In a series of transforming initiatives such as the Cancer Genome Atlas, the genomic landscapes for a variety of malignancies have been characterized in a reliable and robust manner and valuable new insights into ‘driver’ mechanisms responsible for the origin and progression of tumors have been gained. As a result of these efforts, we now have precise information on the genomic alterations present in frequent tumors, including breast, colorectal, ovarian, non-small-cell lung cancer, glioblastoma multiforme and endometrial cancer. It is also anticipated that the findings in other tumor types will be reported soon. The type of valuable information that we are acquiring is well exemplified in breast cancer, where just over the last 12 months a series of landmark studies using massively parallel sequencing to hundreds of breast cancers have provided us a comprehensive catalog on somatic mutations [1–6]. We now know that in breast cancer, somatic mutations are frequent in the TP53, PIK3CA and GATA3 genes, among others, and that the four main breast cancer subtypes have different subsets of genetic and epigenetic abnormalities.
In moving forward towards the goal of personalized and precision medicine, the identification of these potential ‘driver’ genomic alterations provides the opportunity to identify targets for therapy. Although not all targets are ‘druggable’ at this time, the cancer genome reporting has led to a number of hypothesis-driven drug discovery efforts. In addition, large-scale screens with carefully genomically annotated tumor cell lines also result in the identification of genetic, lineage and gene-expression-based predictors of drug sensitivity [7, 8]. These large cell line platforms may also be utilized to address optimal combination therapies and to explore in an unbiased fashion the activation of compensatory cellular responses that may lead to resistance to targeted therapies.
The identification of an increasing number of ‘actionable’ mutations and the availability of specific targeted agents will profoundly change the way we conduct clinical research and practice medicine. Since the list of ‘actionable’ mutations is likely to be large for a number of tumor types, we will need to embrace as routine clinical practice comprehensive tumor genomic profiling in those cases where treatment decisions need to be made. To address this need, a number of academic centers and commercial vendors are pursuing a targeted, massively parallel sequencing approach to detect tumor genomic alterations in formalin-fixed, paraffin-embedded tumor samples [9]. Furthermore, these technologies are suitable for the analysis of tumor suppressor genes, can detect additional types of genomic alterations including copy number alterations and structural rearrangements. They also have greater sensitivity for low allele frequency events and enable rapid addition of newly identified biomarkers.
The application of these sequencing platforms is already having a transformational effect in drug development at several levels. First, it is allowing us to explore correlations between mutational profiles and response to therapy in clinical trials with targeted agents. This is exemplified in the Bolero-2 study, the pivotal study in advanced hormone receptor positive breast cancer that leads to the approval of the combination of the mTOR inhibitor everolimus with exemestane [10]. In a subset of 227 patients representative of the whole patient population, their tumors were analyzed with a next-generation sequencing platform. The key findings included mutations in PI3KCA in close to 50% of cases, cyclin D1 amplification in 30%, p53 mutations or loss in 23% and FGFR1 amplification in 18% of cases [11, 12]. Interestingly, this study established that the presence of PI3K mutations does not result in greater sensitivity to everolimus, as we had predicted based on preclinical models [13]. Other actionable alterations included AKT mutations, estrogen receptor (ESR1) mutations and PTEN loss. These findings are currently being analyzed to generate new hypothesis for combinations of targeted agents in the therapy of hormone receptor-positive breast cancer. For example, they suggest that in patients with FGFR1 amplification we should consider the addition of an anti-FGFR agent.
In addition to retrospective analyses, massive sequencing allows for genotype-driven clinical trials where a number of patients with a specific mutation are prospectively enrolled in a clinical trial with a targeted agent directed at the mutation. In some cases, the mutations and/or genomic alteration are frequent enough that a clinical trial can be readily conducted in a single-tumor type, as in the case of BRAF mutations in melanoma or EGFR mutations in lung cancer. However, there are a number of mutations that are present in multiple tumor types at a low frequency, usually in <5% of cases. In order to be able to study targeted therapies in low-frequency mutations across multiple tumor types, we are conducting a series of ‘basket trials’ in which patients with multiple tumor types and the same genomic alteration are being enrolled. The concept is to enroll under a single clinical trial 10–15 patients per tumor type with a same mutation. A ‘basket trial’ is currently being pioneered for the study of the BRAF inhibitor vemurafenib in patients with tumors harboring BRAF V600E mutations other than melanoma and papillary thyroid cancer. Similarly, the PI3K alpha subunit specific inhibitor BYL719 has been initially studied only in patients with tumors harboring PI3KCA mutations [14] and clinical activity has already been established in this patient population. We are anticipating a number of ‘basket clinical trials’ to be soon underway for a variety of actionable genomic alterations including, in addition to BRAF and PI3K, FGFR amplification, Erbb2 mutations, AKT mutations and PTEN deletions among others. These trials have an inherent degree of flexibility in their design and may evolve as information is gathered during the study. For example, specific tumor cohorts where activity has already been observed can be easily expanded or combinations with new therapies can be added. To this point, in the case of the BRAF ‘basket trial’ mentioned above, in the cohort of patients with colon cancer we have added cetuximab to the BRAF inhibitor vermurafenib based on recently published studies that support the need for simultaneous dual blockade of the EGFR and BRAF in BRAF mutant colon cancer cell lines. Similarly, after early signs of clinical activity with the PI3K alpha inhibitor BYL719 in patients with ER+ PI3KCA mutant breast cancer and having observed in the laboratory a synergistic effect by simultaneous blockade of ER and the PI3K pathway, we have now added hormonal therapy to BYL719 in an ongoing expansion cohort. In summary, this ‘basket trial’ approach allows the testing of a defined biological hypothesis and to study the effect of lineage on drug sensitivity, which, in turn, allows generating additional hypothesis. For example, a non-responding lineage may be due to lineage specific differences in the pattern of co-altered expressed genes (such as in the case of BRAF mutant colon cancer and EGFR activation) or lineage-specific stromal interactions. These protocols do require a culture of multiple team collaboration as different disease teams need to work together.
An alternative approach to genotype-based clinical trials is what has been referred to by David Solit and others as phenotype to genotype studies. The guiding principle here is to identify the genetic basis for rare, extraordinary clinical responses. The identification of the genetic drivers of these clinical responses would then lead to guide trials in select subpopulations. This concept was tested in a patient with metastatic bladder cancer who was enrolled in a clinical trial with the mTOR inhibitor everolimus. In a phase II clinical trial of everolimus in patients with advanced bladder cancer, the drug was felt to be only marginally active. However, there was one patient that had a major and long lasting response to everolimus with no evidence of progression after 24 months of starting therapy. Using whole-genome sequencing a loss-of-function mutation in TSC1 (tuberous sclerosis complex 1), a regulator of mTOR pathway activation, was found in the tumor of the responding patient [15]. Targeted sequencing revealed TSC1 mutations in ∼8% of 109 additional bladder cancers examined, and TSC1 mutation correlated with everolimus sensitivity in all the other patients that had some degree of tumor response. It is anticipated that this approach to ‘mine’ the outliers could also be of great promise in understanding the determinants of response to commonly used chemotherapeutic agents.
Finally, as hinted above, a full genomic characterization of tumors could lead to a better identification of the appropriate combinatorial approaches as well as an understanding of potential mechanisms of resistance to targeted therapies if at the time of progression, tumors are reanalyzed and compared with their baseline features. It will also be important to understand the determinants of response to immunologic check point blockade as these therapies are beginning to display unprecedented responses in some tumor types [16].
In closing, these are exciting times for the oncology and early drug development community and most importantly for our patients. Our goal is the implementation of precision medicine and the delivery of individualized cancer therapy based on the mutational landscape, heterogeneity and lineage of each tumor. This will require integration between our clinical activities and the new sequencing platforms and a new set of clinical trial designs. Strikingly, it may also lead us into a renaissance period for clinical medicine as single observations by astute clinicians could lead to major breakthroughs. So in the era of genomic medicine a single case (n of 1) could be far more valuable than some large clinical trials. For those in the oncology community who are clinicians at heart, welcome back, we need you.
disclosure
The author has declared no conflicts of interest.
