Genome-wide association meta-analysis of 88,250 individuals highlights pleiotropic mechanisms of five ocular diseases in UK Biobank

Abstract

Background: Ocular diseases may exhibit common clinical symptoms and epidemiological comorbidity. However, the extent of pleiotropic mechanisms across ocular diseases remains unclear. We aim to examine shared genetic etiology in age-related macular degeneration (AMD), diabetic retinopathy (DR), glaucoma, retinal detachment (RD), and myopia.

Methods: We analyzed genome-wide association analyses for the five ocular diseases in 43,877 cases and 44,373 controls of European ancestry from UK Biobank, estimated their genetic relationships (LDSC, GNOVA, and Genomic SEM), and identified pleiotropic loci (ASSET and METASOFT).

Findings: The genetic correlation of common SNPs revealed a meaningful genetic structure within these diseases, identifying genetic correlations between AMD, DR, and glaucoma. Cross-trait meta-analysis identified 23 pleiotropic loci associated with at least two ocular diseases and 14 loci unique to individual disorders (non-pleiotropic). We found that the genes associated with these shared genetic loci are involved in neuron differentiation (P = 8.80 × 10^-6) and eye development systems (P = 3.86 × 10^-5), and single cell RNA sequencing data reveals their heightened gene expression from multipotent progenitors to other differentiated retinal cells during retina developmental process.

Interpretation: These results highlighted the potential common genetic architectures among these ocular diseases and can deepen the understanding of the molecular mechanisms underlying the related diseases.

Funding: The National Natural Science Foundation of China (61871294), Zhejiang Provincial Natural Science Foundation of China (LR19C060001), and the Scientific Research Foundation for Talents of Wenzhou Medical University (QTJ18023).

Keywords: Cross-disease genetics; GWAS; Genetic correlation; Ocular diseases; Pleiotropy; Retinal development.

Figure 1

Genetic relationships across five ocular diseases. (a) Manhattan plots of GWAS results among five ocular diseases. The X-axis is the base-pair position, and the Y-axis is the -log₁₀-transformed P-value for each SNP. The red line indicates genome-wide significance (P < 5 × 10⁻⁸), and the blue line represents a suggestive significance (P < 1 × 10⁻⁵). (b) SNP-based genetic correlations (r_g) were estimated between pairs of ocular diseases by GNOVA. The colour and size of each circle indicate the magnitude of the r_g. Asterisks indicate nominal significance (P < 0.05), and double asterisks represent statistical significance after Bonferroni correction (P < 0.05/10). (c) An exploratory factor analysis (EFA) and a confirmatory factor analysis (CFA) were conducted on the GWAS summary statistics using Genomic SEM. Here we showed the standardized estimates. F1_g represents a shared genetic factor among AMD, DR, and glaucoma, while F2_g represents a common genetic factor between RD and myopia. Arrows connecting the factors to the individual diseases represent regression coefficients of the genetic liability for the diseases on the common factor. The arrow connecting the two factors represents their correlation. Two-headed arrows linking the genetic components of the individual ocular diseases to themselves represent residual genetic variances, which can be interpreted as the proportion of heritable variation unexplained by the factors. SEs are shown in parentheses.

Figure 2

Results of cross-traits meta-analysis by ASSET based on 88,250 individuals. (a) Quantile-quantile (QQ) plot of the meta-analysis displaying the observed significance versus the expected significance for each variant. (b) Manhattan plot of the meta-analysis with the X-axis showing genomic position and the Y-axis showing the significance on a -log₁₀ scale for each SNP. The red and blue lines represent the thresholds for genome-wide significance (P_ASSET = 5 × 10⁻⁸) and suggestive associations (P_ASSET = 1 × 10⁻⁵), respectively. (c) Distribution of index SNPs and credible SNPs in functional consequences, minimum chromatin state across 127 tissue and cell types, and RegulomeDB score (The lower the score, the more likely the SNP is to have a regulatory function). (d) Heritability enrichment of 22 functional SNP annotations by stratified LDSC. The X-axis shows the proportion of SNPs in each region, and the Y-axis displays the enrichment, estimated as the proportion of heritability / the proportion of SNPs. The dashed line represents enrichment of 1. Error bars show 95% confidence intervals. TSS, transcription start site; CTCF, CCCTC binding factor; DHS, DNase I hypersensitivity site; TFBS, transcription factor binding site; DGF, digital genomic footprint.

Figure 3

Three most pleiotropic loci associated with all five ocular diseases. (a) Regional association plots of the three loci: 4q21.21 (index SNP rs7678123), 10q26.3 (index SNP rs12570944), and 11q14.2 (index SNP rs9667489). (b) Forest plots with PM-plots show disease-specific effects of the index SNP in each locus. Forest plots display the P-value in METASOFT meta-analysis (Meta P) and the P-value, log(OR) and its standard error of the SNP in the GWAS of individual diseases. PM-plots visualize the posterior probability (m-value, X-axis) of the SNP in each study with disease-specific association significance as -log₁₀(P) (Y-axis). M-values > 0.9 (coloured in red) suggests that the SNP does have an effect on the disease. The dot size represents the GWAS sample size estimated from summary statistics.

Figure 4

Gene mapping of meta-analysis results. (a) Three gene mapping strategies for the index SNP and credible SNPs of each locus. We mapped these SNPs to the protein-coding genes within 10 kb, mapped cis-eQTL markers to their target genes, and mapped the SNPs in the genomic regions interacting with gene promoter regions to corresponding genes. (b) GO pathway enrichment for pleiotropic versus non-pleiotropic loci. This network shows the terms with P < 0.01, a minimum gene count of 3, and an enrichment factor > 1.5. The nodes sizes are scaled with P-value. The genes implicated in pleiotropic loci are most significantly enriched in eye development and neuron differentiation. (c) GTEx tissue-specific enrichment results for 37 loci associated with at least one ocular disease. 55 GTEx tissues were classified as 10 categories, and the retina were coloured in dark blue. The red dotted line represents the P-value threshold after Bonferroni correction (P = 0.05/55 = 9.09 × 10⁻⁴). Genomic risk loci show significant enrichment in genes specifically expressed in retina. (d) Retinal development RNA-seq data from 4.7 PCW to adult was used to plot the normalized expression of 34 genes of pleiotropic loci and 18 genes of non-pleiotropic loci with significant expression variation during development. (e) Left: cell classes marked on scVis map. Arrows represents developmental trajectories. Right: ScVis map displays the expression trajectories of candidate genes during the development of retinal organoids. The genes implicated in pleiotropic loci highly expressed in multipotent progenitors, GCs, HCs, cones, rods, MCs, and RPEs, successively, whereas the genes mapped to non-pleiotropic loci only had high expression in cones and rods.