Abstract
BACKGROUND: Genome-wide association studies of Alzheimer's disease (AD) have identified a number of significant risk loci, the majority of which lie in non-coding regions of the genome. The lack of causal alleles and considerable polygenicity remains a significant barrier to translation into mechanistic understanding. This includes identifying causal variants and the cell/tissue types in which they operate. A fuller understanding of the cell types and transcriptional networks involved in AD genetic risk mechanisms will provide important insights into pathogenesis. METHODS: We assessed the significance of the overlap between genome-wide significant AD risk variants and sites of open chromatin from data sets representing diverse tissue types. We then focussed on macrophages and microglia to investigate the role of open chromatin sites containing motifs for specific transcription factors. Partitioned heritability using LDscore regression was used to investigate the contribution of specific macrophage and microglia transcription factor motif-containing open chromatin sites to the heritability of AD. RESULTS: AD risk single nucleotide polymorphisms (SNPs) are preferentially located at sites of open chromatin in immune cells, particularly monocytes (z score = 4.43; corrected P = 5.88 × 10(- 3)). Similar enrichments are observed for macrophages (z score = 4.10; corrected P < 2.40 × 10(- 3)) and microglia (z score = 4.34, corrected P = 0.011). In both macrophages and microglia, AD risk variants are enriched at a subset of open chromatin sites that contain DNA binding motifs for specific transcription factors, e.g. SPI1 and MEF2. Genetic variation at many of these motif-containing sites also mediate a substantial proportion of AD heritability, with SPI1-containing sites capturing the majority of the common variant SNP-chip heritability (microglia enrichment = 16.28, corrected enrichment P = 0.0044). CONCLUSIONS: AD risk alleles plausibly operate in immune cells, including microglia, and are concentrated in specific transcriptional networks. Combined with primary genetic association results, the SPI1 and MEF2 transcriptional networks appear central to AD risk mechanisms. Investigation of transcription factors targeting AD risk SNP associated regulatory elements could provide powerful insights into the molecular processes affected by AD polygenic risk. More broadly, our findings support a model of polygenic disease risk that arises from variants located in specific transcriptional networks.