New insights into the genetics of primary open-angle glaucoma based on meta-analyses of intraocular pressure and optic disc characteristics

Abstract

Primary open-angle glaucoma (POAG), the most common optic neuropathy, is a heritable disease. Siblings of POAG cases have a ten-fold increased risk of developing the disease. Intraocular pressure (IOP) and optic nerve head characteristics are used clinically to predict POAG risk. We conducted a genome-wide association meta-analysis of IOP and optic disc parameters and validated our findings in multiple sets of POAG cases and controls. Using imputation to the 1000 genomes (1000G) reference set, we identified 9 new genomic regions associated with vertical cup-disc ratio (VCDR) and 1 new region associated with IOP. Additionally, we found 5 novel loci for optic nerve cup area and 6 for disc area. Previously it was assumed that genetic variation influenced POAG either through IOP or via changes to the optic nerve head; here we present evidence that some genomic regions affect both IOP and the disc parameters. We characterized the effect of the novel loci through pathway analysis and found that pathways involved are not entirely distinct as assumed so far. Further, we identified a novel association between CDKN1A and POAG. Using a zebrafish model we show that six6b (associated with POAG and optic nerve head variation) alters the expression of cdkn1a. In summary, we have identified several novel genes influencing the major clinical risk predictors of POAG and showed that genetic variation in CDKN1A is important in POAG risk.

Introduction

In primary open-angle glaucoma (POAG), loss of retinal ganglion cells and nerve fibres manifests itself clinically as optic nerve damage, which leads to visual field loss and, eventually in 15% of these to visual impairment and blindness (1–3). The optic nerve damage is characterized by an increase in cup area, the central portion of the optic nerve head (or optic disc), and/or a decrease in the rim area (the area of the disc occupied by the retinal nerve fibre axons). This damage can be quantified by the vertical cup-disc ratio (VCDR), comparing the vertical diameter of the cup with the vertical diameter of the total optic disc. The VCDR ranges from 0 to 1. In the clinical setting, optic nerve heads with a VCDR above 0.7 or an asymmetry between eyes above 0.2 are considered to be suspect for glaucoma. The interpretation of VCDR depends on the disc area, i.e. discs with larger area have on average a higher VCDR and the cut-off point between a normal and abnormal VCDR might be higher (4).

Elevated intraocular pressure (IOP) is a well-recognized risk factor for POAG and current therapies lower IOP by various mechanisms; eye pressure is considered to be normal between 8 and 21 mmHg, although many eyes with eye pressure in this range can exhibit glaucomatous optic nerve features. Analyses of first-degree relatives of POAG patients have shown that the sib relative risk is 10.4 ± 2.5 (5). Classic twin studies in POAG are lacking and there is no consensus on heritability for POAG, but work by Cuellar-Partida et al. estimated the heritability of POAG based on genome-wide array data at 0.42 ± 0.09 (6). Several genome-wide association studies (GWAS) have identified new POAG genes by examining POAG directly or studying endophenotypes like VCDR and IOP (7–16). Several genes associated with VCDR and IOP - CDKN2B-AS1, SIX6 (VCDR); and CAV1/CAV2, TMCO1, ABCA1 and ARHGEF12 (IOP) - are associated with POAG. Notably, no genes have been genome-wide associated with both VCDR and IOP. Charlesworth et al. previously found a genetic correlation between VCDR and IOP (RhoG = 0.45, P = 0.0012), however, genes underlying this relationship have not yet been identified (17).

The aims of this study were to (1) identify new genes associated with the POAG endophenotypes IOP, VCDR, cup area, and disc area, and ultimately POAG, using the 1000 Genomes imputations reference panel, and (2) investigate the genetic overlap between the different endophenotypes. To accomplish these aims we performed a meta-analysis of GWAS of these four traits within the International Glaucoma Genetics Consortium (IGGC).

Results

Intraocular pressure

We first conducted a meta-analysis of GWAS in the European cohorts. Top hit variants were replicated in the meta-analysis of the Asian cohorts. Next, we performed a meta-analyses of the European and Asians cohorts. After removal of single nucleotide polymorphisms (SNPs) with minor allele frequency (MAF) < 0.01 and low imputation quality, approximately 8 million SNPs were included. Whilst the meta-analysis of individuals of European descent yielded no novel associations, combined meta-analysis of individuals of European and Asian descent (n = 37,930, λ = 1.05; Supplementary Materials, Figs S1A and B, S2B) yielded nine genomic regions reaching genome-wide significance, of which eight genomic regions were already known (Supplementary Materials, Figs S1A and B, S2B, and Table S3) (9,11,13). The peak SNP in the new genomic region was rs55796939 on chromosome 11q25 near ADAMTS8 (Supplementary Materials, Figs S3 and S4).The estimated heritability (h²) of IOP using the LD score regression method (18) and the summary statistics of European-only meta-analysis was 0.13.

Vertical cup-disc ratio

In the meta-analysis of individuals of European descent (n = 23,899, λ = 1.08), 21 genomic regions were genome-wide significant (Supplementary Materials, Figs S5A, S6A and Table S4). Five genomic regions were novel (near to the genes RPE65 on chr. 1p31, F5 on chr. 1q23, PDZD2 on chr. 5p13.3, RREB1 on chr. 6p25, and DGKB on chr. 7p21.2) (Supplementary Materials, Figs S7 and S8); the other genomic regions have been previously associated with VCDR or cup area, two highly correlated traits (19–21). Of the five novel genomic regions, RREB1 (P-value = 4.13 × 10⁻³) was nominally significant in the analysis of individuals of Asian descent (n =  8,373, λ = 1.01). In the combined analysis (n = 32,272, λ = 1.06), another four novel genomic regions, near to the genes VCAN on chr. 5q14.3, PSCA on chr. 8q24.2, ENO4 on chr. 10q25.3, and RBM23 on chr. 14q11.2 (Supplementary Materials, Figs S5B and S6B), were genome-wide significant leading to a total of nine (5 + 4) novel genomic regions associated with VCDR. Of these novel genomic regions, F5 has been associated with the disc area previously (21). Disc area influences the VCDR (22), and therefore we corrected VCDR for disc area in a secondary analysis. After correction for disc area, the β (P-value) decreased from −0.007 (2.15 × 10⁻⁹) to −0.002 (5.60 × 10⁻²) in the subset with disc area available, suggesting that F5 acts primarily on disc area and secondary to VCDR through its relation to disc area. The calculated h² of VCDR using the European -only meta-analysis was 0.31.

Cup area

The meta-analysis of individuals of European descent (n = 22,489, λ = 1.06) yielded 17 genome-wide significant regions of which 14 regions were already implicated for cup area or VCDR (Supplementary Materials, Figs S9A, S10A and Table S5) (20,21). There were three novel associations on chr. 1q42.11 near CDC42BPA, chr. 8q21.11 near CRISPLD1, and on chr. 15q26.3 near FAM169B (Supplementary Materials, Figs S11 and S12). CDC42BPA has previously been associated with disc area and the fact that the association with cup area adjusting for disc area is genome-wide significant suggests an independent effect on cup area. In the combined analysis of European and Asian individuals (n = 29,828, λ = 1.06, Supplementary Materials, Figs S9B and S10B) all loci except FAM169B and CRISPLD1 remained genome-wide significant, and there was one additional genome-wide significant SNPs at chr. 6p21.2 (CDKN1A) and one highly suggestive significant SNP at chr. 9q34.2 (ABO; previously associated with IOP). For cup area the estimated h² was 0.27.

Disc area

The meta-analysis of individuals of European descent (n = 22,504, λ = 1.06) resulted in 13 genome-wide significant regions, of which two were not previously associated with disc area: UGT8 on chr. 4q26 and CTNNA3 on chr. 10q22.2 (Supplementary Materials, Figs S13A, S14A, S15, S16, and Table S6). These SNPs were not significant in the meta-analysis of individuals of Asian descent (n = 7,307, λ = 1.02). An additional four SNPs reached genome-wide significance in the combined meta-analysis (n = 29,811, λ = 1.07): PRDM16 on chr. 1p36.23-p33, GADD45A on chr. 1p31.2, VGLL4 on chr. 3p25.3, and ASB7 on chr. 15q26.3 (Supplementary Materials, Figs S13B and S14B). The estimated h² for disc area was 0.27.

Characterization of the lead association signals

In total, 82 SNPs were associated with one or more of the above endophenotypes. Functional characterization of the 82 SNPs was performed using a range of bioinformatics tools (see Methods). In total, 650 variants in linkage disequilibrium (LD) with the 82 lead SNPs (R²>0.8) were examined for functional annotation. Overall, 61% (50/82) of the associated loci are in LD with variants located in regulatory regions according to the ENCODE data (e.g. DNase I hypersensitive sites, transcription factor binding sites and motifs; see Supplementary Material, Table S7). We investigated the expression levels of the identified candidate genes using the UniGene database (23). Of all reviewed genes, CDKN1A, PAX6 and DUSP1 showed the highest number of transcripts per million in the eye (Supplementary Material, Table S8). According to the Ocular Tissue database (24), CDKN1A is highly expressed in the optic nerve head, as well as DUSP1, which also shows high expression in the trabecular meshwork. Both genes were associated with optic nerve head parameters. PAX6 is highly expressed in the ciliary body and retina; in this study we found it associated with disc area. Other highly expressed genes in the optic nerve include EFEMP1 and ABI3BP, which are associated with cup area and disc area, respectively (Supplementary Material, Table S9). To evaluate whether associated genes are highly expressed in a particular tissue, we performed tissue enrichment analyses for each trait using suggestive SNPs (P-value < 1.00 × 10⁻⁰⁵) in DEPICT (25). No FDR significant tissue enrichment was found for IOP, VCDR or disc area. However, for cup area we found an enrichment for membranes, joints, stem cells and other related connective tissues. Similar results were found when suggestive SNPs associated with VCDR, cup and disc area (the optic nerve parameters) were analyzed together (Table S10).

Table 1.

Single nucleotide polymorphisms (SNPs) that are genome-wide significantly associated with IOP and show and association with vertical cup-disc ratio.

			IOP			VCDR			Cup area			Disc area
SNP	Nearest gene	A1/A2	β	SE	P	β	SE	P	β	SE	P	Β	SE	P
rs10918274	TMCO1	t/c	0.26	0.04	5.64E-12	0.005	0.002	8.38E-03	0.010	0.003	2.47E-03	0.000	0.006	9.49E-01
rs7635832	FNDC3B	g/t	−0.22	0.03	6.61E-13	−0.001	0.001	3.35E-01	−0.004	0.003	1.27E-01	0.002	0.005	7.08E-01
rs10281637	CAV1/CAV2	c/t	0.20	0.03	3.96E-13	0.004	0.001	5.28E-03	0.006	0.003	1.23E-02	−0.002	0.005	6.01E-01
8:78380944*	PKIA	i/r	1.00	0.17	7.54E-09	0.000	0.010	9.74E-01	−0.018	0.017	3.00E-01	0.018	0.031	5.61E-01
rs7815043	PKIA	c/t	−0.10	0.03	4.41E-05	−0.001	0.001	3.13E-01	−0.001	0.002	8.32E-01	−0.002	0.004	5.66E-01
rs7944735	Many genes	c/g	0.19	0.03	6.00E-11	0.001	0.001	4.37E-01	0.006	0.003	3.33E-02	0.000	0.005	9.68E-01
11:120357425	ARHGEF12	d/r	0.18	0.03	2.02E-09	0.001	0.001	6.12E-01	0.001	0.003	6.45E-01	0.001	0.005	8.38E-01
rs12794618	ARHGEF12	c/t	0.17	0.03	7.86E-09	0.001	0.001	4.14E-01	0.002	0.003	4.84E-01	0.004	0.005	4.53E-01
rs55796939	ADAMTS8	t/c	0.36	0.06	2.31E-08	0.003	0.003	3.61E-01	0.006	0.006	3.19E-01	−0.003	0.010	7.95E-01
rs2472496	ABCA1	g/a	0.17	0.02	1.93E-13	0.004	0.001	6.83E-05	0.010	0.002	9.63E-07	0.003	0.004	4.75E-01
rs8176741	ABO	a/g	0.24	0.04	3.47E-10	0.007	0.002	4.51E-05	0.019	0.003	7.12E-08	0.004	0.006	5.42E-01
rs9913911	GAS7	g/a	−0.17	0.02	7.01E-12	−0.006	0.001	1.84E-07	−0.008	0.002	2.48E-04	−0.001	0.004	8.41E-01

For these SNPs, the associations with the other traits are also included. SNPs that are Bonferroni significantly associated with other traits are shown in bold (P-value < 5.37 × 10⁻⁴; 0.05/93). In the first rows, the SNPs genome-wide significantly associated with intraocular pressure (IOP) are shown. Next, the SNPs associated with IOP, vertical cup-disc ratio (VCDR), and cup area are shown. Nearest gene, reference NCBI build37; A1, reference allele; A2, other allele; β, effect size on the endophenotype (IOP, VCDR, cup area or disc area) based on allele A1; SE, standard error of the effect size; i, insertion; d, deletion; r, reference. *This locus was not considered as new because the signal came from only one INDEL (8:78380944).

Gene-based test

To identify new loci not found through per-SNP tests, we performed gene-based testing using VEGAS2. Reflecting the smaller number of tests, our gene-based significance threshold is P_gene-based < 0.05/24,769 = 2.02 × 10⁻⁶ (24,769 genes tested). Using the gene-based test we found several novel loci (Supplementary Material, Table S11). C9 was significantly associated with IOP (P-value 1.61 × 10⁻⁶); RARB (P-value 1.86 × 10⁻⁶) and HORMAD2-AS1 (P-value 1.04 × 10⁻⁶) were associated with VCDR. These genes were previously associated with the disc area, so the novel associations with VCDR could possibly be driven by the influence of the disc area on VCDR (21). In the cup area analysis, the genes LRP10 (P-value 1.20 × 10⁻⁶) and REM2 (P-value 1.55 × 10⁻⁶), and THSD4 (P-value 5.44 × 10 ⁸) were significantly associated. The first two genes are located near to RBM23, which was significant in the per-SNP test. THSD4 is located near to KPNB1, which was associated with VCDR in our previous meta-analysis (20). In the disc area analysis, we found two genes that were significantly associated with disc area: ANKRA2 (P-value 8.42 × 10⁻⁷) and LOC149950 (P-value 3.87 × 10⁻⁷).

Characterizing the overlap in biological pathways involved in glaucoma endophenotypes

In total, 86 SNPs were associated with one or more of the above endophenotypes. The effect estimates and P-values of these SNPs for all four endophenotypes are shown in Table 1–3. ADAMTS8 (IOP and VCDR, Tables 1 and 2B) and ABO (IOP and cup area, Table 1) were genome-wide significantly associated with two traits. Of note is that there were different variants involved in ADAMTS8: rs55796939 for IOP and rs4936099 for VCDR (r² = 0.03 between these SNPs in 1000G European samples). Figure 1 shows the overlap in associations across endophenotypes — we depict annotated genes for which at least one SNP was genome-wide significant in at least one trait — Overlap is defined as nominal significance or stronger for the second trait. The figure shows as expected a strong overlap in variants associated with disc area, cup area and VCDR. Further, overlap is noted in genes associated with IOP, cup area and VCDR. We next explored the genetic overlap between optic disc parameters and IOP using our GWAS results from the European-only meta-analyses, which comprised a larger sample size (n = 22,489–29,578). As expected genetic correlation between optic disc parameters was high; VCDR and cup area showed the highest correlation (RhoG = 0.83, P-value < 1.0 × 10 ⁻³⁰⁸), followed by VCDR and disc area (RhoG = 0.62, P-value = 7.36 × 10⁻²³). The genetic correlation between cup and disc area was 0.31 (P-value = 1.00 × 10⁻⁰⁴). No significant genetic correlation was found between IOP and optic disc parameters; the highest correlation was between IOP and VCDR (RhoG = 0.06, P-value 0.3) (Table S12).

Table 2A.

Single nucleotide polymorphisms (SNPs) that are genome-wide significantly associated with at least one of the optic disc parameters.

			IOP			VCDR			Cup area			Disc area
SNP	Nearest gene	A1/A2	β	SE	P	β	SE	P	β	SE	P	Β	SE	P
rs6804624	COL8A1	c/t	−0.01	0.03	6.54E-01	0.008	0.001	8.63E-12	0.013	0.002	1.99E-08	0.020	0.004	9.67E-07
rs7916697	ATOH7	a/g	0.01	0.03	7.43E-01	−0.018	0.001	2.46E-45	−0.017	0.002	1.32E-12	−0.094	0.004	1.34E-102
10:96008348	PLCE1	d/r	0.01	0.03	5.73E-01	0.007	0.001	4.57E-08	0.013	0.002	1.72E-08	0.015	0.004	2.22E-04
rs324780	TMTC2	g/a	0.03	0.02	2.79E-01	−0.011	0.001	7.16E-23	−0.016	0.002	1.57E-13	−0.029	0.004	8.58E-13
rs4299136	ASB7	c/g	−0.03	0.03	4.22E-01	0.010	0.002	2.68E-12	0.018	0.003	4.09E-10	0.024	0.005	4.02E-06
16:51461915	SALL1	r/i	0.02	0.03	4.34E-01	0.010	0.001	2.62E-13	0.013	0.003	6.78E-07	0.032	0.005	2.38E-12
rs4784295	SALL1	c/g	0.02	0.03	5.63E-01	0.009	0.001	3.93E-13	0.013	0.003	1.63E-07	0.031	0.005	1.12E-11
rs5752773	CHEK2	g/c	0.01	0.03	6.91E-01	−0.012	0.001	1.49E-20	−0.024	0.003	4.12E-21	−0.024	0.005	1.48E-07
rs2092172	CARD10	a/g	0.00	0.03	8.86E-01	0.009	0.001	3.08E-12	0.011	0.003	3.34E-05	0.032	0.005	1.44E-11
rs7717697	VCAN	c/t	0.01	0.02	7.21E-01	−0.007	0.001	6.66E-09	−0.009	0.002	1.19E-05	−0.018	0.004	4.84E-06
rs1681739	ENO4	t/c	0.03	0.02	2.23E-01	0.006	0.001	2.44E-08	0.011	0.002	3.70E-07	0.019	0.004	1.85E-06
rs60779155	ASB7	a/g	−0.02	0.04	6.61E-01	0.010	0.002	3.76E-10	0.019	0.003	3.75E-09	0.030	0.006	8.26E-08
rs1830890	PLCE1	g/a	0.01	0.02	8.14E-01	0.006	0.001	3.02E-08	0.012	0.002	1.06E-07	0.013	0.004	5.51E-04
rs482507	TMTC2	c/t	0.02	0.02	3.48E-01	−0.011	0.001	2.19E-19	−0.017	0.002	2.56E-14	−0.030	0.004	4.49E-13
rs4436712	SIX6	t/g	−0.04	0.02	1.47E-01	0.009	0.001	5.48E-14	0.025	0.002	1.50E-29	−0.018	0.004	6.59E-06
rs738722	CHEK2	t/c	0.02	0.03	3.57E-01	−0.012	0.001	4.94E-20	−0.024	0.003	7.81E-22	−0.021	0.005	2.63E-06
rs2684249	HSF2	c/t	0.03	0.02	2.08E-01	−0.006	0.001	1.64E-07	−0.012	0.002	3.04E-08	−0.015	0.004	1.49E-04
rs34222435	ASB7	t/c	−0.03	0.03	3.86E-01	0.010	0.002	3.07E-12	0.019	0.003	1.07E-10	0.025	0.005	2.98E-06
rs7916410	ATOH7	t/c	0.00	0.03	9.76E-01	−0.018	0.001	1.14E-45	−0.017	0.002	6.11E-12	−0.097	0.004	7.06E-109
rs442376	TMTC2	c/t	−0.03	0.03	3.09E-01	0.011	0.001	1.50E-17	0.017	0.002	3.18E-12	0.032	0.004	4.92E-14
rs1345467	SALL1	g/a	0.01	0.03	6.53E-01	0.009	0.001	4.96E-12	0.012	0.003	1.07E-06	0.032	0.005	6.41E-13
rs5762752	CHEK2	c/g	0.01	0.03	6.61E-01	−0.011	0.001	4.83E-18	−0.021	0.002	6.72E-19	−0.023	0.004	2.26E-08
rs11129176	RARB	a/g	0.02	0.03	4.17E-01	0.005	0.001	3.17E-05	0.010	0.002	1.01E-05	0.023	0.004	3.40E-08
rs1997404	COL8A1	g/t	−0.03	0.03	3.24E-01	0.008	0.001	2.39E-11	0.013	0.002	7.71E-08	0.024	0.004	1.90E-08
rs34935520	SIX6	g/a	−0.04	0.02	1.13E-01	0.009	0.001	7.95E-14	0.025	0.002	6.96E-29	−0.023	0.004	7.61E-08

For these SNPs, the associations with the other traits are also included. Here the SNPs genome-wide significantly associated with at least one of the optic disc parameters and that are Bonferroni significantly associated with the other disc parameters are shown in bold (P-value < 5.37 × 10⁻⁴; 0.05/93).Nearest gene, reference NCBI build37; A1, reference allele; A2, other allele; β, effect size on the effect size on the endophenotype (IOP, VCDR, cup area or disc area) based on allele A1; SE, standard error of the effect size; i, insertion; d, deletion; r, reference.

Table 2B.

Single nucleotide polymorphisms (SNPs) that are genome-wide significantly associated with vertical cup-disc ratio or with cup area

			IOP			VCDR			Cup area			Disc area
SNP	Nearest gene	A1/A2	β	SE	P	β	SE	P	β	SE	P	Β	SE	P
rs1925953	RPE65*	t/a	−0.02	0.02	3.26E-01	0.006	0.001	1.55E-07	0.010	0.002	1.50E-05	0.006	0.004	1.08E-01
rs72759609	PDZD2	c/t	−0.04	0.05	3.50E-01	−0.012	0.002	7.10E-09	−0.020	0.004	1.98E-06	−0.021	0.008	5.62E-03
rs114503346	DUSP1	t/c	−0.12	0.08	1.27E-01	−0.021	0.004	1.31E-08	−0.035	0.007	2.90E-07	−0.035	0.013	5.83E-03
rs4960295	RREB1	a/g	0.02	0.02	4.75E-01	0.007	0.001	2.49E-10	0.009	0.002	3.73E-05	0.012	0.004	3.29E-03
rs10274998	DGKB	t/c	0.02	0.03	4.38E-01	0.008	0.001	4.68E-08	0.012	0.003	8.08E-06	0.011	0.005	2.65E-02
rs2157719	CDKN2B-AS1	c/t	−0.04	0.02	9.81E-02	−0.013	0.001	3.75E-35	−0.024	0.002	3.31E-28	−0.008	0.004	3.03E-02
rs3891783	PLCE1	g/c	0.04	0.02	1.01E-01	0.007	0.001	1.06E-10	0.011	0.002	3.28E-07	0.012	0.004	1.52E-03
rs1346	SSSCA1	t/a	−0.05	0.03	1.20E-01	−0.013	0.002	7.51E-18	−0.019	0.003	9.31E-11	−0.016	0.005	2.10E-03
rs4936099	ADAMTS8	c/a	−0.03	0.03	2.38E-01	−0.007	0.001	6.70E-09	−0.013	0.002	4.96E-08	−0.006	0.004	1.72E-01
13:36629905	DCLK1	d/r	−0.02	0.03	5.70E-01	0.007	0.001	2.98E-08	0.018	0.002	2.20E-14	−0.005	0.004	2.36E-01
rs7323428	DCLK1	t/g	−0.02	0.03	4.13E-01	0.007	0.001	1.86E-08	0.019	0.002	1.67E-15	−0.005	0.004	2.23E-01
rs8015152	SIX6	t/c	−0.06	0.02	2.27E-02	0.010	0.001	2.86E-18	0.024	0.002	8.15E-26	−0.011	0.004	6.18E-03
rs6107845	BMP2	a/g	0.03	0.02	2.80E-01	−0.009	0.001	3.44E-17	−0.017	0.002	2.90E-15	−0.004	0.004	3.27E-01
rs6764184	FLNB	t/g	0.05	0.03	5.03E-02	0.007	0.001	1.89E-08	0.015	0.002	1.30E-10	0.010	0.004	1.92E-02
rs7311936	FAM101A	c/g	−0.03	0.02	1.69E-01	−0.006	0.001	2.48E-09	−0.013	0.002	4.52E-09	0.003	0.004	5.14E-01
14:23388793	RBM23	r/d	0.02	0.03	3.99E-01	0.007	0.001	2.56E-08	0.013	0.003	2.01E-07	0.009	0.005	4.29E-02
rs3794453	RBM23	a/t	0.01	0.02	7.22E-01	0.007	0.001	7.25E-08	0.011	0.002	2.88E-07	0.009	0.004	3.11E-02
rs2252865	RERE	t/c	0.05	0.03	4.11E-02	0.005	0.001	2.66E-05	0.014	0.002	1.33E-09	0.003	0.004	5.08E-01
			IOP			VCDR			Cup area			Disc area
SNP	Nearest gene	A1/A2	β	SE	P	β	SE	P	β	SE	P	Β	SE	P
rs13016883	TRIB2	c/g	0.01	0.03	5.64E-01	0.006	0.001	3.44E-06	0.016	0.002	1.83E-11	0.001	0.004	8.30E-01
rs35084382	DUSP1	c/t	−0.10	0.07	1.32E-01	−0.018	0.003	2.05E-08	−0.033	0.006	2.17E-08	−0.031	0.011	5.51E-03
rs117598310	CRISPLD1*	t/g	−0.05	0.05	3.10E-01	0.009	0.002	1.07E-04	0.021	0.004	1.66E-06	0.022	0.008	5.47E-03
rs1360589	CDKN2B-AS1	c/t	−0.04	0.02	8.42E-02	−0.013	0.001	1.43E-34	−0.024	0.002	2.90E-28	−0.008	0.004	4.45E-02
rs11613189	FAM101A	t/c	−0.03	0.03	2.27E-01	−0.005	0.001	6.04E-06	−0.016	0.002	2.01E-12	0.002	0.004	6.42E-01
rs2251069	DDHD1	c/t	0.01	0.02	7.29E-01	−0.006	0.001	7.41E-08	−0.013	0.002	1.20E-09	0.001	0.004	7.11E-01
rs6598351	FAM169B*	t/c	−0.02	0.03	5.26E-01	0.006	0.001	2.80E-05	0.012	0.003	1.77E-05	−0.004	0.005	3.90E-01
rs11646917	SALL1	t/g	−0.01	0.03	6.65E-01	−0.009	0.001	4.83E-10	−0.015	0.003	4.76E-09	−0.015	0.005	1.30E-03
rs11867840	BCAS3	g/a	0.04	0.03	1.04E-01	−0.006	0.001	4.86E-06	−0.018	0.002	2.35E-13	0.011	0.004	1.00E-02
rs6054375	BMP2	t/g	0.03	0.03	2.45E-01	−0.010	0.001	6.92E-15	−0.018	0.002	1.83E-15	−0.003	0.004	4.74E-01
rs3791679	EFEMP1/PNPT1	g/a	0.04	0.03	1.72E-01	−0.005	0.001	1.17E-04	−0.013	0.002	4.92E-08	0.003	0.004	5.14E-01
rs12494328	FLNB	a/g	0.04	0.03	1.52E-01	0.006	0.001	1.56E-06	0.016	0.002	6.03E-11	0.009	0.004	4.50E-02
6:36592986	CDKN1A	d/r	−0.02	0.03	5.32E-01	0.006	0.001	1.92E-05	0.015	0.003	1.12E-08	−0.006	0.005	2.09E-01
rs72852338	CDKN1A	c/a	−0.02	0.03	5.46E-01	0.006	0.001	3.29E-05	0.014	0.003	3.17E-08	−0.005	0.005	2.97E-01
rs1074407	TRIOBP	t/a	0.11	0.02	4.00E-06	0.006	0.001	3.32E-07	0.012	0.002	1.90E-08	0.008	0.004	3.92E-02

For these SNPs, the associations with the other traits are also included. Here the SNPs genome-wide significantly associated with vertical cup-disc ratio or cup area and that are Bonferroni significantly associated with VCDR or cup area are shown in bold (P-value < 5.37 × 10⁻⁴; 0.05/93).Nearest gene, reference NCBI build37; A1, reference allele; A2, other allele; β, effect size on effect size on the endophenotype (IOP, VCDR, cup area or disc area) based on allele A1; SE, standard error of the effect size; i, insertion; d, deletion; r, reference.

^*Genome-wide associated in the European-only meta-analysis.

Table 2C.

Single nucleotide polymorphisms (SNPs) that are genome-wide significantly associated with vertical cup-disc ratio or with disc area

			IOP			VCDR			Cup area			Disc area
SNP	Nearest gene	A1/A2	β	SE	P	β	SE	P	β	SE	P	Β	SE	P
rs1192414	CDC7/TGFBR3	a/g	0.06	0.03	5.66E-02	0.014	0.001	1.78E-23	0.007	0.003	1.12E-02	0.087	0.005	7.44E-71
rs10753787	F5	t/c	−0.03	0.02	1.69E-01	−0.007	0.001	2.48E-09	−0.005	0.002	2.14E-02	−0.019	0.004	1.60E-06
rs2920293	PSCA	g/c	0.00	0.02	8.57E-01	−0.006	0.001	5.04E-09	−0.007	0.002	9.17E-04	−0.015	0.004	9.94E-05
rs4658101	CDC7/TGFBR3	a/g	0.06	0.03	4.46E-02	0.013	0.001	5.19E-23	0.007	0.003	1.13E-02	0.089	0.005	8.01E-77
1:169530520	F5/SELP	i/r	0.02	0.03	4.22E-01	0.007	0.001	7.20E-07	0.005	0.003	5.44E-02	0.033	0.005	1.49E-12
rs2239854	F5/SELP	a/g	0.03	0.03	2.64E-01	0.006	0.001	8.37E-07	0.005	0.002	5.04E-02	0.030	0.004	7.60E-13
rs9843102	ABI3BP	a/g	0.00	0.03	9.84E-01	−0.006	0.002	2.18E-04	−0.002	0.003	5.88E-01	−0.036	0.005	1.35E-11
8:88744441	DCAF4L2	d/r	−0.01	0.02	6.98E-01	0.006	0.001	6.66E-07	0.006	0.002	4.53E-03	0.026	0.004	2.04E-11
rs6468996	DCAF4L2	t/c	0.00	0.02	9.12E-01	0.005	0.001	2.52E-07	0.006	0.002	2.14E-03	0.025	0.004	5.16E-11
rs61101201	ELP4/PAX6	g/t	0.02	0.03	5.51E-01	0.006	0.001	2.27E-06	0.005	0.002	4.51E-02	0.028	0.004	1.53E-10
rs56385951	CARD10	a/g	−0.06	0.04	9.08E-02	0.011	0.002	1.87E-11	0.008	0.003	8.83E-03	0.047	0.006	1.49E-16
1:3046430	PRDM16	i/r	−0.04	0.04	4.14E-01	0.007	0.002	5.35E-04	−0.002	0.004	7.15E-01	0.044	0.007	1.79E-09
rs12028027	PRDM16	c/t	−0.03	0.04	4.97E-01	0.007	0.002	2.15E-04	−0.001	0.004	8.58E-01	0.043	0.007	1.46E-09

For these SNPs, the associations with the other traits are also included. Here the SNPs genome-wide significantly associated with vertical cup-disc ratio or with disc area that are Bonferroni significantly associated with the VCDR or disc area are shown in bold (P-value < 5.37 × 10⁻⁴; 0.05/93).Nearest gene, reference NCBI build37; β, effect size on effect size on the endophenotype (IOP, VCDR, cup area or disc area) based on allele A1; SE, standard error of the effect size; i, insertion; d, deletion; r, reference.

Table 3.

Single nucleotide polymorphisms (SNPs) that are genome-wide significantly associated with optic nerve head parameters (cup area and disc area)

			IOP			VCDR			Cup area			Disc area
SNP	Nearest gene	A1/A2	β	SE	P	β	SE	P	β	SE	P	Β	SE	P
1:227562773	CDC42BPA	d/r	−0.10	0.05	3.01E-02	0.003	0.002	2.37E-01	0.024	0.004	8.05E-09	−0.055	0.008	3.65E-13
rs73102394	CDC42BPA	t/c	−0.09	0.05	4.34E-02	0.003	0.002	1.62E-01	0.022	0.004	4.16E-08	−0.053	0.007	5.01E-13
rs11811982	CDC42BPA	a/c	−0.12	0.05	1.35E-02	0.004	0.002	5.54E-02	0.027	0.004	2.31E-10	−0.062	0.008	2.02E-15
rs10021731	UGT8	c/t	0.01	0.02	8.23E-01	−0.002	0.001	5.56E-02	−0.002	0.002	2.68E-01	−0.020	0.004	7.48E-07
rs12220165	CTNNA3*	g/c	0.02	0.03	5.88E-01	−0.004	0.002	1.47E-02	−0.004	0.003	1.92E-01	−0.023	0.005	2.51E-05
rs787541	U6, GADD45A	c/g	0.07	0.03	7.08E-03	0.002	0.001	7.47E-02	0.002	0.002	4.82E-01	0.023	0.004	6.66E-08
rs1367187	DIRC3	c/t	−0.07	0.03	9.74E-03	0.002	0.001	2.46E-01	−0.002	0.003	4.87E-01	0.026	0.005	1.03E-08
rs2443724	VGLL4	c/g	0.00	0.02	8.62E-01	−0.003	0.001	1.53E-02	0.000	0.002	9.15E-01	−0.022	0.004	4.72E-08
rs1013830	CTNNA3	t/c	0.00	0.05	9.49E-01	−0.007	0.002	4.80E-03	−0.004	0.005	4.10E-01	−0.046	0.008	5.45E-08

For these SNPs, the associations with the other traits are also included. SNPs that are Bonferroni significantly associated with other traits are shown in bold (P-value < 5.37 × 10⁻⁴; 0.05/93). In the first rows, the SNPs genome-wide significantly associated with cup area are shown. Next, SNPs associated with only disc area, are shown. Nearest gene, reference NCBI build37; A1, reference allele; A2, other allele; β, effect size on the endophenotype (IOP, VCDR, cup area or disc area) based on allele A1; SE, standard error of the effect size; i, insertion; d, deletion; r, reference.

^*Genome-wide associated with the European-only meta-analysis.

Figure 1.

Overlap between the genes associated with one or more endophenotypes. Genes with a genome-wide significant association for at least one trait are shown. These genes are counted as overlapping genes if they are Bonferroni significantly associated with the other trait(s). Chr 11p11.2 (see intraocular pressure circle) means a region on chromosome 11p11.2 that is associated with IOP and has many genes in it; the likely causative gene in this region is not identified yet. Genes in bold are genes associated with primary open-angle glaucoma (POAG) in our meta-analysis of four case-control studies.*Genes associated with familial forms of POAG (e.g. MYOC and OPTN) or found in case-control association studies which did not show an association with the endophenotypes explored in this study.

Table 4.

Association with primary open-angle glaucoma in a meta-analysis of four independent glaucoma case-control studies (ANZRAG, NEIGHBORHOOD, Singapore, and Southampton).

	Nearest gene	A1/A2	OR (95% CI)	OR (R)	P-value	P-value (R)	Direction	I2	P-value of heterogeneity
IOP SNPs
rs10918274	TMCO1	t/c	1.39 (1.30-1.50)	1.39	2.75E-19	1.37E-09	++ ++	38.4	1.82E-01
rs7635832	FNDC3B	g/t	0.89 (0.83-0.95)	0.91	1.41E-03	3.65E-02	—?	33.9	2.20E-01
rs10281637	CAV1/CAV2	c/t	1.13 (1.07-1.20)	1.13	2.32E-05	2.32E-05	++ ++	0	4.89E-01
rs2487048	ABCA1	a/g	1.26 (1.19-1.33)	1.26	2.65E-15	3.82E-03	++ ++	82.9	5.53E-04
rs8176741	ABO	a/g	1.07 (0.99-1.17)	1.04	7.36E-02	5.25E-01	−+-+	58.5	6.51E-02
rs7944735	Many genes (NUP160, PTPRJ)	c/g	1.06 (1.01-1.13)	1.07	2.99E-02	2.99E-02	++ ++	0	8.99E-01
11:120357425	ARHGEF12	d/r	1.16 (1.09-1.23)	1.19	1.52E-06	3.02E-02	++ ++	83.2	4.65E-04
rs55796939	ADAMTS8	t/c	1.07 (0.94-1.24)	1.17	2.72E-01	4.46E-01	+?–	78.6	9.35E-03
rs9913911	GAS7	g/a	0.80 (0.76-0.84)	0.80	1.08E-17	1.08E-17	—-	0	7.50E-01
VCDR SNPs
rs1925953	RPE65	t/a	1.07 (1.02-1.13)	1.10	4.21E-03	2.01E-02	++ ++	46.7	1.31E-01
rs1192414	CDC7/TGFBR3	a/g	1.08 (1.02-1.16)	1.08	9.26E-03	9.26E-03	++ ++	0	7.27E-01
rs10753787	F5	t/c	0.97 (0.93-1.03)	0.97	3.67E-01	3.67E-01	—-	0	9.92E-01
rs6804624	COL8A1	c/t	0.99 (0.94-1.05)	0.99	8.14E-01	8.14E-01	—+	0	8.42E-01
rs72759609	PDZD2	c/t	0.90 (0.83-0.99)	0.91	3.20E-02	3.20E-02	—-	0	9.53E-01
rs114503346	DUSP1	t/c	1.00 (0.80-1.25)	1.00	9.99E-01	8.80E-01	+?-+	42	1.78E-01
rs4960295	RREB1	a/g	0.99 (0.95-1.05)	1.00	9.50E-01	9.09E-01	−+-+	4.6	3.70E-01
rs10274998	DGKB	t/c	1.03 (0.98-1.10)	1.04	2.16E-01	2.16E-01	++-+	0	5.38E-01
rs2157719	CDKN2B-AS1	c/t	0.69 (0.66-0.74)	0.69	1.29E-40	1.29E-40	—-	0	5.67E-01
rs1900005	ATOH7	a/c	1.01 (0.96-1.07)	1.01	6.98E-01	6.77E-01	+- ++	5.1	3.67E-01
10:96008348	PLCE1	d/r	1.02 (0.97-1.09)	1.04	3.38E-01	3.15E-01	+-+?	35.3	2.13E-01
rs1346	SSSCA1	t/a	0.90 (0.85-0.97)	0.91	2.41E-03	2.41E-03	—-	0	9.04E-01
rs4936099	ADAMTS8	c/a	0.94 (0.9-1.00)	0.94	5.75E-02	5.75E-02	—-	0	9.63E-01
rs324780	TMTC2	g/a	0.93 (0.89-0.99)	0.93	1.35E-02	1.35E-02	—-	0	7.69E-01
13:36629905	DCLK1	d/r	0.99 (0.94-1.05)	0.99	7.53E-01	8.00E-01	–+-	6.2	3.62E-01
rs8015152	SIX6	t/c	1.21 (1.16-1.28)	1.19	3.90E-15	7.08E-05	++ ++	62.4	4.62E-02
rs4299136	ASB7	c/g	1.03 (0.97-1.10)	1.03	3.55E-01	3.55E-01	++-+	0	8.29E-01
16:51461915	SALL1	i/r	0.94 (0.89-1.00)	0.94	3.85E-02	3.85E-02	—-	0	7.82E-01
rs6107845	BMP2	a/g	0.89 (0.85-0.94)	0.91	1.02E-05	6.94E-03	—-	43.1	1.53E-01
rs5752773	CHEK2	g/c	0.92 (0.88-0.98)	0.92	4.63E-03	4.63E-03	—-	0	9.12E-01
rs2092172	CARD10	a/g	0.97 (0.92-1.04)	0.98	4.35E-01	4.35E-01	–+-	0	7.76E-01
rs6764184	FLNB	t/g	1.07 (1.02-1.13)	1.02	5.73E-03	7.66E-01	++-+	86.1	8.14E-05
rs7717697	VCAN	c/t	0.98 (0.93-1.04)	0.98	5.26E-01	5.26E-01	—?	0	7.30E-01
rs2920293	PSCA	g/c	1.03 (0.98-1.09)	1.03	2.25E-01	2.25E-01	++-?	0	3.79E-01
rs1681739	ENO4	t/c	1.02 (0.97-1.08)	1.03	3.92E-01	3.99E-01	+–+	49.2	1.16E-01
rs7311936	FAM101A	c/g	0.99 (0.95-1.04)	1.00	8.12E-01	8.59E-01	+—	11	3.38E-01
14:23388793	RBM23	r/d	1.03 (0.98-1.10)	1.03	1.83E-01	1.83E-01	+++?	0	4.61E-01
Cup area SNPs
rs2252865	RERE	t/c	1.11 (1.06-1.18)	1.11	5.76E-05	2.87E-02	++-+	59.3	6.10E-02
rs4846112	DHRS3	a/g	0.95 (0.91-1.01)	0.96	1.18E-01	1.18E-01	—-	0	5.53E-01
1:227562773	CDC42BPA	d/r	0.87 (0.79-0.97)	0.90	1.14E-02	2.11E-01	–+?	48.6	1.43E-01
rs13016883	TRIB2	c/g	1.08 (1.03-1.14)	1.08	4.25E-03	4.25E-03	+++?	0	8.63E-01
rs35084382	DUSP1	c/t	1.04 (0.85-1.29)	1.05	6.72E-01	6.72E-01	+?-+	0	3.91E-01
rs117598310	CRISPLD1	t/g	1.08 (1.00-1.19)	1.09	5.39E-02	5.39E-02	+++	0	8.01E-01
rs1360589	CDKN2B-AS1	c/t	0.69 (0.66-0.73)	0.69	1.90E-42	1.90E-42	—-	0	6.47E-01
rs10998036	ATOH7	c/g	1.01 (0.96-1.08)	1.02	5.42E-01	5.72E-01	+—	26	2.55E-01
10:96008348	PLCE1	d/r	1.02 (0.97-1.09)	1.04	3.38E-01	3.15E-01	+-+?	35.3	2.13E-01
rs1346	SSSCA1	t/a	0.90 (0.85-0.97)	0.91	2.41E-03	2.41E-03	—-	0	9.04E-01
rs482507	TMTC2	c/t	0.94 (0.89-0.99)	0.94	2.03E-02	2.03E-02	—-	0	7.46E-01
rs11613189	FAM101A	t/c	0.99 (0.95-1.05)	0.99	8.25E-01	7.77E-01	++–	18.5	2.98E-01
rs7323428	DCLK1	t/g	0.99 (0.94-1.05)	1.00	7.83E-01	8.87E-01	+-+-	13.6	3.25E-01
rs2251069	DDHD1	c/t	0.95 (0.91-1.00)	0.96	7.62E-02	7.62E-02	–+-	0	4.08E-01
rs4436712	SIX6	t/g	1.24 (1.19-1.31)	1.23	5.77E-18	1.52E-07	++ ++	48.8	1.19E-01
rs6598351	FAM169B	t/c	0.99 (0.93-1.06)	0.99	8.06E-01	8.06E-01	−+–	0	7.11E-01
rs11646917	SALL1	t/3g	0.98 (0.93-1.04)	0.98	5.49E-01	5.49E-01	– ++	0	5.97E-01
rs11867840	BCAS3	g/a	1.06 (1.01-1.13)	1.06	1.83E-02	2.12E-02	++ ++	8.3	3.51E-01
rs6054375	BMP2	t/g	0.89 (0.85-0.94)	0.91	8.52E-06	9.93E-03	—-	47.1	1.29E-01
rs738722	CHEK2	t/c	0.93 (0.89-0.99)	0.93	1.26E-02	1.26E-02	—-	0	9.05E-01
rs3791679	EFEMP1/PNPT1	a/g	0.96 (0.92-1.02)	0.96	2.23E-01	2.23E-01	—-	0	5.51E-01
rs12494328	FLNB	a/g	1.13 (1.07-1.20)	1.13	1.28E-05	5.89E-04	++-+	26.9	2.50E-01
rs6804624	COL8A1	c/t	0.99 (0.94-1.05)	0.99	8.14E-01	8.14E-01	—+	0	8.42E-01
6:36592986	CDKN1A	d/r	1.14 (1.09-1.21)	1.15	7.74E-07	1.04E-04	++ ++	36.6	1.93E-01
rs2684249	HSF2	c/t	0.92 (0.88-0.97)	0.94	1.08E-03	1.66E-01	—+	63.3	4.25E-02
rs8176672	ABO	t/c	1.00 (0.91-1.11)	1.00	9.49E-01	9.49E-01	−+-?	0	3.69E-01
rs4936099	ADAMTS8	c/a	0.94 (0.90-1.00)	0.94	5.75E-02	5.75E-02	—-	0	9.63E-01
rs34222435	ASB7	t/c	1.03 (0.97-1.10)	1.03	3.66E-01	3.66E-01	++-+	0	8.74E-01
rs1074407	TRIOBP	t/a	1.04 (1.00-1.10)	1.04	4.92E-02	8.66E-02	++ ++	32.9	2.15E-01
Disc area SNPs
rs4658101	CDC7/TGFBR3	a/g	1.08 (1.02-1.16)	1.08	7.81E-03	7.81E-03	++ ++	0	7.22E-01
1:169530520	F5/SELP	i/r	1.01 (0.96-1.08)	1.02	5.40E-01	5.40E-01	++-?	0	7.14E-01
rs11811982	CDC42BPA	a/c	0.87 (0.80-0.97)	0.90	1.19E-02	8.28E-02	– ++	20.5	2.87E-01
rs9843102	ABI3BP	a/g	0.92 (0.86-0.98)	0.92	1.37E-02	1.37E-02	—-	0	6.24E-01
rs10021731	UGT8	c/t	1.01 (0.96-1.06)	1.01	6.82E-01	6.82E-01	– ++	0	6.50E-01
8:88744441	DCAF4L2	d/r	1.03 (0.99-1.09)	1.04	0.1225	1.39E-01	++-+	4.9	3.68E-01
rs12220165	CTNNA3	g/c	1.08 (1.01-1.16)	1.09	1.14E-02	1.14E-02	++ ++	0	9.04E-01
rs7916410	ATOH7	t/c	1.00 (0.96-1.06)	1.00	7.63E-01	7.45E-01	+- ++	3.9	3.73E-01
rs61101201	ELP4/PAX6	g/t	1.00 (0.94-1.06)	1.00	9.77E-01	9.77E-01	−+-?	0	9.63E-01
rs442376	TMTC2	c/t	1.04 (0.99-1.10)	1.05	7.94E-02	7.94E-02	− +++	0	6.82E-01
rs1345467	SALL1	g/a	1.07 (1.01-1.14)	1.07	1.86E-02	1.86E-02	++ ++	0	8.73E-01
rs5762752	CHEK2	c/g	0.92 (0.88-0.98)	0.92	4.90E-03	4.90E-03	—-	0	8.29E-01
rs56385951	CARD10	a/g	0.99 (0.92-1.07)	1.00	9.15E-01	9.15E-01	+-+-	0	9.88E-01
1:3046430	PRDM16	i/r	0.97 (0.87-1.10)	0.98	7.13E-01	8.72E-01	+–?	63.9	6.28E-02
rs787541	U6, GADD45A	c/g	0.98 (0.94-1.04)	0.98	6.10E-01	9.06E-01	– ++	50.7	1.08E-01
rs1367187	DIRC3	c/t	0.95 (0.90-1.01)	0.96	1.11E-01	4.12E-01	+-+-	46.1	1.35E-01
rs2443724	VGLL4	c/g	0.91 (0.87-0.97)	0.91	1.04E-03	2.61E-02	–+-	38	1.84E-01
rs11129176	RARB	a/g	0.99 (0.94-1.05)	1.00	8.85E-01	9.93E-01	+—	40.4	1.69E-01
rs1997404	COL8A1	g/t	1.00 (0.95-1.06)	1.00	9.60E-01	9.60E-01	− +++	0	6.18E-01
rs34935520	SIX6	g/a	1.26 (1.20-1.33)	1.26	2.82E-20	6.73E-14	++ ++	21.5	2.81E-01
rs60779155	ASB7	a/g	1.02 (0.96-1.10)	1.03	4.52E-01	4.52E-01	+–+	0	5.02E-01

Results are shown for the most significantly associated single nucleotide polymorphisms from the endophenotype analyses. Nearest gene, reference NCBI build37; A1, reference allele; A2, other allele; OR, estimated odds ratio for allele A1; 95% CI, confidence interval; OR (R), estimated odds ratio for allele A1 in random effect meta-analysis; P-value (R), P-value in random effect meta-analysis; I² statistic measuring heterogeneity on a scale of 0–100%; i, insertion; d, deletion; r, reference.

To further characterize the overlap in biological functions, gene set enrichment of loci associated with IOP and optic disc parameters was performed using DEPICT (25). We first investigated enriched pathways or gene sets using only genome-wide associated SNPs. No significant pathways were found after FDR correction. However, the pathways involved in metabolic processes such as “increased circulating leptin level”, “abnormal fat cell morphology” and “increased insulin sensitivity” were suggestive when we analyzed the list of SNPs associated with VCDR, cup area and disc area (FDR < 0.2, see Supplementary Material, Table S13). We next searched for enriched pathways using suggestive SNPs (P-value <1.0 × 10⁻⁵). We further investigated the potential overlap in pathways across the endophenotypes, and found 57 significant pathways when using VCDR, cup area and IOP variants, and 100 pathways when analysing suggestive VCDR, cup area and disc area variants. Note that in the first analysis we investigated pathways enriched when IOP genes are taken into account, while in the second one we analysed genes influencing the optic nerve head characteristics. Due to a high degree of redundancy between pathways, we clustered the significant pathways into meta-pathways, resulting in 11 meta-pathways for VCDR, cup area and IOP (Fig. 2A, Supplementary Material, Table S14); and 17 for VCDR, cup area, and disc area (Fig. 2B, Supplementary Material, Table S15). Most of the gene sets found in both analyses highlighted pathways involved in cell differentiation, notch signalling, regulatory DNA binding and embryonic development, which reflects the pathways found when VCDR and CA variants are analyzed (Supplementary Material, Fig. S17). Furthermore, we found “abnormal fat cell morphology” and “abnormal liver morphology” significantly enriched; a key gene in these pathways is ABCA1. When IOP genes are included the elongation factor, RNA Polymerase II (ELL2) protein complex” shows an enrichment. When disc area genes are included, pathways such as “blood vessel development”, “protein import into nucleus”, “Thrombospondin 1 (THBS1) and SMAD3 protein complex”, and “abnormal eye morphology” were significant. Key genes in the latter include: CDKN2B, FAT4, LRIG3, SIX6, COL8A1, SOX11, RND3, BOC, WNT2B and CYP26A1.

Figure 2.

Pathways significantly enriched for: (A) Loci associated with the vertical cup-disc ratio, cup area and intraocular pressure (P-value <7.0 × 10-6 in the GWAS). In total 11 meta-pathways were identified after clustering the 57 pathways identified by DEPICT. B) Loci associated with vertical cup-disc ratio, cup area and disc area (P-value <1.0 × 10-5). In total 17 meta-pathways were identified after clustering the 100 pathways identified by DEPICT. In both figures, meta-pathways are represented by nodes coloured according to statistical significance, and edges are scaled according to the correlation between meta-pathways. *The pathway “Abnormal eye morphology” clustered with the meta-pathway “Chordate embryonic development”. ELL2 = Elongation Factor, RNA Polymerase II, DVL3= Dishevelled Segment Polarity Protein 3, THBS1 = Thrombospondin 1, RFX2= Regulatory Factor X, 2. MDFI = MyoD Family Inhibitor.

From endophenotypes to primary open-angle glaucoma

To evaluate the implications of our findings in the context of glaucoma, we examined the association between the genome-wide significant SNPs found in this study and 4 independent POAG studies (n =  6,429 cases and 41,404 controls). In total, 75 independent (i.e. R²<0.8) SNPs associated with one or more of the endophenotypes were assessed in the case/control studies. Of these, 32 were nominal significantly associated with POAG (P-value <0.05; the chance that 32 SNPs of 75 SNPs have a P-value <0.05 is < 2.2 × 10⁻¹⁶), and 11 independent SNPs were Bonferroni significantly associated with POAG (P-value 0.05/75 = 6.67 × 10⁻⁴) (Table 4). Two out of the 11 Bonferroni significant SNPs, the rs2487048 in the ABCA1 gene and the 11:120357425 in the ARHGEF12 showed high heterogeneity (I²). To estimate the common effect size we performed a random effect meta-analysis. The odds ratio (OR) remained almost the same for both variants, although P-values were not significant after adjusting for multiple testing, which is in line with the heterogeneity observed. All other nine SNPs surpassed the Bonferroni threshold for significance in both fixed and random-effect models. The association between CDKN1A and POAG is novel (OR = 1.14, P-value = 7.4 × 10⁻⁷). In our previous paper, the SNP rs6054374 near to BMP2 was already associated with POAG (OR = 0.92, P-value 3.74 × 10⁻³), but the most significantly associated SNP in the current meta-analysis rs6107845 near to BMP2 shows a slightly larger effect on POAG (OR = 0.89, P-value = 8.52 × 10⁻⁶). CDKN1A, the novel associated POAG candidate gene belongs to the cyclin dependent kinase inhibitor (CDKN) gene family as well as CDKN2A, CDKN2B/CDKN2B-AS1, which all lie in a well-known glaucoma associated locus on chr.9p21.

Expression of cdkn1a after knockdown of six6b in zebrafish

Previous studies have shown that the transcription factor SIX6 alters the expression of CDKN2A and CDKN2B (26,27). All three genes (SIX6, CDKN2A and CDKN2B) are in loci associated with POAG, suggesting a possible functional link between these POAG-loci. Given that CDKN1A is part of the same gene family as CDKN2A and CDKN2B we tested whether SIX6 also regulates CDKN1A. To assess the potential for functional regulation of CDKN1A through SIX6, we first performed in silico analyses and observed that SIX6 binds to CDKN1A (core score = 0.812). Then we used a previously studied zebrafish model (27), in which we tested whether knockdown of six6b alters the expression of cdkn1a in vivo. Knockdown of six6b was achieved using morpholino technology (27). 85% of the knockdown embryos showed a small eye phenotype, reduced optic nerve thickness and an up-regulation of the expression levels of cdkn2a/cdkn2b, as observed in previous studies (n = 220) (27,28). In zebrafish, there is only one gene which is analogous to the human CDKN2A and CDKN2B and it is referred to in this paper as cdkn2a/cdkn2b. We evaluated the expression levels of cdkn1a in eyes of six6b deficient embryos by RT-qPCR. A 41-fold overexpression of cdkn1a in the eye of six6b knockdown embryos was found (P-value = 0.001) (Fig. 3), showing that in vivo downregulation of six6b affects the expression levels not only of cdkn2a/cdkn2b but also of cdkn1a, likely by binding to their sequence, repressing their expression.

Figure 3.

cdkn1a mRNA expression change Overexpression of cdkn1a and cdkn2a/cdkn2b in response to six6b depletion is shown. All samples expression were normalized to the control gene sdha. Relative expression was calculated by setting the wild-type expression level at 1. Values represent mean ± standard error of the mean. *P < 0.05; **P < 0.005.

Discussion

This meta-analysis within the IGGC identified a novel genomic region associated with IOP, nine genomic regions associated with VCDR, five with cup area, and six with disc area. Eleven genomic regions were associated with POAG. Of these regions, the association between CDKN1A and POAG is novel.

We identified some specific loci that underlie the genetic correlation between IOP and VCDR described earlier (17). ADAMTS8 and ABO were genome-wide significant for both IOP and VCDR or cup area. Variants found close to ABO (rs8176672 for cup area and rs8176741 for IOP) are in LD (r² >0.85) with rs12216891, which lies in an enhancer and promoter histone mark, suggesting a potential regulatory mechanism in that region. Furthermore, TRIOBP is genome-wide significant for cup area, and reached a P-value of 3.42 × 10⁻⁶ for IOP. Interestingly, TRIOBP is approximately 180 kb away from CARD10 which is associated with disc area. There is a large overlap between VCDR/cup area and disc area. Since VCDR is related to disc area, it might be that the effect found for VCDR is due to the effect of disc area. Most of these overlapping genes are still Bonferroni significant in the cup area analysis in which we corrected for disc area. Only CDC7/TGFBR3 and F5 are genome-wide significant for VCDR as well as for disc area, but the effect is negligible after correction for disc area, suggesting that these two genes play primarily a role in disc area.

When suggestive SNPs (P-value <1.0 × 10⁻⁵) for VCDR and cup area were analyzed together using DEPICT, we found an enrichment of pathways involved in cell differentiation, development, regulatory DNA binding and Notch signalling. Including disc area SNPs to the VCDR and cup area analysis reveals additional joint pathways: 1) eye and blood vessel development, 2) cancer, 3) protein import into nucleus, and 4) thrombospondin 1 and SMAD3 complexes, related to the extracellular matrix. The extracellular matrix pathway has been previously implicated in optic nerve degeneration (20). ADAMTS8, COL8A1 and the novel identified gene VCAN (versican) are involved in the metabolism and composition of the extracellular matrix. Interestingly, mutations in VCAN have been implicated in several ophthalmologic disorders (29). Of interest, known POAG genes also fit in the pathways identified in this study, for example GAS7 and SIX6 play a role during development (27,30), TGFBR3 has been implicated in extracellular matrix regulation (31) and in cancer as well as GMDS (32).

Common variants in CDKN2B and its antisense CDKN2B-AS1 have been robustly found associated with POAG. The gene CDKN1A, also known as p21, CIP-1 or WAF-1, belongs to the same family as CDKN2B and also encodes a cyclin-dependent kinase inhibitor. Upregulation of CDKN1A causes G1 arrest and inhibits proliferation of the cell. Herein, for the first time, we provide genome-wide significant evidence for association of CDKN1A variants with cup area. Two prior small cohort studies suggested a possible role of CDKN1A in POAG. Tsai et al. (33) found an association between a codon 31 polymorphism in CDKN1A and POAG in 58 patients and 59 controls from China (OR = 2.39 [1.14–5.01]). Saglar et al. found no statistically significant association between the codon 31 polymorphism and POAG in 75 patients and 119 controls from Turkey (OR = 1.70, P-value = 0.25) (34). Our study provides strong evidence for the role of CDKN1A in POAG risk in a large sample consisting of 6,429 cases and 41,404 controls and shows the first convincing evidence for an association of CDKN1A and POAG in individuals of European descent. A potential functional interaction between POAG-loci (i.e. SIX6 and CDKN genes) has been shown previously (26,27). We performed in vivo studies in embryonic zebrafish eye and found that knockdown of six6b upregulates both cdkn2a/cdkn2b and cdkn1a. The mechanism behind the genetic variation in SIX6, and the CDKN genes remains obscure. However, in a recent study, Skowronska-Krawczyk et al. showed that SIX6 regulates the expression of CDKN2A in the context of IOP elevation (26). More comprehensive studies at the individual tissue level e.g. retinal ganglion cell layer or optic nerve should be performed to evaluate the consequences of the genetic variation associated with POAG; for example, if associated variants in SIX6 lead to up-regulation of CDKN genes.

Genetic variation in CDK inhibitor genes has been found associated with various diseases and traits. SNPs close to CDKN1A, for example, have been associated with electrocardiographic measures (35) and colorectal cancer (36). While the CDKN2A/CDKN2B and CDKN2B-AS1 locus has been found in GWAS of Type 2 Diabetes (37), fasting glucose (38), intracranial aneurysm (39), myocardial infraction (40), coronary heart disease (41) and various cancers, including basal cell carcinoma (42), breast (43) and prostate (44) cancer, and glioma (45). Interestingly, the same allele in CDKN2B that decrease risk of POAG, was found associated with increased risk of glioma. The functional significance of the genetic variation in the CDK genes remains to be determined, possible mechanism in POAG might be related with the disruption of a delicate balance between proliferation and apoptosis.

It has been suggested that p53 plays a role in POAG, especially in POAG with paracentral visual field loss (46). Intriguingly, p53 has been also related to the CDK-inhibitors and to four of the new genes pointed out by this study (GADD45A, PDZD2, RREB1 and PSCA). GADD45A is involved in growth arrest through p53-dependent and independent mechanisms (47,48) and can interact via CDKN1A (49). PDZD2 is a tumour suppressor gene, reported to activate p53 by transcriptional regulation (50). RREB1 has an effect on p53 by binding to its promotor and transactivates its expression (51). While PSCA expression correlates with the expression of p53 in choriocarcinoma, suggesting a role of PSCA in cell growth through p53-related pathways (52). Other genes play a role in apoptosis or cell growth via other mechanisms than p53: VGLL4 inhibits Bax- and TNFa-induced apoptosis (53) and DGKB is a regulator of diacylglycerol, which is important for cell growth and differentiation. UGT8 plays a role in the biosynthesis of the sphingolipids of myelin membranes of the central and peripheral nervous system; sphingolipids are also implicated in apoptosis (54).

Another interesting novel gene is RPE65 (retinal pigment epithelium -specific protein 65kDa). This gene has been associated with retinitis pigmentosa (RP) (55,56) and Leber congenital amaurosis type 2 (LCA2) (57). As the name implies, the encoded protein is located in the retinal pigment epithelium (58). Both diseases (RP and LCA2) are not characterized by an excavation of the optic nerve head. However, we have checked several online databases for expression in different tissues. In the eye, it is also highly expressed in the optic nerve head (Tables S8 and S9) suggesting that this gene could be involved in other ocular processes. Little expression is found in the brain, with no expression in other tissues or organs in the body. Future studies are necessary to confirm our finding.

Of the genes identified by gene-based testing, C9 (complement component 9) is especially interesting. Its protein is part of the membrane attack complex (MAC), together with the proteins C5b, C6, C7, and C8. This complex activates several steps that lead to cell death, and cells protect themselves by removing the complex through endocytosis. Caveolin is one of the proteins involved in endocytosis and the CAV1/CAV2 genes are associated with IOP and POAG. It has been shown that inhibition of caveolin-1 inhibits the endocytosis of MAC (59).

To our best knowledge, this meta-analysis is the largest study of IOP and optic nerve head parameters to date, using well-characterized datasets from populations world-wide. A limitation of our study is the lack of an available dataset for replication of the novel associations detected by combined European and Asian ancestry samples. However, the heterogeneity of these novel genomic regions is generally low in the meta-analysis. For VCDR, cup area, and disc area we have identified novel SNPs in the analysis of individuals with European ancestry. Of the nine novel associations found in these populations (RPE65, PDZD2, RREB1, DGKB for VCDR; CDC42BPA, CRISPLD1 and FAM169B for cup area; and CTNNA3 and UGT8 for disc area), only RREB1 was nominally significant in the individuals with Asian ancestry. Five of the seven non-significant SNPs in the individuals with Asian ancestry had an effect estimate in the same direction. As the analysis in individuals with Asian ancestry contains a smaller number of individuals, this could be due to lack of power.

We have identified 21 genetic variants associated with POAG endophenotypes: IOP, VCDR, cup area, and disc area. The effect estimates of single SNP associations are small. We expected small effects of many SNPs, since glaucoma endophenotypes are highly polygenic traits (60). Although individual SNP effects are small, the joint effect of the various genetic variants could yield a clinical relevant parameter. These association results do not imply that the variants described here have a causal effect. Fine-mapping and functional studies are required to identify the causal variants tagged by our findings and the exact molecular mechanisms involved in POAG. Although the overlap between IOP-loci and the optic disc parameters-loci is not large, this is the first study showing a genome-wide significant evidence of the genetic correlation between IOP and VCDR; we expect that larger sample sizes and improved imputation accuracy may help to find more of the loci underlying the genetic correlation between these two endophenotypes. Of the novel associations, CDKN1A is strongly associated with POAG, This finding is in line with other studies (26), pointing to the CDK-inhibitor genes as key players in the development of POAG. The p53 pathway has been implicated in POAG and interact in different cellular contexts with four of the new genes pointed out by this study (GADD45A, PDZD2, RREB1 and PSCA). Functional studies need to be performed to assess the role of p53 and CDK-inhibitors in the pathophysiology of POAG. A more comprehensive study of these mechanisms may inform the development of new therapies for POAG.

Materials and Methods

Study design

We performed a meta-analysis on directly genotyped and imputed SNPs to the 1000 Genomes reference panel. We analyzed four outcomes: IOP, VCDR, cup area, and disc area; the same approach was used for all four outcomes. In the first stage, we conducted a meta-analysis of GWAS including 22,489–29,578 individuals of European ancestry. Subsequently, we evaluated the genome-wide significant SNPs from the first stage in a meta-analysis of GWAS including 7,307–8,373 individuals of Asian ancestry. Finally, we performed a mega/meta-analysis of GWAS from all individual studies including individuals of European and Asian ancestry. We subsequently tested the effect of all genome-wide significant SNPs on POAG in four independent case-control studies (n = 6,429/41,404). These cases are not part of our primary meta-analyses which focused on quantitative traits in the general population (IOP, VCDR, cup area and disc area). Three of the case-control studies were of European ancestry (7,61,62) and one of Asian ancestry (63).

Subjects, phenotyping and genotyping

All 19 studies included in this meta-analysis are part of the IGGC (Table S1A). Details for each individual study can be found in Supplementary Material and Tables S1B and C and S2. The ophthalmological examinations included measurements of IOP and optic nerve head assessment. All 19 studies contributed to the IOP mega/meta-analysis, 18 to the VCDR and 16 to the cup area and disc area mega/meta-analysis. Validation of mega/meta-analysis results was performed in four independent POAG case-control studies. Studies performed genomic imputation using 1000 Genomes phase 1 reference samples. Study-specific quality control can be found in the Supplementary Material. All studies were performed with the approval of their local medical ethics committee, and written informed consent was obtained from all participants in accordance with the Declaration of Helsinki.

Statistical analysis

In the IOP analysis, individuals who underwent IOP-lowering laser or surgery were removed from the analysis; in individuals receiving IOP-lowering medication, the IOP value was multiplied by 1.3 to estimate a pre-medication IOP value (64). The mean IOP, VCDR, cup area, and disc area of both eyes was used for the analyses. SNPs with MAF < 0.01 and imputation quality scores <0.3 (proper-info of IMPUTE) or R² < 0.3 (MACH) were removed from the analyses. Each individual study performed a linear regression between each endophenotype (IOP, VCDR, cup area, and disc area) and the SNPs, under the assumption of an additive model for the effect of the risk allele. Analyses were adjusted for age, sex and the first five principal components (for population-based studies) or family structure (for family-based studies).

We performed an inverse variance weighted fixed-effect meta-analysis with METAL software (65). P values for heterogeneity were calculated by using the Cochran’s Q-test for heterogeneity. SNPs with a P-value for heterogeneity <0.001 were removed from the results, as well as SNPs only present in three studies. We used the ‘genomic control’ option in METAL to correct the standard error of each individual study for estimated genomic inflation. To evaluate if our results exhibit signs of population stratification we estimated the intercept using the LD score regression method (18) in the European-only meta-analyses. The intercept, which estimates inflation after removing polygenic signals, was: 1.0142 for IOP, 1.0199 for VCDR, 1.0074 for cup area and 1.0167 for the disc area (Table S16).

In the meta-analyses of individuals with European ancestry, a P-value <5.0 × 10⁻⁸ (the genome-wide threshold of association) was considered significant. In the second stage, these genome-wide significant SNPs were validated in individuals with Asian ancestry, and in this look-up a p value <0.05 was considered significant. Finally, in the mega/meta-analysis of individuals with European and Asian ancestry a P-value of <5.0 × 10⁻⁸ was considered significant. In total, we identified 75 independent SNPs across different genomic regions for all the traits together. Therefore, the significance level after Bonferroni correction in the meta-analysis of POAG cohorts was = 6.67 × 10⁻⁴ (0.05/75 independent SNPs). To estimate the common effect size of the top SNPs associated with IOP, optic disc parameters and their effect in the look-up in the POAG cohorts a random-effect meta-analysis was performed using plink (66) http://pngu.mgh.harvard.edu/purcell/plink/ parameter –meta-analysis. Manhattan, regional and forest plots were made using R (67) and LocusZoom (68).

The genetic correlation between optic disc parameters (VCDR, cup and disc area) and IOP was estimated using the LD score regression method (18,69). GWAS summary statistics from the Europeans-only meta-analyses were used to calculate the genetic overlap. To restrict the analyses to well imputed SNPs, we included only SNPs with MAF > 0.01 that were present in HapMap3. GWAS data were harmonized using the “munge_sumstats.py” function. For analyses, we used the pre-computed LD scores for Europeans “eur_w_ld_ch” available at https://github.com/bulik/ldsc/wiki. The ldsc.py function (with all default settings) was used to calculate the genetic correlation between traits. In addition, we calculated the heritability (h²) estimates from the GWAS meta-analysis using the same function.

Gene-based test using VEGAS

A gene-based test was performed using the VEGAS2 software (70), with a 50kb gene boundary. We used the parameter ‘-top 100’ (default) to perform gene-based tests. This parameter considers association test statistics of all variants mapped to a gene to compute gene-based test statistics. The 1000 Genomes European and Asian populations were used as a reference to calculate LD for European and Asian ancestry data respectively. After calculation of gene-based test statistics for Asian and European ancestry populations separately, meta-analyses were conducted using Fisher’s method for combining P-values.

Functional characterization, expression data, zebrafish and gene-set enrichment

We investigated for evidence of regulatory functions of associated loci HaploReg version 2 (71) and Regulomedb version 1.1 (72). We investigated the expression of the associated genes using NCBI’s UniGene (23) and The Ocular Tissue Database (24). We also investigated the expression of cdkn1a in a six6b knockdown zebrafish and used DEPICT to investigate gene-set enrichment. More information about these analyses can be found in the Supplementary Material.

Data access

Access to the summary association statistics for IOP and optic disc parameters is available at https://www.dropbox.com/sh/3j2h9qdbzjwvaj1/AABFD1eyNetiF63I5bQooYura?dl=0.

Supplementary Material

Supplementary Material is available at HMG online.

Funding

Terri L. Young, MD, MBA is supported by National Institutes of Health/ National Eye Institute (NIH/ NEI) 1R01EY018246-01, NIH/ NEI R01 EY014685, Research to Prevent Blindness, Inc. and The University of Wisconsin School of Medicine and Public Health Centennial Scholars Fund. Blue Mountains Eye Study was supported by the Australian National Health & Medical Research Council (NH&MRC), Canberra Australia (974159, 211069, 457349, 512423, 475604, 529912); the Centre for Clinical Research Excellence (Translational Clinical Research in Major Eye Diseases, 519923); NH&MRC research fellowships (358702, 632909 to J.J.W); and the Wellcome Trust, UK as part of Wellcome Trust Case Control Consortium 2 (A. Viswanathan, P. McGuffin, P. Mitchell, F. Topouzis, P. Foster) for genotyping costs of the entire BMES population (085475/B/08/Z, 085475/Z/08/Z, 076113). The Southampton acknowledges funding from the UK and Eire Glaucoma Society and the International Glaucoma Association (in association with the Royal College of Ophthalmologists). EY015473 (LRP) supported generation of the cohort at risk for POAG in Nurses Health Study (NHS, UM1 CA186107) and Health Professionals Followup Study (HPFS, UM1 CA167552). HG004728 (LRP) supported genotyping in a subset of the Neighborhood consortium, specifically NHS, HPFS and Massachusetts Eye and Ear Infirmary (MEEI) cases and controls. We acknowledge funding from National Institutes of Health NIH R01 EY022305 (JLW) and NEI grant K08EY022943 (RW). Genetic analyses for the Orkney Complex Disease Study (ORCADES) were supported by the MRC HGU “QTL in Health and Disease’ core programme.