Peripheral blood was obtained from the patient, a 9-year-old girl with COD, in September 2020 at Tianjin Eye Hospital (Tianjin, China). During the same hospital visit, peripheral blood samples were also collected from the parents, who had not been diagnosed with any associated eye disease. The patient underwent eye examinations, including fundus photography, optical coherence tomography (OCT) and electroretinography (ERG). Follow-up was performed once yearly.

High-throughput sequencing of the samples was performed by Novogene Co., Ltd. DNA libraries were prepared for sequencing using the NEBNext^® Ultra™ DNA Library Prep Kit (E7370L; New England Biolabs, Inc.). The quality of the samples was verified by the 5400 Fragment Analyzer System (Agilent Technologies Inc.). Exome sequencing was performed using an NovaSeq 6000 Sequencing System (Illumina, Inc.) using paired-end 150-bp reads. The NovaSeq 6000 S4 Reagent Kit v1.5 (Illumina, Inc.) was used for sequencing. Library concentrations were assessed using the CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories Co., Ltd.) and an Invitrogen Qubit 4 Fluorometer (Thermo Fisher Scientific, Inc.).

The data files obtained from high-throughput sequencing were analyzed by Base Calling and converted into Sequenced Reads, and then sequencing data were aligned with the human reference genome (Genome Reference Consortium Human Build 37) using BWA (version 0.7.8-r455; https://github.com/). Variant calling was performed using SAMtools (version 1.21) (6) and the resulting VCF files were annotated using ANNOVAR (version 2022-01-13) (7).

Pathogenic variants were screened based on the following criteria: i) Variants with an allele frequency of #x003C;1% in the 1000 Genomes (8) Exome Sequencing Project 6500 (https://www.ebi.ac.uk/ena), Genome Aggregation databases (9) and ChinaMap (http://www.mbiobank.com/) were retained; ii) only variants located in the coding (exonic) or in the splice site (±10 bp) regions were retained; iii) nonsynonymous mutations in highly conserved regions (grep++ score >2) and those affecting splicing (dbscSNV score >0.6) were retained, whereas small (#x003C;10 bp) fragments of non-displaced indel mutations located in the repeat region were excluded; iv) variants predicted to be deleterious using two of the following four functional annotation algorithms were retained: Sorting Intolerant from Tolerant (SIFT) (10), Polymorphism Phenotyping v2 (PolyPhen-2) high-diversity model (11), MutationTaster (12) and Combined Annotation Dependent Depletion (13); and v) alignment with autosomal recessive or compound heterozygous inheritance patterns.

Genomic DNA was extracted from peripheral blood samples using the Magnetic Blood DNA Extraction Kit (Vazyme Biotech Co., Ltd.) and primers were designed using Primer3 (https://primer3.org/). The primer sequences are listed in Table SI.

To analyze the mutant allele c.566_567insT:p.R189fs, PCR was performed to amplify the target locus. The thermal cycling protocol consisted of initial denaturation at 95˚C for 5 min, followed by 35 cycles of denaturation at 95˚C for 30 sec, annealing at 57˚C for 4 min and extension at 72˚C for 4 min. The resulting PCR products were cloned into the pEASY-Blunt Zero Cloning vector and transformed into Trans1-T1 receptor cells using the pEASY-Blunt Zero Cloning Kit (TransGen Biotech Co., Ltd.) following the manufacturer's instructions. Sanger sequencing was performed on individual clones using M13 forward and reverse primers to confirm the presence of the mutation in the patient samples.

The pcDNA3.1 plasmid containing the CNGA3 cDNA sequence (Table SII) with a C-terminal FLAG tag (Table SIII) was synthesized by Beijing Tsingke Biotech Co., Ltd. The CNGA3 sequence in the National Center for Biotechnology Information database (accession number, NM_001298.3) was used. This matched the sequences of the alleles from the father and mother of the proband, confirming they carried wild-type alleles. The CNGA3 wild-type plasmid was designated pcDNA3.1-CNGA3 (Fig. S1).

To introduce a p.S334F mutation, the Mut Express II Fast Mutagenesis Kit V2 (Vazyme Biotech Co., Ltd.) was used. The pcDNA3.1-CNGA3-p.S334F (c.C1001T) mutant plasmid was generated using pcDNA3.1-CNGA3 as a template. Mutation-specific primers were designed as follows: Forward primer (p.S334F-F): 5'-GACAGACTtCTGGGTCTACCCAAACATCTCAA-3' (plasmid positions, 1,984-2,015), which corresponds to bases 993-1,024 in the CNGA3 coding sequence (CDS), with the lower case ‘t’ indicating the mutation site between bases 1,000 and 1,002; and reverse primer (p.S334F-R): 5'-AGACCCAGaAGTCTGTCCCAAAACCAATGAACT-3' (plasmid positions, 1,968-2,000), which corresponds to the reverse complementary sequence to bases 977-1,009 in the CNGA3 CDS, with lower case ‘a’ representing the mutation site. The amplified product was treated with DpnI to remove the methylated template plasmid, and ExNase II was used to cyclize the linear DNA. Subsequently, the resulting pcDNA3.1-CNGA3-p.S334F plasmid was transformed into E.coli DH5α competent cells (Takara Bio Inc.), inoculated into Luria-Bertani (LB) solid medium (Thermo Fisher Scientific Inc.) supplemented with ampicillin (Amp+) (Beyotime Biotechnology) and subjected to 12 h of inverted culture. Single clones were selected and inoculated into LB liquid medium (Amp+) at 37˚C for amplification. Plasmid DNA was then extracted from the bacteria using an EndoFree Plasmid Midi Kit (Jiangsu CoWin Biotech Co., Ltd.), and the experimental steps were as described by the manufacturer.

A pcDNA3.1-CNGA3-p.R189fs mutant plasmid was also generated. The mutation-specific primers were as follows: Forward primer (p.R189fs-F): 5'-ATTTGCAGtGGCCTGTTTCGATGAGCTGCAGT-3' (plasmid positions, 1,550-1,580) which corresponds to bases 559-589 in the CNGA3 CDS, with lower case ‘t’ indicating the mutation site between 566 and 567; and reverse primer (p.R189fs-R): 5'-AACAGGCCaCTGCAAATAAGCAGATACCAGTTATAGA-3' (plasmid positions, 1,530-1,565), which corresponds to the reverse complementary sequence to bases 539-574, with lower case ‘a’ indicating the mutation site. The plasmid construction process was analogous to that described for pcDNA3.1-CNGA3-p.S334F. All constructed plasmids were verified by sequencing.

293T cells (Guangzhou Ubigene Biosciences Co., Ltd.) were cultured at 37˚C in a 5% CO₂ atmosphere in Dulbecco's modified Eagle's medium (Thermo Fisher Scientific Inc.), supplemented with 10% fetal bovine serum (Thermo Fisher Scientific Inc.) and 1% penicillin-streptomycin. When cell confluence reached 70-80%, the plasmids were mixed with polyethyleneimine (PEI) to form a DNA-PEI complex. DNA (2 µg) was added to the cells and incubated at 37˚C for 48 h prior to the execution of subsequent experiments.

After rinsing with PBS, total protein was extracted from the cells by treatment with RIPA buffer (Beijing Solarbio Science & Technology Co., Ltd.; cat. no. R0020) and PMSF (cat. no. P0100). The total protein concentration was then determined using the BCA method with BSA as the standard. A total of 20 µg total protein/lane was separated using SDS-PAGE (separation gel 10%, concentrating gel 5%) and transferred onto a polyvinylidene fluoride (PVDF) membrane. The membrane was blocked with 5% skimmed milk powder for 1 h at 37˚C to prevent nonspecific binding. FLAG-tagged CNGA3 protein was then detected using a FLAG antibody (product no. F3165-1MG; 1:5,000; Sigma-Aldrich; Merck KGaA) and incubated at 4˚C for 14 h. GAPDH served as a loading control and was detected using a GAPDH antibody (cat. no. 2118S; 1:5,000; Cell Signaling Technology) and incubated at 4˚C for 14 h. The blot was then incubated with a horseradish peroxidase-conjugated secondary antibody (cat. no. SE134; 1:5,000; Beijing Solarbio Science & Technology Co., Ltd.) at room temperature for 1 h. Subsequently, a luminescent solution was applied to the PVDF membrane. NcmECL Ultra Luminol/Enhancer Reagent (A) was mixed with NcmECL Ultra Stabilized Peroxide Reagent (B) in equal proportions to make a luminescent solution (New Cell & Molecular Biotech Co., Ltd.). Images were captured using NiceAlliance Q9 software (Uvitec Ltd.). Experiments were repeated three times.

Protein distribution data were retrieved from the Human Protein Atlas (HPA; https://www.proteinatlas.org/). CNGA3 RNA levels in the brain were retrieved from the HPA Brain Resource (https://www.proteinatlas.org/ENSG00000144191-CNGA3/brain), which integrates and normalizes RNAseq data from the Genotype-Tissue Expression (GTEx) project for the brain and retina; the GTEx project collects and analyzes human postmortem tissues, with RNA-sequencing performed on 36 tissue types using RSEMv1.3.0 (v8) (https://deweylab.github.io/RSEM/). The GTEx retina data are based on EyeGEx data reported by Ratnapriya et al (14), with transcript abundance estimated using Kallisto v0.48.0 (https://pachterlab.github.io/kallisto/) with Ensembl version 109 as the reference genome. The RNA tissue expression results for CNGA3 were obtained by searching the consensus dataset in the HPA Tissue Resource for ‘CNGA3’ (https://www.proteinatlas.org/ENSG00000144191-CNGA3/tissue). The consensus dataset in the HPA integrates transcriptomics data from HPA and GTEx datasets for 55 tissue types through an internal normalization pipeline. CNGA3 single-cell expression results were obtained by searching the Single Cell Type and Single Cell Resource sections of the HPA for ‘CNGA3’ (https://www.proteinatlas.org/ENSG00000144191-CNGA3/single+cell). This analysis integrates single-cell information for 31 human tissues from various databases, including the Single Cell Expression Atlas, Human Cell Atlas, Gene Expression Omnibus, Tabula Sapiens, Allen Brain Atlas and European Genome-phenome Archive. Transcript per million (TPM) values were rescaled to a sum of one million TPM following the removal of non-coding transcripts. Within each data source, the TPM values of all samples were normalized separately using the trimmed mean of the M values to allow between-sample comparisons. The final normalized transcript expression values, denoted as normalized parts per million (nTPM) markers, were calculated for each gene in each sample, with nTPM values #x003C;0.1 excluded from visualization in the HPA. Prediction of protein stability after CNGA3 mutation was performed using the DEZYME tool (https://soft.dezyme.com/query/create/pop), with ΔΔG #x003C;0 meaning a stabilizing mutation.

Protein expression results in the in vitro experiment are presented as the mean ± SD. The difference between wild-type and CNGA3:p.S334F groups was analyzed using an unpaired Student's t-test. P#x003C;0.05 was considered to indicate a statistically significant difference. In order to assess whether the distribution of CNGA3 mutation sites in the structural and non-structural domains was statistically significant, the distributional characteristics of the mutation sites were systematically analyzed in this study using a one-sample hypothesis test. The null hypothesis (H0) of the test posited that there was no statistically significant difference in the distribution of mutation sites between the structural and non-structural domains. By contrast, the alternative hypothesis (H1) predicted that the proportion of mutation sites in the structural domains would be significantly higher than that in the non-structural domains. The significance level of the test was set at α=0.05. The statistical analysis was conducted using GraphPad Prism (version 10.4) and R software (version 4.1.2).

The patient was a 9-year-old girl with COD whose parents exhibited no signs of any associated eye disease. The child was first diagnosed with COD in September 2020. She had a visual acuity of 20/100 in both eyes and mild nystagmus. Fundus photography revealed a loss of foveal reflex in both eyes (Fig. 1A and B). OCT revealed low ganglion cell thickness and a retinal nerve fiber layer thinner than normal (Fig. 1C-F) (15). ERG demonstrated decreased dark-adapted b-wave and a-wave amplitudes and approximately normal oscillatory potential amplitudes in both eyes (Fig. 2). The patient was prescribed glasses for full correction of hyperopia and astigmatism. However, her visual acuity improved by no more than one line during follow-up. Upon re-examination on December 5, 2024, her visual acuity was unchanged at 20/100 in each eye. Her electroretinogram exhibited a slight improvement, although it remained markedly abnormal (Fig. S2).

Whole-exome sequencing of the proband and her parents (Fig. 3A) was performed (16). Genetic analysis revealed compound heterozygous CNGA3 mutations in the proband, including a known missense single-nucleotide variant (CNGA3:c.C1001T) and a previously unreported frameshift mutation (CNGA3:c.566_567insT) (Table I). The mother was a carrier of the c.C1001T mutation, which results in substitution of serine with phenylalanine at codon 334 (p.S334F). The novel c.566_567insT frameshift mutation was inherited from her father, who was a carrier. Fig. S3 presents the nucleic acid sequence following the CNGA3:c.566_567insT:p.R189fs mutation, highlighting the insertion of base T between positions 566 and 567 in the cDNA sequence. This mutation leads to a TGA premature stop codon at position 194, thereby reducing the number of amino acids in the reading frame from 694 to 193, resulting in a shorter, less stable protein. These mutations were validated using Sanger sequencing, which confirmed the CNGA3:c.C1001T:p.S334F mutation in the patient and her mother (Fig. 3B) and the CNGA3:c.566_567insT:p.R189fs mutation in the patient and her father (Fig. 3C). Clonal sequencing of CNGA3:c.566_567insT:p.R189fs yielded results consistent with this mutation (Fig. 3D). Bioinformatics analysis reveals that CNGA3 is highly expressed in the brain (Fig. S4A), retina (UniProtKB, Q16281; Fig. S4B), neuronal protrusions, axons and cone photoreceptor cells (Fig. S4C).

The p.S334F and p.R189fs mutations are located on the ion-transport (ion-trans) structural domain of CNGA3 protein. Notably, p.R189fs results in a truncated protein product that only retains the ion-trans domain and lacks the cysteine-rich CAP domain-extended domain (CAP-ED) and cyclic nucleotide-gated ligand-binding zinc finger-like (CLZ) domain (Fig. 3E). The conservativeness of the two mutant loci, p. R189fs and p.S334F, was evaluated, and the mutations were found to be located in regions that are highly conserved across species (Fig. 3F). Mutations in highly conserved protein regions may impact their functionality (17). The p.S334F mutation affects a conserved residue in the ion-trans domain and is predicted to be harmful based on SIFT and PolyPhen analyses (SIFT 0.003, damaging; Polyphen2_HVAR 0.863, possibly damaging; PolyPhen2_HDIV 0.965, probably damaging). The protein structure and stability were then predicted to gain insights into the impact of the mutation on the protein. The p.S334F mutation was predicted to decrease protein stability, with a ΔΔG value of 0.74 kcal/mol. A schematic showing the location of the CNGA3 mutation in the molecular structure is presented in Fig. 3G. Of the total 151 CNGA3 mutations reported in humans, 40 are associated with COD (Table II). Further hypothesis testing of the relationship between the mutations and structural domains of CNGA3 proteins yielded P#x003C;0.0011 (α=0.05), indicating the enrichment of mutations in structural domains (Fig. 4).

To further investigate the effects of these mutations on protein expression, wild-type and mutant CNGA3 proteins were expressed in 293T cells and analyzed by western blotting (Fig. S5). The FLAG-tagged wild-type CNGA3 protein appeared as two distinct bands, specifically, a main band at 81.7 kDa and an additional band of higher molecular weight (~100 kDa), likely representing a post-translationally modified protein. The S334F mutation resulted in significantly increased CNGA3 protein expression compared with that of wild-type CNGA3, but notably lacked the ~100 kDa band (Fig. 5). Given that S334 is a potential phosphorylation site, we hypothesize that the S334F mutation might prevent phosphorylation at this position. By contrast, the c.566_567insT:p.R189fs mutation produced a truncated protein with a molecular weight of 16.0 kDa, consistent with the predicted premature termination of protein synthesis.