Disease asymmetry and hyperautofluorescent ring shape in retinitis pigmentosa patients

Retinitis pigmentosa (RP) is described as a bilateral disease with inter-eye symmetry that presents on short-wavelength fundus autofluorescence (SW-AF) imaging with hyperautofluorescent (hyperAF) rings with an ellipsoid shape and regular borders. Nevertheless, both asymmetry and irregular ring morphologies are also observed. In this retrospective study of 168 RP patients, we characterize the degree of inter-eye asymmetry and frequency of irregular hyperAF ring morphologies according to mode of inheritance and disease-causing gene by using SW-AF imaging and spectral-domain optical coherence tomography (SD-OCT) scans. We observed that from 336 eyes, 290 (86%) presented with regular hyperAF rings and 46 (14%) presented with irregular shapes. From the 168 patients, 23 (14%) presented with asymmetric disease, with 16 (70%) of these patients also presenting with irregular ring shapes. Patients with autosomal dominant RP (adRP) had the highest proportion of irregular ring shapes (21%) and disease asymmetry (23%) in comparison to other modes of inheritance. Furthermore, both RP1 and RHO-adRP had the highest proportions of both disease asymmetry and irregular ring morphology. Our results suggest that in patients presenting with either irregular ring shapes or asymmetric disease, emphasis should be placed in targeted gene sequencing of genes known to cause adRP, such as RHO and RP1.

Image analysis. The SW-AF images from the patients that met the inclusion criteria for the study were analyzed independently by two different graders (RJ and LC). The hyperAF rings were grouped into two categories based on morphology: 1) regular and 2) irregular. Regular rings were defined as closed rings with an ellipsoid/ round shape and regular borders, while irregular rings included any ring morphologies that deviated from the above, including open rings, closed rings with irregular borders, and closed rings with non-ellipsoid shapes. Measurements of the horizontal and vertical diameters of the hyperautofluorescent ring, along with the width of the ellipsoid zone (EZ) line from the SD-OCT scans, were performed by the two graders on closed rings only, as not all of these parameters were always well-defined in open rings. A total of 151 patients (302 eyes) from the total cohort of 168 presented with closed rings. The measurements on both eyes of each patient were performed using a built-in measurement tool in the Spectralis HRA + OCT software. The horizontal diameter was defined as the longest distance between the nasal and temporal borders of the ring, while the vertical diameter was perpendicular to the defined horizontal diameter. The external boundary of the ring, which is better defined than the internal boundary, was used as the borderline for the diameter measurements.
Statistical analysis. The statistical analyses were performed using Stata 12.1 (StataCorp, College Station, Texas, USA) software. The Pearson correlation was calculated for the measurements of both independent graders. Given the high correlation between the two graders (r = 0.99, p < 0.001 for all parameters measured), the average of the two values obtained from the graders was calculated and used for subsequent analysis. For each parameter, the difference between both eyes was calculated and descriptive statistics for the horizontal, vertical diameters, and EZ line width were calculated (see Supplementary Table S1). Given the categorical nature of our data, chi-squared tests were used to compare disease asymmetry and ring morphology among the different modes of inheritance and to assess for an association between categorical variables.
Non-syndromic arRP 78 Syndromic arRP 22 www.nature.com/scientificreports www.nature.com/scientificreports/ Asymmetry analysis. Patients were defined to have asymmetric disease if they exhibited a difference greater than the 95 th percentile cut-off in one or more of the parameters. Additionally, if patients exhibited a different ring morphology in each eye, they were also considered to have asymmetric disease. For patients with open rings,   www.nature.com/scientificreports www.nature.com/scientificreports/ only the best-defined parameter was considered in the analysis. Patients were analyzed as an entire cohort and as sub-cohorts based on mode of inheritance of the disease (autosomal dominant, autosomal recessive, and X-linked recessive). Patients were also divided into sub-cohorts based on the identified disease-causing gene, and only those cohorts with ten or more patients were analyzed.

Results
Patients. In total, 168 patients with RP were analyzed for this study. Among the 168 patients, 57 (34%) presented with adRP, 100 (60%) with arRP, and 11 (6%) with XLRP. From the arRP patient cohort, 22 patients presented with syndromic disease: 6 with Usher syndrome type I, 13 with Usher type II, 2 with Usher type III, and 1 with Bardet-Biedl syndrome. The average age of the patients was 40 years old, which was similar to the average age of the sub-cohorts, except for XLRP, where the average age was 25 years. The most common disease-causing gene was RHO for adRP (37%), USH2A for arRP (42%), and RPGR for XLRP (100%). Patient demographics and genetic characterization are summarized in Table 1 (for more complete demographic and genetic characterization, see Supplementary Table S2).

Morphology of the hyperautofluorescent rings.
When analyzing the morphology of the hyperAF rings, the majority of patients presented with regular shapes (86% of eyes), while the rest presented with irregular shapes (Fig. 1). When segregating the patients by mode of inheritance, we observed that irregular ring www.nature.com/scientificreports www.nature.com/scientificreports/ morphology is more common in the autosomal dominant (21%) as compared to the autosomal recessive (10%, P = 0.007) and X-linked (9%, P = 0.191) forms. We also analyzed the occurrence of regular and irregular ring shapes by segregating patients into cohorts by disease-causing gene. We found that there is an association between the disease-causing gene and whether the ring presents with regular or irregular morphology (P = 0.004). We observed that disease caused by EYS, RP1, and RHO presented with the highest proportions of irregular rings (29%, 23%, and 21%, respectively), whereas PDE6A/B, USH2A, and RPGR have the highest proportions of regular rings (100%, 96%, and 91%, respectively). This information is summarized in Table 2.

Discussion
With the recent advances of ocular gene therapy as a promising treatment modality for retinal dystrophies, it is important to study the level of asymmetry in these diseases, as the contralateral eye is often used as a control for the treatment eye due to the assumption of disease symmetry [16][17][18][19] . Although some studies have studied asymmetry in RP, they are either limited to a certain population of RP patients or are lacking in genetic characterization, a crucial aspect in a patient's diagnosis due to the mutation-specific nature of gene therapy. In a study by Fakin et al., for example, 54 patients with Usher Syndrome type I and II were characterized with SD-OCT and SW-AF, and they report asymmetry in 10% of their patients. A different study by Sujirakul et al. used SW-AF to measure the horizontal and vertical diameters of the hyperAF ring, and they reported asymmetry in approximately 14% of their patients 12 . Nevertheless, genetic characterization was only available for 30 out of the 88 (34%) patients they analyzed 12 . In our study, we observed asymmetry in 14% of our patients, which is similar to what the above two studies reported. Moreover, the complete genetic characterization of our patient cohort allowed us to not only analyze asymmetry in RP, but to also correlate asymmetry with mode of inheritance.
We observed that adRP presents with both higher proportions of patients with irregular hyperAF ring shapes (24%) and asymmetry (23%), as compared to arRP (10% and 9%, respectively) and XLRP (2% and 9%, respectively). We theorize that this higher proportion of disease asymmetry in adRP might be related to genetic factors and variable expressivity of the diseased allele. As compared with other forms of inheritance, adRP is known to frequently present with variations in expressivity, and multiple studies have analyzed variable expressivity and incomplete penetrance in genes that cause adRP such as PRPF8 and PRPF31 2,20,21 . A similar study to ours analyzed asymmetry in the disease progression of Stargardt disease, where the authors report that lower inter-eye correlations are more likely to be found on late-onset Stargardt disease 22 . Similar to adRP as compared to arRP or XLRP, late-onset Stargardt is milder than the other forms of the disease, such as early-onset 23 . Thus, these results suggest that disease asymmetry might be associated with mild disease severity. Of note, despite observing a higher proportion of patients with asymmetry and irregular AF ring shapes in adRP as compared to XLRP, we did not observe a statistically significant difference when comparing the two. We believe that this is due to the low number of XLRP patients in our cohort, as asymmetry is uncommon in RP and our cohort only contains 11 (6%) XLRP patients.   www.nature.com/scientificreports www.nature.com/scientificreports/ We categorized the hyperAF rings into two groups: regular and irregular ring shapes. The vast majority of patients presented with regularly shaped rings (86% of eyes analyzed). Similar to the analysis of disease asymmetry, adRP also presented with the greatest proportion of rings with irregular shapes. Currently, studies using quantitative autofluorescence (qAF) suggest that the mechanism for ring formation involves accelerated bisretinoid formation in actively degenerating receptor cells 24 . Clinically, we also know how the hyperAF ring relates to a patient's vision. The signal for SW-AF (488 nm excitation) is derived mostly from RPE lipofuscin, which is formed in the photoreceptors as a byproduct of all-trans-retinal reactions [25][26][27] . The inner border of the hyperAF in RP patients corresponds to the lateral end of the EZ line on SD-OCT, and as disease progresses, the EZ line shortens along with constriction of the hyperAF ring 24,28 . Previous studies have shown that the point at which the EZ line disappears corresponds to the edge of the patient's visual field and marks the boundary between healthy and unhealthy retina [28][29][30] . The results from these studies help us to conclude that not only is asymmetry observed in SW-AF and SD-OCT images, but also in functional vision parameters such as visual fields. Visual acuity, however, should not be affected by the asymmetric process, as RP starts on the periphery and affects central vision during the later stages of the disease. In fact, we observed no difference in the degree of asymmetry in www.nature.com/scientificreports www.nature.com/scientificreports/ visual acuity between fellow eyes in our cohort of patients with asymmetric disease as compared to those with symmetric disease (P = 0.255). The presentation of patients with asymmetric RP was also similar to patients with symmetric RP in regards to age of disease onset and severity of disease. It has been widely reported that XLRP is more severe than adRP 31 . This was also observed in our patient cohort, regardless of whether the patient presented with asymmetric disease or not. As expected, patients with RPGR-XLRP, for example, presented with more aggressive disease and earlier onset as compared to patients with RHO-adRP.
In addition to mode of inheritance, we also decided to analyze disease asymmetry and ring morphology when stratifying by the disease-causing gene. From our patient cohort, we observe that although every gene exhibits a regular ring shape more frequently, genes like EYS, RP1, and RHO have a higher proportion of patients with irregular ring morphology (29%, 23%, and 21% of eyes, respectively). This is in contrast to genes where irregular ring morphologies were observed infrequently, such as the PDE6A/B family, USH2A, and RPGR (0%, 4%, and 9%, respectively). Similarly, we observed disease asymmetry proportions were the highest in RP1 (31%) and RHO (24%), which generally have milder presentation than disease caused by the PDE6A/B family, USH2A, and RPGR.
We were able to characterize asymmetry based on the hyperAF rings observed in SW-AF imaging. Similarly, hyperAF rings are also observed in near-infrared autofluorescence imaging (NIR-AF) 14,32 . Previous studies from our group have reported that the hyperAF rings appear larger in SW-AF imaging and that similar rates of disease progression are observed in both modalities 14,32 . In the patient cohort we present for this study, only a few patients had been imaged with both modalities. Yet, in those patients, we were able to observe disease asymmetry in both NIR-AF and SW-AF imaging (Fig. 3). Future studies with larger cohorts of patients with both SW-AF and NIR-AF imaging should analyze the extent to which NIR-AF imaging demonstrates asymmetry as compared to SW-AF.
In conclusion, our study suggests genotype-phenotype correlations that can help the clinician in the diagnosis and management of RP patients. Based on our results, there is a relationship between symmetry of disease and ring morphology. Thus, if a patient presents with asymmetric disease or irregular rings on SW-AF, a diagnosis of adRP is more likely. Among the different methods of genetic sequencing, such as clinical exome or clinical genome, targeted sequencing has greater availability to patients, as its price is lower and results are obtained in less time. If either clinical exome or clinical genome sequencing cannot be obtained for a patient with asymmetric disease, emphasis should be placed in targeted sequencing of genes known to cause adRP, such as RP1 and RHO. Furthermore, given the higher likelihood of adRP disease, emphasis should be placed in screening other family members. Our results also beget the important question of whether disease progression is asymmetric between fellow eyes in those patients with asymmetric disease, which future studies should address.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.