A Genome-Wide Association Study Provides Evidence for Association of Chromosome 8p23 (MYP10) and 10q21.1 (MYP15) with High Myopia in the French Population

Abstract

Purpose.: Myopia is a complex trait affected by both genetic and environmental factors. High myopia is associated with increased risk of sight threatening eye disorders such as retinal detachment. The purpose of this genome-wide association study was to identify susceptibility genes contributing to high myopia in the French population.

Methods.: High myopic cases were genotyped using Affymetrix SNP 6.0 chips and population controls were selected from the GABRIEL French dataset, in which samples were genotyped by Illumina Human610 quad array. The association study was conducted using 152,234 single nucleotide polymorphisms that were present on both manufacturers' chips in 192 high myopic cases and 1064 controls to identify associated regions. Imputation was performed on peak regions.

Results.: Associations were found at known myopia locus MYP10 on chromosome 8p23 and MYP15 on chromosome 10q21.1. Rs189798 (8p23), and rs10825992 (10q21.1) showed the strongest associations in these regions (P = 6.32 × 10⁻⁷ and P = 2.17 × 10⁻⁵, respectively). The imputed results at 8p23 showed two peaks of interest. The first spanned 30 kb including rs189798 between MIR4660 and PPP1R3B with the most significant association at rs17155227 (P = 1.07 × 10⁻¹⁰). The second novel peak was 4 kb in length, encompassing MIR124-1 and the MSRA gene, with the strongest association at rs55864141 (P = 1.30 × 10⁻⁷). The peak of imputed data at 10q21.1 was 70 kb in length between ZWINT and MIR3924, with rs3107503 having the lowest P value (P = 1.54 × 10⁻⁷).

Conclusions.: We provide evidence for the association of MYP10 at 8p23 and MYP15 at 10p21.1 with high myopia in the French population and refine these regions of association.

Introduction

Myopia is one of the most common visual disorders in the world. Its prevalence is over 25% in Western Europe and is higher in some East Asian countries. ^1,2 Despite being a non-life threatening disease, myopia presents significant social and economic burdens. In the United States alone, the estimated cost for refractive error (including myopia and hyperopia) in adults over 40 years old was $5.5 billion in 2004. ³ High myopia, normally defined as refractive error less than −6 diopters (D), is an important contributor to legal blindness, particularly in the population of working age. ⁴  

Multiple studies have confirmed that both genetic and environmental factors are involved in the progression of myopia. It is estimated that genetic factors account for over 50% of the population variation. ⁵ So far, 18 loci named MYP1 to MYP18 have been suggested to relate to myopia by genome-wide linkage scans. ⁶ Candidate genes such as PAX6, MYOC, MMP, UMODL1, MFRP genes, and collagen genes have been studied by genotyping intragenic single-nucleotide polymorphisms (SNPs). ⁷ However, no candidate gene has been consistently replicated, and the pathophysiological mechanism of myopia is still not understood fully. 

Since 2009, there have been six genome-wide association studies (GWAS) on myopia or high myopia in a range of populations. The first GWAS on high myopia was reported using Japanese cohorts and a region on 11q24.1 was identified. ⁸ Li et al. ⁹ then located a high myopia locus on 5p15 using Japanese and Singaporean Chinese samples. Furthermore, genetic variants at 4q25 and 13q12.12 were revealed to be associated with high myopia in Han Chinese population. ^10,11 Treating refractive error as a quantitative trait, Soluki et al. ¹² and Hysi et al. ¹³ each discovered a locus for myopia at 15q14 and 15q25 in populations of European descent, respectively. 

In order to identify genetic loci associated with high myopia in French population, we designed this GWAS study using French high myopic patients and French population controls. To our knowledge, this is the first case-control GWAS of myopia in the white European population. 

Materials and Methods

Participants

The 192 high myopic cases used in the study were recruited from the Department of Ophthalmology, Purpan Hospital, Toulouse, France. Values of refractive error and axial length were collected by clinical ophthalmologists. Only high myopic individuals with both eyes less than −6 D were included. All were considered to have axial myopia according to a linear regression analysis, which evaluates the values of refractive error and axial length. ¹⁴ Patients were excluded from the study if there was evidence of any lens or corneal lesions which might affect refractive error or any previous history of ocular procedures which might change refraction. All patients gave informed consent. All DNA samples and data in this study were handled anonymously. The 1064 population controls used in the study were from the GABRIEL French study of asthma, which was obtained through European Genome-phenome Archive. The original GABRIEL dataset contains 1903 individuals, including family members, 839 of whom were removed for relatedness. Full details of the GABRIEL French samples were reported by Moffatt et al. ¹⁵  

This project was approved by the ethics committee of Purpan Hospital, Toulouse, France, and conformed to the tenets of the Declaration of Helsinki. 

Microarray Genotyping and Quality Control

DNA was genotyped using Affymetrix DNA SNP 6.0 chips (Atlas Biolabs, Berlin, Germany). Genotyping was carried out using the Birdseed 2.0 algorithm, which is integrated in Genotyping Console v4.0 (Affymetrix, Santa Clara, CA). Default quality control thresholds were applied. The controls were genotyped using Illumina Human610 (Illumina, Inc., San Diego, CA) quad array by GABRIEL researchers, as described in Moffatt et al. ¹⁵ The data obtained in PLINK (Center for Human Genetic Research, Massachusetts General Hospital, Boston, MA) format already succeed genotyping quality control for Illumina chips (Illumina, Inc.). 

Statistical Analysis

The flowchart for quality control and analysis of association is shown in Figure 1. Samples that passed quality control measures for base-calling underwent further analysis for relatedness, heterozygosity indicative of contamination and low genotyping rate using PLINK v1.07 (Center for Human Genetic Research, Massachusetts General Hospital). Only founders were included, and one of any pair of individuals related by a proportion of greater than 0.05 was removed. Principal components analysis (PCA) of eigenvectors was carried out using EIGENSOFT v3.0 (http://genepath.med.harvard.edu/∼reich/Software.htm, Broad Institute of Harvard and MIT, Cambridge, MA) and individuals with outlying ancestry were excluded from further analyses. SNPs with a genotyping rate of greater than 95%, minor allele frequency greater than 0.01 and a P value for deviation from Hardy-Weinberg equilibrium in controls less than 1×10⁻⁴ were included. SNPs on the X and Y chromosomes and mitochondrial SNPs were excluded from analyses. 

Figure 1. 

 

Flowchart of quality control and analysis of association.

Genome-wide allelic χ² tests were applied to the SNPs that were common to both microarray chips. A P value of less than 10⁻⁴ was considered to be suggestive of association. Due to the risk of systematic genotyping differences between the two manufacturer's platforms, individual SNPs with an association P value of less than 10⁻⁴, but with no supporting neighbor SNPs that were significant at the same level for 10 SNPs upstream or downstream were excluded as probable artifacts. A table listing all SNPs associated with a P value less than 10⁻⁴ is provided (see Supplementary Material and Supplementary Table S2). Imputation of peak regions identified was performed in cases (725,579 SNPs) and controls (536,638 SNPs) separately using SHAPEIT and IMPUTE2, referenced to The 1000 Genomes Project phase I dataset. ^16,17 After merging the datasets, the allelic χ² tests were again performed to check statistical association. Haploview (Broad Institute of MIT and Harvard) was used to generate Manhattan plots. 

Results

The ocular characteristics of the cases are summarized in the Supplementary Table S1 (see Supplementary Material and Supplementary Table S1). The cohort contains 68 female and 124 males with an overall mean age of 51.82 ± 12.31. Three cases failed to pass genotyping quality control. After merging cases and controls and performing PCA, two cases were eliminated for outlying ancestry (leaving 187 cases and 1064 controls). The genomic inflation factor (λ = 1.05) indicated a slight inflation of the GWAS results. The corresponding Q–Q plot generated is shown in Supplementary Figure S1 (see Supplementary Material and Supplementary Fig. S1). 

The number of SNPs in common on the Affymetrix SNP 6.0 chip (Atlas Biolabs) and Illumina Human610 (Illumina, Inc.) quad array was 152,234. Allelic χ² tests were performed using these data and a Manhattan plot was generated by Haploview (Broad Institute of MIT and Harvard) (Fig. 2). A region on chromosome 8p23 was associated with high myopia, and there was support for possible association of a region on chromosome 10q21.1. Table 1 summarizes the SNPs within these regions, which differed in frequency between cases and controls. We imputed these two loci to obtain more information. Despite a peak region between MIR4660 and PPP1R3B with strongest P value of 1.07 × 10⁻¹⁰ at rs17155227, the imputed region of 8p23 revealed a new peak between MIR124-1 and MSRA gene spanning 4 kb with rs55864141 showing the lowest P value of 1.30 × 10⁻⁷ (Fig. 3). The imputed 10p21.1 region showed that the peak remained in the interval between ZWINT and MIR3924, with the strongest association of 1.54 × 10⁻⁷ at rs3107503 (Fig. 4). The top SNPs of imputed 8p23 and 10q21.1 regions are shown in Table 2. 

Figure 2. 

 

Allelic χ²–association tests of common 152,234 SNPs identified peak regions at 8p23 and at 10q21.1.

Figure 3. 

 

Imputed 8p23 region. Distance between MIR4660 and PPP1R3B: 87kb. Distance between MIR124-1 and MSRA: 150kb.

Figure 4. 

 

Imputed 10q21.1 region. Distance between ZWINT and MIR-3924: 943kb.

Table 1. 

 

Top SNPs at 8p23 and 10q13.1 Regions in a GWAS of 152,234 Common SNPs

Region	SNP	Position	Gene	Minor Allele in Case	Frequency in Cases	Frequency in Controls	P Value*	OR
8p23	rs189798	8990577	Intergenic	T	21.1%	34.2%	6.32 × 10−7	0.52
8p23	rs9949	8994512	PPP1R3B	G	20.9%	33.9%	8.61 × 10−7	0.52
8p23	rs19334	9009906	Intergenic	T	25.0%	36.1%	5.44 × 10−5	0.60
8p23	rs4840437	9606264	TNKS	G	26.5%	38.1%	1.67 × 10−5	0.59
8p23	rs6989782	9610538	TNKS	T	19.5%	30.2%	2.43 × 10−5	0.56
8p23	rs656319	9814411	Intergenic	T	32.1%	45.1%	2.69 × 10−6	0.58
8p23	rs6983332	9977918	MSRA	T	34.0%	45.4%	3.90 × 10−5	0.62
8p23	rs12678938	10016969	MSRA	T	53.2%	42.1%	5.99 × 10−5	1.57
10q13.1	rs1394108	58846575	Between	C	39.3%	28.6%	3.26 × 10−5	1.61
10q13.1	rs4589208	58853784	ZWINT	G	35.6%	25.5%	5.27 × 10−5	1.61
10q13.1	rs11005665	58913651	and	G	35.6%	25.1%	2.55 × 10−5	1.64
10q13.1	rs10825992	58948207	MIR3924	G	46.3%	34.8%	2.17 × 10−5	1.61

*  The P values are genomic control adjusted.

Table 2. 

 

Top SNPs at Imputed 8p23 and 10q21.1 Regions

Region	SNP	Position	Gene	Minor Allele in Case	Frequency in Cases	Frequency in Controls	P Value*	OR
8p23	rs17155227	8961990	Between	T	21.5%	9.81%	1.07 × 10−10	2.52
8p23	rs10111263	8969433	MIR4660	G	25.9%	39.4%	5.86 × 10−7	0.54
8p23	rs7830804	8970762	and	G	26.8%	41.2%	1.57 × 10−7	0.52
8p23	rs12155586	8973858	PPP1R3B	A	23.9%	40.6%	2.25 × 10−9	0.50
8p23	rs4841103	8974338		C	23.9%	40.7%	1.87 × 10−9	0.48
8p23	rs1822951	8975103		C	23.7%	40.1%	4.64 × 10−9	0.46
8p23	rs10109064	8977198		A	22.6%	37.6%	1.30 × 10−7	0.49
8p23	rs10092662	8980000		A	23.5%	37.6%	1.58 × 10−7	0.51
8p23	rs189798	8990577		T	21.1%	34.2%	6.32 × 10−7	0.52
8p23	rs13271489	9803712	Between	C	36.4%	51.1%	2.31 × 10−7	0.55
8p23	rs13264066	9804075	MIR124-1	C	35.1%	48.8%	1.26 × 10−6	0.57
8p23	rs9650622	9804292	and	G	35.8%	50.2%	3.30 × 10−7	0.56
8p23	rs56243511	9805695	MSRA	A	35.3%	49.8%	2.20 × 10−7	0.55
8p23	rs58689676	9806492		T	35.2%	49.8%	2.19 × 10−7	0.55
8p23	rs55975067	9806824		C	35.2%	50.0%	1.43 × 10−7	0.54
8p23	rs55864141	9807022		T	35.2%	50.1%	1.30 × 10−7	0.54
10q13.1	rs3105329	58845672	Between	T	37.5%	26.0%	4.12 × 10−6	1.71
10q13.1	rs11814751	58864949	ZWINT	G	33.0%	21.8%	2.86 × 10−6	1.77
10q13.1	rs36042895	58901118	and	C	31.7%	19.9%	1.84 × 10−6	1.87
10q13.1	rs3107503	58918112	MIR3924	C	46.0%	32.1%	1.54 × 10−6	1.80

*  The P values are genomic control adjusted.

Discussion

In this study, we provided additional evidence for the associations of two known myopia loci (MYP10 and MYP15) that were previously reported in family association studies, with high myopia in the French population. 

The 8p23 region was first reported in a genome-wide linkage scan using 506 UK twin samples by Hammond et al. ¹⁸ The region was the second highest peak in this study with a logarithm of the odds (LOD) score of 4.1. Later, Stambolian et al. ¹⁹ also replicated the linkage of this region with myopia in a study involving 34 extended myopic families in the Old Order Amish population. The 8p23 region was named MYP10 by the Hugo Gene Nomenclature Committee. In our study, two sub peaks were identified at 8p23 region. The first spans 110kb from MIR4660 to PPP1R3B. PPP1R3B encodes the catalytic subunit of the serine/theonine phosphatase, protein phosphatase-1 and has been reported to affect circulating lipid levels and C-reactive protein level as well as type 2 diabetes. ^20–22 PPP1R3B has not been reported in relation to any eye disorder, although the gene is expressed in ocular tissue. ²³ MicroRNAs are short non coding RNAs that participate in post transcriptional regulation of gene expression in multicellular organisms by affecting both the stability and translation of mRNAs. ²⁴ MIR4660 is a low abundance microRNA, which was recently identified in breast tumor tissue and its functional role is not clear ²⁵ ; it has not been reported in eye tissues. The second peak in this region encompasses MIR124-1 and MSRA, both of which are eye related genes. MIR124-1 has been studied extensively, and was previously characterized as neuron specific. ²⁶ However, it has been reported to be expressed in the lens, both in animals and in humans. ^27,28 In addition, MIR124 is highly expressed in the human retina and affects retinal gene expression. ^29,30 It can efficiently restrict transgene expression to retinal pigment epithelium in mice and pigs. ³¹ Furthermore, MIR124 regulates early neurogenesis in the optic vesicle, and is essential in regulating the intrinsic temporal changes in retinal ganglion cell growth cone sensitivity. ^32,33 The adjacent gene, MSRA is important for lens cell viability and resistance to oxidative stress. ³⁴ Silencing of the MSRA gene results in loss of mitochondrial membrane potential and increased reactive oxygen species production in human lens cells. ^35,36 Deletion of mouse MsrA can cause hyperbaric oxygen-induced cataract. ³⁷ MSRA may also play an important role in protecting macular retinal pigment epithelial cells from oxidative damage. ³⁸ In GWAS of myopia by Li et al., ⁹ SNPs in the MSRA region showed significant associations with high myopia (the strongest at rs11774836 with P = 5.0 × 10⁻⁴) in Singaporean Chinese, though this group did not carry this association through to their replication study. The peak identified by imputation between MIR124-1 and MSRA is a strong candidate for the functional source of the statistical association. 

Computational analysis of the predicted effect of the associated SNPs using SNP-NEXUS revealed that rs56243511 between MIR124-1 and MSRA is predicted to be a transcription factor binding site for PAX6 in human and in mice. ³⁹ PAX6 gene is important in developing eyes and its protein contains a transcription regulating homeo box domain that binds to DNA. ⁴⁰ PAX6 locates in the 11p13 and was the highest peak in linkage scan in twins by Hammond et al. ¹⁸  

The 10q21.1 region of possible association overlaps the previously reported MYP15 locus. ⁴¹ Nallasamy et al. ⁴¹ first reported this region by using a large Hutterite family from South Dakota of 29 members. However, by directing sequencing the ZWINT and PCDH15 in the 10q21.1 region, they did not find any causative mutations to explain the phenotype. PCDH15 is expressed in the neurosensory epithelium of the eye. ⁴² Immunohistochemistry also showed that PCDH15 is expressed in the photoreceptors of human and monkey retinas, predominantly in the outer photoreceptor segments. ⁴³ Mutations of PCDH15 have been consistently reported to be causative of USHER syndrome type 1F, in which patients are born deaf and develop progressive pigmentary retinopathy leading to blindness in the second to fourth decades of life. ^44,45 To date, other genes in the region including ZWINT, GAPDHP21, MTRNR2L5, and LOC100419872 have no clear evidence relating to ocular diseases. According to our results, the peak region is intergenic between ZWINT and MIR3924 around 110 kb in length. MIR3924 was discovered in a female reproductive tract tissue using next generation sequencing and its main function is not known. ⁴⁶  

Our study may have been insufficiently powered to replicate other loci that have been reported to associate with myopia. We did not replicate the two myopic loci on 15q14 and 15q25 identified by Soluki et al. ¹² and Hysi et al., ¹³ which were conducted in multiple UK and European populations. The two teams applied quantitative methods to identify myopia genes with additive effects and they used general population samples containing hyperopia, normal eye sight, myopia, and high myopia (maximum 5% in the population). We adopted a case-control study approach to high myopia, while these two studies used a cohort approach to investigate the quantitative trait of myopia. The genetic mechanism for high myopia and low myopia may not be exactly the same. ⁴⁷ A recent large meta-analysis confirmed 15q14 as a myopia locus but did not replicate 15q25. ⁴⁸  

Nakanishi et al. ⁸ reported the first case-control GWAS on high myopia and found BLID and LOC399959 in 11q24.1 were associated with high myopia in Japanese. It is of interest to note that miR-125b and let-7 are encoded within LOC399959, and both were found in mammalian and squid lenses. ^27,49 However, the locus failed to be replicated by a Chinese cohort, ⁵⁰ and there is no supporting evidence from our study for this association. Later, multiple GWAS on high myopia in Chinese population have suggested genetic variants at 4q25, 5p15, and 13q12.12 are associated with high myopia. ^9–11 Associations at 4q25 and 5p15 have been successfully replicated by other Chinese teams, but no replication study on 13q12.12 has been published so far. ^51,52 Recently, genetic variants on chromosome 1q41 were identified to influence ocular axial length and high myopia. ⁵³ Shi et al. ⁵⁴ identified that mutations in the ZNF644 might cause monogenic high myopia using exome sequencing. The loci of 11q24.1, 4q25, 5p15, 13q12.12, and 1q41 did not show significance in our French dataset. The different definition of high myopia in these GWAS might represent different subgroups of high myopia and caused various findings. The odds ratio (OR) of reported SNPs in these GWAS indicated quite low or moderate effects on high myopia ranging from 0.74 to 1.71, while in our study, the strongest protective effect could be provided by rs55864141 at 8p23 with an OR of 0.54. 

To increase the likelihood of a positive finding, our case population was enriched for severe or high myopic individuals, and non axial myopic individuals were excluded using linear regression analysis. ¹⁴ The power of this study was low due to limited number of cases, so it is likely that some myopic loci may not be detected. The use of a mixed cohort of asthmatic and non asthmatic controls offers potential for confounding. However, neither 8p23 nor 10p21.1 was reported to be an asthma region by the GABRIEL study. ¹⁵ Due to limited resources, we used French GABRIEL data to be our controls. The same strategy adapted by Nakanishi et al. ⁸ Although two genotyping platforms were adopted in the study, genotyping and post genotyping quality control measures were applied to limit compatibility issues. Principal components analysis showed the case and control populations to be highly compatible. More sporadic positive SNPs were mainly due to compatibility issues between the genotyping platforms. Thus, to avoid false positive findings, we concentrated on regions with multiple clustered positive SNPs. Our study suggests it is reasonable for researchers to use population controls genotyped by a different platform (when data genotyped by the same platform is not available) with a price of increasing the risk of type 2 errors. We found potentially novel associations on regions of chromosomes 9 and 11, but are not currently able to conduct further studies to validate these loci. 

In summary, we provide evidence that MYP10 at chromosome 8p23 is associated with high myopia in the French population. Our findings also support possible association of MYP15 at 10p21.1. We have significantly refined the MYP10 and MYP15 regions of association with myopia. This is the first case-control GWAS of myopia in a population of European descent. The additional evidence and fine localization of these two loci provide new insights into the pathogenesis of high myopia. This study points to the possibility that variation affecting eye-related microRNAs may underlie the aetiology of some high myopia. In the future, candidate genes identified in the two loci, including MIR124-1, should be investigated to better understand their relationships with high myopia. 

Supplementary Materials

pdf S2 

tif S1 

xls S3 

Acknowledgments

The authors thank all the patients who were involved in this study and clinical staff who contributed in sample collection. They also thank the European Genome-phenome Archive, which granted them the access of the GABRIEL French asthma genome-wide dataset. 

Footnotes

 Supported by grants from the European Community's Sixth Framework Programme through the Marie Curie Research Training Network ‘MY EUROPIA' (MRTN-CT-2006-034021).

Footnotes

²  These authors contributed equally to the work presented here and should therefore be regarded as equivalent authors.

Footnotes

 Disclosure: W. Meng, None; J. Butterworth, None; D.T. Bradley, None; A.E. Hughes, None; V. Soler, None; P. Calvas, None; F. Malecaze, None