High Resolution Adaptive Optics In Vivo Autofluorescence Imaging in Stargardt Disease

Abstract

IMPORTANCE:

Targeting the early pathogenic steps in Stargardt disease (STGD1), is critical to advance our understanding of this condition and to develop potential therapies. A hypothesis is that lipofuscin precursors accumulate within photoreceptors, leading to photoreceptor damage and preceding RPE cell death. Fluorescence adaptive optics scanning light ophthalmoscopy (FAOSLO) can provide autofluorescence (AF) images in vivo with microscopic resolution to elucidate the cellular origin of AF abnormalities in STGD1.

OBJECTIVE:

To study the spatial distribution of photoreceptor, RPE, and AF abnormalities in STGD1 patients at a cellular level.

DESIGN:

Observational study using FAOSLO to compare the cones, rods, and RPE cells between STGD1 patients and normal control.

SETTING:

Imaging sessions were conducted at the University of Rochester. Further image analyses was performed at Beijing Tongren Eye Center and the University of Pittsburgh.

PARTICIPANTS:

Two women and one man, ages 21–38, with the macular atrophy phenotype of STGD1 and visual acuity loss ranging from 20/30 to 20/150; and one unaffected woman, age 34, with 20/20 visual acuity.

MAIN OUTCOMES AND MEASURES:

Structural appearance of cones, rods and autofluorescence structures at different retinal locations.

RESULTS:

Cone and rod spacing was increased in all patients at all locations where photoreceptors were detectable; most cones had a dark appearance. Autofluorescence (AF) was low contrast but contained structures consistent with RPE cells in the periphery. In the transition zone peripheral to the foveal atrophic lesion, the structural pattern of AF was more consistent with photoreceptors than RPE cells. The microscopic AF was disrupted within areas of clinically detectable atrophy.

CONCLUSION and RELEVANCE:

Adaptive optics high resolution images of cones, rods, and RPE cells at the leading disease front of STGD1 macular atrophy shows an AF pattern that appears to colocalize with photoreceptors, or may result from a combination of AF signals from both RPE cells and photoreceptors. This in vivo observation is consistent with histologic reports of fluorescence arising from photoreceptors in STGD1. The detection of bisretinoid accumulation in the photoreceptors may represent an early pathologic step in STGD1 and can provide an in vivo imaging tool to act as a biomarker of disease progression.

Introduction

Stargardt disease (STGD1), the most common form of early onset macular dystrophy, is associated with mutations in the ATP-binding cassette transporter (ABCA4) gene¹ and is characterized by a spectrum of retinal phenotypes ranging from retinal pigment epithelial (RPE) atrophy in the macula to photoreceptor degeneration across the entire retina.² Loss of ABCA4 function is associated with RPE lipofuscin accumulation and photoreceptor degeneration in mouse models.³ Histological investigations have demonstrated increased accumulation of lipofuscin in the RPE and accumulation of lipofuscin precursors within photoreceptors.^4,5 Lipofuscin is a complex aggregate of many different bisretinoids, including A2E.⁶ The immediate A2E precursor, A2-PE,⁷ is generated in photoreceptor outer segments.⁸ The phagocytosis of shed photoreceptor outer segment discs in normally functioning RPE is thought to prevent A2-PE from accumulating within the photoreceptors.⁵ However, A2-PE has been shown to accumulate in the photoreceptors of animals that are unable to phagocytose shed outer segments.⁹ A2-PE is autofluorescent (AF) at short wavelengths and it may account for the lipofuscin-like fluorescence of photoreceptors seen histologically in STGD1.^4,7 A pathogenic sequence of lipofuscin accumulation leading to RPE cell damage and subsequent photoreceptor death has been proposed.¹⁰ An alternative hypothesis is that lipofuscin precursors accumulate within photoreceptors, leading to photoreceptor damage simultaneous to or preceding RPE cell death.

Clinical imaging methods, such as confocal scanning laser ophthalmoscopy (cSLO), have been used to examine the short wavelength fundus autofluorescence (FAF) pattern thought to arise from lipofuscin to assess the status of the RPE in STGD1.¹¹ In macular atrophy, the most frequently observed phenotype, increased autofluorescence (AF) accumulation surrounding the margin of RPE atrophy, designated the leading disease front, is considered to represent the earliest disease change in STGD1.¹² Although FAF in cSLO provides a wide field view of the regional distribution of lipofuscin AF, morphometric analysis of the RPE cell mosaic is not available using currently available commercial FAF imaging systems because they lack the resolution to identify individual cells. Using AF imaging methods in fluorescence adaptive optics scanning light ophthalmoscopy (FAOSLO), single cell resolution of RPE cells has been achieved in both healthy¹³ and diseased eyes^14–16 in the living human retina, recent advances in our group having greatly increased the efficiency of the approach.^17–19 Here we use FAOSLO to characterize the cellular level retinal phenotype of rods, cones and RPE cells in three patients with the macular atrophy phenotype of STGD1.

Methods

This study was designed at beginning of 2015 and approved by the University of Rochester Research Subjects Review Board. Written informed consent was obtained and participants received financial compensation. Three patients with STGD1 were examined at the end of 2015 and in 2016 respectively; an age-matched participant from a previous study of normal eyes served as a control for comparison. A complete ophthalmic examination, including Snellen visual acuity measurement and dilated fundus examination, was performed. Clinical images were obtained, including: color fundus photography (FF450 plus; Carl Zeiss Meditec), conventional FAF in confocal scanning laser ophthalmoscopy and spectral domain optical coherence tomography (SD-OCT) (Spectralis HRA+OCT, Heidelberg Engineering and Cirrus OCT, Carl Zeiss Meditec).

Confocal near-infrared reflectance (λ_max=790 FWHM = 19 nm) and short wavelength auto-fluorescence (excitation = 532 nm; emission = 575–725 nm) images were acquired using a technique described previously with recently demonstrated improvements.^16–17 Photoreceptors were marked semi-automatically using previously established methods.²⁰

Results

Clinical Findings:

All three patients demonstrated the macular atrophy phenotype of STGD1, with central visual acuity loss. Conventional FAF showed central hypo-AF with surrounding hyper-AF, and uniform AF peripherally. SD-OCT confirmed atrophy of the RPE and outer retina centrally, with a surrounding annular transition zone, and normal retinal and RPE layers peripherally. Multimodal clinical imaging results of Patient 2, a 34-year-old woman with 20/30 visual acuity, are shown in Figure 1. Multimodal clinical imaging results of Patient 1, a 38-year-old man with 20/150 visual acuity, and Patient 3, a 21-year-old woman with 20/150 visual acuity, are shown in the supplementary data (eFigure 1).

Figure 1. Multimodal Imaging of Patient 2.

A, Color fundus photograph shows macular atrophy with peripheral flecks. B, Autofluorescence (AF) shows central hypo-AF with surrounding annular hyper-AF and also homogeneous peripheral AF. Yellow lines indicate position of spectral-domain optical coherence tomography (OCT) scans shown in C and D. Blue boxes indicate the locations of the adaptive optics scanning light ophthalmoscopy images shown in Figure 2 and Figure 3. C and D, Spectral-domain OCT shows atrophy of the retinal pigment epithelial and outer retina at the foveal center, with a surrounding transition zone, and normal layers in the periphery. Blue arrowheads indicate locations of adaptive optics scanning light ophthalmoscopy imaging.

AOSLO Imaging

Within the macular atrophic lesion, no clearly discernible photoreceptors were identifiable in confocal reflectance images in the STGD1 patients (Figure 2, eFigure 2). In patients 2 and 3, no continuous RPE cell mosaics were identifiable in FAOSLO images near the fovea; only scattered hypo- and hyper-AF was seen at this retinal eccentricity (Figure 2D, eFigure 2D). In Patient 1, who had the smallest area of macular atrophy in this series, RPE cell-like structures were identified in reflectance images (eFigure 2A). AOSLO AF was abnormal, lacking the characteristic appearance of the RPE mosaic and its structure was more consistent in appearance with photoreceptor reflectance (eFigure 2B).

Figure 2. Adaptive Optics Scanning Light Ophthalmoscopy (AOSLO) Imaging Near the Fovea (0.5 mm Eccentricity).

A and C, Confocal reflectance AOSLO. B and D, Fluorescence AOSLO. Patient 2 shows no detectable photoreceptor or retinal pigment epithelial cells (C and D) compared with a matched location in a normal eye (A and B). The location marked by blue box 1 in Figure 1 corresponds to the AOSLO imaging location here in C and D.

At the transition zone surrounding the macular atrophic lesion, AOSLO showed cones were abnormally dark, enlarged, and sparse (Figure 3, e-Figure 3). Cone spacing (Figure 3A) was increased by 50% compared to normal eyes at 1.8 mm (16.2 μm vs. 10.7 ± 2.9 μm in normals ¹⁷ (mean ± 2SD)). Rod spacing at the same location was increased by ~20% compared to normal eyes (rod spacing = 3.85 μm in patients vs. 3.2 μm in normal), consistent with our previous findings in STGD1 patients.²¹ The AF signal appeared abnormal (Figure 3, e-Figure 3), lacking the characteristic appearance of the normal RPE mosaic, and appearing to co-localize with the photoreceptor reflectance pattern. Video 1 shows an overlay comparison of Figures 3C and D. These structures may be composed of photoreceptor and RPE cell debris; some hyper-AF features in this area may represent lipofuscin laden macrophages.

Figure 3. Adaptive Optics Scanning Light Ophthalmoscopy (AOSLO).

Imaging at the Transition Zone of Patient 2A and C, Reflectance images show abnormally enlarged rods and cones. B and D show abnormal autofluorescence structure appearing more consistent in appearance with photoreceptor reflectance than that of retinal pigment epithelial cells. The location imaged in A and B is indicated by blue box 2 in Figure 1, and Cand D are box3.

In the retinal periphery, photoreceptor structures were more consistent with the normal expected photoreceptor mosaic appearance and spaicing. FAOSLO images exhibited an AF pattern consistent with the RPE cell mosaic. However, RPE cell contrast was lower than in normal eyes and the central hypo-AF region of each cell (thought to derive from the absence of lipofuscin in the cell nucleus)¹³ appeared qualitatively smaller, with the hyper-AF at the cell margin appearing thicker and less distinct.

Discussion:

This study demonstrates the microscopic distribution and accumulation of the short wavelength autofluorescence (SWAF) that underlies the patterns seen in conventional wide field FAF images of STGD1 patients. Near the fovea, we observed some persisting AF in AOSLO in areas corresponding to markedly reduced FAF in cSLO. However, the pattern of AF was not characteristic of the normal RPE cell mosaic and these areas were devoid of the normal photoreceptor pattern seen in reflectance AOSLO, suggestive of photoreceptor loss and consistent with SDOCT findings. In the transition zone outside the clinically apparent macular atrophic lesion, cones were sparse, enlarged, and their reflectance diminished; rods were continuous but with increased spacing, consistent with our previous study.¹⁹ Although near the resolution limit of our imaging system and thus difficult to quantify, the rod spacing increase appeared to be driven by an enlarged cell diameter, similar to the enlarged cones that we have shown previously. ¹⁹

Comparing the photoreceptor layer findings to RPE cell morphometry has yielded additional insight into the temporal sequence of structural abnormalities that accompany cell loss in STGD1. We had previously shown that AOSLO photoreceptor findings were suggestive that photoreceptor loss preceded the clinically detectable RPE loss seen in FAF. Here we can see that despite uniform AF in wide field FAF images, the underlying RPE morphometry can be quite abnormal, demonstrating that cellular level AF imaging with AOSLO can reveal RPE morphometric abnormalities that are not detectable in conventional FAF instruments.

The microscopic structural pattern of AF we observed at the transition zone, or leading disease front, just outside the area of hypo-AF seen in wide field FAF, appeared to correspond to areas where photoreceptors, particularly rods, are present in the confocal reflectance AOSLO images (Figure 3A and B) rather than the characteristic pattern of RPE morphology typically seen. A similar morphology was seen in the area of a hyper-AF ring seen in the widefield FAF image for Patient 1 and in an area of heterogeneous AF in widefield FAF in patient 2 demonstrating that the widefield AF pattern does not uniformly correspond to the abnormalities seen at a microscopic scale. In Patient 1 near the fovea, we can see RPE-like structures in the reflectance images similar to a previous report by Roorda et al²⁰ in cone-rod dystrophy patients; however in the AF image at the same retinal location, normal RPE structures are not apparent but rather hyper-AF signal, suggesting the RPE cells are not healthy RPE cells but cells filled with lipofuscin-like AF material. The AF pattern seen in AOSLO suggests that the AF signal may arise either from the photoreceptors or represent a combination of AF signal arising from both RPE cells and photoreceptors. A loss of ABCA4 function causing the failure of transport of retinoids from the photoreceptors to the RPE would explain our findings that are suggestive of AF signal arising from photoreceptors. This finding is consistent with previous histological evidence of lipofuscin-like AF in photoreceptors⁴ and represents evidence of this in vivo.

In the retinal periphery, a more typical pattern of RPE cell morphometry was seen with AOSLO suggesting the RPE cells are present and relatively normal at this eccentricity. However, there were some differences in the appearance of the RPE cells in these areas: there was less contrast from the central hypo-AF region of the cell and cell borders were indistinct. These subtle morphological differences may relate to an RPE phenotype that is indicative of early RPE dysfunction. The reduced contrast of RPE cells in these areas may also be due to photoreceptor AF, but at a reduced level relative to that seen at more central locations where the disease process is more advanced. However, clear photoreceptor disruption was observed at these retinal eccentricities suggesting that the photoreceptor disruption may occur earlier, or be easier to detect than subtle changes in RPE AF.

Limitations of this study include the small number of STGD1 patients imaged, and the cross-sectional study design. Imaging of additional patients to confirm these findings, and longitudinal evaluations with FAOSLO may further elucidate the spatiotemporal dynamics and the sequence of cell death in STGD1.

Our results provide valuable information to identify viable cells, which will play a central role in the selection of patients amenable to gene replacement or other emerging therapies.^21–23 Treatments targeting photoreceptor preservation or the reduction of the accumulation of bisretinoids such as A2-PE in photoreceptors should be considered for the treatment of STGD1.

Conclusion:

AOSLO can be used to image both photoreceptors and RPE cells in STGD1 patients. High resolution AF imaging with AOSLO provided insight into the microscopic pattern of AF that underlies the morphologies seen in widefield FAF. We show here that uniform hyper-AF and heterogenous AF in wide field FAF images can underlie similar patterns of FAF at a microscopic scale that are suggestive of AF arising from both photoreceptors and RPE cells. In the peripheral retina, cone and rod photoreceptors are disrupted while RPE cells still appear relatively normal suggesting photoreceptor damage mediated by the accumulation of lipofuscin precursors such as A2-PE may precede RPE cell death in STGD1.

Figure 4. Adaptive Optics Scanning Light Ophthalmoscopy (AOSLO).

Imaging in the Periphery A, Reflectance images show photoreceptor structures more consistent with normal photoreceptors. B, Autofluorescence images exhibited a pattern consistent with the retinal pigment epithelial cell mosaic. Adaptive optics scanning light ophthalmoscopy image location is denoted by the blue box 4 shown in Figure 1.

Key Points.

Question:

What structures are the sources of the autofluorescence (AF) signals in Stargardt disease (STGD1)?

Findings:

Here we provide an observational study of high resolution in vivo images of cones, rods and retinal pigment epithelial (RPE) cells in STGD1. The AF pattern imaged in the transition zone adjacent to clinically detectable RPE atrophy may arise from photoreceptors or represent a combination of AF signals from both RPE cells and photoreceptors.

Meaning:

These results suggest that the hyperautofluorescence at the leading disease front in STGD1 is mediated by the accumulation of lipofuscin precursors in photoreceptors and may represent an early disease step in STGD1.