Genetic Factors for Choroidal Neovascularization Associated with High Myopia

Abstract

Purpose.: Nonsyndromic high myopia, defined by a refractive error greater than −6 diopters (D), is associated with an increased risk of macular choroidal neovascularization (CNV), a vision-threatening complication. The aim of this study was to investigate whether genetic factors associated with age-related macular degeneration (AMD) are related to myopic CNV.

Methods.: We conducted a case-control study, including 71 cases with myopic CNV and 196 myopic controls without CNV, from Creteil and Toulouse, France, and Boston, MA. Single nucleotide polymorphisms (SNPs) from 15 genes reported to be related to AMD were selected for association testing in this study.

Results.: In univariate analysis, the rs10033900 SNP located in CFI was associated with myopic CNV (P = 0.0011), and a SNP in APOE was also related (P = 0.041). After adjustment for age, sex, and degree of myopia, SNPs in three genes were significantly associated, including CFI (odds ratio [OR] 2.1, 95% confidence interval [CI] 1.3–3.37, P = 0.0023), COL8A1 (OR 1.88, 95% CI 1.18–2.98, P = 0.0076), and CFH (OR 1.65, 95% CI 1.02–2.66, P = 0.04). After correction for multiple testing, only CFI remained significantly related to high myopic CNV (P = 0.045).

Conclusions.: We report the first genetic associations with choroidal neovascularization (CNV) in a high myopic Caucasian population. One SNP (rs10033900) in the CFI gene, which encodes a protein involved in the inflammatory pathway, was significantly associated with myopic CNV in multivariate analysis after correction for multiple testing. This SNP is a plausible biological marker associated with CNV outgrowth among high myopic patients. Results generate hypotheses about potential loci related to CNV in high myopia, and larger studies are needed to expand on these findings.

Introduction

High myopia or pathologic myopia is defined by an axial length higher than 26 mm or by a refractive error more than −6 diopters (D) with pathological modifications of the posterior pole of the retina, including staphyloma, lacquer cracks, and myopic conus. High myopia is a common vision-threatening disease that affects 0.5% to 5.0% of the worldwide population. ^1–3 Choroidal neovascularization (CNV) is the most common cause of visual loss related to this disorder, with an estimated prevalence of 4% to 11% among high myopic patients, and there is a 2-fold higher risk among women in some studies. ^4,5  

Genetic factors have been described in nonsyndromic high myopia through linkage analysis, genome-wide association analysis, or candidate gene case-control studies. ^6–25 However, the genetic factors influencing the risk of CNV in eyes with high myopia have not been extensively investigated. ^26,27 Several genetic factors have been strongly associated with exudative age-related macular degeneration (AMD), another degenerative retinal disease characterized by a neovascular process developing from the choroid beneath the neurosensory retina located in the macular area of the retina. ^28–50 Therefore, we hypothesized that genes associated with exudative AMD could be considered as candidate genes for myopic CNV. 

Materials and Methods

Participants

High myopic patients with axial myopia more than −6 D and pathologic myopic retinal degeneration were recruited from three different centers (Créteil, Toulouse, and Boston). Cases had high myopia with CNV in one or both eyes. The control group was defined as high myopic patients without CNV with visual acuity of 20/32 or better in both eyes. Only subjects of European/Caucasian ancestry were included. Demographic data and ocular characteristics of cases and controls are shown in Table 1. 

Table 1. 

 

Characteristics of High Myopic Patients with CNV (Cases) and High Myopic Patients without CNV (Controls)

Variable	Cases (n = 71)	Controls (n = 196)	P
Age at diagnosis, y	53.9 ± 14.9 (13 to 85)	40.9 ± 13.9 (19 to 88)	3.2 × 10−9
Sex, % female (n)	73.2% (52)	69.4% (136)	0.65
Refractive error, diopters
OD	−12.1 ± 5.3 (−6 to −27)	−9.2 ± 3.0 (−6 to −22)	1.4 × 10−4
OS	−11.9 ± 5.0 (−6 to −23)	−9.2 ± 3.2 (−6 to −21)	1.3 × 10−4
Site
France	62 (87%)	182 (91%)	0.24
United States	9 (13%)	14 (7%)

 Values denote means ± SDs and ranges or percentages.

P  values for age and refractive error are calculated by two-tailed t-test.

P  values for sex and site are calculated by χ ² test.

All patients with myopic CNV underwent complete clinical examination, including visual acuity assessment, dilated fundus examination, and fluorescein angiography. The diagnosis of CNV was based on fundus examination showing a macular scar or a lesion with subretinal hemorrhages in the absence of drusen in either eye, and/or by staining and leakage on early and late phases of fluorescein angiography. An indocyanine green angiography (ICG) scan and an optical coherence tomography (OCT) scan were also performed to confirm the diagnosis of CNV in some cases. On OCT scans, CNV appeared as a hyper-reflective lesion located beneath the neurosensory retina, usually associated with a hyporeflective intraretinal or subretinal accumulation of fluid. On ICG, CNV could be seen as a network in the early phase or as a hyperfluorescent macular lesion in the late phase, sometimes spreading from lacquer cracks or an atrophic patch that appeared hypofluorescent on the late phase. Myopic patients without CNV underwent visual acuity assessment and a dilated fundus examination. 

Patients with clinical features of AMD, including drusen or pigmentary changes, or other retinal diseases (i.e., diabetic retinopathy, ocular histoplasmosis syndrome, lacquer cracks due to trauma) related to CNV were excluded. To avoid potential spurious findings due to population admixture, non-Caucasian subjects were also excluded. 

Written informed consent was obtained for each individual participating in this study, and all clinical investigations were conducted according to the principles expressed in the Declaration of Helsinki. 

SNP Selection

We selected a total of 30 SNPs located in 15 candidate genes or genetic regions related to AMD based on previous reports of whole-genome linkage studies, genome-wide association studies (GWAS), and molecular and functional studies. ^28–51 We chose SNPs that were reported previously or SNPs tagging each candidate gene/region of transcription. ⁵² Tagging SNPs with a minor allele frequency (MAF) greater than 10% and with a minimum r ² of 0.8 for the tagging region were selected by tagger (http://www.broad.mit.edu/mpg/tagger/) based on the HapMap data from the population of Utah residents with ancestry from northern and western Europe (phase II, http://www.hapmap.org). 

Genotyping

Genotyping was performed in the Clinical and Translation Research Center Core Laboratory, Tufts Clinical and Translational Science Institute, Boston, MA, using Applied Biosystem (ABI) BioTrove OpenArray Genotyping Platform and ABI 7900HT Sequence Detection System (Life Technologies, Carlsbad, CA). OpenArray is a new platform designed for medium-throughput genotyping by ABI. SNPs were genotyped on a metal-based array of negatively charged wells in which DNA samples were amplified with TaqMan genotyping chemistry. Per the manufacturer instruction, the samples were loaded onto the assay plate, followed by PCR amplification and imaging on the Open Array NT Imager (Life Technologies). The results were analyzed with TaqMan Genotyper Software (Life Technologies). Three SNPs failed in assay design or genotyping assay on the OpenArray platform and these markers were then genotyped with TaqMan SNP genotyping assay on ABI 7900HT SDS. 

Statistical Analyses

Quality control, allele frequency estimation, and tests for association were performed for each SNP using PLINK 1.07. ⁵³ SNPs failing the following quality control criteria were excluded from the analyses: missingness rate more than 0.1, minor allele frequency (MAF) less than 0.01, deviant from Hardy-Weinberg equilibrium (HWE) (P < 0.001), or with significantly different missingness in case and control groups (P < 0.001). The resultant SNP set for analysis contained a total of 29 SNPs that complied with the above quality control criteria. Details of these SNPs are shown in Table 2. For each SNP, the genotypes were coded as 0, 1, or 2 by copies of the minor allele based on an additive genetic model. Univariate analysis was performed using logistic regression. Differences in the distribution of each potential confounding factor between cases and controls were tested using a t-test (for continuous age, refractive error) or χ2 test (for sex, site). To adjust for potential confounding factors of age, sex, and degree of myopia, SNPs were further tested using a multivariate logistic regression model. To adjust for multiple testing, corrected P values were calculated by the max (T) procedure (10,000 permutations) in PLINK. ^53,54 The association between APOE haplotypes and myopic CNV were analyzed by the haplotype-based association tests with logistic models in PLINK. 

Table 2. 

 

Candidate SNPs Tested for Association with Myopic CNV

SNP	Chr	Genomic Location (hg19)	Gene	Minor Allele Frequencies			OR* (95% CI)	P Value*	Corrected P Value†	OR (95% CI) Adjusted for Covariates‡	P Value Adjusted for Covariates‡	Corrected P Value Adjusted for Covariates§
				Cases	Controls	Allele
A. SNPs significantly associated with myopic CNV in 1 or more analyses
rs10033900	4	110,659,067	CFI	0.51	0.35	T	1.95 (1.31–2.92)	0.0011	0.021	2.1 (1.3–3.37)	0.0023	0.045
rs669676	3	99,448,852	COL8A1	0.56	0.47	A	1.42 (0.96–2.08)	0.08	0.86	1.88 (1.18–2.98)	0.0076	0.15
rs1061170	1	196,659,237	CFH	0.42	0.34	C	1.47 (0.98–2.21)	0.064	0.79	1.65 (1.02–2.66)	0.04	0.63
rs769455	19	45,412,040	APOE	0.039	0.005	T	5.44 (1.07–27.56)	0.041	0.61	4.54 (0.91–22.75)	0.066	0.82
B. Other SNPs tested for association with myopic CNV
rs10737680	1	196,679,455	CFH	0.37	0.42	C	0.78 (0.51–1.18)	0.24	1	0.77 (0.48–1.26)	0.3	1
rs1410996	1	196,696,933	CFH	0.38	0.43	A	0.76 (0.5–1.17)	0.21	1	0.77 (0.48–1.25)	0.29	1
rs7645305	3	99,363,985	COL8A1	0.28	0.22	G	1.39 (0.89–2.16)	0.15	0.98	1.26 (0.76–2.09)	0.37	1
rs793494	3	99,508,768	COL8A1	0.29	0.31	A	0.91 (0.6–1.38)	0.64	1	1.14 (0.71–1.82)	0.6	1
rs9332739	6	31,903,804	C2	0.028	0.043	C	0.63 (0.2–1.93)	0.41	1	0.46 (0.14–1.53)	0.2	1
rs641153	6	31,914,180	CFB	0.103	0.14	A	0.72 (0.4–1.3)	0.27	1	0.59 (0.3–1.16)	0.12	0.97
rs699947	6	43,736,389	VEGFA	0.5	0.49	A	1.06 (0.71–1.56)	0.79	1	1.11 (0.72–1.72)	0.63	1
rs13207351	6	43,737,794	VEGFA	0.51	0.49	A	1.1 (0.75–1.63)	0.62	1	1.18 (0.77–1.81)	0.44	1
rs735286	6	43,744,621	VEGFA	0.27	0.29	T	0.89 (0.57–1.4)	0.61	1	0.82 (0.49–1.36)	0.44	1
rs2146323	6	43,745,095	VEGFA	0.35	0.35	A	0.98 (0.65–1.48)	0.92	1	1.03 (0.64–1.64)	0.92	1
rs3025021	6	43,749,163	VEGFA	0.32	0.34	T	0.88 (0.58–1.34)	0.54	1	0.91 (0.56–1.46)	0.69	1
rs3025039	6	43,752,536	VEGFA	0.13	0.11	T	1.22 (0.67–2.2)	0.52	1	1.22 (0.61–2.44)	0.57	1
rs4711751	6	43,828,582	VEGFA	0.48	0.46	C	1.07 (0.72–1.59)	0.75	1	1.17 (0.73–1.85)	0.52	1
rs25648	6	43,846,955	VEGFA	0.2	0.17	T	1.21 (0.74–2)	0.45	1	1.35 (0.75–2.42)	0.32	1
rs1999930	6	116,387,134	FRK/COL10A1	0.36	0.33	G	1.14 (0.77–1.68)	0.53	1	1.07 (0.67–1.68)	0.79	1
rs12196141	6	116,489,550	COL10A1	0.35	0.29	G	1.27 (0.85–1.91)	0.25	1	1.17 (0.73–1.86)	0.52	1
rs1883025	9	107,664,301	ABCA1	0.33	0.30	T	1.16 (0.77–1.74)	0.48	1	1.28 (0.8–2.05)	0.3	1
rs10490924	10	124,214,448	ARMS2	0.23	0.19	T	1.3 (0.81–2.08)	0.28	1	1.42 (0.83–2.42)	0.2	1
rs10468017	15	58,678,512	LIPC	0.24	0.3	T	0.71 (0.45–1.13)	0.15	0.98	0.81 (0.48–1.36)	0.42	1
rs493258	15	58,687,880	LIPC	0.43	0.49	C	0.79 (0.54–1.17)	0.24	1	0.76 (0.48–1.19)	0.22	1
rs3764261	16	56,993,324	CETP	0.25	0.31	A	0.76 (0.49–1.17)	0.21	1	0.68 (0.41–1.12)	0.13	0.97
rs2230199	19	6,718,387	C3	0.21	0.17	C	1.28 (0.78–2.08)	0.33	1	1.32 (0.76–2.28)	0.33	1
rs429358	19	50,103,781	APOE	0.11	0.13	C	0.86 (0.46–1.61)	0.64	1	1.02 (0.5–2.06)	0.96	1
rs7412	19	50,103,919	APOE	0.09	0.06	T	1.53 (0.76–3.06)	0.23	1	1.41 (0.62–3.18)	0.41	1
rs9621532	22	33,084,511	TIMP3	0.06	0.049	C	1.17 (0.48–2.83)	0.73	1	1.02 (0.38–2.75)	0.97	1

P  values in bold are significant (P < 0.05).

 Chr, chromosome; OR, odds ratios per allele comparing allele frequencies between myopic patients with CNV with myopic patients without CNV.

*  ORs and P values for univariate logistic regression model.

†   P values for univariate analysis corrected for multiple testing by 10,000 permutations.

‡  ORs and P values for multivariate logistic regression model adjusting for age, sex, and spherical equivalent of worse eye.

§   P values for multivariate analysis corrected for multiple testing by 10,000 permutations.

Results

As shown in Table 1, cases included 71 patients with high myopia in both eyes with a refractive error greater than −6 D in both eyes and CNV in one or both eyes. Controls consisted of 196 individuals with high myopia not complicated by CNV. Cases tended to be older than controls: mean ± SD age at diagnosis for cases was 53.9 ± 14.9 years for cases, and 40.9 ± 13.9 years for controls (P = 3.2 × 10⁻⁹). Sex distributions were balanced between cases and controls (P = 0.65). Cases had a higher degree of myopia than controls (P = 1.4 × 10⁻⁴ for OD and P = 1.3 × 10⁻⁴ for OS). On average, there was a difference of 3D between cases and controls. 

Table 2 shows the allele frequencies, odds ratios (OR) and P values for the candidate SNPs and associations with high myopic CNV. In univariate logistic regression analyses, two of the SNPs were significantly associated with high myopic CNV: rs10033900 in the CFI gene (P = 0.0011, OR = 1.95 [95% confidence interval (CI) 1.31–2.92]) and rs769455 in the APOE gene (P = 0.041, OR = 5.44 [95% CI 1.07–27.56]). Only CFI rs10033900 remained significantly associated with myopic CNV after correction for multiple testing (P = 0.021) in the univariate model. 

In the multivariate model, with adjustment for age, sex, and degree of myopia, the T allele of rs10033900 in the CFI gene was significant (P = 0.0023, OR = 2.1 [95% CI 1.3–3.37]) and remained significant after correction for multiple testing (P = 0.045). Two other SNPs, rs669676 in COL8A1 (P = 0.0076, OR = 1.88 [95% CI 1.18–2.98]) and rs1061170 in CFH (P = 0.04, OR = 1.65 [95% CI 1.02–2.66]) were associated with myopic CNV in the multivariate analysis but were not significant after correction for multiple testing. Rs769455 in APOE was significant in the uncorrected univariate analysis, but was no longer significant (P = 0.066) after adjustment for the covariates. The APOE haplotypes (E2, E3, E4) were not significantly associated with myopic CNV in either univariate or multivariate analysis (see Supplementary Material and Supplementary Table S1). 

Discussion

Despite some similarities between AMD and myopic CNV, including macular CNV, subretinal location of the CNV outgrowth, and some degree of atrophy, no genetic variants have been previously reported to be associated with myopic CNV among Caucasians. In this study, we evaluated several AMD genetic variants that could also be involved in myopic CNV development. Following a candidate gene approach, this case-control study identified one SNP in the CFI gene significantly associated with myopic CNV even after adjusting for confounding factors and multiple testing. Other SNPs that showed suggestive evidence for association are worthy of further exploration as well. 

The T allele of rs10033900, which is located 2781 bp upstream of the 3′ untranslated region of the complement factor I gene (CFI), has been associated with increased risk of exudative AMD in Caucasians. ^44,47 Both complement factors I and H are complement regulatory proteins. CFH encodes factor H, the most important alternative pathway discriminator that binds C3b and prevents the formation of C3 convertase and acts as a cofactor of factor I to cleave C3b in iC3b. ⁴⁹ CFI is expressed by hepatocytes, macrophages, lymphocytes, endothelial cells, and fibroblasts and encodes factor I, a regulator protein of the three complement pathways. ⁵⁰ By cleaving of C3b and C4b, factor I reduces the formation of the C3 and C5 convertase enzymes. ⁵⁰ Other genes in the complement cascade pathway are associated with exudative AMD, including CFH, C2, CFB, and C3. ^{34–36,39–43} Interestingly, CFH (rs1061170, P = 0.04) and CFI (rs10033900, P = 0.0023) appeared to be related to myopic CNV in this study after adjustment for age, sex, and degree of myopia, and CFI remained significant after adjustment for multiple testing. The possible difference in effect of CFI compared with CFH on myopic CNV risk may be related to the fact that factor I is involved in the classic, lectin and alternative pathways, while factor H is only involved in the regulation of the alternative pathway. ⁵⁵ It is interesting to note that the gene on chromosome 10, ARMS2/HTRA1, was not related to myopic CNV in this study even though this gene is more strongly associated with CNV compared with geographic atrophy in AMD ^47,56 and also more strongly associated with all AMD subtypes when compared with the CFH at-risk common variant. ⁵⁷ It is also noteworthy that the SNPs in the VEGF gene were not related to myopic CNV, given that VEGF rs4711751 is related to advanced dry and exudative AMD. ⁴⁷  

The intronic SNP rs669676 in the COL8A1 gene was associated with myopic CNV after adjustment for age, sex, and degree of myopia (P = 0.0076). This gene encodes one of the two alpha chains of type VIII collagen, a major component of basement membranes of Bruch's membrane and choroidal stroma. ⁵⁸ The intronic SNP rs13095226 in this gene is associated with advanced AMD in our previous studies. ^45,47 The SNP rs669676 of COL8A1 might lead to direct or indirect structural alterations of the Bruch's membrane as frequently observed during high myopia (Fig.), which is a risk factor for myopic CNV. ⁵⁹  

Figure. 

 

High myopic patient (−12 D) with CNV in the right eye. (A) Fluorescein angiography showing CNV (white arrow). (B) Indocyanine green angiography showing the CNV in the early phase (white arrow). (C) Indocyanine green angiography showing a lacquer crack in the late phase (white arrow).

The association between the E4 APOE haplotype and a reduced risk of AMD has been described in two independent case-control studies, ^32,33 and supported by other studies or in meta-analyses. ^60–62 The lipid component of soft drusen observed in AMD and the genotypic correlations between APOE and macular pigment ⁶³ could possibly support the hypothesis of a genetic association between this gene and CNV due to high myopia or AMD. However, associations between AMD and APOE are not consistent ⁴⁸ and the APOE gene is known to be linked to human longevity. ⁶⁴ In a murine model, apoE4 mice showed a more severe AMD-like pathological phenotype and also developed marked CNV, a hallmark of exudative AMD. ⁶⁵ In this study, although we did not find significant evidence supporting the association between E2, E3, and E4 APOE haplotypes and myopic CNV (see Supplementary Material and Supplementary Table S1), we found suggestive evidence of association between myopic CNV and another SNP (rs769455), which is a rare variant of the APOE gene previously implicated with type III hyperlipoproteinemia. ⁵¹ This was not significant after correction for multiple testing, however. Additional studies are required to replicate this result and elucidate the roles of this gene in myopic CNV etiology. 

A limitation of this study is the relatively small sample size, especially for cases with myopic CNV; however, the study is strengthened by at least two factors. First, all participants came from a similar ethnic background, which reduced the chances of heterogeneity in different populations. Second, both the cases and controls were high myopic patients with high myopic genetic profile, which enhanced the ability to detect susceptible loci for myopic CNV. In contrast, comparing myopic CNV patients with nonmyopic controls may be confounded by the genetic and environmental factors influencing the risk of myopia. 

To our knowledge, this study is the first to explore specific genetic effects influencing risk of CNV in high myopic patients compared with controls who are also highly myopic. This study suggests that the inflammatory pathway may be associated with myopic CNV, a vision-threatening complication of high myopia, through CFI. Larger studies are needed to analyze this gene and other candidate loci for this important vision-threatening complication of high myopia. 

Supplementary Materials

doc S1 

Acknowledgments

We thank Joelle Dumas, Jérôme Barré, and Patrick Ledudal for providing technical assistance and Mark Daly for helpful suggestions. 

Footnotes

 Supported in part by Grant RO1-EY11309 from the National Institutes of Health, Bethesda, Maryland; Massachusetts Lions Eye Research Fund, Inc., New Bedford, Massachusetts; an unrestricted grant from Research to Prevent Blindness, Inc., New York, New York to Tufts University School of Medicine; and the Macular Degeneration Research Fund of the Ophthalmic Epidemiology and Genetics Service, New England Eye Center, Tufts Medical Center, Tufts University School of Medicine, Boston, Massachusetts.

Footnotes

⁸  These authors contributed equally to the work presented here and should therefore be regarded as equivalent authors.

Footnotes

 Presented at the annual meeting of the Association for Research in Vision and Ophthalmology in Fort Lauderdale, Florida, May 9, 2012.

Footnotes

 Disclosure: N. Leveziel, None; Y. Yu, None; R. Reynolds, None; A. Tai, None; W. Meng, None; V. Caillaux, None; P. Calvas, None; B. Rosner, None; F. Malecaze, None; E.H. Souied, None; J.M. Seddon, None