Massively parallel characterization of context specific regulatory risk elements across psychiatric disorders in human-induced pluripotent stem cell-derived glutamatergic neurons

There is an urgent need to decipher the complex polygenic risk architecture of neuropsychiatric and neurodevelopmental disorders. Most disorder-associated common variants are non-coding, and their functional impact remains unknown. Fine-mapping methods predict casual variants in putative regulatory elements that may modulate transcriptional activity of key genes contributing psychiatric risk. Recently, these methods have expanded to predict tissue, cell-type, and context-specific activity of risk variants. While fine-mapping approaches have led to large-scale identification of regulatory sequences and highlighted their importance as context-specific drivers of genetic risk, validation of the causal SNPs and their downstream functional impact remains an active area of exploration. By pairing contemporary advances in high-throughput sequencing and large-scale screening techniques, such as Massively Parallel Reporter Assays (MPRAs) and single-cell CRISPR screens, with hiPSC-models of neurodevelopment we can validate our prediction models at a meaningful scale in a context and cell-type specific manner. Here we quantify of the impact of ∼7,450 neuropsychiatric candidate regulatory sequences (CRS) on transcriptional activity using a lenti-MPRA approach in mature hiPSC-derived glutamatergic neurons (iGLUTs). CRS were predicted using the Psychiatric Genomics Consortium's GWAS summary statistics for 8 neuropsychiatric traits (AD, ADHD, AN, ASD, BIP, MDD, PTSD, SCZ) and Dorsolateral Pre-frontal Cortex (DLPFC) derived expression quantitative trait loci (eQTLs) from the Common Mind Consortium (CMC) using two approaches: (1) Bayesian co-localization (using coloc2) and (2) transcriptomic imputation (S-PrediXcan). At baseline, we found that ∼10% of our CRS were transcriptionally active in iGLUTs (Benjamin-Hochberg corrected pval To identify risk variants within brain-related CRS that confer greater susceptibility, or resilience, to environmental stressors, we performed context-dependent lenti-MPRAs. Inflammation and stress during pregnancy results in immune activation, inducing pro-inflammatory agents, neuropoietic cytokines, and glucocorticoids, such as Interleukin-6 (IL-6), Interferons (IFN) and cortisol (hCort). This immune activation is associated with negative mental health outcomes in the offspring. To determine the impact of inflammation and stress on transcriptional patterns during neurodevelopment we performed lenti-MPRAs in iGLUTs treated with IL-6 (60ng/mL), INFα2β (500ng/ml), and hCort (1000nM). Exposure to different physiological cues altered the level of transcriptional activity – of note, INFα2β treatment resulted in an increase in transcriptional activity (∼ 25% active CRS vs. ∼10%) matching a bias towards up-regulation in gene expression resolved by paired RNA sequencing. Additionally, hCort exposure revealed stress-dynamic regulatory regions. For example, of the tested CRS at the ELOVL1 locus, some showed (i) no significant allelic shifts, (ii) significant allelic shifts without pronounced context-specificity, (iii) and significant allelic shifts with hCort-specificity, identifying context-specific eQTLs. Overall, we will show how the results from lenti-MPRA studies in hiPSC-derived models can shed insight into the accuracy of fine-mapping approaches and can reveal context-specific interactions with genetic risk at a cell-type specific resolution.