Personalizing Cancer

From total remission to recurrence or no reaction at all, every possible response to treatment – including a wide variety of adverse reactions – can be expected.

The implications of this wide array of possible outcomes are immense and everybody involved – the patient, their families, the physician, the pharmaceutical companies developing drugs, and the insurance companies paying for the treatment – is negatively impacted by this uncertainty.

This differentiates cancer from many other diseases, where once a diagnosis is made a treatment can readily be devised and the outcome is mostly predictable. Cancer seems to defy the laws of causality with its apparently random response to drugs that are the results of many years of scientific and clinical research. Or does it?

The difficulties of cancer treatment start with the diagnosis since early detection of cancer in most cases significantly improves the prognosis but is hampered by the lack of suitable screening tests. Therapy selection and prognosis provide additional challenges, as seemingly identical tumor types can be caused by different mutations affecting different pathways and reacting differently to individual therapies. Biomarkers that flag the differences in the individual cancer genomes have become the holy grail of cancer research. Several different classes of biomarkers are of potential value. These include genetic markers (for example, mutations and SNPs), epigenetic markers (the status of chromatin modifications or DNA methylation), expression markers (levels of gene expression) and protein markers1.

The Pursuit of Biomarkers

One of the many drivers that led to the initiation of the Human Genome Project (HGP), which entailed a massive sequencing undertaking, was the promise of personalized medicine.

Personalized medicine is indeed a strange buzz word. Many would argue that Hippocrates was the first advocate of personalized medicine – and that physicians ever since practice medicine by tailoring their therapies to the individual patient. Many of us are probably fortunate enough to know physicians who are practicing this very successfully. However, the advancement of molecular biology and genetics holds the promise that the tools and knowledge base that physicians use to judge the efficacy of treatment choices is getting vastly expanded.

While this is all good news, the transition into the clinic seems slower than most of us would wish for. Still, in the last few years, progress has been made towards truly personalized treatment decisions of potent drugs. A great example is Herceptin (trastuzumab), used as an adjuvant treatment of HER2+ node positive breast cancer. Herceptin is a monoclonal antibody that blocks a growth factor receptor responsible for tumor growth. Blocking the associated pathway leads to significant reduction of recurrence. Nevertheless, only those patients whose tumor displays an overexpression of the growth factor receptor are susceptible to the treatment. Fortunately, the assessment of this molecular marker is fairly straightforward. Tests that directly measure the protein on the cell surface or indirect genetic tests that measure the receptor gene copy number through FISH provide a reliable assessment of the HER2 status. Through the use of these testing methods the approximately 30 percent of patients who can benefit from the treatment can be selected.

This model could provide a case study for the development of potent drugs targeted towards a specific subgroup of patients which are defined by the molecular status of one or multiple biomarkers. Despite the nature of this example, single nucleotide polymorphisms (SNPs) and insertion/deletion mutations provide an important class of biomarkers, especially for potential drug resistance effects. The occurrence of those mutations can lead to non-response to treatment or later relapses after treatment or enable treatment.

Today, there seems to be a level of acceptance for the fact that there may be no silver bullet for many cancer types, but that rather different targeted drugs need to be employed for treatment, depending on the molecular status of the patient’s tumor. One implication of this is the need for accurate patient stratification in clinical trials as the accurate targeting of subgroups, and understanding of limitations becomes of paramount importance for the success of cancer drug development.

Recent Progress

Earlier this year, a group of leading scientists from world renowned institutions published an important breakthrough in the quest for developing tools to generate a snapshot of the genetic profile of tumors. The multi-center study led by scientists from the Dana-Farber Cancer Institute was published in this year’s February issue of Nature Genetics.

The group profiled a large number of mutations (238) in 17 oncogenes over more than 1000 human samples of 17 different tumor lineages. The oncogenes in this study included prominent genes like EGFR, HRAS, ERBB2, KIT and BRAF. Mutations in these genes are known to affect the response to chemotherapy. For example, mutations in EGFR can lead to resistance to gefitinib treatment, and mutations in KIT can lead to resistance to imatinib.

The underlying assumption of this work is that a detailed knowledge of prevalent somatic mutations in the main cancer lineages will advance the development of targeted cancer therapeutics and aid in treatment decisions.

The authors point out that our success in discovering genetic alterations in cancer now has clearly shifted the bottleneck towards the translation of this knowledge into direct benefit to the patient.

In their experience, at least one of the difficulties associated with this task seems to be technical in nature. Most initiatives in the discovery of mutations in cancer cells have relied heavily on sequencing technologies. However, routine analysis of clinical specimens by conventional sequencing may not be the ideal tool due to insensitivity and cost of this method.

To test this hypothesis, the authors compare results between conventional Sanger sequencing and two different genotyping technologies: Sequenom’s MassARRAY and Pyrosequencing.

Sequencing obviously has a distinct advantage for the discovery of de novo mutations and when random mutations are expected in a genomic sequence. A significant disadvantage of conventional sequencing is a rather low sensitivity for heterozygous alleles, especially when they fall below a 50 percent distribution. Also, genotyping methods like MassARRAY allow for a multiplexed analysis of several different, independent mutations, even if they are far apart or on different chromosomes; a cost and workflow advantage that conventional sequencing can not deliver.

As the authors observe, between 16-44 individual mutations in RAS, EGFR and BRAF cover approximately 99 percent of the mutations thus far observed in human tumors, so that genotyping becomes an efficient tool to assay them. Thus, in order to overcome the limitations of sequencing, the authors employed genotyping techniques for their study, reasoning that these methods provide the means to robustly and sensitively detect multiple mutations at acceptable cost.

Clinical specimens are also often difficult to analyze with sequencing as malignant and non-malignant cell are often mixed, which is further complicated by copy number and ploidy changes in tumor cell genomes. Genotyping, especially when high-sensitive methods such as the MALDI-TOF detection based MassARRAY are employed, has clear sensitivity advantages over conventional sequencing. A greatly reduced background allows for the detection of alleles even at a level of 10 percent in a mixture.

When comparing the results between sequencing and genotyping for EGFR mutations in 22 primary lung tumor samples both genotyping methods detected 12 mutations, whereas sequencing only found nine. The three missing mutations were low frequency events at 16, 12 and 9 percent. This further validates the superior sensitivity of genotyping for mutation detection in mixed samples.

One illustrative example from this study is the identification of KIT mutations in gastro-intestinal stromal tumors (GIST). One GIST specimen was found to carry two KIT mutations including a mutation that previously had been demonstrated to be associated with imatinib resistance. The individual from which this sample was derived had a tumor that relapsed after imatinib treatment, suggesting a predictive value of the assay for treatment response.

Overall, the study revealed at least one of the queried mutations in 30 percent of the analyzed samples; most of them either heterozygous or admixed with stromal DNA. Often, oncogene mutations that activate the same downstream pathways are found to appear mutually exclusive in tumor samples. This study, however, also revealed several co-occurring mutations that had not previously been described. Further experimental data suggest that several of these co-occurring mutations have complementary rather than redundant effects on tumorigenesis.

The Path Forward

One important application of this study might be the use for patient stratification in clinical trials to facilitate the development of highly targeted, thus more efficient drugs and improved predictability of treatment. For all involved this would present a major improvement in the treatment of cancer. The availability of highly accurate, sensitive and flexible solutions for genotyping and mutation analysis has the potential to greatly enhance the progress in these programs. Specifically, flexibility is an important and sometimes overlooked aspect. As research progresses and new biomarkers are discovered, the ability to quickly develop, change and adjust existing panels becomes an important aspect by itself and seems to be at the very core of personalized medicine.

On the emerging field of epigenetics and its role in cancer and common disease see Sequenom’s online article ‘Lamarckism Revisited’ at http://www.ngpharma.com/pastissue/article.asp?art=270871&issue=212

Today, there seems to be a level of acceptance for the fact that there may be no silver bullet for many cancer types, but that rather different targeted drugs need to be employed for treatment, depending on the molecular status of the patient’s tumor