The locus for recessive STGD was mapped to chromosome 1 (1p21–p13).³ Mutations in ABCA4 have been described in autosomal recessive STGD (arSTGD),⁴ autosomal recessive retinitis pigmentosa (arRP),⁵ autosomal recessive cone–rod dystrophy (arCRD)⁶ and age-related macular degeneration (AMD).⁷

Up to now, ~500 disease-causing mutations have been identified in ABCA4. The mutation spectrum ranges from single base substitutions to deletions of several exons, although the majority of reported changes are missense mutations. As has been described, mutations in this gene account for 66–80% of STGD-associated chromosomes.^{8 9}

Despite all the efforts of many research teams, there is a variable mutation detection rate, and this is often not satisfactory because there are still many undiscovered mutations. Therefore, the combination of different technical approaches, together with knowledge of the genetic background for a given population is important for the determination of novel mutations and for the genetic characterisation of these patients.

The frequency of mutated alleles found with the microarray was 43.5% (27/62; 33.8% (21/62) for STGD patients, 4.8% (3/62) for arCRD patients and 4.8% (3/62) for arRP patients). However, the frequency of mutated alleles detected with microarray and dHPLC was 54.8% (34/62) for STGD patients and 6.4% (4/62) for arCRD patients. In contrast, the mutation detection rate for arRP cases was not increased by dHPLC scanning, as no additional disease-associated alleles were found.

The p.Arg1129Leu mutation was found to be the most frequent missense variant, representing 8% (5/62) of the total pathogenic alleles (table 1). In previous studies of the Spanish population, the p.Arg1129Leu variant was identified as a major mutant allele which accounted for 24% of the STGD alleles.¹⁴ This variant has been postulated to have a moderately severe effect and has predominantly been associated with a STGD phenotype.¹⁴ In contrast, the prevalence of this mutation in patients from North America was less than 1%.¹⁵ Interestingly, we identified one 30-year-old patient (ARDM-247), double heterozygous for the p.Arg1129Leu and p.Cys2137Tyr alleles, who presented a CRD phenotype. This p.Cys2137Tyr change was located more towards the amino terminus. Moreover, in other study, the results showed that the changes located in this zone appear to result in altered processing of the protein and to be associated with an earlier onset of disease.¹⁶ The p.Cys2137Tyr change in combination with the p.Arg1129Leu allele produced a CRD phenotype. Therefore, we speculate that the novel p.Cys2137Tyr variant could be a severe allele which is modifying the patient’s phenotype.

The second most prevalent missense disease-associated allele was p.Gly1961Glu (3/62; 4.8%), presenting a lower frequency than other European populations.¹⁰ This mutation was found in three STGD patients, two of them presenting at least one putative null mutation (c.1030insT and p.Gln2187X), who presented symptoms at the age of 24–25 years. Thus, we could suspect a moderate effect for the p.Gly1961Glu allele, as previously reported.⁷

In relation to null alleles, the most frequent mutation was c.3211insGT, identified in four independent STGD families out of 31 (6.4%). In two of these cohorts (ARDM-125 and ARDM-158), the second mutated allele was found by dHPLC (p.KNLFA1876dup and c.4537delC, respectively). In both patients, first symptoms of the disease appeared at the age of 9 years old. Nevertheless, in patients from families ARDM-165 and ARDM-167, in whom the second disease-associated allele was not detected, the age of onset of the disease was at the third decade of life (table 1). Therefore, we speculate that the second—not yet detected—mutation might be of a moderate effect, as the disease started later in life. The c.3211insGT allele has been previously described in other studies.^17–19. Rozet et al¹⁷ reported two STGD patients who were compound heterozygous for the c.3211insGT allele and another sequence change, whose age of onset was before 12 years old.¹⁷ In other population study performed by Briggs et al,¹⁸ two STGD patients harbouring the heterozygous c.3211insGT allele were described. These probands presented reduced central visual acuity before the age of 30. Interestingly, Paloma et al¹⁹ described one CRD case, compound heterozygous for the c.3211insGT and p.Arg212Cys variants, showing symptoms at the age of 9. Considering these previously reported genotype–phenotype correlations, together with the genetic and clinical findings described in our Spanish patients, we could hypothesise that the c.3211insGT variant is predominantly associated with STGD, while the second mutated allele might have a modulating effect on the patient’s phenotype.

Two additional null alleles, associated with STGD, were identified. The patient from family ARDM-125 presented the novel p.KNLFA1876dup, located at the second transmembrane region of the ABCR protein, while the patient from family ARDM-158 harboured the frameshift c.4537delC variant that causes a premature stop codon 12 amino acids downstream.

Nevertheless, as the c.3211inGT, p.KNLFA1876dup and c.4537delC are null alleles, and some patients presented symptoms early in life (9 years), we cannot rule out that they could develop a more severe phenotype (arCRD) in the future. Indeed, a similar situation has been previously reported in one Spanish patient homozygous for the c.2888delG allele,²⁰ whose clinical phenotype changed from STGD (age of onset 10 years) to arCRD (age of diagnosis 26 years), during the course of the study.

Two novel splice-site alterations were described, IVS21-2A>T (ARDM-90) and IVS38+5G>A (ARDM-181). Both patients presented STGD phenotype, and the age of onset of the disease corresponded to the first and second decade of life, respectively. In these cases, the total inactivation of the splice site is unlikely, although these changes could alter the splicing of corresponding exon, creating cryptic sites, thus producing a longer protein.²¹

In two patients diagnosed as having arCRD and arRP phenotypes respectively, we identified the missense p.Val1433Ile mutation, located in the second transmembrane domain. This sequence change has been previously associated with age-related macular degeneration (AMD)⁷ and has also been reported in one arCRD patient, although it was questioned whether the nature of this amino acid change was pathological or not.²² We investigated this change further in the Spanish control population, and it was not found in 124 ethnically matched chromosomes, thus suggesting a pathogenic effect.

Family RP-959 was previously analysed by the genotyping microarray, and the p.Ile156Val allele was detected.²⁰ This variant has been associated with the STGD phenotype. In the affected individual, the second mutated allele was not found by dHPLC. Segregation analysis of the mutation within the family did not allow this gene to be excluded. Thus, the presence of RP phenotype and this mutation could be explained because the second mutation would be a severe allele or would be due to fortuitous association. Moreover, in the RP-773 and RP-1058 families, it was not possible to detect the second mutated allele, so the implication of mutations in this gene in the development of RP is still not clear as was previously reported.²⁰

To establish the nature of the three non-reported changes located in the introns, we used a bioinformatics tool to analyse and compare the wild type and mutated sequences. The results demonstrated that these variants probably did not modify the splicing process.

By MLPA analysis, no deletions or duplications were found in patients. Indeed, it has been suggested that these alterations contribute to only a small fraction of disease-associated alleles for ABCA4.²³

The genotyping microarray is a comprehensive screening tool for genetic variation in patients with ABCA4-associated retinal pathology with a high reliability. Despite the relative small number of families, a huge number of ABCA4 disease-associated alleles were identified, representing an increment of 4% of novel variants described in this gene, as ~500 sequence changes have been described so far. To the best of our knowledge, this optimal screening strategy represents extensive mutational spectrum identification in Spanish patients with ABCA4-associated phenotypes.

Moreover, these new mutations could be added to the new versions of the ABCR400 microarray, thus increasing the detection and validity of the array. The combination of these three different techniques of screening (microarray, dHPLC and MLPA) allowed us to reach a complete diagnosis of 14 of 31 patients with only one mutant allele identified, because of the detection of the second disease-associated allele with by dHPLC. In addition, it is probable that a great number of mutations might reside in parts of the gene (eg, the promoter region or the introns) that have not yet been identified.

Given an estimated prevalence of STGD (1:10 000), the Hardy–Weinberg equilibrium would indicate that the heterozygous state can be expected in about 1/50.²⁴ So many variants, such as p.Val552Ile, could be really a mutation found in normal population. Thus, expression analysis could be an interesting tool with which to discriminate the pathological effect of these changes.