Unravelling the Complexities of Breast Cancer Genomics

May 31, 2014

Jorge Reis-Filho, MD, PhD, of the Memorial Sloan Kettering Cancer Center chaired the Education Session, “Frontiers in Precision Medicine: Practical Application of Genomic Tools in Breast Cancer Management,” held Friday. The session began with Dr. Reis-Filho’s presentation, “Genomics 101: Introduction to and Definitions of Current Genomic Tools in Breast Cancer.” We have learned from breast cancer gene expression analyses that breast cancer is not a single disease, but is made up of multiple subtypes such as luminal A, luminal B, HER2-enriched, and basal-like. Molecular subtypes of triple-negative breast cancer include basal-like I, basal-like II, basal mesenchymal, mesenchymal stem-like, immunomodulatory, and luminal androgen receptor. In addition, the Molecular Taxonomy of Breast Cancer International Consortium has further classified breast cancer tumors using molecular signatures. Fifteen years of microarray analysis has also led to the realization that estrogen receptor-positive (ER+) and ER-negative (ER-) tumors have fundamental differences. For example, ER+ cancer outcomes can be predicted by proliferation-related genes, whereas ER- breast cancer prognosis is associated with immune response-related genes. “Microarrays, sadly, did not result in ways for us to define the best therapy in individual patients,” Dr. Reis-Filho noted.

Precision medicine refers to the use of available data to determine disease characteristics in the pursuit of optimal treatment for individual patients. The current management of breast cancer patients consists of tumor biopsy, histopathology, gentotyping, pathway analysis, treatments based on these findings, and subsequent real-time monitoring using repeat biopsies, PET scanning, and analysis of circulating tumor cells. Dr. Reis-Filho believes that precision medicine is now possible in breast cancer due to two major developments: targeted treatments (small molecule inhibitors and monoclonal antibodies) and massive parallel sequencing of tumor genomes, also known as next-generation sequencing (NGS). By using these platforms, a great deal of information can be obtained in a very short time.

Cancer cells are dependent on oncogenic expression to maintain malignancy, and these can be used as predictive biomarkers. Examples include oncogene addiction as in HER2 amplification in breast cancer, the KIT mutation in gastrointestinal stromal tumors, epidermal growth factor receptor mutations and/or amplification in nonsmall cell lung cancer, and the BRAF V600E mutation in melanoma. Data obtained from sequencing tumors from patients with breast cancer has revealed a high level of intertumor and intratumor heterogeneity, with very few highly recurrent mutations. It has also become apparent that not every tumor has an identifiable driver mutation, and not all drivers of metastatic disease or resistance have been identified. “It is only by harnessing the knowledge from intra- and intertumor heterogeneity that we will be able to realize the application of precision medicine,” Dr. Reis-Filho concluded.

Available Testing Platforms

Banu Arun, MD, of The University of Texas MD Anderson Cancer Center, gave an overview of the current genetic testing platforms available to clinicians in her presentation, “Current Use of Genomics in Clinical Practice: What Is Useful, What Is Not, and How to Communicate Results to Patients.” MammaPrint is a test used for prognostication of patients with stage I, II node-negative breast cancer with tumors less than 5 cm. Fresh or frozen samples, or paraffin-embedded tissue are required, with at least 30% tumor cell content. Supervised analysis of 25,000 genes resulted in 70 genes that predict prognosis. MammaPrint was shown to predict benefit from chemotherapy in a group of patients who had a poor prognosis. PAM 50 is a 50-gene predictor that uses microarray and quantitative reverse transcription polymerase chain reaction (RT-PCR) from 189 untreated patients. This predictor was demonstrated to provide prognostic significance and predicted benefit in chemotherapy. The Breast Cancer Index (BCI) was developed from a randomized, prospective study using the algorithmic combination of HOXB13:IL17BR ratio (H/I) and the Molecular Grade Index. BCI was validated in two retrospective cohorts to predict risk of early (0-5 years) and late recurrence (5-10 years), and risk of overall recurrence (0-10 years). EndoPredict is an RNA-based multigene score that was validated in two cohorts to predict early and late distant recurrence. The Oncotype DX Recurrence Score (RS) model is an RT-PCR assay of 21 selected genes (16 cancer-related and 5 reference genes) that was demonstrated to have predictive value in breast cancer recurrence. An RS score of less than 18 is considered low risk, a score between 18 and 31 is intermediate risk, and a score over 31 is classified as high risk.

Dr. Arun eagerly awaits the results of several ongoing prospective studies that are further evaluating some of these tools. The Trial Assigning Individualized Options for Treatment (Rx; TAILORx) is being conducted in patients with node-negative, ER+ breast cancer. Patients were evaluated with the Oncotype DX assay and randomly assigned to receive treatment based on their RS. Patients at low risk (RS ≤ 10) were entered into the hormone therapy registry, those with intermediate risk (RS 11-25) were randomly assigned to receive hormone therapy or chemotherapy plus hormone therapy, and those at high risk (RS > 25) received chemotherapy with hormone therapy. The RxPONDER Trial is a large study being conducted in patients with node+, HER2- breast cancer. Patients with an RS less than or equal to 25 were randomly assigned to receive either chemotherapy with appropriate endocrine therapy or appropriate endocrine therapy alone, stratified by RS, menopausal status, and axillary node dissection versus sentinel node biopsy. Alternative clinical trials were discussed with patients high-risk who have an RS over 25. The European Organisation for Research and Treatment of Cancer 10041/BIG3-04 MINDACT trial is being conducted in patients with node- or three or fewer involved lymph nodes. Patients are assessed using MammaPrint and Adjuvant!, an online breast cancer risk estimation tool. Patients with a good prognosis by both methods did not receive chemotherapy, whereas those with a poor prognosis by both methods were administered chemotherapy. Patients with discordant results were randomly assigned to either chemotherapy based on the Adjuvant! or MammaPrint prognosis.

Figure 1

Click to Expand.

Dr. Arun concluded her talk with a brief description of FoundationOne, an NGS tool that is in clinical use. FoundationOne identifies somatic mutations by interrogating 236 cancer-related genes and 47 introns from 19 genes that are frequently rearranged in solid tumors. It is possible that actionable mutations can be identified, and patients can be given targeted therapy or enrolled in clinical trials. Dr. Arun cautioned that nonactionable mutations or incidental mutations in hereditary genes may also be identified. In these cases, she advised that patients be referred for genetic counseling and undergo standard clinical germline testing.

Potential Implications

The implementation of precision medicine in the treatment of metastatic breast cancer faces many challenges, but there are several ways that genomic testing could help improve breast cancer treatment. The first is the identification of the oncogenic drivers that lead to malignancy, progression, and death. According to Fabrice Andre, MD, PhD, of the Institut Gustave Roussy, France, there are currently 10 to 20 candidate actionable oncogenic drivers, but most of these occur in a fraction of the patients with metastatic disease and few have been tested in clinical trials. The classification of genes based on their probability of being drivers is an ongoing area of biology, genomics, and bioinformatics research. Identification of potential drivers, as well as refining the definition of exactly what constitutes a driver, are areas that need additional research. The possibility of coexisting drivers further complicates the picture, as this may ultimately lead to resistance.

Genomic tests may also help predict resistance. High levels of heterogeneity within the tumor and genomic instability increase the likelihood that these tumors will develop resistance and elude targeted therapies. One approach is to identify lethal subclones within the primary tumor using deep sequencing. Monitoring circulating DNA may aid in the identification of subsequent lethal clones that are resistant to therapy. In addition, studying patterns of mutations in sequenced tumors may also help to decipher the mutational processes and DNA repair defects in individual patients. In addition, finding immune system defects in patients with metastatic breast cancer could determine the individuals who might benefit from adjuvant immunotherapy.

Stratified medicine has typically been used to test whether an agent will improve outcomes in metastatic breast cancer. In this approach, genomics is used to screen patients, cohorts are generated based on screening results, and then a drug is tested in the appropriate groups of patients (Fig 1). Given the rarity of some of the mutations in metastatic disease, this methodology may not be feasible for genomic medicine trials. One of the problems with this model is low trial accrual. For example, if a mutation was present in 4% of patients with metastatic breast cancer, more than 5,000 patients would need to be screened to find 200 patients for the study. Another approach would be to use a study design where molecular screening tests an algorithm rather than a particular treatment. This model takes all participants and patients are randomly assigned to receive either standard care or personalized medicine (Fig. 1). Dr. Andre concluded by noting that ensuring equal access to genomic testing remains another large undertaking affecting the future of precision medicine.