The Seven Signatures of Triple Negative Breast Cancer
By: Carol A. Rouzer, VICB Communications
Published: June 24, 2011
Once viewed and treated as a single disease, triple negative breast cancer is revealed to comprise seven subtypes with distinct therapeutic sensitivities.
Due to improved treatment and early diagnosis, the mortality rate for breast cancer (Figure 1) has been decreasing by 1.8% to 3.3% per year since 1990. However, African American women and women of other racial minorities have not benefited from this improved survival. One reason for the ethnic disparity is the higher incidence among minority patients of a particularly aggressive form of the disease known as triple negative breast cancer (TNBC). TNBC is named for the fact that the malignant cells in these tumors do not express the estrogen receptor, progesterone receptor, or HER2, a growth factor receptor expressed at high levels in some breast cancers. Since many of our most effective therapies are directed against these proteins, their absence in TNBC makes treatment particularly difficult. Consequently, the 5 year survival rate for TNBC is less than 30%, and almost all patients eventually die of their disease. Although TNBC has often been viewed and treated as a single clinical entity, growing evidence suggests that cancers meeting the triple negative criteria vary considerably with regard to other characteristics. This led Vanderbilt Institute of Chemical Biology investigator Jennifer Pietenpol and her laboratory to develop a bioinformatics approach that has identified 7 distinct subtypes of TNBC, each with its own characteristics and sensitivities to different therapeutic modalities. [Lehmann et al. (2011) J. Clin. Invest., published online June 1, DOI: 10.1172/JCI145014].
Figure 1. Figure 1. Gross specimen of a large human breast cancer. Figure reproduced from Wikimedia Commons under the GNU Free Documentation License.
The Pietenpol lab’s approach was based on the hypothesis that similar cancers should exhibit similar patterns of gene expression (GE), and that distinctly different GE patterns would form the basis for classifying disease subtypes. To test their hypothesis, they compiled 21 publicly available datasets encompassing comprehensive GE analysis of 3,247 primary breast cancers. They first analyzed their data to identify those cancers that did not express the estrogen or progesterone receptors or HER2, resulting in the classification of 587 tumors as TNBC. They then divided the data into a training set, encompassing 14 datasets with 386 TNBCs and a validation set, encompassing 7 datasets with 201 TNBCs. A statistical method known as k-means clustering identified seven distinct GE patterns among the TNBCs in the training set. A second approach, consensus clustering, yielded the same results, providing support for the existence of 7 subtypes of TNBC (Figure 2). When the investigators searched the validation set data for these same GE patterns, they found a similar distribution of tumors across the subtypes, indicating that the classification system was robust and generally applicable to TNBCs.
Figure 2. Graphic depiction of the results of statistical analysis of GE data from TNBC tumors. Each sphere represents a tumor, and the colors indicate the tumor subtype. Reproduced with permission from Lehmann, et al. (2011) J. Clin. Invest., published online June 1, DOI: 10.1172/JCI145014. Copyright 2011, Lehmann, et al.
Gene set enrichment analysis of the GE patterns of each of the TNBC subtypes revealed the biochemical pathways that were highly expressed in each subtype. This led to the characterization of the subtypes as:
- Basal-like 1 (BL1) - Tumors that have some characteristics of normal breast basal cells, but strongly express genes involved in cell division, the cell cycle, and the DNA damage response.
- Basal-like 2 (BL2) - Tumors that have some characteristics of normal breast basal cells, but strongly express genes involved in growth factor signaling, glycolysis, and gluconeogenesis.
- Immunomodulatory (IM) - Tumors that strongly express genes involved in immune system regulation.
- Mesechymal (M) - Tumors that strongly express genes involved in cell motility, growth, and differentiation.
- Mesenchymal stem-like (MSL) - Tumors that strongly express genes involved in cell motility, growth, and differentiation plus growth factor signaling and angiogenesis.
- Luminal androgen receptor (LAR) - Tumors that bear some resemblance to normal breast luminal cells, and which strongly express the androgen receptor and genes related to its signaling.
- Unstable - Tumors that did not meet the criteria of any of the other subtypes.
The GE patterns correlated well with protein expression data from the different tumors. For example, BL1 and BL2 tumors, characterized by GE profiles suggesting high rates of cell proliferation, contained high levels of Ki-67, a protein found in rapidly dividing cells (Figure 3). High levels of androgen receptor protein were present in LAR subtype TNBCs, but not in tumors belonging to the other subtypes (Figure 4).
Figure 3. Photomicrograph of TNBC tumor subtypes stained for the cell proliferation marker Ki-67 (brown color). Reproduced with permission from Lehmann, et al. (2011) J. Clin. Invest., published online June 1, DOI: 10.1172/ JCI145014. Copyright 2011, Lehmann, et al.
Figure 4. Photomicrograph of TNBC tumor subtypes stained for the androgen receptor (brown color). Reproduced with permission from Lehmann, et al. (2011) J. Clin. Invest., published online June 1, DOI: 10.1172/JCI145014. Copyright 2011, Lehmann, et al.
GE data revealed that 27 out of 30 TNBC cultured cell lines fit the criteria for classification into one of the seven subtypes. This provided an opportunity to investigate the differential response of each subtype to specific antitumor therapeutic agents. The high proliferation rate of cells in the BL1 and BL2 subclasses suggested that they would be particularly sensitive to DNA damaging agents, such as cisplatin. The importance of androgen receptor signaling in the LAR subclass led to the hypothesis that these cells would be susceptible to inhibitors of androgen receptor function, such as bicalutamide. Cell motility pathways were important in both the M and MSL subclasses, suggesting that therapeutic agents, such as dasatinib, that target the nonreceptor tyrosine kinase src would be effective against these tumors. Finally, the importance of the phosphatidylinositol 3-kinase (PI3K) pathway in M, MSL, and LAR subtypes suggested that agents such as NVP-BEZ235, that target PI3K signaling, would be effective therapies for those tumors. Subsequent tests of these hypotheses both in cell culture and in tumor xenografts in mice confirmed the predictions.
Together these findings demonstrate that TNBC is not one form of cancer. Rather, the triple negative phenotype is shared by up to seven different subtypes of the disease, each of which is driven by a unique set of genetic and biochemical drivers. Thus, it is not surprising that the TNBC subtypes respond differently to various therapeutic modalities. This important discovery provides the foundation for a new personalized approach to TNBC therapy that will identify the most effective therapy while limiting the patient’s exposure to toxic drugs that will likely be of little value.