Associations of 6p21.3 Region with Age-related Macular Degeneration and Polypoidal Choroidal Vasculopathy

Neovascular age-related macular degeneration (AMD) and polypoidal choroidal vasculopathy (PCV) are leading causes of blindness in aging populations. This study was conducted to investigate the associations of chromosome 6p21.3 region, including CFB-SKIV2L-TNXB-FKBPL-NOTCH4 genes, with both neovascular AMD and PCV. Six single nucleotide polymorphisms (SNPs) in this region and two known AMD-associated SNPs in CFH (rs800292) and HTRA1 (rs11200638) were genotyped in a Han Chinese cohort composed of 490 neovascular AMD patients, 419 PCV patients and 1316 controls. Among the SNPs, TNXB rs12153855 and FKBPL rs9391734 conferred an increased susceptibility to neovascular AMD (P = 2.8 × 10−4 and 0.001, OR = 1.80 and 1.76, respectively), while SKIV2L exerted a protective effect on neovascular AMD (P = 2.2 × 10−4, OR = 0.49). Rs12153855C and rs9391734A alleles could further increase the susceptibility to AMD in subjects with rs800292, rs11200638 and rs429608 risk alleles. However, only the association of SKIV2L rs429608 remained significant after adjusting for rs800292, rs11200638 and the other 5 SNPs. The protective haplotype AATGAG exhibited significant association with neovascular AMD (permutation P = 0.015, OR = 0.34). None of the SNPs in this region was associated with PCV. Association profiles of 6p21.3 region showed discrepancy between neovascular AMD and PCV, indicating possible molecular and pathological differences between these two retinal disorders.

Clinically, it remains controversial as to whether PCV is a subtype of neovascular AMD or a separate disease entity 10 . Similarities between PCV and neovascular AMD have been observed in demography, pathology and manifestation, but differences have been noted in clinical features, histopathology and response to treatment 12 . While genetically, whether a true association exists between neovascular AMD and PCV is still poorly understood. The high temperature required factor A1 gene (HTRA1) and the complementary factor H gene (CFH) were reported to be significantly associated with AMD in various populations, which were later replicated in PCV 13 . On the other hand, several AMD susceptible genes, including cholesterylester transfer protein (CETP), superkiller viralicidic activity 2-like (SKIV2L), complement component 3 (C3), elastin (ELN), and apolipoprotein E (APOE) showed different association profiles between AMD and PCV 11,[14][15][16][17] . These evidences, however, are neither sufficient nor cogent enough to interpret the relationship between neovascular AMD and PCV. It was, therefore, the purpose of this study to investigate whether the differences in clinical presentations between these two retinal disorders can be attributed to other genetic factors that may reveal different underlying pathogenic mechanisms.
Complement component 2 (C2) and complement factor B (CFB) were paralogous genes located 500 bp from the major histocompatibility complex (MHC) class III region on chromosome 6p21 18 . C2 and CFB function as activators of classical and alternative complement cascade, respectively. Single nucleotide polymorphisms (SNPs) in the C2 and CFB genes were reported to be associated with AMD, while these associations were inconsistent and inconclusive among different populations [19][20][21][22][23] . Recently, a genome-wide association study (GWAS) and several case-control association studies demonstrated that the superkiller viralicidic activity 2-like gene (SKIV2L), adjacent to the C2-CFB region, conferred a protective effect of AMD 15,21,24 . In our previous study, rs429608 in the SKIV2L gene has also been found to be associated with AMD 25 . SNPs in the TNXB-FKBPL-NOTCH4 region, downstream of the C2-CFB-SKIV2L region, have been identified to be associated with AMD in the UK population through a GWAS 26 . More recently, the Genetics of AMD in Asians (GAMA) Consortium reported that a functional SNP rs12661281 (Asp47Val) in the SLC44A4 gene, upstream of the C2-CFB-SKIV2L region, was associated with AMD in East Asians 27 . All of these SNPs, including rs12661281, rs541862, rs429608, rs12153855, rs9391734, rs2071277 and rs3132946, were located at chromosome 6p21.3 region. Taking these evidences together, it is possible that more genetic variants in this region may be associated with AMD and PCV.
Functionally, SKIV2L plays a role in the degradation of RNAs and autophagy, increased autophagy of RPE has been associated with drusen formation [28][29][30] . Tenascin-X, encoded by the TNXB gene was shown to involve in the collagen and elastin networks, both collagen and elastin are present in the Bruch's membrane 31,32 . In addition, FKBPL has been linked to proangiogenic hypoxic signals, hypoxic stress in the micro-environment of the photoreceptor/RPE/Bruch's membrane/choriocapillaris complex has been reported to be an underlying cause of pathophysiology of AMD 33,34 . Moreover, NOTCH4 is one of the four cell surface receptors of the Notch signaling pathway, Notch signaling involves in the development of retinal vasculature by regulating the specification of endothelial cells into stalk and tip cells 35 . As stated above, the functions of these genes could possibly lead to the relevance between the 6p21.3 region and neovascular AMD/PCV.
In the present study, we have genotyped these 6 SNPs in the 6p21.3 region (including rs541862, rs429608, rs12153855, rs9391734, rs2071277 and rs3132946) and 2 major AMD associated SNPs (rs800292 in CFH and rs11200638 in HTRA1) using SNaPshot method, and tested their associations with both neovascular AMD and PCV in a Han Chinese population composed of 490 neovascular AMD patients, 419 PCV patients and 1316 controls. As for rs12661281, since we have been part of the previous GAMA Consortium study, the AMD and PCV samples used in the GAMA Sichuan replication cohort were almost the same as the samples used in the present study, so we did not genotype rs12661281. Instead, we re-evaluated the data, because we have not stratified them into neovascular AMD and PCV sub-groups before.

Results
Association Study of the CFB-SKIV2L-TNXB-FKBPL-NOTCH4 Region. As shown in Table 1, a total of 2225 subjects were recruited for genotyping, including 490 neovascular AMD patients, 419 PCV patients and 1316 unrelated controls. There were more males in case and control groups, therefore, we adjusted gender in the association studies using binary logistic regression. Notably, the mean age of control group was older than that of both case groups, this was due primarily to that we purposely recruited subjects older than 60 as controls to eliminate the confounding effects from younger subjects, thus age was not adjusted in the association analysis. None of the 8 SNPs showed significant deviation from Hardy-Weinburg equilibrium in the three groups (P > 0.001, Supplementary Table 1). Allelic associations and minor allele frequencies were shown in Table 2. Two major variants, rs800292 and rs11200638, were significantly associated with both AMD and PCV as expected (Table 2). However, rs429608 (P = 2.2 × 10 −4 , OR = 0.49, 95% CI: 0.34-0.72), rs12153855 (P = 2.8 × 10 −4 , OR = 1.80, 95% CI: 1.31-2.48) and rs9391734 (P = 0.001, OR = 1.76, 95% CI: 1.27-2.50) showed significant association only with neovascular AMD, but not with PCV ( Table 2). The minor allele of rs429608 (A) showed a protective effect for neovascular AMD, and this was consistent with our previous study 25 . Meanwhile, the minor allele of rs12153855 (C) and rs9391734 (A) conferred an increased risk of neovascular AMD. Rs541862 (CFB), rs3132934 (NOTCH4) and rs2071277 (NOTCH4) were neither associated with AMD nor with PCV in this study ( Table 2). In addition, when we compared neovascular AMD with PCV, significant differences were observed in rs12153855 (P = 0.001, OR = 2.10, 95% CI: 1.35-3.26), rs9391734 (P = 0.001, OR = 2.33, 95% CI: 1.43-3.80) and rs11200638 (P = 0.001, OR = 1.40, 95% CI: 1.14-1.72). It was noteworthy that rs12153855 and rs9391734 appeared to be risk factors for neovascular AMD (OR = 1.80 and 1.76, respectively), but they showed a protective trend for PCV (OR = 0.84 and 0.87, respectively).
The genetic effects of rs12153855 and rs9391734 were further evaluated under the allelic model in the context of rs11200638, rs800292 and rs429608. The stratification of rs11200638, rs800292 and rs429608 were defined by the dominant model with the risk allele as reference (Table 4). In the strata of rs11200638AA+ AG, rs800292CC+ CT and rs429608GG+ AG genotypes, rs12153855C and rs9391734A alleles showed significantly increased risk effects on neovascular AMD (Table 4). In comparison, in the strata of rs11200638GG, rs800292TT and rs429608AA genotypes, neither rs12153855 nor rs9391734 showed significant association with neovascular AMD (P > 0.05). These data indicated that rs12153855C and rs9391734A alleles would further increase the susceptibility to neovascular AMD in subjects carrying rs11200638, rs800292 and rs429608 risk alleles.
Multiple logistic regression analysis was conducted with all of the 8 genotyped SNPs and gender as a covariate (Table 5). Only SKIV2L rs429608, in the CFB-SKIV2L-TNXB-FKBPL-NOTCH4 region, remained significant association with neovascular AMD when conditioning on other covariates (P = 0.016, Table 5), indicating its independent effects on neovascular AMD.

SNP Chromosome Position Gene
Minor allele  Table 3). So in order to investigate whether rs12661281 showed significant association only with neovascular AMD/PCV, we stratified the subjects by sub-groups and re-evaluated the data of rs12661281 by using different genetic models (Supplementary  Table 4). Interestingly, rs12661281 was also only associated with neovascular AMD, the association was observed under allelic model (P = 0.028, OR = 1.29, 95% CI: 1.02-1.62), homo model (P = 0.00066, OR = 4.21, 95% CI: 1.72-10.31) and recessive model (P = 0.00073, OR = 4.15, 95% CI: 1.70-10.14). But for PCV, no association was revealed under any of these genetic models (P > 0.05).

Discussions
We conducted a comprehensive evaluation of the 6p21.3 region in both neovascular AMD and PCV. Two genetic variants in this region, TNXB rs12153855 and FKBPL rs9391734, are significantly associated with neovascular AMD in a Han Chinese population. As expected, associations were also observed for CFH rs800292 and HTRA1 rs11200638 with both AMD and PCV, and for SKIV2L rs429608 with AMD. CFB were expressed in the neural retina, RPE, choroid and Bruch's membrane. Plasma concentration of CFB in AMD patients was significantly higher than that in normal controls 18,36 . Therefore, it is plausible that genetic variants of CFB gene may be associated with the susceptibility to AMD. Cipriani et al. 26 reported that CFB rs541862 was significantly associated with AMD in the UK population through a genome-wide association  study. In the same year, Nakata et al. noted that rs541862 was associated with typical AMD and PCV in a Japanese cohort 21 . Tanaka et al. took a step further, they categorized AMD patients into typical AMD, PCV and retinal angiomatous proliferation (RAP) groups, then categorized the PCV group into polypoidal CNV and typical PCV subgroups. They found that rs541862 was associated with polypoidal CNV but not with typical PCV, indicating that PCV might be genetically divided into polypoidal CNV and typical PCV 37 . In the current study, however, rs541862 neither associated with neovascular AMD nor with PCV in the Han Chinese cohort. SKIV2L was thought to function as RNA helicase, involving exosome recruitment and activation. Also, SKIV2L was reported to play a role in the mechanism of autophagy, and increased autophagy of RPE was related to drusen formation [28][29][30] . Previously, SKIV2L rs429608 has been associated with AMD in US, UK and Han Chinese populations, all of these studies indicated the minor allele of rs429608 as a protective factor [24][25][26] . Liu et al. reported that rs429608 was only associated with neovascular AMD but not with PCV in a Chinese study group 15 . The results of this study were in accordance with the previous genetic studies by these researchers.
TNXB rs12153855 and FKBPL rs9391734 were highly linked. On the one hand, TNXB gene encodes tenascin-X, tenascin-X appears to control cell adhesion and migration, and also shows to involve in the maturation and maintenance of collagen and elastin networks 31,32,38 . Both collagen and elastin are present in the Bruch's membrane, which could possibly lead to the relevance between TNXB gene and AMD. On the other hand, FKBPL encodes the FKBP-like protein, which plays an essential role in zebrafish and murine blood vessel development 33 .  In human body, FKBPL is specifically downregulated by proangiogenic hypoxic signals 33 . Hypoxic RPE cells can produce angiogenic substances, such as vascular endothelial growth factor (VEGF), to stimulate the growth of new vessels from choriocapillaris complex, resulting in CNV 34 .In the original UK GWAS 26 , rs12153855 and rs9391734 were significantly associated with AMD, minor alleles of these two polymorphisms were risk factors.

SNP Genotypes
In the current study, both rs12153855 and rs9391734 only associated with neovascular AMD but not with PCV. NOTCH 4 is a member of the Notch signaling pathway, which can mediate the development of retinal vasculature 35 . Notch signaling also contributed to angiogenic homeostasis by providing a counterbalance to proangiogenic pathways such as that mediated by VEGF 39 . In the UK GWAS 26 , NOTCH4 rs2071277 was significantly associated with AMD; and rs3132934 showed nominal association in the discovery cohorts (P = 0.0027) but demonstrated no association in the English and Scottish replication cohorts 26 . As for the present study, our results suggested rs2071277 and rs3232934 were not associated with neovascular AMD or PCV.
In the current study, SKIV2L rs429608, TNXB rs12153855 and FKBPL rs9391734 presented different association profiles between neovascular AMD and PCV. These SNPs were significantly associated with neovascular AMD, but not with PCV. When we compared PCV with neovascular AMD by binary logistic regression, significant differences were showed in SNPs of rs429608, rs12153855, rs9391734 and rs11200638 (Table 2). It was noteworthy that rs12153855 and rs9391734 appeared to be risk factors for neovascular AMD (OR = 1.80 and 1.76, respectively), but they showed a protective trend for PCV (OR = 0.84 and 0.87, respectively). However, the minor (B) For AMD, 5 haplotypes were observed. AGCAGG conferred an increased susceptibility to AMD (P = 0.014, OR = 1.56), but it could not withstand permutation procedure (permutation P = 0.10). AATGAG showed a protective effect on AMD (P = 0.002, OR = 0.34), and it remained statistically significant after correction for multiple testing (permutation P = 0.015). (C) LD was measured using combined PCV case and normal control data. (D) For PCV, 6 haplotypes were observed, but no significant association was detected.
Scientific RepoRts | 6:20914 | DOI: 10.1038/srep20914 allele frequencies of rs12153855 (0.042) and rs9391734 (0.041) were a little low in the Han Chinese population, thus their effect on genetic differentiation of neovascular AMD and PCV might be limited. Although risk effects of HTRA1 rs11200638 were detected on both neovascular AMD and PCV, the effect was more pronounced in neovascular AMD than that in PCV (OR = 2.31 and 1.65, respectively). Furthermore, 2 haplotypes generated by the 6 SNPs of the CFB-SKIV2L-TNXB-FKBPL-NOTCH4 region showed significant association with neovascular AMD. On the other hand, no haplotypes in this region were associated with PCV (Fig. 1).
In addition, we found a very interesting correlation between rs12661281 and AMD. According to the previous GAMA Consortium study 27 , SLC44A4 gene has been found to be associated with AMD in East Asians through a genome-wide and exome-wide association study, and it may represent Asian-specific genetic association for AMD. The researchers conducted stratified analysis in the discovery samples (ref . 27 Supplementary Table 3). In the Hong Kong discovery cohorts, rs12661281 neither associated with typical neovascular AMD (tAMD) nor with PCV; in the Japan discovery cohort, rs12661281 associated with both tAMD and PCV; while in the Singapore discovery cohort, rs12661281 only associated with tAMD but not with PCV. Due to the potential importance to stratify the AMD samples into neovascular AMD and PCV, we have re-evaluated the data of the Sichuan replication cohort in the GAMA Consortium study. Similar to the results of the Singapore cohort, rs12661281 only associated with neovascular AMD in our samples (Sichuan Samples).
Taken the results of previous studies 15,25,26 and this study together, it suggested that the chromosome 6p21.3 region was associated with neovascular AMD and this region might play an important role in pathogenisis of neovascular AMD in the Han Chinese population. Genetic variants in the TNXB and FKBPL genes would further increase the risk of neovascular AMD in subjects carrying HTRA1 rs11200638, CFH rs800292 and SKIV2L rs429608 risk alleles. Association profiles of the 6p21.3 region showed discrepancy between neovascular AMD and PCV, indicating possible molecular and pathological differences between these two retinal disorders. The SNPs of rs12153855 and rs9391734 could be applied as biomarkers to differentiate the neovascular AMD and PCV. Further research for the effects of the 6p21.3 region on neovascular AMD and PCV will provide additional insights.

Methods
Study Participants. All study subjects were unrelated Han Chinese recruited from the ophthalmology clinic at Sichuan Provincial People's Hospital. All participants underwent a standard ophthalmic examination protocol, including optical coherence tomography (OCT), indocyanine green angiography (ICGA), ocular tonometry, slit-lamp biomicroscopy, best-corrected visual acuity measurement, color fundus photographs, and fluorescein angiography. All AMD cases selected for this study had at least one eye affected by neovascular AMD. All PCV cases were diagnosed using ICGA. Individuals with other causes of CNV, or with both CNV and PCV lesions in the same or fellow eye, were excluded. All controls were matched by geographic area and were given complete ophthalmic examinations, they were included on the following criteria: 1) 60 years or older; 2) no signs of early AMD or macular degeneration of any cause; and 3) no other major eye diseases, except for mild senile cataracts or mild refractive errors.
For the association study on the CFB-SKIV2L-TNXB-FKBPL-NOTCH4 region, a total number of 490 neovascular AMD patients, 419 PCV patients and 1316 normal matched controls were recruited. Characteristics of the study subjects were listed in Table 1. And for the re-evaluation of the SLC44A4 rs12661281, the data were gained from the previous replication study in the GAMA Consortium research. The characteristics of the Sichuan replication cohort were reported in ref. 27. This study was approved by the Ethnics Committee on Human Research of Sichuan Provincial People's Hospital. All participants signed informed consent prior to participation in the study. The study procedures were performed in accordance with the tenets of the Declaration of Helsinki. SNP Selection and Genotyping. Previous GWAS in the UK population identified 5 novel SNPs in the chromosome 6p21.3 region, including rs541862 in CFB, rs12153855 in TNXB (tenascin XB), rs9391734 in FKBPL (FK506 binding protein like), rs2071277 and rs3132946 in NOTCH4 26 . These SNPs were chosen as candidate variants in the present study. Additionally, because rs429608 in SKIV2L, rs800292 in CFH and rs11200638 in HTRA1 have been previously reported to be associated with AMD in Han Chinese, they were also included in this study. Venous blood of each subject was drawn and collected in an EDTA-containing tube. Genomic DNA was extracted from the blood using a Gentra Puregene Blood DNA kit (Gentra, Minneapolis, MN). SNP genotyping was conducted by the dye terminator-based SNaPshot method (Applied Biosystems, ABI, Foster City, CA), with success rate and accuracy greater than 99%, as judged by random re-genotyping of 10% of the samples in the subject group.
The genotype data for rs12661281 in SLC44A4 were obtained from ref. 27. Because the neovascular AMD and PCV samples were highly overlapped between the present study and the previous replication study in the GAMA Consortium research, we did not genotype rs12661281 in this study. Instead, we re-evaluate the data by stratifying them into neovascular AMD and PCV sub-groups.
Statistical Analysis. All statistical analyses were conducted using SPSS software version 17.0 (SPSS, Inc., Chicago, IL). Hardy-Weinberg equilibrium (HWE) of each SNP was tested with a standard observed-expected chi-square test (χ 2 test). The genetic association analysis was carried out by constructing 2 × 3 tables of the genotype counts and 2 × 2 tables of the allele counts for each SNP in both patient and control groups. Subsequently, Pearson χ 2 statistics were calculated and P values were computed by comparing the statistic to a χ 2 distribution with 1 or 2 degrees of freedom for the allelic and genotypic tests. The odds ratios (ORs) and corresponding 95% confidence interval (CI) were estimated with the minor allele as reference, and were calculated using the χ 2 test. The overall association study was adjusted for gender with binary logistic regression. For the significance threshold of P values, Bonferroni correction was applied to adjust the P values by the number of strata. In the overall allelic association study, as 8 SNPs were chosen, a P value of less than 0.00625 (0.05/8) was considered statistically significant.
The genetic effects of rs12153855 and rs9391734 were further evaluated under the allelic model in the context of rs11200638, rs800292 and rs429608. The stratifications of rs11200638, rs800292 and rs429608 were under the dominant model with the risk allele as reference. For Bonferroni correction, as 2 SNPs (rs12153855 and rs9391734) were analyzed under 6 strata (2 strata for each of rs11200638, rs800292 and rs429608), a P value less than 0.0042 [0.05/ (2 SNPs × 6 strata)] was considered statistically significant. Multiple logistic regression analysis was conducted under the allelic model using binary logistic regression, with all of the 8 SNPs and gender as a covariate.
Haplotype analysis was carried out with Haploview 4.2 software (Daly Lab at the Broad Institute, Cambridge, MA), in order to assess linkage disequilibrium (LD) patterns and haplotype association statistics. Haplotype blocks were defined using the "custom" option implemented in Haploview software, all of the 6 SNPs in the CFB-SKIV2L-TNXB-FKBPL-NOTCH4 region were placed within 1 haplotype block. LD values were displayed in D' and r 2 , respectively. To correct for multiple testing in the haplotype association analysis, 10,000 permutations were run using Haploview software. The OR and 95% CI for each haplotype were calculated using SPSS.