化疗前后肿瘤的单细胞转录组比较揭示化疗耐受性骨肉瘤细胞的治疗弱点

Relapse remains the leading cause of treatment failure and mortality in osteosarcoma (OS). There is currently no reliable clinical approach to predict relapse at diagnosis or to inform early, personalised treatment strategies aimed at avoiding relapse. This study aims to address this gap by profiling OS tumors with single-cell RNA sequencing (scRNA-seq) at diagnosis (Dx), after chemotherapy (Rx), and at relapse (Rel). By mapping gene-expression changes across these stages, we identify resistance-associated genes and pathways, define persister cell populations, and prioritise translational combination therapies for high-risk patients. We performed scRNA-seq on 17 primary OS tumor samples (11 Dx, three Rx, and three Rel) from 12 patients including two Dx/Rx and three Dx/Rel pairs. After removing low-quality cells, batch effects were corrected in Seurat using Harmony. Cell types were annotated based on canonical markers, and malignant cells were inferred from copy-number variation (CNV) profiles using CONICSmat. To further increase our sample size and statistical power, we integrated an additional 12 diagnostic and 15 after chemotherapy OS scRNA-seq data from 27 patients using publicly available datasets, applying Harmony to correct for sequencing platform differences and minimise inter-dataset batch effects. Differential expression in malignant cells was assessed between Dx and Rx/Rel using pooled and paired analyses. Genes upregulated at Rx/Rel were prioritised using four filters: patient prognostic association in NCI TARGET, supporting evidence from published studies, gene dependency in DepMap, and small molecule inhibitors suitable for in vivo studies. Five candidate genes were selected for functional validation: SNHG6, PFKFB3, MRE11, PLCbeta4, and S100A13. PFKFB3 and PLCbeta4 were tumor-specific and induced by cisplatin and doxorubicin. Their silencing by siRNA or pharmacologic inhibition with the selective inhibitor sensitised OS cells to chemotherapy. Gene set enrichment identified epithelial-mesenchymal transition (EMT) as the top pathway in Rx/Rel, consistent with therapy-tolerant states. SB-431542, a TGF-beta-induced EMT inhibitor, sensitised cisplatin-resistant OS cells to cisplatin and reduced migration of OS cells. Taken together, these findings define chemoresistant tumor states and identify actionable targets and pathways, including PFKFB3 and EMT, for rational combination therapies. They support a scRNA-guided framework for risk stratification and precision treatment selection in high-risk OS.