TCGA's Breast Cancer Project May Yield Important Therapeutic Benefits, but It's Too Early to Be Sure

The Cancer Genome Atlas (TCGA) project represented a herculean effort to examine the molecular underpinnings of common yet heterogeneous diseases. Using a suite of tools encompassing the cutting-edge of omics technologies, the project cataloged all somatic mutations, differential gene expression, methylation, and protein expression in more than 1,000 breast cancers (among many other cancers).[1] As detailed in the excellent review by Ma and Ellis, the project successfully identified a plethora of key pathways that are either mutated and/or dysregulated in breast cancer.[2] However, the jury is still out on whether this genomic catalog of treatment-naive samples will transform breast cancer treatment.

A major aspect of the TCGA project was identification of significantly mutated genes (SMGs) in breast cancer. These SMGs are genes that are mutated at the DNA base-pair level at a higher rate than would be expected based on the background mutation rate. These include commonly mutated genes such as PIK3CA, TP53, MAP3K1, MAP2K4, GATA3, MLL3, and CDH11. Therapeutic exploitation of common mutations in breast cancer is currently underway, in particular with the phosphoinositide 3 kinase (PI3K) inhibitors[3] and the mammalian target of rapamycin (mTOR) inhibitor everolimus, approved by the US Food and Drug Administration in 2012.[4] The presence of PIK3CA mutations, particularly in estrogen receptor–positive (ER+) disease, was already known, and inhibitors were already in clinical trials prior to the publication of TCGA data.[3] The question remains whether and how novel therapeutic targets derived from TCGA’s analysis will be pursued. A potential novel therapeutic target derived from TCGA’s analysis is forkhead box M1 (FOXM1). This transcription factor, known for its role in mediating metastasis, was demonstrated to be upregulated and hyperactivated in luminal B and basal-like breast cancer.[1,5,6] In a network analysis, TCGA showed that FOXM1 is a hub protein driving proliferation of these cancers. Of note, inhibitors of FOXM1 are currently in preclinical development, and shutting down this central node may prove efficacious.

Another application of drug target discovery using TCGA data is the identification of genes that are uncommonly mutated but that may serve as powerful drug targets. An example is the recent identification of somatic human epidermal growth factor receptor 2 (HER2) mutations in non–HER2-amplified breast cancers. While rare, many of these are activating mutations with preclinical sensitivity to the kinase inhibitor neratinib (although not to trastuzumab or lapatinib).[7] This work has led to an ongoing clinical trial testing neratinib in patients with metastatic HER2-mutated, non–HER2-amplified breast cancer (National Cancer Institute ClincalTrials.gov ID NCT01670877).

Along the same lines of rare but targetable mutations, a notable analysis missing in TCGA is identification of expressed gene fusions. The lack of this analysis is surprising, given the identification of recurrent microtubule-associated serine-threonine (MAST) kinase and notch family translocations in breast cancers.[8] While rare, these gene fusions represent potentially targetable entities. Taken together, these common and rare targetable mutations may support targeted therapy beyond ER and HER2 for many breast cancer patients.

Another aspect of using genomics data is the derivation of novel classifications or subtypes. With regard to novel classifications of breast cancers informing therapeutic targeting, the TCGA project unfortunately provided little information that was new. Using complementary DNA (cDNA) microarrays more than a decade ago, Perou et al identified four intrinsic subtypes of breast cancer (luminal A, luminal B, HER2-enriched [HER2-E], and basal-like).[9] Now, in a combination of cutting-edge genomic, transcriptomic, epigenomic, and proteomic analyses, the data published by TCGA investigators in Nature redemonstrate the same intrinsic subtypes occurring at a very similar frequency.[1]

This lack of novel classifications may not be a failure of technological platforms, but of the analysis of the data. This is best demonstrated by the work of Curtis et al, which used the combination of germline and somatic copy number variation, single nucleotide polymorphisms (SNPs), and expression data to derive 10 subtypes of breast cancer, known as the integrated clusters or IntClust.[10] While overlapping with the intrinsic subtypes, the IntClust classification shows divergent subtypes, characterized primarily by large somatic copy number variants. Interestingly, in both papers, the input data that the investigative groups needed to derive each other’s classification were present, but each research group determined its own classification. This suggests not only potential bias in these classification analyses, but also that the way in which the data are dissected is key to determining the final endpoint. Another major barrier to the use of TCGA data for molecular classification is the lack of long-term survival and treatment data for development of new predictive and prognostic signatures. Future clinical updates to TCGA should aid new analyses.

One novel analysis that TCGA did perform was MEMo (Mutual Exclusivity Modules in cancer), which leveraged the multi-platform data at its disposal.[11] This analysis uncovered striking molecular-level similarities between basal-like breast cancer and ovarian cancer. While the link between germline BRCA1 mutation carriers predisposing patients to basal-like breast cancer and ovarian cancer has been known for years,[12] this analysis demonstrates a molecular similarity between sporadic and non–BRCA-mutated disease also. This includes similar frequencies of mutations in TP53, AKT3, MYC, CCNE1, and RB1, as well as BRCA1 and BRCA2, along with similarities in gene expression and copy number variation.[1] The strong molecular similarity between these two cancers lends evidence to a potentially shared therapeutic strategy that may extend beyond a mutual sensitivity to DNA-damaging agents.

Taken together, TCGA provides an extraordinary molecular encyclopedia of breast cancer. Its therapeutic outputs have not yet been realized, and its potential is limited by shortcomings that include the lack of clinical data on, and genomic profiling of, treatment-resistant tumors. Nonetheless, we agree with Drs. Ma and Ellis that TCGA’s data strongly suggest that a precision medicine approach-one that encompasses both common and rare mutations, amplifications, deletions, and gene fusions; and which considers each cancer as an N-of-1-is well warranted. Future clinical trials that employ this approach will determine whether genomics-based precision medicine is superior to the standard of care. Lastly, the strongest aspect of TCGA is that the data are publically available, fueling the input needed for unparalleled discovery. As the broader scientific community continues to analyze and integrate TCGA data with their own datasets, it is highly likely that breast cancer patients will benefit.

Financial Disclosure:The authors have no significant financial interest or other relationship with the manufacturers of any products or providers of any service mentioned in this article.

References:

1. Comprehensive molecular portraits of human breast tumours. Nature. 2012;490:61-70.

2. Ma CX, Ellis MJ. The Cancer Genome Atlas: clinical applications for breast cancer. Oncology (Williston Park). 2013;27:1263-79.

3. Rodon J, Dienstmann R, Serra V, Tabernero J. Development of PI3K inhibitors: lessons learned from early clinical trials. Nat Rev Clin Oncol. 2013;10:143-53.

4. Baselga J, Campone M, Piccart M, et al. Everolimus in postmenopausal hormone-receptor-positive advanced breast cancer. N Engl J Med. 2012;366:520-9.

5. Wierstra I, Alves J. FOXM1, a typical proliferation-associated transcription factor. Biological Chem. 2007;388:1257-74.

6. Raychaudhuri P, Park HJ. FoxM1: a master regulator of tumor metastasis. Cancer Res. 2011;71:4329-33.

7. Bose R, Kavuri SM, Searleman AC, et al. Activating HER2 mutations in HER2 gene amplification negative breast cancer. Cancer Discov. 2013;3:224-37.

8. Robinson DR, Kalyana-Sundaram S, Wu YM, et al. Functionally recurrent rearrangements of the MAST kinase and Notch gene families in breast cancer. Nature Med. 2011;17:1646-51.

9. Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumours. Nature. 2000;406:747-52.

10. Curtis C, Shah SP, Chin SF, et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature. 2012;486:346-52.

11. Ciriello G, Cerami E, Sander C, Schultz N. Mutual exclusivity analysis identifies oncogenic network modules. Genome Res. 2012;22:398-406.

12. Sorlie T, Tibshirani R, Parker J, et al. Repeated observation of breast tumor subtypes in independent gene expression data sets. Proc Natl Acad Sci U S A. 2003;100:8418-23.