Decoding Cellular Stress States for Toxicology Using Single-Cell Transcriptomics

Abstract

We applied the TempO-LINC(R) platform to generate single-cell transcriptomic (SCTr) profiles of [~]40,000 HepaRG cells exposed to etoposide, brefeldin A, cycloheximide, rotenone, tBHQ, troglitazone, and tunicamycin at three concentrations for 24 hours. SCTr enabled a detailed analysis of adaptive stress response pathways (SRPs), including the unfolded protein response (UPR), oxidative stress response (OSR), heat shock response (HSR), and DNA damage response (DDR). Troglitazone upregulated lipid metabolism genes (PLIN2, ACOX1) along with HSR and UPR activation, with co-expression of DNAJA1, HSP90AA1, and DDIT3 in subsets of cells. Brefeldin A and tunicamycin strongly induced UPR markers (HSPA5, SYVN1, LMF2, PDIA4) in subsets of cells, with some also expressing apoptotic (DDIT3, CASP8) and autophagic (SQSTM1) genes, indicating diverse stress responses. Rotenone activated GDF15, TRIB3, and DDIT3 in a fraction of cells, accompanied by PLIN2 and mild UPR induction, reflecting heterogeneous mitochondrial stress responses. We scored individual cells using literature-derived SRP gene signatures to characterize overall stress phenotypes and clustered them using a generalized Jaccard metric. The clustering revealed five phenotypic groups spanning cell states associated with homeostasis, adaptive responses, terminal outcomes, autophagy, and apoptosis. By systematically analyzing the distributions of cells in different states across treatments, we visualized dynamic shifts in cellular subpopulations responding to chemicals, revealing early stress responses and potential transitions to cell death. Our findings suggest the utility of SCTr in decoding stress states that could provide possible insights into transitions between cellular adaptive and terminal transitions involved in toxicity.