Single-cell splicing analysis with ISSAC links cell type specific and cell state-dependent sQTLs to neurological disorders

Single-cell RNA sequencing enables comprehensive profiling of gene expression and splicing at cellular resolution, revealing cell type-specific and cell state-dependent regulation (variation within cell types based on their functional states). While genetic studies of expression (eQTLs) in single cells are well established, the genetic regulation of alternative splicing in single cells remains challenging. Existing single-cell splicing QTL (sQTL) studies perform pseudobulk aggregation using bulk analysis methods, which reduces power to detect cell type-specific sQTLs and cannot capture cell state-dependent splicing regulation. Here, we introduce ISSAC to directly quantify metacell-level splice site usage and map cell type- and cell state-specific sQTLs through generalized linear mixed models. In real-world benchmarking on peripheral blood single-cell data, ISSAC identified 1.4- to 2.5-fold more cell type-specific sQTLs than pseudobulk sQTL analysis, and uniquely enabled cell state-dependent sQTL discovery. We applied ISSAC to a harmonized aging brain resource consisting of approximately 3 million dorsolateral prefrontal cortex (DLPFC) single nuclei from 722 donors. ISSAC identified 31,318 independent cis-sQTLs across seven major cell types and 16,861 independent cis-sQTLs across 67 subcell types, with [~]67% of sGenes showing no overlap with eGenes. We identified 369 independent sQTLs whose genetic effects were mediated by various cell states such as dendrite development and synaptic signaling. Additionally, we uncovered 194 Alzheimers-biased and 207 sex-biased sGenes, as well as 142 risk genes that colocalized with neurological disorders including Alzheimers disease, Neuroticism, Amyotrophic lateral sclerosis, Parkinsons disease, Lewy body dementia and Schizophrenia. Specifically, we functionally validated a causal variant rs11549690 modulating TRPT1 exon 7 skipping to influence neuroticism risk.