Circulating Blood-Based Proteins in Psychopathology and Cognition: A Mendelian Randomization Study

Abstract

Importance: Peripheral (blood-based) biomarkers for psychiatric illness could benefit diagnosis and treatment, but research to date has typically been low throughput, and traditional case-control studies are subject to potential confounds of treatment and other exposures. Large-scale 2-sample mendelian randomization (MR) can examine the potentially causal impact of circulating proteins on neuropsychiatric phenotypes without these confounds.

Objective: To identify circulating proteins associated with risk for schizophrenia (SCZ), bipolar disorder (BD), and major depressive disorder (MDD) as well as cognitive task performance (CTP).

Design, setting, and participants: In a 2-sample MR design, significant proteomic quantitative trait loci were used as candidate instruments, obtained from 2 large-scale plasma proteomics datasets: the UK Biobank Pharma Proteomics Project (2923 proteins per 34 557 UK individuals) and deCODE Genetics (4719 proteins per 35 559 Icelandic individuals). Data analysis was performed from November 2023 to November 2024.

Exposure: Genetic influence on circulating levels of proteins in plasma.

Main outcomes and measures: Outcome measures were summary statistics drawn from recent large-scale genome-wide association studies for SCZ (67 323 cases and 93 456 controls), BD (40 463 cases and 313 436 controls), MDD (166 773 cases and 507 679 controls), and CTP (215 333 individuals). MR was carried out for each phenotype, and proteins that showed statistically significant (Bonferroni-corrected P < .05) associations from MR analysis were used for pathway, protein-protein interaction, drug target enrichment, and potential druggability analysis for each outcome phenotype separately.

Results: MR analysis revealed 113 Bonferroni-corrected associations (46 novel) involving 91 proteins across the 4 outcome phenotypes. Immune-related proteins, such as interleukins and complement factors, showed pleiotropic effects across multiple outcome phenotypes. Drug target enrichment analysis provided support for repurposing of anti-inflammatory agents for SCZ, amantadine for BD, retinoic acid for MDD, and duloxetine for CTP.

Conclusions and relevance: Identifying potentially causal effects of circulating proteins on neuropsychiatric phenotypes suggests potential biomarkers and offers insights for the development of innovative therapeutic strategies. The study also reveals pleiotropic effects of many proteins across different phenotypes, indicating shared etiology among serious psychiatric conditions and cognition.

Figure 1. MR analysis workflow.

Figure 2. Manhattan plot showing findings from MR analysis for schizophrenia, bipolar disorder, major depressive disorder, and cognitive task performance, employing Cis- pQTLs from UKB-PPP and deCODE dataset as instrumental variables

Note: Manhattan plots: (Top panel) proteins with positive predictive beta-values (Bottom panel) proteins with negative predictive beta-values relative to the traits investigated. X-axis: genomic coordinates/chromosomes; Y-axis: -log10p values of associations. Traits: Bipolar Disorder (turquoise), Cognitive ability (red), MDD (purple), Schizophrenia (orange). MR/Meta-Analysis: Meta-P (circle), deCODE: Results for deCODE specific MR analysis (triangle), UKB- PPP: Results for UKB-PPP specific MR analysis (square). Proteins based on more than one method, on multiple traits may be represented in the figure.

Figure 3. Manhattan plot showing findings from MR analysis for schizophrenia, bipolar disorder, major depressive disorder, and task performance employing Trans-pQTLs from UKB-PPP and deCODE datasets as instrumental variables

Note: Manhattan plots: (Top panel) proteins with positive predictive beta-values (Bottom panel) proteins with negative predictive beta-values relative to the traits investigated. X-axis: genomic coordinates/chromosomes; Y-axis: -log10p values of associations. Traits: Bipolar Disorder (turquoise), Cognitive ability (red), MDD (purple), Schizophrenia (orange). MR/Meta-Analysis: Meta-P (circle), deCODE: Results for deCODE specific MR analysis (triangle), UKB- PPP: Results for UKB-PPP specific MR analysis (square). Proteins based on more than one method, on multiple traits may be represented in the figure.

Figure 4. Cis-pQTL MR results across traits

A. Venn Diagram showing shared significant (P-Bonf<0.05) cis-protein across four phenotypes B. Cis-Proteins significantly associated with at least one trait are presented in the figure. Note: The strength of MR association p-values is represented by a blue gradient across data values. The strength of association is further annotated for easy reference, nominally significant p < 0.05 (circle), FDR significant P_scz < 1.58 x 10^-3 P_BIP < 7.3 x 10^-4, P_MDD < 4.9 x 10^-3, P_CTP < 1 x 10^-3 (square), and Bonferroni significant p < 1.31x 10^-5 (triangle) are annotated accordingly. C. Venn Diagram showing shared significant (P-Bonf<0.05) trans-protein across four phenotypes D. Trans-Proteins significantly associated with at least one trait are presented in the figure. Note: Proteins significantly associated with at least one trait are presented in the figure. The strength of MR association p-values is represented by a blue gradient across data values. The strength of association is further annotated for easy reference, nominally significant p < 0.05 (circle), FDR significant P_scz < 1.8 x 10-4, P_BIP < 3.81 x 10^-5, P_MDD < 1.90 x 10^-5, P_CTP < 2 x 10-4 (square) and Bonferroni significant p < 7.3 x 10^-6(triangle) are annotated accordingly.