Association of Irregular Pigment Epithelial Detachment in Central Serous Chorioretinopathy with Genetic Variants Implicated in Age-related Macular Degeneration

We evaluated phenotype and genotype correlation of central serous chorioretinopathy (CSC) patients with or without irregular pigment epithelial detachment (PED) on optical coherence tomography (OCT). For CSC, a flat, irregular protrusion of retinal pigment epithelium (RPE) with hyper-reflective sub-RPE fluid on OCT was defined as an irregular PED. Participants were classified into 5 subgroups; (1) total CSC (n = 280) (2) CSC with irregular PED (n = 126) (3) CSC without irregular PED (n = 154) (4) typical choroidal neovascularization (CNV) (n = 203) and (5) polypoidal choroidal vasculopathy (PCV) (n = 135). Ten known major AMD-associated single-nucleotide polymorphisms (SNPs) were analyzed. Age, sex adjusted logistic regression was performed for the association between subgroups. Association analysis between CSC without irregular PED and CNV revealed that significant difference for rs10490924 in ARMS2, rs10737680 in CFH, and marginally significant difference for rs800292 in CFH. Between CSC without irregular PED and PCV, rs10490924, rs10737680, and rs800292 were significantly different. In contrast, CSC with irregular PED and CNV revealed no SNP showing significant difference. Between CSC with irregular PED and PCV, only rs10490924 was significantly different. CSC with irregular PED and CSC without irregular PED revealed significant difference for rs800292, and marginal difference for rs10737680. These findings suggest CSC patients with irregular PED are genetically different from those without irregular PED and may have genetic and pathophysiologic overlap with AMD patients.

Results characteristics of patients. The demographic details of the study population are shown in Table 1.

Discussion
Based on the age, sex adjusted association analysis with 10 AMD-associated SNPs, patients without irregular PED showed significant difference with CNV patients in 2 SNPs, marginally significant difference in 1 SNP, and significant difference with PCV patients in 3 SNPs. In contrast, patients with irregular PED revealed no SNP showing significant difference with CNV patients, and only one SNP showing significant difference with PCV patients. In the association analysis between patients with irregular PED and those without irregular PED, the rs800292 in CFH were significantly different, and the rs10737680 in CFH showed marginally significant difference. These findings suggest not CSC patients but CSC patients with phenotype of irregular PED may have genetic and pathophysiologic overlap with neovascular AMD patients. The age, sex adjusted association analysis revealed that patients with irregular PED were genetically different from patients without irregular PED. In regards to genetic profile, patients with irregular PED were closer to patients with CNV or PCV than those without irregular PED. These findings suggest CSC patients with irregular PED may have genetic overlap with neovascular AMD patients. The findings imply that patients with irregular PED may have more tendency to have hidden CNV or to develop into CNV or PCV. This implication is similar to the previous studies that showed irregular PED in CSC or pachychoroid disease appears to be a risk factor of CNV 11,14 . One study using OCTA showed that CNV was detected more in chronic CSC eyes with irregular PED (13/31 eyes; 41.9%) than in those with regular PED (1/18 eyes; 5.6%) 11 . Given the fact that CSC is included in the pachychoroid spectrum disease, another study using OCTA was similar to the previous one and it revealed that type 1 neovascular tissue was visualized in most of the study eyes (21/22 eyes; 95%) with pachychoroid spectrum diagnoses and shallow irregular PED 14 . In the study of Dansingani KK et al. 18   www.nature.com/scientificreports www.nature.com/scientificreports/ ARMS2, three in CFH) showing significant difference with pachychoroid without neovascularization among 12 AMD-associated SNPs that were analyzed. Pachychoroid without neovascularization revealed 7 SNPs showing significant difference with neovascular AMD in their study. Although the study was different from our study because we excluded CSC patients with any evidence of choroidal neovascularization, it was similar to the present study in that two subgroups of pachychoroid disease were genetically different and one subgroup was relatively closer to the neovascular AMD.
Considering that CSC with irregular PED is more likely to develop into neovascular change, findings in the current study are in line with the previous studies that implied CSC resides within a spectrum of diseases with a pachychoroid-driven process including PPE, CSC, PNV, and PCV 4 . The current finding that CSC patients with irregular PED may have genetic overlap with neovascular AMD patients implies irregular PED in CSC might be the prodromal change for neovasularization. There have been some studies on genetic overlap between CSC and AMD. De Jong et al. showed that one SNP in ARMS2 and three SNPs in CFH were significantly associated with chronic CSC among 19 AMD-associated SNPs that were analyzed 5 . Recent study revealed that the G allele of rs800292 and T allele of rs10490924 are significantly associated with the CNV development in patients with CSC 19 . Considering the fact that CSC with irregular PED is more likely to have hidden CNV, the result of the study corresponds with the findings of the present study which showed the frequency of the G allele of rs800292 or T allele of rs10490924 was higher in CSC with irregular PED than that in CSC without irregular PED (Table 2). Another studies demonstrated that a significant association between CSC and some of the SNPs in the CFH gene 20,21 . Miyake et al. investigated the genotypic difference between PNV and neovascular AMD 22 . The study revealed that genetic susceptibility of PNV patients to AMD was significantly lower than that of neovascular AMD; the frequency of the ARMS2 rs10490924 minor allele (AMD-risky allele) was lower and the frequency of the CFH rs800292 minor allele (AMD-protective allele) was higher in PNV patients.
In addition to the genetic profile, patients with irregular PED were demographically different from patients without irregular PED. The proportion of female and the mean age of the patients were relatively higher in patients with irregular PED than those without irregular PED. Given the fact that CSC is more prevalent in men [23][24][25] and the mean age is relatively younger, ranging 39-51 years 25,26 , the above demographic findings suggest that patients with irregular PED have different phenotype from the typical CSC patients than those without irregular PED have.
Association analysis between CSC without irregular PED and PCV revealed that significant difference for . Similarly, association analysis between CSC without irregular PED and typical CNV, rs10490924 in ARMS2, rs10737680 in CFH were significantly different (P = 1.45 * 10 −3 , OR = 0.30; P = 3.65 * 10 −3 , OR = 2.87, respectively). Rs800292 in CFH showed marginally significant difference between the two groups (P = 6.12 * 10 −3 , OR = 2.68). Although some of the p-values did not reach the significant difference, the directions of the ORs were similar for the comparison between total CSC and typical CNV ). Overall, the association of CFH (rs800292, rs10737680) with total CSC or CSC without irregular PED was risky and the association of ARMS2 (rs10490924) was protective, compared to typical CNV or PCV. The minor allele of the rs800292 was known to be protective for exudative AMD in Korean 27 and Chinese 28 . This minor allele was also known to be risky for CSC in Japanese 20 and Western European 5 population. These findings can explain the risky association of the minor allele of the rs800292 for CSC compared to typical CNV or PCV in the current study. The minor allele of the rs10737680 was known to be protective for AMD in East Asian 29 and Chinese 30 . There has been no previous study analyzing the association of rs10737680 with CSC. Hence, the risky association of the minor allele of the rs10737680 for CSC compared to typical CNV or PCV in the current study cannot be directly explained. The observation that some SNPs of the CFH have opposite effects in CSC compared to AMD was also revealed in the previous studies 5, 20  the exact rationale and the mechanism are unclear. CFH is an inhibitor of the alternative complement pathway of the complement system. CFH also binds and interacts with adrenomedullin which has been shown to induce vasodilation of choroidal vascular beds, affect choroidal blood flow, and increase microvascular permeability [31][32][33][34] . The exact mechanisms for risk association of some SNPs in CFH with CSC are not clear, but the interaction of CFH with adrenomedullin may be the possible explanation of the association of CFH with CSC in that the main pathogenesis of the CSC is related with the choroidal vasodilation and hyperpermeability. In our study, the minor alleles of the rs800292 and rs10737680 were risky for only CSC without irregular PED, not for CSC with irregular PED, compared to typical CNV or PCV. Further studies on the exact mechanisms for the risky association of some SNPs in CFH only with CSC without irregular PED (and not with CSC with irregular PED) are warranted. The minor allele of the rs10490924 was known to be risky for exudative AMD in Korean 27 and for AMD in East Asian 29 . This minor allele was also reported to be protective for CSC in Western European 5 . These findings can explain the protective association of the minor allele of the rs10490924 for CSC compared to typical CNV or PCV in the current study. A recent study revealed that ARMS2 interacts with components of the extracellular matrix (ECM) 35 . Disturbances in the ECM at the level of RPE and/or choroid could affect the susceptibility to cellular detachments, which may be the possible mechanisms for the protective association of ARMS2 with CSC 5 . Our study has some limitations. First is its cross-sectional study design, which may have resulted in inherent limitations. Second, we analyzed only 10 known AMD-associated SNPs and could not cover whole genome. Considering that genetic association may be affected with ethnicity, we analyzed limited number of main SNPs based on previous studies 27,29 showing the replicated association with AMD in East Asian or Korean. Third, OCTA was performed only in small portion of the CSC patients with irregular PED (Fig. 2). To reveal the association between CSC with irregular PED and CNV (or PCV), further long term follow up study with more OCTA data of CSC patients with irregular PED is needed. Finally, there has been no consensus or established definition of irregular PED. We newly defined the 'irregular PED' as a flat, irregular non-dome-shaped protrusion of RPE with 'at least partially hyper-reflective' sub-RPE fluid and analyze the genetic difference between CSC patients with irregular PED and those without irregular PED.
Among 10 known AMD-associated SNPs, CSC without irregular PED showed significant difference with typical CNV in 2 SNPs and with PCV in 3 SNPs. In contrast, CSC with irregular PED revealed no SNP showing significant difference compared with typical CNV, and only one SNP showing significant difference with PCV. In the association analysis between CSC with irregular PED and CSC without irregular PED, rs800292 in CFH were significantly different, and rs10737680 in CFH showed marginally significant difference. In conclusion, these findings suggest patients with irregular PED are genetically different from those without irregular PED and may have genetic and pathophysiologic overlap with AMD patients.

Methods
The present study is a cross-sectional study. This study was approved by the Institutional Review Board (IRB) of Seoul National University Bundang Hospital (SNUBH; IRB No.B-1105/127-014, Seongnam, South Korea) and adhered to the tenets of the Declaration of Helsinki. Written informed consent was obtained from all subjects before participation in the study.
The CSC group was defined as a group of patients with following characteristics: serous subretinal fluid on optical coherence tomography, focal leakage spot (ink blot) or smokestack pattern or ≥1 area of multifocal diffuse www.nature.com/scientificreports www.nature.com/scientificreports/ leakage, or presence of irregular retinal pigment epithelium window defects on fluorescein angiography and corresponding hyperfluorescence on indocyanine green angiography [36][37][38] . Patients receiving PDT before the first spectral-domain (SD) OCT at our institute or patients receiving exogenous corticosteroids were excluded. Patients showing evidence of diabetic retinopathy, retinal vascular diseases, or other diseases that can cause macular exudation, such as exudative age-related macular degeneration (AMD), PCV were excluded in the CSC patients.
The typical CNV group was determined by demonstration of choroidal neovascular membrane on FA without any evidence of PCV on ICGA. According to the recent studies suggesting PCV belongs to pachychoroid spectrum diseases 4,39,40 , we analyzed the typical CNV and PCV patients separately. Cases having evidence of retinal-retinal or retinal-choroidal anastomosis on ICGA were defined as retinal angiomatous proliferation (RAP). Patients with RAP having choroidal neovascular membrane were included in the typical CNV group. The PCV group was defined as hyperfluorescent polypoidal choroidal vasculature with branching vascular network on ICGA with concomitant exudation or hemorrhage 41 .

Definition of Irregular PED and subgroups of CSC.
Based on the OCT findings in patients with CSC 42 , a dome-shaped protrusion of RPE with sub-RPE fluid was defined as a 'regular PED' . A flat, irregular non-dome-shaped protrusion of RPE with 'at least partially hyper-reflective' sub-RPE fluid was defined as an'irregular PED' . According to the optical density of the sub-RPE fluid of flat PED, there were also categories of flat PED with 'hypo-reflective' sub-RPE fluid and flat PED with 'indeterminate optical density' (i.e., flat PED with <20 μm height of the sub-RPE fluid; optical density of the sub-RPE fluid could be evaluated only when the height  www.nature.com/scientificreports www.nature.com/scientificreports/ of the sub-RPE fluid ≥20 μm.) Small irregular protrusion of RPE without sub-RPE fluid was defined as a 'RPE bump' . Using high speed mode with the spectralis OCT, we checked the 30° * 20° field of view (8.7 mm * 5.8 mm) obtained from the 25 serial B scan to classify the PED. Representative OCT images of regular PED, flat PEDs, and RPE bump were shown in Fig. 3. CSC patients were divided into two subgroups; (1) CSC with 'irregular PED' , (2) CSC without 'irregular PED' . CSC without 'irregular PED' included cases with 'regular PED' , flat PED with 'hypo-reflective' sub-RPE fluid, or with 'indeterminate optical density' , RPE bump, or only sub-retinal fluid (SRF).
Finally, study participants were classified into five subgroups; (1) total CSC (i.e., CSC with or without 'irregular PED') (2) CSC with 'irregular PED' (3) CSC without 'irregular PED' (4) typical CNV and (5)  Statistical analysis. Demographics of the study population were compared among subgroups using the t-test, χ2 test and the one-way analysis of variance (ANOVA). Genotype data cleaning and analysis was performed using PLINK (http://pngu.mgh.harvard.edu/purcell/plink). Among the genotyped SNPs, twelve major AMD-associated SNPs including 9 SNPs replicated in the previous studies of East Asian 29 , and 2 SNPs replicated in the previous studies of Korean 27 were analyzed to investigate the difference of genetic profile and association with AMD among subgroups. The SNP data was cleaned by removing SNPs with a low genotyping pass rate (greater than 5% of genotypes missing from the entire cohort), SNPs with a low minor allele frequency (less than 0.2% of MAF), and SNPs in those subjects without disease that were not in Hardy Weinberg Equilibrium (HWE p < 10 −6 ). After quality control, among 12 major AMD-associated SNPs, 10 SNPs (including 8 SNPs replicated in the previous studies of East Asian, and 1 SNP replicated in the previous study of Korean) remained for further analysis. The remained 10 SNPs were rs800292, rs1061170, rs10737680 in CFH, rs6795735 in ADAMTS9, rs4698775 in CFI, rs429608 in C2-CFB, rs943080 in VEGFA, rs334353 in TGFBR1, rs10490924 in ARMS2-HTRA1, and rs3764261 in CETP. After quality control, 280 patients with CSC, 203 patients with CNV, and 135 patients with PCV remained for the statistical analysis.
The OCT-based phenotyping for CSC was performed independently by two observers (S.C.C and N-K.R); 1) regular PED, 2) flat PED with 'at least partially hyper-reflective' sub-RPE fluid (=i.e., irregular PED), 3) flat PED with 'hypo-reflective' sub-RPE fluid, 4) flat PED with 'indeterminate optical density' , 5) RPE bump, and 6) only SRF. All measurements were performed based on 1:1 micron images. The inter-observer agreement of the phenotyping was evaluated with Cohen's kappa (k = 0.916). When discrepancies arose for the OCT-based phenotyping, the two observers discussed their evaluations and came to an agreement for final phenotype.
The analysis of 10 SNPs was performed in each pair of subgroups: (1) between CSC with 'irregular PED' and CSC without 'irregular PED'; (2) between total CSC with and typical CNV (or PCV); (3) between CSC with 'irregular PED' and typical CNV (or PCV); and (4) between CSC without 'irregular PED' and typical CNV (or PCV). For the SNP analysis, the minor allele, or less frequent allele, for each SNP was tested for association with disease of the case for each pair of the subgroups mentioned above using the chi square test in PLINK. Logistic regression was also performed to investigate the age, sex adjusted association of the genotype with disease of the case. A Bonferroni correction was performed, and the individual test for the 10 SNPs were considered significant at P < 0.005, corrected for 10 tests. The p-value of the age, sex adjusted analysis using logistic regression was finally used to determine the significant association.