DISCLOSURES: Dr. Cortés has received compensation as a consultant from Roche, Celgene, and Novartis and for lectures, including service on the speakers' bureaus, from Roche, Celgene, Novartis, and Eisai.
Currently, treatments for cancer and basic, translational, and clinical cancer research are essentially designed based on the organ of the body in which the cancer is originally observed. However, global gene expression and genomic and epigenomic analyses using high-throughput technologies have helped to elucidate the molecular characteristics of cancer, providing a new angle to oncology. We are observing that individual tumors generated within the same organ show an enormous diversity of biologic behavior, clinical prognosis, sensitivity to treatments, and therapeutic targets. Moreover, tumors from different sites share important features relative to oncogenic mutations and tumor microenvironments of crucial relevance for cancer treatment.
Here, we propose a different approach to cancer classification based on the molecular characteristics of cancer, and this will have important implications for cancer treatment and research. Thousands of patients are entered into organ-based clinical trials in which, on many occasions, modest benefits, if any, are obtained. This is not acceptable, and novel and innovative perspectives have to be evaluated to accelerate and improve our endeavor to treat cancer. Throughout this report, we propose a new cancer classification based on the molecular characteristics of tumors, and we describe specific examples of how the novel classification could impact cancer research and treatment.
Driver Mutations Determining Response to Treatment
Despite tumoral complexity, increased knowledge about the molecular characteristics of cancer provided hope for the development of better therapeutic approaches. Gene mutations present in tumors were classified into driver and passenger mutations. Driver mutations confer a selective advantage to the tumor and, hence, the inactivation of these mutations should affect tumorigenesis. However, during cancer progression, other somatic mutations arise that do not confer a selective fitness advantage. These passenger mutations could theoretically be reversed with little or no effect on tumor progression. Following this definition, driver mutations constitute therapeutic targets, whereas passenger mutations might be irrelevant for therapeutic purposes.
Even when a tumor contains several driver mutations, a single driver mutation sometimes can have a crucial role in the viability of the tumor. In this scenario, the inactivation of one driver mutation can suppress cancer. The tumor is in some way addicted to one mutation. This phenomenon, termed “oncogene addiction,” provides a rationale for molecular-targeted therapy. V-erb-b2 avian erythroblastic leukemia viral oncogene homolog 2 (HER2), v-raf murine sarcoma viral oncogene homolog B (BRAF), epidermal growth factor receptor (EGFR), PATCHED, and others are examples of oncogene addiction.[2-4]
Different tumor types tend to have an increased propensity for specific driver mutations, such as HER2 amplifications in breast cancer or point mutations and indels in genes, such as BRAF in melanoma, EGFR in nonsmall cell lung carcinoma (NSCLC), and PATCHED in medulloblastoma. However, alterations in the same genes are detected in a variety of other tumors. Breast and gastric tumors share alterations in HER2[7, 8]; NSCLC and gastric tumors share alterations in EGFR; and NSCLC, anaplastic large cell lymphoma, and renal medullary carcinoma share alterations in anaplastic lymphoma receptor tyrosine kinase (ALK).[10-12] Moreover, driver mutations of the same genes can be present with low frequency in many different tumor types; for example, mutations in HER2 can also be found in 8% of glioblastomas and in 2% of NSCLC adenocarcinomas.
Hence the presence of driver mutations does not depend on the organ in which the tumor arises, and tumors of different organs can be “addicted” to the same oncogene. More important, different tumor types with the same oncogene addiction respond to similar therapeutic strategies.
A Novel Classification of Cancer
Following this rationale, we propose a different approach to the classification of cancer that may have important implications for treatment and research. If driver mutations present in tumors determine the treatment of cancer, then we should study and classify cancer based on its molecular features rather than by the organ of origin of the tumor. Thus, comparisons between different tumor types with common molecular characteristics not only could help us understand the biology of these diseases but also would have crucial research and therapeutic implications. New clinical trials should include patients with similar tumors, as defined at the molecular level, independent of the organ of the tumor. Some examples are discussed below.
Same Driver Mutations in Different Tumor Types
In the case of human epidermal growth factor receptor 2 (HER2)-positive tumors, instead of working individually with HER2-positive breast cancer, HER2-positive gastric tumors, or HER2-positive adenoid cystic carcinomas of the parotid gland, we propose to work on HER2-positive malignant tumors altogether. We could use the term “HER2-dependent tumors,” reflecting that HER2 is a critical oncogene for the proliferation and survival of tumors. HER2-positive tumors are more biologically homogeneous than, for example, sarcomas or colorectal tumors, which are currently being included in their respective tumor-type–directed studies. Trastuzumab was approved for the treatment of metastatic breast cancer 10 years ago but only recently was approved for gastric cancer. Moreover, pertuzumab was approved in HER2-positive metastatic breast cancer after demonstrating an improvement in progression-free survival. Efficacy of pertuzumab has just begun its testing in gastric cancer and is quite unlikely to be explored in other HER2-positive tumor types. Following our proposed approach, patients with HER2-positive breast cancer, gastric cancer, or other HER2-driven types of cancers would have been randomized to receive the standard of care with or without trastuzumab, and patients might have benefitted from trastuzumab therapy years before they actually did. The same may happen with all of the new, upcoming anti-HER2 approaches (pertuzumab, trastuzumab-DM1 [ado-trastuzumab emtansine], and others). Our proposed new clinical trial designs are schematized in Figure 1.
In the HER2-dependent context, it is worth noting that not all tumors have the same alteration; although most tend to bear HER2 gene amplifications, a small proportion of lung tumors have been shown to harbor point-activating mutations in HER2 (similar to the EGFR mutation pattern). A case report of a patient who had NSCLC with a point mutation in HER2 that resulted in a response to trastuzumab raises the issue that there might be different alterations that lead to HER2 dependency and that these should also be considered in the same group. It is noteworthy that HER2-positive tumors are just one of the multiple examples that support our novel approach.
Sonic hedgehog homolog is one of the proteins in the mammalian signaling pathway family called hedgehog. Aberrant activation of the hedgehog signaling pathway is strongly implicated in the development of some cases of medulloblastoma. In addition, deregulated hedgehog signaling is the pivotal molecular abnormality underlying basal cell carcinomas.[18-20] Vismodegib, a new hedgehog pathway inhibitor, has shown activity in both tumor types. It is noteworthy that vismodegib was approved for the treatment of basal cell carcinoma,[21-23] and trials evaluating the role of vismodegib in patients with advanced medulloblastoma are still recruiting patients. Hedgehog signaling is frequently activated in other human cancers. Thus, clinical trials enrolling patients who have tumors with aberrant activation of the hedgehog signaling pathway, independent of tumor location, may help accelerate the approval of new drugs for a broader spectrum of patients.
BRCA1 and BRCA2 are well known tumor-supressor genes. The term “BRCAness” identifies a broad range of tumors with dysfunction of DNA repair homologous recombination. Once again, poly (ADP-ribose) polymerase (PARP) inhibitors have been studied separately across different tumor types. For example, olaparib (AZD-2281) was studied in BRCA-deficient advanced breast cancer on 1 hand and in BRCA-associated ovarian cancer on the other. Not surprisingly, the activity in terms of overall response rates (41% and 33% in breast and ovarian cancer, respectively) and median time to progression (5.7 months and 5.8 months in breast and ovarian cancer, respectively) was overlapping in both trials.[25, 26] Further development of this molecule will definitely assess the role of PARP inhibitors in these “different” tumor types. We suggest that trials including tumors with BRCAness will be more beneficial to the patient.
Our proposal is based on the finding that the joint study of tumors with similar molecular characteristics (which might be very difficult to define sometimes) will be more beneficial for the understanding of cancer biology and will lead to an improvement in cancer therapy. However, several important concepts have to be considered to promote the fast advance of cancer research and treatment. These considerations are relevant for our proposed model for cancer classification and also for the organ-based classification.
Tumors are heterogeneous and contain several genetic clones that, in some cases, exhibit an important degree of evolutionary divergence. The analysis of distant metastasis also reveals additional clonal evolution, and metastasis may arise from different clones. These observations exemplify the large genomic complexity of cancer and have important implications in the identification of the critical driver mutation that determines the treatment of a tumor.
However, a way to circumvent intratumoral heterogeneity might be the identification of root mutations, which are present in all the different clones because they arose early in tumor progression. The identification of von Hippel-Lindau tumor suppressor, E3 ubiquitin protein ligase (VHL) as a root mutation in renal cancer may suggest that compounds targeting VHL or its pathway might be beneficial independent of the clonal diversity of the tumor mass.
Another aspect to consider is that, along with relevant therapeutic driver mutations, tumors have other driver and passenger mutations that might affect response to treatment. The way different mutations arise during tumorigenesis and how they network together are relatively unknown areas yet may predict treatment failure. In fact, under certain selective pressure circumstances, such as targeted treatments, a passenger mutation or a mutation present in a minor population might become a driver and be selected. Examples include the T790M EGFR mutation present in lesions before erlotinib/gefitinib treatment and the KRAS mutations present in extremely low proportions in colorectal tumors that are eventually selected and cause resistance to cetuximab treatment.[30, 31] Therefore, the thorough characterization of mutations in cancer with a focus on the mutations involved in mechanisms of resistance will be highly relevant to make correct therapeutic decisions. In these cases, assessment of mutations in circulating-free DNA or circulating tumor cells in patients undergoing treatment holds promise to provide early detection of resistance-associated alleles.[32-35]
Impairing oncogene-addicted pathways with targeted treatments results in clinical response, although resistance eventually arises. In many cases, the original targeted pathway is reactivated in the resistant tumor population (eg, the T790M mutation in EGFR after gefitinib/erlotinib, RAS mutations after BRAF V600E inhibitors, restoration of BRCA competence after PARP inhibitors, MET amplification in NSCLC).[36-40] A better understanding of these processes will allow the design of treatment strategies that take into account frequently emergent mechanisms of resistance. In addition, the epigenetic landscape of a tumor cell can also determine the response to a treatment. The same mutation might have different therapeutic implications, depending on the epigenomic regulation of genes (and, hence, the state of differentiation of the tumor cell). This is a concept to be explored and considered when classifying tumors based on genetic characteristics.
Tumor cells have an intense interaction with nontumoral components of the host and the stroma, including endothelial cells, immune/inflammatory cells, and cells from the surrounding tissue. The differential presence of nontumoral cells in tumors might determine the biology and treatment of cancer. The characterization of the stroma should be considered when classifying tumors based on molecular characteristics.
The enormous increase in knowledge achieved during the last years in the field of cancer genomics has revolutionized our concept of cancer and provided several successful treatment options. The present strategies, however, are not achieving the expected efficacy in healing or chronicity, and only an innovative change in our vision of cancer will allow us to deal with the emerging knowledge of its complexity.
It is no longer sustainable to build a research-based clinical structure based on organ-classified cancer. Although there are some studies that enroll patients according to mutational events and not just based on tumor type, many of these are phase 1 or exploratory phase 2 studies. In addition, to the best of our knowledge, none of these trials have the objective of exploring activity in a combined matter. Moreover, each of the different cohorts enroll patients with a specific tumor type, and the go/no-go decision will be based on each of these groups. For this reason, these trials are considered by some investigators as real different phase 2 trials. For example, the National Institute for Health Research Cancer Research Network NCRN396 VE BASKET study is a phase 2 study of vemurafenib in patients with BRAF V600 mutation-positive cancers. However, patients with different tumor sites are being enrolled independently.
We need to be able to offer the same therapies to the same molecularly defined malignant cells irrespective of their primary location. In this way, we will be able to have a high enough number of patients with the same tumor characteristics to observe a desirable effect and improve the time to approval of new drugs that will benefit all patients in clinical. For example, Mano proposes that tumors carrying abnormal ALK as an essential growth driver should collectively be termed “ALKomas.”
However, our proposal raises an important technological challenge. Fast and easy ways to assess the molecular characteristics of tumors are required. Although some recent technological breakthroughs allow the fast characterization of tumors, we expect great news in this context in the near future.
We are proposing a novel way in which to study and deal with cancer treatment based on the finding that differences among tumors within the same organ are larger than among tumors with the same molecular aberrations. This novel approach will lead to a change in the therapy of cancer and might accelerate success in the fight against this deadly disease.