Shiyou Wei, Yulan Deng, Jinho Lee, Hongbin Lan, Zhenyu Yang, Marilyne Labrie, Courtney B. Betts, Hao Duan, Benjamin Tate, Joanna Pucilowska, Dennie Frederick, Aleigha R. Lawless, Tatyana Sharova, Yuanzhong Yang, Wanming Hu, Georgia M. Beasley, Lynn M. Schuchter, Xiuqi Wang, Wei Xu, Gen Yong, Megan E.G. Vandenberg, Drew A. Torigian, Shamilene Sivagnanam, Kuang Du, Eric Sugarman, Suzanne McGettigan, Cathy Zheng, Rami N. Al-Rohil, Maria A. Selim, Michael B. Datto, Giselle Y. López, Smita K. Nair, David M. Ashley, Xiaowei Xu, Ravi K. Amaravadi, Giorgos C. Karakousis, Donald M. O'Rourke, Steven Brem, Bert W. O'Malley, Gokhan Demirkan, Shuangxing Yu, Yiling Lu, Todd Camp, Janice A. Patterson, Zhi Wei, Christopher Corless, Dmitry I. Gabrilovich, Yonggao Mou, Keith T. Flaherty, Lisa M. Coussens, Genevieve Boland, Meenhard Herlyn, Gordon Mills, Lunxu Liu, Gao Zhang
Despite advances in immune checkpoint blockade, resistance in metastatic melanoma remains a major challenge. To decode resistance mechanisms, we generate a comprehensive longitudinal, multi-omic, and spatial atlas of 45 tumor samples across 10 patients. Analysis reveals resistant tumors undergo convergent evolution toward a shared, spatially organized immunosuppressive ecosystem. We identify a structural mechanism characterized by spatial partitioning of immune checkpoints, where B7-H3 dominates MITF-high niches while IDO1 characterizes MITF-low zones. Furthermore, integrated single-cell and spatial analysis identifies a specific malignant subclone (c1) and a distinct architectural niche (RCN3), both exhibiting aberrant PI3K-mTOR signaling. Notably, c1 promotes the "ignored tumor" phenotype via FN1-ITGB1 and GDF15 signaling. Validated across independent cohorts, these spatial and molecular signatures predict poor survival and point to actionable targets. Ultimately, our study elucidates the spatial logic of resistance and provides a rationale for translating multi-omic discoveries into actionable, personalized therapeutic strategies.