Background: The prognostic value of tumor infiltrating lymphocytes (TILs) has been extensively reported in several primary cancers. The broad prognostic utility of TIL profiling in primary solid tumors was further demonstrated by a number of retrospective large-scale meta-analyses. These studies were enabled by in silico methods to assess leukocyte infiltration and T-cell clonality from transcriptome data. However, in contrast to primary cancers, the landscape of immune infiltration within metastases is largely unknown. This represents a significant gap in knowledge as immune evasion is widely recognized as a hallmark of metastatic cancer. To identify genetic and functional mechanisms that enable immune escape of individual tumors and to better understand tumor-host immune interactions in metastasis we characterize immune infiltration across 500 metastatic cancer patients.

Methods: We obtained integrative molecular data, comprising whole-exome and transcriptome sequencing, for metastatic cancer patients enrolled under the MI-Oncoseq program. We utilized existing and novel computational methods to quantify the level of immune infiltration (TIL-score) and the predicted composition of the infiltrate within each sample. Further, to identify facets of tumor-host immune interactions we characterized the joint expression profiles of immune receptors, ligands, cytokines, and HLA molecules.

Results: Application of TIL-scores revealed a large heterogeneity of lymphocyte infiltration levels, both within and across cancer types. Consistent with data from primary tumors, the highest levels of immune infiltration were observed in melanoma and kidney cancer, but highly infiltrated samples were observed for most cancers. However, multiple lines of evidence suggest that over 50% of tumors lack significant T-cell infiltration and that HLA expression is almost completely lost in lymph node metastases for several cancer types. Based on the breakdown of immune cells, cancers were most strongly typified by the ratios of M2 to M0 macrophages and CD8+ to CD4+ T-cells. Unsupervised analysis revealed coherent expression of immune ligands and their cognate receptors (CD80/CD86 and CD28), but anticorrelation between stimulatory (CD80/CD86) and suppressive (PD-L1/PD-L2) ligands.

Conclusions: To develop the most effective immunotherapeutic strategies for metastatic cancer, it is essential to dissect the specifics of immune evasion in each tumor, determine the regulatory network that links these mechanisms, and identify biomarkers that predict sensitivity to specific immune therapies. This study highlights the heterogeneity of metastatic immune infiltration and represents the first step in the development of immunophenotypic biomarkers in metastatic cancer.

Citation Format: Marcin P. Cieslik, Yi-Mi Wu, Lisha Wang, Xuhong Cao, Dan Robinson, Arul M. Chinnaiyan. The landscape of immune infiltration and mechanisms of immune escape across 500 metastatic cancers. [abstract]. In: Proceedings of the AACR Special Conference on Tumor Immunology and Immunotherapy; 2016 Oct 20-23; Boston, MA. Philadelphia (PA): AACR; Cancer Immunol Res 2017;5(3 Suppl):Abstract nr B15.