conjugated secondary antibodies were from Molecular Probes (Thermo Fisher). Sections were imaged using either a Zeiss Axiovert S100 fluorescent microscope or a Zeiss LSM 510 Laser Confocal equipped with a 40X C-Apochromat water immersion objective (NA 1.2).
Terminal deoxynucleotide transferase (TdT)-mediated dUTP nick-end labeling (TUNEL) assay was performed on retinal cryosections using the In Situ Cell Detection kit (TMR red, Roche, Basel, Switzerland), as per the manufacturer’s protocol, and co-labeled for rods (4C12).
Immunoblotting was performed as previously described [77,78] with the following modifications. Three individuals of each genotype were used for the analysis. Two retinas from an adult fish were dissected in cold 80% Hanks salt solution, pooled and homogenized in 50 mM of Tris-buffered saline containing 0.5% Triton X-100 and a protease inhibitor cocktail, and incubated for 3 h at 4 °C. Insoluble material was pelleted by centrifugation at 12,000 ×g. In total, 10 μg of the soluble protein per sample was diluted in a sodium dodecyl sulfate (SDS) gel-loading buffer and separated on 12% acrylamide gels by SDS–polyacrylamide gel electrophoresis (PAGE), followed by electro-transfer to nitrocellulose. Duplicate blots were immunolabeled with the 1D1 mab against Rho diluted at 1:1,000 in a blocking buffer or the mouse anti-Β-actin mab (Cat# A2228; Sigma-Aldrich, St. Louis, MO) diluted at 1:1,000, followed by peroxidase-conjugated goat anti-mouse secondary antibody (Sigma-Aldrich) diluted at 1:3,000. Labeling was detected using enhanced chemiluminescence (ECL; GE Health Life Sciences, Marlborough, MA) with exposure on Fugi Rx film (MIDSCI, St. Louis, MO). Quantitative densitometry of immunolabeling was performed using a Hewlett-Packard Scanner (model Scanjet 4850; Hewlett Packard, Palo Alto, CA) in combination with Quantiscan software (Biosoft, Cambridge, UK; 78). Rho labeling was standardized to actin labeling as an internal control. Standardized immunodensity values were then compared across genotypes using the Kruskal–Wallis statistical analysis with a significance value of p<0.05.
Our goal was to generate lines of zebrafish harboring mutations of rh1–1 that mimic disease-causing alleles in humans. In zebrafish, rh1–1 is located on chromosome 8. As in other teleosts, the gene is represented by a single exon (Figure 1A), but the protein sequence shares considerable homology with other vertebrate RHO proteins (Figure 1B). A second rho gene, rh1–2, is located on chromosome 11, but RNA in situ hybridization shows a spatially and temporally restricted pattern of expression [79,80]; therefore, we focused more effort on rh1–1. Zebrafish rh1–1 was targeted for disruption using CRISPR/Cas9 and gRNA targeting PAM sequences near conserved motifs in the 5′- and 3′-coding sequences (Figure 1C). One- or two-cell staged zebrafish embryos were microinjected with in vitro transcribed mRNA encoding Cas9 and a single gRNA complementary to either the 5′ rh1–1 region encoding the N-linked glycosylation consensus sequence at N15 or the 3′ rh1–1 region just upstream of the sequences encoding VSPA intracellular trafficking moiety. At 24 h post-fertilization (hpf), DNA was extracted from injected embryos and uninjected controls and used as a template for PCR. The efficiency of the gRNA and Cas9 pair was analyzed by RFLP analysis (Figure 1D-G) and validated by Sanger sequencing. Targeting the 5′-region of rh1–1, indels were detected by the loss of an NciI restriction site and retention of the 253-bp amplicon (Figure 1D,E arrow). Alterations to the chromatograms were consistent with mosaicism in the template DNA sequence starting at the predicted Cas9 cleavage site (data not shown). Indels at 3′ rh1–1 were detected following digestion of a 325-bp 3′-rh1–1 amplicon with BbsI by the gain of a 284-bp band (Figure 1F,G). Baseline noise in the chromatograms downstream of the Cas9 cleavage site was interpreted as a low level of mosaicism (data not shown).
To identify the germline transmission of mutant rh1–1 alleles, clutches of injected embryos were grown to adults. G0 adults were inbred or outcrossed to uninjected animals. DNA isolated from F1 embryos was screened by RFLP and Sanger sequencing, as described above. The germline transmission of novel alleles affecting both targeted regions was readily identified in the F1 consistent with the mosaicism observed in the chromatograms of injected embryos. F1 progeny were grown to adults and genotyped by fin clipping to identify specific rh1–1 alleles. Four mutations in the 5′-coding sequence and six mutations in the 3′-coding sequence were recovered, including in-frame deletions and a putative null allele (Table 1). The phenotypes for five of the alleles are described below. All analyses were performed on larvae from multiple generations, and the phenotypes did not vary across generations.
The zebrafish rhofl6 allele encodes a non-sense mutation at codon 17 (T17*), predicted to form a truncated product or a null allele (Figure 2A). If a null allele, it is anticipated that heterozygous animals will express approximately half of the level of Rho as WT animals . Therefore, we compared the level of the Rho expression in retinas from heterozygous rhofl6/+ and homozygous rhofl6/fl6 fish to that of WT adults (n=3 for each genotype). Immunolabeling for Rho using the 1D1 mab was detected as a ladder with major bands at 32, 64, and 98 kDa (Figure 2B). The densitometry of the immunoblots indicated that Rho labeling was significantly different (p<0.025, Kruskel-Wallace test) in heterozygous rhofl6/+ and homozygous rhofl6/fl6 samples compared to homozygous WT samples, with approximately half the labeling in the heterozygotes and virtually no detectable labeling in the homozygous rhofl6/fl6 retinas, consistent with rhofl6 being a null allele (Figure 2C).
Differential immunolabeling of a histological section with the 1D1 mab and the 4C12 mab, which recognizes an epitope on the rod plasma membrane , allowed for the analysis of rod presence independent of the opsin expression. Sections of 5 or 6 dpf WT larvae immunolabeled with 1D1, 4C12, or Zpr-1 were consistent with the previously described arrangement of rods and cones across the ONL. Immunolabeling with either 4C12 or 1D1/Rho showed the typical asymmetric patterning of rods with the greatest intensity of labeling in the ventral retina and more sparse cells in the central and dorsal retina (Figure 2D,E). 1D1 selectively labels the outer segment and 4C12 labels the entire rod photoreceptor, revealing the position of their cell bodies at the innermost region of the ONL. Of the four cone subtypes, red and green wavelength-sensitive cones can be immunolabeled with the Zpr-1 mab, which recognizes Arr3a [75,76]. Immunolabeling for cones shows uniform palisades across the length of the ONL (Figure 2D,E). Rod and cone distributions in heterozygous rhofl6/+ larvae appear similar to those in the WT. In homozygous rhofl6/fl6 retinas, Rho immunostaining is completely absent, but the immunolabeling of rods was observed close to the retinal margin (Figure 2 D,E), the site of continual neurogenesis in teleosts . These data are consistent with rhofl6/fl6 as a likely null allele. Furthermore, the lack of rod immunolabeling in the central retina suggests the Rho expression in zebrafish is essential for rod survival. Labeling of red/green cones in rhofl6/fl6 was indistinguishable from that of the WT, indicating that the genetic alteration resulted in a rod-specific defect.
Two of the N-terminal mutations were predicted as in-frame alterations that disrupt an N-linked glycosylation consensus sequence (Figure 3A). The N-linked glycosylation of proteins occurs at NXS/T consensus sequences, where X can be any amino acid other than proline. Human mutations at N15 and T17 disrupt a conserved glycosylation sequence and are associated with adRP. The zebrafish rhof10 and rhofl7 alleles disrupt the conserved consensus sequence that includes N15 (Figure 3A,C). Zebrafish rhof10 encodes an in-frame deletion of T17 while leaving the remaining coding sequence unaltered (Figure 3A). The zebrafish rhofl7 allele deletes both A16 and T17 and inserts a proline at codon 16 (Figure 3C). WT siblings immunolabeled for rod and cone markers at 6 dpf display the expected labeling across the ONL (Figure 3B,D). In heterozygous rhofl7/+ and rhofl10/+ larvae, rod and Rho immunolabeling is confined to the peripheral retinas, with no labeling of the central retina. Immunolabeling for cones was unaltered from the controls (Figure 3B-E). The phenotype of the heterozygous and homozygous mutants was similar with labeling restricted to the retinal margin.
Several alleles were recovered using a gRNA that targeted Cas9 to a site just upstream of the DNA encoding the conserved VSPA sorting sequence and phosphorylation sites. The zebrafish rhofl8 allele results in an in-frame deletion of A338 and S339, as well as the insertion of T338, V339, W340, and T341, disrupting a putative phosphorylation site at S339 (Figure 4A). Similar to that observed for previous alleles, immunolabeling of 6 dpf retinal sections for Rho and rods was limited to a few cells near the retinal margin, whereas the red/green cones remained intact (Figure 4C).
The zebrafish rhofl9 allele results in a non-sense mutation at codon 347 (S347*), eliminating the final eight amino acids of the Rho C-terminal tail, including the conserved VXPX targeting signal (VAPA in humans, VSPA in zebrafish; Figure 1B). Heterozygous rhofl9/+ larvae display loss of rods and Rho staining from the central retina, with no obvious effect on the immunolabeling of red/green cone photoreceptors (Figure 4C). In humans, rodents, and frogs, disruption of the C-terminal VXPX consensus sequence results in the mistrafficking of Rho to the plasma membrane [8,9,43-46]. To determine whether the rhofl9 mutation in zebrafish disrupts Rho trafficking, the rhofl9 was placed on the Xops:EGFP reporter line background. The EGFP expression fills the entire rod cell body, terminals, inner segment, and, to a lesser extent, the outer segment . In WT siblings, immunolabeling for Rho specifically localizes to the outer segment (Figure 4D). However, in rhofl9/+ heterozygous animals, immunolabeling for Rho was not restricted to the outer segment, but localized to the plasma membrane of the cell body, and many rods lack overt labeling of an outer segment (Figure 4D).
The fact that in zebrafish, rods are not essential for viability allows for the analysis of phenotypes in adults carrying rh1–1 mutations. The extent of the changes in the number and morphology of the rods in rhofl7/+ and rhofl9/+ adult retinas were clear following immunolabeling with the 4C12 mab (Figure 5A-C). At this age in WT retinas, the photoreceptor cell nuclei are tiered with rod cell bodies, forming a row of several cells in thickness vitread to the elongated cone nuclei (Figure 5A). The rod inner segments project distally past the row of cone nuclei, and the outer segments interdigitate with the pigmented epithelium (Figure 5A). In rhofl7/+ (Figure 5B) and rhofl9/+ (Figure 5C) retinas, fewer rods were present, their spacing and morphologies were less regular, and no outer segments were labeled. In both mutants, TUNEL was observed in the ONL adjacent to the OPL (arrows, Figure 5B,C), but no TUNEL was observed in WT retinas (Figure 5A). Immunolabeling for Rho confirmed the lack of outer segments and altered trafficking in rhofl7/+ and rhofl9/+ adult retinas. In WT adults, immunolabeling of Rho was restricted to the outer segments located at the most distal region of the retina that interdigitate with the pigmented epithelium (Figure 5D, top of the panel). In heterozygous rhofl7/+ or rhofl9/+ adults, the rod cell bodies were immunolabeled for Rho, and little outer segment material was observed (Figure 5E,F). Immunolabeling for red and green cones (Figure 5G-I) showed no differences between WT and mutant samples.
This study reports the characterization of novel zebrafish models of rod degeneration produced by the CRISPR/Cas9 gene disruption of two separate targets within the rh1–1 gene. We isolated alleles that behave as disease-causing mutations observed in the clinic and in other well-characterized animal models. We showed evidence that N-terminal and C-terminal mutations resulted in rapid rod degeneration in larval zebrafish; however, similar to our previous report , the loss of rods did not appear to affect cones in adults up to 7 months of age. The goal in targeting the endogenous zebrafish locus is that the alleles will recapitulate transcriptional regulation, mRNA processing, and post-translational modifications, which may underlay the variation in the onset of visual deficits observed in clinically relevant disease alleles. Ultimately, these lines in zebrafish will provide much-needed models for high throughput genetic and small molecule screens in an effort to identify compounds or pathways that may lessen the initial pathological sequences leading to rod degeneration.
TALENs and CRISPR/Cas9 have opened the door to the development of a range of innovative tools and approaches to study gene function in vitro and in vivo. While many have taken advantage of the ability to tag genes with reporter knock-ins [66,67] or to induce precise modifications through homology-directed repair [68-70], we took a relatively simpler approach to isolate novel germline mutations and retain those that have phenocopied mutations of the human locus. Given the ease and efficiency of gene editing in zebrafish using CRISPR/Cas9, an extension of the approach is that gRNAs could be designed to tile any region of a gene to induce a library of in-vivo mutations to screen for effects on gene function or novel phenotypes.
The targeted disruption of 5′ rh1–1 yielded a putative null allele, rhofl6, and two alleles, rhofl7 and rhofl10, encoding in-frame mutations and altering the conserved NXT glycosylation consensus sequence at N15. rhofl7 removes XT and inserts a proline, whereas rhofl10 deletes T17. In heterozygous larvae, both alleles result in rod-specific cell death and a reduced opsin expression. Human mutations N15S and T17M affecting the N15 RHO glycosylation consensus sequence are associated with adRP [17,18]. RHO N15S and T17M alleles are classified as class II mutations, which are expressed at lower levels than WT opsin, fail to reconstitute with chromophore, and are retained in the trans-Golgi network in heterologous in-vitro expression systems [2-4]. Transgenic animal models overexpressing these class II adRP alleles exhibit shortened ROSs, ONL thinning, and rod death [30,36], confirming that glycosylation at N15 is essential.
The putative null allele, rhofl6 (T17*), is the only mutation in our set that results in a recessive phenotype. A reduced expression by immunoblot and differential immunolabeling with the Rho-specific mab and a mab that labels an antigen expressed on the rod plasma membranes and the outer segment confirm that this is a null allele. While most human RHO mutations cause adRP, several alleles result in arRP, including null alleles of RHO [44,83,84]. Similar to null alleles in humans and mice [82,85-87], homozygous rhofl6 zebrafish exhibit rod degeneration consistent with the opsin expression being essential for rod survival, whereas heterozygous animals display no rod death but lower levels of immunolabeling by the Rho-specific antibody. Multiple polymorphisms at each V345 and P347, as well as deletion of the distal RHO C-terminus (Q344ter) likely affect vectorial sorting of RHO to the ROS [8,9]. Disruption of a putative phosphorylation site at S339 rhofl8 results in rod death and the lack of the full VXPX sorting signal (rhofl9) exhibits opsin mislocalization, consistent with mouse, rat, rabbit, frog, and pig models [9,42-45]. In all, the cones appear unaltered consistent with our previous report , which has allowed for an analysis of regeneration and circuitry remodeling in the absence of cone death [58-61].
The rapid rod degeneration observed in the rh1–1 mutant zebrafish described in this report compared to other animal models and human RP is notable. Regardless of the lesion, mutations of the endogenous zebrafish rh1–1 result in a lack of rod labeling in the central and ventral retina by 6 dpf. Rod-specific labeling was limited to the retinal margin, the site of continuing retinal neurogenesis in teleosts . This pattern of cell death is similar to that observed in a transgenic line of zebrafish overexpressing a membrane-targeted CFP (mCFP) in rod photoreceptors . Overexpression of mCFP under the control of the Xenopus opsin promoter resulted in Rho mislocalization to the plasma membrane and the retention of a Rho–GFP fusion protein in the endoplasmic reticulum or Golgi . It remains unclear what underlies the more rapid rod degeneration in zebrafish models. In mice and rats, data suggest that variation in the rate of photoreceptor degeneration may be dependent on the copy number of the mutant opsin transgene or the relative amount or ratio of the mutant to WT Rho expressions [45,50]; however, only the most aggressive rodent models show similar rapid cell death. One possibility is that the swift death observed in our models may be a consequence of the pace of differentiation in the rapidly developing zebrafish larvae. In zebrafish embryos and during photoreceptor regeneration, the expressions of the transcription factors Crx, Nrl, and Nr2e3 precede terminal mitosis, and final cell division is symmetric, generating two cells of the same photoreceptor subtype [60,88-90], suggesting that neuroblasts are primed for rapid differentiation. The rho expression first appears in the ventral retina between 50 and 52 hpf, a time frame within hours of terminal mitosis [76,90-94]. Rod numbers increased rapidly within the ventral retina, and they extended into the nasal and temporal regions of the retina by 70 hpf [90,92,93]. The first outer segments appear at 60 hpf [95,96], and larvae exhibit their first visual responses at approximately 70 hpf [97,98]. Taken together, the rapid onset and high level of protein expressions destined for the outer segment may contribute to the rapid rod death observed in the rh1–1 mutants. In contrast, in the mouse, the expression of rod-specific transcription factors occurs after terminal mitosis , and the Rho expression for later-born rod precursors is delayed for 5.5–6.5 days following terminal mitosis . The rapid degeneration in zebrafish may prove advantageous as a model for high-throughput in-vivo drug screens to identify small molecules that may slow the degeneration phenotype.
A recent report by Feehan et al.  uses a similar gene-targeting approach to generate endogenous mutations in the Xenopus rho genes, and they found phenotypes consistent with dominant and recessive forms of RP, including rod cell loss and opsin mislocalization. X. laevis are tetraploid organisms, and the three Rho gene products are highly conserved, with either silent or conservative amino acid differences among them. Rod degeneration was associated with larger in-frame indels; however, frame shift mutations or loss of the initiation methionine, resulting in a lower opsin expression, were not associated with degeneration. The zebrafish rh1–2 amino acid sequence is 78% conserved with zebrafish rh1–1, with a putative N-linked glycosylation consensus sequence at N15 (NES compared with NAT in Rho), but no C-terminal VXPX sorting signal. Unlike rh1–1, the rh1–2 expression is delayed until 5 dpf, at which time the expression is restricted to the ONL adjacent to the peripheral retina, where it remains at 175 dpf [79,80]. Our data for gene targeting rh1–2 did not reveal any rod degeneration (data not shown). Furthermore, the rod death in our putative null allele of rh1–1 suggests the absence of genetic compensatory effects, consistent with the late onset and limited expression pattern of rh1–2.
In summary, we have taken advantage of recent advances in genome editing to generate novel alleles of the zebrafish rh1–1 locus, which displays inheritance patterns and molecular features similar to major classes of human alleles associated with disease. Moreover, histological analyses showed patterns of rod dysfunction and degeneration analogous to those contributing to human RP. These lines provide a small yet useful collection of zebrafish models for use in high-throughput in-vivo small molecule screens or genetic screens to identify compounds or gene targets to slow the rod degeneration associated with RP.
This work was supported by grants funded by the Foundation Fighting Blindness, TA-NMT-0614–0654-FSU, and The National Institutes of Health R21EY025410. We thank the members of the Fadool laboratory. We thank the staff of the Department of Biological Science Imaging Resource Facility, and Sequencing and Molecular Cloning Laboratory and the University of Utah Mutation Generation and Detection Core for their assistance designing and constructing the gRNA targeting vectors.