Integrative functional genomic search for regulatory DNA sequence polymorphisms influencing DNA methylation and mRNA expression in hippocampal brain tissue

Neuropsychiatric disorders have a strong genetic predisposition, but their genetic basis remains elusive. Genome-wide association studies (GWASs) have mapped more than 2,000 susceptibility loci that were shown to increase the risk of common brain disorders. However, the majority of these susceptibility loci reside in non-coding regions and their functional consequences are unknown. The present study addresses the question whether regulatory sequence variants, affecting DNA methylation and gene expression, may be causal susceptibility alleles. I used an integrative functional genomics approach to investigate epigenetic regulation phenomena in human hippocampal brain of 115 European patients with pharmacoresistant mesial temporal lobe epilepsy. High-density SNP genotypes were correlated with genome-wide quantitative CpG methylation and mRNA expression levels using the Human Methylation450 array (HM450) and the Human HT-12 v3 array. Subsequently, a genome-wide map of methylation quantitative trait loci (meQTLs) and expression quantitative trait loci (eQTLs) was used to dissect regulatory SNPs (rSNPs) that confer susceptibility to common brain disorders at 488 known GWAS hits (P < 5.0 x 10-8). This is the first meQTL study of brain tissue applying the high-density HM450 array in specimens of fresh frozen human brain tissue obtained by epilepsy surgery at large scale. Linear regression analysis of this study implementing a correction for cell-type heterogeneity, identified 19,954 (8.5% of 362k CpGs) cis-acting meQTLs at a false-discovery rate (FDR) of 5%, which is a six-fold increase compared to previous meQTL studies that all investigated postmortem brain tissue. Specifically, cis-meQTLs were strongly enriched upstream of the gene promoter region (TSS201-1500; P = 7.7 x 10-61), highlighting the functional impact of this 5´-regulatory region that harbors binding sites of enhancers and insulators. Some of the most significant cis-meQTLs affected high-ranking candidate genes (ADARB2, HDAC4, NAPRT1, MAD1L1, PTPRN2 and RIMBP2) for neurodevelopmental disorders. To explore tissue specifity, the same approach was repeated in an additional meQTL analysis of whole blood cells originating from 496 German population controls without neuropsychiatric disorders. Results show that 65% of the meQTLs in brain tissues were also present in whole blood cells (Spearman’s Rank coefficient = 0.42). The present database of cis-meQTLs in brain and blood cells provides a key to select accessible epigenetic biomarkers for brain disorders in whole blood cells. The performed eQTL study identified 734 out of 31k mRNA probes at which expression levels were significantly influenced by cis-acting SNPs (FDR < 5%). Apart from meQTL and eQTL analyses, additionally a CpG methylation to gene expression correlation analysis was performed. This represents the first systematic delineation of methylation-driven genes in fresh frozen brain tissue. Both inverse correlations (73%) and positive correlations (27%) were observed, whereby the strongest inverse correlations were detected at NAPRT1, the gene encoding Nicotinate Phosphoribosyltransferase. Furthermore, the NAPRT1-associated meQTLs and eQTL were both genetically regulated by SNP rs9657360. The minor C allele of that very SNP was significantly associated with high methylation levels in the NAPRT1 promoter region and simultaneously associated with low gene expression of NAPRT1. Both, the tumor-specific hypermethylation of a promoter CpG island as well as loss of NAPRT1 expression have been previously proposed as predictive biomarkers for the therapy of carcinomas using NAMPT inhibitors. The additionally genetic risk constellation which has been identified by my approach – combining meQTLs and eQTLs to unravel the translational impact of epigenetic regulation of gene expression – is of high clinical relevance. It enables a diagnostically driven clinical strategy in tumorigenesis including the selection of patients which likely benefit from the administration of NAMPT inhibitors. To dissect imprinted meQTLs (imeQTLs) exhibiting differential methylation in a Parent-of-Origin (PofO) dependent manner, the CpG methylation states of blood cells in groups of 269 individuals stratified by parentally inverse heterozygous genotypes of nearby SNPs were compared. The imeQTL analysis revealed 177 CpGs at 31 genomic loci of which 22 were previously unknown. The strongest PofO effects were observed at loci harboring neurodevelopmental genes and on chromosome 3p21.1, which is a GWAS candidate region for mood disorders. Genes at genomic loci that show imprinting effects are promising candidate genes because of their potentially monoallelic gene expression which may unmask recessive susceptibility alleles. Enrichment analyses of genes associated with cis-meQTLs revealed an overrepresentation of genes implicated in GWAS hits of brain disorders (P = 5.8 x 10-4). Potential rSNPs at the GWAS candidate loci 1q31.2 (RGS1 gene locus) and 3p21.1 (PRBM1 gene locus) were identified. The allelic alteration of transcription factor binding sites by potential rSNPs is likely to result in changes of gene transcription or splicing processes which could contribute to pathogenic pathways underlying neuropsychiatric disorders. As exemplified in this thesis, the created database of autosomal meQTLs, imeQTLs and eQTLs in brain tissue provides a valuable resource to dissect rSNPs at GWAS hits and to decipher their functional effects