Integration of GWAS and cerebrospinal fluid proteogenomics identifies proteins for AD and shared dysregulated mechanisms in other neurological traits

Abstract Background Cerebrospinal fluid (CSF) is a valuable resource for the study and diagnosis of neurological diseases, but few studies have comprehensively characterized the genetic determinants of CSF protein levels that may contribute to the development of disease. These quantitative trait loci (QTL) have proven vital to identifying candidate genes for disease treatment and monitoring. Here, we utilize our largest‐to‐date CSF protein QTL atlas to prioritize potentially causal proteins for 14 neurological traits and examine the unique and overlapping disease mechanisms observed using CSF proteins. Method We generated the largest CSF proteogenomic dataset (3,506 samples, 6,361 proteins) to date. We performed proteome‐wide association study (PWAS), colocalization, and Mendelian Randomization (MR) to identify potentially causal proteins for 14 neurodegenerative and psychiatric diseases or disorders. We then performed pathway analysis to identify dysregulated mechanisms underlying disease and used RHOGE to assess the overlap of proteogenomic profiles between diseases. We are repeating these analyses using blood plasma to assess overlap with a more accessible tissue. Result Across the 14 diseases, we identified 171 protein‐disease associations in CSF through at least two of PWAS, COLOC, and MR. For Alzheimer’s disease (AD), we identified 38 proteins enriched in immune‐relevant pathways (TREM2, CD33, IL34) and lysosomal function (GRN, CLN5, TMEM106B). Many proteins were found for schizophrenia (38), multiple sclerosis (27), and bipolar disorder (22), among others. We identified negative genetic correlations between PD and neuroticism and between anorexia nervosa and most other psychiatric traits. We observed limited overlap between CSF and plasma pQTLs (1,225/4,767 colocalize; 25.7%) with a moderate correlation in pQTL effect size ( R = 0.576). Disease‐specific analyses in plasma are ongoing. Conclusion We have identified numerous novel associations for neurological diseases that may represent therapeutic or biomarker targets in disease. We highlighted substantial differences in proteogenomic regulation between CSF and plasma, demonstrating the importance of investigating novel tissues in genetic studies. Through the integration of analyses from both tissues (ongoing), we will uncover disease‐relevant changes specific to neurological tissues or shared between tissues that may allow for diagnosis or prioritization of individuals for treatment based solely on proteomics generated from accessible tissues.