Optic Atrophy and Inner Retinal Thinning in CACNA1F-Related Congenital Stationary Night Blindness

Hemizygous pathogenic variants in CACNA1F lead to defective signal transmission from retinal photoreceptors to bipolar cells and cause incomplete congenital stationary night blindness in humans. Although the primary defect is at the terminal end of first-order neurons (photoreceptors), there is limited knowledge of higher-order neuronal changes (inner retinal) in this disorder. This study aimed to investigate inner retinal changes in CACNA1F-retinopathy by analyzing macular ganglion cell layer-inner plexiform layer (GCL-IPL) thickness and optic disc pallor in 22 subjects with molecularly confirmed CACNA1F-retinopathy. Detailed ocular phenotypic data including distance and color vision, refraction and electroretinogram (ERG) were collected. Distance vision was universally reduced (mean: 0.42 LogMAR), six had abnormal color vision and myopia was common (n = 15; mean: −6.32 diopters). Mean GCL-IPL thickness was significantly lower in patients (55.00 µm) compared to age-matched controls (n = 87; 84.57 µm; p << 0.001). The GCL-IPL thickness correlated with scotopic standard (p = 0.04) and bright-flash (p = 0.014) ERG b/a ratios and photopic b-wave amplitudes (p = 0.05). Twenty-one patients had some degree of disc pallor (bilateral in 19). Fifteen putative disease-causing, including five novel variants were identified. This study establishes macular inner retinal thinning and optic atrophy as characteristic features of CACNA1F-retinopathy, which are independent of myopia and could impact potential future treatment strategies.


Introduction
Congenital stationary night blindness (CSNB) is a group of genetically and phenotypically heterogeneous retinal disorders that follow autosomal dominant, autosomal recessive or X-linked patterns of inheritance and can be broadly categorized into those with a largely normal, or an abnormal, fundus appearance [1]. CSNB with a largely normal fundus can be ≤ −0.5 and > −6.00D (mean SE − 3.22D) and (3) high myopia [37] group: 17 eyes with SE ≤ −6.00D (mean SE −8.38D). In addition, peripapillary RNFL thickness parameters derived from optic disc cube scans with good automatic segmentation were available in 6 of the 22 CACNA1F patients. A representative RNFL thickness analysis is shown in Figure 1C,D.  retinal nerve fiber layer (RNFL) thickness was determined using the Cirrus inbuilt analysis tool; (C) RNFL thickness map and (D) RNFL thickness graph (case 14; average RNFL thickness 61 µm; temporal RNFL thickness 44 µm). (E) Control ERG: The electroretinogram a-wave was measured from the prestimulus baseline to the trough, and the b-wave amplitude from the a-wave trough to the subsequent peak. (F) Electronegative ERG (case 9): the b-wave amplitude is smaller than the a-wave amplitude (b/a ratio < 1). (G-J) Color fundus photos were graded by a pediatric neuroophthalmologist as exhibiting (G) no disc pallor (case 10); (H) two clock hours of disc pallor (case 12); (I) three clock hours disc pallor (case 22) and (J) five clock hours of optic disc pallor (case 16).
Full-field ERG data performed in accordance with the accepted International Society for Clinical Electrophysiology of Vision Standards were available in the study eye of 22 patients [38,39]. The amplitude of the a-wave (photoreceptor origin) was measured from the prestimulus baseline to the negative-peak or "trough", and that of the b-wave (bipolar cell origin) was measured from the a-wave trough to the subsequent positive-peak ( Figure 1E). The ratio of the b-wave to a-wave amplitude (b/a ratio) was calculated for the DA standard flash (3.0 or 2.29 cd·s·m −2 ), DA bright flash (10.0 or 7.6 cd·s·m −2 ) and LA standard flash (3.0 or 2.29 cd·s·m −2 ). In the earlier years, the lab used 2.29 and 7.6 cd·s·m −2 stimuli for standard and bright flash ERG but later switched to 3.0 and 10.0 cd·s·m −2 , respectively [38,39]. A reduced b/a ratio denotes generalized bipolar cell dysfunction and a b-wave amplitude smaller than the a-wave amplitude (b/a ratio < 1) is termed an electronegative ERG ( Figure 1F).
True-color optic disc photos, available for 21 patients, were qualitatively evaluated by an experienced pediatric neuro-ophthalmologist noting the presence or absence of pallor of the disc substance as reflected in the neuroretinal rim. The extent of pallor was documented according to the number of clock-hours of the disc, which appeared pale, as was the presence or absence of a peripapillary crescent [40].
Data from one eye per patient was included in the study analysis. OCT and all ophthalmic data were collected from the same eye; the right eye was selected (n = 17) when the macular OCT was of sufficient quality and the left eye was included otherwise (n = 5).
Statistical analyses were performed using R version 3.6.2 (R Foundation for Statistical Computing, Vienna, Austria). Analysis of variance (ANOVA) was used to compare the differences among group means with post-hoc Tukey test to determine statistical significance. A two-sample t-test was used to compare subject age between patients and controls, and to compare GCL-IPL thickness between the groups of patients with, and without, RNFL measurements. Linear regression was used to evaluate the effect of SE on GCL-IPL thickness and on ERG parameters, the interaction between these parameters in the presence of SE was examined using two-way ANOVA. A p value of ≤0.05 was taken to be statistically significant.

Patient Demographics and Clinical Features
Twenty-two males with molecularly confirmed hemizygous mutations in CACNA1F were included in the study. The average age of the patient cohort was 14.3 years (range 6-58 years). Nystagmus occurred in 50% (7 of 14) of patients for whom its presence or absence was documented. Similarly, nyctalopia and photophobia were reported in 50% (8 of 16) and 44% (7 of 16) of documented cases, respectively. The BCVA was reduced in all patients (mean 0.42 logMAR, median 0.40 and range 0.10-0.80 logMAR). The mean SE refractive error was −6.32D (median −7.38 and range −20.50 to + 2.50D), with myopia being common (68%; 15 of 22). Color vision was normal in the majority (71%; 15 of 21), whereas six patients showed red-green color defects, which were classified as mild (n = 4), medium (n = 1) or strong (n = 1) on HRR. Only 23% (5 of 22) CACNA1F-subjects had normal contrast sensitivity; the mean contrast sensitivity was 1.36 log units (range 0.90-1.80 log units) ( Table 1). Two of the three patients who had MRI to evaluate for optic atrophy were found to have unremarkable optic nerves (cases 11 and 14) and one displayed bilateral, mild changes compatible with optic nerve atrophy (case 16).
For every diopter decrease in spherical equivalent there was a decrease in GCL-IPL thickness of 0.81 µm, (R 2 adjusted = 0.86, p << 0.01), which was not significantly different between CACNA1F-iCSNB and control subgroups ( Figure 2B). After correcting for refractive error, however, macular GCL-IPL thickness in the CSNB group was on average thinner by 26.37 µm compared to the control group (F (2.106) = 326, p << 0.001). Genes 2021, 12, x FOR PEER REVIEW 8 of 18  As the majority of the study (21/22) and control (85/87) subjects were under 25 years of age, we excluded the three older subjects from any analysis including age as a variable. After the exclusion of one CACNA1F-subject with age > 25 years (case 6, 58 years) and 2 control subjects (aged 34 and 53 years), age was negatively correlated with GCL-IPL thickness in both CSNB and normal controls (R 2 adjusted = 0.79, p = 0.03) as previously reported [47][48][49][50]; however, the rate of thinning was not significantly different between the groups (p = 0.15).
Six patients in the CACNA1F cohort (mean age 12.5, range: 6-17 years) had good quality optic disc cube scans, and RNFL thickness was determined using the Cirrus inbuilt analysis tool ( Table 2). Mean (SD) average peripapillary RNFL thickness was 68.67 (4.72) µm and mean (SD) temporal RNFL thickness was 51.33 (13.81) µm. Both mean values were lower than similar populations reported in literature (Table 3) [51][52][53][54][55][56][57]. The average RNFL values for all six CACNA1F patients were below the lower range reported in normal children, and two of four myopic patients had values outside the lower range reported for normal myopes. The temporal RNFL measures were below the lower range or 5th percentile for pediatric and myopic populations in four of six study subjects. The mean (SD) average macular GCL-IPL thickness for these six iCSNB subjects was 54.50 (3.67) µm, which was similar to the rest of the cohort (55.19 (7.16) µm; p = 0.77). The OCT values for each study eye are shown in Table 2.  [58]); WWP, white-without-pressure. GCL-IPL thickness, RNFL thickness and ERG b/a ratios are shown for the study eye; the right eye was selected when macular optical coherence tomography was of sufficient quality.

Full-Field Electroretinogram Results
The After controlling for SE, the macular GCL-IPL thickness was weakly correlated with the DA standard flash b/a ratio (R 2 adjusted = 0.41, p = 0.04; Figure 2C), DA bright flash b/a ratio (R 2 adjusted = 0.45, p = 0.014; Figure 2D) and LA standard flash b-wave amplitude (R 2 adjusted = 0.41, p = 0.05; Figure 2F). The correlation with GCL-IPL thickness did not reach significance for LA standard flash b/a ratio (R 2 adjusted = 0.42, p = 0.08) or LA standard flash a-wave amplitude (R 2 adjusted = 0.34, p = 0.56; Figure 2E). Individual values for b/a ratio and amplitude measurements are shown in Table 2.

Optic Disc Evaluation and Fundus Findings
Twenty-one patients had color fundus photos available; 19 had some degree of disc pallor in both eyes (measured in observable clock hours) and the remaining 2 had unilateral disc pallor (Table 2). In the study eye, one disc (5%) was graded as having no pallor, three discs (14%) with two clock hours of pallor, seven (33%) with three clock hours, three (14%) with four clock hours and seven (33%) with five clock hours of pallor. Representative disc photos from four subjects are shown in Figure 1G-J. There was no significant correlation between macular GCL-IPL thickness and number of clock hours of disc pallor (R 2 adjusted = 0.03, p = 0.22; Figure 2G). Fifteen subjects showed myopic retinal changes (bilateral fundus tessellation (n = 14), white without pressure (n = 4), posterior staphyloma (n = 2), peripheral degeneration (n = 2) and/or unilateral Forster-Fuchs spot (n = 1)). Four cases displayed normal fundi and five had hypopigmented fundi more typical of Åland island eye disease, a phenotypic variant of CACNA1F iCSNB [20,29,59]; two in the latter group also had myopic changes. Foveal hypoplasia was identified in six cases on fundus photographs or OCT; the grading [58] is shown in Table 2.

Genetic Results
Fifteen putative disease-causing variants were identified in the study (Table 1); these included missense (n = 5, 6 subjects), nonsense (n = 4, 5 subjects), frameshift (n = 3, 7 subjects), splice site (n = 2, 3 subjects) and silent variants predicted to affect splicing (n = 1, 1 subject). Five of the variants are novel, which were p.(Gln561*), p.(Val611Glyfs*32), p.(Gln1157Gln), c.3741+2T>C and p.(Pro1492Ser). The details of the novel variants, includ-ing the predicted pathogenicity, conservation scores and ACMG classification are summarized in Table 4. None have previously been identified in large population databases (gnomAD) [60]. Both p.(Gln561*) and p.(Val611Glyfs*32) are expected to produce truncated protein or the mRNA produced are likely to be targeted for nonsense-mediated decay. The silent variant p.(Gln1157Gln) affects the last nucleotide of exon 28, is predicted to alter the natural splice donor site 1 base pair downstream and is thereby likely to cause exon 28 skipping. The c.3741+2T>C substitution is located on the canonical donor splice site of intron 30, and is expected to cause exon 30 skipping. The p.(Pro1492Ser) variant affects a highly conserved residue and is predicted pathogenic by all in silico tools; further, there are other substitutions at and near position 1492 (p.(Pro1492Ala) and p.(Gly1494Arg)) that have been reported to cause iCSNB [1,42]. The mutation segregated with the disease phenotype in all tested cases (n = 18); among the four cases where segregation analysis was unavailable, two harbored previously reported mutations (case 19: p.(Pro1492Ala) [1,42] and case 21: p.(Arg1502*) [42]) and two harbored novel disease-causing variants (case 9: p.(Val611Glyfs*32) and case 17: p.(Gln1157Gln)). The macular GCL-IPL thickness did not differ regardless of the type of mutation ((F (3,18) = 0.68, p = 0.57; Figure 2H)).

Discussion
This is the first comprehensive study to have reported marked macular GCL-IPL thinning in subjects with molecularly confirmed CACNA1F-related iCSNB. All subjects in the current cohort demonstrated at least some degree of optic disc pallor in at least one eye. The majority of patients were myopic and the peripapillary RNFL thickness was reduced below average in all six tested individuals. Previously, only rare instances of optic nerve pallor or atrophy had been reported in CACNA1F-retinopathy [18][19][20] and since many CACNA1F subjects have high myopia [10], this has been proposed to account for the optic disc pallor, often at the temporal disc, in these patients [7,72,73]. The present study identified disc pallor together with thinner GCL-IPL and RNFL compared to a myopic control cohort and values reported in the literature suggesting that both findings are unrelated to myopia, and the optic disc pallor in CACNA1F-patients is due to atrophy. Our findings also indicate that inner retinal and optic disc changes are far more common than originally thought in this condition, traditionally described as having a normal fundus.
The incidence of nystagmus (50%), nyctalopia (50%) and photophobia (44%) in our cohort was similar to those identified in two large studies of CACNA1F-related iCSNB, which documented rates of 44-65%, 58-60% and 50%, respectively [10,73]. The BCVA was reduced in all study subjects. The mean BCVA in this study (0.42 LogMAR) was similar to that reported by Allen et al. (0.40 LogMAR) [74] and the range of BCVA in our cohort (0.10-0.80) was similar to two previous studies, which had reported 0.10-1.00 and 0.10-1.30 LogMAR, respectively [10,73]. Myopic refractive error was common in the current cohort (73%), as reported in other series in literature (81%-85%) [10,73]. Only red-green color defects were observed in the current study (29%, HRR testing); a previous group reported color vision defects in 46% (21/46, using one of D15, Ishihara or HRR) of their cohort [10]. Only a small proportion of patients in either study had strong color vision defects (5% in the current study vs. 13% in Bijveld et al.) [10], however, a different study reported strong color defects in all tested family members (n = 6; D15 and Ishihara) [75].
The average macular GCL-IPL thickness was severely reduced in our CACNA1F cohort compared to controls with high myopia (p << 0.001). Further, most tested subjects had thinner average RNFL compared to myopic [55][56][57] or pediatric control cohorts [51][52][53][54]. A previous study suggested thinner GCL-IPL and normal RNFL thickness in three cases of CACNA1F-related iCSNB, although thickness measurements were not reported [21]. Further, selective inner retinal thinning, as evidenced by reduced average GCL-IPL thickness (range 59-65 µm), was reported in three patients with GRM6-related cCSNB; the authors hypothesized reduced bipolar and/or ganglion cell numbers or altered inner retinal synaptic structure to be the cause [76]. It is notable that average GCL-IPL thicknesses in our study are similar (55 (6.17) µm; range: 46-70 µm) to those reported in GRM6 cCSNB patients. Additionally, of note are various mouse models for iCSNB (Cacna1f, Cabp4 and Cacna2d4) primarily demonstrating abnormal synapses and thinning in the outer plexiform layer with no apparent changes in the inner retina including the GCL and IPL [77][78][79]. This may suggest that inner retinal changes are perhaps unique to human CACNA1F-phenotype in comparison to known animal models of the disease. One of the patients in the present cohort was additionally observed to have abnormal synapses in the outer retina [30].
The DA standard and bright-flash ERGs were electronegative in the majority of subjects in the current study with mean b/a ratios of 0.59 and 0.56, respectively, indicative of severe generalized rod ON-bipolar cell dysfunction. An electronegative ERG (or reduced b/a ratio) to DA standard flash or higher intensities is a characteristic feature of CACNA1F (median b/a ratio: 0.70) [10] and other forms of CSNB [1,10,73,74]. The mean LA standard flash ERG b/a ratio was 1.18 in our cohort; a similar average b/a ratio of 1.35 in the cohort reported by Bradshaw et al. was significantly reduced compared to controls (2.31) [80], indicative of generalized cone ON-and OFF-bipolar cell dysfunction. In the present study, GCL-IPL thickness showed weak correlation with DA standard flash and bright flash ERG b/a ratios, and LA standard flash b-wave amplitudes, with thinner GCL-IPL corresponding to lower b/a ratios or b-wave amplitudes, respectively. These results are novel and might indicate some superseding inner retinal (bipolar cell) dysfunction in addition to the signal transmission defect at the terminal end of photoreceptors in CACNA1F disorder. Future studies, however, are needed to further validate this observation.
All subjects in the current cohort had at least 2 clock hours of optic disc pallor in at least one eye and most had bilateral disc pallor. To date, a few cases in the literature have documented optic disc pallor or atrophy in CACNA1F-related iCSNB [18][19][20] and these reports are limited to the clinical appearance, rather than the graded measurement of optic disc pallor or ganglion cell structural integrity. All seven reported cases of iCSNB associated with RIMS2, a regulator of synaptic membrane exocytosis localized to rod photoreceptors and the outer plexiform layer, from 3 unrelated, ethnically diverse families demonstrated clinically appreciable optic disc pallor [15]. Inner retinal thinning was seen in the three subjects for whom OCT was available, and RNFL thinning was shown in two of them [15]. Whilst the authors advise optic disc and inner retinal changes should be interpreted with caution in the context of myopia, it is noted that none of the cases had high myopia (range: +6.00 to −4.50 D; only three were myopic) [15]. In 1983 Heckenlively et al. reported optic disc anomalies (interpreted as representing atrophy, dysplasia, or both) in subjects with CSNB; five had tilted discs with a lack of visible temporal disc tissue, two had dysplastic nerves and three had disc pallor without tilt [81]. Further, we showed that inner retinal thinning in CACNA1F-related iCSNB patients is in excess of controls with the same degree of myopia, strongly suggesting the possibility that optic pallor is consistent with optic atrophy, which is likely overlooked in CACNA1F-related iCSNB as a whole. Optic disc pallor has not been described in cases of iCSNB due to CABP4 or CACNA2D4 mutations [13,14]. The clock hours of appreciable disc pallor did not correlate with GCL-IPL thickness. This could in part be attributed to clinical difficulty in fundoscopy in identifying optic atrophy in the presence of significant myopia because of a prominent white scleral crescent and very little visible temporal disc tissue in the presence of a very obliquely exiting optic nerve (tilted disc) in many patients [82][83][84]. OCT measures of RNFL atrophy and decreased GCL-IPL thickness, on the other hand, are objective markers of axonal atrophy and support the presence of nerve atrophy within the disc in this study cohort.
CACNA1F encodes the primary subunit (α 1f ) of the L-type voltage-gated calcium channel, a 1977 amino acid protein with four homologous transmembrane domains flanked by intracellular components (N and C termini) [85]. Fifteen putative disease-causing variants including missense (33%), nonsense (27%), frameshift (20%) and splice-site (20%) were identified in our cohort. A previous study with 26 pathogenic mutations reported a higher proportion of nonsense and frameshift variants (27% each) followed by missense variants (23%). Amongst the five novel mutations, two each were nonsense (p.(Gln561*) and p.(Val611Glyfs*32)) and splice-site ((c.3741+2T>C) and p.(Gln1157Gln)) variants. Both Gln561 and Val611 are in the second transmembrane domain and a premature stop codon will lead to severely truncated protein missing the 3rd and 4th transmembrane domains and the C termini. Further, the mRNA produced is likely removed by nonsense mediated decay. The synonymous variant (c.4474C>T; p.(Gln1157Gln)) affects the last nucleotide in exon 28 and is predicted to alter the splice donor site. Recently, two other exonic synonymous variants in CACNA1F (c.646C>T; p.(Leu216Leu) and c.1719G>A; p.(Thr573Thr)) were reported to affect splicing using a minigene approach [42]. In addition, there are reported instances of pathogenic synonymous exonic variants affecting the last nucleotide of an exon leading to a splice site defect in other disorders [86,87]. Although the average macular GCL-IPL thickness in the CACNA1F-subjects was reduced regardless of the mutation class, there appeared to be variability within each mutation class ( Figure 2F). This is perhaps not surprising as others have shown considerable intra-and inter-familial variability in phenotypic features (including refractive error, visual acuity and dark adaptation thresholds) in subjects harboring the same CACNA1F mutation [10,73].
The Cacna1f gene is well expressed in wildtype mouse and rat retina; whilst most studies confirm its expression at the photoreceptor terminal in the outer plexiform layer [45,[88][89][90][91][92][93], some report expression in the inner part of inner nuclear layer and GCL [20,45,90,91]. Further, there is electrophysiological evidence from Cacna1f -deficient mouse lines to suggest that this subunit contributes to calcium influx in the retinal bipolar cells (Qi Lu PhD dissertation, Wayne State University-referenced in [85]). Hence the findings of GCL-IPL thinning and optic disc pallor observed in the current study may either be due to defective CACNA1F function in the inner retinal layers or perhaps even due to transsynaptic changes hypothesized by some in CSNB [76,81]. In either case, our findings are important in the context of novel approaches under development to treat CACNA1F-retinopathy [94], as it is likely that inner retinal changes contribute to the visual deficit in the disorder in humans and as such, treatment targeted at a defect in the photoreceptor synapse alone may not be sufficient to restore vision.
To summarize, this study identified inner retinal thinning and optic atrophy as characteristic features of iCSNB due to hemizygous pathogenic mutations in CACNA1F. The inner retinal thinning was independent of myopia and mutation type; although there was variability in the presence of myopia in our cohort, GCL-IPL thickness was uniformly reduced, and thinner than a myopic control population. The macular OCT was useful to objectively differentiate optic atrophy from disc pallor alone, such as that associated with high myopia. A prospective natural history study would help identify whether these inner retinal changes are stationary or progressive in nature.

Conflicts of Interest:
The authors declare no conflict of interest.