T32. A LARGE-SCALE PLASMA PROTEOMICS STUDY TO PREDICT TREATMENT RESISTANCE IN MAJOR PSYCHIATRIC DISORDERS

Abstract

Background Schizophrenia (SCZ), bipolar disorder (BD) and major depressive disorder (MDD) are leading causes of disability, still only few biomarkers are available to optimise treatment prescription. As a matter of fact, about a third of patients suffering from these diseases do not respond to multiple pharmacological therapies and develop treatment resistance (TR), resulting in an increased risk of chronic course, quality of life deterioration, suicide, and poor physical health. TR may have shared biological mechanisms across SCZ, BD, and MDD, in line with the observation that these diseases share pathogenetic mechanisms. Based on this hypothesis, the Psych-STRATA project aims to identify biomarkers of TR across major psychiatric disorders, to facilitate the early identification of patients at risk of TR. A key component of the project is proteomic profiling of patients with SCZ, BD, or MDD, characterised for their treatment response profile, alongside with genome-wide analyses. Methods We performed a large-scale exploratory study to identify protein biomarkers of TR in each disorder of interest and across disorders. The new Olink® Explore HT panel including ∼5,400 proteins was used for the analysis of plasma samples from a total of 860 patients (310 with BD, 247 with SCZ, 303 with MDD), with information on TR/response status, defined based on number of medications prescribed, clozapine prescription in case of SCZ, and standard symptom scales. Regression models were run to identify associations between protein levels and TR in each disorder, adjusting for age, sex, centre, delivery box, median protein levels. Treatment class was also included in the models. As we were interested in finding shared associations across disorders, we focused on signals shared in two or more analyses (FDR < 0.1 and |log 2(fold change)| > 0.5). To further characterise TR using a data-driven approach, we applied consensus clustering to identify patients’ subgroups and Weighted Gene Correlation Network Analysis (WGCNA) to discover protein modules with distinct abundance patterns associated with these. Results A total of 854 samples (∼54% TR) with measurements for 5348 proteins were included after quality control. A total of 164 proteins were increased or decreased in at least two comparisons, with the largest proportion of concordant protein changes between MDD and BD (n=36). Examples of proteins associated with TR across disorders independent of the direction of change were CACYBP, ITGA6, APEX1, EDA2R, and FCHO1. A total of 44 proteins were identified as module hubs in WGCNA, with IRF3 shared across the three disorders. Discussion Proteins with trans-diagnostic associations with TR were mostly involved in immune system processes, intracellular signalling, and cell adhesion. Previous findings from genome-wide association studies include CACYBP, associated with hospitalization rate in serious mental illnesses, and FCHO1, associated with neuroticism (GWAS Catalog). The next steps of this work include the identification of protein quantitative trait loci (pQTL), comparison with protein levels prediction from genetic data, and a large validation study of the top 90 proteins identified in over 5000 samples using bead-based affinity proteomics.