Sensitized chemical mutagenesis screen of Crb1^rd8 mice

Identification of variants contributing to the severe dysplastic phenotype in STOCK Crb1^rd8 mice by classic genetical means of recombinant mapping proved to be inconclusive, likely due to multigenic effects and gene interactions. Therefore, to identify genetic modifiers of Crb1^rd8 dysplasia, two sensitized mutagenesis screens were considered initially: mutagenesis of STOCK Crb1^rd8 mice and screen for decreased severity of the fundus spotting phenotype, or mutagenesis of B6 rd8, an incipient congenic strain with a mild phenotype [19], and screen for increased severity of the phenotype. Mutagenesis of STOCK Crb1^rd8 males failed to yield pups in repeated attempts with varying concentrations and timing of ENU injections. Thus, mutagenized B6 rd8 males were bred according to an established strategy to detect both dominant and recessive mutations [46], and the resulting G₃ population screened for an increased spotting phenotype (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Breeding strategy for chemical mutagenesis.

B6 rd8 G₀ males were treated with N-ethyl-N-nitrosourea (ENU). After about three months when G₀ males regained fertility, they were mated to B6 rd8 females to produce G₁ progeny. G₁ males were mated to produce G₂ progeny, of which four females were backcrossed with the same G₁ male to produce G₃ progeny. G₃ mice were screened by indirect ophthalmoscopy for evidence of increased retinal dysplasia. G₃ founders were developed into TVRM models by testing for heritability and further backcrossing to remove unlinked mutagenized genes. The increased spotting phenotype typical of retinal dysplasia is evident by comparing the fundus images of a B6 rd8 mouse (G₀) and a G₃ founder at 12 weeks of age. Homozygosity of the Crb1^rd8 allele is designated as rd8/rd8; mutant alleles of candidate modifier loci are designated as m/+ (heterozygous) or m/m (homozygous).

https://doi.org/10.1371/journal.pgen.1009798.g001

Identification of mutations

Two mutant lines, Tvrm266 and Tvrm323, exhibited a heritable, bilateral increase in fundus spots in the inferior nasal retina compared to the parental B6 rd8 mice (Fig 2A) and were backcrossed for a minimum of five generations to unmutagenized B6 rd8 to remove unlinked mutations prior to characterization. A high fraction of affected progeny during initial intercrosses to expand the colony suggested that the phenotype in both strains might be semi-dominant. The modifier loci were mapped by crossing these strains with B6N mice, which are homozygous for the Crb1^rd8 allele [36] and have diverged sufficiently from B6 to allow for chromosomal mapping of the modifying loci without contributing to the observed disease phenotypes. For Tvrm266 mice, high-throughput whole-exome sequencing of affected individuals was first performed to identify likely mutation candidates. For Tvrm323 mice, a genome-wide recombinational mapping screen (S1 Fig and S1 Data) of 103 backcross progeny ((Tvrm323 × B6N) F₁ × B6N) revealed that the interval yielding the highest frequency of affected heterozygotes (72%) was a 15.3 Mbp region of Chromosome 3 (Chr 3; 03–021837059 to 03–037129353). Putative causative mutations were identified in Arhgef12, by Sanger sequencing confirmation of candidate gene variants detected from the whole-exome sequences of Tvrm266 mice, and in Prkci, by candidate gene sequencing within the confidence interval identified by mapping Tvrm323 mice (Figs 2B and S2). T to A point mutations occurred in both models resulting in an early stop signal at codon 24 of Arhgef12 (Fig 2B, asterisk) and substitution of asparagine for tyrosine at codon 136 of Prkci. Co-segregation of disease phenotype with the mutations identified was statistically significant by endpoint genotyping analysis of 16 progeny from a Tvrm266 mapping cross (Tvrm266 × B6N) F₁ × B6N; p = 0.010, Fisher’s Exact Test) and by sequencing 12 progeny from the Tvrm323 mapping cross (p = 5.4 × 10⁻⁵, Fisher’s Exact Test), indicating that these mutations are causative (S2 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Identification of mutations and protein expression in modifier strains.

A. Fundus photodocumentation of B6 rd8 and Tvrm266 mice at 10 weeks of age, and Tvrm323 mice at 3 months of age. Right (OD) and left (OS) eyes of a single mouse are shown. The lower portion of each image corresponds to the inferior retina based on the position of the mouse head relative to the camera during imaging. B. Sequence and amino acid translation of the mutant alleles. The DNA sequence corresponds to a portion of the forward strand of each gene, numbered as in the mRNA. Mutations are indicated in bold above the sequence. The predicted amino acid sequence is given beneath the DNA sequence using the single-letter code and the predicted translational effect of the mutation is indicated in bold below. C. Enhanced chemiluminescence western analysis. Posterior eyecup lysates from B6 control and modifier strains (n = 4) were electrophoresed and transferred to nitrocellulose. Tvrm266 blots were probed with ARHGEF12 antibody and Tvrm323 blots with PRKCI antibody. After imaging, blots were probed with GAPDH antibody and reimaged to control for protein recovery and loading. Molecular weights of protein standards are indicated in kDa.

https://doi.org/10.1371/journal.pgen.1009798.g002

To assess the effect of the mutations on expression of the encoded proteins, western analysis was performed on posterior eyecups. Analysis of ARHGEF12 in control B6 (Figs 2C and S3) and B6 rd8 (S3 Fig) samples revealed a single band near the predicted molecular weight of ARHGEF12 (171 kDa predicted for isoform X1, XP_017169054.1). Similar analysis of PRKCI in B6 mice revealed a 66,000 M_r band consistent with the 65 kDa predicted for isoform X1 (XP_006535474). Quantitative western analysis (S4 Fig and S1 Data) indicated that the level of ARHGEF12 in Tvrm266 eyes was no more than 0.037 of that in B6 eyecups (Student’s t-test, p = 0.0015); this value is likely to be an overestimate arising from non-specific staining of other eyecup proteins in the blot region analyzed. Little to no expression is expected due to the premature stop codon caused by the mutation. The expression of ARHGEF12 in both B6 and B6 rd8 mice (S3 Fig) suggested that the homozygous Crb1 mutation was not the cause of the dramatic decrease in ARHGEF12 expression in Tvrm266 mice. PRKCI levels in Tvrm323 mice were 0.57 of those in B6 mice but this effect was not statistically significant (Student’s t-test, p = 0.052). Comparison of B6 rd8, Tvrm266, and Tvrm323 to B6 control eyecups by qRT-PCR (S1 Data) indicated statistically significant differences in Arhgef12 mRNA levels among strains (one-way ANOVA of ΔC_t values, F[3,20] = 6.869, p = 0.0023). Dunnet’s post-hoc test revealed a significant decrease in Arhgef12 mRNA levels (0.74 of B6, p = 0.0018) but not in other strains. This decrease may reflect nonsense-mediated decay of a subset of Arhgef12 transcripts in Tvrm266 mice. Parallel analysis of Prkci mRNA did not indicate significant post hoc differences among the strains compared to B6 controls (S1 Data). Taken together, these results establish novel alleles of two genes as candidate modifiers of the Crb1^rd8 fundus phenotype. The full designation for these models is B6.Cg-Crb1^rd8 Arhgef12^Tvrm266/Pjn and B6.Cg-Crb1^rd8 Prkci^Tvrm323/Pjn; for simplicity, we retain the original strain designations Tvrm266 and Tvrm323, respectively, for the remainder of the paper, and unless otherwise indicated the genotype is homozygous for both Crb1^rd8 and the modifier.

Epistasis of Crb1^rd8 modifier genes

We postulated that the increased retinal spotting phenotypes of Tvrm266 and Tvrm323 may be due to the new mutations acting alone or through an interaction with Crb1^rd8. To distinguish these possibilities, we tested for epistasis, classically defined as the masking of a phenotype associated with a variant at one locus by a variant at a second [50]. To examine epistatic interactions, each modifier strain was outcrossed to B6 mice, which are wild-type for Crb1, and F₁ progeny subsequently intercrossed to yield F₂ mice. In the case of Tvrm266, F₂ mice progeny were genotyped and intercrossed to produce F₃ mice of select genotypes for additional analysis. An increase in fundus spotting was observed in heterozygous or homozygous Arhgef12^Tvrm266 or Prkci^Tvrm323 mice in the presence of the homozygous, but not heterozygous Crb1^rd8 allele (Fig 3A). The fundus phenotype was more severe in the presence of two alleles (homozygous) than with one allele (heterozygous) of either Arhgef12^Tvrm266 or Prkci^Tvrm323. These results were extended by indirect ophthalmoscopy to all nine genotypes expected from the epistasis intercross (Fig 3B and 3C). Phenotypes of heterozygous Crb1^rd8 mice were similar to those of B6 mice, which rarely exhibited spots (Fig 3A). Interaction modeling of indirect ophthalmoscopy data (S1 Data) from epistasis crosses of both mutants indicated a significantly improved fit using a model that accounted for gene-gene interaction (Fig 3B and 3C) compared to one that was purely additive (adjusted R² value for interacting vs. additive models: Tvrm266 cross, 0.78 vs. 0.62; Tvrm323 cross, 0.89 vs. 0.69). An F-ratio test of the F-statistic derived from the two statistical models was significant, indicating superior performance of the interacting model for both mutants (F-statistic and p values: Tvrm266 cross, 26.1 and p = 6.3 × 10⁻¹⁶; Tvrm323 cross, 97.9 and p = 5.4 × 10⁻⁴⁶). Both analyses revealed a significant effect of the heterozygous allele on the phenotype (Tvrm266 cross, p = 1.9 × 10⁻⁵; Tvrm323 cross, p = 1.6 × 10⁻⁸), which was smaller than that of the homozygous allele, confirming the semi-dominant mode of inheritance. Taken together, these results reveal a semi-dominant epistatic interaction between Arhgef12 or Prkci mutant alleles and the Crb1^rd8 mutation leading to increased fundus spotting.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3.

A. Fundus photographs of B6, B6 rd8, and F₃ progeny from the Tvrm266 epistasis cross at 10 weeks of age or F₂ progeny from the Tvrm323 epistasis cross at 2.5–3.7 months of age. Genotypes are indicated by m/+ (heterozygous) or m/m (homozygous), where m is the mutant allele symbol and, for comparison with panels B and C, by the values a, b at the upper right of each fundus image, which indicate the number of mutant alleles at the modifier and Crb1 loci, respectively (0, wild-type; 1, heterozygous; 2, homozygous). At least 3 females and 3 males of each genotype were imaged bilaterally. B. 3D plot of the indirect ophthalmoscopy phenotypes of F₂ and F₃ progeny from the Tvrm266 epistasis cross (9–15 weeks of age) against the nine genotypes expected (n = 6–21 for each genotype). C. Corresponding plot of F₂ phenotypes from the Tvrm323 epistasis cross (2.5–3.7 months of age; n = 10–48 for each genotype). Indirect ophthalmoscopy annotations were converted to ordinal phenotype scores. Scores for individual mice of each genotype examined are shown in blue, with fit statistical models represented by red circles. Genotypes values are indicated as in panel A.

https://doi.org/10.1371/journal.pgen.1009798.g003

Pathological features of modifier strains

Having established that the Arhgef12 and Prkci alleles modify the Crb1^rd8 phenotype, we sought to gain insights into disease pathogenesis by studying the modified strains. Analysis was performed on homozygous Tvrm266 and Tvrm323 mice, as the phenotype could be observed at one month of age. To determine whether increased fundus spotting was correlated with dysplastic lesions and other pathological changes, eyes collected from B6, B6 rd8, Tvrm266, and Tvrm323 mice were examined histologically (Fig 4). We defined dysplastic lesions as retinal folds or perturbations resulting in the displacement of photoreceptor nuclei by more than three nuclear diameters from either boundary of the outer nuclear layer (ONL). B6 rd8 retinas (Fig 4B) were morphologically similar to B6 retinas (Fig 4A) except for rare dysplastic lesions in the inferior retina (Fig 4B, asterisk; also compare the higher magnification images in Fig 4E and 4F). By contrast, retinal morphology in the inferior retina was disrupted more extensively in the modifier strains, as reported previously in STOCK Crb1^rd8 mice [19,23,34] and other Crb1 mutant models [20]. Lesions were also evident in the superior retina of Tvrm266 and Tvrm323 mice near the optic nerve head. Dysplastic lesions were larger than those in B6 rd8 mice (compare Fig 4G and 4H with Fig 4F) and differed between the two modifier strains. In Tvrm266 mice, the largest lesions (Fig 4C, asterisks; Fig 4G) extended toward the inner nuclear layer (INL) and resembled the pseudorosettes described in STOCK Crb1^rd8 mice and other models [19,20,23,34]. By contrast, Tvrm323 lesions were characterized by displacement of retinal cell layers toward the RPE (Fig 4D, asterisks; Fig 4H) and often contained tubular structures that reached and perturbed cells of this layer (Fig 4H, red arrowhead). These structures may represent retinal neovessels, a frequently reported attribute of Crb1 mutant mice and rats [35,41,51]. A novel feature discovered in modifier strains at one month of age was the presence of apparent voids in the ONL near dysplastic lesions (Fig 4G and 4H, yellow arrowheads). One sample with ONL voids was also observed among nine B6 rd8 mice examined (Fig 4I).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Histological analysis at one month of age.

A–D. Superior (S) and inferior (I) retinas cropped from digital images of B6, B6 rd8, Tvrm266 and Tvrm323 whole eye sections stained with hematoxylin and eosin and imaged with a slide scanner. Dysplastic lesions, asterisks. Scale bar, 0.5 mm. Sites from A-D detailed in panels E–H, respectively, are marked by arrowheads. E. Normal retinal layers in B6 mice. INL, inner nuclear layer; ONL, outer nuclear layer; RPE, retinal pigment epithelium. F–H. Dysplastic lesions in B6 rd8, Tvrm266, and Tvrm323 mice, respectively. Voids in panels G and H, yellow arrowheads. H. Lesion containing possible vascular structure, red arrowhead. Scale bar, 50 μm; applies to E–H. I. Incidence of pathological features. Mice showing one or more of the identified features among 17–21 reviewed histological sections were scored as affected. Asterisks denote statistical significance (Fisher’s exact test against B6 rd8 mice, p < 0.05). DL, dysplastic lesions; NV, neovessels. J. Photoreceptor nuclei count within a 50 μm length of retina as a function of distance from the ONH. Values represent mean ± standard error of the mean (SEM); n = 5–8.

https://doi.org/10.1371/journal.pgen.1009798.g004

To assess the statistical significance of these findings, histological sections from each independent sample were reviewed for each pathological feature, scoring as affected any samples in which the feature occurred at least once. Compared to B6 rd8 mice, significant differences in incidence were observed only for dysplastic lesions among B6 mice (Fisher’s exact test, Bonferroni post hoc correction, p = 0.0021), and for neovascular lesions among Tvrm323 mice (Fisher’s exact test, Bonferroni post hoc correction, p = 0.0026; Fig 4I, asterisks; S1 Data). To determine whether these pathological changes were associated with photoreceptor degeneration, photoreceptor nuclei were counted in regions corresponding to a fixed length of the retinal perimeter, which were positioned at regular intervals over the entire retina. At one month of age, no significant effect of strain on the number of photoreceptor nuclei was observed among B6 mice or modifier strains compared to B6 rd8 mice (Fig 4J, repeated measures mixed-effects model, F (3, 22) = 1.8, p = 0.18; S1 Data). These results show that retinal dysplasia and associated pathological changes occur at an early age prior to significant degeneration. The topographical distribution of lesions in the inferior retina and the superior retina near the ONH matched that of fundus spots, indicating that retinal dysplasia accounts for the increased fundus spotting phenotype.

Age-dependent pathological changes in modifier strains

Most forms of CRB1 retinal disease are progressive in humans [1–6]. To examine whether progressive changes occurred with age in the modifier strains, histological analysis was repeated at 12 months of age (Fig 5). B6 rd8 retinal morphology was similar to that of B6 retinas in aged mice (Fig 5A and 5E), except for rare dysplastic lesions as observed at one month of age (Fig 5B and 5F). By contrast, in addition to a continued presence of dysplastic lesions at 12 months of age (Fig 5C and 5D, asterisks), retinas in both modifier strains exhibited areas of complete photoreceptor loss, which were more extensive in Tvrm323 than in Tvrm266 mice (Fig 5C and 5D, solid lines). ONL voids were not observed, but neovascular lesions resulting in RPE perturbation were apparent in both modifier strains at this age (Fig 5G and 5H, red and white arrowheads). The incidence of neovascular lesions was statistically significant in both modifier strains compared to B6 rd8 (Fig 5I and S1 Data). Analysis of photoreceptor nuclei counts indicated a statistically significant effect of strain (mixed-effects model, F (3, 36) = 21.42, p < 0.0001). In post hoc analysis of data from both modifier strains, a statistically significant decrease in nuclei counts was observed near the optic nerve head and extended more peripherally in the inferior than in the superior retina (Fig 5J and S1 Data). Importantly, degeneration was limited to areas where dysplastic lesions were observed at one month of age (Fig 4C and 4D) and the peripheral retina was largely spared, as previously reported in STOCK Crb1^rd8 mice [19]. No significant difference in nuclei counts was observed in post hoc analysis comparing B6 and B6 rd8 at 12 months of age, although an age-related decrease in counts throughout the retina was evident in both strains (compare Figs 4J and 5J). Taken together with the histological findings at one month of age, these results establish that dysplastic lesions, neovascular structures, and photoreceptor cell loss occur and progress to different extents among B6 rd8, Tvrm266, and Tvrm323 mice, supporting the hypothesis that modifier genes contribute to Crb1^rd8 retinal disease variability.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Histological analysis at 12 months of age.

A–D. Superior (S) and inferior (I) retinas cropped from representative digital images of B6, B6 rd8, Tvrm266 and Tvrm323 whole eye sections stained with hematoxylin and eosin and imaged with a slide scanner. Dysplastic lesions, asterisks. Areas devoid of photoreceptor nuclei, solid lines. Scale bar, 0.5 mm. Sites from A-D detailed in panels E–H, respectively, are marked by arrowheads. E. Normal retinal layers in B6 mice, labeled as in Fig 4. F. Dysplastic lesion in B6 rd8. G, H. Photoreceptor loss at neovascular lesions (red arrowheads) in Tvrm266, and Tvrm323 mice, respectively. In G, red blood cells are observed within a large vessel at the RPE. I. Incidence of pathological features among mice. Samples showing one or more of the identified features among 10–21 sections were scored as affected. DL, dysplastic lesions; NV, neovessels. Asterisks denote statistical significance (Fisher’s exact test against B6 rd8 mice, p < 0.05). J. Photoreceptor nuclei count within a 50 μm length of retina as a function of distance from the ONH. Values represent mean ± SEM; n = 6–8. Statistical significance colored according to the strain compared to B6 rd8 in post hoc tests. p < 0.01, asterisk; p < 0.0001, hash symbol.

https://doi.org/10.1371/journal.pgen.1009798.g005

Cellular disorganization at dysplastic lesions

Retinal dysplasia associated with mutant Crb1 alleles in mice and rats is characterized by mislocalized photoreceptor cells, shortened photoreceptor outer segments, dysmorphic Müller cells, and the mobilization of immune cells, including microglia and possibly infiltrating macrophages, to the subretinal space [19,35,38,43,51]. Mislocalized rod photoreceptor cells were readily identified in histological sections, as the nuclei of these cells in adult mice are compact and stain intensely with hematoxylin (Fig 4). To assess rod outer segment integrity and to test if other retinal cells were abnormally distributed, immunofluorescence analysis of retinal sections was performed using antibody markers for photoreceptor outer segments, Müller glia, and microglia (Fig 6). Rod and cone outer segments were correctly localized to the outer retina in B6 mice (Fig 6A) and in intact retinal regions of B6 rd8, Tvrm266, and Tvrm323 mice (Fig 6B–6D). By contrast, outer segments were absent from the RPE interface at dysplastic lesions in all three mutant strains. Small foci of rhodopsin (RHO) and M opsin (OPN1MW) staining were observed throughout the dysplastic areas, consistent with mislocalization of photoreceptor cells.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Ectopic retinal cells in dysplastic lesions at one month of age.

Retinal sections of 7–9 B6 (A, E, I, M), B6 rd8 (B, F, J, N), Tvrm266 (C, G, K, O), and Tvrm323 (D, H, L, P) mice were examined for each panel of markers studied. A–D. Staining with rhodopsin (RHO, red) and M cone opsin (OPN1MW, green) antibodies indicated a normal location of rod and cone outer segments in B6 mice (A) but an ectopic location at dysplastic lesions of Crb1^rd8 mutants (B–D; white and yellow arrowheads, respectively). E–H. Glutamine synthetase (GLUL, green) antibodies revealed Müller cell soma at the center of the INL in B6 mice (E, detailed in I) but displaced toward the ONL at lesions of Crb1^rd8 mutants (F–H, white arrowheads, detailed in J–L). M–P. Purinergic receptor P2RY12 antibodies (green) revealed microglia at their normal location in the OPL of B6 mice but mobilized to the ONL and subretinal space among dysplastic lesions of Crb1^rd8 mutants (arrowheads). Nuclei were detected with DAPI (blue). Scale bars: H, 50 μm, applies to A–H and M–P; L, 25 μm, applies to I–L.

https://doi.org/10.1371/journal.pgen.1009798.g006

In some instances, intact cone outer segments were observed in the interior of pseudorosettes (Fig 6C, arrowheads), suggesting infolding of the intact outer retina. In B6 mice, Müller glia were observed with soma positioned at the center of the INL and retina-spanning processes extending to endfeet at the internal limiting membrane (ILM) and ELM (Fig 6E and 6I). Similar morphology was observed in unaffected regions of B6 rd8, Tvrm266, and Tvrm323 retinas (Fig 6F–6H), but at dysplastic lesions, Müller cell bodies were occasionally displaced toward the ONL (Fig 6F–6H and 6J–6L arrowheads). Müller cell displacement was observed in 2/4 lesions that were imaged in three B6 rd8 mice, 4/12 lesions in seven Tvrm266 mice, and 14/18 lesions in seven Tvrm323 mice. These results suggest that Müller cell displacement is a common feature of dysplastic lesions, particularly those that are more severe as in Tvrm323 mice. Microglia were found in their normal locations on the outer surface of the INL in B6 mice (Fig 6M) and in unaffected regions of the three Crb1^rd8 strains (Fig 6N–6P). By contrast, at dysplastic lesions of these strains, microglial cells accumulated in the subretinal space or appeared to be displaced toward this region (Fig 6N–6P). P2RY12 staining between the outer plexiform layer and subretinal space indicated displacement of microglial cells and was observed in 4/5 lesions imaged from five B6 rd8 mice, 7/9 lesions in six Tvrm266 mice, and 24/24 lesions in eight Tvrm323 mice. This observation provides evidence for a strong correlation between lesions and displaced microglia. As in other Crb1 mouse and rat models [19,35,38,43,51], these results emphasize an enhanced focal pathology in Tvrm266 and Tvrm323 mice that is limited to dysplastic lesions, in which photoreceptor cells are displaced toward the INL from their normal locations and Müller cells and microglia are mobilized toward the outer retina and RPE.

ELM fragmentation

Pan-retinal disruption of the ELM is a distinct feature of homozygous Crb1^rd8 mice thought to arise from a defect in the formation and maintenance of cell-cell adhesive interactions, which depend on CRB1 [19]. This feature is observed in retinal sections of Crb1^rd8 mutant mice immunostained with antibodies against components of the cell adhesion machinery, such as β-actin, cadherins, PALS1, or TJP1 [19,40], which reveal a fragmented appearance characterized by small gaps in the normally continuous ELM. It is unknown whether retinal dysplasia requires ELM fragmentation or is instead the result of an independent pathological process. To test whether fragmentation was affected by the modifier mutations, we stained retinal sections to detect TJP1. The ELM was continuous in the inferior (Fig 7A) and superior retina (Fig 7E) of B6 mice but exhibited gaps in the corresponding regions of B6 rd8 (Fig 7, 7B and 7F) and Tvrm266 mice (Fig 7, 7C and 7G). Strikingly, however, ELM fragmentation was rare in Tvrm323 retinas (Fig 7, 7D and 7H), despite an increase in the number and size of dysplastic lesions. The decrease in the number of gaps relative to B6 rd8 was particularly noticeable in the superior retina of Tvrm323 mice, which at one month of age had no dysplastic lesions (compare Fig 7H and 7F). Note that the ELM is distorted at dysplastic lesions (for example, Fig 7B) and therefore fragmentation cannot be evaluated at these sites. Gap counts normalized to ELM length in areas unaffected by dysplasia were determined as a measure of fragmentation (Fig 7I and S1 Data). A statistically significant difference in this measure was observed in one or more strains compared to B6 rd8 (one-way ANOVA, F (3, 12) = 38.90, p <0.0001; S1 Data). In post hoc analysis corrected for multiple comparisons by Dunnet’s test, statistically significant differences in fragmentation were observed in B6 and Tvrm323 (p < 0.0001), but not Tvrm266 (p = 0.091) mice, compared to B6 rd8 mice (Fig 7I, asterisks; S1 Data). These results provide further evidence that genetic modifiers affect Crb1^rd8 phenotypes differentially and raise the possibility that dysplasia and ELM fragmentation may be independent pathological changes arising from the Crb1^rd8 mutation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. ELM fragmentation in modifier strains.

Retinal sections from mice at one month of age were stained with antibodies against TJP1 (red) to detect the ELM and DAPI (blue) to detect nuclei. A–D. Inferior retina. E–H. Superior retina. A, E, B6; B, F, B6 rd8; C, G, Tvrm266; D, H, Tvrm323. Gaps are indicated by yellow arrowheads. Images are representative of n = 8 samples from each strain. H. Scale bar, 50 μm, applies to all image panels. I. Count of gaps over the full length of ELM except at dysplastic lesions, normalized to ELM length (n = 3–4). Bars indicate mean ± SEM. Asterisks, p < 0.0001, post hoc analysis.

https://doi.org/10.1371/journal.pgen.1009798.g007

We also tested whether the homozygous Arhgef12^Tvrm266 mutation might result in an ELM phenotype in the absence of the homozygous Crb1^rd8 allele. Compared to Tvrm266 mice, the gap number per mm of ELM in mice bearing the homozygous Arhgef12^Tvrm266 mutation and a heterozygous Crb1^rd8 allele was significantly reduced to a value near zero (S5 Fig). This result indicates that the ELM fragmentation phenotype of Tvrm266 mice requires homozygous Crb1^rd8 and suggests that the Arhgef12 mutation does not result in gross changes to the ELM.

Effect of modifiers on neonatal retinal progenitor cell development

Studies of severely dysplastic Crb1 mutant mice that also carry conditional mutations in Crb2 targeted to photoreceptor cells, retinal progenitor cells, or Müller glia have revealed defects in early retinal development [43]. In these strains, mitotic progenitor cells of the retinal neuroblastic layer (NBL), which at P1 are normally localized within a narrow band near the RPE, are distributed ectopically throughout the NBL and even reach the ganglion cell layer [43]. To test for ectopic localization of mitotic progenitor cells in our Crb1^rd8 modifier strains, we examined retinal sections at P0 from B6, B6 rd8, Tvrm266, and Tvrm323 mice stained with antibodies against phosphohistone H3 (PHH3) (Fig 8), which identifies cells at mitotic prophase and metaphase. Mitotic progenitor cells were similarly positioned at the NBL apical surface of all strains (Fig 8A–8H). Analysis of the PHH3 fluorescence distribution of P0 mice (Fig 8I) revealed no significant difference due to strain (one-way ANOVA, F (3, 19) = 1.48, p = 0.25, n = 3, 6, 5, and 7 for B6, B6 rd8, Tvrm266, and Tvrm323, respectively). Similarly, TJP1 staining, which at P0 detects the prospective ELM, revealed few gaps (Fig 8E–8H). There were no statistically significant differences in the number of gaps per mm of ELM (Fig 8I) due to strain (one-way ANOVA, F (3, 12) = 0.92, p = 0.47, n = 3, 4, 4, and 5 for B6, B6 rd8, Tvrm266, and Tvrm323, respectively). Thus, our analysis does not reject the hypotheses that mitotic progenitor cell localization and prospective ELM integrity at P0 are unaffected by the homozygous Crb1^rd8 mutation alone or in combination with the Arhgef12 or Prkci modifier mutations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Distribution of mitotic progenitor cells at P0.

A–D. Representative ocular sections of B6 (n = 5), B6 rd8 (n = 6), Tvrm266 (n = 5), or Tvrm323 (n = 7) mice, respectively, were probed with antibodies against phosphohistone H3 (PHH3, red) to detect mitotic cells and DAPI (blue) to detect nuclei. A similar distribution of PHH3-positive staining within the outer NBL (A, arrow) was observed in all strains. E–H. Detail of P0 retinas stained as in A–D and with antibody against TJP1 (red). The order of samples is the same as in A–D. Placement of progenitor cells undergoing mitosis close to the apical boundary of the NBL, delineated by TJP1, was observed in all strains. NBL, neuroblastic layer. Scale bars: D, 0.25 mm, applies to A–D; H, 50 μm, applies to E–H. I. Quantitation of PHH3 fluorescence in a 20 μm-wide band at the apical NBL relative to the total NBL. J. Quantitation of gaps per mm of the prospective ELM. Bars in I and J indicate mean ± SEM.

https://doi.org/10.1371/journal.pgen.1009798.g008

Loss of visual function in modifier strains

Loss of vision in CRB1-associated retinal disease occurs congenitally in LCA patients or progresses rapidly during early life in RP and cone-rod dystrophy patients. In comparison, mouse strains carrying a single homozygous Crb1 mutation exhibit little to no progressive reduction in visual function as measured by electroretinography (ERG) [20,22,23], limiting their utility as models of CRB1-associated vision loss. Functional loss in mouse models is accelerated in strains combining homozygous Crb1 mutations with modifier mutations, such as those in Crb2 [42,43,52]. To assess whether a decline in visual function was accelerated in Crb1^rd8 modifier strains, B6, B6 rd8, Tvrm266, and Tvrm323 mice were examined by ERG. As no significant loss of photoreceptors was noted histologically at one month of age, scotopic and photopic ERG response amplitudes as a function of flash intensity were measured at four, eight, and 12 months of age (Fig 9A). Scotopic a- and b-wave amplitudes provide a measure of rod photoreceptor and secondary neuron function while the photopic b-wave amplitude measures secondary neuron responses to cone stimulation. Responses at each age were compared statistically to those of B6 rd8 by mixed-effects analysis, a repeated-measures approach that allows datasets with occasionally missing values. By this analysis (S1 Data), statistically significant differences in mean scotopic a-wave amplitude were detected in one or more strains compared to B6 rd8 mice (4 months, p = 0.0306; 8 months, p = 0.0022; 12 months; p = 0.0004); in mean scotopic b-wave amplitude (8 months, p = 0.0009; 12 months; p = 0.0002); and in mean photopic b-wave amplitude (8 months; p = 0.0034; 12 months; p = 0.0021). Post hoc analysis with Dunnet’s correction for multiple comparisons revealed a statistically significant decrease in Tvrm266 a-waves amplitude at a single flash intensity (4 months; -1.8 log cd s/m²; p = 0.046) and in Tvrm323 at multiple intensities (12 months; -1.4, -1.0, and-0.6 log cd s/m²; p = 0.017, 0.032, and 0.35, respectively). Parallel analysis indicated an unusual and statistically significant increase in mean Tvrm266 b-wave amplitude (8 months; -1.4, -1.0, and-0.6 log cd s/m²; p = 0.031, 0.013, and 0.0070, respectively) as well as statistically significant decreases in mean Tvrm323 b-wave amplitude (8 months; 1.8 log cd s/m²; p = 0.049; 12 months; -2.6, -2.2, -1.8, -1.0, and-0.6 log cd s/m²; p = .0098, 0.013, 0.029, 0.013, and 0.0098, respectively). No other comparisons, including those of B6 to B6 rd8 were found to be statistically significant (S1 Data). These results indicate a transiently increased ERG response in Tvrm266 mice at eight months of age, which is unexpected in light of the decreased number of photoreceptor cells predicted from histological analysis. They also indicate a loss of photoreceptor function in Tvrm323 mice as indicated by the statistically significant decrease in scotopic a- and b-wave amplitudes at 12 months of age.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. ERG analysis of B6, B6 rd8, Tvrm266, and Tvrm323 mice at four, eight, and 12 months of age.

A. Scotopic and photopic response amplitudes as a function of flash luminance. Asterisks show statistical significance in a comparison to B6 rd8 and are colored as in the legend to indicate the strain compared. * p < 0.05; ** p < 0.01. Values indicate mean ± SEM; n = 6–26. B. Response amplitudes at the highest flash luminance as a function of age, normalized to the mean amplitude at four months of age. Legend is the same as in panel A. Asterisks indicate statistical significance compared to the value at four months of age and are colored as in the legend in panel A. * p < 0.05, ** p < 0.01; *** p < 0.001; **** p < 0.0001.

https://doi.org/10.1371/journal.pgen.1009798.g009

As an alternative measure of progressive functional changes within each strain, we reanalyzed ERG response amplitudes obtained at the highest scotopic and photopic flash intensities as a function of age (Fig 9B and S1 Data). In this analysis, amplitude values at each age were normalized and compared to the mean value at four months of age. Statistically significant differences in response amplitudes at one or more ages were detected in all strains except B6 (S1 Data; two-way ANOVA; scotopic a- and b-wave, photopic c-wave; p < 0.0001). Post hoc analysis with Dunnet’s correction for multiple comparisons revealed a statistically significant decrease in mean B6 rd8 response amplitudes (12 months; scotopic a-wave; p = 0.0082; scotopic b-wave, p = 0.0020; photopic b-wave, p = 0.013); a statistically significant increase in Tvrm266 amplitudes (8 months; scotopic a-wave; p = 0.0001; scotopic b-wave, p < 0.0001) and a statistically significant decrease in Tvrm323 amplitudes (8 months; scotopic a-wave; p = 0.0007; scotopic b-wave, p < 0.0001; photopic b-wave, p = 0.017; 12 months; scotopic a-wave; p < 0.0001; scotopic b-wave, p < 0.0001; photopic b-wave, p < 0.0001). Comparison of the mean values indicated a progressive and statistically significant decrease in ERG response amplitudes in B6 rd8 mice of 30–40% from four to 12 months of age, which was exacerbated in Tvrm323 mice to 44–63% over the same period, consistent with degeneration of rod and cone photoreceptor cells (Fig 9B). Tvrm266 mice showed increased scotopic responses of 45–51% at eight months of age, a difference which was not sustained at 12 months of age. In summary, the ERG data provide evidence for functional changes in the modifier strains, which differ from each other and from B6 rd8 mice, supporting the hypothesis that Crb1^rd8 modifiers differentially affect retinal function.

Ethics statement

All procedures utilizing live mice were approved by the JAX Institutional Animal Care and Use Committee (protocol ACUC 99089).

Colony management and ENU mutagenesis

STOCK Crb1^rd8 mice were backcrossed to C57BL/6J (JAX stock number 000664) mice for seven generations to produce an incipient congenic strain, B6.Cg-Crb1^rd8/Pjn (N₇), which was bred to homozygosity. Males from this incipient congenic colony were administered weekly intraperitoneal injections of ENU (Millipore Sigma, Burlington, MA, USA; N3385) for three consecutive weeks at a concentration of 80 mg/kg per treatment [109,110]. Following return to fertility, the mutagenized G₀ males were backcrossed to B6.Cg-Crb1^rd8/Pjn females, producing G₁ mice which were subsequently outcrossed to unmutagenized B6.Cg-Crb1^rd8/Pjn (N₇) and resulting female G₂ mice were backcrossed to G₁ sires to produce G₃ mice [46].

G₃ mice were screened at 12 weeks of age using indirect ophthalmoscopy. Heritability of novel ocular phenotypes was established by outcrossing to B6.Cg-Crb1^rd8/Pjn (N₇) mice, producing F₁ progeny to test for dominance. If no phenotype was observed, F₁ mice were then intercrossed, to produce F₂ progeny, and screened for recessive phenotypes. Once heritability was established, mutant mice were assigned a number in the TVRM program [46] and lines were bred and maintained in the JAX Research Animal Facility. Mutant strains were backcrossed five generations (N₅) to the founder strain, B6.Cg-Crb1^rd8/Pjn, to remove potential unlinked mutations before samples for characterization were collected. The full designations for the modifier strains are B6.Cg-Crb1^rd8 Arhgef12^Tvrm266/Pjn and B6.Cg-Crb1^rd8 Prkci^Tvrm323/Pjn, which are abbreviated as Tvrm266 and Tvrm323, respectively, and refer to a genotype that is homozygous for both Crb1^rd8 and the modifier allele unless otherwise indicated.

Mice were provided with NIH 6% fat chow diet and acidified water, with a 12:12 hour dark:light cycle in pressurized individual ventilation caging and were monitored regularly to maintain a pathogen-free environment.

Fundus photodocumentation

Mouse pupils were dilated using 1% Cyclomydril (Alcon) prior to imaging. Isoflurane (Kent Scientific) was delivered at 2–4% with an oxygen flow rate of 0.5 liter/min to anesthetize mice during imaging. Color fundus videos (100 frames) were acquired using a Micron IV Retinal Imaging Microscope (Phoenix Technology Group), and registered, averaged, and sharpened in Fiji [111] using the ImageStabilizeMicronStack macro as described [112]. Images were further processed using a custom Fiji macro that applies the Polynomial Shading Corrector plugin (degree x = 2, degree y = 2, regularization = 50) to flatten brightness across the image and resets the range of each channel to optimize color balance.

Gene identification

Based on observations during initial breeding, the mutations in Tvrm266 and Tvrm323 appeared to segregate as semidominant alleles. To determine the map position of Tvrm323, affected mice (enhanced spots) were outcrossed to C57BL/6NJ (B6N, JAX stock number 005304) and the resulting F₁ progeny were backcrossed to B6N. All backcross progeny homozygous for the Crb1^rd8 mutation were phenotyped by indirect ophthalmoscopy. DNA from 103 backcross progeny (59 unaffected, 44 enhanced) were genotyped with 149 single nucleotide polymorphisms (B6/B6N SNP panel) spanning the genome (JAX Fine Mapping Facility). We identified genomic regions where >70% of the B6 allele co-segregated with the enhanced phenotype. Candidate genes for Tvrm323 were identified in the minimal confidence interval on Chr 3. Gene-specific PCR amplification and subsequent Sanger sequencing confirmed a missense mutation in Prkci. To identify the causative mutation in Tvrm266, high-throughput exome sequencing was performed on a whole-exome library created from Tvrm266 genomic DNA on a HiSeq 2000 Sequencing System (Illumina) as previously described [113]. Whole-exome sequencing of Tvrm266 identified a nonsense mutation in Arhgef12, which was subsequently confirmed by Sanger sequencing. An abbreviated mapping cross, with 16 progeny (8 unaffected, 8 enhanced) confirmed the cosegregation of the affected Tvrm266 phenotype with the Arhgef12 mutation on Chr 9.

For strain maintenance, modifier mice were genotyped using DNA isolated from tail tips incubated with 50 mM sodium hydroxide at 95°C for about 20 min before being neutralized with 1M Tris-Cl, pH 8.0, to a final concentration of 50 mM. PCR amplification was performed in a 10 μl reaction volume containing 1X PCR Buffer (New England Biolabs), 10 μM dNTP and 0.05 U Taq DNA polymerase (New England Biolabs). An initial denaturation step at 97°C for 2 minutes was followed by 45 cycles of 10 seconds at 95°C, 15 seconds at 55°C, and 30 seconds at 72°C with a final 3 minute extension at 72°C. Prkci genotypes were determined using an allele-specific protocol combining the following primers (Eurofins Genomics): Prkci-F, TCCAGCAGTAAGTATGGGAAAC; Prkci-WT-R, GTGTGGCCATTTGCACAACA; and Prkci-MUT-R2, GTTTGGCTTGAAAAACGTGGCCATTTGCACAGTT. Depending on zygosity, this protocol yielded amplified products of 145 bp (WT), 159 bp (mutant), or both. Arhgef12 genotypes were determined by direct sequencing. DNA samples were amplified with primers (Arhgef12-266seqF, CACACACACGTCACTGTAAA; mArhgef12-266seqR, GAGTGCCTCAATCCACATAAG). The PCR products were purified and sequenced using the reverse primer.

Western analysis

B6, B6 rd8, Tvrm266, and Tvrm323 mice (n = 4) adult eyes at approximately two months of age were dissected in ice-cold PBS with EDTA-free protease inhibitor cocktail (Roche) to remove the anterior eye and lens, and snap-frozen in eppendorf tubes on dry ice. Eyecups were combined with 100 μl 1× NuPAGE LDS sample buffer/reducing agent (Invitrogen), homogenized immediately using a hand-held pellet pestle motor (Kontes) on ice, and briefly sonicated (Qsonica). Lysates were then centrifuged briefly at 16,000g. A dilution series of the B6 samples was used to standardize the western signal. For eyecups at one month of age (S3 Fig) were combined with ice-cold RIPA buffer (1% NP-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate) containing EDTA-free protease inhibitor cocktail, homogenized, and centrifuged for 20 min at 16,000g. The supernatant was collected and protein concentration determined by Protein Assay reagent (Bio-Rad) using a DU530 spectrophotometer (Beckman Coulter). For each sample, 20 μg was analyzed by gel electrophoresis. Samples were denatured at 100°C for 10 min and electrophoresed on a 10% Mini-Protean TGX gel (Bio-Rad). Electrophoresed proteins were transferred onto a nitrocellulose membrane using Turbo-blot Transfer system (Bio-Rad). Membranes were pre-blocked in Blotto A (5% w/v milk powder in Tris-buffered saline, 0.05% Tween [TBS-T]), and incubated with antibodies against ARHGEF12 (Santa Cruz, sc-15439), PRKCI (Novus, NBP1-84959), or GAPDH (Cell Signaling, #2118) at 4°C overnight. Membranes were washed in TBS-T and incubated with HRP-conjugated anti-goat IgG (R&D, HAF017) or anti-rabbit IgG (Cell Signaling, 7074S) antibody in Blotto A (1:1000) for 2 h at room temperature. Following washes in TBS-T, immunoblots were visualized using the Clarity ECLWestern Blot Detection System (Bio-Rad). Protein band images in 16-bit tif format were captured using an Azure c600 bioimaging system (Azure Biosystems). The raw integrated densities of the ARHGEF12 or PRKCI bands, and the densities for GAPDH bands on the same blots, were determined using Fiji following subtraction of a constant background value evaluated between lanes. The density values for the standard series were plotted against load volume and fit to a hyperbolic curve in Prism, which was used to calculate the relative load volumes for all samples. Relative expression was determined by dividing the equivalent load volumes of the ARHGEF12 or PRKCI bands by those of the corresponding GAPDH bands. The resulting values were normalized by dividing by the mean ARHGEF12 or PRKCI values from all B6 samples.

RNA extraction and analysis

Total RNA was extracted from B6, B6 rd8, Tvrm266, and Tvrm323 mice (n = 5) at one month of age by homogenizing posterior eyecups in TRIzol (Thermo Fisher) using a gentleMACS dissociator (Milteny Biotec). Chloroform was then added and the resulting aqueous phase transferred to 100% ethanol. RNA clean-up was performed using an RNeasy spin column kit (Qiagen). DNase (Qiagen) treatment was performed according to the manufacturer’s instructions. Total RNA was eluted in RNase-free water, quantified using a NanoDrop ND1000 spectrophotometer (Thermo Fisher) and examined for quality using the TapeStation 4150 system (Agilent Technologies). cDNA was synthesized using the Superscript IV Firststrand Synthesis kit (Thermo Fisher).

To design primers for use in real time quantitative PCR, gene and coding sequences were downloaded from mouse Ensembl (https://useast.ensembl.org). Forward and reverse primers amplifying cDNA sequences of 150–250 bp and separated by large intronic regions in the genomic sequence were designed using Primer3 software (http://bioinfo.ut.ee/primer3-0.4.0/), which yielded Arhgef12 rt-F3 (AGAGCCATCAGTCACTGGACA) and Arhgef12 rt-R3 (CAGCCGTTCCTGTTCCTTC) for Arhgef12, and Prkci rt-F1 (GGAGTGAGGAGATGCCGAC) and Prkci rt-R1 (TCATAATATCCCCGCGGTAG) for Prkci. For quantitative RT-PCR, amplification using these primers was performed with iTaq Universal SYBR Supermix (Bio-Rad Laboratories) using the CFX96 Real-time PCR Detection System (Bio-Rad). The relative fold change of gene transcripts was calculated using the comparative CT method (ΔΔC_T), and 2^-ΔΔCT values were normalized to levels of Actb, an internal control calibrator. Melting curve analysis was evaluated to confirm accurate amplification of the target genes.

Histological analysis

Following carbon dioxide euthanasia, eyes were enucleated, placed in chilled methanol:acetic acid:phosphate-buffered saline (PBS) (3:1:4), fixed overnight at 4°C, and stored in 70% ethanol. Fixed eyes were embedded in paraffin and 5 μm sections were stained with hematoxylin and eosin or left unstained for immunohistochemistry (see below). Eyes were oriented to yield sections parallel to the superior-inferior axis by marking the superior eye with a non-soluble paint prior to enucleation. Eyes were also oriented using the long posterior ciliary arteries, which serve as landmarks [114] and could be viewed from the back of embedded eyes illuminated through the cornea. Histosections were scanned using a NanoZoomer 2.0-HT digital slide scanner (Hamamatsu) at 40× magnification; typically 18–21 sections were captured per slide. Sections on the slide were reviewed using NDP.view 2 software (Hamamatsu) to ensure correct orientation based on the position of the central retinal artery to one side of Bruch’s membrane opening at the optic nerve head. All sections were reviewed for the presence of pathological features, including dysplastic lesions, voids, or neovessels, and mice were graded as affected if the feature was observed in one or more sections. To address whether voids were an artifact of histological fixation, we reviewed NanoZoomer images from B6 eyes collected at 0.9–1.1 months of age over a period of nine years and found that no voids were present. To assess photoreceptor cell loss from scanned images, a custom Fiji macro was first used to extract .tif images of single sections from full-resolution NanoZoomer output files. A separate macro was created to draw tilted rectangular regions of interest (ROIs) encompassing a 50 μm length of retina and spaced at 0.3 mm intervals starting from the optic nerve head. Photoreceptor cell loss was determined from .tif images by manually counting photoreceptor cell nuclei within each ROI.

Immunohistochemistry

Unstained sections of eyes were deparaffinized through a xylene/alcohol gradient series. Following deparaffinization, antigen retrieval was performed by microwave treatment for 6 min in 10 mM sodium citrate. Samples were treated with blocking solution (1:50 normal donkey serum:PBS, 0.3% v/v Triton X-100) for 30 min at room temperature and then incubated overnight at 4°C with primary antibodies against rhodopsin (MS-1233-R7, Thermo Fisher), green opsin (AB5405, EMD Millipore), glutamine synthetase (ab176562, Abcam), phosphohistone H3 (PHH3; ab14955, Abcam), P2YR12, (AS-55043A, AnaSpec), and TJP1 (339100 and 61–7300, Thermo Fisher), all diluted 1:100 in blocking solution. Slides were washed using 1× PBS prior to application of secondary antibodies at 1:200 in blocking solution for two h at room temperature. Nuclei were stained using DAPI (10 μg/ml in PBS) for 5 min. Negative controls, in which primary antibodies were omitted during the incubation stage, were performed in parallel. All samples were mounted in Vectashield (Vector Laboratories), coverslipped and imaged using a Zeiss Axio Observer.z1 fluorescence microscope (Carl Zeiss Microscopy) with an ApoTome2 attachment. To count ELM gaps, ApoTome2 stacks encompassing the full thickness of the stained section and covering the entire retina were acquired and processed to yield a maximum intensity projection using Zen 2.6 software (Carl Zeiss). The resulting images were viewed in Fiji and gaps counted manually. The ELM was manually fit to a spline curve, avoiding dysplastic regions identified in the DAPI channel. The gap count was divided by the total ELM length calculated from the fitted curves. To assess the distribution of mitotic nuclei at P0, images were acquired with a 10× objective using the Zeiss microscope and ApoTome, and analyzed in Fiji to compare the intensity of PHH3 fluorescence at the apical NBL to that in the full NBL as described previously [115].

ERG analysis

ERG analysis was carried out using a Espion V6 ColorDome Multifocal system (Diagnosys) essentially as described [116]. Mice at four, eight, and 12 months of age were dark-adapted overnight and pupils were dilated using 1% atropine sulfate (Akorn) or 1% cyclopentolate (Akorn). Mice were anesthetized using ketamine (Covetrus) and xylazine (Akorn) diluted in normal saline and administered by intraperitoneal injection at a final dose of 80 and 16 mg/kg of body weight, respectively. Anesthetized mice were placed on the heated platform of the Espion system. Gold loop electrodes were placed on the corneal surface of each eye along with Goniovisc hypromellose ophthalmic solution (HUB Pharmaceuticals), Gonak (Akorn), or Refresh (Allergan). Small needle electrodes beneath the skin at the top of the head and at the tail served as reference and system ground, respectively. Eyes of dark-adapted mice were presented with a six-step protocol of increasing light intensity measuring mainly rod-dependent responses, followed by light adaptation and a three-step protocol measuring cone-dependent responses. The flash luminance (cd s m^-2), number of sweeps, and time between sweeps (s) for the dark-adapted series was 0.0025, 5, 5; 0.006, 5, 5; 0.016, 5, 5; 0.04, 5, 10; 0.1, 5, 15; and 0.25, 5, 15, respectively. Following a 10-min light adaptation at 110 cd/m², the corresponding parameters for the light-adapted series were 0.63, 10, 1; 4, 10, 1; and 10, 20, 1 on a background illumination of 110 cd m^-2. Response amplitudes were determined from the average of multiple sweeps as follows: scotopic a-wave, pre-flash baseline to the negative trough; scotopic b-wave, negative trough to positive peak within 40–120 ms following the flash; photopic b-wave, negative trough to positive peak within 40–60 ms following the flash. The mean response amplitude from both eyes of each mouse was used for statistical analysis; if data were missing from one eye, only one eye was used.

Statistical analysis

Genotype-phenotype associations to identify the causative mutation and to determine the association of strain with pathological features were assessed with Fisher’s Exact test using JMP, Version 15 or 16 (SAS Institute). Quantitative immunoblotting and qRT-PCR data were analyzed by one-way ANOVA and Student’s t-test, respectively, using Prism (GraphPad). Genotype-phenotype association in epistasis crosses were analyzed by linear regression modeling. Heterozygous and homozygous genotypes were independently coded to assess allelic dominance in both main and interaction effects. Significance was estimated by the ratio of F-statistics between interactive and additive models, with p-values derived from the Fisher-Snedecor distribution. ERG results were analyzed in Prism using a mixed effect model to accommodate missing data and to allow the use of repeated measures; further analysis of normalized data at the highest ERG flash intensities was performed in Prism using two-way ANOVA. A significance threshold of p < 0.05 was used for all experiments.