Ning Zhang, Tong Wan, Siyue Zhang, Yunzhen Jiang, Peng Su, Jixin Liu, Zhitong Chen, Jie Hao, Haoyu Wang, Bing Chen, Wenjing Zhao, Lijuan Wang, Tao Xing, Qihuang Zhang, Ulf Schmitz, Zhiyong Ding, Nicola Crosetto, Bingqiang Liu, Qifeng Yang
Ductal carcinoma in situ (DCIS) is a non-invasive precursor of invasive ductal carcinoma (IDC), yet the biological mechanisms underlying the transition from DCIS to IDC remain incompletely understood. Here, we integrate spatial transcriptomics, single-cell RNA sequencing, and single-cell DNA sequencing on coexisting DCIS and IDC samples to characterize cellular and microenvironmental alterations. Integrated analyses reveal differential molecular characterizations between coexisting DCIS and IDC and identify candidate genes (MGP, PLAT, and SERPINA3) potentially limiting the progression from DCIS to IDC. Malignant epithelial meta-programs (MPs) delineate distinct transcriptional states, with development-associated MP1 enriched in DCIS and cell cycle-related MP5 enriched in IDC. Further analysis reveals varied microenvironmental features in DCIS and IDC, with invasion-associated Mph_SPP1 and development-associated iCAFs_HOPX enriched in DCIS, while immunosuppressive Mph_PRDM1 and metabolism-related tCAFs_BNIP3 predominate in IDC. We then construct a machine learning model to identify DCIS at high risk of progression, which is externally validated across independent bulk RNA-sequencing cohorts (mean AUC = 0.903). Collectively, this study provides an integrative framework for understanding the molecular and spatial features that distinguish DCIS from IDC, with the potential to inform more precise risk stratification and clinical management for DCIS progression.