MURPHREE FESTSCHRIFT: RESEARCH REPORT Functional imaging of mitochondria in genetically confirmed retinal dystrophies using flavoprotein fluorescence

Background: Whether by indirect oxidative stress or direct genetic defect, various genetic retinal dystrophies involve mitochondrial stress. Mitochondrial flavoprotein fluorescence (FPF), reported as either average signal intensity or variability (heterogeneity), may serve as a direct, quantifiable marker of oxidative stress. Materials and Methods: This observational study enrolled patients with genetically confirmed retinal dystrophies between January and December 2021. Patients with concomitant maculopathy and ocular hypertension were excluded. Patients were FPF imaged with OcuMet Beacon® third generation device during routine outpatient visit. Results: The final analysis cohort included 242 images from 157 patients. Mean FPF intensity was significantly increased between age matched controls and patients with confirmed rod-cone dystrophy, Stargardt disease, Bardet-Biedl syndrome (BBS), and Mitochondrial ATP synthase mutation (P ≤ 0.007). Mean FPF heterogeneity was significantly increased between age matched controls and patients with confirmed rod-cone dystrophy, Stargardt disease, and BBS (P ≤ 0.011). FPF lesions were noted to correlate with Fundus Autofluorescence (FAF) lesions in diseases examined. Conclusions: FPF intensity and heterogeneity significantly increased in patients with retinal dystrophies. The correlation of FPF lesions with FAF lesions implies FPF may be a clinically useful biomarker in patients with IRDs.


Introduction
As a highly metabolic tissue, the human retina relies heavily upon normal and healthy mitochondrial function. Mitochondrial oxidative stress is a major consequence of the disease process in degenerative retinal dystrophies (1,2). Multiple mechanisms drive these changes: genetic alterations generate oxidative stress on the tissue and hinder mitochondrial function as seen in Rod-Cone dystrophies, or direct genetic changes in mitochondrial DNA (mtDNA) impair mitochondrial function as seen in mitochondrial retinal dystrophies. The common final pathway is increased oxidative stress from mitochondrial dysfunction, leading to cellular stress and ultimately death and atrophy of the tissue. Environmental factors such as smoking, UV light, and air pollution can further exacerbate these phenomena ( Figure S1) (3)(4)(5).
Mitochondrial flavoproteins are proteins containing riboflavin derivatives that serve essential roles in mitochondrial electron transport (3,4). In a pro-oxidant environment, flavoproteins display properties of autofluorescence (FPF), emitting green light (520-540 nm) when excited by a blue spectrum light (455-470 nm) (6,7). The green emission signal can then be quantified and used as a direct marker of oxidative stress and mitochondrial dysfunction (8). FPF may be reported as intensity, a cumulative value reflecting global signal strength, and/or heterogeneity, which quantifies the variation of the relative intensity of points in the image. Together, these signals are meant to facilitate early disease detection for a given patient. Twelve peer reviewed clinical studies have demonstrated that FPF intensity and heterogeneity increase in various ocular pathologies including diabetic retinopathy, glaucoma, central serous chorioretinopathy, and AMD (8)(9)(10)(11)(12)(13)(14). Elner et al. examined FPF intensity in one patient with retinitis pigmentosa (RP) and found increased FPF intensity compared to a similarly aged control patient (12). However, this was a single case report, leaving the utility of FPF in patients with retinal dystrophies largely unknown.
Due to the high retinal stress and distinctive lesions characteristic of retinal dystrophies, this patient population may provide new insights into the utility of FPF imaging. In addition, retinal dystrophies resulting from mtDNA damage might be suited to early detection with this form of imaging intended to target mitochondrial dysfunction. Herein, we sought to assess both FPF intensity and heterogeneity in a large patient cohort of genetically confirmed inherited retinal dystrophies (IRDs).

Materials and methods
This observational study was conducted following IRB approval from Cleveland Clinic Institutional Review Board. Study-related procedures were performed in accordance with good clinical practice (International Conference on  Harmonization  of  Technical  Requirements  of  Pharmaceuticals for Human Use E6), applicable FDA regulations, the Health Insurance Portability and Accountability Act, and the Declaration of Helsinki. Participants were patients aged 12 years or older examined at the Cleveland Clinic Cole Eye Institute from 1 January 2021 to 21 December 2021. Inclusion criteria were as follows: dilated fundus examination (DFE), diagnosis of a retinal dystrophy by board-certified ophthalmologist and clinical geneticist, and molecular genetic testing that supports the clinical diagnosis. Exclusion criteria were as follows: ocular hypertension including glaucoma, concomitant macular abnormalities or retinopathy not resulting from the IRD, history of uveitis, hydroxychloroquine use, history of retinal detachment, vitreoretinal surgery in the previous three months, and fluorescein imaging on the day of examination. In cases of bilateral involvement, each eye was considered separately. Control patients had no retinopathy on clinical exam and lacked exclusion criteria. We included patients in the five following categories: Rod-Cone dystrophy (Retinitis pigmentosa, Usher syndrome), Stargardt disease, Bardet-Biedl syndrome (BBS), Mitochondrial encephalomyopathy, lactic acidosis, and stroke-like episodes (MELAS), and Mitochondrial ATP synthase subunit 6 mutation (MT-ATP6).
The OcuMet Beacon ® third-generation device (for investigational use only) was used for FPF imaging (14)(15)(16). Patients provided written informed consent prior to study participation. Each eye was imaged three times, 5-10 minutes after dilated fundus examination. Raw images generated by the device were securely exported to OcuSciences ® , developers of the OcuMet Beacon ® ; for quality control to ensure proper focus, centering around the fovea, and exclusion of shadow artifacts (e.g, eyelid or eyelash). Only the highest quality of the three images taken per eye was used for analysis. Eyes with no good quality images were excluded from the cohort, which occurred in 4 out of the 125 dystrophy eyes (3.2%). Control eyes were age matched from a database of 552 eyes. Proprietary software was used to calculate FPF parameters for each image. FPF intensity was calculated as the average FPF intensity over a 5.5 mm diameter region centered around the Fovea. FPF heterogeneity was calculated as the integral of the peak areas of FPF intensity in the region of interest, yielding a measure of variability.
The electronic medical record (EMR) was reviewed for demographic information, pertinent systemic comorbidities and ocular comorbidities that would exclude the patient from the study. Clinical exam information was also acquired including best visual acuity (BVA), intraocular pressure, macular examination, media opacities, and lenticular abnormalities. Continuous variables were reported as mean (95% confidence interval), while categorical variables were described as frequency (percentage). BVA was converted from Snellen to ETDRS letters using the formula ETDRS = 85 + 50 x log 10 (Snellen Fraction) (17). Cases were matched to controls using caliper matching, creating an age matched data set of 121 retinal dystrophy eyes and 121 control eyes. Analysis of variance was used to assess FPF differences across various retinal dystrophies (Tables 1 and 2, Figures 1 and 2). Multivariate logistic regression controlling for confounders including age, smoking, gender, and race was used to model FPF intensity, heterogeneity, and visual acuity (Tables S3-S5) (16). FPF heterogeneity was plotted against intensity to visualize FPF metrics amongst disease states studied ( Figure S2). Statistical analysis was performed using R (Foundation for Statistical Computing, Vienna, Austria, version 3.6.2), and P < 0.05 was taken to be statistically significant.  Table S1. Of the 121 eyes with Retinal dystrophies, the distribution across categories was as follows: 78 eyes with Rod-Cone dystrophy (Retinitis pigmentosa, Usher syndrome), 31 eyes with Stargardt disease, 4 eyes with BBS, 6 eyes with MELAS, and 2 eyes with MT-ATP6. List of confirmed genes reported in supplemental Table S2.

Results
To assess for associations between clinical dystrophy and FPF value, patients were age-matched using caliper matching, and stratified based upon dystrophy category which included control, Rod-cone dystrophy, Stargardt disease, BBS, MELAS and MT-ATP6. Mean FPF intensity was significantly increased in patients with Rod-cone dystrophy, Stargardt disease, BBS, and MT-ATP6 compared to control patients (p = 0.003, p < 0.001, p < 0.001, and p = 0.007, respectively (Table 1, Figure 1). Mean FPF heterogeneity was significantly increased in patients with Rod-cone dystrophy, Stargardt disease, and BBS compared to control subjects (p < 0.001, p < 0.001, and p = 0.011, respectively (Table 2, Figure 2). No significant differences were present between MELAS patients and control subjects with respect to FPF intensity or heterogeneity. To visually assess FPF intensity and FPF heterogeneity amongst disease states, individual cases were plotted in a scatter plot ( Figure S2).
Multivariate regression analysis of FPF intensity showed Rod-cone dystrophy, Stargardt disease, BBS, and MT-ATP6 correlated with significantly increased FPF intensity (p < 0.001 for all diseases, respectively, Table S3). Multivariate regression analysis of FPF heterogeneity showed that Rod-cone dystrophy, Stargardt disease, and BBS positively correlated with significantly increased FPF heterogeneity (p < 0.002 for all diseases, respectively (Table S4). Multivariate regression analysis of BVA showed that Rod-cone dystrophy, Stargardt disease, and BBS were correlated with significantly decreased visual acuity (p < 0.001 for all diseases, respectively (Table S5). FPF intensity and heterogeneity were not significant predictors of visual acuity.
FPF images were visually compared to fundus auto fluorescence (FAF) image areas of retinal dystrophy by MWR and verified by JCM, CAR, EIT, and RPS, in order to associate areas of flavoprotein fluorescence with areas of retinal dystrophy in select cases (Figure 3). FPF images were scaled to FAF images and cases of disease states were highlighted. In a case of RP, perimeters of increased FPF signal as measured by pixel intensity were outlined and compared to FAF intensity, finding greater FPF signal in the lesion perimeter relative to FAF. In a case of Stargardt disease, atrophic lesions were designated r1, r2, and r3. Lesions r2 and r3 were found to have greater FPF intensity relative to r1 than was observed on FAF. FPF signal was compared in a case of MT-ATP6, finding differences in signal intensity around macular lesions which was not present on FAF imaging.

Discussion
This study found a significant increase in FPF intensity or heterogeneity in patients with IRDs including rod-cone dystrophies, Stargardt disease, BBS, and MT-ATP6. These findings indicate a quantifiable increase in mitochondrial dysfunction and oxidative stress in patients with these IRDs, which is consistent with current theories regarding the natural history of retinal dystrophy progression (2,18,19). When aggregated, this large dystrophy cohort supports and expands upon the finding by Elner et al. as FPF intensity was significantly increased in one patient with retinitis pigmentosa (Rod Cone dystrophy) compared to a control patient (12). The present study also adds perspective regarding FPF and imaging of patients with Stargardt disease, BBS, MELAS, and MT-ATP6 upholding the concept that oxidative stress can be measured in a variety of IRDs.
Regression modeling attempting to control for known confounders of FPF imaging further confirmed Rod Cone dystrophy, Stargardt disease, BBS, and MT-ATP6 to be positively correlated with increased FPF intensity or heterogeneity, further validating associations seen between FPF signal and disease state. Encouragingly, this association was observed to persist despite known confounders such as age, smoking, gender, and race. As expected, a separate regression model determined that diagnoses of rod-cone dystrophy, Stargardt disease, and BBS to be significant negative predictors of BVA. Interestingly, FPF intensity and heterogeneity were not found to be significant predictors of BVA. Multiple explanations are possible for this finding. First, BVA as measured in clinic is preserved by a rather small portion of retina photoreceptors in the central fovea. Because FPF captures a much larger area than required to preserve vision, increases in FPF may not reflect central fovea involvement. Additionally, as each image represents a discrete snapshot, FPF evolution with respect to  longitudinal disease evolution within the same patient is not conferred. Consequentially, it may be possible that BVA correlates with FPF change in time, however this information was not elucidated by this study's design. Along similar lines, BVA decline is most often observed in later disease stages in some of these diseases, as retinal atrophy involves the foveal area. Further research will be needed to examine these associations. Because patients with MT-ATP6 mutations have errors in ATP synthase resulting in dysfunctional mitochondria, this subset of patients had potential to serve as a positive control for detection of mitochondrial dysfunction with increased metrics of FPF intensity (20,21). This hypothesis proved to be correct at the eyes studied with MT-ATP6 had significantly increased FPF intensity. The MELAS cohort examined was also hypothesized to be a potential positive control for FPF imaging. Interestingly, no significant differences were seen between patients with MELAS and age-matched control patients. Multiple reasons may explain this finding. Only 6 eyes from patients with MELAS were included in this study, and these eyes were only found to have mottling noted on fundus exam and color fundus photography. As mottling noted in these patients is often reflective of abnormal pigment deposition, there is reason to believe a pigmentary change alone may not be representative of retinal stress and mitochondrial dysfunction (22). Because of this, it is possible these patients did not have severe dysfunction and ocular pathology to be detected by FPF imaging. Because patients with MELAS have mutations in tRNA and are non-protein encoding, it is plausible that these patients will not have decreases in cellular respiration and increases in ROS necessary to elicit an FPF signal (23). Conversely, MT-ATP6 mutations directly impact mitochondrial cellular respiration by affecting ATP synthase,  making increased FPF more plausible in theory. Longitudinal imaging and expansion of these cohorts may further reveal how trends in FPF values correspond to long term visual function.
When FPF images were scaled and compared with FAF images, FPF signal intensity was observed to be surrounding dystrophic lesions, suggesting increases in oxidative stress and mitochondrial dysfunction in these areas ( Figure 3). This observation is in line with the current theory of FPF imaging, as tissue undergoing increased oxidative stress and environmental pressure is necessary for flavoproteins to emit fluorescence. Most convincing evidence of this unique signal resides in analysis of eyes with MT-ATP6 mutations (Figure 3). These eyes displayed areas of increased intensity which was not seen on FAF imaging, suggesting FPF signal may be unique from FAF in eyes with genetically dysfunctional mitochondria. FPF signal was also observed to mirror the hyper-autofluorescent ring seen on FAF imaging of a patient with RP, a studied prognostic indicator for visual decline in these patients (Figure 3) (24). Encouragingly, these findings suggest FPF imaging accurately reflects FAF signal in space. Further, it can provide information about the time to deterioration of retinal deterioration by highlighting active mitochondria stress. These touted benefits remain to be investigated, however, the potential merits exploration. Clinically, this has implications for low vision aid implementation and may be used in targeting CRISPR-CAS-9 therapeutics to these areas. Currently, FAF is commonly used to visually assess retinal stress in expanding dystrophic lesions (25,26). FAF utilizes an excitation filter around 590 nm and a barrier filter centered around 680 nm, allowing fluorescence detection up to 680 nm.
Conversely, FPF utilizes a narrow excitation filter around 467 nm and a narrow pass filter around 535 nm. Since FAF is largely a signal of lipofuscin which has a peak emission spectra of 600 nm and FPF aims to quantify peak emission of flavoproteins around 520-540 nm ( Figure S3), the technology has the potential to provide a new pertinent dataset (27). However, lipofuscin emission spectra may interfere with FPF signal due to its broad emission spectra of 500-750 nm ( Figure S3). As Stargardt disease is characterized by deposits of lipofuscin in retinal RPE cells, it offers a valuable in vivo comparison of these competing signals (28). FPF imaging metrics of patients with Stargardt disease showed marked increases in both FPF intensity and heterogeneity, even when compared to other retinal dystrophies examined (Figures 1 and 2, Tables 1 and 2). These increases in signal may be due to increases in emission spectra from FPF and lipofuscin in these patients, suggesting lipofuscin emission may be enhancing the FPF signal as appreciated by similarities between FAF and FPF imaging in a case example of Stargardt disease ( Figure 3). Areas with increased lipofuscin deposition may also correlate with increased metabolic tissue stress, suggesting these modalities may complement each other. However, further comparative image analysis suggests a more nuanced picture. Most notably, when comparing atrophic regions in patients with Stargardt disease on FPF and FAF imaging, signal was detected on FPF imaging where absences were noted on FAF (r2 & r3, Figure 3). This suggests FPF imaging may be detecting unique signal, even when there is lack of FAF signal suggesting RPE stress leading to atrophy. These findings are plausible as FPF signal is presumably generated from all layers of retina tissue, not solely RPE which is the primary contributor to FAF. Further studies aimed to quantify lipofuscin emission in a similar manner to FPF will be needed to clarify the enhancement of FPF signal by lipofuscin.
Advantages of the study at hand reside in its large cohort of patients with retinal dystrophies, use of an appropriately age matched control cohort, multivariate analysis, and association of FPF activity to retinal dystrophic lesions. Limitations of this study include ongoing methods to adjust for cataract status, lack of formal image grading or analysis when comparing FPF to FAF, and potential unknown confounders of the FPF signal. For example, natural lens fluorophores may be interfering with FPF. Supporting this notion, pseudophakic lens status found to be a significant negative predictor of FPF intensity. However in the present study, pseudophakic patients were only present in our IRD cohorts, which may have falsely lowered the true FPF value in these patients, potentially diminishing effect size observed. Future studies correcting for and validating lens contribution to FPF signal are essential for removing this interference. FPF signal was only measured around the macula for this patient cohort, which is often spared in retinal dystrophies with more notable peripheral involvement which may have resulted in falsely low FPF scores for cases examined. While age-matching was performed to control for age-related increases in FPF signal, authors recognize the wide heterogeneity that exists in IRDs with respect to age of disease onset, severity of disease, and progression. Future studies examining FPF amongst specific genotypes of IRDs over time may provide a more comprehensive dataset. Other limitations include the exclusion of eyes without good quality images, which may have introduced inadvertent sampling bias. The authors also recognize and have accounted for implicit bias of working with OcuSciences® as per best practices.
The conclusions in this study are also limited in that photographs provide a point in time impression of a long clinical course in patients with IRDs. For this reason, the study can only suggest future utility but cannot provide definite information on the significance of discrete scores for disease prognosis. Longitudinal imaging of patient cohorts will provide an understanding of FPF's clinical utility in retinal dystrophies and beyond by elucidating FPF signal's correlation with BVA, visual fields and microperimetry, disease evolution, and comparability to lesions seen of FAF.
This study is the first to examine FPF in a large cohort of patients with IRDs and reveals increases in FPF intensity and heterogeneity in patients with retinal dystrophies compared to age-matched controls. The data herein suggest that FPF has the potential to serve as an indicator of retinal metabolic health and disease in retinal dystrophies. Longitudinal tracking of disease evolution, FPF signal, and relationship with common metrics of disease progression are needed to bridge current observations to the ultimate clinical potential of this technology.

Disclosure statement
RPS reports personal fees from Genentech/Roche, personal fees from Alcon/Novartis, grants from Apellis and Graybug, personal fees from Zeiss, personal fees from Bausch + Lomb, personal fees from Regeneron Pharmaceuticals, Inc. All other authors report no disclosures.

Funding
CAR is an employee of OcuSciences Inc. maker of OcuMet Beacon®. The device was provided by Ocuscience,Inc for the purposes of the research.