Personalized genetic counseling for Stargardt disease: Offspring risk estimates based on variant severity

Summary

Recurrence risk calculations in autosomal recessive diseases are complicated when the effect of genetic variants and their population frequencies and penetrances are unknown. An example of this is Stargardt disease (STGD1), a frequent recessive retinal disease caused by bi-allelic pathogenic variants in ABCA4. In this cross-sectional study, 1,619 ABCA4 variants from 5,579 individuals with STGD1 were collected and categorized by (1) severity based on statistical comparisons of their frequencies in STGD1-affected individuals versus the general population, (2) their observed versus expected homozygous occurrence in STGD1-affected individuals, (3) their occurrence in combination with established mild alleles in STGD1-affected individuals, and (4) previous functional and clinical studies. We used the sum allele frequencies of these severity categories to estimate recurrence risks for offspring of STGD1-affected individuals and carriers of pathogenic ABCA4 variants. The risk for offspring of an STGD1-affected individual with the “severe|severe” genotype or a “severe|mild with complete penetrance” genotype to develop STGD1 at some moment in life was estimated at 2.8%–3.1% (1 in 36–32 individuals) and 1.6%–1.8% (1 in 62–57 individuals), respectively. The risk to develop STGD1 in childhood was estimated to be 2- to 4-fold lower: 0.68%–0.79% (1 in 148–126) and 0.34%–0.39% (1 in 296–252), respectively. In conclusion, we established personalized recurrence risk calculations for STGD1-affected individuals with different combinations of variants. We thus propose an expanded genotype-based personalized counseling to appreciate the variable recurrence risks for STGD1-affected individuals. This represents a conceptual breakthrough because risk calculations for STGD1 may be exemplary for many other inherited diseases.

Graphical abstract

Introduction

An autosomal recessive disease displays a classical Mendelian inheritance pattern when all the pathogenic variants in one implicated gene have a consistent effect on the clinical phenotype. Assuming that variants in only one gene underlie a clinically well-recognizable condition, the sum allele frequency of disease-causing variants in the general population can be deduced from the disease prevalence on the basis of the Hardy-Weinberg principle, or, if they are not too rare, by analysis of their frequency in the general population. In case of normal reproductive fitness and non-consanguinity, the recurrence risk for the same autosomal recessive disease can thus be calculated. Several human diseases however display an atypical autosomal recessive inheritance due to the existence of variants that result in different levels of residual protein function, meaning that different combinations of variants can thus correlate with a spectrum of phenotypes or even the absence of a clinical manifestation. These varying effects of gene variants can complicate recurrence risk calculations. For example, different combinations of pathogenic USH2A variants can result in Usher syndrome type 2, consisting of retinitis pigmentosa (RP) and hearing impairment, non-syndromic RP (USH2A [MIM: 276901]), or even no phenotype.^{1, 2, 3} Another important illustration of autosomal recessive inheritance with a wide phenotypic spectrum was observed for CFTR-associated diseases. Cystic fibrosis (CF [MIM: 219700]) is caused by bi-allelic variants in CFTR. Different combinations of CFTR variants, many of which result in residual CFTR function, have been associated with CFTR-related disorders (CFTR-RDs), the mildest of which is congenital absence of the vas deferens (CAVD [MIM: 277180]) in males with no evidence of lung disease.^4,5 For several CFTR variants, functional assays have demonstrated the resulting residual CFTR activity,^{6, 7, 8} but for many missense variants, the functional effect is unknown. Finally, different combinations of ABCA4 variants (GenBank: NM_000350.2) can result in multiple retinal phenotypes, i.e., early-onset cone-rod dystrophy—sometimes diagnosed as retinitis pigmentosa (retinitis pigmentosa 19 [MIM: 601718])—panretinal cone-rod dystrophy (cone-rod dystrophy 3 [MIM: 604116]), due to the early degeneration of both cone and rod cells; classic Stargardt disease (STGD1); or late-onset macular degeneration (STGD1 [MIM: 248200]).⁹ ABCA4-associated STGD1 will be the focus of this study.

STGD1 represents the most prevalent inherited maculopathy, estimated to occur in 1 in 10,000 individuals.¹⁰ Although originally considered a juvenile macular degeneration, individuals with STGD1 may experience initial visual complaints between the first and the eighth decade of life.^{9,11, 12, 13, 14, 15} STGD1 results in visual impairment due to central or pericentral vision loss, impaired color vision, distorted vision, and/or visual field defects, with legal blindness after a median of 12 disease years.^12,16,17 Currently, in the absence of a treatment for STGD1, a major part of its management involves counseling about prognosis, e.g., to help make career choices, and recurrence risk, e.g., to aid in family planning.

The risk of passing a disease to (future) children is a common concern of persons with STGD1 and their relatives. This concern needs careful consideration in STGD1 particularly because of the high frequency of pathogenic ABCA4 variants in the general population. High ABCA4 variant carrier frequencies are illustrated by common observations of pseudodominant inheritance^{18, 19, 20, 21, 22} and different combinations of disease-causing ABCA4 variants among siblings.^17,23

The large difference in the age of onset of STGD1 between individuals with STGD1 is hypothesized to be mainly caused by the variable amount of residual ABCA4 activity. The combination of ABCA4 variants of different severity, ranging from null (severe) alleles to moderately severe or mild alleles, influences the clinical expression.^21,24,25 Especially relevant to recurrence risk is that combinations of pathogenic ABCA4 variants may or may not cause STGD1 depending on the variant severity. Two mild alleles, in principle, do not cause STGD1. Another layer of complexity was added to the existing ABCA4 genotype-phenotype correlation model, and thus genetic counseling, by evidence of incomplete penetrance.^26,27 We calculated that the very frequent hypomorphic variant c.5603A>T (p.Asn1868Ile) (allele frequency [AF] of 4.2% in the Genome Aggregation Database [gnomAD]), when in trans with a null allele, has a very low penetrance (<5%).^17,28 Moreover, a female bias among individuals with STGD1 was observed for this variant and for another frequent variant, c.5882G>A (p.Gly1961Glu).^17,29 These findings led to the hypothesis that ∼25% of persons with STGD1 display multifactorial or polygenic inheritance in which two ABCA4 alleles are a prerequisite to develop STGD1 but additional genetic or non-genetic modifiers play a significant role.

The ABCA4 mutational landscape has further been expanded by the discovery of causative deep-intronic variants and structural variants. Targeted whole-gene sequencing has revealed 42 causal deep-intronic variants with variable effects on RNA splicing in ∼5% of ABCA4 alleles and ∼10% of STGD1 affected individuals.^{30, 31, 32, 33, 34, 35} Moreover, 46 different structural variants have been identified in ∼2% of STGD1 affected individuals.^{32, 33, 34} Finally, at least 53 complex ABCA4 alleles have been identified, consisting of multiple disease-causing variants in cis,^33,36 illustrating the importance of establishing the phase of the variants found in diagnostic studies.

Despite the advances in genetic testing and increased knowledge of pathogenic variants, most variants are of uncertain significance.⁹ As shown by the proposed ABCA4 genotype-phenotype correlation model, knowledge of not only the pathogenicity but also the severity of ABCA4 variants is crucial in estimating recurrence risk. Existing guidelines for the interpretation of pathogenicity of variants are not tailored to assess variant severity. The commonly used 5-tier American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG-AMP) system addresses the likelihood of a variant’s pathogenicity on a scale from benign (1) to pathogenic (5), which in no way relates to its severity.³⁷ For instance, the frequent mild ABCA4 variant c.5882G>A (p.Gly1961Glu) is considered “likely pathogenic” according to the ACMG-AMP guidelines, whereas variants classified as variants of uncertain significance by ACMG-AMP may well cause a severe phenotype. As genetic counseling is mainly oriented toward the ACMG-AMP classification, the latter poses challenges for recurrence risk assessments in a clinical context. Therefore, genetic counseling of STGD1 families demands a genotype-based approach that considers variant severity.

For all conditions characterized by autosomal recessive inheritance with a wide spectrum of phenotypes due to differences in effects of genetic variants it is quite complex to estimate the risk for the offspring of affected persons to develop the same disease. This is particularly true for diseases due to genetic variants for which there is no or very limited information on their effect. Unlike most other recessive diseases, ABCA4-related STGD1 has been well studied, both on the molecular level as well as on the population level.⁹

Therefore, we propose an expanded genotype-based counseling approach for STGD1 that considers several variant severity categories instead of a binary benign/pathogenic categorization. In this study, we thereby aimed to calculate the risk for offspring of individuals with STGD1 and carriers to develop STGD1. We used published ABCA4 genotype data from individuals with STGD1 and ABCA4 AF data from the general population to assess the severity of 1,619 ABCA4 variants. The sum frequencies of ABCA4 alleles of different severities were calculated in the general population matched to the genetic ancestries of the population with STGD1. Using Hardy-Weinberg’s principle, we assessed recurrence risk for offspring of individuals with STGD1 and ABCA4 variant carriers.

Subjects and methods

Currently non-existing terminology limits the use of inclusive language. By “offspring” and “child,” we refer to the genetic offspring and child, and by “partner,” we refer to the other genetic parent of the offspring.

Dataset compositions: The bi-allelic affected person (BAP) dataset and the genetic ancestry-matched gnomAD (GAM-gnomAD) dataset

Recurrence risk calculations require knowledge of all ABCA4 variants that contribute to the disease in the given population, i.e., which variants are disease causing and how severe those variants are. For all null alleles from gnomAD, such as frameshift, stop, and canonical splice site variants, pathogenicity was assumed because none of those are known to be benign at the moment. For other variants, the recurrence risk calculations require knowledge of the sum AF of each severity category in the general population matched to the genetic ancestries of the population with STGD1 under study. In order to obtain this information, an STGD1 and a control dataset must be established.

Selection of the BAP dataset

We collected all publications until January 1, 2021 that contain ABCA4 variants in individuals with STGD1 with an autosomal recessive retinal dystrophy (supplemental material and methods) as well as data from 145 bi-allelic persons with STGD1 from Prevention Genetics (M. Pantrangi, personal communication). ABCA4 variant data as well as other published data such as sex, ethnicity, and age at onset, were uploaded to the LOVD-ABCA4 database (see web resources). Reported variants in bi-allelic persons with STGD1 were collected per individual. Only bi-allelic individuals with STGD1 were selected because mono-allelic individuals with STGD1 are less likely to have ABCA4-associated retinopathy. Individual records with the same variants that had been published by overlapping authors were considered duplicates and were removed from the dataset when there was no conflicting data on sex, ethnicity, or age at onset (n = 751). Furthermore, we excluded affected family members from the dataset (n = 58) to prevent a bias in variant frequency. The resulting dataset (n = 5,579) will be referred to as the bi-allelic person dataset (BAP dataset).

Composition of the GAM-gnomAD dataset

We derived ABCA4 AF data from the gnomAD, downloaded on April 13, 2021, and we used this data to establish a control dataset that matches the BAP dataset in terms of genetic ancestry on the basis of reported ethnicities of individuals with STGD1: the genetic ancestry-matched gnomAD (GAM-gnomAD) dataset (supplemental material and methods).

The sum AFs of ABCA4 alleles categorized by severity: Severe, moderately severe, mild with complete penetrance (mild^CP), and mild with incomplete penetrance (mild^IP)

The combination of test results from the following three tests were used for the severity category assignment: AF test, homozygosity test, and severity odds ratio test.

AF test: Benign variants test

To identify benign variants, we compared the frequency of each variant in the BAP dataset to the frequency of that variant in the GAM-gnomAD dataset by using Fisher’s exact and creating an odds ratio (OR) for each variant. An OR < 1 points toward a benign nature of the variant.

Homozygosity test: Mild variants test

As described previously, mild variants causing STGD1 in a homozygous configuration are only very rarely observed.^25,36 Therefore, a lower homozygous frequency in the BAP dataset than expected based on the AF in the general population indicates that a variant is mild. We compared the observed homozygous frequency of each variant in the BAP dataset to the expected homozygous frequency based on the AF in the GAM-gnomAD dataset assuming they would cause disease in a homozygous state by using Fisher’s exact: assuming a disease prevalence of 1:10,000, every person in the BAP dataset represents a population of 10,000 people. To calculate the expected homozygous occurrence of each variant in the BAP dataset, we therefore multiplied the GAM-gnomAD AF squared with the inverse of the assumed STGD1 prevalence (1:10,000) and with the BAP dataset size

$${\left(\mathrm{v}\mathrm{a}\mathrm{r}\mathrm{i}\mathrm{a}\mathrm{n}\mathrm{t}\mathrm{A}{\mathrm{F}}_{\mathrm{G}\mathrm{A}\mathrm{M}-\mathrm{g}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{A}\mathrm{D}}\right)}^2\times \mathrm{10,000}\times \left(\mathrm{B}\mathrm{A}\mathrm{P}\mathrm{d}\mathrm{a}\mathrm{t}\mathrm{a}\mathrm{s}\mathrm{e}\mathrm{t}\mathrm{s}\mathrm{i}\mathrm{z}\mathrm{e}\right).$$

An OR was calculated for each variant. A low OR equals a low frequency of homozygous configuration and therefore points toward a mild nature of the variant.

Severity odds ratio: Severe variants test

According to the model introduced by Maugeri et al., mild and moderately severe variants are not expected to cause STGD1 disease when in trans with a mild variant.²⁵ Therefore, to distinguish severe variants from mild and moderately severe variants, we compared the ratio of each variant occurring in trans with previously established mild variants and previously established severe variants (supplemental material and methods) to the same ratio in a reference group of previously established severe variants in the BAP dataset by using Fisher’s exact. An OR was calculated for each variant. A high severity OR (>0.8) means that a variant has a relatively high frequency with mild variants and therefore points toward a severe nature of the variant, whereas a low severity OR (<0.8) points toward a mild or moderately severe nature of the variant. Due to random fluctuations in variant combinations, a cut-off value had to be chosen between 1—the expected odds ratio for severe variants—and 0.7—the median OR for the 15 known moderately severe variants for which data was available. The cut-off value of 0.8 has been chosen arbitrarily.

Severity category assignment

In order to get a best estimate of sum AFs per severity category, we assigned variants to the categories “benign,” “mild^IP,” “mild^CP,” “moderately severe,” and “severe” or multiple on the basis of the outcomes of the three aforementioned tests; the steps are described in detail in the supplemental material and methods and Figure S1.

Sum AFs per severity category

Initial sum AFs were created per severity category. Because of the subset of variants (n = 192) that were inconclusively allocated to two severity categories, we made an underestimate and an overestimate of severity AFs. In the underestimate AF, all ambiguous variants were allocated to the least severe category, while those same variants were allocated to the most severe category in the overestimate AF. The sum AFs of variants that were categorized as one of three or more categories were proportionally divided over the three pathogenic severity categories for “causative of unknown severity” and over benign, mild^CP, moderately severe, and severe for “not categorized.” For this, the sum AF of benign variants only included those variants with an AF < 0.01 across all gnomAD populations, as we considered the more frequent benign variants as erroneously included in studies despite a high AF. The sum AF of 65 null alleles from gnomAD that were absent from the BAP dataset were added to the sum AF of the severe category. The sum AFs were further adjusted according to the factors described below.

Consideration of frequent complex alleles

Several frequent mild alleles occur in cis with other pathogenic ABCA4 variants. Therefore, the AF was corrected for c.[769−784C>T; 5603A>T] (p.[=, Leu257Aspfs^∗3; Asn1868Ile]), c.[1622T>C; 3113C>T] (p.[Leu541Pro; Ala1038Val]), c.[2588G>C; 5603A>T] (p.[Gly863Ala, Gly863del; Asn1868Ile]), c.[5461−10T>C; 5603A>T] (p.[Thr1821Aspfs^∗6, Thr1821Valfs^∗13; Asn1868Ile]). In the calculation of the sum AF of mild^IP, variants c.769−784C>T and c.2588G>C were only partially considered (25% [Genome of the Netherlands]³⁸ and 10%,²⁵ respectively) because these are likely benign if c.5603A>T is not located on the same allele. Also, the AF of the severe c.5461−10T>C variant was subtracted from the AF of c.5603A>T because they are in linkage disequilibrium.³³

Likewise, c.3113C>T was only partially considered in the sum AF of mild^CP alleles because it has a confirmed severe effect when in cis with c.1622C>T. Eighty-five percent of the AF of c.1622C>T was therefore subtracted from the sum AF of mild^CP alleles with complete penetrance because 85% of all c.1622C>T alleles in our BAP data also contains c.3113C>T and this is expected to be a similar proportion in the general population.

Recurrence risk calculations

ABCA4 genotype frequencies of unaffected individuals in the GAM-gnomAD dataset

We assessed the ABCA4 genotype frequencies in the GAM-gnomAD dataset on the basis of the Hardy-Weinberg principle (p² + 2pq + q² = 1). Consequently, genotypes with two alleles from the same severity category have an occurrence of that category’s sum AF squared, while genotypes of two variants of two different categories have an occurrence of two times the multiplication of the categories’ sum AFs. We assume that the frequencies of severe, moderately severe, mild^CP, mild^IP (p.Asn1868Ile), and wild-type alleles add up to a total of 100%, that allele and genotype frequencies remain rather stable throughout time, that the STGD1 prevalence is equal in all analyzed populations, and that genetic variation was spread evenly through the population, even though this is not the reality. Recurrence risks are calculated for situations where one of the biological parents has an unknown ABCA4 genotype, a situation that applies to most unaffected individuals in the general population. However, some individuals with a pathogenic ABCA4 genotype may be clinically diagnosed after their child is born and are therefore most likely considered an unaffected individual at the time of risk assessment. The mean age of biological parents of children born in 2018 in the European Union was 32 years according to Eurostat and CBS (see web resources). Individuals with STGD1 with the genotypes severe|mild^CP and moderate|moderate are generally diagnosed with STGD1 around this age. Therefore, only the proportion of individuals with these genotypes and a diagnosis age > 32 in our STGD1 cohort (E.H.R., unpublished data) was taken into consideration in the calculations of recurrence risk involving an unaffected parent; this was 21% of the moderate|moderate genotype frequency and 44% of the mild^CP|severe genotype frequency.

Penetrance of c.5603A>T (p.Asn1868Ile): 5% to 65%

It is likely that incomplete penetrance occurs for several—if not all—known mild variants. However, in this study, only the allele with the strongest evidence of incomplete penetrance was considered: the non-complex p.Asn1868Ile. We implemented a penetrance of 5% when this allele is passed on from an unaffected carrier, as its penetrance is estimated at 5% in the general population.^17,28 An alternative penetrance of 65% is implemented when an individual with STGD1 passes p.Asn1868Ile to their child, which is the penetrance rate we observed among relatives in a study cohort of 27 families.^17,28 For comparison, we included both a 65% and a 5% penetrance calculation when the p.Asn1868Ile variant was inherited from an unaffected ABCA4 variant carrier. The higher penetrance rate should be considered when the carrier has first-degree relatives with STGD1, most often a sibling or a parent with STGD1.

Results

Composition of datasets

In total, data on ABCA4 variants in 5,579 persons with STGD1 were collected (BAP dataset), consisting of 1,619 unique variants. Genetic ancestry data showed the following distribution: non-Finnish European (67.8%), East Asian (9.7%), Latino/Admixed American (7.0%), African (6.8%), other (4.5%), South Asian (4.1%), Finnish (0.02%), and Ashkenazi Jewish (0.02%) (Table S1). A GAM-gnomAD dataset was constructed with corresponding genetic ancestry ratios containing 132,890 alleles.

Cumulative AF of severe, moderately severe, and mild alleles

Of the 1,619 variants in the BAP dataset, 1,285 variants (79.4%) could be categorized into one or multiple severity groups (Table 1, Table S2A, and Table S2B). 1,044 variants were categorized to a single severity group, and the other 241 corresponded to more than one potential severity category. In total, 83 variants were categorized as benign (supplemental materials and methods and Figure S1, categorization steps 3, 4, and 14). Furthermore, 65 null alleles from the GAM-gnomAD dataset that were absent from the BAP dataset were added to the sum AF of severe variants. Finally, the sum AF of potential pathogenic alleles in the GAM-gnomAD dataset was 7%–8%.The total sum AFs per severity category per genetic ancestry group can be found in Table S3.

Table 1.

Variant categorization

Number of categories	Severity category	Number of unique variants per severity category
1	benign	83
	mild	1IP, 26CP
	moderate	76
	severe	923
2	benign/mild	2
	mild/moderate	140
	moderate/severe	42
3	mild/moderate/severe	57
4	benign/mild/moderate/severe	334

All ABCA4 variants from the BAP group were categorized into severity categories and additional null variants from gnomAD were added to the total. Of these, one variant was considered mild with incomplete penetrance (mild^IP; step 1). Twenty-six variants were categorized as mild with complete penetrance (mild^CP; steps 1, 5, 9, and 15). Seventy-six variants were considered moderately severe (steps 1, 6, and 8). One hundred and forty variants were considered either mild^CP or moderately severe (step 12). Concerning the severe variants, 858 variants were categorized as severe (step 1, 2, and 10; no variants were categorized in step 7). Another 65 null variants were added from the gnomAD data that were absent in the BAP dataset list. Forty-two variants were considered to be either moderately severe or severe (step 11). Two variants were categorized as either benign or mild^CP (step 13). Fifty-seven variants were considered to be a causative variant of unknown severity (step 16). Finally, the sum AF of the 334 variants that could not be classified was proportionally divided over the known categories benign, mild^CP, moderately severe, and severe.

Genotype-based recurrence risk

The genotype frequencies per severity category in the general population based on the GAM-gnomAD dataset are shown in Figure 1. It shows that ∼10% of individuals carry at least one p.Asn1868Ile allele and ∼5% of individuals carry at least one (potentially) pathogenic ABCA4 allele other than p.Asn1868Ile. The recurrence risks for five genotype scenarios were calculated on the basis of the GAM-gnomAD dataset (Figure 2 and Table S4). Results per gnomAD genetic ancestry population can be found in Figures S2–S8.

Figure 1.

Estimates of ABCA4 genotype frequencies in the general population

Mild^CP, mild ABCA4 variant with complete penetrance; mod., moderately severe effect; N1868I, mild ABCA4 variant with incomplete penetrance, c.5603A>T (p.Asn1868Ile); sev., severe effect; WT, wild type.

Sum AFs are presented of different severity categories in the genetic ancestry-matched gnomAD dataset. Data include the sum AFs of causative variants of unknown severity divided over mild^CP, moderately severe and severe, and variants with unknown pathogenicity divided over benign, mild^CP, moderately severe, and severe. ↓ represents the underestimate and ↑ the overestimate of the sum AFs and the corresponding genotype frequencies. Genotype frequencies are calculated with Hardy-Weinberg’s principle, p² + 2 × p × q + q² = 1.

Red shading indicates a genotype that results in an STGD1 phenotype; orange shading indicates a genotype that may result in an STGD1 phenotype; light blue shading indicates a genotype that is unlikely to result in an STGD1 phenotype but might contribute to a pathogenic genotype in offspring; white shading indicates a genotype does not result in an STGD1 phenotype and does not contribute to a pathogenic genotype in offspring.

Figure 2.

Genotype estimates for offspring of a person with STGD1 or a carrier of a pathogenic ABCA4 variant and an unaffected partner with an unknown genotype

(A and B) The blue boxes present the genotype scenarios for an individual with STGD1 (A) and a known ABCA4 variant carrier (B) who have a child with an unaffected individual harboring an unknown ABCA4 genotype (white box on the left side). Ranges indicate the estimates based on the pathogenic sum allele frequency underestimate and overestimate in the genetic ancestry-matched gnomAD. The risks are divided by genotype in the white boxes on the right side. The numbers in the blue boxes on the right represent the total risk of having affected offspring. Two different penetrance rates of p.Asn1868Ile (hereafter: N1868I) alleles were implemented: the N1868I allele in the general population has a penetrance of approximately 5%, whereas the N1868I allele within families with affected individuals shows higher penetrance, roughly estimated at 65%. The higher penetrance is therefore probably only applicable if the offspring inherited the N1868I allele from the affected parent (N1868I allele with underscore) or from the unaffected carrier of an ABCA4 variant in whose (first degree) family that same allele had been penetrant.

The “}” specifies the risk of offspring with a severe|severe or severe|moderately severe genotype, who most likely manifest STGD1 in childhood.

∼, an individual with the severe|mild^CP or moderately severe|moderately severe genotype may be unaffected at the age they conceive a child and be diagnosed with STGD1 shortly or a long time thereafter. The presented risk estimates do not apply to situations where both (future) parents have STGD1, in which case it is generally feasibly to narrow down the possible inheritance scenarios on the basis of known ABCA4 genotypes of both (future) parents. Based on our data from Dutch individuals (E.H.R., unpublished data), 44% of STGD1-affected individuals with a severe|mild^CP genotype and 21% with the moderate|moderate genotype received the diagnosis after the average age of European parents of children born in 2018 (32 years). The allele frequencies for these genotypes are corrected for those percentages.

N1868I, mild (p.Asn1868Ile) ABCA4 variant with incomplete penetrance. Mild^CP, mild ABCA4 variant with complete penetrance. Mod., ABCA4 variant with moderately severe effect on ABCA4 function. WT, wild-type ABCA4 allele.

The risk of STGD1 for the offspring of an individual with STGD1 highly depends on the genotype of the individual with STGD1 (Figure 2A). As the combination of a severe ABCA4 variant with any other pathogenic ABCA4 variant will cause STGD1, the recurrence risk for offspring of a person with STGD1 with two severe alleles and a non-tested unaffected individual is highest, estimated at 2.8%–3.1% (1 in 36–32) in the population represented by the BAP group. This is twice the risk of STGD1 for offspring of an individual with STGD1 who harbors a severe and a mild^CP allele: 1.6%–1.8% (1 in 62–57). Of important note: the risk that offspring will develop STGD1 already in childhood (0.68%–0.79%, 1 in 148–126, for a person with STGD1 who harbors two severe variants) is estimated 2- to 4-fold lower than the risk that offspring will develop STGD1 at any moment in life (2.8%–3.1%, 1 in 36–32, for an individual with STGD1 who harbors two severe variants). This is because the genotype severe|mild—which typically does not lead to an onset in childhood—is most prominently present in any offspring genotype scenario.

The risks for offspring of unaffected ABCA4 variant carriers varies tremendously, and again the severity of the ABCA4 allele is the most important determinant. When an ABCA4 variant carrier from an STGD1-affected family (for instance, a child or genetically tested sibling of an individual with STGD1) will have a child with an unaffected non-tested individual who does not have relatives with STGD1, the recurrence risk varies between 0.13% (1 in 782, for carriers of a mild^IP allele) and 1.5% (1 in 65, for carriers of a severe allele; Figure 2B). The risk that offspring develops STGD1 in childhood is estimated at practically zero for a known carrier of a mild allele to 0.34%–0.40% (1 in 296–252) for a known carrier of a severe allele.

Discussion

Genetic counseling in STGD1 demands a personalized approach. Generalized counseling in autosomal recessive diseases based on the premise that a genotype consisting of two pathogenic alleles causes disease is inadequate in STGD1 because of the considerable genotypic and phenotypic variability. Two main aspects highlight the relevance of a personalized, genotype-directed counseling approach. First, the recurrence risk of STGD1 highly depends on the severity of the ABCA4 alleles the parent carries. Accordingly, risks in the investigated population were estimated to range from 0.68%–3.1% (1 in 148–32) when one parent was an individual with STGD1 and 0.13%–1.5% (1 in 782–65) when one parent was a known carrier of an ABCA4 variant from an STGD1 family. As such, a sibling or child of a person with STGD1 should be informed that, depending on the severity of the allele they carry, the recurrence risk of STGD1 at any moment in life for their children is either negligible or quite real. Second, people who seek counseling for family planning often involve an individual with an early onset of STGD1. They might want to be counseled specifically for the risk that their offspring develops STGD1 at a young age. The recurrence risks for genotypes that generally lead to STGD1 in childhood are 2- to 4-fold lower than the total STGD1 recurrence risks.

These personalized risk estimates enable persons with STGD1 and their family to make informed decisions regarding carrier testing of a healthy partner and/or reproductive decisions. Consequently, one may consider ABCA4 testing of an unaffected partner. Today, whole-gene sequencing will reveal up to 95% of causal variants in clinically well-characterized STGD1-affected individuals. If no putative causal ABCA4 variants are identified in the partner, the risk of STGD1 for the offspring accordingly will decrease up to 20-fold. When a (likely) causal variant is found, prenatal or pre-implantation genetic diagnostics or prenatal testing could be discussed with the prospective parents, taking into account knowledge of the pathogenicity as well as the severity of the variant.

Our study provides data on variant severity for many variants in ABCA4. This data can help clinicians interpret the consequences of a specific ABCA4 variant for an individual with STGD1 only if the robustness of the data underlying the variant categorization is carefully considered. It is likely that a proportion of variants, allocated on the basis of lesser robust data, have been assigned to the wrong severity category. However, the methods we used did not create a bias of variant allocation into one category over the other and thus provide an adequate estimate of the overall sum AF per severity category. Due to insufficient data, 391/1,619 unique ABCA4 variants could not be assigned to one specific or either of two severity categories. Their total AF in the GAM-gnomAD dataset was only 0.29%. Excluding these variants from the analysis completely would only be justified if each of these was benign, which is unlikely. Finally, stop-gain variants may not always lead to a null allele,³⁹ but of all the stop-gain variants that had a low severity odds ratio, none were found homozygously less often than expected in the BAP dataset, indicating that they are moderately severe or severe. Genetic counseling for individuals carrying a variant that was assigned a severity category on the basis of the lesser robust data should especially consider that the recurrence risk lies on a range of estimates provided by several possibly relevant genotypic scenarios.

STGD1 recurrence risks are greatly influenced by differences in penetrance due to the high frequency of the incompletely penetrant p.Asn1868Ile allele. The estimated penetrance rates of 5% for the general population to 65% for familial cases are based on the only data available to date: population data and data of a small number of STGD1-affected families, respectively.^17,28 When p.Asn1868Ile is inherited from an unaffected first degree relative of an STGD1-affected individual, the penetrance might be lower than 65% as also the total number of genetic modifiers outside ABCA4 may be lower, as compared to the situation where the p.Asn1868Ile allele has been penetrant in a first degree relative. We recommend counselors to carefully consider these potential differences in penetrance based on the family history of both (future) parents.

When estimating the risk of STGD1 for the offspring of STGD1-affected individuals or unaffected carriers of ABCA4 alleles, several other factors need to be considered. First, given the hypothesized polygenic or multifactorial nature of STGD1 for a significant fraction of affected individuals (∼25%), the risk estimates may vary depending on the culture, population, and country. Second, ABCA4 variant carriership reported in literature varies tremendously (6%–20%),^25,40,41 and accordingly, the chances of meeting an unaffected partner who carries a (potentially) pathogenic ABCA4 allele may vary. This variability is highly influenced by genetic testing method, variant interpretation, new insights over time, and possibly by population. In our study, the AF of (potentially) pathogenic alleles totals 7%–8%, meaning that according to Hardy-Weinberg calculations, 14%–15% of individuals in the general population carries at least one potentially pathogenic ABCA4 allele (2pq + q²).

We calculated recurrence risks on the basis of the AFs in different genetic ancestry subpopulations, which in theory would result in a more precise, personalized estimate. Risks in subpopulations differed up to 230% from the general population that was ethnically matched to the BAP group. However, risk estimates for separate subpopulations should be considered with special caution. First and most importantly, genetic variation within populations is likely much larger than between populations.⁴² Moreover, one’s self-reported ethnicity might not correspond with genetic ancestry grouping on the basis of genetic similarities. Local AFs can highly impact the risk assessment, e.g., c.5882G>A, p.Gly1961Glu is extremely frequent in the Somali population with an allele frequency of 11.3%, potentially increasing the risk of STGD1 in this population.⁴³ Local AFs may have to be incorporated into the risk assessment when they differ highly from the published data. Another source of uncertainty concerns insufficient knowledge on complex alleles, especially in understudied populations. Recurrence risk is overestimated when unrecognized complex alleles are included in sum AFs in duplicate. Furthermore, the causal variants as well as the severity of variants are less studied especially in populations outside Europe and North America. Some populations are underrepresented in both the BAP group as well as the control group. Consequently, causal variants in these populations may be underreported and unidentified, leading to an underestimation of the total recurrence risk in these populations. Finally, prevalence studies in understudied populations might help indicate whether the risk of STGD1 and of passing on STGD1 to offspring indeed varies across populations or whether there still remain a lot of disease variants to be identified.

Implications for other autosomal recessive disorders with a wide spectrum of phenotypes

The genetic risk estimates for offspring of individuals with STGD1 represent a paradigm for many other autosomal recessive human genetic diseases with a wide spectrum of phenotypes. Two prevalent inherited human diseases that resemble STGD1, i.e., with a high variant prevalence in the general population, a broad spectrum of variants, and accompanying clinical effects, are the previously mentioned spectrums of diseases due to pathogenic variants in CFTR or USH2A. For these conditions, a large number of causal variants has been collected in open-access databases such as LOVD-USH2A and CFTR2 (see web resources), and population-specific allele frequencies are known. Similarly as presented in this study for STGD1, it is now feasible to establish a genotype-based disease risk for offspring of individuals with STGD1 or heterozygous carriers with variants in CFTR and USH2A, allowing for prediction of the different risks for specific phenotypes associated to the gene. These calculations are at present less reliable for rare conditions, as there currently are insufficient variant prevalence data in general populations.

In this study, we also used statistical methods that allowed us to provide an initial prediction of the severity of ABCA4 missense variants and other small protein variants. To provide an accurate disease risk assessment for future offspring, it will be important to provide in silico variant severity predictions until functional tests provide accurate data. This becomes increasingly important, as it is expected that preconception carrier screening of disease-associated variants, not only in families with genetic conditions, will increase.^44,45

The tremendous advances in genetic testing and knowledge of ABCA4 made over the past decade should now be followed by the improvement of genetic counseling to individuals with STGD1 and their families. In the future, recurrence risk estimations need to be optimized on the basis of new studies on the functional effect of ABCA4 variants and increased knowledge on genotype frequencies worldwide as well as differences between populations and within populations, haplotypes, allele severity, and penetrance. For individuals with STGD1 and their families today, the presented genotype-directed approach can help inform them about their personalized recurrence risks more reliably than generalized autosomal recessive risk counseling.

Published: February 3, 2022

Data and code availability

The BAP dataset generated during this study has been uploaded to the LOVD-ABCA4 database. The specific dataset is available from the corresponding author on request.

Web resources

• CBS, https://www.cbs.nl/en-gb/news/2011/39/one-in-six-first-time-fathers-over-40

• CFTR2, https://cftr2.org/mutations

• Eurostat, https://appsso.eurostat.ec.europa.eu/nui/submitViewTableAction.do

• gnomAD, https://gnomad.broadinstitute.org/

• LOVD-ABCA4, https://databases.lovd.nl/shared/genes/ABCA4

• LOVD-USH2A, https://databases.lovd.nl/shared/genes/USH2A

• OMIM, http://www.omim.org/