Characterization of Macular Fundus Autofluorescence Changes in Patients with Retinitis Pigmentosa

Retinitis pigmentosa (RP) is an inherited retinal degeneration caused by many different genetic variants that promote rod photoreceptor degeneration [1]. The loss of rods results in reduced vision when there is dim illumination often described as night blindness [2]. This can be documented by electrophysiological testing that shows prolonged dark adaptation and/or reduced scotopic responses on full-field electroretinography (ERG). Initially, ERGs show greater reduction in rod function than cone function, but loss of ERG function often precedes other functional losses so that when many patients present, both cone and rod ERG function are severely reduced [2]. In rodent models of RP, cone degeneration begins after the majority of rods have died, starts around the optic nerve, and progresses from central to peripheral retina [3, 4]. In humans, cone photoreceptor loss starts in the mid-periphery of the retina and extends posteriorly and anteriorly [5]. In the middle stages of the disease, visual fields show midperipheral scotomata that enlarge over time resulting in constriction of the central field and gradual reduction in the size of peripheral islands of remaining vision [2]. This constellation of symptoms and signs constitute the RP phenotype, which is critical to document because RP is a clinical diagnosis. The clinical diagnosis is supported if genetic testing demonstrates one or more RP disease-causing variants, but there are many patients for whom disease-causing variants have not been identified and their diagnosis rests solely on clinical presentation. In addition, genetic variants that usually result in RP, can in some patients cause another type of inherited retinal degeneration [6, 7]. Therefore, documentation of the RP phenotype is important even in patients presumed to have RP because they harbor genetic variants that commonly cause RP.

Fundus autofluorescence (FAF) is an imaging test that allows visualization of fluorophores within the retina. In eyes of patients with RP, FAF often shows a hyper-autofluorescence (hyperAF) ring surrounding the fovea [8, 9], and therefore the presence of a hyperAF ring supports the diagnosis of RP and is a component of the RP phenotype. However, the absence of a hyperAF ring does not exclude the diagnosis of RP. In a study of 34 patients with RP, 20 patients (59%) had a hyperAF ring [10] and in a study of 109 eyes of RP patients, 63 eyes (57.8%) had a ring [11]. Little is known regarding non-ring phenotypes or other features present in eyes that have a ring. Also, a detailed description of ring characteristics and how they are similar or different among RP patients is not available. This study was designed to identify and catalog the FAF features seen in the macula of eyes of patients with RP and to correlate FAF features with loss of ellipsoid zone (EZ) observed on SD-optical coherence tomography (OCT). Use of the grading scheme in RP patients entered in interventional trials will help determine if any of the FAF features have predictive value with regard to therapeutic response.

The protocol for this study, a retrospective review of FAF images obtained as part of standard care of RP patients at the Wilmer Eye Institute, was approved by the Johns Hopkins Institutional Review Board. Subjects of the study were patients with a clinical diagnosis of RP based upon history, funduscopic examination, visual fields, OCT, and in most cases ERG, for whom FAF images were available. Sample size was determined by the number of images that were available. We obtained each patient’s age, sex, race, ethnicity, visual acuity (VA) and the identified gene from their electronic health records. The FAF images available for evaluation included 66 ultrawide-angle FAF images obtained with an Optomap camera (Optos Inc., Marlborough, MA; 532 nm green laser excitation and a barrier filter with a 615 nm–715 nm bandpass region) on 33 patients with RP and 130 30° images obtained with a Spectralis (Heidelberg Engineering, Inc., Franklin, MA; 488 nm blue laser excitation and a 500 nm barrier filter) on 66 RP patients (images were available from only 1 eye in 2 patients). The FAF images were downloaded from their platform in a deidentified manner and assigned a study identification number. Some of the authors (M.J.K., Z.R., F.B.A., M.C.M., I.S., M.S., X.K., P.A.C.) examined the images and described the characteristics seen in the macula of each image. During meetings, the authors developed a consensus regarding features that were seen and nomenclature for describing them using terminology from previous RP studies as much as possible. A grading scheme was developed to list the characteristics seen on each image. The images were analyzed many times by each author individually and as a group to refine the grading scheme and then to use it to grade each of the images.

For complete hyperAF rings within the macula for which there was EZ truncation within the scan window, the internal ring diameter along the horizontal meridian through the fovea was measured by two independent graders. This was done using the measurement tool on the respective software that each image was retrieved from. Any differences between graders were adjudicated.

SD-OCT Heidelberg Spectralis images taken on the same dates as the autofluorescence images were used to the measure the EZ width. The measurement was taken through the foveal center using the measurement tool in Heidelberg Spectralis for all eyes. The measurement corresponded and aligned with the ring diameter. For the eyes in which the EZ width extended beyond the window of the scan, we noted that their measurements have been underestimated.

Demographic and clinical characteristics of the participants were initially summarized. To compare each outcome measure by phenotype, linear mixed-effects models were employed with a random intercept to account for potential between-eye correlation. Similar models were utilized to compare subphenotypes within phenotype 2 eyes. Given the presence of multiple groups for comparison, adjusted p values were derived from the raw p values generated by the linear mixed-effects models. Tukey’s method was applied to adjust for multiplicity in pairwise comparisons (https://cran.r-project.org/web/packages/emmeans/vignettes/comparisons.html). All statistical analyses were conducted using R version 4.4.0. The linear mixed-effects models were fitted using the “lme4” package, and adjustments for multiple comparisons were executed using the “emmeans” package.

Detailed evaluation of the images by the authors (except AA) led to identification of four primary phenotypes: (1) hyperAF arcs, (2) hyperAF rings, (3) hyperAF rings with abnormal hyperAF within the ring, and (4) central hyperAF. The second phenotype was most common, occurring in 86% of ultrawide angle images and 80% of 30° images (Table 1). There were subgroups of phenotype 2 because rings differed in several respects including completeness, width, and location. The most common subgroup was complete hyperAF rings seen in 65% of ultrawide field images and 70% of 30° images. Most often ring thickness (distance from the inner to the outer border of the ring) was less than 0.5 disc diameters (DD) in which case the ring was designated as a narrow hyperAF ring (Fig. 1a, b). Less often ring thickness was greater than 0.5 DD (wide hyperAF ring, Fig. 1c, d) or part was less than 0.5 DD and part was greater than 0.5 DD (mixed narrow and wide hyperAF ring, Fig. 1e, f). A complete ring is continuous or has only small regions of discontinuity constituting less than 10% of the circumference, and if there were larger areas of discontinuity, the ring was designated incomplete. Similar to complete rings, there were narrow (Fig. 2a, b), mixed narrow and wide (Fig. 2c, d), and wide (Fig. 2e) incomplete hyperAF rings in the macula. Some hyperAF rings extended outside the macula (Fig. 3a–c). Noting when a ring extends outside the macula conveys information about its location and also suggests that it may be a large ring.

The first phenotype, hyperAF arcs, was much less common than the second phenotype and consisted of segments of a ring that differed from incomplete rings in that more than 30% of the ring was missing (Fig. 3d–f). Retina located within the arc appeared normal.

The third phenotype was hyperAF rings with abnormal hyperAF within the ring. This differs from phenotype 2 in which there is hypoAF within the ring from macular pigment often surrounded by gradually increasing AF to the inner border of the ring. In phenotype 3, there is a hyperAF ring and there may be the appearance of background hypoAF with irregular hyperAF superimposed on the hypoAF background (Fig. 4a, b). The irregularly diffuse hyperAF within a hyperAF ring seen in phenotype 3 must be distinguished from focal hyperAF in the fovea or perifoveal area that corresponds with overlying cystoid spaces which can occur in phenotype 2 (Fig. 4c, d). The fourth phenotype is central hyperAF in which there is not a definite ring, but rather more uniform hyperAF in the macula (Fig. 4e, f).

In addition to the features that defined the four primary macular FAF phenotypes, several secondary characteristics were identified. Each of the secondary characteristics appear in multiple images in Figures 1-5, but to simplify and conserve space, only one example is cited; however, online supplementary Table 2 (for all online suppl. material, see https://doi.org/10.1159/000543082) lists all of the images in which each secondary FAF phenotypic characteristic appears. Some images showed focal areas of hypoAF within the macula that were not contiguous with peripheral hypoAF. There were two types of posterior hypoAF, distinct, which was dense with sharp borders (Fig. 4d) and indistinct, which was less dense with indistinct borders (Fig. 1f). OCT line scans through areas of distinct posterior hypoAF showed loss of RPE and photoreceptors. Indistinct posterior hypoAF was associated with media opacities, either vitreous opacities or posterior subcapsular cataract. Posterior hypoAF sometimes formed complete or incomplete distinct (Fig. 5a) or indistinct rings (Fig. 5b). Similar to hyperAF rings, a complete hypoAF ring is continuous or has only small regions of discontinuity constituting less than 10% of the circumference. Another secondary characteristic is the presence of diffuse hyperAF (Fig. 2a) or focal hyperAF (Fig. 2b) in remaining macula outside hyperAF/hypoAF rings. In some images, there was no surrounding AF to assess because the outer border of the hyperAF ring abutted peripheral hypoAF (Fig. 3a).

Peripheral hypoAF due to pigment and/or atrophy occurs relatively early in RP and over time extends posteriorly. The extent of posterior progression of peripheral hypoFAF into the macula may be a measure of disease progression and is worthwhile capturing as a component of the macular FAF phenotype of RP patients. Some images showed no peripheral hypoAF extending inside the region of the major arcade vessels between the disc and the temporal border of central hypoAF from macular pigment (Fig. 1c). These were characterized as having no encroachment of peripheral hypoAF into the macula. If there was hypoAF extending within arcade vessels but not by more than 0.25 DD, this was characterized as minimal encroachment of peripheral hypoAF into the macula (Fig. 1e). If peripheral hypoAF extended well past major arcade vessels but not within 1.5 DD of the fovea, this was characterized as moderate encroachment (Fig. 1b) and if peripheral hypoAF extended within 1.5 DD of the fovea this was characterized as severe encroachment (Fig. 4f).

Almost all images showed peripapillary hypoAF unless the disc or a substantial portion of the disc was not included in the image or was lacking clarity. Online supplementary Table 1 is a grading form that can be used to document all of the described fundus AF features seen in the macula of patients with RP. When grading wide field images, peripheral changes are also documented using the grading scheme of Hariri et al. [12].

In general, most of the characteristics described above could be assessed on either ultrawide or 30° FAF images, but the limited field of view in the latter sometimes made it difficult to assess whether hyperAF rings extended outside the macula, the extent of encroachment of peripheral hypoFAF, and the presence or absence of peripapillary hypoAF. For these reasons, the compilation of macular FAF phenotypic characteristics was done separately for ultrawide and 30° FAF images.

The EZ is a high interference band in OCT scans generated by photoreceptors inner and outer segments. In patients with RP, the horizontal EZ width through the fovea is a measurement from the point at which the EZ is lost temporal to the fovea to the point it is lost nasal to the fovea and therefore provides an assessment of the region of retina on each side of the fovea in which photoreceptors have intact inner and outer segments. A SD-OCT obtained on the same day as the FAF image was used to measure the horizontal EZ width through the fovea (online suppl. Fig. 1A). Almost all SD-OCT scans were obtained using a 20 × 20 window which provides visualization of 6,000 µm of central retina. In some cases, intact EZ extended to the edge of the window on one or both sides of the fovea in which case the measured EZ width was less than the actual EZ width. This occurred for 15/16 (94%) of phenotype 1 images, 20/161 (12%) of phenotype 2 images, and none of the phenotype 3 or 4 images. Despite the measured EZ width being less than the actual EZ width for all but one of the phenotype 1 images and 12% of phenotype 2 images, the mean measured EZ width for phenotype 1 was significantly greater than each of the other 3 phenotypes and that for phenotype 2 was greater than that for phenotypes 3 and 4 (Fig. 6a). Of the 20 phenotype 2 images for which intact EZ extended to the edge of the OCT window on one or both sides of the fovea, 14 occurred in 2 subgroups, incomplete hyperAF rings and hyperAF rings extending outside the macula. Mean measured EZ width was significantly higher in these 2 subgroups compared with the subgroup of complete hyperAF rings within the macula (Fig. 6b). For the 128 complete hyperFAF rings for which the measured horizontal EZ width through the fovea was equal to the actual EZ width, the horizontal inner diameter of the ring was measured (online suppl. Fig. 1B) and potted versus EZ width for each image. The values were very similar with a correlation coefficient of 0.9397 (Fig. 7).

Genetic testing was done in 89 of the 99 patients included in the study. Two patients had autosomal dominant RP with siblings who had a RHO pathogenic variant and chose not to be tested because it is reasonable to assume that they also have RHO pathogenic variants. In the 89 patients tested, pathogenic variants were identified in 69 plus the two with presumed RHO variant gives 140 images from patients with known a pathologic variant (images were available from only 1 eye for 2 patients). There were six phenotype 1 images from patients for whom a pathologic variant was identified and all were RHO pathologic variants (online suppl. Table 3). Autosomal recessive inheritance predominated in other phenotypes, including the 3 subgroups of phenotype 2.

In 2003, Robson et al. [8] reported a dense hyperAF ring in the macula of 33 patients with RP with VA of 20/30 or better and found that pattern ERG P50 amplitude correlated with ring radius suggesting that hyperAF ring size might be an indicator of disease status. Using a variety of functional tests, subsequent studies confirmed that many hyperAF rings provide a border between relatively good functioning retina with intact EZ within the ring and poorly functioning retina lacking the EZ outside the ring [9, 10, 13‒15]. In 44 consecutively referred patients with RP, 34 had retained fixation and in those, Murakami et al. [10] determined that 59% had a hyperAF ring, 18% had central hyperAF, and 24% had neither. Hariri et al. [12] described a grading system to characterize FAF features on ultrawide field FAF images which was quite detailed with regard to FAF characteristics in the periphery and used the Murakami grading scheme for the macula. In a study evaluating disease asymmetry and hyperAF ring shape in RP patients, Jauregjui et al. [16] found that of 336 eyes, 290 (86%) presented with a regular hyperAF ring and 46 (14%) with an irregular shaped ring. Regular rings were defined as closed rings with an ellipsoid/round shape and regular borders, while irregular rings included any other ring morphology, including open rings, closed rings with irregular borders, and closed rings with non-ellipsoid shapes. In this study, we attempted to describe all of the FAF characteristics seen the macule of 99 patients with RP and organize them into a grading scheme.

Four primary RP FAF phenotypes were seen. The most common, phenotype 2, was a hyperAF ring with no abnormal hyperAF within the ring. This phenotype was seen in 86% of ultrawide field images and 80% of 30° images. There were many differences among rings resulting in several subgroups based upon the completeness of the ring (complete or incomplete), the thickness of the ring (narrow or wide), and the location (contained within the macula or extending outside the macula). Incomplete rings were often seen when peripheral hypoAF abnormalities were present in only one or two quadrants. The first phenotype, hyperAF arcs, was also frequently seen in the setting of segmental or mild hypoAF changes in the peripheral retina. The incomplete ring or arc generally occurred on the side of the macula bordering quadrants of peripheral retina with hypoAF abnormalities and the open segments of the arc or ring lacking hyperAF occurred adjacent to normal appearing peripheral retina. This suggests that incomplete hyperAF rings or arcs may be associated with less severe degeneration or an earlier stage of degeneration. This is supported by EZ width measurements which showed that the mean horizontal EZ width through the fovea was greater for hyperAF arcs and incomplete hyperAF rings than complete hyperAF rings. Longitudinal studies are needed to determine if hyperAF arcs evolve into incomplete hyperAF rings and then into complete hyperAF rings over time. The size and location of hyperAF rings may also have clinical significance. Longitudinal studies have shown that hyperAF rings tend to decrease in size over time [11, 15, 17], suggesting that large hyperAF rings that extend outside the macula may be an indicator of early stage disease. This was supported by relatively high mean EZ width in this subgroup.

The clinical significance of hyperAF ring width is unclear. Understanding its significance is hampered by the lack of understanding of the molecular pathogenesis of ring formation. Excessive lipofuscin in RPE cells is one source of hyperAF and has been postulated to be the source of hyperAF rings in RP [14, 18]; however, bisretinoids in photoreceptors may also contribute and could be a source of oxidative stress through photooxidation [19]. This could mean that eyes with thicker hyperAF rings may have greater oxidative stress and might be more likely to benefit from treatments designed to reduce oxidative stress. FAF in rings and surrounding retina can be quantified by quantitative FAF utilizing a confocal scanning laser ophthalmoscope equipped with an internal fluorescent reference [20]. This could be very useful, but the equipment is not widely available and might not be applicable for large interventional trials. In addition, many of the images analyzed in this study had indistinct areas of hypoAF due to media opacities that perturb quantitative FAF. However, studies using quantitative FAF protocols have provided important insights. They have shown that photopigment-related unmasking of RPE AF is unlikely to account for the presence of the AF rings [20]. Also, abnormal SW-AF within the ring cannot be attributed to accelerated phagocytosis of photoreceptor outer segments (as has been suggested) since formation of bisretinoid lipofuscin occurs before phagocytosis.

Rings usually enclosed areas of hypoAF or areas of hypoAF surrounded by a gradient of gradually increasing AF, but some rings contained diffuse and/or irregular hyperAF and were categorized into a separate phenotype, phenotype 3. Quantitative FAF has indicated that abnormal AF within rings cannot be attributed to accelerated phagocytosis of photoreceptor outer segments (as has been suggested) since formation of bisretinoid lipofuscin occurs before phagocytosis [20]. Phenotype 4 is central hyperAF in which there was no definite ring, unlike phenotype 3, and instead there was diffuse relatively uniform hyperAF in the macula.

The high correlation between horizontal EZ width and the horizontal diameter of hyperAF rings supports previous studies suggesting that the ring demarcates retina with intact EZ from that lacking EZ [10, 14, 16]. The potential for photoreceptors lacking inner and outer segments to recover is uncertain, but it may be possible and could be influenced by FAF characteristics in this area of retina outside the hyperAF ring. Also, the status of retina surrounding a hyperAF ring might influence the viability and/or function of neighboring photoreceptors within the ring. For instance, it could indicate retina and/or RPE experiencing high stress that could release cytokines that have a negative effect on posterior retina within the hyperAF ring. If bisretinoids in photoreceptors contribute to hyperAF surrounding hyperAF rings as well as to thickness of hyperAF rings, these characteristics could identify eyes with a higher level of oxidative stress that could have greater benefit from treatments that reduce oxidative stress.

Some eyes with RP showed focal distinct posterior areas of hypoAF outside hyperAF rings which sometimes formed incomplete or complete hypoAF rings. These regions corresponded with areas of RPE and photoreceptor dropout seen with OCT. This could negatively impact outcomes. In contrast, focal indistinct posterior hypoAF from media opacities should not affect outcomes. These hypotheses can only be tested in future trials if these FAF features are documented.

It was found that peripapillary hypoAF occurred in almost all eyes with RP. In rodent models of RP, cone degeneration starts in the peripapillary region and spreads anteriorly [4, 21]. Peripapillary degeneration may also occur early in humans but anterior spread may be delayed, particularly on the temporal side of the disc due to the high density of cones in the adjacent macula in humans which is not present in rodents. The high prevalence of peripapillary hypoAF in RP suggests that its absence should raise suspicion of a possible phenocopy.

This study has limitations because it is retrospective. There was no prespecified protocol for obtaining images and there was some variability in image quality. The study is cross-sectional; future longitudinal studies will be useful and determine if there is progression from one phenotype to another over time. It is also limited because it provides more questions than answers. As noted above, the significance of many of the FAF features described is unknown.

There are some similarities and many differences in the FAF features seen in the macula in patients with RP. We do not understand the significance of the differences, but one way to gain more insights regarding this is to carefully document and categorize them so that changes over time and impact on outcomes in interventional trials can be evaluated. We hope that this study provides a resource to accomplish this goal.

The authors thank Janet Sparrow, PhD and Stephen Tsang MD, PhD Edward S. Harkness Eye Institute, Columbia University Medical Center for critically reviewing the manuscript and providing valuable suggestions.

This study, which involved a retrospective review of patient data, was conducted in accordance with ethical standards and was approved by the Institutional Review Board (IRB)-1 of Johns Hopkins University (Approval No.-IRB00381006). Given the retrospective nature of the study, the requirement for informed consent was waived by the IRB. As this study involved a retrospective analysis of previously collected data, a waiver of written informed consent was granted by the Johns Hopkins Institutional Review Board (Approval No. [IRB00381006]).

All federal and state laws and Johns Hopkins policies regarding patient privacy and HIPAA were followed. All study team members have completed all clinical training modules and research modules regarding patient privacy and HIPAA. There were no study deviations or unidentified problems.

None of the authors have a conflict of interest related to this study. Not related to study P.A.C.: Personal financial interest – Aerpio, Allegro Ophthalmics, AsclepiX, Ashvattha, Bausch + Lomb, Boehringer Ingelheim Pharmaceuticals, Inc. Celanese, Clearside Biomedical, CureVac, Exegenesis Bio, Exonate, Genentech, Inc./Roche, Graybug, Regeneron, Merck, Novartis, Perfuse, Wave Life Sciences; Financial support – Aerpio, Ashvattha, Genentech, Inc./Roche, Graybug, Mallinckrodt, Oxford Biomedica, Regeneron, Regenxbio, Sanofi Genzyme.

This study was supported by 4 grants from the National Eye Institute: 1UG1EY033286 which provided salary support for M.J.K., Z.R., M.S., and PAC; UG1EY033293 which provided salary support for X.K., R01EY033103 which provided salary support for M.S., and P30EY001765 which provided salary support for J.L., This support allowed the authors to spend the necessary time needed to design and carry out the study (1UG1EY033286, UG1EY033293, and R01EY033103) and perform statistical analyses (P30EY001765).

Muhammad Jehanzeb Khan: conceptualization, data curation, formal analysis, and writing – original draft. Zainab Rustam: methodology, data curation, and writing – review and editing. Faiqa Aamir: data collection, project administration, and writing – review and editing. Maria Chairez Miranda and Imad Shaikh: investigation, data collection, and writing – review and editing. Anam Akhlaq: data collection and writing – review and editing. Jiawen Liu: statistical analysis and writing – review and editing. Mandeep Singh: conceptualization and writing – review and editing. Xiangrong Kong: conceptualization, writing – review and editing, and supervision of statistical analysis. Peter A. Campochiaro: conceptualization, supervision, writing – review and editing, final draft, and funding acquisition.

Muhammad Jehanzeb Khan and Zainab Rustam contributed equally to this work.

The datasets generated and analyzed during the current study are available as supplementary data included with this publication. These datasets are accessible to editors, reviewers, and readers without restriction. In instances where specific data are not publicly available due to legal or ethical considerations, access may be granted upon reasonable request to the corresponding author, in accordance with the policies and guidelines of the Institutional Review Board at Johns Hopkins University.