Genome-wide association study and meta-analysis of intraocular pressure

Abstract

Elevated intraocular pressure (IOP) is a major risk factor for glaucoma and is influenced by genetic and environmental factors. Recent genome-wide association studies (GWAS) reported associations with IOP at TMCO1 and GAS7, and with primary open-angle glaucoma (POAG) at CDKN2B-AS1, CAV1/CAV2, and SIX1/SIX6. To identify novel genetic variants and replicate the published findings, we performed GWAS and meta-analysis of IOP in >6,000 subjects of European ancestry collected in three datasets: the NEI Glaucoma Human genetics collaBORation, GLAUcoma Genes and ENvironment study, and a subset of the Age-related Macular Degeneration-Michigan, Mayo, AREDS and Pennsylvania study. While no signal achieved genome-wide significance in individual datasets, a meta-analysis identified significant associations with IOP at TMCO1 (rs7518099-G, p = 8.0 × 10(-8)). Focused analyses of five loci previously reported for IOP and/or POAG, i.e., TMCO1, CDKN2B-AS1, GAS7, CAV1/CAV2, and SIX1/SIX6, revealed associations with IOP that were largely consistent across our three datasets, and replicated the previously reported associations in both effect size and direction. These results confirm the involvement of common variants in multiple genomic regions in regulating IOP and/or glaucoma risk.

Fig. 1

Relationship of the untreated and treated IOP values in two sample sets from NEIGHBOR. a The 114 cases from the CIGTS set, consisting of 30 cases treated surgically, with or without medication, and 84 cases treated with medication. b The 117 cases from the University of Michigan set: 70 cases treated surgically (shown in red, with regression line y = 0.59x; Pearson's correlation coefficient r = 0.31) and 47 cases treated with medication (shown in black, with y = 0.69x; r = 0.46). The observed relationship of the untreated and treated IOP values, with a slope of 0.67, justified the extrapolation procedure we adopted, and as reported previously (van der Valk et al. 2005). The regression equations and Pearson's correlation coefficients are shown

Fig. 2

NEIGHBOR, GLAUGEN, and AMD-MMAP MI meta-analysis results over ∼467,000 SNPs. a Genome-wide −log₁₀ (p value) plot. The horizontal line marks the 1.1 × 10⁻⁷ threshold of genome-wide significance. b Quantile–quantile plot of observed vs. expected −log₁₀ (p value) of meta-analysis without correction by genomic control. c–g Regional plots for TMCO1, CDKN2B-AS1, GAS7, CAV1/CAV2 and SIX1/SIX6, respectively. The lead SNP in each region is indicated by the solid diamond

Fig. 2

NEIGHBOR, GLAUGEN, and AMD-MMAP MI meta-analysis results over ∼467,000 SNPs. a Genome-wide −log₁₀ (p value) plot. The horizontal line marks the 1.1 × 10⁻⁷ threshold of genome-wide significance. b Quantile–quantile plot of observed vs. expected −log₁₀ (p value) of meta-analysis without correction by genomic control. c–g Regional plots for TMCO1, CDKN2B-AS1, GAS7, CAV1/CAV2 and SIX1/SIX6, respectively. The lead SNP in each region is indicated by the solid diamond

Fig. 3

Comparison of the effect size and direction across datasets for the top SNPs in our meta-analysis in TMCO1, CDKN2B-AS1, GAS7, CAV1/CAV2, and SIX1/SIX6. In these “forest plots”, point sizes are proportional to the weights (sample sizes) used in the meta-analysis; and the horizontal lines indicate the 95 % confidence interval of regression beta values. a TMCO1 top SNP, rs7518099(G), b CDKN2B-AS1 top SNP, rs1412829(G), c GAS7 top SNP, rs12150284(A), d CAV1/CAV2 top SNP, rs4236601(A), e SIX1/SIX6 top SNP, rs10483727(A). In a and c, the beta values from van Koolwijk et al. (2012) are included

Fig. 4

Comparison of the effect size and direction across datasets for SNPs significantly or suggestively associated with IOP in van Koolwijk et al. (2012). a rs7555523(C) at TMCO1, b rs11656696(A) at GAS7, c rs1826598(A) at the ADAMTS18 and NUDT7 locus, d rs9841621(G) at SATB1, e rs21117760(A) at FOXP1. Some of the SNPs were not genotyped in NEIGHBOR and/or GLAUGEN and are not shown for these datasets

Fig. 5. Comparison of the effect size and direction across datasets for the top SNPs in two regions significantly associated with POAG in Burdon et al. (2011). a TMCO1, rs4656461(G), b TMCO1, rs7518099(G), c CDKN2B-AS1, rs4977756(G), d CDKN2B-AS1, rs10120688(G)

Fig. 6. Comparison of the effect size and directions across datasets for two lead SNPs in the CAV1/CAV2 region significantly associated with POAG in Thorleifsson et al. (2010). a rs4236601(A), b rs1052990(G)

Fig. 7. Comparison of the effect size and directions across datasets for SNPs significantly associated with POAG in Wiggs et al. (2012b). CDKN2B-AS1 SNP, rs2157719(G). The other signal, at SIX1/SIX6, rs10483727(A), is shown in Fig. 3e