CELL INTELLIGENCE · 肿瘤细胞治疗研究

BACKGROUND: In human epidermal growth factor receptor 2 (HER2)-positive early breast cancer (EBC), we investigated tumor and immune changes during neoadjuvant treatment and their impact on residual disease (RD) biology and prognostic implications across four neoadjuvant studies of trastuzumab with or without lapatinib, and with or without chemotherapy: CALGB 40601, PAMELA, NeoALTTO, and NSABP B-41. PATIENTS AND METHODS: We compared tumor and immune gene expression changes during neoadjuvant treatment and their association with event-free survival (EFS) by uni- and multivariable Cox regression models in different cohorts and timepoints: 452 RD samples at baseline including 169 with a paired RD, and biomarker changes during neoadjuvant therapy, evaluating model performance via the c-index. RESULTS: Analysis of 169 paired tumor samples revealed a shift in intrinsic subtype proportions from HER2-enriched at baseline (50.3%) to normal-like (49.1%) followed by luminal A (18.9%) in RD. This luminal phenotypic change was supported by decreased correlation to the HER2-enriched centroid, ERBB2, and HER2 amplicon genes and increased correlation to the luminal A centroid (Wilcoxon test P < 0.001). Additionally, RD showed relative immune activation marked by significant increases in B-cell, CD8 T-cell, and natural killer cell signatures (Wilcoxon test P < 0.05). In multivariable Cox models, intrinsic subtypes at baseline provided more prognostic information, while immune gene expression signatures provided more prognostic information in RD. Notably, the best multivariable EFS model (c-index = 0.77) integrated the immunoglobulin G signature from RD samples (adjusted hazard ratio 0.45, 95% confidence interval 0.30-0.67, adjusted P = 0.002). CONCLUSIONS: In patients with HER2-positive EBC and RD, tumor biomarkers provide more prognostic information at baseline. In contrast, immune biomarkers perform better for EFS prognosis in RD.