2.1 Patient cohort, clinical examination, and ethical considerations

Patients with a clinical diagnosis of ACHM were recruited over a period of 25 years (1996–2020) in various ophthalmic centers predominantly located in Europe (Austria, Belgium, Czech Republic, Denmark, France, Germany, Greece, Hungary, Italy, the Netherlands, Spain, Sweden, and United Kingdom), but also in Brazil, Israel, Palestine, Turkey, and in the USA. In total, our entire Tübingen ACHM cohort of genetically confirmed ACHM patients includes 1060 affected individuals diagnosed with ACHM originating from 889 independent families.

Clinical diagnosis was established by standard clinical ophthalmologic examination and depended on the local clinical protocol (i.e., color vision testing, best-corrected visual acuity, electroretinography [ERG], fundus photography, and optical coherence tomography).

The study was conducted in accordance with the principles of the Declaration of Helsinki. Blood, saliva, or DNA samples of the ACHM patients and family members were collected over a period of 25 years after informed consent approved by the respective local research and ethical boards and/or dependent on the local regulatory bodies and regulations. Specifically, the study was approved by the Ethics Board of the University of Tübingen under the study no. 349/2003V and 116/2015BO2.

2.2 Variant screening

Peripheral blood, saliva or DNA samples from patients diagnosed with ACHM and family members were sent to the Institute for Ophthalmic Research, Tübingen (Germany) for genetic testing. Extraction of whole genomic DNA was performed according to standard procedures. Molecular genetic screening of the CNGA3 gene was executed as previously described (Kohl et al., 1998; Wissinger et al., 2001). Primers used for PCR amplification and Sanger sequencing are provided in Table S1. If two potential disease-causing CNGA3 alleles were identified in cases that were analyzed within a research setting, no further genetic analysis in other ACHM genes was performed. In case DNA samples from additional family members were available, segregation analysis was performed to validate the presence and the independent segregation of two mutant CNGA3 alleles using either restriction fragment length polymorphism analysis or Sanger sequencing.

2.3 Variant nomenclature and establishment of the CNGA3 variant database

Variant nomenclature of the identified CNGA3 sequence alterations was based on the NCBI reference sequence NM_001298.2 and the Ensembl transcript reference sequence ENST00000272602.7 (GRCh38/hg38 genome assembly) comprising seven coding exons. The first nucleotide of the translation initiation codon ATG is denoted as nucleotide 1. Moreover, all variants located in the alternatively spliced exon 3b (formerly named exon 2b; Wissinger et al., 2001; now renamed exon 3b) and its flanking intronic sequences were denoted based on NCBI reference sequence XM_011510554.2 (GRCh38/hg38). Validation of the variant nomenclature was performed using the web-based Mutalyzer name checker tool (https://mutalyzer.nl/name-checker).

To establish a comprehensive CNGA3 variant database, a literature search was conducted on the free search engine PubMed® (https://pubmed.ncbi.nlm.nih.gov/) (last queried on January 11, 2021). In addition, web-based databases including the Human Gene Mutation Database (HGMD®) (http://www.hgmd.cf.ac.uk/ac/index.php) (Stenson et al., 2017), the Leiden Open Variation Database (LOVD) for CNGA3 (https://www.lovd.nl/CNGA3), and the public archive of human genetic variants ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/) (Landrum et al., 2014) were searched for entries regarding CNGA3 disease-associated variants (searches conducted on January 11, 2021 and March 18, 2021). Variants retrieved from public databases (i.e., LOVD and ClinVar) that have not been associated with a retinal phenotype (i.e., ACHM or allied disease) were not included and not validated. Finally, all CNGA3 variants identified in our ACHM cohort were entered into our in-house CNGA3 variant database. Variants identified in the ACHM cohort genetically characterized in Tübingen that were not yet published in a journal listed on PubMed® are referred to as “novel.” All novel variants were submitted to the ClinVar database with accession numbers from SCV001571262 to SCV001571322. Additionally, these variants will be submitted to the Global Variome shared LOVD database and can be accessed using the following URL: databases.lovd.nl/shared.

2.4 In silico variant assessment

Sequence alterations listed in the CNGA3 variant database were subjected to in silico analysis to evaluate their pathogenic impact. The Combined Annotation Dependent Depletion (CADD) tool was used to score the deleteriousness of both coding and noncoding single-nucleotide variants in the CNGA3 gene (https://cadd.gs.washington.edu/). CADD provides PHRED-scaled scores ranging from 1 to 99. A scaled CADD score of 10 thereby indicates a variant being amongst the top 10% of deleterious variants in the human genome whereas a score of 20 is used to describe variants that are amongst the top 1% of deleterious variants and so on (Kircher et al., 2014). To score the identified variants, we used the CADD model GRCh38-v1.6 and applied a PHRED-scaled score of 20 as a cutoff for deleteriousness as suggested by the authors (Rentzsch et al., 2021). Three in silico prediction algorithms for the evaluation of missense variants were used as embedded in Alamut Genova v.1.4/Alamut Visual v.2.14.0: SIFT (Sorting Intolerant From Tolerant, v4.0.3/v6.2.0) applies a score range of 0.0 for a deleterious effect to 1.0 for the substitution being tolerated (http://sift.bii.a-star.edu.sg/) (Kumar et al., 2009), whereas MutationTaster (v2013) classifies the variants either as “disease-causing” or as a “polymorphism” with the p value indicating the certainty of the prediction (http://www.mutationtaster.org/) (Schwarz et al., 2010). The third prediction algorithm, Align-GVGD (v2007), provides prediction classes defining a spectrum of classifications with C0 suggesting least likely to interfere with protein function to C65 being most likely to affect functionality of the protein (https://agvgd.iarc.fr/index.php) (Tavtigian et al., 2006). In addition, phyloP scores and the Grantham distance were computed as embedded in Alamut Genova v.1.4/Alamut Visual v.2.14.0 to further elucidate the possible pathogenic effect of missense variants. PhyloP determines the evolutionary conservation and acceleration of a given nucleotide. Positive scores are assigned to sites predicted to be conserved whereas negative scores indicate a fast-evolving site (Pollard et al., 2010). The Grantham distance scores missense substitutions regarding the physicochemical difference between the exchanged amino acids with scores ranging from 0 to 215 (Grantham, 1974).

To further assess the conservation of amino acid residues of missense substitutions, we, in addition, performed a protein sequence alignment of Homo sapiens CNGA3 against its orthologues from seven other species, the Homo sapiens CNGA1 protein and its orthologues from additional five species using Clustal Omega (https://www.ebi.ac.uk/Tools/msa/clustalo/) (Sievers et al., 2011) and the alignment visualization tool BoxShade version 3.21 (https://embnet.vital-it.ch/software/BOX_form.html) (Figure S2).

Minor allele frequencies were retrieved from the Genome Aggregation Database (gnomAD) (v2.0.2/v2.1) (https://gnomad.broadinstitute.org/).

In silico evaluation of variants affecting either canonical or noncanonical splice sites was conducted using the splicing algorithms SpliceSiteFinder-like [0-100] (Shapiro & Senapathy, 1987), MaxEntScan [donor sites: 0-12 & acceptor sites: 0-16] (Yeo & Burge, 2004), NNSplice [0-1] (Reese et al., 1997), GeneSplicer [0-15] (Pertea et al., 2001) as embedded in Alamut Genova v.1.4/Alamut Visual v.2.14.0, and SpliceAI [0-1] (Jaganathan et al., 2019) (https://spliceailookup.broadinstitute.org/). Furthermore, variants affecting noncanonical exonic splice site sequences (cutoff: first/last five nucleotides of an exon) were also evaluated using the above-mentioned splicing algorithms. All in silico tools were used following the guidelines for the use of prediction tools (Vihinen, 2013).

2.5 Variant classification

The pathogenic impact of the CNGA3 variants listed in this study was automatically and/or semi-automatically classified according to the standards and guidelines provided by the American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP) (Richards et al., 2015) using the web-based variant interpretation tool Franklin (Genoox Ltd, https://franklin.genoox.com/). The criterion PP1 was thereby used for sequence alterations showing co-segregation of the variant(s) with disease in multiple affected family members, either performed in-house or documented in previously published studies. CNGA3 variants which were functionally analyzed in previously published studies and proven to have a damaging effect on protein function were assigned the criterion PS3. The criterion PP3 was used for missense variants that were predicted to be deleterious by CADD and MutationTaster, and for canonical/noncanonical splice site variants unanimously predicted to affect splicing using the above-mentioned in silico tools. If the applied in silico tools did not suggest a pathogenic effect, the criterion BP4 was assigned. For CNGA3 variants having a minor allele frequency of <0.1% on gnomAD, the criterion PM2 was applied. In contrast, variants with a minor allele frequency of >0.1% were assigned the criterion BS1.

3.1 Spectrum of CNGA3 gene variants in our ACHM study cohort genetically characterized in Tübingen

Our total ACHM cohort genetically characterized in Tübingen comprises 1060 affected individuals from 889 independent families carrying “likely disease-causing” variants in the six genes associated with ACHM. Among these, 385 patients from 304 distinct families carry likely biallelic ACHM-related variants in the CNGA3 gene, hereby revealing that CNGA3 is the second most commonly mutated gene in our study cohort (36.3% patients/34.2% families, Figure S1). Of note, 216 of these families have not been published previously.

In total, we detected 221 distinct genotypes of CNGA3 sequence alterations in our patient cohort (Table S2). Segregation analysis was performed in both parents for 181 affected individuals and in a single parent and/or in other family members (i.e., siblings, cousins, or offspring) for 88 ACHM patients confirming the presence and independent inheritance of two mutant biallelic CNGA3 alleles. In 116 cases no relatives were available for segregation analysis which corresponds to 116 families out of 304 families in the cohort missing segregation analysis (38.2%). One hundred and forty-seven affected individuals carry possibly disease-causing CNGA3 variants in an (apparent) homozygous state whereas 238 solved cases harbor two (likely compound) heterozygous variants expected to explain the ACHM phenotype. Homozygosity for the c.1641C>A/p.(Phe547Leu) or the c.847C>T/p.(Arg283Trp) CNGA3 variant were by far the most common genotypes in our cohort (n = 23 and n = 22 affected individuals, respectively; Table S2).

Moreover, we observed one family in our study cohort with pseudo-dominant inheritance involving two mutant CNGA3 alleles. In this family, the affected mother was homozygous for the c.847C>T/p.(Arg283Trp) variant and the clinically unaffected spouse was a heterozygous carrier of the c.488C>T/p.(Pro163Leu) variant. Both affected children in this family were compound heterozygous for the two CNGA3 variants inherited from the parents (Table S2).

One patient in our study cohort was shown to harbor two heterozygous missense variants (c.353A>G/p.(Gln118Arg) and c.1615G>A/p.(Val539Met); Table S2), of which the c.353A>G variant was classified as likely benign and the c.1615G>A alteration as a variant of uncertain significance (VUS) according to ACMG/AMP guidelines. Notably, this male patient was in the meantime shown to carry an alteration at the OPN1LW/OPN1MW gene cluster compatible with a genetic diagnosis of blue cone monochromacy, an X-linked incomplete form of ACHM. In one patient, only a single heterozygous CNGA3 variant was identified. Of note, in this patient screening of all other ACHM-associated genes did not identify any other likely causative variants. Both above-mentioned cases were excluded from the group of genetically solved ACHM cases.

In the ACHM cohort genetically characterized in Tübingen, we identified 140 different potential disease-associated variants. The ten most common CNGA3-ACHM-related alleles detected in our patient cohort are presented in Table 1. All but one of these are missense sequence alterations. The two missense substitutions c.1641C>A/p.(Phe547Leu) and c.847C>T/p.(Arg283Trp) are the by far most commonly observed CNGA3-associated alleles for ACHM in our cohort, both with a similar allele frequency (n = 86 chromosomes for c.1641C>A/p.(Phe547Leu) and n = 84 chromosomes for c.847C>T/p.(Arg283Trp)). Both alleles are frequently seen in ACHM patients originating from Germany (c.1641C>A/p.(Phe547Leu): n = 21 affected individuals; c.847C>T/p.(Arg283Trp): n = 22) and in the case of the c.1641C>A/p.(Phe547Leu) variant also in the Middle East (n = 15). Further common CNGA3 alleles associated with ACHM include the missense substitution c.829C>T/p.(Arg277Cys) (n = 48 chromosomes) which is mainly seen in German (n = 18) and US ACHM patients (n = 6). Similarly, we observed some descent-specific enrichment for c.1279C>T/p.(Arg427Cys) (total n = 26 chromosomes) in US patients (n = 9), as well as for the c.848G>A/p.(Arg283Gln) variant (total n = 20 chromosomes) observed in 5 out of 15 individuals who originated from Great Britain. Furthermore, the c.1495C>T/p.(Arg499Ter) variant (total n = 19 chromosomes) is the only nonsense variant detected in the spectrum of commonly observed CNGA3-ACHM-related alleles and was predominantly identified in French patients (n = 6) (Table 1 and Figure 1).

3.2 Comprehensive CNGA3 variant database

To create a comprehensive CNGA3 variant database, we combined all CNGA3 variants identified in our ACHM cohort with all CNGA3 sequence alterations obtained after an extensive literature search on PubMed®, and database search in HGMD®, LOVD and ClinVar. The so established CNGA3 variant database currently comprises 316 variants (Tables 2 and S3). Of these, four variants are located in the alternate exon 3b and its flanking intronic sequences. As the functional relevance of this sequence remains unclear and no association with disease has ever been reported, these variants were not further analyzed and were classified as “likely benign.”

Following the standards and guidelines of ACMG/AMP, 67 out of the remaining 312 variants were classified as pathogenic, 152 as likely pathogenic, and 71 were categorized as variants of uncertain significance (VUS). Furthermore, 12 out of 312 sequence alterations were classified as likely benign and 10 as benign (Tables 2 and S3).

The 316 CNGA3 variants listed in our CNGA3 variant database were revised manually and expert-categorized into “likely benign,” “VUS,” and “likely disease-causing” applying the following criteria. Variants categorized as “likely benign” (n = 44 variants; Tables 2 and S3) included: (1) variants classified as likely benign or benign according to ACMG/AMP guidelines, (2) synonymous variants that are not predicted to affect splicing, (3) missense variants with a minor allele frequency (MAF) higher than 0.1% and predicted to be tolerated by at least two in silico algorithms, (4) missense variants predicted to be tolerated by at least two in silico tools and only observed in a heterozygous state in the absence of a second “likely disease-causing” allele, (5) noncanonical splice variants not predicted to affect splicing and (6) variants located in the alternate exon 3b and its flanking intronic sequences.

Variants classified as “VUS” (n = 28; Tables 2 and S3) comprised missense variants that were classified as VUS following the ACMG/AMP guidelines and were predicted to be tolerated by at least two in silico tools. In addition, 15 variants were grouped in this category since no genotype could be established from the literature or database data. If the following criteria for a “VUS” were fulfilled, we considered patients carrying such variants with respect to the phenotype as solved cases: (1) observation of the variant in a heterozygous state with a (known) “likely disease-causing” variant and/or (2) segregation analysis demonstrated independent segregation of two variants and/or co-segregation of the variant with the disease.

All variants in the CNGA3 variant database not meeting the above-mentioned criteria were categorized as “likely disease-causing” (n = 244; Tables 2 and S3). Thus, we included the following variants in this “likely disease-causing” category: (1) variants classified as pathogenic or likely pathogenic following the ACMG/AMP guidelines, (2) variants representing likely null alleles (nonsense, start loss, canonical splice site, and frameshift variants), (3) missense variants classified as VUS following ACMG/AMP guidelines, but predicted to be deleterious by at least three in silico tools, (4) missense variants for which functional analysis revealed a deleterious effect on protein function, (5) noncanonical splice site variants classified as VUS following ACMG/AMP guidelines, but predicted to have an effect on splicing by at least three in silico tools, and (6) in-frame deletions affecting (completely) conserved amino acid residues. Consequently, all patients carrying CNGA3 variants categorized as “likely disease-causing” were designated as solved cases with respect to the ACHM phenotype.

Among 316 CNGA3 variants listed in the CNGA3 variant database, 251 variants were already published in the literature and in public databases. From these, we classified 196 variants as “likely disease-causing”, 23 as “VUS” and 32 as “likely benign” following manual and expert-based revision (Table S3). Only a small number of previously published missense variants were categorized as “likely benign” including the variants p.(Thr20Ile), p.(Pro48Leu), p.(Thr153Met), and p.(Val540Ile) based on the following observations: (1) All but one of these are located in the N-terminal region affecting residues which are neither conserved within CNGA3 orthologues nor the rod homolog CNGA1—only the p.(Val540Ile) variant affects a conserved residue in the CNBD. (2) All above-mentioned missense variants have a MAF higher than 0.1% and were predicted to be tolerated by in silico tools underlining their presumably nonpathogenic impact.

Within our CNGA3 variant database, we report 48 novel “likely disease-causing” variants detected in our ACHM cohort with six variants classified as pathogenic, 36 as likely pathogenic, and six as VUS following ACMG/AMP classification (Table S3). These previously unpublished variants comprise 24 missense substitutions, 14 small indel variants, and 6 nonsense substitutions creating a premature stop codon.

Overall, out of the 244 “likely disease-causing” variants in the CNGA3 variant database, missense substitutions were the most frequently observed variant type (66.8%) followed by small indel variants (16.8%) and nonsense alterations (12.3%) creating a premature stop codon (Figure 2). An overview of the mutation types and the respective frequencies of these “likely disease-causing” CNGA3 variants within our database is provided in Figure 2.

In addition, we observed one novel start loss variant affecting the first nucleotide of the translation initiation codon ATG (c.1A>T/p.(?)) and one copy number variation leading to the complete deletion of exon 8 (NC_000002.12:g.98393909_98399093del (GRCh38/hg38)). Actually, this deletion is the first large structural variation/copy number variation observed and reported for the CNGA3 gene and was identified in compound heterozygous state with the variant c.847C>A/p.(Arg283Trp) on the second allele in one patient. Following the standard variant screening, the variant c.847C>A/p.(Arg283Trp) was initially identified apparent homozygously in this index patient. Segregation analysis failed to detect the variant in the father of the patient. Microsatellite marker analysis confirmed paternity and deletion mapping via PCR and primer walking identified the heterozygous 5.19 kb deletion.

Copy number variations (CNVs) are an important mutation type associated with ACHM and are commonly seen in the CNGB3 gene as well as in GNAT2 and ATF6. Whereas ten out of 98 disease-associated CNGB3 variants identified in a large cohort of ACHM patients are CNVs, the mutation spectrum of ATF6 and GNAT2 includes two CNVs each (Felden et al., 2019; Kohl et al., 2002; Lee et al., 2020; Mayer et al., 2017). In contrast, it seems that this kind of structural variation is not common in CNGA3. This notion is further supported by the fact, that the gene had not been pinpointed by a CNV-study focusing on genes in inherited retinal dystrophies (Van Schil et al., 2018). In addition, single heterozygous cases are rarely found in CNGA3 compared to the CNGB3 gene where 44 out of 552 patients with CNGB3 variants were identified to harbor only a single heterozygous pathogenic variant following standard variant screening and CNV analysis (Mayer et al., 2017). This further suggests that most variants in CNGA3 are exonic indicating that there are not many hidden pathogenic variants—such as deep intronic variants—left.

Regarding the localization of “likely disease-causing” CNGA3 variants, 236 sequence alterations are located in the coding exons whereas only eight variants are located in the intronic sequences of the gene (Figure 3). Of these intronic variants, seven variants represent single nucleotide substitutions affecting either the canonical splice sites (i.e., the first two and last two nucleotides of an intron) (n = 5) or are located in the vicinity of a highly conserved splice acceptor site (n = 2). All variants affecting the canonical splice sites are predicted to lead to the loss of the respective donor or acceptor site based on five in silico algorithms. Additionally, the previously published variant c.396-11C>G/p.(?) and the novel sequence alteration c.450-15T>G/p.(?), both located in the vicinity of a highly conserved splice acceptor site, are predicted to activate an intronic cryptic acceptor site potentially affecting splicing (SpliceAI Δ scores: acceptor loss 0.87 and 0.65, respectively; acceptor gain 0.70 and 0.53, respectively; Table S3). Furthermore, one small duplication is localized at the exon/intron border of exon 5 (c.396-2_398dup/p.(?)). Although only one of the in silico tools predicts a minor effect on transcript splicing, the variant is still considered to be “likely disease-causing” (SpliceAI Δ score: acceptor loss 0.00; acceptor gain 0.22). The novel missense variant c.566G>A/p.(Arg189Lys) affecting the last nucleotide of exon 6 was also predicted to have an effect—though minor—on transcript splicing (SpliceAI Δ score: donor loss 0.08; donor gain 0.32). Consequently, it remains unclear whether this variant needs to be considered a splicing or missense variant, as it may also act both ways.

The majority of the exonic ACHM-associated CNGA3 variants classified as “likely disease-causing” within the CNGA3 variant database are located in exon 8 including 140 missense substitutions, 46 nonsense/small indel variants, and the single copy number variation (deletion of complete exon 8) (Table 2 and Figure 3). The remaining 49 variants are distributed across exon 2 to exon 7, with exons 6 and 7 (encoding TM1 and TM2) each harboring triple the number of variants as exons 2 to exon 5. On protein level, the majority of CNGA3 variants (n = 82) are spread across the six transmembrane domains and their connecting loops. Another major group of “likely disease-causing” CNGA3 variants (n = 57) clusters within the CNBD whereas only few variants (n = 8) reside in the adjacent CLZ domain more closely located to the C-terminus. The C-linker of the CNGA3 subunit and the pore region harbor a similar fraction of “likely disease-causing” CNGA3 variants (n = 36 and n = 30, respectively). The pore region spanning around 46 amino acids, is thereby rendered to have the highest mutation density within the CNGA3 subunit. Whereas the overall majority of variants within the CNGA3 protein are missense substitutions, the sequence changes within the N-terminal region mainly comprise nonsense variants or small indels (17 out of 22) (Figure 3). Of 163 “likely disease-causing” missense variants listed in our database, 94 affect amino acid positions which are completely conserved across CNGA3 orthologues and its rod equivalent CNGA1 (Table S3 and Figure S2).

3.3 Prevalence of CNGA3 variants in ACHM

From our in-house genetic investigations, we report disease-associated CNGA3 variants in 385 affected individuals descending from 304 independent families representing 36.3% of our ACHM cohort which predominantly comprises Western European and North American patients (Figure S1). This observation is in line with our previous and other studies reporting a prevalence of 5%–41% for CNGA3 variants in cohorts of Dutch and British ACHM patients making CNGA3 the second most frequently mutated gene in European ACHM patients or patients of European descent (Johnson et al., 2004; Thiadens et al., 2009; Wissinger et al., 2001).

The vast majority of all “likely disease-causing” CNGA3 variants listed in our CNGA3 database represent amino acid substitutions (66.8%) which are completely in line with several previously conducted studies where missense variants showed a frequency of 61.5%–71.4% (Li et al., 2014; Liang et al., 2015; Wissinger et al., 2001). In contrast, 16 out of 98 sequence changes (16%) in the CNGB3 subunit of the cone photoreceptor CNG channel identified in a large cohort of ACHM patients are missense variants (Mayer et al., 2017). This high prevalence of missense substitutions in the CNGA3 gene may indicate a reduced tolerance of CNGA3 for amino acid substitutions regarding the functional maintenance of the channel protein.

All “likely disease-causing” variants listed in our CNGA3 database except the c.682G>A/p.(Glu228Lys) variant have a minor allele frequency (MAF) lower than 0.1% in gnomAD (Table S3). In contrast, the mentioned variant displays a MAF of 0.13% (Table S3). In the South Asian population, the c.682G>A/p.(Glu228Lys) variant displays an even higher allele frequency of 0.99% in concordance with CNGA3 variants being more common here than in the European population. The c.682G>A/p.(Glu228Lys) variant has been observed in four patients from two families of South Asian origin in our cohort, but since it is part of complex alleles in these patients (homozygous for c.[67C>T;682G>A] or c.[682G>A;1315C>T]) the contribution of the c.682G>A/p.(Glu228Lys) variant to the ACHM phenotype in these families remains unclear—it may exert a cumulative effect (Table S2). In addition, one ACHM patient carrying the c.682G>A/p.(Glu228Lys) variant homozygously was described in a previously published study (Sun & Zhang, 2019). Furthermore, the variant was described to be part of a triallelic inheritance pattern involving the CNGB3 c.1208G>A/p.(Arg403Gln) variant (Burkard et al., 2018; Thiadens et al., 2010). Due to the deleterious effect predicted by two out of three in silico analysis tools and functional studies demonstrating impaired functionality of CNG channels harboring CNGA3_Glu228Lys subunits (Reuter et al., 2008; Täger et al., 2021), we classified this variant as “likely disease-causing.” A complete overview of the outcomes of the in silico prediction algorithms and previously published functional studies are included in Table S3.

3.4 Functional studies on CNGA3 variants

As a first gene therapy trial for ACHM resulting from defects in CNGA3 has already been completed (Fischer et al., 2020), and several trials are ongoing (ClinicalTrials.gov Identifier: NCT02610582, NCT03278873, NCT03758404, and NCT02935517), the genotypes of patients with CNGA3 missense variants classified as “VUS” unfortunately hamper the inclusion of such patients eligible for these gene therapeutic trials. Therefore, functional studies are needed to validate the possible pathogenic effect of the respective variants. Commonly used methods to characterize the functionality of mutant CNG channels involve calcium imaging and patch-clamp recordings of heterologously expressed mutant channels (Burkard et al., 2018; Dai & Varnum, 2013; Koeppen et al., 2008, 2010; Kuniyoshi et al., 2016; Liu & Varnum, 2005; Matveev et al., 2010; Meighan et al., 2015; Muraki-Oda et al., 2007; Patel et al., 2005; Reuter et al., 2008; Shaikh et al., 2015; Täger et al., 2018; Tränkner et al., 2004). Furthermore, a recently developed functional assay proposes a fluorescence-based, high-throughput calcium influx-based approach for analyzing mutant CNGA3 channels (Jacobson et al., 2019). In total, 54 missense substitutions, two nonsense and one canonical splice site variant have already been functionally characterized (Table S3), and this information was used to classify 44 variants as pathogenic, 11 as likely pathogenic, one as VUS—due to contradictory results in functional approaches—and one as benign according to ACMG/AMP guidelines.

Following functional characterization of heterologously expressed mutant channels in previously conducted studies, 27 out of 54 missense variants completely abolished formation of functional channels (Koeppen et al., 2008, 2010; Liu & Varnum, 2005; Matveev et al., 2010; Muraki-Oda et al., 2007; Patel et al., 2005; Shaikh et al., 2015; Tränkner et al., 2004). Seventeen previously analyzed missense substitutions allowed the formation of functional channels but with impaired/altered functionality such as impaired protein folding and/or channel trafficking or altered biophysical properties, for example, altered apparent cGMP-sensitivity or ion selectivity (Burkard et al., 2018; Dai & Varnum, 2013; Koeppen et al., 2008; Liu & Varnum, 2005; Meighan et al., 2015; Muraki-Oda et al., 2007; Patel et al., 2005; Reuter et al., 2008; Tränkner et al., 2004). Furthermore, an improvement in the mutant channel functionality was achieved in a total of ten missense variants by using chemical chaperones or co-expressing the CNGB3 subunit and/or reducing the cell cultivation temperature from the standard 37°C to 27/28°C, a procedure that is known to enhance protein folding and trafficking (Koeppen et al., 2010; Kuniyoshi et al., 2016; Patel et al., 2005; Reuter et al., 2008). A complete overview of CNGA3 variants previously analyzed in functional studies and the respective channel functionality is summarized in Table S3.