Abstract 6149: Prognostic tumor microenvironment subtypes in triple-negative breast cancer

Abstract

Abstract Triple-negative breast cancer (TNBC) is an aggressive subtype of breast cancer in which the tumor microenvironment (TME) is prognostic, with increased tumor-infiltrating lymphocytes predicting a good prognosis. Positive clinical trials of immune checkpoint inhibitors in TNBC further highlight the TME’s importance, but its single-cell composition and relationship to patient survival has not been well described. Therefore, we generated highly-multiplex cyclic immunofluorescence (CyCIF) data from 59 TNBC in tissue microarrays. Our CyCIF dataset served as a discovery cohort; the data included 35 protein markers and clinical outcomes for 18 patients. For a validation cohort, we analyzed 32 markers imaged with multiplex ion beam imaging (MIBI) in 38 TNBC patients (Keren et. al., Cell 2018). In the CyCIF cohort, we used markers’ single-cell mean intensity to generate a Umap manifold, which we clustered with the unsupervised Leiden algorithm. We annotated the resulting 28 cell types as epithelial, immune or stromal. In the MIBI cohort, we similarly clustered and annotated 21 cell types as epithelial, immune or stromal. Next, we clustered the patients based on the fraction of epithelial, immune or stromal cells in the tissue. In the CyCIF cohort, patients fell into three subtypes: epithelial-rich, stroma-rich and immune-rich. The patients in the epithelial-rich group had significantly shorter survival than the other two subtypes (log-rank p-value = 0.0008) and had shorter recurrence-free survival (log-rank p-value = 0.048). In the MIBI cohort, patients fell into the same three subtypes. As in the CyCIF cohort, the epithelial-rich patients had shorter survival than the other subtypes (log-rank p-value = 0.021), although there was no difference in recurrence-free survival (log-rank p-value = 0.47). Finally, we identified protein biomarkers with significantly different expression levels between the subtypes in one or both cohorts. The poor-prognosis, epithelial-rich subtype had higher CD31 expression in both cohorts, and in the CyCIF cohort, Glut1 was elevated, indicating a hypoxic, angiogenic TME. The immune infiltrate was primarily CD68+ macrophages and the stroma was activated, with higher levels of Ki67 in both cohorts. In contrast, the stroma-rich, good prognosis subtype had low immune infiltration, including fewer macrophages, low CD31 expression and a quiescent, non-proliferating stroma. Lastly, the good-prognosis, immune-rich subtype had high T and B cells levels, as well as markers of immune regulation, memory cells, and antigen presentation cells. In conclusion, we generated and analyzed a new multiplex CyCIF dataset with clinical annotation and confirmed our analytical findings in a second multiplex imaging dataset. We delineated three TME subtypes in TNBC with different outcomes, cell types and biomarker expression, potentially leading to better patient stratification and informing new drug targets. Citation Format: Jennifer R. Eng, Elmar Bucher, Zhi Hu, Melinda Sanders, Bapsi Chakravarthy, Summer Gibbs, Koei Chin, Jennifer Pietenpol, Joe W. Gray. Prognostic tumor microenvironment subtypes in triple-negative breast cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 6149.