Precision cancer medicine entails treating patients based upon the molecular characteristics of their tumor. One could argue that we have been tailoring therapeutic regimens based upon tumor characteristics for years, whether it be treating patients based upon disease subtypes determined by histology or assessing markers by immunohistochemistry. However, precision medicine implies that therapeutics are directed precisely toward the identified molecular defect.
Mixed Success
The first example of precision medicine in oncology is the use of imatinib to treat chronic myeloid leukemia (CML) with the BCR-ABL translocation. Other examples have followed, but they have had mixed success, and only a few of them have impacted the clinical management of patients.
With the announcement of a precision medicine initiative by the National Institutes of Health, there will be renewed vigor in this endeavor. However, in attempting to replicate the model of imatinib in CML, it is important to understand the biologic context behind its great success, specifically the impact clonal evolution has on precision cancer medicine.
Imatinib: Targeting the Root of the CML Evolutionary Tree
In 1976, Peter C. Nowell, MD, proposed a model of clonal evolution in Science magazine in which he suggested “tumor progression results from acquired genetic variability within the original clone allowing sequential selection of more aggressive sublines.” As we approach the 40th anniversary of this theory and stand upon the doorstep of being able to therapeutically target acquired genetic alterations in a variety of diseases, this concept of clonal evolution becomes increasingly important.
The theory itself has been validated in countless studies showing that the sequential acquisition of mutations results in gains in evolutionary fitness. Furthermore, tumor initiator clones (also often referred to as cancer stem cells) have been identified in a subset of cancers and highlight the potential for a genetically “simple” tumor cell progenitor to propagate disease relapse. There is perhaps no disease with greater evidence of this than CML.
The initiating event in CML is acquisition of the t(9;22)(q34;q11) translocation, which creates a fusion between the BCR and ABL1 genes. This alteration can be found in hematopoietic stem cells and traced through the lineage to CML cancer cells. Secondary genetic alterations, such as mutations of TP53, RB1, and CDKN2A, can be acquired after the BCR-ABL translocation and may play a role in progression of CML from an early chronic phase to a more aggressive blast phase.
The introduction of imatinib, a small molecule inhibitor of ABL family kinases including the BCR-ABL fusion gene, revolutionized the way that CML is managed and dramatically improved outcomes for these patients. But why was imatinib so successful when therapies targeting other important “driver mutations” in other diseases have failed to produce similar results? An important factor contributing to the success of imatinib is that it targets the initiating event in the clonal evolution of CML. This means that all daughter cells that evolve following this initial event (ie, every cell in the clonal pool) also carry the BCR-ABL translocation and are susceptible to the effects of imatinib. The efficacy of imatinib is confounded by stromal interactions that protect the tumor-initiating clone and prevent clearance of the disease. However, the fact that this event is acquired early in clonal evolution and is in every evolved CML clone makes this disease manageable for extended periods, until an escapee mutant inhibits or circumvents the drug’s activity.
Is Every ‘Actionable’ Mutation Worth Acting On?
The terms “driver mutation” and “passenger mutation” were coined to discriminate between (1) those mutations that play an active role in disease pathogenesis (ie, driver mutations) and (2) those mutations that do not contribute to disease pathogenesis but undergo clonal expansion alongside one that does (ie, passenger mutations). However, the model for clonal evolution implies that not all driver mutations are created equal but rather are acquired in an ordered hierarchy. That is, some driver mutations occur as early events during clonal evolution and play a role in disease genesis (early drivers), whereas others occur as later events during clonal evolution and play a role in disease progression (late drivers/accelerators). Early driver mutations that have a role in disease genesis, such as the BCR-ABL translocation, will therefore be present in every tumor cell, whereas late driver mutations may only be present within a subset of tumor cells (ie, in a subclone).
To further complicate understanding and measurement of the clonal origin of mutations, each driver mutation will confer a variable boost in evolutionary fitness, which will cause them to overtake less-fit clones at different rates. This means that some driver mutations, despite occurring as late events in disease evolution, may appear to be present in the majority of tumor cells because they provide a significant boost to clonal fitness, and the subclone has therefore expanded prior to the time of sampling. However, in many cases, late driver mutations can be identified by variant allele frequencies (the representation of mutant alleles as a proportion of all sequencing reads over that nucleotide), which indicate they are subclonal events.
Currently, the majority of cancer subtypes lack a repertoire of “actionable” mutations that are matched to targeted therapies for oncologists to routinely make treatment decisions based upon tumor genotype. However, the perceived future of precision medicine is one in which patients will receive a report of “actionable” mutations matched to suggested therapeutics or clinical trials, which their oncologist can use in the formulation of their treatment strategy.
The question will then be is every “actionable” mutation worth acting on? In addition to the questions of efficacy and toxicities that are asked of all therapies, the clonal origins of each mutation and the desired outcome of the treatment will be important factors in this decision for “actionable” mutations. This is because early mutations are more likely to reside with a tumor progenitor cell compartment that propagates relapse, whereas late mutations may not. Assuming that these therapies eliminate every cell that carries the targeted mutation, the targeting of an early driver mutation that is clonally dominant may therefore provide the best chance of cure. In contrast, targeting a late driver mutation that has expanded to become clonally dominant may clear a large number of tumor cells and induce remission but may not eradicate tumor-initiating cells that propagate a relapse.
This situation may result in a clinical management strategy akin to a game of subclone whack-a-mole, but it could potentially extend patient survival. However, targeting a subclonal mutation may have little or no effect on tumor burden and could waste valuable time. Understanding the hierarchical order in which somatic mutations are acquired in each disease will therefore be an important consideration in ranking therapeutic targets for drug development, but this is a complex scientific undertaking. At the very least, the reporting of mutations in molecular genetic tests should incorporate a measure of clonal representation (allelic frequency) for each mutation, so that oncologists can be more informed about the biology of the tumor they are treating.
Concluding Remarks
Clonal evolution and clonal dominance will be important considerations in precision medicine as an increasing repertoire of mutations becomes “actionable.” However, it is necessary to discriminate between the importance of subclonality with regard to precision medicine and prognostication.
A mutation that is associated with an inferior outcome but is present only in a subclone may still maintain its prognostic value. This is because investigators often do not discriminate between clonal and subclonal events when identifying associations with prognosis. In addition, adverse mutations residing within a subclone are likely to be associated with inferior outcome because of their role in therapeutic resistance and/or increased clonal fitness, which are functions that can facilitate clonal expansion and disease progression regardless of the initial size of the subclone within the total tumor cell pool.
Accurate assessment of the clonal representation of mutations will require a method for estimating the total tumor content of samples that are submitted for sequencing, so the variant allele frequency can be normalized to the estimated fraction of tumor cells present within the sample. This can be achieved by pathology review of sections prior to DNA extraction, but this approach is cumbersome and therefore unlikely to become standard practice. New approaches are therefore required for automated tumor content assessment, but, until the time that such technology is available, variant allele frequencies should nonetheless be reported and used in the interpretation of sequencing results. Furthermore, the credential of “early driver mutations” should be used to prioritize therapeutic targets for drug development and “actionable” mutations for clinical decision-making. ■
Disclosure: Dr. Green reported no potential conflicts of interest.
Dr. Green is Associate Professor at Eppley Institute for Research in Cancer and Allied Diseases at the University of Nebraska Medical Center in Omaha.
Disclaimer: This commentary represents the views of the author and may not necessarily reflect the views of ASCO.
