Monitoring Lesion Area Progression in Stargardt Disease: A Comparison of En Face Optical Coherence Tomography and Fundus Autofluorescence

Abstract

Purpose: To compare longitudinal changes in en face spectral domain-optical coherence tomography (SD-OCT) measurements of ellipsoid zone (EZ) and retinal pigment epithelium (RPE) loss to changes in the hypoautofluorescent and hyperautofluorescent (AF) areas detected with short-wavelength (SW)-AF in ABCA4-associated retinopathy.

Methods: SD-OCT volume scans were obtained from 20 patients (20 eyes) over 2.6 ± 1.2 years (range 1–5 years). The EZ, and RPE/Bruch's membrane boundaries were segmented, and en face slab images generated. SubRPE and EZ slab images were used to measure areas of atrophic RPE and EZ loss. These were compared to longitudinal measurements of the hypo- and abnormal AF (hypoAF and surrounding hyperAF) areas.

Results: At baseline, the en face area of EZ loss was significantly larger than the subRPE atrophic area, and the abnormal AF area was significantly larger than the hypoAF area. The median rate of EZ loss was significantly greater than the rate of increase in the subRPE atrophic area (1.2 mm²/yr compared to 0.5 mm²/yr). The median rate of increase in the abnormal AF area was significantly greater than the increase in the hypoAF area (1.6 mm²/yr compared to 0.6 mm²/yr).

Conclusions: En face SD-OCT can be used to quantify changes in RPE atrophy and photoreceptor integrity. It can be a complementary or alternative technique to SW-AF with the advantage of monitoring EZ loss. The SW-AF results emphasize the importance of measuring changes in the hypo- and abnormal AF areas.

Translational Relevance: The findings are relevant to the selection of outcome measures for monitoring ABCA4-associated retinopathy.

Introduction

Autosomal recessive Stargardt disease (STGD1) is the most common form of inherited macular dystrophy.¹ This genetically and clinically heterogeneous disease is caused by mutations in the ABCA4 gene.^2–5 The clinical characteristics vary widely. They range from an early-onset form with progressive loss of central visual function and atrophy in the central macular area accompanied by yellow-white flecks at the posterior pole or mid-periphery on funduscopic examination to a milder late-onset form often with foveal sparing.^6,7 This phenotypic heterogeneity of STGD1 is matched by its genetic heterogeneity. To date more than 2200 disease-causing variants have been identified in the photoreceptor-specific ABCA4 gene, which encodes the ATP-binding cassette transporter (www.lovd.nl/ABCA4).⁸ The ABCA4 transporter facilitates the clearance of both 11-cis and all-trans-retinaldehyde.⁹ When the transport activity is reduced or absent, the result is an accelerated accumulation of bisretinoids associated with RPE lipofuscin.^10,11 This can be visualized as lesions with increased fundus autofluorescence or hyper-autofluorescence. Although research on the role of bisretinoids in degenerative retinal diseases affecting the macula is ongoing, there is evidence in STGD1 that the accumulation of bisretinoids causes toxicity and leads to RPE and photoreceptor cell degeneration and to a characteristic appearance on fundus autofluorescent imaging (i.e., low or hypo-autofluorescent signals).^12,13 

The increasing number of clinical trials for STGD1 highlights the need for robust outcome measures. Currently several noninvasive imaging modalities are used for the clinical diagnosis and monitoring of disease progression and thus are candidates for outcome measures. These include short wavelength fundus autofluorescence (SW-AF), near-infrared autofluorescence (NIR-AF), ultra-wide field AF using green excitation (532 nm), and spectral domain optical coherence tomography (SD-OCT). SW-AF imaging reveals the previously described areas of decreased autofluorescence or hypoautofluorescence (hypoSW-AF) reflecting disease-specific RPE atrophy, and areas of hyperautofluorescence (hyperSW-AF) in the form of flecks or rings. Recent studies have used SW-AF and ultra-wide field AF to monitor progression; the rate of progression being based on the growth of the area of definitely decreased AF or questionably decreased AF.^14–19 Based on the results of these longitudinal studies, it has been suggested that fundus autofluorescent imaging is a suitable tool for monitoring areas of hypoAF in interventional trials that aim to slow down disease progression.^15,16 Similarly, NIR-AF has been used to diagnose and monitor disease progression^20–22; the hypoAF areas indicative of RPE atrophy appear to be larger and more widespread on NIR-AF when compared to SW-AF.²² 

The areas of RPE atrophy have also been monitored using serial OCT scans. Methods include measurements of the width of sub-RPE hypertransmission in OCT B-scans²³ and of the atrophic areas using en face analysis.²⁴ The majority of recent studies monitoring disease progression have focused on methods for measuring RPE degeneration, primarily by measuring changes in hypoSW-AF areas. However, a few have investigated photoreceptor degeneration over time in terms of ellipsoid zone (EZ) loss using SD-OCT.^25–27 The latter evaluated the reliability of using the EZ band to track disease progression either by measuring transverse EZ width loss or the area of EZ loss on en face SD-OCT analysis.^25–27 Studies that have used en face SD-OCT analysis to follow disease progression in patients with STGD1 demonstrate its potential benefits for quantifying both RPE atrophy and photoreceptor layer integrity.^24,25 

Here we assess disease progression in STGD1 patients using en face SD-OCT analysis and SW-AF imaging. We compare the morphological changes on en face images derived from SD-OCT scans qualitatively and quantitatively to changes in the hypo and hyperAF areas detected with SW-AF to further evaluate the use of en face subRPE and EZ area loss as anatomic markers for monitoring disease progression. 

Methods

This is a retrospective observational study of 20 patients (20 eyes, 10 male, 10 female). All patients were diagnosed with recessive STGD1 confirmed by sequencing of the ABCA4 gene. Two expected disease-causing ABCA4 variants were detected in all 20 patients. Protein products for the following variants were determined by vitro studies (see Supplementary Table S1): c.5461-10T>C,4 c.1938−619A>G2 and c.2588G>C. All other listed proteins are predicted consequences of the cDNA variant and thus presented in parentheses. Genotypes were determined to be biallelic based on familial segregation in 8 patients (P3, P7, P8, P9, P10, P14, P15, P16 and P17) and observed co-occurrence in gnomAD in P5, P6, P12 and P19 (See Supplemental Table, https://gnomad.broadinstitute.org/variant-cooccurrence). 

Clinical, demographic, and genetic characteristics of the study cohort are summarized in Tables 1A and 1B. All study procedures were defined under protocols AAAI9906 and Study00002299, approved by the Institutional Review Board at Columbia University Irving Medical Center and The University of Texas Southwestern Medical Center, respectively and adhered to tenets set out in the Declaration of Helsinki. Patients had a complete ocular examination. Eyes were excluded from the study if there was evidence of significant ocular media opacities, refractive errors greater than ± 6 diopter (D) sphere or ± 2D cylinder, elevated intraocular pressure > 21 mm Hg, and a history or diagnosis of any other significant ocular comorbidities. The eye with the higher image quality value for the OCT cube scan was included in the study. All scans had a quality index ≥ 20 dB. Best corrected visual acuity in the tested eye at baseline ranged from 0.0 to 1.0 logMAR (20/20–20/200, Snellen acuity). Full-field scotopic and photopic electroretinograms (ERGs) were obtained at the baseline visit from both eyes of 20 patients according to the International Society for Clinical Electrophysiology of Vision standards.²⁸ Based on the full-field ERG results, patients were classified into one of three groups: group 1 if the scotopic and photopic ERG amplitudes were within normal limits (13 patients, 13 eyes), group 2 if only the photopic ERG amplitudes were decreased (seven patients, seven eyes), or group 3 if the scotopic and photopic ERG amplitudes were both decreased.²⁹ In addition, tested eyes were assigned to one of three clinical disease stages.³⁰ Eight eyes were classified as stage 1. This is characterized by pigmentary changes in the macula ranging from nonspecific pigment mottling to the presence of small atrophic-appearing foveal lesions with localized parafoveal or perifoveal flecks. Eight eyes were classified as stage 2, characterized by the presence of numerous yellowish-white flecks throughout the posterior pole, and four as stage 3, characterized by extensive macular atrophy with diffuse flecks throughout the fundus (see Table 1B). 

Table 1A.

 

Selected Clinical, Demographic and Genetic Characteristics of Study Patients

 

Table 1B.

 

Selected Clinical Characteristics of Study Patients

 

Multimodal Imaging

After pupil dilation with 1% tropicamide and 2.5% phenylephrine short wavelength-autofluorescence (SW-AF; 488 nm excitation) 30° and 55° images were acquired using the Spectralis HRA+OCT (Heidelberg Engineering, Heidelberg, Germany; acquisition software versions 6.8.1.0, 6.10.7.0) after a 20-second bleach of the photopigments in AF mode.³¹ Care was taken to obtain high-quality images with maximum field uniformity. In addition, 6-mm and 9-mm high-resolution horizontal line scans through the fovea were obtained. The SD-OCT scans were automatically registered to a simultaneously acquired near-infrared reflectance (NIR-R) fundus image. Horizontal volume scans (6 × 6 mm, and 9 × 6 mm) centered at the fovea consisting of 97 B-scans (spaced ≈ 60 µm apart) for the purposes of en face analysis were obtained from each of the 20 eyes. They were obtained using the automatic retinal tracking mode and automatic registration was used for all follow-up scans. The SD-OCT images were analyzed with the Heidelberg software (Heidelberg Eye Explorer viewing module 6.4.8.116 [SPX1504]; Heidelberg Engineering). After manual correction of the automated segmentation of the EZ and RPE/BM boundaries, the built-in software was then used to generate en face slab images based on the average reflectance intensity for slabs of constant thickness. To obtain optimal visualization of the morphological changes, we chose two en face slabs: a subRPE slab with a thickness of 100 to 200 µm below the RPE/BM boundary and an EZ slab positioned on the EZ with a thickness of 5 to 15 µm. The thicknesses of the slabs were chosen based on previous studies.^32,33 The subRPE slab image was used to measure the central RPE lesion. It is an en face visualization using only the light penetrating below the RPE into the choroid and sclera. The rationale is that in the healthy regions of RPE, the pigment in the RPE will block most of the light penetrating into the choroid, whereas in atrophic RPE areas the underlying choroid will be illuminated. On the subRPE slab image the atrophic RPE area was indirectly visualized as a “hyper-reflective area.” The hyper-reflectivity or increased OCT signal is attributable to scleral reflectance because of penetration of light into the choroid as a result of loss of the RPE and atrophy of the choriocapillaris. The EZ slab image was used to measure the area of EZ loss. The RPE lesion area on the subRPE slab and area of EZ loss were measured in mm² by two independent observers (VCG and DSC) using ImageJ64 (National Institutes of Health, Bethesda, MD, USA; available at imagej.nih.gov/ij/). The following steps were used: the measurement scale was set, the image was converted to 8-bit, the freehand drawing tool in the ImageJ toolbar was used to outline the area(s), and then the selected area(s) were measured automatically via the Analyze tab in ImageJ. In cases where the quality of the en face EZ slab images affected the clarity of the boundaries of the area of EZ loss the locations chosen to mark the boundary were checked by comparing them to the locations of EZ loss on the corresponding individual B-scans. The subRPE lesion area and area of EZ loss were then compared to measurements of the areas of hypoAF and abnormal AF (i.e., the abnormal AF area includes the hypoAF and hyperAF ring encircling the hypoAF area but excludes adjacent flecks on the SW-AF image), respectively. The latter were also measured using ImageJ64. Baseline and follow-up measurements of the en face areas and hypo- and abnormal AF areas on SW-AF images were each analyzed twice in a masked fashion by the two independent graders (V.C.G. and D.S.C.) to assess reliability. The rates of lesion enlargement (mm²/year) were determined for the areas of hypoAF, abnormal AF, subRPE hyper-reflective area, and area of EZ loss by subtracting the respective areas measured (in mm²) at the last visit from the areas measured at baseline divided by time between visits (in years). 

Statistical Methods

Statistical analysis was performed using PRISM 9 (GraphPad Software, Inc., La Jolla, CA, USA) and SPSS software. Normality of the data was assessed, and appropriate parametric and nonparametric tests were used. The reliability of the measurements made by the two independent graders (D.S.C. and V.C.G.) was calculated in SPSS (version 28.0.1.1; IBM, Armonk, NY, USA) using inter-operator reliability intraclass correlation coefficients (ICC) and their 95% confidence intervals (CI). Bland-Altman plots were generated to assess the agreement between the en face and SW-AF measurements at baseline. The statistical significance of the differences between the subRPE and EZ areas, and between the hypoAF and total abnormal AF areas was tested with the Wilcoxon matched pairs signed rank test. 

Results

At baseline the average age of the patients was 38.3 ± 19 years (range 10.8–70.3 years). The average duration of follow-up was 2.6 ± 1.2 years (range 1–5 years). The average number of follow-up visits was 2 (range 1–4). Clinical, demographic, and genetic characteristics of the study cohort are summarized in Table 1A and Table 1B. 

SW-AF

All 20 eyes exhibited changes on SW-AF. At baseline, the SW-AF images for 10 of the 20 eyes showed a central area of hypoAF of variable size (Figs. 1A, 1B yellow arrow) surrounded by a well-defined border of hyperAF (Figs. 1A, 1B; red arrow) and adjacent hyperAF flecks (Fig. 1B). The white arrow (Figs. 1A, 1C, 1D) indicates foveal sparing. In two of the 20 eyes, a pattern of hyperAF flecks of varying size and shape were observed in the perifoveal region (Fig. 1C; red arrows). Widespread hypoAF and hyperAF changes extending beyond the arcades were observed in eight of the 20 eyes (Fig. 1D). The areas of abnormal AF (i.e., the hypoAF and surrounding hyperAF areas, excluding flecks), and the areas of central hypoAF or absent AF were measured at baseline and at the final follow-up visit (see examples in Figs. 2A and 2B inserts). In agreement with a previous study³⁰ the abnormal AF area at baseline was significantly larger than the hypoAF area (median = 8.3 mm², first quartile (Q1) 4.6, third quartile (Q3) 16.5 mm² compared to median = 3.5 mm², Q1 2.1, Q3 8.2 mm², Wilcoxon matched pairs signed rank test, P < 0.0001). At the final follow-up visit, the hypoAF and abnormal AF areas were increased for all eyes (see examples in Figs. 2A–C). The rate of lesion enlargement was significantly greater for the abnormal AF area, (median = 1.6 mm²/yr, Q1 0.4, Q3 2.4 mm²/yr, Wilcoxon matched pairs signed rank test P < 0.0002), than for the hypoAF area (median = 0.6 mm²/yr, Q1 0.3, Q3 1.4 mm²/yr). 

Figure 1.

 

Examples of SW-AF images from four patients at baseline. (A) The central area of hypoAF (yellow arrow) is surrounded by a well-defined ring of hyperAF (red arrow). The area of abnormal AF includes the area of hypoAF and the surrounding ring of hyperAF. Foveal sparing is indicated by the white arrow. (B) The central area is surrounded by a mottled pattern of white flecks (red arrow). (C) The fovea is encircled by hyperAF and hypoAF flecks in the perifovea of varying size and shape. (D) Widespread hypo- and hyperAF changes.

Figure 2.

 

(A–C) Examples of longitudinal changes in the hypo- and hyperAF regions in SW-AF images from three patients P3, P5, P17. The central hypoAF area on the SW-AF image for P3 and the central area of abnormal AF are outlined in the insets.

OCT En Face Slabs

The subRPE slab images for eyes with a central hypoAF area at baseline and follow-up showed a central hyperreflective area (Figs. 3A, 3C, 3E) with foveal sparing identifiable as a hyporeflective area (Figs. 3A, 3C, 3E white arrow). The central hyperreflective areas were similar in size and shape to the central hypoAF areas on SW-AF (Figs. 2A–C). The corresponding en face EZ slab images showed a relatively hyporeflective area, surrounded in some cases by a hyperreflective border (Figs. 3B, 3D, 3F). Again, in agreement with a previous study³⁰ the en face EZ loss area was significantly larger at baseline than the subRPE area reflecting RPE atrophy (median = 8.3 mm², Q1 = 5.4 mm², Q3 = 11.8 mm²; median = 3.4 mm², Q1 = 2.1 mm², Q3 = 7.8 mm²; Wilcoxon signed rank test P < 0.0001). At follow-up all eyes showed increases in the subRPE and EZ loss areas (see examples in Fig. 3). The rate of lesion enlargement was significantly greater for the EZ area than for the subRPE area (median = 1.2 mm²/yr, Q1 0.7 mm², Q3 1.8 mm²/yr; median = 0.5 mm²/yr, Q1 0.3 mm², Q3 1.5 mm²/yr; Wilcoxon matched pairs signed rank test P < 0.0001). The rates of lesion enlargement (mm²/yr) for each of the 20 eyes derived from both the en face and SW-AF measurements are shown in Figure 4. 

Figure 3.

 

(A–F) The corresponding subRPE and EZ en face slab images for P3, P5, P17. The subRPE images show a central hyperreflective area (white outline) and foveal sparing (white arrow). In contrast the EZ images show a relatively hyporeflective area and foveal sparing (white arrow).

Figure 4.

 

Rates of EZ, abnormal AF, RPE, and hypoAF lesion area increase for individual patients. The median, Q1 and Q3 for EZ, abnormal AF, RPE and hypoAF are indicated by the horizontal lines.

Comparisons Between SW-AF and OCT En Face Slab Measurements

Agreement between the two techniques for quantifying RPE atrophy and EZ loss was assessed by comparing the measurements obtained at baseline. Measurements of the central hypoAF area on SW-AF images were compared to the central hyperreflective area on the subRPE en face image, and measurements of the total abnormal AF area on SW-AF images were compared to the central abnormal reflective area on the EZ en face image. The scatter plots (Figs. 5A, 5B) illustrate the similarity of the measurements for the two techniques. To investigate differences between these measurements we used the Bland-Altman method. The difference plots (Figs. 6A, 6B) again illustrate the similarity of the measurements. On average the SW-AF measurements of the hypoAF area were slightly larger than the en face subRPE measurements (mean difference [bias] 0.5 mm²), and the SW-AF measurements for the abnormal AF area were also slightly larger than the en face EZ measurements (mean difference [bias] 1.3 mm²). However, there were no significant differences between these measurements with the Wilcoxon matched pairs signed rank test. The reliability of the SW-AF and en face area measurements made by the two observers (V.C.G. and D.S.C.) was evaluated by calculating ICC for hypoAF (ICC = 0.97; 95% CI, 0.96–0.98), abnormal AF (ICC = 0.98; 95% CI, 0.97–0.99), en face subRPE (ICC = 0.98; 95% CI, 0.98–0.99) and en face EZ loss (ICC = 0.92; 95% CI, 0.86–0.95) areas. The ICCs showed good to excellent agreement between the two independent observers (V.C.G. and D.S.C.) for all the area measurements (range 0.92–0.98). 

Figure 5.

 

(A) Scatter plot comparing the hypoAF and subRPE en face areas, Spearman r = 0.98. (B) Scatter plot comparing the abnormal AF and EZ en face area, Spearman r = 0.91.

Figure 6.

 

(A) Bland-Altman plot showing differences in measurements of the central hypoAF area on the SW-AF image to the central hyperreflective area on the subRPE en face image. (B) Bland-Altman plot showing differences in the size of the measurements of the abnormal AF area on the SW-AF image to the central hyporeflective area on the EZ en face image. The mean difference (bias) is indicated by the dashed horizontal line and the limits of agreement within which 95% of the differences lie, by the horizontal dotted lines (upper limit = mean + 1.96 × SD and lower limit = mean − 1.96 × SD).

Discussion

Recent studies of disease progression in STGD1 have focused on methods for measuring RPE atrophy, using SW-AF and ultra-wide field AF (“green” excitation, 532 nm) imaging. The rate of increase in the atrophic lesion was based on measurements of the increase in the hypoAF areas.^14–16 However, this outcome measure does not reflect the changes in visual function that occur in STGD1. For example, a recent study using fundus-controlled perimetry (a.k.a. microperimetry) demonstrated that visual function deficits extend outside the hypoAF lesion.³⁴ Despite the relevance to visual function of photoreceptor layer integrity in STGD1 only a limited number of studies have monitored photoreceptor integrity in terms of changes in EZ loss.^25–27 Yet the measurement of EZ loss has proved to be an effective and precise method for tracking disease progression in slowly progressive retinal dystrophies such as retinitis pigmentosa and is now well established as an outcome measure for this group of inherited eye diseases.^35,36 In an earlier study we demonstrated that the en face SD-OCT approach had the potential to be a useful clinical tool for detecting and monitoring changes in EZ loss and RPE atrophy in STGD1.³³ We used en face SD-OCT analyses and SW-AF imaging to quantify changes in photoreceptor integrity and RPE atrophy. In the current study to further evaluate the en face approach for monitoring disease progression, we compared longitudinal changes in en face SD-OCT measures of EZ loss and RPE atrophy to SW-AF measures of the abnormal and hypoAF areas. 

In both our earlier and current study, the en face subRPE slab image showed a central hyper-reflective area attributable to RPE atrophy. It was similar in size and shape to the hypoAF area in the central macula in the SW-AF image. As in our earlier study we found good agreement between the baseline en face subRPE measurements and the hypoAF area measurements on SW-AF and between the en face EZ measurements and the abnormal AF measurements.³³ This indicates that the en face SD-OCT technique can be used as a complementary technique to SW-AF. However, a comparison between en face SD-OCT measurements of the area related to EZ loss and of the area attributable to RPE atrophy showed that the area of EZ loss was significantly larger. In addition, our SW-AF results indicated that the area of abnormal AF was significantly larger than the hypoAF area. These findings are consistent with our earlier study and with observations by Sodi et al.³² In the current study, not only did we find that the area of EZ loss was significantly larger than the area of RPE atrophy, the rate of EZ loss or lesion enlargement was greater. Similarly for the SW-AF results the rate of lesion enlargement was greater for the abnormal AF area than for the hypoAF area. To interpret these results and gain some insight into the underlying pathophysiology, we also need to consider the results of previous studies comparing fundus AF modalities (SW- and NIR-AF) and SD-OCT in STGD1 patients. In these studies, the area of decreased NIR-AF was reported to be larger than the area of hypo-SW-AF, and more closely related to both the area of abnormal AF and to the extent of the EZ loss observed in SD-OCT scans.^20,33,37 Because the NIR-AF signal is produced, for the most part, from RPE melanin, absence of NIR-AF is indicative of loss of RPE cells. Thus the results of the current and previous studies suggest that photoreceptor cell loss or impairment has occurred in the presence of degenerating or dysfunctional RPE cells. 

The limitations of the study include its retrospective design, modest sample size, and the heterogeneity of the group of patients in terms of age, disease onset, follow-up, and genetics. A limitation of the en face approach is that high-density raster scans are needed to create the en face images, and some patients may have difficulty maintaining fixation during the OCT procedure. Also, it is dependent on reliable segmentation of the EZ and RPE/BM boundaries. We found that we had to correct some of the automated segmentation of these boundaries manually, particularly in the region of the lesion. Manual correction of the automated segmentation is time consuming. In addition, the quality of some of the en face EZ slab images affected the clarity of the boundaries of EZ loss. For these cases, we checked the accuracy of our en face EZ loss area measurements by comparing them to the extent of EZ loss on the corresponding individual B-scans. 

Our results have implications regarding the choice of outcome measures for future longitudinal studies designed to monitor disease progression in STGD1. We have demonstrated that the en face SD-OCT technique can be used to quantify changes in RPE atrophy and photoreceptor integrity. It not only can be used as a complementary test to SW-AF for measuring changes in RPE atrophy but also more importantly measurements of the area of EZ loss provide valuable information in terms of photoreceptor loss. The measurement of the area of EZ loss, although more time consuming than the measurement of EZ width, has the following advantages: the reconstructed en face view provides a more comprehensive view of EZ loss than the two dimensions of retinal structure provided by single line scans, and it facilitates comparisons between SD-OCT measures of photoreceptor integrity and visual field loss. Also, the SW-AF results we obtained further emphasize the relevance of measuring changes in the entire area of abnormal autofluorescence, not just the hypoAF area, because the former corresponds more closely to the area of EZ loss. 

Acknowledgments

Supported by grants from the National Eye Institute/NIH EY009076, EY028954, EY024091, EY028203, EY029315; from the Foundation Fighting Blindness; and a grant from Research to Prevent Blindness to the Department of Ophthalmology, Columbia University. 

Disclosure: V.C. Greenstein, None; D.S. Castillejos, None; S.H. Tsang, Abeona Therapeutics, Inc. (F), Emendo (F); W. Lee, None; J.R. Sparrow, None; R. Allikmets, SpliceBio (C), Belite Bio (C), Intergalactic Therapeutics (C), Rectify Pharmaceuticals (C), Shape Therapeutics (C); D.G. Birch, Nacuity Pharmaceuticals (C), Editas (C), ONL (C), AGTC (C, F), Novartis (C), Biogen (F), Ocugen (F), PYC (F), DTx (F); D.C. Hood, Heidelberg Engineering (R, F), Topcon, Inc. (R, F)