P5-14-07: Novel Interactions between Immunity and Proliferation in Breast Cancer Prognosis.

Abstract

Abstract Highly proliferative breast tumors are phenotypically and clinically heterogeneous and exhibit a significant metastatic tendency. However, not all highly proliferative breast cancers will progress to distant metastasis, suggesting that other factors may play a role in limiting their clinical progression. With an interest in identifying genes that predict metastatic capacity of breast cancers in a proliferation-dependent context, we recently compiled a database of ∼2,000 tumor expression profiles using microarray data from multiple large breast cancer cohorts. We randomized the dataset into training and testing sets to allow discovery and validation of novel gene-survival (ie, distant metastasis-free survival; DMFS) associations. We employed Cox regression analysis to identify genes with statistically significant associations with DMFS. Then, we hierarchically clustered the expression patterns of the significant genes to identify gene clusters lacking correlation with proliferation genes. A large cluster of genes with known functions in innate and adaptive immunity emerged. Cross referencing these genes to a published microarray dataset of pan-leukocyte expression profiles revealed cell type-specific structure within the immunity cluster, likely reflective of distinct populations of tumor-infiltrating immune cells. Specifically, 3 immune gene subclusters identified within the tumors showed predominant expression in B-cells, T-cells and natural killer (T/NK) cells, or monoctytes and dendritic (M/D) cells. To evaluate the prognostic relevance of these immune gene clusters, we averaged the gene expression measurements within each cluster to generate “metagenes” that could be used to divide breast cancer cases into population tertiles based on the low, intermediate and high expression of the immune genes. The same was done for a large cluster of proliferation genes known to reflect the proliferative capacity of breast cancer cells. Strikingly, we found that that prognostic power of the B-cell, T/NK and M/D metagenes was not consistent among the proliferation tertiles, but rather, was exclusively statistically significant in the highest proliferation tertile, enriched for basal-like, luminal B and HER2+-like tumors. In these tumors, all 3 immune metagenes stratified cases into low, intermediate and high risk of recurrence with high statistical significance (p<0.0001), wherein the highest expression tertile of the immune metagenes corresponded to reduced risk of metastatic recurrence (at 5 and 10 years), consistent with a long-term anti-metastatic immune response. Comparative analysis of the B and T/NK metagenes revealed a highly significant prognostic interaction between the two, whereby the low expression tertile of either the B-cell or T/NK metagene portends a poor prognosis that is not overcome by a high expression tertile of the other. Our results suggest that cell type-specific immune signatures in breast cancer can predict survival in highly proliferative subtypes of breast cancer, with potential to spare some patients from aggressive adjuvant treatment while identifying others who will require more than standard of care. Furthermore, our results provide novel insight into the important roles of T and B cell-mediated immunity in breast cancer progression. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P5-14-07.