Background Checkpoint inhibitors can induce robust and durable responses in a subset of patients. Extending this benefit to more patients could be facilitated by better understanding of how interacts with immune cells with the tumor microenvironment, which is a critical barrier to control both local and systemic disease. The composition and pattern of the immune infiltrate associates with the likelihood of response to immunotherapy. Inflamed tumors that exhibit a brisk immune cell infiltrate are responsive, while those in which immune cells are completely or partially excluded are not. Transforming growth factor β (TGFβ) is immunosuppressive and associated with the immune excluded phenotype.

Methods Using an immune competent mammary tumor derived transplant (mTDT) model recently developed in our lab, exhibits inflamed, excluded or deserts immune infiltrate phenotypes based on localization of CD8 lymphocytes. Using whole transcriptome deep sequencing, cytof, and PET-CT imaging, we evaluated the tumor, microenvironment, and immune pathway activation among immune infiltrate phenotypes.

Results Three distinct inflamed tumors phenotypes were identified: ‘classically’ inflamed characterized by pathway evidence of increased CD8+ T cells and decreased PD-L1 expression, inflamed tumors with pathways indicative of neovascularization and STAT3 signaling and reduced T cell mobilization, and an inflamed tumor with increased immunosuppressive myeloid phenotypes. Excluded tumors were characterized by TGFβ gene expression and pro-inflammatory cytokine signaling (e.g. TNFα, IL1β), associated with decreased leukocytes homing and increased immune cell death of cells. We visualized and quantified TGFβ activity using PET-CT imaging of 89Zr-fresolimumab, a TGFβ neutralizing antibody. TGFβ activity was significantly increased in excluded tumors compared to inflamed or desert tumors, which was supported by quantitative pathology (Perkin Elmer) of its canonical signaling target, phosphorylated SMAD2 (pSMAD2). pSMAD2 was positively correlated with PD-L1 expression in the stroma of excluded tumors. In contrast, in inflamed tumors, TGFβ activity positively correlated with increased F4/80 positive macrophages and negatively correlated with expression of PD-L1. CyTOF analysis of tumor and spleen immune phenotypes revealed increased trafficking of myeloid cells in mice bearing inflamed tumors compared to excluded and deserts.

Conclusions The immunocompetent mTDT provides a model that bridges the gap between the immune landscape and tumor microenvironment. Integration of these high-dimensional data with further studies of response to immunotherapies will help to identify tumor features that favor response to treatment or the means to convert those that are unresponsive.