HOW CAN GENOME SEQUENCING ADVANCE CANCER TREATMENT?
Because every tumor is different and has its own genetic makeup, it is ideally suited for personalized medical management based on genomic information. Although not yet a routine part of care, when genomic analysis is performed, the tumor genome is typically sequenced at more than 80-fold coverage, that is, every base is sequenced 80 times on average (Figure 1); many researchers recommend 200 fold or even higher coverage. Normal DNA isolated from blood or saliva of the same patient also is often sequenced. The sequencing might be of the whole genome, the exome, or a large set of genes implicated in cancer (e.g., Foundation Medicine sequences a panel of 315 genes*). The latter approaches (exome sequencing; gene panel sequencing) are not as ideal for detecting structural variants, but because they allow deeper sequencing, they may detect variants present in only a subset of the tumor cells. Variants may be present in only a portion of tumor cells because tumors are heterogeneous. Although tumors arise from a single abnormal cell, as the cells in the tumor rapidly divide it is common for new mutations to arise leading to subpopulations of tumor cells with distinct genomic profiles. New mutations accumulate because tumor cells often carry mutations that adversely affect the controls on genome integrity. Deeper sequencing also may provide more sensitive detection of variants in cases where a great deal of normal tissue is present in the tumor sample.
Figure 1. Normal tissue (typically cells from blood) and cancer tissue are sequenced to 30-fold and greater than 80-fold average coverage, respectively. The genetic differences between the samples are compared to find somatic changes (somatic mutations) in the cancer cells. Courtesy of The Cancer Genome Atlas, National Cancer Institute and National Human Genome Research Institute.
Somatic mutations unique to the tumor cells are revealed by identification of variants (i.e., changes) in the tumor cells compared with the normal DNA. The biggest challenge is to identify which variants are most likely to be “driver mutations,” that is, mutations that actively contribute to the growth of the tumor. This task can be especially daunting when whole genome sequencing is performed on tumor cells from advanced cancers because there can be many thousands of somatic mutations compared with the normal DNA from the same person. Indeed, some types of cancer typically have tens of thousands of variants! One typically looks for mutations in known proto-oncogenes or tumor suppressor genes—of highest interest are those mutations that affect targets of FDA-approved drugs or drugs in clinical trials, that is, druggable targets. Examples include EGFR and PDGFR; these are inhibited by erlotinib and sunitinib, respectively. As an example, we sequenced the genome of a metastatic colon cancer patient and found increased copies of the EGFR gene (Figure 2). This patient was treated with an EGFR inhibitor as part of his therapy. Examples of the many targeted drug treatments often used for non-small-cell lung cancer are shown in Table 1.
Figure 2. Sequencing of a patient with metastatic colon cancer revealed amplification of the EGFR and CDK6 regions. This patient was treated with an inhibitor of the EGFR signalling pathway. (Hanlee Ji and Michael Snyder)
Table 1. Examples of Drugs Used to Guide Treatment of Non-small-Cell Lung Cancer Based on Genetic Changes. (Reference: National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines in Oncology.)
Genetic Variations Targeted Therapy Agents
EGFR mutations Gefitinib, erlotinib, afatinib
HER2 mutations Trastuzumab, afatinib
BRAF mutations Vemurafenib, dabrafenib
MET amplification Crizotinib
RET rearrangements Cabozantinib
In addition to finding somatic changes in a person’s cancer, usually by sequencing a person’s normal DNA, the presence of common germline mutations that increase the risk of cancer is also investigated. Although these germline mutations are usually not “druggable,” the information may be useful for determining prognosis and is valuable for family members of the patient in determining their own cancer risk. Indeed, there have now been a number of instances where sequencing the normal and cancer DNA of one family member has revealed a germline mutation that has alerted other family members of their own increased cancer risk.
Genomic analysis might reveal the presence of a potential driver mutation that is not targeted by standard treatment for a given type of cancer. It may be that a drug targeting the effects of that driver mutation is commercially available, but only approved by the U.S. Food and Drug Administration (FDA) for another form of cancer. Or, it may be that a drug targeting the effects of that driver mutation is not commercially available, but is accessible to patients through a clinical trial. The oncologist might use the information from the genomic analysis to select a nonstandard, personalized treatment for the patient. In the scenario described here, the oncologist might use the evidence of the driver mutation to treat the patient with an off-label targeted therapy. “Off-label” refers to the use of drugs outside of their FDA-approved prescribing information. Physicians have discretion to prescribe off-label, but should have a scientific rationale for doing so. As an example, a drug that targets mutant forms of EGFR and that is approved only for the treatment of lung cancer may be prescribed by an oncologist for colon cancer if testing reveals that the colon cancer cells carry a mutated EGFR and the clinician feels that the patient would be an appropriate candidate for the drug. Physicians’ liability risk may be greater when prescribing an off-label use, depending on how strong the scientific evidence is to support that use. Insurance companies and other third party payers (e.g., Medicare) might require documentation from the physician of the scientific rationale before they will agree to pay for off-label use. In an alternative resolution to the scenario described here, the oncologist might use the evidence of the driver mutation to recommend the patient enroll in a clinical trial in which the drug under investigation targets the effects of the driver mutation.
The potential for genomic analysis to help guide personalized treatment for cancer is generating considerable excitement. The hope is that drugs will no longer be used against cancer in situations where they have an extremely low probability of success, and, more importantly, targeting drivers of uncontrolled growth present in a given tumor in a timely manner will be highly effective with reduced side effects. It is important, however, to note that most of these therapies are usually targeted at late-stage cancer patients.
Most patients follow the normal course of treatment that has been well established, and many of the drugs used are new and have adverse side effects. As drugs become more specific and their efficacy better known, however, it is expected that these targeted approaches and therapies will be administered to earlier stage patients and become more commonplace.
Another area of active research is the use of genomic and gene expression data to determine when to start anticancer treatment. All anticancer treatments carry a risk of side effects. For the patient (and the physician), it would be extremely useful to know how aggressive the cancer they are dealing with is so that they can make a well-informed decision about whether to start treatment immediately or delay treatment and just keep a close eye on tumor growth (“active surveillance”). This information would be especially useful in cases where the cancer is quite likely not to be aggressive, whereas treatments may have serious, life-altering side effects. For example, early stage prostate cancer often progresses very slowly over the course of many years, and some treatments may cause incontinence or impotence. Thus distinguishing aggressive versus nonaggressive prostate cancer would be valuable for knowing when and what type of treatment to use for that cancer.