Targeted Therapies in Triple-Negative Breast Cancer: Failure and Future

“We see the future of development of personalized therapy for triple-negative breast cancer as based on biology-oriented comprehensive approaches.”

Triple-negative breast cancer (TNBC) is defined by lack of estrogen receptor (ER) and progesterone receptor (PR) expression and by normal levels of HER2. TNBC represents 15–20% of all breast cancers and is associated with poor patient outcomes [1]. While patients with HER2-positive disease can benefit from trastuzumab and patients with ER-positive disease can benefit from endocrine therapy, the only recommended treatment for TNBC is standard chemotherapy; there is no targeted therapy specifically recommended for patients with TNBC. Although efforts so far to identify targeted therapy for TNBC have not been successful, we are hopeful that new approaches will lead to identification of targeted therapies for TNBC in the future.

Need for TNBC classification to develop therapy strategy

TNBC is a heterogeneous disease comprising a spectrum of cancers with distinct activated biological pathways, various levels of chemosensitivity, and different propensity for metastasis. Clinically relevant biomarkers need to be identified to guide development of targeted therapies.

Significant efforts have been made to better understand the clinical impact of the activated biological pathways in TNBC. Many classifications have been proposed based on metabolic characteristics, immunohistochemical biomarkers and gene signatures. Out of the classifications proposed to date, two based on gene-expression profiles are the most commonly cited: the PAM50 classification, for breast cancer in general and Lehmann's seven subtypes, for TNBC only [2,3]. In both classifications, each subgroup is characterized by dominant biological functions or pathways; these include DNA repair deficiency for basal TNBC, epithelial-to-mesenchymal transition for mesenchymal and claudin-low TNBC, immune pathway for immune-regulatory TNBC, androgen receptor for luminal-apocrine TNBC and HER2 pathway for HER2–enriched TNBC. Each of them is associated with significantly different outcomes.

Unfortunately, only limited driver mutations have been identified in TNBC, which suggests that treatment strategies based on gene mutation or amplification are unlikely to come up in the near future. However, the high frequency of p53 mutations (80%) in basal-like breast cancer – which represents a large number of TNBC – have prompted a new effort to make p53 mutations drug-gable [4]. The drugs able to restore or activate the p53 pathway have promising results in various neoplasms: gendicine (adenovirus-mediated transfer of human wild-type p53) has already demonstrated efficacy in head and neck and lung cancer [5]; PRIMA-1^MET (p53-binding molecule to restore p53 wild-type transcriptional properties) and MDM2 inhibitors (inhibitors p53-negative regulators) are currently investigated in Phase I clinical trials [6].

The next most commonly mutated genes in TNBC were BRCA1 and BRCA2 (20% of patients have mutations in one or both genes) and PI3KCA (9%) [4]; PARP inhibitors and PI3K/AKT/mTOR inhibitors are being tested for breast cancer patients with these mutations in clinical trials [7,8].

Classification of patients with TNBC on the basis of chemosensitivity is another approach that has been proposed. TNBC is potentially heterogeneous in terms of response to chemotherapy. If a pathological complete response or a residual cancer burden I is achieved, TNBC patients have survival similar to that of patients with non-TNBC. However, TNBC patients with residual disease after chemotherapy have an extremely poor prognosis [9,10]. Could dividing TNBC into chemo sensitive and chemoresistant disease be sufficient to personalize TNBC treatment and select TNBC subpopulation who could deserve from additional therapy?

Need for interinstitutional collaboration

To develop personalized therapies for TNBC, it is necessary to fully understand the molecular basis of the oncogenic pathways and microenvironmental changes that promote the disease. The better we understand the biology of TNBC, the more we are likely to split TNBC into multiple subgroups. However, proliferation of subtypes could yield many ‘orphan’ subgroups of TNBC for which it would be diffcult to design clinical trials with optimal statistical power (sufficient numbers of patients). For example, SHIVA European clinical trial (NCT01771458) is a proof of concept: if a molecular abnormality is identified for which an approved targeted agent is available, patients are randomized between this targeted therapy and the conventional therapy [11]. The National Cancer Institute also investigates the utility of genetic sequencing to determine therapy, independent of tumor histology, with the M-PACT (study of 20 mutations for four drugs proposed) and the MATCH pilot studies (large screening to match more than 20 US FDA-approved investigational agents) [12,13].

“Negatve results do not always mean ‘ineffective drug;’ often negative results could mean ‘misused drug’.”

To overcome this problem, interinstitutional collaboration should be encouraged from the first step of drug evaluation. For example, only 4% of TNBC cell lines harbored an FGFR2 amplification, which would make cells highly sensitive to FGFR inhibitor [14]. Thus, to design a powerful TNBC trial using FGFR amplification as an inclusion criterion, a huge number of patients (if the study needs 40 patients, 1000 patients with TNBC) would need to be screened. Such a screen could only be achieved with the involvement of many institutions. This means that not only is there is an urgent need to identify clinically relevant targets but also there is a need to design trials to ensure rapid accrual of sufficient numbers of patients with the molecular subtypes or targets of interest.

Need for identification of clinically relevant biomarkers

Matching the right drug to a particular cancer subtype is challenging, especially given the vast number of targeted therapies brought to us every year by the pharmaceutical industry. Most of the current trials of systemic therapy for breast cancer are designed for the overall breast cancer population or for all patients with TNBC. However, as previously explained, response of TNBC to treatment could be very heterogeneous due to the molecular heterogeneity. So, if a prospective study in breast cancer patients or patients with TNBC in general has negative findings, can we really conclude that the tested drug is ineffective? Should we not rethink the targeted population? For example, FDA approval of bevacizumab in breast cancer was recently withdrawn because of insufficient benefit in the overall breast cancer population. However, a meta-analysis suggested a 35% reduced risk of relapse and a 19% response rate, albeit without an overall survival improvement, for the subset of patients with metastatic TNBC [15]. Although a subgroup seems to benefit from this targeted therapy, so far, no biomarkers of bevacizumab efficiency have been approved to help us accurately select this population. There is an obvious lack of predictive biomarkers of treatment efficacy for actual targeted drugs. Most drugs are tested in clinical trials before standardization, assessment of repro-ducibility, cost–effectiveness testing, and prospective validation of efficacy biomarkers [16]. This step should become mandatory in the design of current Phase III clinical trials to ensure proper selection of the patients most likely to benefit from targeted drug and obtain the most meaningful results. Negative results do not always mean ‘ineffective drug;’ often negative results could mean ‘misused drug’.

Carboplatin provides a good example. This old DNA cross-linking agent, commonly used in the treatment of other neoplasms, was only recently identified as beneficial in patients with TNBC. Interestingly, carboplatin was used for high-dose chemotherapy with autologous stem cell transplant, and efficacy of platinum in TNBC was strongly suggested almost 20 years ago. Good response rates to such agents were associated with germline or sporadic DNA repair deficiency, also termed ‘BRCAness signature’ [17]. Platinum-based chemotherapy for TNBC with BRCAness signature is associated with significant improvement of the pathological complete response rate from 40 to 54%. [17,18] In patients with metastatic disease, this BRCAness signature (as estimated by homologous recombination deficiency assay) was predictive of platinum response [19]. Thanks to a well-defined target population, the efficacy of this old-fashioned drug in TNBC was demonstrated. The best drug for a particular cancer subtype is not always the newest drug.

Need to examine biomarker concordance between primary tumors & recurrent disease

Another issue, not specific to TNBC or even to breast cancer in general, is that using primary tumor molecular signatures to guide treatment selection in patients with recurrent disease may not be appropriate. For the past decade, biomarker discordance between primary tumors and recurrent disease has been widely studied. Three prospective studies showed changes in biomarker status between primary tumors and recurrent disease. These studies showed discordance in ER status in 10–16% of patients, in PR status in 24–40% of patients and in HER2 status in 7–10% of patients. These changes led to therapeutic changes in up to 20% of patients [20]. Consequently, the National Comprehensive Cancer Network recommends obtaining biopsy samples of recurrent disease and assessing the expression of ER, PR and HER2 in these samples if the status of these biomarkers in the primary tumor was unknown, negative or underexpressed [21]. The status of newer potential biomarkers such as PTEN loss and PI3KCA mutations was also found to be discordant between primary tumors and recurrent disease [22]. However, even though we now have proof of adaptive response to endocrine or anti-HER2 therapy, biopsies of recurrent disease are not always performed in clinical practice. Repeat biopsy to determine the molecular signature at progression or relapse seems mandatory to ensure accurate targeted treatment [20].

Furthermore, monitoring changes in tumor characteristics during treatment could also be relevant. Biopsies of the tumor before and after anti-EGFR therapy showed that although most tumors had EGFR expression and pathway activation, cetuximab blocked expression of the EGFR pathway in only a few cases, suggesting an alternate mechanism for pathway activation. All these data highly suggest biological changes due to treatment exposure or natural evolution of the neoplasm [23].

In the recent prospective SAFIR-001 trial, using whole-genome arrays (e.g., array comparative genomic hybridization and DNA sequencing), the investigators were able to detect targetable genomic alterations in 70% of biopsy samples from patients with metastatic breast cancer, and these alterations were used to personalize targeted treatment. Updated results regarding the percentage of patients who benefited from targeted therapy based on genomic screening results are expected [24]. Genomic analysis to determine a molecular target at relapse seems feasible and could substantially improve tailoring of treatment.

More easily assessable biomarkers: circulating tumor cells & circulating free tumor DNA

Although repeat biopsy at the time of progression or relapse could improve the efficacy of treatment for TNBC, convincing heavily pretreated patients to undergo another invasive biopsy could be difficult. Moreover, the risks of tumor seeding and adverse events associated with invasive procedures, mostly for patients receiving anti-angiogenic treatment, could contraindicate traditional biopsy.

“Monitoring cfDNA could allow oncologists to change ineffective treatment before imaging detection of relapse.”

Development of more easily assessable biomarkers to monitor metastatic disease and treatment response is a hot topic in numerous type of cancer, including breast cancer. Declining levels of circulating tumor cells (CTCs) after chemotherapy have been correlated with increased progression-free survival and overall survival (p < 0.0001) and thus potentially with chemosensitivity [25]. More recently, another type of ‘liquid biopsy’ has been developed to monitor cancer genetics: measurement of circulating free tumor DNA (cfDNA). Detection of cfDNA has prognostic implication [26], and a recent study showed that it was possible to track the genomic evolution of metastatic disease in response to treatment by analyzing cfDNA [27]. Monitoring cfDNA could allow oncologists to change ineffective treatment before imaging detection of relapse.

However, for both kinds of ‘liquid biopsies’ – monitoring of CTCs and monitoring of cfDNA – the techniques need to be standardized (only one CTC platform has been approved by the FDA so far), and clinical validation is needed before daily use.

Recommendations

Our group is also attempting to design a comprehensive treatment strategy for TNBC. First, baseline molecular profiling and imaging during standard chemotherapy will identify a chemoresistant population who can benefit from additional treatment. Second, patients will be divided on the basis of their BRCA status (BRCA mutant and BRCAness signature versus none). Then, patients without BRCA mutation or BRCAness signature will be segregated into mesenchymal TNBC or non-mesenchymal TNBC subgroups. Each of these subgroups could benefit from an additional investigational treatment: PARP inhibitors for carriers of BRCA1 or BRCA2 mutations, DAT [doxil (doxorubicin HCL liposome injection), avastin (bevacizumab), temsirolimus] for patients with mesenchymal TNBC, and EGFR inhibitors or androgen receptor inhibitors for patients with nonmesenchymal TNBC. In parallel, multiple biomarker-driven clinical trials with integrated biomarker assessment are under way. This treatment strategy is only one example of possible future TNBC treatment strategies. Identification of clinically relevant mutations or gene amplifications may lead to further personalization of breast cancer treatment. Moreover, functional signaling pathway enrichment, such as based on Lehmann's classification, could bring new treatment approaches [3].

We see the future of development of personalized therapy for TNBC as based on biology-oriented comprehensive approaches. The first step is to develop robust biomarkers (or collections of biomarker profiles) that reflect the biological behavior of TNBC. However, these biomarkers may not always be mutation or amplification. The second step is to determine whether these molecular targets or adaptive changes of the targets by treatments will predict response to treatments. The third step is whether the lead out may not be always the shrinkage of tumors; rather, it could be control of metastasis. The last step is to avoid treatment resistance by closely monitoring the evolution of the tumor under treatment pressure and altering the choice of treatment as needed.

Acknowledgements

The authors thank the MD Anderson Moon Shot programs for their support. They also acknowledge the Scientific Publications Department at MD Anderson for their helpful review of the manuscript.