Evaluating the associations between intelligence quotient and multi-tissue proteome from the brain, CSF and plasma

Intelligence quotient is a vital index to evaluate the ability of an individual to think rationally, learn from experience and deal with the environment effectively. However, limited efforts have been paid to explore the potential associations of intelligence quotient traits with the tissue proteins from the brain, CSF and plasma. The information of protein quantitative trait loci was collected from a recently released genome-wide association study conducted on quantification data of proteins from the tissues including the brain, CSF and plasma. Using the individual-level genotypic data from the UK Biobank cohort, we calculated the polygenic risk scores for each protein based on the protein quantitative trait locus data sets above. Then, Pearson correlation analysis was applied to evaluate the relationships between intelligence quotient traits (including 120 330 subjects for 'fluid intelligence score' and 38 949 subjects for 'maximum digits remembered correctly') and polygenic risk scores of each protein in the brain (17 protein polygenic risk scores), CSF (116 protein polygenic risk scores) and plasma (59 protein polygenic risk scores). The Bonferroni corrected P-value threshold was P < 1.30 × 10^-4 (0.05/384). Finally, Mendelian randomization analysis was conducted to test the causal relationships between 'fluid intelligence score' and pre-specific proteins from correlation analysis results. Pearson correlation analysis identified significant association signals between the protein of macrophage-stimulating protein and fluid intelligence in brain and CSF tissues (P _brain = 1.21 × 10^-8, P _CSF = 1.10 × 10^-7), as well as between B-cell lymphoma 6 protein and fluid intelligence in CSF (P _CSF = 1.23 × 10^-4). Other proteins showed close-to-significant associations with the trait of 'fluid intelligence score', such as plasma protease C1 inhibitor (P _CSF = 4.19 × 10^-4, P _plasma = 6.97 × 10^-4), and with the trait of 'maximum digits remembered correctly', such as tenascin (P _plasma = 3.42 × 10^-4). Additionally, Mendelian randomization analysis results suggested that macrophage-stimulating protein (Mendelian randomization-Egger: β = 0.54, P = 1.64 × 10^-61 in the brain; β = 0.09, P = 1.60 × 10^-12 in CSF) had causal effects on fluid intelligence score. We observed functional relevance of specific tissue proteins to intelligence quotient and identified several candidate proteins, such as macrophage-stimulating protein. This study provided a novel insight to the relationship between tissue proteins and intelligence quotient traits.