Haploview version 3.2 plot showing a pairwise linkage disequilibrium map for single-nucleotide polymorphisms within a 100–kilobase (kb) flanking region centered on TGFB1 rs4803455 in the Singapore Cohort Study for Risk Factors in Myopia cohort. The r2 algorithm was used. Single-nucleotide polymorphisms are identified by their dbSNP rs numbers, and their relative positions are marked by vertical lines. Numbers within the diamonds indicate the r2 value expressed as a percentile. Increasing degrees of r2 values are denoted by squares of a darker shade.
Khor CC, Fan Q, Goh L, Tan D, Young TL, Li Y, Seielstad M, Goh DLM, Saw SM. Support for TGFB1 as a Susceptibility Gene for High Myopia in Individuals of Chinese Descent. Arch Ophthalmol. 2010;128(8):1081-1084. doi:10.1001/archophthalmol.2010.149
Zha et al1 reported a haplotype-tagging single-nucleotide polymorphism (SNP) (rs4803455) within the transforming growth factor β1 (TGFB1 [OMIM 190180]) gene and observed evidence of association between it and severe myopia (spherical equivalent [SE] ≤−8.0 diopters [D]) using a case-control sample of Chinese individuals (N = 600; P < .001). They also provided evidence of replication for a previously reported neighboring SNP (rs1800470) in moderate linkage disequilibrium with rs4803455.2 For both SNPs, the minor allele was at decreased risk for high myopia and their effect was strongest within a model of recessive inheritance. Exhaustive analysis by Zha and colleagues showed the best correlate for high myopia susceptibility within TGFB1 to be rs4803455. We thus sought to assess this association within Chinese children in the Singapore Cohort Study for Risk Factors in Myopia.
The Singapore Cohort Study for Risk Factors in Myopia cohort has been previously described.3,4 The students enrolled in the study are examined annually, and serial eye measurements are taken using standardized protocols. These include cycloplegic refraction and axial length measurement of the eyeball. To remove ethnicity as a potential source of population heterogeneity, we only included children of Chinese descent in this genotyping exercise (n = 978). Phenotypic classification of the children into those with severe myopia, those with nonsevere myopia, and nonmyopic controls was made at visit 4 when the children were aged 10 to 12 years. The SE was defined as sphere plus half-negative cylinder. High myopia was defined as an SE of −5.0 D or less; mild to moderate myopia was defined as an SE between −5.0 D and −0.5 D; and nonmyopic controls included those with an SE greater than −0.5 D. The axial length of the globe was measured by contact ultrasound A-scan biometry as previously described.3,4 The SNP rs4803455 was analyzed in an opportunistic but hypothesis-driven manner as the data were available from an ongoing genome-wide association study using the Illumina HumanHap 550 Beadchips (Illumina, Inc, San Diego, California; http://www.illumina.com). Rigorous quality-control steps were performed, including genotype success rate, missingness, population stratification, departure from Hardy-Weinberg equilibrium in controls, monomorphism, excess heterozygosity, cryptic relatedness, and sex discrepancy. Data analysis was performed using SPSS version 17 statistical software (SPSS Inc, Chicago, Illinois). Pairwise linkage disequilibrium between markers was computed based on the squared Pearson correlation coefficient (r2) using the overall data set. We used the linkage disequilibrium information to select a set of 4000 independent autosomal markers (r2 < 0.16) with approximately equal intermarker distance (approximately 670 kilobases [kb]) across the genome. This set of markers was used to examine sample relationships with the Graphical Representation of Relationships program (Center for Statistical Genetics, University of Michigan, Ann Arbor; http://www.sph.umich.edu/csg/abecasis/GRR/) and to examine population structure with Structure software (Department of Statistics, University of Oxford, Oxford, England).5 Additional evaluative testing of population structure was performed with the Eigenstrat method6 using all markers with no Hardy-Weinberg equilibrium deviation and with greater than 1% minor allele frequency. For related samples identified by Graphical Representation of Relationships, we retained those with a higher call rate. Meta-analysis of all available data was performed using inverse-variance weighting as previously described.7
We observed a marginally significant overall genotypic association at TGFB1 rs4803455 when children with any myopia vs nonmyopic children were compared (n = 348 controls, 630 cases; P = .01) (Table). When analyzed specifically for high myopia, the association became more pronounced (n = 348 controls, 107 cases; P = .007). Children who were recessive for the minor T allele were markedly underrepresented among the high myopia cases (7.5%) compared with nonmyopic controls (14.9%) (P = .046) (Table). The relationship between recessivity of the minor T allele and myopia-related quantitative traits in the entire Singapore Cohort Study for Risk Factors in Myopia cohort (n = 978) was then assessed. Children who were recessive for the minor T allele had less myopic SEs (mean, −1.54 D) and shorter axial lengths (mean, 23.94 mm) compared with wild-type children (mean SE, −1.95 D; mean axial length, 24.15 mm) (P = .04 and P = .05, respectively).
When assessed in the context of the entire genome-wide association study, 2618 SNPs had single-point P values that exceeded the significance of TGFB1 rs4803455 (lowest P = 7.9 × 10−6 at chromosome 11). Full correction for multiple testing in light of more than 400 000 independent tests rendered the TGFB1 rs4803455 observation insignificant.
However, as our study was hypothesis-based with regard to TGFB1, we proceeded to assess TGFB1 rs4803455 in the context of the nearby flanking SNPs that were also genotyped. A linkage disequilibrium plot (Figure) using 32 SNPs genotyped within the 100-kb flanking region localized rs4803455 to a small block that included 3 other neighboring SNPs (rs2241715, rs2241714, and rs1046909). All 3 showed marginal evidence of association (P = .06, P = .04, and P = .02, respectively), which did not exceed that observed with rs4803455 (P = .007). No other SNP within the 100-kb flanking region upstream and downstream of TGFB1 rs4803455 showed any evidence of association, thus localizing the signal to TGFB1.
Zha et al1 identified a new marker within TGFB1 that proved to be more informative in predicting the risk of high myopia in individuals of Chinese descent over and above the previously reported rs1800470.2 We provide evidence in another Chinese population in support of the observations by Zha and colleagues. We show new data linking TGFB1 rs4803455 and myopia-related quantitative traits. When our data are interpreted in the context of earlier results,1,2 association of TGFB1 rs4803455 was reproduced in both the broad genotypic model and specific recessive models, thus showing further consistency with previous studies. Even if this is not the functional variant(s) involved, the TGFB1 SNP rs4803455 is very likely the closest correlate. We are mindful of 2 previous reports that show no association between TGFB1 genetic variation and high myopia.8,9 Although rs4803455 was not genotyped in these 2 reports, Wang et al8 did genotype rs1800470, which was moderately correlated with rs4803455 (r2 = 0.56), and failed to observe a significant association (recessive: odds ratio = 0.83; 95% confidence interval, 0.56-1.23). Despite not being significant, the disease odds ratio observed by Wang and colleagues is in keeping with that observed by the 2 previous studies,1,2 and meta-analysis of all available studies1,2,9 on the rs1800470 variant revealed suggestive evidence of an association with reduced susceptibility to high myopia (P = 3.0 × 10−4). Meta-analysis of all available data (by Zha and colleagues and our study) on rs4803455 resulted in a slightly stronger effect estimate (P = 9.88 × 10−5).
We acknowledge that the association observed with TGFB1 rs4803455 is not significant after being subjected to more than 400 000 independent tests and that the meta-analysis does not yield a genome-wide significant association. However, given the consistent evidence for the role of TGFB1 and high myopia1,2 and considering this in light of new evidence by Zha and colleagues and our study, the overall evidence thus far suggests support for an association between TGFB1 and high myopia. Additional samples should be genotyped for the SNPs that have been implicated in this locus to increase power to detect a significant effect of this locus.
Correspondence: Dr Khor, Division of Infectious Diseases, Genome Institute of Singapore, 60 Biopolis St, Genome, Singapore 138672 (firstname.lastname@example.org).
Financial Disclosure: None reported.
Funding/Support: This work was supported by grant 06/1/21/19/466 from the Biomedical Research Council, Agency for Science, Technology, and Research (a statutory board of the Singapore government) (Dr Saw). Dr Khor is a scholar of the Agency for Science, Technology, and Research.
Additional Contributions: We are grateful to all of the participants involved in the Singapore Cohort Study for Risk Factors in Myopia study cohort.