A Functional Polymorphism at 3′UTR of the PAX6 Gene May Confer Risk for Extreme Myopia in the Chinese

Abstract

Purpose.: The paired box 6 (PAX6) is involved in eye development and associated with several ocular diseases. Conflicting results have been reported regarding the association between PAX6 polymorphism and myopia. This case–control study and functional assay were conducted to identified a functional risk polymorphism for myopia.

Methods.: The study cohort included 1083 cases (≤ −6.0 D) and 1096 controls (≥ −1.5 D) from a Chinese population residing in Taiwan. Four common tag single-nucleotide polymorphisms (SNPs) and an SNP at the 3′ untranslated region (UTR) were selected. Secondary analyses were conducted in which cases and controls were redefined based on different spherical refractions. Permutation was used to adjust for multiple testing. The luciferase reporter assay was conducted for the 3′UTR SNP to assess the allelic effect on gene expression.

Results.: SNPs rs644242 and rs662702 had marginal significance (P = 0.063), and further analyses showed that these SNPs were associated with extreme myopia (≤ −11 D). The OR for extreme myopia was 2.1 (empiric P = 0.007) for the CC genotype at SNP rs662702 at the 3′UTR. The functional assay for SNP rs662702 demonstrated that the C allele had a significantly lower expression level than did the T allele (P = 0.0001). SNP rs662702 was predicted to be located in the microRNA-328 binding site, which may explain the differential allelic effect on gene expression.

Conclusions.: In this study, a functional SNP was identified at the 3′UTR that influences the risk for extreme myopia. The functional assay suggested that the risk allele can reduce PAX6 protein levels which significantly increases the risk for myopia.

Myopia is a common eye condition worldwide, and its prevalence varies widely among populations and ages and between the sexes. ^{1 –3} Myopia is extremely common in Taiwan. When the definition of < −6 D is used, the prevalence of high myopia is 18% among young Taiwanese men and 24% among young Taiwanese women ³ ; both rates are even higher than the 13.1% reported among young men in Singapore. ² Furthermore, the frequency of high myopia (< −6.0 D) has increased in young Taiwanese people: 10.9% in 1983 and 21% in 2000. ⁴ Although studies have found several environmental risk factors, twin studies have indicated a strong genetic influence on refractive errors, with estimates of heritability between 58% and 90%. ^{5 –8} Several studies have also shown that a family history of myopia is a significant risk factor. ^{9 –13} Recently, studies have reported several genes that increase susceptibility to nonsyndromic myopia, but most of the findings could not be replicated. ^{14 –20} Genetic association studies are subject to type I error, especially when the sample size is small. Therefore, replication of the genetic effects in large and independent samples is an important way to reduce false-positive findings. 

The paired box 6 (PAX6) gene belongs to a highly conserved family of transcription factors containing the paired and homeobox DNA-binding domains. PAX6 is involved in the development of the central nervous system and the eye. It plays significant roles during the induction of lens and retina differentiation, and has been considered the master gene in eye development. ²¹ In humans, mutations in PAX6 are associated with a variety of human ocular diseases including aniridia, foveal hypoplasia, presenile cataract, and aniridia-related keratopathy (reviewed by Tsonis and Fuentes ²¹ ). In addition to the biological plausibility, a genome-wide linkage study ¹⁴ revealed a strong linkage of refractive error to the PAX6 locus. Accordingly, PAX6 has been proposed as a candidate gene for the development of myopia. 

Conflicting data have been reported from recent genetic association studies regarding the association between PAX6 and myopia. ^{14 –16,22 –24} To further clarify this relationship, we conducted a large case–control study of more than 2000 subjects and a functional assay, to obtain a more reliable result. 

Materials and Methods

Subjects

The present study participants were enrolled from the general population between the ages of 16 and 45 years. The enrollment was conducted in southern Taiwan between 2003 and 2009. All the participants were of Chinese descent. All the cases had myopia in both eyes and had a spherical refraction ≤ −6.0 D in at least one eye. A subject with a spherical refraction ≥ −1.5 D in the more myopic eye was defined as a control. We used negative cylindrical powers in all subjects. In addition, none of the controls had received any previous refractive surgery. The refractive error was measured without cycloplegia in subjects aged ≥18 years and with cycloplegia in those aged <18 years. The refractive error was measured with autorefractometers (KR-8100 or RM-8800; Topcon, Tokyo, Japan) for all eyes. A written informed consent was given by each subject or custodian (if the age of the participant was <18 years). The study was approved by the Institutional Review Board at the Kaohsiung Medical University Hospital. The research adhered to the tenets of the Declaration of Helsinki. 

SNP Selection and Genotyping

We first selected the tag single-nucleotide polymorphisms (tSNPs) at the PAX6 gene from the release 2.0 phase II data of the HapMap Project (www.hapmap.org), ²⁵ using the Tagger Pairwise method. ²⁶ tSNPs were chosen according to the following criteria: r ² ≥ 0.8 and the minor allele frequency (MAF) ≥ 10% in the Han Chinese population. Four tSNPs met the selection criteria: rs628224 (intron 7), rs644242 (intron 11), rs2071754 (intron 11), and rs3026393 (intron 12). tSNP rs644242 was found to be significant, and we then searched for potential functional SNPs that are tagged by rs644242. We found that SNP rs662702 at the 3′UTR is tagged by rs644242. Because the 3′UTR can be related to mRNA stability, we then included SNP rs662702 and tested for its association with high myopia. Genotyping was performed by using gene analysis technology (TaqMan; Applied Biosystems, Inc. [ABI], Foster City, CA). Briefly, PCR primers and minor groove binder (MGB) probes were designed, and reactions were performed in 96-well microplates on a thermal cycler (model 9700; ABI). Fluorescence was measured with real-time PCR (model 7500; ABI) and analyzed with allied software (System SDS, ver. 1.2.3; ABI). 

Statistical Analysis for Genetic Polymorphism Studies

The allele frequency was obtained by direct gene counting. Hardy-Weinberg equilibrium (HWE) was tested in the controls ²⁷ by using the χ² test for each SNP. According to the myopic status and three genotypes of each SNP, the χ² test for a 2 × 3 contingency table or Fisher exact test was performed. The genotype specific odds ratio (OR) was first checked to test for the allele dominance. If one allele is dominant over the other allele, the two genotypes containing the dominant allele is combined to increase the statistical power. When there is no evidence of dominance, we prefer to not collapse heterozygotes with minor homozygotes unless the number of minor homozygotes is too small. Because the genetic effect may be more prominent for extreme myopia as indicated by a previous study, ¹⁵ we also conducted exploratory analyses by reclassifying cases with different refractive errors. 

To account for multiple testing, we also performed permutations to obtain an empiric P value. In the present study, we used the minP permutation procedure to calculate the empiric P values. The minP is a distribution-free statistic that uses random label-swapping to generate the null distribution in a population. The null distribution was obtained with the following steps: (1) swapping refractive errors randomly among all study subjects, (2) performing χ² tests (or the Fisher test, if a cell contains four or fewer samples) on the permuted data based on all redefined case and control status using different refractive error cutoffs to obtain the minimum P value of each run, and (3) repeating steps 1 and 2 to obtain 10,000 permutations. The empiric P values were calculated by the following formula: (R + 1)/(N + 1), where R is the number of runs in which the minimum P value is less than the observed, and N is the number of permutations. We performed the permutation only for the most significant and biologically relevant SNP, rs662702, since other SNPs were either not significant in the first place or were unlikely to be causal variants. 

Construction of Reporter Plasmids

The luciferase assay was used as the reporter system to assess functional consequences for the significant 3′UTR SNP rs662702. We first synthesized double-stranded oligonucleotides that contain 17 bp surrounding SNP rs662702 in three tandem copies and used the restriction enzymes SpeI and MluI for cloning sites. The oligonucleotides were cloned into a reporter vector (pMIR-REPORT miRNA Expression Reporter Vector System; ABI) by T4 DNA ligase (New England BioLabs, Boston, MA). One reporter construct carries risk allele C (denoted as rs662702C), and the other carries protective allele T (denoted as rs662702T). The primer sets are: rs662702C F 5′-TTACTAGTGTGGATCCTAGCCCTTCTCTGACAGGTTAGCCCTTCTCTGACAGGTTAGCCCTTCTCTGACAGACGCGTTT-3′, R 5′-AAACGCGTCTGTCAGAGAAGGGCTAACCTGTCAGAGAAGGGCTAACCTGTCAGAGAAGGGCTAGGATCCACACTAGTAA-3′; and rs662702T F 5′-TTACTAGTGTGGATCCTAGCCCTTTTCTGACAGGTTAGCCCTTTTCTGACAGGTTAGCCCTTTTCTGACAGACGCGTTT-3′, R 5′-AAACGCGTCTGTCAGAAAAGGGCTAACCTGTCAGAAAAGGGCTAACCTGTCAGAAAAGGGCTAGGATCCACACTAGTAA-3′. All the constructs were confirmed by DNA sequencing. 

Transient Transfection and Luciferase Assays

The human retinal pigment epithelial (RPE-19) cell line (American Type Culture Collection, Manassas, VA) was cultured in Dulbecco's modified Eagle's medium (Gibco-BRL, Gaithersburg, MD) supplemented with 10% bovine fetal serum (Gibco-BRL), 2 mM l-glutamine (Sigma-Aldrich, St. Louis, MO), and 1 mM pyruvate (Sigma). The cells were maintained at 37°C in an atmosphere of 5% CO₂. The constructs (400 ng) were transfected into the RPE-19 cells (Lipofectamine 2000; Invitrogen, Carlsbad, CA). The luciferase assay was then performed according to the manufacturer's protocol. The cells were lysed in the passive lysis buffer (Promega, Madison, WI) for 24 hours after transfection, then luciferase activity was measured (Luciferase Assay System; Promega). The pEGFP-C1 vector was also co-transfected into the RPE-19 cells, and the fluorescence level was used as the internal control. Each experiment was independently repeated three times, and each sample was studied in triplicate. A two-sided P < 0.05 was considered significant, according to the results of the Mann-Whitney test. 

Results

A total of 1083 cases and 1096 controls were included in the present study. The mean age was 21.3 years for cases and 21.0 years for controls. The spherical refractions ranged from −6.0 to −17.0 D, with a mean of −8.0 D and an SD of 1.8 D for cases. For the controls, the spherical refractions ranged from 0.75 to −1.5 D, and the mean ± SD was −0.4 ± 0.6 D. The call rate ranged from 93% to 97%. 

Single SNP Results

The frequencies of the genotypes and the associations between high myopia (≤ −6.0 D) and the four tSNPs are shown in Table 1. All the SNPs were in HWE in the controls, (Table 1), and the linkage disequilibrium plot is shown in Figure 1. The difference in genotype distribution between cases and controls was marginally significant (P = 0.063 from the χ² analysis) for SNP rs644242. We further explored analyses of this SNP by changing the cutoff values to redefine cases and controls. The results showed that the frequency of the common C allele as well as the CC genotype increased with the severity of myopia (Table 2). The most significant result was from the comparison between subjects with spherical refractions ≤ −11 and ≥ −0.5 D. The CC genotype had an OR of 2.3 (nominal P = 0.009) for extreme myopia when compared to the combination of AA and AC genotypes. 

Table 1.

 

Four tSNPs and Their Relationships to High Myopia

SNP (Major/Minor)	Major Homozygote	Heterozygote	Minor Homozygote	MAF (%)	P *	Call Rate	P †
rs628224 (G/A)
Case	516 (52.0)	390 (39.3)	87 (8.8)	28.4	0.61	92
Control	497 (50.4)	391 (39.7)	98 (9.9)	29.8		90	0.11
rs644242 (C/A)
Case	625 (62.8)	301 (30.2)	70 (7.0)	22.1	0.063	92
Control	622 (61.2)	343 (33.8)	51 (5.0)	21.9		93	0.68
rs2071754 (C/T)
Case	316 (30.2)	485 (46.3)	247 (23.6)	46.7	0.75	97
Control	311 (29.4)	507 (47.9)	240 (22.7)	46.6		97	0.23
rs3026393 (A/C)
Case	325 (33.0)	466 (47.4)	193 (19.6)	43.3	0.98	91
Control	332 (32.6)	485 (47.6)	202 (19.8)	43.6		93	0.30

 Data are the number of cases/controls (% of total group). Refraction in cases was ≤ −6 D and in controls was ≥ −1.5 D.

*   P for the comparison of the genotype distribution between cases and control (2 df).

†  Hardy-Weinberg test.

Figure 1.

 

The linkage disequilibrium plots for the five SNPs. Top plot shows the D′ and the bottom plot r ² (%) between any pair of SNPs. The dark gray cells in the top plot indicate strong linkage disequilibrium and a log of odds (LOD) of 2 or greater. In the bottom panel, the gray-scale spectrum indicates pairwise r ² values ranging from black (r ² = 1) to white (r ² = 0).

Table 2.

 

The Genotype Distribution of rs644242 According to Different Refractive Errors

Refractive Error	CC	CA	AA	Freq. of C Allele (%)	P *	OR (P)†
Controls
−0.5 D (Reference)	374 (60.4)	216 (34.9)	29 (4.7)	77.9		1.0
−1.0 D	527 (60.9)	293 (33.8)	46 (5.3)	77.8
−1.5 D	622 (61.2)	343 (33.8)	51 (5.0)	78.1
Cases
−6 D	625 (62.8)	301 (30.2)	70 (7.0)	77.9	0.042	1.1 (0.348)
−7 D	433 (63.8)	199 (29.3)	47 (6.9)	78.4	0.039	1.1 (0.213)
−8 D	297 (65.3)	129 (28.4)	29 (6.4)	79.5	0.053	1.2 (0.104)
−9 D	154 (66.7)	65 (28.1)	12 (5.2)	80.7	0.176	1.3 (0.094)
−10 D	84 (68.9)	30 (24.6)	8 (6.5)	81.1	0.076	1.4 (0.079)
−11 D	43 (78.2)	9 (16.4)	3 (5.4)	86.4	0.012‡	2.3 (0.009)

 Data are the number of individuals (% of the total group). Freq., frequency.

*   P value for 2 × 3 table to compare the genotype distribution between cases and controls (≥ −0.5D).

†  OR for CC versus (CA + AA) using subjects with refraction ≥ −0.5D as controls.

‡  Fisher exact test.

Since rs644242 is a tSNP without prior known or predictable function, we searched for potentially functional SNPs that are tagged by significant SNP rs644242. From the HapMap data, we found that SNP rs662702 (r ² = 0.88 with rs644242) at the 3′ UTR appeared to be a good candidate, because variants at the 3′UTR may influence mRNA stability, and SNP at this region may have functional consequences. Using the same statistical approach, this 3′UTR SNP revealed a result similar to that obtained for tSNP rs644242, and the risk allele C was significant only for extreme myopia (Table 3). To adjust for multiple testing, we performed permutation and analyzed the data, using different cutoffs of spherical refraction for both cases and controls. Therefore, a total of 18 (three definitions of controls and six definitions of cases) association tests were examined for set of permuted data. The empiric P value was 0.0448 for the association between extreme cases (≤ −11 D) and controls (≥ −0.5D) for rs662702. Therefore, our result remains significant after a family-wise multiple testing correction. 

Table 3.

 

The Genotype Distribution of rs662702 According to Different Refractive Errors

Refractive Error	CC	CT	TT	Freq. of C Allele (%)	Nominal P *	Empirical P *
−0.5 D	368 (60.4)	212 (34.8)	30 (5.0)	77.8
−1.0 D	521 (60.9)	292 (34.2)	43 (5.1)	78
−1.5 D	615 (61.4)	338 (33.8)	49 (4.9)	78.3
−6 D	590 (61.7)	300 (31.4)	66 (6.9)	77.4	0.1489	0.5109
−7 D	407 (62.3)	204 (31.2)	42 (6.4)	77.9	0.2651	0.7351
−8 D	284 (64.7)	127 (28.5)	28 (6.4)	79.2	0.1082	0.3939
−9 D	149 (66.8)	62 (27.8)	12 (5.4)	80.3	0.1673	0.5502
−10 D	84 (68.9)	30 (24.6)	8 (6.6)	81.1	0.0863	0.3005
−11 D	44 (77.2)	9 (15.8)	4 (7.0)	85.1	0.0074†	0.0448

 Data are the number of individuals (% of the total group). Individuals with refractive errors ≥ −0.5 D were used as the reference group. Freq., frequency.

*   P value for 2 × 3 table to compare the genotype distribution between cases and controls (≥ −0.5 D).

†   P value was calculated by the Fisher exact test.

Luciferase Assays

To understand the functional significance of rs662702C/T change, we used a reporter assay with luciferase. As shown in Figure 2, the rs662702T clone had a significantly higher expression level of firefly luciferase than did the rs662702C clone (P = 0.0001). These results suggested that the risk C allele had lower protein expression levels than did the protective T allele. 

Figure 2.

 

(A) The plasmid construct carries the CMV promoter region in a 5-to-3 orientation in the pMIR-REPORT luciferase vector. Double-stranded oligonucleotides (containing rs662702) were cloned into the CMV promoter–luciferase plasmid to construct two different reporter constructs carrying the protective T allele or risk C allele. (B) Luciferase activity of the two reporter constructs that include SNP rs662702 in the RPE-19 cell line. Relative change (P = 0.0001) was measured as luciferase activities of the reporter constructs, using the data (mean ± SD) from three independent transfection experiments.

Discussion

We systematically investigated four tSNPs at PAX6 in a large Chinese population residing in Taiwan. The common CC genotype of SNP rs644242 was significant when we analyzed subjects who had extreme myopia (≤ −11 D). Further analysis showed that the functional SNP rs662702 at 3′UTR tightly linked to the tag SNP rs644242 was also significant. The risk C allele of SNP rs662702 had a lower expression level than did the protective T allele. The frequency of the risk CC genotype of SNP rs662702 increased with the severity of myopia in our population, which indicates a genetic dose-dependent effect. The present study demonstrates that a novel functional SNP at PAX6 may influence an individual's susceptibility to extreme myopia. Our finding for the PAX6 gene is consistent with two previous reports on high ^16,24 and extreme myopia ¹⁵ in the Chinese population. 

Although 3′UTR SNP rs662702 would not alter the structure of the PAX6 protein, it may influence stability of PAX6 mRNA. Recent evidence implies that a microRNA can act through base pairing to the complementary segment within the 3′UTR mRNA of the target gene leading to translation inhibition and/or mRNA degradation. MicroRNAs are evolutionarily conserved, small, noncoding RNAs known to regulate posttranscriptional expression of target genes. We used microRNA software, including miRanda, ²⁸ PITA, ²⁹ and RNA22, ³⁰ to search for potential microRNAs that target the region where SNP rs662702 is located, and microRNA-328 (miR-328) was predicted by all software. miR-328 was reported to be expressed in many tissues, although its expression in the eye was not investigated. ³¹ The position of rs662702 is next to the seed region of miR-328 and the C-to-T substitution leads to a mismatch between miR-328 and PAX6 mRNA paring. Our functional assay indicates that the risk C allele causes a lower PAX6 expression level than the T allele, which can be due to the knockdown effect by miR-328 on the C allele. If so, SNP rs662702 also belongs to a new category of SNPs called mirSNP. ³² A decreased expression of PAX6 can result in an increased expression of MMP2, ³³ and an increase of MMP2 expression has been implicated in the pathogenesis of myopia. ³⁴ Accordingly, our initial finding based on the statistical significance can be further supported by these functional studies. 

The relationship between PAX6 expression levels and myopia is inconsistent. Both human and animal studies have suggested a decrease in PAX6 level in myopia. A recent study reported that the PAX6 mRNA level in chicken retinas decreased during form-deprivation myopia but did not have a significant effect on the negative-lens–induced myopia. ³⁵ Another line of evidence to support a low PAX6 level as a cause of myopia is from the human mutation analysis ²² in which PAX6 haploinsufficiency was found in extreme myopia. However, conflicting observations have also been reported. A recent paper showed that longer lengths of dinucleotide repeats in the promoter of PAX6 increase gene expression and are statistically associated with high myopia in Chinese. ²⁴ Zhong et al. ³⁶ studied rhesus monkeys and reported a small but significant upregulation in PAX6 mRNA after negative-lens–induced myopia, but no change in PAX6 mRNA in form-deprivation myopia. Therefore, the role of PAX6 in myopia may depend on the species and the methods of myopia induction. However, the statistical results in the present study should be replicated in large and independent samples. Without further validation, readers are advised to regard the current results with caution. 

Tsai et al. ¹⁵ investigated a Chinese population residing in Taiwan and reported a significant association between the common C allele of rs667773 at PAX6 and extreme myopia (≤ −10 D), whereas no significant association was found for high myopia, defined as ≤ −6.0 D. Although we did not genotype rs667773, it can be tagged by our initial tSNP rs644242 (r ² = 0.92) and our functional SNP rs662702 (r ² = 0.86) according to the Han Chinese data in HapMap. In concert with the finding of Tsai et al., our data suggest that both rs644242 and rs662702 are more related to extreme myopia but less associated with high myopia. We conducted a meta-analysis using data from Tsai et al., and the present studies, and the results showed that the common alleles had an OR of 2.9 (P < 0.0001) for extreme myopia in the Chinese population. Using the family data set, the Han et al. ¹⁶ study revealed a weak association (P = 0.0458) between rs667773 and high myopia (≤ −6.0 D). However, they did not specifically examine the association with extreme myopia. On the other hand, they did report the most significant SNP, rs3026393, with a P < 0.001, which was not significant in our study. 

The PAX6 gene is a strong candidate gene for myopia according to both biological and linkage data. The first statistical evidence to support PAX6 as a candidate gene was from the genomewide linkage study in a British population. ¹⁴ However, their study did not find any SNP at PAX6 associated with the refractive error in the same study subjects. A subsequent study ²³ failed to show either linkage or association between PAX6 and myopia in subjects of different ethnic groups. Similarly, another study using 596 British subjects could not demonstrate an association between PAX6 SNPs and myopia. ³⁷ However, including the present study, a total of four independent studies indicate genetic polymorphisms of the PAX6 gene are likely to influence susceptibility to high or extreme myopia in the Chinese. ^15,16,24 It is likely that PAX6 is only involved in severe myopia, which is more prevalent in the Chinese. Notably, induced myopia in the above-mentioned experimental studies ^35,36 also had different findings regarding PAX6 expression level in myopia. Cross-population studies are needed to elucidate whether PAX6 also has differential effects on different ethnic groups. 

Our study design has strengths and limitations. We used a large number of study subjects to reduce both type I and II error rates. The statistical finding is further supported by the functional experiments, and the significant 3′UTR SNP can be one of the causal variants. Our study population was relatively homogenous in terms of ethnicity, geographic location, and age. As a result, several undetectable confounding factors can be minimized. A previous study investigating the population admixture of Taiwanese also indicated high homogeneity among the Taiwanese subpopulations. ³⁸ In addition, we have assessed population stratification in subjects recruited from the same geographic locations by our team, and the inflation factor was 1. ³⁹ Therefore, population stratification is unlikely to be an explanation for our association results. Using permutation, we also found a significant P value under the consideration of multiple testing. However, the main finding of a significant association was based on only 55 cases of extreme myopia. A meta-analysis is warranted to provide a sufficient power to validate our finding. 

In conclusion, we conducted a large-scale study to systematically evaluate the genetic effect of the PAX6 gene. We found that the functional SNP rs662702 at the 3′UTR is associated with extreme myopia but not with high myopia among Chinese living in Taiwan. This functional SNP is located at the miR-328 binding site, and the risk allele may be downregulated by miR-328, leading to a lower level of PAX6 protein. 

Footnotes

 Supported by Grant NSC95-3112B037-003 from the Taiwan National Science Council.

Footnotes

 Disclosure: C.-L. Liang, None; E. Hsi, None; K.-C. Chen, None; Y.-R. Pan, None; Y.-S. Wang, None; S.-H.H. Juo, None