18Age-related macular degeneration (AMD) is a leading cause of visual impairment in ageing 19 populations and has no radical treatment or prevention. Although genome-wide association studies 20 (GWAS) have identified many susceptibility loci for AMD, the underlying causal genes remain 21 elusive. Here, we prioritized nine putative causal genes by integrating expression quantitative trait 22 locus (eQTL) data from blood ( = 2,765) with AMD GWAS data (16,144 cases vs. 17,832 controls) 23 and replicated six of them using retina eQTL data ( = 523). Of the six genes, altering expression of 24 cnn2, sarm1 and bloc1s1 led to ocular phenotype, impaired vision and retinal pigment epithelium 25 (RPE) loss in zebrafish. Essential photoreceptor and RPE genes were downregulated in cnn2-and 26 sarm1-knockdown zebrafishes. Through integration of GWAS and eQTL data followed by functional 27 validation, our study reveals potential roles of CNN2, SARM1 and BLOC1S1 in AMD pathogenesis 28 and demonstrates an efficient platform to prioritise causal genes for human complex diseases. 29 2 Introduction 30 Age-related macular degeneration (AMD) is an incurable blinding disorder caused by dysfunction of the 31 retinal pigment epithelium (RPE) and progressive loss of photoreceptors in the macula 1 . It results in visual 32 impairment of central vision and disability of daily life activities, such as reading, walking and face 33recognition. The prevalence of AMD is 8.69% in the age range of 45-85 years globally, and it is projected 34 to affect 196 million people worldwide in 2020 2 . As such, AMD is highly endorsed as a major health and 35 social problem for both individuals and communities, especially in elderly populations 3 . 36 37 AMD is one of the most genetically well-defined complex diseases. Genome-wide association studies 38 (GWAS) with increasing sample sizes have identified 52 susceptibility loci which together explain more 39 than 50% of the heritability of liability 4-6 . These findings provide important clues for understanding the 40 genetic architecture of the disease, but the causal genes at those susceptibility loci and underlying 41 mechanisms remain largely unclear. For example, a nonsynonymous variant, CFH p. Arg1210Cys (allele 42 frequency = 0.00017 in ExAC), increases AMD risk by >20-fold 7 , but there has been no evidence showing 43 its functional impact on the regulation of gene expression, structural and functional integrity of the protein-44 coding region, or interplay with the genes nearby 8 . This is partly because of linkage disequilibrium (LD) 45 between single-nucleotide polymorphisms (SNPs) and causative variants that GWAS mapping resolution 9 . 46This could also be the reason that the trait-associated variants, especially those residing in non-coding 47