Large‐scale pQTL analysis of over 3,000 CSF samples

Abstract

AbstractBackgroundExpression quantitative trait locus (eQTL) studies have been influential in identifying gene targets for many complex diseases. However, levels of mRNA often correlate poorly with levels of the protein that they encode, due to rates of degradation, translation, or alternative splicing, among other reasons. By correlating protein levels to genetic variation, pQTLs can elucidate a more accurate genetic architecture that underlies disease. While few large‐scale studies have considered cerebrospinal fluid (CSF) protein levels, CSF biomarker levels are one of the primary diagnostic tools in Alzheimer’s disease (AD), highlighting CSF’s relevance to brain aging and AD pathology. Here, we present CSF pQTL analysis of over 3,000 individuals.MethodWe generated CSF levels of 7,584 analytes for 3,065 samples using the aptamer‐based SOMAscan 7k platform. CSF levels of 4,884 analytes for additional 1,158 samples were also included. After stringent quality control, 3,107 unrelated European samples were selected for analysis. We performed pQTL analysis using a linear model with protein levels as the outcome variable, using a three‐stage study design: discovery, replication, and meta‐analysis. We performed colocalization of our pQTLs with eQTLs and GWAS loci, and compared our results to pQTLs from plasma studies to identify CSF‐specific pQTLs. We used Mendelian Randomization (MR) to identify proteins potentially causative for AD and other traits.ResultThis study represents the first use of the SOMAscan 7k platform in CSF, as well as the largest pQTL analysis of CSF to date. Of 3818 analytes currently tested in our discovery set, 1480 have at least one SNP associated with analyte level at p < 5×10−8. Analyses are ongoing.ConclusionWe previously performed pQTL analysis of CSF using the SOMAscan 1.3k platform, where 425 pQTLs were identified and 3 proteins (PTP1B, Siglec‐3, and SLAF5) were found through MR to be involved in AD risk, as well as 48 additional protein associations for other neurodegenerative diseases. This study expands on those findings by including 6,000 more proteins and much larger samples, allowing for greater power. This study will be vital for identifying novel proteins involved in AD and elucidate novel drug targets for AD and other neurodegenerative disorders.