Abstract PD9-08: Breast cancer liver metastases vascularize by vessel co-option, not angiogenesis, and have a desert immune phenotype: A histopathological and gene expression study

Abstract

Abstract Phase 3 trials of bevacizumab combined with chemotherapy in metastatic breast cancer have consistently failed to demonstrate a survival benefit for the addition of bevacizumab. When cancers metastasize to highly vascular organs (including the liver), they can utilize vessel co-option, instead of angiogenesis, as a mechanism to obtain a vascular supply (1). We have repeatedly shown by histopathological analyses that almost all (95%, 2 cohorts) breast cancer liver metastases utilize vessel co-option instead of angiogenesis to vascularize (2,3). The prevalence of vessel co-option in breast cancer could explain, at least in part, why anti-angiogenic therapy has been a disappointing therapeutic approach in metastatic breast cancer. Animal models of non-angiogenic liver and lung metastases also displayed resistance to anti-VEGF treatment (3,4). We have now undertaken a gene expression study (mRNA sequencing) of targeted samples at the tumor-liver interface to discover gene expression patterns and signaling pathways that are associated with non-angiogenic growth of metastatic cancer in the liver (n = 70). A network to detect biological themes of non-angiogenic growth was built by gene set enrichment analysis. Key components of this network are: cancer cell motility and invasion, epithelial-to-mesenchymal transition, stemness and proliferation. This contrasts with the network of angiogenic liver metastases of which the most important components are inflammation and ECM remodeling. Semi-automated image analyses of CD8-immunostained section of liver metastases confirms that non-angiogenic liver metastases have a significantly lower density of CD8-positive cytotoxic T-lymphocytes at the tumor-liver interface when compared with angiogenic liver metastases (300 cells/mm2 and 1000 cells/mm2, respectively (p<0.0001)). In addition, a clear CXCL13-driven B-cell gene expression signature is associated with angiogenic growth of liver mets but is absent in non-angiogenic growth of breast cancer liver metastases. Gene expression patterns that may be play a role in vessel co-option are the up-regulation of LAMA3, LAMB3, LAMC2, coding for the 3 subunits of laminin-5, and of ITGA3, ITGB1, ITGA6 and ITGB4, coding for both integrin-receptors of laminin-5. This supports the concept of 'adhesive' vessel co-option during which cancer cells use the basement membrane of the co-opted blood vessels as a soil (5). In addition, the claudin-2 gene (CLDN2) is significantly overexpressed in non-angiogenic liver metastases which is consistent with earlier reports on the role of claudin-2 during breast cancer metastasis to the liver (6). In conclusion, we provide evidence, based on morphology and gene expression, for the almost exclusive non-angiogenic growth of breast cancer liver metastases. In addition, non-angiogenic breast cancer liver metastases are characterized by a desert immune phenotype. Both observations can have an impact on the treatment strategy of patients with metastatic breast cancer.