Molecular classification of triple-negative tumors

Abstract

Triple-negative breast cancers (TNBCs) are among the most clinically challenging because of their poor prognosis and paucity of treatment options. In part through our genomic profiling studies, breast cancer is now appreciated as being composed of multiple diseases. One of these diseases, the basal-like breast cancer (BLBC) subtype, is now known to represent a unique disease entity with a distinct etiology and biology. Over the years, BLBC has become more commonly known as TNBC because the majority of these tumors lack expression of ER, PR and HER2; however, not all TNBC are BLBC, and not all BLBC are TNBC. Recently, we discovered that a significant subset of TNBC is comprised of a new subtype, the claudin-low, which is important because it is biologically distinct from BLBC and has a number of features reminiscent of mammary stem cells [1]. In addition, luminal A, luminal B, and HER2-enriched tumors are also identified within TNBCs in various small proportions, which highlights the complexity of the clinically based classification. We have explored the treatment sensitivity of the various intrinsic subtypes to neoadjuvant anthracycline/taxane-based chemotherapy using a large publicly available dataset [2]. Across all patients, and within TNBC, basal-like tumors were found associated with a higher likelihood of achieving a complete pathological response (pCR) than the rest of the subtypes, including the claudin-low. In multivariate logistic regression models for pCR prediction, we observed that the intrinsic molecular subtypes virtually always make the final model, even if clinical variables and other genomic predictors are included. In addition, our analyses show that those tumors that achieve a pCR showed a better survival outcome than those that did not, regardless of their molecular subtype; this effect is much larger within the basal-like subtype, which is concordant with previous findings. This intriguing association between residual disease after therapy and poor outcome in basal-like and claudin-low tumors points to intratumor cell heterogeneity as a possible explanation, where resistant and aggressive cell clones might already exist in the pretreated tumor. Our preliminary analyses using a combination of fluorescent activated cell sorting and global gene expression on numerous preclinical models of basal-like breast cancers including cell lines and primary tumor xenografts suggest the existence of at least two cell populations in many BLBC models. These different cell populations are currently being tested for tumor-initiating cell activities, and additional studies focusing on these populations changing with treatment are also being performed.