Over 200 ABCA4 sequence variants have been reported so far in patients with STGD and other retinopathies1,5,6,7,8,9,10,11,12,13,14,15,16,17. We have examined 33 missense mutations, 3 small in-frame deletions and 1 frameshift near the carboxy terminus (Table 1 and Fig. 2), including those mutations most commonly encountered in STGD patients1,5,6,7,8,9,10,11 and several that were reported in AMD patients15. As an initial step in assessing protein folding and stability, we analysed each ABCR variant by immunoblotting and azido-ATP labelling (Fig. 3). Mutations that cause small deletions (delVVAIC1681 and delPAL1761) or introduce charged amino acids into predicted transmembrane domains (G851D and G1886E) produce greatly reduced amounts of protein. Among the ABCR variants that are expressed with normal or nearly normal yield, azido-ATP labelling revealed a subset that is defective in ATP binding. A variety of mutations that lie outside of the nucleotide-binding domains (NBDs) can impair azido-ATP labelling, including L541P, predicted to reside adjacent to a transmembrane domain, and W1408R, which resides between the homologous halves of ABCR (Fig. 2). These data suggest that ATP binding to the NBDs is allosterically coupled to conformational changes in or near the transmembrane regions. Moreover, some mutations within either of the two NBDs abolish or nearly abolish all azido-ATP labelling, as seen, for example, with variants T971N, L1971R, G1977S and E2096K, implying allosteric coupling between the two NBDs, as described for P-glycoprotein22,23.Table 1 Naturally occurring ABCR variants produced in transfected 293 cellsFull size tableFigure 2: Locations of 37 naturally occurring ABCR sequence variants and 4 synthetic mutations.The predicted transmembrane topography and domain structure of ABCR is based on the hydropathy profile and sequence alignment with other ABC transporters. The cytosolic face of the membrane is downward. NBD, nucleotide binding domain; HH, highly hydrophobic domain shared with other members of the ABC1/ABCR subfamily of ABC transporters. A, B and C indicate the sequence motifs characteristic of nucleotide binding folds. Asterisks denote the four synthetic mutations.Full size imageFigure 3: Protein yield and ATP-binding capacity of 37 naturally occurring ABCR variants produced in transiently transfected 293 cells.Membranes were analysed by immunoblotting with affinity-purified anti-ABCR antibodies (top) and photoaffinity labelling with α-32P azido-ATP (bottom). We loaded 1 μg (immunoblotting) or 2.5 μg (azido-ATP labelling) of total membrane protein, as determined by Bradford assay, per track. The mutations that reside in NBD-1 and NBD-2 are indicated above the corresponding lanes. The relative levels of the different variant proteins and the extent of azido-ATP labelling were observed to be highly reproducible in multiple independent experiments. ABCR (large arrowhead); an endogenous 55-kD protein (small arrowhead) serves as an internal control for azido-ATP labelling. Molecular mass standards are shown on the left in kD.Full size imageThe combination of immunoblotting and azido-ATP labelling revealed defects in more than 75% of the variants tested. Among the variants with reduced yield and/or ATP binding are G863A and delG863, the two protein products of a guanosine2588→cytosine mutation that both generates a glycine-to-alanine substitution at codon 863 and activates a cryptic splice acceptor site in exon 17 that results in the removal of codon 863 from approximately 50% of the transcripts10. This is the most common allele among STGD patients in Northern Europe, representing roughly 20% of disease-associated alleles. It is also present at a frequency of approximately 3% in the general population in Northern Europe and approximately 1% in the United States population7,9,10. Genotype-phenotype correlations suggest that it is a mild allele and that it leads to STGD only when paired with a more severe allele10. Relative to wild type, the G863A variant is subtantially impaired and the delG863 variant is mildly impaired (Fig. 3).

The plots of fluorescence anisotropy changes of 11-cis-retinal with wild type and mutant NBD1 protein titrations are shown in Fig. 6, A–C. The binding of mutant G863A was significantly attenuated as indicated by the drastic right shift of the binding isotherm as well as its inability to achieve saturation in the presence of a high concentration of protein (Fig. 6A). Nonlinear regression analysis gave Kd of 8.0 ± 1.3 × 10−8 m, 8.0 ± 2.0 × 10−6 m, and 4.0 ± 1.4 × 10−6 m for the wild type and R943Q and P940R mutations, respectively (Fig. 6 and Table 1). As a consequence of Stargardt disease mutations, 100-fold (R943Q) to 50-fold (P940R) decreases in the binding affinity of the NBD1 domain for 11-cis-retinal were observed. The retinal binding of the G863A mutant was severely attenuated, and the Kd was ≥1.0 × 10−5 m.

The measurement of ATPase activity has been the only assay available to study the effects of mutations on ABCA4 function. We employed this assay to examine the effects of [L541P; A1038V], R602W and C1490Y mutations on in vitro ATP hydrolysis. Constructs containing wild-type and mutated ABCA4 cDNAs, tagged with the eight amino acid bovine opsin C-terminal epitope (1D4), were expressed in COS7 cells and proteins were purified on a 1D4 affinity column. CHAPS-solubilized ABCA4 was incubated subsequently with ATP, and the hydrolysis rate was estimated with the charcoal method (31).The rate of ATP hydrolysis of the complex allele [L541P; A1038V] was decreased to 68.1% of wild-type ABCA4 (Fig. 3).

A 37-year-old East Asian woman

G818EER51 ± 1363 ± 5111 ± 825 ± 312 ± 32Moderate/Severe

Analysis of the rates of hydrolysis of ATP and CTP in the mutant protein demonstrated that they were significantly reduced (Fig. 3). The results presented in Fig. 3A indicated that the ATPase function of G863A mutant protein was reduced ∼3-fold as compared with NBD1wt, indicating ∼70% of inhibition of the ATPase activity. TheVmax (ATPase) for G863A mutant was 128 pmol/min/mg and that of the wild-type NBD1 was 584 pmol/min/mg (Table II). A time-course analysis of ATP hydrolysis using 2.5 μg of protein (Fig. 3C) suggested the actual rates of ATP hydrolysis were attenuated 3-fold as a consequence of the mutation. We have earlier reported that the NBD1wt has significantly higher CTPase than ATPase activity (1717.Biswas, E.E.Biochemistry. 2001; 40:8181-8187CrossrefScopus (25)PubMedGoogle Scholar). However, in the mutant G863A protein the CTPase activity was reduced (Fig.3B). In this case, the CTP hydrolysis of G863A was reduced to ∼30% of the activity of NDB1wt. The Vmaxfor G863A mutant was 104 pmol/min/mg and that of the wild-type NBD1 was 376 pmol/min/mg (Table II).

Complete deletion