Plasma proteomics predicts pathological complete response and reveals LIF as a potential mediator of resistance to neoadjuvant immunochemotherapy in resectable NSCLC

作者：Feng S, Feng Z, Xie S, Zhou Y, Sun W, Lin N, Lin Z, Wang Q, Zhao Z, Lin Y, Long H · 发表于：Journal for immunotherapy of cancer · 年份：2026 · DOI：10.1136/jitc-2026-015312 · 被引用次数：31 · 研究领域：Neoadjuvant Therapy、Lung Neoplasms、Proteomics、Carcinoma, Non-Small-Cell Lung、Leukemia Inhibitory Factor、Immunotherapy、Biomarkers, Tumor、Humans、Pathologic Complete Response、Animals、Female、Drug Resistance, Neoplasm

BACKGROUND: Neoadjuvant immunochemotherapy (nICT) is increasingly used for resectable non-small cell lung cancer (NSCLC), yet a substantial proportion of patients fail to achieve pathological complete response (pCR). Clinically scalable, minimally invasive biomarkers that predict response and point to actionable resistance mechanisms remain needed. METHODS: Pretreatment plasma samples collected prior to the first dose were profiled using the Olink proximity extension assay in a Sun Yat-sen University Cancer Center cohort (n=86), randomly split into a training set (n=65) and an internal validation set (n=21). Differentially abundant proteins associated with pCR were identified, and predictive models were developed using logistic regression, random forest, and extreme gradient boosting (XGBoost), followed by internal validation and independent external validation in the Jiangmen Central Hospital cohort (n=46). Survival associations were evaluated by Cox regression. Mechanistic analyses integrated tumor immunohistochemistry/multiplex immunofluorescence, bulk RNA sequencing with immune deconvolution, and functional validation in subcutaneous and orthotopic murine lung cancer models with pharmacologic leukemia inhibitory factor (LIF) blockade. RESULTS: Four candidate proteins (LIF, CXCL1, CX3CL1, and NT-3) showed modest single-marker discrimination for pCR (area under the curve (AUC), 0.611-0.673). Multiprotein models improved prediction, with XGBoost achieving the highest perfo...