Mendelian randomization identifies proteins involved in neurodegenerative diseases

Proteins are involved in multiple biological functions. High-throughput technologies have allowed the measurement of thousands of proteins in population biobanks. In this study, we aimed to identify proteins related to Alzheimer's disease, Parkinson's disease, multiple sclerosis and amyotrophic lateral sclerosis by leveraging large-scale genetic and proteomic data. We performed a two-sample cis Mendelian randomization study by selecting instrumental variables for the abundance of >2700 proteins measured by either Olink or SomaScan platforms in plasma from the UK Biobank and the deCODE Health Study. We also used the latest publicly available genome-wide association studies for the neurodegenerative diseases of interest. The potentially causal effect of proteins on neurodegenerative diseases was estimated based on the Wald ratio. We tested 13 377 protein-disease associations, identifying 169 associations that were statistically significant (5% false discovery rate). Evidence of co-localization between plasma protein abundance and disease risk (posterior probability > 0.80) was identified for 61 protein-disease pairs, leading to 50 unique protein-disease associations. Notably, 23 of 50 protein-disease associations corresponded to genetic loci not previously reported by genome-wide association studies. The two-sample Mendelian randomization and co-localization analysis also showed that APOE abundance in plasma was associated with three subcortical volumes (hippocampus, amygdala and nucleus accumbens) and white matter hyper-intensities, whereas PILRA and PILRB abundance in plasma was associated with caudate nucleus volume. Our study provided a comprehensive assessment of the effect of the human proteome that is currently measurable through two different platforms on neurodegenerative diseases. The newly associated proteins indicated the involvement of complement (C1S and C1R), microglia (SIRPA, SIGLEC9 and PRSS8) and lysosomes (CLN5) in Alzheimer's disease; the interleukin-6 pathway (CTF1) in Parkinson's disease; lysosomes (TPP1), blood-brain barrier integrity (MFAP2) and astrocytes (TNFSF13) in amyotrophic lateral sclerosis; and blood-brain barrier integrity (VEGFB), oligodendrocytes (PARP1), node of Ranvier and dorsal root ganglion (NCS1, FLRT3 and CDH15) and the innate immune system (CR1, AHSG and WARS) in multiple sclerosis. Our study demonstrates how harnessing large-scale genomic and proteomic data can yield new insights into the role of the plasma proteome in the pathogenesis of neurodegenerative diseases.