13. DIRECT CRISPR-EDITING OF RS4702 DECIPHERS CELL-TYPE SPECIFIC EPISTATIC INTERACTIONS WITH SCHIZOPHRENIA POLYGENIC RISK

Background Schizophrenia (SCZ) has a complex, heterogeneous, and highly polygenic genetic architecture. Polygenic risk scores (PRS) assume additive genetic effects and may not capture epistatic or gene-by-environment interactions. However, PRS analyzed in conjunction with functional genomic approaches using CRISPR editing and stem cell–derived brain models allow us to study these interactions and elucidate epistatic mechanisms. Previously, we functionally validated the top causal SCZ variant at the FURIN locus, rs4702. We also identified interactions beyond additive expectations when perturbing FURIN alongside other SCZ risk genes. However, these studies were limited to 1-2 control lines and did not account for polygenic background. Methods Here, we characterize transcriptomic differences between neurotypical individuals with high or low SCZ PRS (n=12) across hiPSC-derived astrocytes, GABAergic and glutamatergic neurons (iASTROs, iGABAs, iGLUTs), and neural progenitor cells. Using CRSIPR base-editing, we examined how polygenic liability interacts with rs4702 dosage (n=6) by performing transcriptomic and phenotypic assays on isogenic rs4702 AA vs GG lines with varying SCZ PRS (2 high 2 mid, 2 low) following differentiation into iASTROs, iGABAs, and iGLUTs. Results SCZ PRS interacts with cell-type and developmental stage causing significant differences in gene expression, neuronal excitation, and synaptic morphology between low and high lines. In rs4702 edited iGLUTs, FURIN expression was significantly decreased by the SCZ risk allele (GG) in mid (log2FC FURIN =-1***) and high (FURIN=0.72***), but not low, PRS lines (FURIN=-0.04) – directly demonstrating epistatic interactions influencing FURIN expression. The degree of FURIN downregulation correlated with the degree of overall transcriptomic dysregulation (DEG: mid=9188, high=6631, low=621) and demonstrated time-dependent effects. For example, significant upregulation of key glutamatergic genes (GRIA2, SLC17A6, SLC17A7) and increases in neuronal excitability and neurite length (p=2 × 10-6) at 1 wk that were no longer present at 3 wks. However, this ablation of differential effects was not true in high PRS lines (1wk FURIN=-0.72***, DEG=6631 vs 3wk FURIN=-0.53***, DEG=3939), suggesting that high SCZ liability prolongs the period of neuronal dysregulation caused by rs4702. To identify pathways driving these epistatic interactions, we identified PRS-associated WGCNA modules within cell-types, calculated transcriptomic risk scores based on gene membership, and performed phenome-wide association studies in 2 biobanks (Mount Sinai and Vanderbilt University). We found cell-type x PRS co-expression networks that interact with rs4702 and have unique functional enrichments and associations with psychiatric diagnoses. Discussion Overall, we measured transcriptomic and phenotypic epistatic interactions between rs4702 and SCZ PRS with larger implications for inclusion of genetic background in functional genomics and assumptions of additivity in modeling genetic risk. We found that the magnitude of effect of the rs4702 variant is mediated by polygenic background – with donor lines with the highest genetic liability showing sustained dysregulation and alterations in neuronal morphology that are otherwise ablated or absent in mid or low PRS lines at later developmental stages. We further identified cell-type x PRS co-expression networks that highlight biological programs that may underlying specific risk pathways.