Quantitative Fundus Autofluorescence and Optical Coherence Tomography in PRPH2/RDS- and ABCA4-Associated Disease Exhibiting Phenotypic Overlap

Abstract

Purpose.: To assess whether quantitative fundus autofluorescence (qAF), a measure of RPE lipofuscin, and spectral-domain optical coherence tomography (SD-OCT) can aid in the differentiation of patients with fundus features that could either be related to ABCA4 mutations or be part of the phenotypic spectrum of pattern dystrophies.

Methods.: Autofluorescence images (30°, 488-nm excitation) from 39 patients (67 eyes) were acquired with a confocal scanning laser ophthalmoscope equipped with an internal fluorescent reference and were quantified as previously described. In addition, horizontal SD-OCT images through the fovea were obtained. Patients were screened for ABCA4 and PRPH2/RDS mutations.

Results.: ABCA4 mutations were identified in 19 patients (mean age, 37 ± 12 years) and PRPH2/RDS mutations in 8 patients (mean age, 48 ± 13 years); no known ABCA4 or PRPH2/RDS mutations were found in 12 patients (mean age, 48 ± 9 years). Differentiation of the groups using phenotypic SD-OCT and AF features (e.g., peripapillary sparing, foveal sparing) was not reliable. However, patients with ABCA4 mutations could be discriminated reasonably well from other patients when qAF values were corrected for age and race. In general, ABCA4 patients had higher qAF values than PRPH2/RDS patients, while most patients without mutations in PRPH2/RDS or ABCA4 had qAF levels within the normal range.

Conclusions.: The high qAF levels of ABCA4-positive patients are a hallmark of ABCA4-related disease. The reason for high qAF among many PRPH2/RDS-positive patients is not known; higher RPE lipofuscin accumulation may be a primary or secondary effect of the PRPH2/RDS mutation.

Recessive Stargardt disease (STGD1), caused by mutations in the gene ABCA4,¹ follows an autosomal recessive mode of inheritance and usually manifests in the first two decades of life,² although late-onset forms of STGD1 also exist.³ Typical phenotypic features of STGD1 include intensely fluorescent foci (flecks) that emerge with disease progression, mottling of the retinal pigment epithelium (RPE), and central atrophy.^4,5 These fundus changes can be confined to the macula or extend beyond the vascular arcades. Another fundus feature that is considered pathognomonic for ABCA4-related disease is relative sparing of the peripapillary area from flecks and atrophy.^6,7 However, peripapillary sparing does not necessarily persist with ABCA4-related disease progression.⁸ Foveal atrophy is common in STGD1 patients, even at young ages, although in some patients the fovea is spared.⁹ On fluorescein angiography, STGD1 patients also often, but not always, exhibit a “dark choroid.”¹⁰ 

Despite the frequency with which STGD1 can be defined by these features, the phenotype is not necessarily exclusive to STGD1. For instance, peripapillary sparing can also be present in cases of pattern dystrophy (PD) without ABCA4 mutations,¹¹ and both STGD1 and PD can present with pigmentary changes identified as flecks and mottling.¹² Pattern dystrophy refers to a genetically heterogeneous group of retinal disorders, with mutations in PRPH2/RDS being the most common cause.⁴ Pattern dystrophy is usually inherited in an autosomal dominant manner and usually manifests in midlife with mild-to-moderate vision loss.^13,14 In PD with an absence of ABCA4 mutations, foveal atrophy is generally found only in late stages of the disease.¹³ 

Histopathological studies, spectrophotometric analysis, and studies in mouse models have demonstrated over many years that the formation of RPE cell lipofuscin is augmented and pathogenic in ABCA4-associated disease.^15–18 Retinal pigment epithelium lipofuscin, a mixture of fluorophores with spectral features that reflect its bisretinoid constituents,¹⁹ is the primary source of fundus AF, with a minor contribution from Bruch's membrane.²⁰ We recently demonstrated that the increase in RPE lipofuscin in STGD1 can be measured indirectly by quantifying fundus autofluorescence intensities in images acquired by confocal scanning laser ophthalmoscope (cSLO).^21,22 Quantitative fundus autofluorescence (qAF) was also shown to be a valuable tool for distinguishing ABCA4-related from non-ABCA4-related bull's-eye maculopathy (BEM).²³ 

Here we present a patient cohort that, because they presented with clinical features of STGD1, were referred to our medical center for ABCA4 genetic screening. In all patients the clinical evaluation included qAF imaging and analysis. We report that some of these patients were ultimately found to carry mutations in PRPH2/RDS. The inheritance pattern of PD can often not be readily discerned because incomplete penetrance²⁴ and variable expression can mask dominant inheritance. Whether PRPH2/RDS patients exhibit increased lipofuscin levels has been a matter of debate.^4,25 Thus our objective was to consider whether qAF can assist in distinguishing ABCA4-related disease from disease caused by mutations in PRPH2/RDS and other unknown genes. Given the phenotypic overlap between some cases of ABCA4- and PRPH2/RDS-associated disease,^4,5 we also assessed AF and spectral-domain optical coherence tomography (SD-OCT) images qualitatively to attempt to identify characteristic fundus features that may aid in the differentiation of these diseases. PRPH2/RDS-associated PD is generally considered to have a better prognosis than ABCA4-associated disease; thus it is important to differentiate the two conditions clinically. 

Methods

Patients and Genetic Testing

The cohort consisted of 39 patients (67 eyes) from 35 families. All patients were prospectively recruited at Columbia University, were examined by a retina specialist (SHT), and had clear media except for some floaters. Cases included those for which STGD1 could be considered as a potential diagnosis, with the final diagnosis depending on the results of genetic testing. Recruitment to the cohort was based on the presence of central atrophy, mottling, and/or macular flecks in fundus AF (488-nm excitation) images. Patients included in this cohort were (1) those presenting with fundus features that could be observed in both confirmed ABCA4-associated disease and in patients considered to carry the PD phenotype (central lesion with jagged border, butterfly-shaped lesion in the macula); (2) patients with a phenotype having some features that were atypical for ABCA4-associated STGD1 (no peripapillary sparing; atypical size, shape, and distribution of flecks); and (3) patients in whom ABCA4 mutations were not detected despite ABCA4-like fundus features but in whom additional genetic screening revealed PRPH2/RDS mutations. The study cohort was compared to 374 healthy eyes of 277 subjects with no family history of inherited retinal dystrophies²⁶ and 36 of 42 previously reported patients²⁷ (57 eyes) with confirmed ABCA4 mutations that did not have a PD-like phenotype. Visual acuity data, recorded as logMAR, was obtained using the recently obtained refractive correction and a Snellen chart. 

Screening with the ABCA4 array was performed on most study subjects followed by direct Sanger sequencing to confirm identified changes, as previously described.²⁸ Since recruitment proceeded over many years, versions of the ABCA4 chip included the least representative (∼300 mutations) to the most recent version of the array (>600 variants). When array screening identified only one mutated ABCA4 allele or no ABCA4 mutations, next-generation sequencing (NGS) was carried out. In the latter case, the 50 exons and exon–intron boundaries of the ABCA4 gene were amplified (Illumina TruSeq Custom Amplicon protocol; Illumina, San Diego, CA, USA), and then submitted to NGS on the Illumina MiSeq platform with analysis using the variant discovery software NextGENe (SoftGenetics LLC, State College, PA, USA) and reference genome GRCh37/hg19. Variants were confirmed by Sanger sequencing and analyzed with Alamut software (http://www.interactive-biosoftware.com [in the public domain]). To assess cosegregation of the new variants with disease status, family members were studied if available. 

Genetic analyses were originally performed according to the following scheme. Patients whose phenotypes were more likely compatible with ABCA4-associated disease were first screened for variants in the ABCA4 gene. If two or more ABCA4 variants were identified that segregated with the disease, the case was considered an ABCA4-associated disease. All patients in whom complete sequencing of the gene identified only one or no disease-associated ABCA4 alleles were sequenced for variants in the PRPH2/RDS gene. Conversely, some patients presenting with likely PRPH2/RDS-associated phenotypes were first sequenced for this gene, and if a causal mutation was identified, the case was considered genetically solved and ABCA4 was not sequenced. 

This scheme was followed because the statistical probability that an individual carries mutations in both the ABCA4 and PRPH2/RDS genes is very low. Since ABCA4 is located on chromosome 1 and PRPH2/RDS on chromosome 6, variants in the two genes are inherited independently. While the exact population frequency of PRPH2/RDS-associated disease is not known, it comprises between 10% to 20% of all cases of autosomal dominant PD.^29,30 If we consider that the population frequency of autosomal dominant PD is 1:50,000 (which is likely an overestimate), then no more than 1 in 250,000 people are affected by autosomal dominant PD due to PRPH2/RDS mutations. The carrier frequency of disease-associated ABCA4 alleles is estimated to be 1:20; therefore only 1 in 5 million patients with PRPH2/RDS-associated disease would also carry an ABCA4 allele. Despite the extremely low probability of finding disease-causing mutations in both genes in one individual, we sequenced both genes in all patients and confirmed that no patients carried disease-causing mutations in both genes (Table). 

Table

 

Summary of Demographic, Clinical, and Genetic Data

 

All procedures adhered to the tenets of the Declaration of Helsinki, and written informed consent was obtained from all subjects after a full explanation of the study procedures had been provided. The study was approved by the Institutional Review Board of Columbia University and complied with the Health Insurance Portability and Accountability Act of 1996. 

Electrophysiology

A Diagnosys Espion Electrophysiology System (Diagnosys LLC, Littleton, MA, USA) was used to perform electroretinography (ERG) according to the International Society for Clinical Electrophysiology of Vision standards.³¹ For all recordings, tropicamide (1%) and phenylephrine hydrochloride (2.5%) were used for mydriasis before full-field ERG recording, and a drop of proparacaine (0.5%) was used to anesthetize the corneas. Full-field ERGs were performed using silver-impregnated fiber electrodes (DTL; Diagnosys LLC) with a ground electrode on the forehead. 

Image Acquisition

Autofluorescence images (30° × 30° field, 768 × 768 pixels; 488-nm excitation; barrier filter 500–680 nm; beam power < 260 μW) were obtained with a scanning laser ophthalmoscope (cSLO; Spectralis HRA+OCT; Heidelberg Engineering, Heidelberg, Germany) modified by incorporation of a fluorescent reference internal to the imaging device such that the reference is displayed in the AF image. As discussed below, evaluating gray levels (GL) in the AF image relative to the GL of the reference accounted for variable laser power and detector gain.²¹ Previously published protocols were followed to ensure the acquisition of high-quality AF images that permit meaningful quantification.^21,22,26,32 Briefly, pupils were dilated to at least 7 mm (1% tropicamide and 2.5% phenylephrine); rhodopsin was bleached in AF mode for 20 to 30 seconds²¹; detector sensitivity was adjusted so that the GLs were within the linear range of the detector (GL < 175)²¹; and focus and alignment of the camera were adjusted to produce a uniform fundus signal of maximal intensity. Two or more images (each of 9 frames; video format) were recorded in high-speed mode (8.9 frames/s), followed by a second imaging session for estimating the reproducibility. All videos were reviewed, and frames without localized or generalized decreased AF signal²¹ were aligned, averaged, and saved in “nonnormalized” mode (two images per session). The data presented in this study were based on AF images of 67 eyes, with 44 of these eyes having a second AF imaging session; an additional 13 imaging sessions were excluded because of insufficient image quality. Autofluorescence images from all eyes included in this study are presented in the Supplementary Material. 

A horizontal 9-mm SD-OCT image through the fovea, registered to a simultaneously acquired AF or near-infrared reflectance (NIR-R) image, was recorded in high-resolution mode as an average of 100 individual images for each eye. The optical depth resolution in the Spectralis is currently ∼7 μm. In cases in which the SD-OCT image had been registered to an NIR-R image, i2kRetina software (DualAlign LLC, Clifton Park, NY, USA) was used to align the AF image of the same field to the NIR-R image. Assignment of retinal layers was based on published nomenclature.³³ 

AF Image Analysis

As formerly reported,^22,23 AF images were analyzed under the control of an experienced operator with dedicated image analysis software written in IGOR (WaveMetrics, Lake Oswego, OR, USA) to determine qAF.²¹ Briefly, the software recorded the mean GLs of the internal reference and of eight circularly arranged segments positioned at an eccentricity of approximately 7° to 9° (Fig. 1) that were scaled to the horizontal distance between the fovea and the temporal edge of the optic disc. Quantitative AF was calculated from the mean GLs of each segment taking into account the Reference Calibration Factor, the internal reference GL, zero GL, magnification, and optical media density from normative data on lens transmission spectra.^21,34 For each image, the average qAF from the eight segments (qAF₈)²¹ was generated. 

Figure 1

 

Quantitative fundus autofluorescence image analysis. P36. Mean gray levels (GLs) are recorded from the internal reference (white rectangle, top of image) and from 8 circularly arranged segments (black). The segments are scaled to the distance between the temporal edge of the optic disc (white vertical line) and the center of the fovea (white cross). The average of qAF values of the 8 segments is defined as qAF₈.

Statistical Analyses

Analyses were performed using Prism 5 (GraphPad Software, La Jolla, CA, USA) and Stata (version 12.1; College Station, TX, USA). Where appropriate, mixed-effects models were used to account for data from two eyes and familial and twin relationships. When comparing qAF between groups, age and race were included in the model. To evaluate between-session repeatability of the measurements and coefficient of agreement between right and left eyes, the method of Bland and Altman³⁵ was utilized. For each qAF₈ value, z-scores were also calculated and plotted as a measure of the numbers of standard deviations that describe each value relative (negative and positive) to the mean of healthy eyes. 

Results

ABCA4 mutations were detected in 19 patients (31 eyes) from 18 families, including 11 cases (58%) with two disease-causing ABCA4 variants. Of those patients, 17 were white and 2 were Indian. Mutations in PRPH2/RDS were identified in 8 patients (16 eyes) from 5 families, all of whom were white. No mutations in ABCA4 or PRPH2/RDS were found in 12 patients (20 eyes). Of those patients, 11 were white and 1 was black. Demographic, clinical, and genetic information of the patients is presented in the Table. 

The group with ABCA4-associated disease (mean age, 37 years; range, 20–60 years) was younger than both the PRPH2/RDS-positive group (mean age, 48; range, 30–62 years; mixed-effects regression, P = 0.03) and patients without mutations in ABCA4 or PRPH2/RDS (mean age, 48 years; range, 28–58 years; P = 0.01). There was no difference in best-corrected visual acuity among the three groups of patients (mixed-effects regression, P > 0.19). Electroretinogram results (digital scans or clinical interpretations) were available for 30 out of 39 (30/39) patients: 20 patients (12 ABCA4 positive, 3 PRPH2/RDS positive, 5 ABCA4 and PRPH2/RDS negative) exhibited focal disease with a normal full-field ERG; 9 patients (4 ABCA4 positive, 2 PRPH2/RDS positive, 3 ABCA4 and PRPH2/RDS negative) had generalized cone ERG abnormalities, and 1 patient (1 ABCA4 positive) exhibited generalized cone and rod ERG abnormalities. Mean scotopic and photopic values and latency are summarized in Supplementary Table S1. 

Representative AF images from patients with ABCA4-associated disease are shown in Figure 2. These patients either had a central lesion with a jagged border and no flecks in the periphery, a phenotype that can be found in both ABCA4-associated and non-ABCA4-associated disease, or had fundus features that may be considered atypical for STGD1, for instance, foveal sparing or peripapillary involvement. 

Figure 2

 

Representative fundus AF images of ABCA4-positive patients P32 (A), P31 (B), P25 (C), P38 (D), P14 (E), P9 (F), P27 (G), and P39 (H). Fundus features include foveal atrophy with adjacent flecks resembling a butterfly-shaped lesion (P32); subtle foveal disturbance (P31); a ring of flecks surrounding a spared fovea (P25, P38); mottled atrophy in the central macula surrounded by flecks (P14), which extend beyond the vascular arcades in the case of P9; a large atrophic lesion surrounded by flecks with foveal sparing and no distinct peripapillary sparing (P27); and numerous macular flecks in the absence of atrophy (P39).

Figure 3 illustrates the phenotypic range within AF images of PRPH2/RDS-positive PD patients. Prior to genetic screening these patients had been misdiagnosed as STGD1. The AF image of patient (P)35 appears qualitatively normal, while in the image of the father (P34), a central area of atrophy and flecks are present. Patient 11 and P8 also have a central area of atrophy, which spares the fovea. There is a similar radial arrangement of flecks visible around the fovea in P20, P12, P4, and P13. In P12, the central macular region, including the peripapillary region, was less affected than peripheral rod-dominated fundus areas. These topographical differences are further illustrated in Figure 4, where an AF fundus montage of the other eye of P12 is shown. 

Figure 3

 

Fundus AF images of PRPH2/RDS-positive patients P35 (A), P20 (B), P12 (C), P4 (D), P34 (E), P11 (F), P13 (G), and P8 (H). As in ABCA4 disease, fundus AF images from PRPH2/RDS-positive patients can present with few or numerous flecks exhibiting a high or, once resorbed, low AF signal. In some atrophic lesions, the fovea is spared (P11, P8), in others not (P34).

Figure 4

 

Montage from multiple fundus AF images of P12, illustrating the distribution of AF changes across the fundus.

In Figure 5, examples of AF images from patients without known ABCA4 or PRPH2/RDS mutations are shown. For some of these patients, there is a remarkable phenotypic similarity to the ABCA4-positve cases presented in Figure 2. 

Figure 5

 

Fundus AF images of patients without ABCA4 or PRPH2/RDS mutations: P33 (A), P6 (B), P5 (C), P29 (D), P7 (E), P22 (F), P37 (G), and P17 (H).

Relative sparing of the peripapillary region in terms of flecks and atrophy is often regarded as a characteristic feature of STGD1 disease.⁶ However, as illustrated in Figure 6, within our cohort some ABCA4-positive cases (upper row) did not have complete peripapillary sparing, as these patients had flecks in close proximity to the optic nerve. Conversely, some PRPH2/RDS-positive patients (lower row) presented with “classic” peripapillary sparing. Foveal sparing, another fundus feature that could potentially help to discriminate patients with ABCA4-associated disease from patients presenting with PD due to mutations in other genes, was also observed in a subgroup of patients irrespective of whether they were PRPH2/RDS positive or ABCA4 positive or patients without mutations in either ABCA4 or PRPH2/RDS. 

Figure 6

 

Assessment of peripapillary sparing. ABCA4-positive patients (A–D) with irregular flecks (white arrows) in close proximity to the optic nerve. PRPH2/RDS-positive patients (E–H) with “classic” peripapillary sparing.

Quantitative Fundus AF

For qAF₈ of right and left eyes (n = 28), the Bland-Altman coefficient of agreement was 16.4%. The between-session Bland-Altman coefficient of repeatability was 9.3% (44 eyes of 26 patients). 

As shown in Figure 7A, the group of patients with ABCA4-associated disease had higher qAF₈ than those with healthy eyes (mixed-effects regression, P < 0.001), with 16/19 patients (26/31 eyes) having higher qAF₈ than the 95% confidence interval for the 374 healthy eyes (adjusted for age and race). As shown in Figure 7B, the group of patients with ABCA4-associated disease had, however, lower qAF₈ than other previously published²⁷ ABCA4-positive patients with non-PD-like phenotypes (P = 0.002). 

Figure 7

 

Quantitative fundus autofluorescence. (A) qAF₈ of ABCA4-positive (ABCA4 +) patients (red circles with filled square in center) for whom PD may be considered as a potential diagnosis, PRPH2/RDS-positive (PRPH2/RDS +) PD patients (green squares), and PD patients without mutations in ABCA4 or PRPH2/RDS (ABCA4 − PRPH2/RDS −; blue triangles) plotted as a function of age. Values from both eyes of a patient are plotted when both images were available. (B) qAF₈ values of ABCA4-positive (ABCA4 +) patients from this study (red circles with filled square in center) plotted with qAF₈ values of other ABCA4-positive (ABCA4 +) previously reported²⁷ patients that did not have a PD-like phenotype (red diamonds) as a function of age. Each value is mean of left eye and right eye when images from both eyes were available; otherwise the value is from a single eye. Mean (solid line) and 95% confidence intervals (dashed lines) are shown for white, healthy eyes (n = 125; small gray filled triangles).

As shown in Figure 7A, the PRPH2/RDS-positive PD group had higher qAF₈ than those with healthy eyes (P < 0.001), with 4/8 patients (9/16 eyes) having higher qAF₈ than the 95% confidence interval for the healthy eyes. The increases in RPE autofluorescence measured as elevated qAF in some PRPH2/RDS-positive patients are unlikely to be due to a “window defect”³⁶ resulting from photoreceptor cell degeneration since visual pigment was bleached before imaging. The PRPH2/RDS-positive PD group had lower qAF₈ than the group of patients with ABCA4-associated disease (P = 0.016). The group of PD patients without mutations in ABCA4 or PRPH2/RDS had qAF₈ that was not different from the normal range for age (P = 0.69), with 3/12 patients (5/20 eyes) having high qAF₈ and 3/20 eyes having qAF₈ below the 95% confidence interval for healthy eyes (Fig. 7A). 

These differences in qAF levels between patients are further demonstrated in Figure 8. In the upper row, AF images from a healthy subject and three age-similar patients (age: ∼50 years) with different genotypes are shown. In the lower row, the corresponding color-coded qAF maps are presented. While the healthy subject and the PD patient without mutations in ABCA4 or PRPH2/RDS (P21) have similar qAF levels, qAF levels are clearly elevated in the PRPH2/RDS-positive patient (P12) and most dramatically increased in the ABCA4-positive patient (P26). 

Figure 8

 

Color-coded qAF images. AF images of a healthy subject (A) and three age-similar patients (B–D) with different genotypes are shown. Patient 21 (B) has no mutations in ABCA4 or PRPH2/RDS; P12 (C) has a PRPH2/RDS mutation; and P26 (D) has a mutation in ABCA4. In the corresponding color-coded qAF maps (E–H), colors have been assigned to each pixel based on its qAF value. Lower qAF levels are indicated as blue and higher qAF levels as orange colors (see color code scale below).

Of interest is whether it is possible to use qAF to discriminate patients with ABCA4-associated disease from patients presenting with PD due to mutations in other genes. Initially, we approached this question using raw qAF values, but the result gave poor discrimination (d' = 1.09, sensitivity = 0.94, specificity = 0.33 at a criterion of qAF = 355). Next, we investigated the ability to discriminate patients with ABCA4-associated disease using qAF values that were adjusted for age and race. To do that, we determined the z-score based on the fit of the healthy eye data (information provided in Table 2 of Greenberg et al.²⁶). The z-score of a given value indicates how many standard deviations above or below the mean that value is.³⁷ The qAF z-score provided better discrimination (d' = 1.52, sensitivity = 0.71, specificity = 0.83 at a criterion of qAF = 2.5 standard deviations) than the raw qAF (Fig. 9). Note that since d' combines sensitivity and specificity into a single metric that describes³⁸ the ability to distinguish the signal (i.e., presence of ABCA4 mutation) from noise (absence of ABCA4 mutation), a z-score criterion of 2.5 indicates the value above which the qAF z-score is likely to be indicative of ABCA4-associated disease. 

Figure 9

 

qAF z-scores of patients in this study corrected for race and age. ABCA4-positive (ABCA4 +) patients (red circles with filled square in center), PRPH2/RDS-positive (PRPH2/RDS +) PD patients (green squares), and PD patients without mutations in ABCA4 or PRPH2/RDS (ABCA4 − PRPH2/RDS −; blue triangles). The z-score is based on healthy eyes (small open gray circles). Values for both eyes of each patient are plotted. A z-score equal to 0 represents a value equal to the mean. A z-score of 1 is 1 standard deviation greater than the mean.

Spectral-Domain Optical Coherence Tomography

As illustrated in Figure 10, there was a wide range of changes observed by SD-OCT within the group of PRPH2/RDS-positive patients (nomenclature of anatomic landmarks in SD-OCT images was based on the proposed lexicon by Straurenghi et al.³³). Patient 35 did not have any obvious fundus changes. In P20, flecks exhibiting a high AF signal were present in the macula. These flecks traversed the photoreceptor ellipsoid zone (EZ) and external limiting membrane (ELM), and one fleck even extended into the outer nuclear layer (ONL). The choroidal signal under the flecks was attenuated. Patient 13 and P8 both had significant outer retinal loss temporally, while outer retinal damage was less pronounced nasally and at the fovea. Increased choroidal reflectivity, a feature that is generally attributed to RPE atrophy, was also more distinct temporally than nasally in P13 and P8. Patient 8 had parafoveal cystic-like spaces in the inner retina. Patient 11 had the most pronounced fundus changes. Except for a spared foveal island, the choroid was severely thinned and RPE and outer retina were atrophic. The inner retina seemed mostly intact and had prolapsed into areas of outer retina and RPE loss, leaving the impression of optical empty spaces bordering areas of preserved outer retina and RPE. 

Figure 10

 

Spectral-domain optical coherence tomography (SD-OCT) images of PRPH2/RDS-positive patients. AF images (left) and corresponding SD-OCT images (right) of P35 (A), P20 (B), P13 (C), P8 (D), and P11 (E). Horizontal axis and extent of SD-OCT image indicated by white arrow in AF image. Outer nuclear layer (ONL), external limiting membrane (ELM), photoreceptor ellipsoid zone (EZ), interdigitation zone (IZ), and retinal pigment epithelium/Bruch's membrane (RPE).

In Figure 11, examples of SD-OCT images from ABCA4-positive patients are presented. The retina of P39 appeared intact except for the presence of some flecks. In P18 the flecks were located closer to the fovea, enclosing an area where ELM and EZ appeared pronounced. In P2, flecks disrupted the EZ at the fovea, and hyperreflective dots were visible at the level of the ELM and ONL. In P23 and P24, there was central EZ loss in tandem with increased choroidal reflectance. Patient 23 also had hyperreflective dots at the level of ONL and ELM, similar to P2. In P2, P23, and P24, the ONL appeared thinned centrally. 

Figure 11

 

Spectral-domain optical coherence tomography (SD-OCT) images of ABCA4-positive patients with a PD phenotype. AF images (left) and corresponding SD-OCT images (right) of P39 (A), P18 (B), P2 (C), P23 (D), and P24 (E). Horizontal axis and extent of SD-OCT image indicated by white arrow in AF image.

Discussion

In some patients, differentiating ABCA4-associated disease from retinal phenotypes stemming from mutations in other genes such as PRPH2/RDS can be challenging. This difficulty is related to phenotypic similarities on funduscopy and the lack of diagnostic tests that can be used for better differentiation. For example, the full-field ERG in PRPH2/RDS-associated disease is often normal,^39–41 but can also show profound variability even within families carrying the same mutation.³⁰ In ABCA4-associated disease, scotopic and photopic responses can also be normal.^42,43 

In the present study, we assessed whether qAF, an indirect in vivo measure of RPE lipofuscin, could aid in differentiating these patients. Our data showed, as expected, that ABCA4-positive patients had the highest qAF₈ levels within the cohort. However, it is notable that PRPH2/RDS-positive patients also had significantly increased values. Patients without mutations in ABCA4 and PRPH2/RDS had the lowest qAF₈ values within the cohort, and in most cases qAF₈ levels were either within or below the normal range for age. It was possible to discriminate patients with ABCA4-related disease from PD patients without ABCA4 mutations with an accuracy of 78% in our sample of 67 eyes using the age- and race-adjusted qAF z-score. 

In STGD1, the mechanism for accentuated RPE lipofuscin formation is known. Insufficient ABCA4 protein activity leads to a buildup of bisretinoids in the outer segment disc membranes; this material includes an adduct of glycerophosphoethanolamine and two vitamin A molecules (A2-GPE), all-trans-retinal dimer and A2E as just some of the components.¹⁹ These bisretinoids are transferred to the RPE by phagocytosis where they form the lipofuscin of these cells. It is likely that the process of outer segment membrane shedding and phagocytosis by RPE serves to protect photoreceptor cells. If allowed to accumulate in outer segments, these bisretinoid fluorophores would trap incoming photons and compete with visual pigment photoisomerization. Importantly, bisretinoids are also photosensitizers that initiate photooxidation reactions.¹⁹ Yet because photoreceptor cells are specialized for phototransduction, the outer segments are not equipped with the levels of reducing equivalents and antioxidant defense systems expressed by RPE.^44–46 

But what is the basis for an increase of RPE lipofuscin in the presence of PRPH2/RDS mutations? PRPH2/RDS is a photoreceptor-specific glycoprotein that is essential for the construction and preservation of the outer segment (OS) rim region of rods and cones. PRPH2/RDS functions by assembling into protein complexes that include rod OS membrane protein-1 (ROM-1). In the inner segment, PRPH2/RDS and ROM-1 form tetramers that are held together by noncovalent interactions between the second intradiscal (D2) loop of the two proteins.^47–49 These complexes are then trafficked to the OS where they form higher-order oligomeric structures, both hetero- and homoligomers⁵⁰ that interact by means of disulfide bonds.^50,51 Mouse models have revealed that rods and cones have different constraints with respect to PRPH2/RDS. In the absence of PRPH2/RDS, rods fail to form OS⁵² while cones form open OS that lack rims and flattened membranous lamellae but are still capable of mediating function.⁵³ Nonetheless, why some PRPH2/RDS mutations result in rod-dominant retinal disease (such as autosomal dominant RP) while others, such as amino acid substitutions at position 172 in the D2 loop (R172W, R172Q), are associated with a cone-associated phenotype (autosomal dominant macular dystrophy and cone–rod dystrophy)²⁴ has so far eluded explanation. One issue is that the architecture of OS in cones and rods is different: In rods the stack of OS discs is enclosed and separated from the plasma membrane, while in cones the membrane evaginations that form the OS do not become surrounded by plasma membrane and instead are exposed to extracellular space. Some investigations indicate that haploinsufficency may underlie rod-dominant disease while gain-of-function defects lead to cone-related disease.⁵⁴ Even then, however, it is not understood how photoreceptor defects associated with some PRPH2/RDS mutations translate into secondary disease features such as RPE atrophy.⁵⁵ Whether loss of PRPH2/RDS could affect retinoid handling in OS as part of the primary disease process, or whether retinaldehyde adducts form secondarily due to photoreceptor cell dysfunctioning and degeneration, remains to be investigated. Lipofuscin formation as secondary effect could explain why PRPH2/RDS-positive patients exhibit fundus changes similar to STGD1 patients (high qAF and geographic atrophy) but later in life than STGD1 patients. The damaging effects of RPE lipofuscin would be expected to be the same, but the increase in lipofuscin could generally occur less rapidly. 

We observed other phenotypic features in PRPH2/RDS-positive patients that deserve further discussion. It was our impression that flecks in PRPH2/RDS-positive patients appeared to be smaller and nonconfluent (e.g., P34; Fig. 3) while those in STGD1 often exhibited irregular profiles with adjacent flecks appearing to become contiguous. This may be an observation deserving further study. Additionally, at least in some PRPH2/RDS-positive patients, topographical differences in the disease were evident. In the SD-OCT images of P13 and P8 (Fig. 9), there were more disease-related changes present temporally than nasally. Also, in P12, P4, and P13 (Fig. 3) there were more fundus changes in the periphery and relatively less disease-related changes in the macular area, including the peripapillary area. It is interesting to consider that some PRPH2/RDS mutations are more specific to rods versus cones. 

We observed that ABCA4-positive patients for whom PD may be considered as a differential diagnosis tended to have lower qAF than previously reported patients²⁷ with non-PD-like STGD1 phenotypes and also tended to be more represented at older ages (Fig. 7B). ABCA4-positive patients exhibiting a BEM phenotype were also reported to have lower qAF than ABCA4-positve patients with other phenotypes, including those with extramacular disease.²³ In the case of ABCA4-positive BEM, many of the patients carried the G1961E mutation, which was also the case for 9/19 of the ABCA4-positive patients in this study. In future studies we will examine for further phenotype–genotype correlations. 

Color-coded qAF maps allow a rapid assessment of qAF levels and may therefore be valuable in the clinical assessment and follow-up of patients. However, since qAF also increases with age in healthy subjects, current comparisons of qAF maps can be made only between age-similar subjects. To compare patients regardless of age, it may be useful to develop color-coded qAF maps based on fold increases relative to healthy subjects of the same race and age. This could be done using the z-score approach that we used to discriminate patients with ABCA4-associated disease from patients presenting with PD due to mutations in other genes. The z-score approach furthermore allows for a comparison of qAF levels among patients regardless of their race. 

As with any novel imaging technique, the clinical utility of qAF awaits further study. Nevertheless, the data presented here provide evidence that qAF can be used to establish genotype–phenotype correlations and can help to better understand the underlying pathomechansims of retinal dystrophies. Quantitative AF can expose phenotypic variation among patients even when conventional imaging fails to detect differences. For the workup of retinal dystrophy patients, qAF could prove especially valuable at the initial visit of a patient. Judging the likelihood of ABCA4-associated disease based on qAF levels will likely result in less unnecessary sequencing of non–disease-causing genes. Ongoing longitudinal qAF studies will clarify whether high qAF levels are also correlated with fast disease progression. When treatment options for STGD1 and PD become available (e.g., gene and small molecule),^56–59 qAF can help to monitor treatment endpoints and should be considered as an outcome measure in clinical trials. 

Acknowledgments

Supported in part by grants from the National Eye Institute/National Institutes of Health EY024091, EY021163, EY019007; Foundation Fighting Blindness; and a grant from Research to Prevent Blindness to the Department of Ophthalmology, Columbia University. The authors alone are responsible for the content and writing of the paper. 

Disclosure: T. Duncker, None; S.H. Tsang, None; R.L. Woods, None; W. Lee, None; J. Zernant, None; R. Allikmets, None; F.C. Delori, None; J.R. Sparrow, None