Background Ovarian cancer is the most lethal gynecologic malignancy worldwide, yet factors leading to therapy resistance are poorly understood, partly hindered by the complexity and heterogeneity of the tumor microenvironment (TME). This study aims to construct a single-cell and spatial atlas of epithelial ovarian cancer, to comprehensively characterize the immune cell states and ecotypes across the disease spectrum, thereby providing insights for more effective individualized therapies.

Methods 16 publicly available and in-house generated single-cell RNA sequencing datasets were collected, which contained 394 samples from 169 individuals. Following data integration, a total of 89 distinct TME cell states were identified, spanning both immune and stromal cell compartments. Subsequently, six unique immune ecotypes were defined using non-negative matrix factorization. Further validations of the novel discoveries were performed using spatial transcriptomic data.

Results A total of 89 different TME cell types and states were identified, including 33 T/NK, 15 B/plasma, 18 myeloid, and 23 stromal cell types and states. These TME cells showed distinct distribution patterns among samples of different pathological types, disease stages, metastatic sites, and treatment statuses. Based on 306 samples with an intact CD45⁺ compartment, we defined six immune ecotypes. Ecotype 1 was mainly comprised of naïve and naïve-like lymphocytes, enriched in malignant ascites and peripheral blood. Ecotype 2 was formed by immune cells highly expressing interferon-stimulating genes, along with 4-1BB⁺ regulatory T cells, exhausted CD8 T cells, and multiple plasma cells, indicating a tumor-reactive microenvironment. Ecotype 3 was highly enriched in metastatic tumors, including multiple types of memory lymphocytes. Ecotype 4 was comprised of dendritic cells, effective CD8 T cells, NK cells, as well as M1-like macrophages, representing a unique ecotype in post-treatment samples, indicating immune activation in response to anti-cancer therapies. Ecotype 5 was highly abundant in normal tissues, cancer-related benign conditions (e.g. ovarian endometrioma), and tumor-draining lymph nodes, including multiple tissue-resident memory lymphocytes, mast cells, as well as various cell types in stressed states. Ecotype 6 was dominated by different tumor-associated macrophages and neutrophils, indicating an immune-suppressive microenvironment. These findings were also recapitulated in additional ovarian cancer samples in situ using spatial transcriptomic data.

Conclusions In summary, this study provided a comprehensive single-cell and spatial immune landscape of epithelial ovarian cancer and highlighted the spatiotemporal heterogeneity of the immune microenvironment across tumor sites, patients, and treatment phases. These findings underscore the importance of personalized immune-modulating therapies in the systemic treatment of ovarian cancer.

Acknowledgements We appreciate all the patients who generously provided their samples for this research. We also acknowledge the support of the High Performance Computing for Research Facility at the University of Texas MD Anderson Cancer Center for providing computational resources that have contributed to the bioinformatics analysis conducted.

Ethics Approval This study inlcuded publicly available datasets, and in-house generated datasets. For the in-house generated datasets, the study was approved by the Institutional Review Board of The University of Texas MD Anderson Cancer Center (Approval number: 2017-0264). All the participants gave informed consent before taking part in this study.