HIGH-THROUGHPUT ANTIBODY-BASED PROFILING OF SERUM IN SCHIZOPHRENIA AND BIPOLAR DISORDER PATIENTS: AN INTEGRATIVE GENOMICS-PROTEOMICS PILOT STUDY

Background The identification of biological markers in the peripheral blood for the prediction of the individual clinical outcome has an enormous potential in the frame of personalized medicine. Previous studies have already tried to create predictive models for major psychiatric disorders based on serum protein profiles though with inconclusive results. This study leverages the rich repertoire of biosamples available in the KFO241/PsyCourse cohort to perform a high-throughput antibody-based serum protein profiling. The aim of this pilot study is to establish the feasibility of such an analysis in the serum of these patients. This study also capitalizes on the availability of genomic data to carry out an integrative (genomics/proteomics) profiling. Methods Serum samples of 113 Schizophrenia (SCZ) and 125 Bipolar Disorder (BD) patients belonging to the German KFO241/PsyCourse cohort (www.kfo241.de; www.PsyCourse.de) were included in this study. All biomaterials in this cohort (blood, serum, plasma, DNA, RNA), available at different time points given its longitudinal design, are part of a high-end biobank. The expression of a selected panel of ~100 serum proteins was determined using a set of 155 antibodies in a high-throughput antibody-based assay. This suspension bead array technology enables a multiplexed protein profiling and it leverages the information generated by the Human Protein Atlas (www.proteinatlas.org). Median fluorescent intensities were log-transformed / standardized for downstream analyses. DNA samples were genotyped (Infinium PsychArray) and underwent genotype imputation (1000 Genomes Phase 3 ref. panel). Polygenic Risk Scores (PRS) were calculated by summing up the weighted effect of each SNP contributing to the PRS to obtain an individual estimate of SCZ genetic risk burden. Age, sex and ancestry principal components were used as covariates. Results Multivariate profiles based on the whole panel of quantified proteins did not significantly discriminate between both diagnoses (SCZ and BD). Single-protein ANCOVA analyses revealed, after Bonferroni correction, significant differences in C4B, CFB, VWF, NRG1, BACE1 and other proteins between SCZ and BD samples. High-risk versus Low-risk patients (according to PRS thresholds 5E-8, 5E-2 and 1) did not show a differential level of expression of these proteins. Likewise, polygenic risk did not correlate with the serum level of any of the proteins. Suggestive serum pQTL loci were identified. Discussion Our results confirm the feasibility and potential of this approach using serum samples of psychiatric patients. The observed differences in protein levels between diagnoses warrant replication in independent samples controlling the effect of medication. These results lay the groundwork for large-scale analyses for the discovery of predictive biomarkers of course, outcome and treatment response in SCZ and BD. Such an approach, joining the longitudinal phenotyping and biosample availability at four different time points in the KFO241/PsyCourse cohort with state-of-the-art biomarker quantification methodologies, holds promise for the generation of clinically-relevant predictive models in complex psychiatric disorders.