Precision Medicine in Cancer Diagnosis and Treatment
Precision or Personalized Medicine is today an integral part of modern cancer therapy. It does not involve tailoring treatment to a person’s specific genome sequence but rather tailoring treatment to mutations found in the patient’s tumor. Much of the buzz about what personalized medicine is based on speculations in areas of medicine where personalized medicine has not yet arrived. My purpose here will be to describe what precision medicine involves today in oncology, how publicly funded genomics research has gotten us this far, and what some of the key challenges for the future are.
Human Genome Project
The Human Genome Project (HGP) was a large initiative to determine the sequence of DNA that constitutes the human genome. The DNA is the program which describes how each protein is built Proteins do the work of cells and the protein molecules present in a cell determine the tissue “phenotype” of the cell and how the cell will interact with other cells. Proteins also interact with chemicals entering the cell and with the DNA itself, influencing further protein production.
During and subsequent to the completion of the HGP, new technologies for DNA sequencing were developed and the cost of sequencing has dropped dramatically. The cost of the HGP was about $2.7 billion in 1991 dollars. Today a human genome can be sequenced for about $1000 to $5000. This dramatic reduction in the cost of genome sequencing led to two new kinds of large genomics projects in the field of cancer.
Whole Genome Association Studies
The first type of new project is the whole genome association study (WGA)which aims to find inherited genetic changes associated with cancer. WGA’s involve genotyping the DNA in blood cells for a large number of patients with a particular kind of cancer and an equal control group. Genotyping involves determining the DNA code at through specific “polymorphic” genomic locations that vary among individuals.
Although many associations between polymorphisms and cancer incidence, usually the associations have been too weak to use for counseling individuals on their disease risk.1 These studies have also so far not led to personalizing treatment for patients with an inherited risk polymorphism.
Tumor Sequencing Studies
The other type of large genomic studies that followed the HGP are Tumor Sequencing The Tumor Cell Genomic Atlas (TCGA) project sponsored by the National Cancer Institute has attempted to sequence the DNA of at least 200 tumors of each of the major types of cancer.5 An international consortium has conducted similar studies.6 These studies have generally sequenced the DNA of the approximately 25,000 genes.. The portion of the genome consisting of genes is called the “exome”. The exome consists of less than 1% of the genome, the remainder consisting of regulatory regions and sequence of unknown function. With new methods of sequencing called next generation sequencing (NGS), whole exomes of tumors can be sequenced very economically. The exome contains the sequence of all the genes and the genes contain the DNA code for all of the human proteins.
Tumors can result from protein malformations caused by mutations in key genes. The development of a tumor is considered a multi-stage process with each stage being triggered by a mutation in a somatic cell. A somatic cell is a cell of a tissue in the body other than the egg cells or sperm cells. A mutation in a somatic cell such as a lung cell may provide a growth advantage to that cell. Somatic cells can divide and the mutations they contain are inherited by their daughter cells. Additional mutations can occur in a member of the descendant cells of those containing the original mutation providing additional growth advantage. .7
All Tumors are Unique
The tumor sequencing studies have been rich in new discoveries. It was found that in many cases, tumors of the same primary site (e.g. lung cancer) are heterogeneous with regard to the somatic mutations that they contain. However most of the mutations do not change the protein structure. Consequently, most of the differences among tumors are not functional or exploitable therapeutically.
As a tumor population expands from a single cell to about 1011 cells, the size needed for clinical diagnosis, it may undergo an enormous number of total cell divisions. For any position in the genome, the probability of a mutation for normal mammalian cells is approximately one mutation in 109 cell divisions. These mutations result from the thermodynamics of cell division and are what drive species evolution. In an expanding tumor population which may have undergone 1013 mutations by the time of clinical diagnosis, every location in the genome is likely to have undergone one or more mutations. These neutral mutations have been called “passenger mutations” because they are not driving the development or progression of the tumor. Many of the tumor sequencing studies have identified hundreds of mutations per tumor and one of the challenges is to distinguish the “driver” mutations from the “passenger” mutations.
Development of Drugs for Inhibiting Driver Mutations
Some genes have been found to be mutated in many tumors of the same primary site. For example, the BRAF gene is mutated in about 50% of melanoma tumors, the KRAS gene is mutated in about 55% of pancreatic cancer, and the P53 gene is mutated in over 50% of all cancers. Some types of cancer are characterized by the presence of a specific mutation. For example Chronic Myelogenous Leukemia is characterized by the presence of a translocation bringing together the ABL gene with the regulatory region of the BCR gene and Burkitt’s lymphoma is characterized by a mutation in the MYC gene.
There was some personalization of treatment based on biological characterization of the tumor even before the HGP. For example, breast tumors that expressed estrogen receptors on their surface were treated with anti-estrogen drugs such as Tamoxifen which block those receptors.. .. On the other hand, treating an estrogen receptor negative tumor with Tamoxifen has no effect. The selective effect of Tamoxifen was an early form of precision medicine
For breast cancer, prior to the HGP it was discovered that about 25% of the tumors have lots of HER2 protein receptors on the cell surface and that the patients with these HER2 positive tumors had somewhat shorter survival than other breast cancer patients. The over-expression of HER2 appears to be caused by amplification of the HER2 gene in the tumor genome. The development of an HER2 receptor antibody, Herceptin that blocked the activation of the protein, was a landmark in cancer therapeutics.8 This was unlike the earlier development in CML where all patients had a specific gene translocation.
Today breast cancer patients are classified with regard to their stage of disease (how far it has spread) and with regard to whether the tumor expresses estrogen receptors, progesterone receptors, HER2 or is “triple negative.” Hence actionable tumor genomics which influences treatment decisions is firmly established in current practice guidelines.
The tumor sequencing studies have played a prominent role in cancer drug development. As mentioned above, a single point mutation in the BRAF gene was found in about 50% of melanomas. BRAF is also a kinase gene and the mutation was at a location that put the protein in a permanently “on” position. That on switch stuck in the on position propagated a signal to activate cell division which drove tumor proliferation. Medicinal chemists know how to design kinase inhibitors, but in this case a young biotech company, Plexicon, used innovative structural methods to design an inhibitor with inhibitory effects that were relatively specific for the mutated form of the protein.10 All drug treatment involves a trade-off between killing the target cancer cells without killing essential normal cells. Antibiotics are very effective because bacterial cells are very different from our mammalian cells and we can use molecular drug targets which are crucial for bacterial proliferation but do not even exist in human cells. Hence, very high doses of antibiotics can be used without causing undue toxicity and the bacterial cells can be eliminated before resistant clones are selected for. Such selectivity rarely exists with cancer treatment; resistance is common because the tumor cannot be safely treated with a high enough dose. The Plexicon drug was able to be administered at a dose that shut-down oncogenic signaling because of its selectivity. Their drug was highly effective, a real breakthrough for melanoma (EGFR) which is considered one of the most drug unresponsive of tumors. Unfortunately, in treating patients with advanced metastatic disease, resistance often develops eventually and the drug, although prolonging survival, is not curative.
Other examples of precision medicine in oncology are the use of drugs to block mutated and constituatively activated epidermoid growth factor receptors in lung cancer and the determination that anti-EGFR antibodies were only effective in colorectal cancer if the KRAS protein was not mutated11.
As a final example, a translocation in the ALK gene was found in approximately 4% of patients with non-small cell lung cancer (NSCLC). ALK is another kinase and an effective inhibitor, Crezotinib, was developed to turn off the ALK switch, providing extended survival for these patients.12 It used to be conventional wisdom that pharmaceutical companies wanted only blockbuster treatments and hence would not be interested in treatments for small subsets of traditionally defined diseases.
Now however, these corporations appear increasingly interested in tapping into the relatively new, and alluring field of personalized treatment.
While precision has made its debut into the field of oncology, it is mostly based on genomics of the tumor, not the inherited genetics of the patient The new paradigm of drug development over the past decade has been to develop drugs with companion diagnostics to inhibit mutated oncogenes. This has led to an expansion of the number of diagnostic tests that patients should undergo in order to properly diagnose their disease. This increased complexity is causing problems for some payers in their decisions of what drugs and what tests to reimburse under what clinical circumstance.
The expansion of the number of diagnostic tests used for guiding drug selection has also created an industry for testing for somatic alterations in tumors with regard to a panel of actionable mutations. This is a highly regulated area as the DNA testing must be very accurate and because some genomic alterations in a gene may be actionable while others may not. The panels may involve sequencing some or all of up to several hundreds of genes.
In many cases the treatments that have been developed in this way are much more effective than previous chemotherapy, but in most cases they are not curative. It is hoped that by combining such molecularly targeted drugs in clever ways, cures can be achieved. The ability to design effective combination regimens rationally is currently limited by understanding of basic tumor biology.
Another challenge to precision oncology is the scarcity of treatments for many commonly occurring mutations. For example, the family of RAS proteins are frequently mutated in multiple types of cancer. These proteins are not kinases, however, and in spite of great investments by the pharmaceutical industry, effective blockage of constitutive RAS activation in tumors is not yet possible. Similarly, the P53, Rb and MYC transcription factors are inactivated by mutation in many tumors. P53 and Rb are important “tumor suppressors” but restoring activity of a tumor suppressor has proved much more difficult that inactivating a kinase stuck in the on state. This has been a persistent roadblock because it is too high risk for both academic investigators and industry.
Medical studies of new treatments is one of the few areas where we as a society have the benefit of good randomized experiments to determine what works and what does not.. With the finding that tumors of most primary sites are heterogeneous with regard to their driver mutations, however, a new generation of clinical trial designs has come into use.14 These clinical trials focus eligibility on more homogeneous subsets of patients who share a common driver mutation in the gene which is the molecular target of the test drug. This approach has resulted in an improved success rate and larger average treatment effects. For early phase I and II studies, new “basket” designs are being used in which the mutations in the patient’s tumor are matched against the targets of a panel of drugs and the one is tested which provides the greatest likelihood of benefit.15
Although genomic driven precision medicine is bringing about significant changes in the kinds of clinical trials that are being used to develop more effective treatments, the situation is far different from proposals that have been made to conduct therapeutic research based on standard medical practice. The proposal is usually to establish large databases of treatments and outcomes for patients treated in standard practice, to characterize the patients by genotyping them and to try to sort out what treatments work for what types of patients. Clearly this has severe limitations for cancer research. First it is tumor genomics, not germ-line genetics, which generally determines outcome. Secondly, the current generation of cancer drugs is not adequate for the kind of success we seek. New drugs and new drug combinations generally cannot be used in standard medical practice and so electronically collecting the standard practice experience will limit us to learning about how to use existing treatments. Finally, unless treatment effects are large, controlled trials have generally been needed to reliably compare treatment regimens.
The class of diseases called cancer has turned out to be more biologically complex and more difficult to treat effectively than previously imagined. The battle is being waged by scientists and clinical investigators in a mixture of types of organizations: academic investigators with public or charity support, and the pharmaceutical and biotech industries. There are many challenges, both scientifically, which I have elaborated upon, and politically. However, good policy choices require knowledge of the substantive issues by the policymakers, and I hope that this article can contribute to the accumulation of that knowledge.
ABOUT THE AUTHOR
Richard Simon is Chief of the Biometric Research Program at the National Cancer Institute and head of the Computational and Systems Oncology Branch. Dr. Simon holds a doctoral degree in applied mathematics and computer science from Washington University in St. Louis, Mo. He is a fellow of the American Statistical Association and a former member of the Oncologic Drug Advisory Committee of the FDA. He is the architect of BRB Array Tools software and author of Using Genomics in Clinical Trials and Predictive Medicine (Cambridge U. Press 2013). He is the recipient of the 2013 Karl Peace award of the American Statistical Association “for contributions that have played a pivotal role in bridging the gap among statistics, clinical research, and translational medicine to improve human health."
- McCarthy MI, Abecasis GR, Cardon LR, Goldstein DB, Little J, Ioannidis JPA, Hirschhorn JN. Genome-wide association studies for complex traits: consensus, uncertainty and challenges. Nature Reviews Genetics. 2008; 9:356-69.
- King MC, Marks JH, Mandell JB. Breast and ovarian cancer risks due to inherited mutations in BRCA1 and BRCA2. Science. 2003; 302:643-6.
- Li FP, Fraumeni JF, Mulvihill JJ, Blattner WA, Dreyfus MG, Tucker MA, Miller RW. A cancer family syndrome in twenty-four kindreds. Cancer Research. 1988; 48:5358-62.
- Knudson AG. Mutation and cancer: Statistical study of retinoblastoma. Proceedings of the National Academy of Sciences of the U.S.A. 1971; 68:820-23.
- Thomas RK, Baker AC, DeBiasi RM, et al. High-throughput oncogene mutation profiling in human cancer. Nature Genetics. 2007; 39:347-51.
- Stratton MR. Exploring the genomes of cancer cells: Progress and promise. Science. 2011; 331:1553-58.
- Hahn WC, Weinberg RA. Modeling the molecular circuitry of cancer. Nature Reviews Cancer. 2002; 2:331-41.
- Slamon D, Pegram M. Rationale for trastuzumab (Herceptin) in adjuvant breast cancer trials. Seminars in Oncology. 2001; 28:13-19.
- Druker BJ. ST1571 (Gleevec) as a paradigm for cancer therapy. Trends in Molecular Medicine. 2002; 8:S14-S18.
- Bollag G, Tsai J, Zhang J, Zhang C, Ibrahim P, Nolop K, Hirth P. Vemurafenib: the first drug approved for BRAF-mutant cancer. Nature Reviews Drug Discovery. 2012; 11:873-86.
- Lievre A, Bachet JB, Le Corre D, et al. KRAS mutation status is predictive of response to cetuximab therapy of colorectal cancer. Cancer Research. 2002; 66:3992-5.
- Shaw AT, Kim DW, Nakagawa K, et al. Crizotinib versus chemotherapy in advanced ALK-positive lung cancer. The New England Journal of Medicine. 2013; 368:2385-94.
- Pardoll DM. The blockade of immune checkpoints in cancer immunotherapy. Nature Reviews Cancer. 2012; 12: 252-64.
- Simon R. Genomic Clinical Trials and Predictive Medicine. Cambridge University Press, 2013.
- Simon R, Polley E. Clinical trials for precision oncology using next-generation sequencing. Personalized Medicine. 2013; 10:485-95.