The predictive assay puzzle

March 2006
Feature Story

William Check, PhD

It seems like last month’s Winter Olympics were designed to emphasize the inherent riskiness of predicting great things for high performers. So many American Olympians didn’t live up to their promise. In oncology, predicting long-term accomplishment based on early results may also be risky. Initial success with targeted therapies such as trastuzumab for breast cancer and gefitinib and erlotinib for non-small-cell lung cancer, along with assays to select patients who are most likely to benefit from these agents, has given rise to optimistic forecasts for the new discipline of theranostics—the use of molecular tests to predict response to specific inhibitors in solid tumors. How sound are these expectations? Will theranostics prove to be a flop, or will it produce multiple medals?

To explore the future of theranostics, molecular pathologist Karin Berg, MD, MS, organized a session on this topic at the meeting last November of the Association for Molecular Pathology. "We decided to put theranostics on the program because of the success of a few drugs in the marketplace right now that are targeted and based on molecular assays," Dr. Berg told CAP TODAY. "There have been a few successes, and obviously there are a ton of drugs in the pipeline." The aim of the session was to take a closer look at one of the successes and to get an overview of where the field is now.

As an example of a successful application of theranostics, Dr. Berg chose the use of assays to identify patients with non-small-cell lung cancer who have mutations in the gene for epidermal growth factor receptor, or EGFR. Such tumors are more likely to respond to gefitinib or erlotinib. Dr. Berg called this application "a great success," and said several groups are looking at both mutations and gene amplification to predict drug response. "These assays show a way to confine a relatively low-toxicity therapy to those who will benefit and to get other people on another therapy in a more timely fashion," she says. Targeting therapy with this assay could save the health care system considerable money, because specific inhibitors are, in her words, "extraordinarily expensive."

To discuss assays for EGFR mutations, Dr. Berg invited Neal Lindeman, MD, associate pathologist for clinical chemistry and molecular diagnostics in the Department of Pathology at Brigham and Women’s Hospital and instructor of pathology at Harvard Medical School. For the general overview, she selected Kathleen M. Murphy, PhD, director of the molecular diagnostics laboratory at the Johns Hopkins Medical Institutions.

In Dr. Lindeman’s view, the EGFR/non-small-cell lung cancer situation is not a special case. "It would not shock me at all if other tumors had this kind of association with EGFR or other similar molecules," he says. To Dr. Lindeman a more practical question in molecular pathology now is which assay to use for EGFR. "The big question for us is whether we should continue to do complete DNA sequence analysis [for the EGFR gene] or to make more of a shortcut assay to look only for known mutations," he says. "That is my biggest dilemma."

Dr. Murphy is a bit more reserved in her evaluation of theranostics. At the end of her AMP presentation she concluded, "We may be in the era of targeted molecular therapy, but it remains unclear what role molecular assays will play in predicting sensitivity or resistance to these drugs."

"Individualized medicine is so talked about and hyped these days," Dr. Murphy told CAP TODAY, "but we are just at the beginning of accumulating data on this topic. It is too early to tell whether the EGFR story will prove to be the model or an exception." Her hope is that it will be a model, but she says solid tumors are genetically complex. Even in leukemias, targeted therapy "has not been exactly a home run," she notes, due partly to the emergence of second-line mutations that confer resistance to imatinib. "It is not going to be a home run in solid tumors either," she says. "We face years of collecting data and figuring out what it means. Right now we don’t know how big an impact theranostics will have."

Dr. Berg predicts that the non-small-cell lung cancer story "will be played out in other disease states." She foresees a situation in solid tumors in which a targeted therapy does relatively well, then the disease overcomes the drug by resistance mechanisms. "It will be very much a moving target," she says. Resistance will mandate combination treatment. Dr. Berg envisions many cancers that benefit from multi-agent targeted therapy, like conventional combination therapy but with drugs that are more targeted and less toxic. However, she cautions that there are some data that molecular testing doesn’t always predict how a drug will affect a cancer. "In some cases, even if a patient has a specific abnormality in the phosphatidylinositol 3-kinase [PI3K] pathway, that doesn’t predict that an inhibitor of that kinase will be effective," she says.

For pathologists, the uncertainties these experts raise pose important questions about whether and when to bring predictive assays into the laboratory, and which ones to adopt.

The EGFR story started with a clincal observation, Dr. Lindeman says. "Some patients—about 10 percent—had dramatic responses to gefitinib even though statistically the drug did not pan out." Gefitinib was found to regress non-small-cell lung cancers in which the EGFR gene is mutated so that it is constitutively active, just as trastuzumab is effective against breast cancers in which the HER2 gene is always turned on (Lynch TJ, et al. N Engl J Med. 2004;350:2129-2139). In the case of EGFR, the mutations alter the active site of the enzyme—EGFR is a transmembrane tyrosine kinase—so that it becomes overactive even in the absence of its ligand. "Probably other oncogenes are activated as well," Dr. Lindeman says. "So that doesn’t say EGFR is the dominant growth factor driving oncogenic transformation. But it is an important player and interrupting its function is important."

It is becoming standard to use the tyrosine kinase inhibitor erlotinib (which is largely replacing gefitinib) only in non-small-cell lung cancer patients whose tumors carry EGFR mutations. The current assay for EGFR mutations is bidirectional sequencing of the entire EGFR gene. However, two mutations account for 90+ percent of all found mutations. "We may someday initially screen for those two mutations and do the broad mutation assay only if the clinical scenario doesn’t fit the result," Dr. Lindeman says. His argument for limited mutation analysis is that it is cheaper, faster, and more sensitive than complete sequencing. On the other hand, looking for only two mutations makes it impossible to find mutations that are not yet known. "In the past year we found two new mutations we had not known about, both associated with resistance," Dr. Lindeman says. "I believe that someone should offer full sequencing." But perhaps not all laboratories need to do so.

An alternative to both full-sequence analysis and focused mutation assays is denatured high-performance liquid chromatography, or DHPLC. In this method, instead of looking for individual mutations, one screens for whether any EGFR mutation is present. In DHPLC, DNA from the EGFR gene is denatured so that the individual strands are separate. If only normal cells are present, when the DNA is cooled the strands will reanneal to their original partners, forming normal duplexes. But a mixture of normal cells carrying no point mutation and tumor cells bearing a point mutation will form heteroduplexes, which separate on HPLC. Heteroduplexes can be isolated from the chromatography column and analyzed for mutations. "With this technique we can screen out all normals," Dr. Lindeman says. "If we find a mutation is present, we can go back and try to identify it."

Another advantage of DHPLC is that it is suitable for mixed samples. "Many biopsy samples we get are heterogeneous," Dr. Lindeman says. "For sequencing, I spend my time manually dissecting these slides to enrich the population for tumor cells, since you need at least half tumor cell DNA or you will have trouble finding mutations. So I have to dissect with a razor and a steady hand." With DHPLC this is not necessary. In his talk, Dr. Lindeman showed Transgenomics’ Wave procedure for DHPLC of the EGFR gene. On that slide he had this footnote: "I have no conflict to report, other than that I am torn between wanting to continue to offer this test and not wanting to spend the rest of my life scraping tissue off glass slides."

Dr. Murphy agrees it may turn out to be preferable to use a scanning technique such as heteroduplex analysis and then do limited sequencing for positive samples. "Many techniques are cheaper and have higher throughput than sequencing," she says. "But they are not 100 percent sensitive." On the other hand, sequencing does not achieve its potential high sensitivity if the sample is too heterogeneous.

Dr. Lindeman cited other aspects of the debate over methodology. A group of investigators at the University of Colorado Health Sciences Center, Denver, has published data on gene amplification suggesting that increased EGFR gene copy number assessed by fluorescence in situ hybridization may also predict response to erlotinib (Hirsch FR, et al. J Clin Oncol. 2005;23:6838-6845). To make things even more complicated, a group at the University of Toronto has reported that it was unable to verify the predictive value of either sequencing or FISH (Tsao MS, et al. N Engl J Med. 2005;353:133-144). The authors wrote, "Among patients with non-small-cell lung cancer who receive erlotinib, the presence of an EGFR mutation may increase responsiveness to the agent, but it is not indicative of a survival benefit."

Methodological differences may explain the discrepant findings of the Toronto group. For instance, they used laser capture microdissection and found many mutations never before described. "We haven’t seen that," Dr. Lindeman says. "Are they selecting very small subclones within tumors that may have secondary mutations at a very low frequency?" he wonders. Another possible explanation is artifact of PCR, which can occur in small samples prepared by laser capture microdissection. "Because of these possibilities, particularly when a novel mutation is detected, I think it’s important to confirm all positive results before making a clinical recommendation," he says. All positives are confirmed in Dr. Lindeman’s laboratory. Dr. Berg reinforces the importance to pathologists of these methodological questions. "For molecular anatomic pathologists, doing appropriate testing to predict therapeutic response is very complicated," she says. The debate about method "likely represents a major question for the molecular laboratory," she says. "Will we test for amplification by FISH or mutation by a molecular technique? Or both? We don’t yet understand the physiology well enough to know how we should be testing." She adds: "Will we be able to get reimbursed for both tests?"

Dr. Berg confirms the need for a pathologist to be involved in molecular anatomic pathology because of the makeup of tumors. "There is a tremendous amount of non-tumor tissue in any piece of tumor—inflammatory cells, stromal cells. These bystanders can contaminate what you are trying to test. So pathologists should be involved in this process to sample appropriately."

In her overview of theranostics, Dr. Murphy noted that targeted therapy of solid tumors has lagged behind hematopoietic tumors because chromosomal and mutational analysis tends to be more complex in solid tumors. That makes it more difficult to sort out which genetic changes are important and need to be targeted for therapy. She calls the HER2 breast cancer story "the poster child for a solid tumor assay that helps predict whether a therapeutic will be effective." Yet even after years of accumulated data in that context, people still argue about whether to measure HER2 with FISH or immunohistochemistry. "It takes a long time to figure out what these assays are telling us," Dr. Murphy says.

The same will be true for solid tumors: As we discover new targets in solid tumors, it will take a while to sort out whether to assay at the level of protein, DNA, or RNA. "Which will be most predictive?" Dr. Murphy asks. "And what exactly are these assays telling us about the patient?" Even with established kinase inhibitors in solid tumors, it is not clear what predicts whether a patient will respond. "Some of the new angiogenesis inhibitors are effective, but they are hitting more targets than we thought," she says. "That makes it difficult to define from a theranostic viewpoint."

In patients who respond to specific small-molecular inhibitors, secondary resistance mutations often develop. "Do we need to test for those?" Dr. Murphy asks. "Molecular diagnostic assays may end up playing a role here." An intriguing observation is that secondary resistance mutations occur in analogous amino acid positions in different kinases in different cancers, such as EGFR in non-small-cell lung cancer, BCR-ABL in chronic myeloid leukemia, and KIT in gastrointestinal stromal tumors. Dr. Murphy believes that this implies a role for molecular diagnostics in helping to determine second-line therapy.

Another complication is that not all secondary mutations cause resistance to newer inhibitors now under development (Debiec-Rychter M, et al. Gastroenterology. 2005;128:270-279). "Many people are now thinking of two to four kinase inhibitors given as a cocktail," Dr. Murphy says, "analogous to antiretroviral drug cocktails, to prevent resistance from arising. From a clinical laboratory standpoint, you have to decide where you are going to spend time and money in developing assays that will continue to be useful six months to a year down the road." If combinations of kinase inhibitors become routine therapy, perhaps qualifying assays will become irrelevant.

The future of qualifying assays is complicated by the fact that specific inhibitors are being developed to a number of oncogenic targets besides receptor tyrosine kinases. "Receptor kinases make nice targets because they sit on the cell surface," Dr. Murphy says. "So you can use a small-molecular inhibitor or a monoclonal antibody against them." However, companies are also looking at molecules downstream of receptor tyrosine kinases that are activated by the kinases’ signals for growth and proliferation. It is appealing to inhibit downstream pathways because such inhibitors may be effective in cancers that have mutations in different receptor tyrosine kinases—HER2 or EGFR or KIT. "They all signal through the same pathway, so a downstream inhibitor would have broader applicability," Dr. Murphy says. A particularly promising target is PI3K, which is mutated in several cancers (Samuels Y, et al. Cell Cycle. 2004;3:1221-1224).

"One problem with all these new drugs that inhibit parts of this pathway is that we don’t have good assays to understand what we should be measuring," Dr. Murphy says. "How do we know they are hitting their targets?" Her laboratory is developing assays that can be used in clinical trials of new inhibitors.

In addition to developing assays for new inhibitors, Dr. Murphy is interested in assays that predict sensitivity/resistance to standard anti-cancer drugs. "Many standard therapeutics were put into use before we knew about genetics and mutations in tumors," she says. "We don’t know why they work or why only some patients respond." Research showed recently that ovarian and pancreatic cancers are susceptible to crosslinking agents such as cisplatin when they bear mutations in genes in the Fanconi anemia pathway, which is basically a DNA repair pathway (van der Heijden MS, et al. Clin Cancer Res. 2005;11:7508-7515). "This finding has biological plausibility," Dr. Murphy says. If a tumor has a DNA repair defect and you give an agent that causes DNA strand breaks, the tumor can’t repair them and will be sensitive. Clinical trials are needed to address these issues. One trial that is being planned at Johns Hopkins will test pancreatic cancer patients for mutations in the BRCA2 gene, which is a member of the Fanconi anemia pathway, to determine whether to use cisplatin.

"Choosing which assays to do is really a struggle," Dr. Murphy says. "Molecular laboratories have always had to deal with clinicians calling and demanding the latest and greatest reported assay. Then a few months later no one cares about that mutation any longer and they are interested in the newest mutation." One of the things she’s done with the molecular laboratory at Johns Hopkins to keep it "afloat" and to avoid developing assays that become outdated is to work with oncologists doing assays in drug trials and to ensure there is a payment mechanism through grant money so they know they have a guaranteed volume for at least a couple of years. "I don’t think you can do development work outside that context," she says.

Merging the ideal with the real in theranostic testing is a problem that Dr. Berg, too, has thought about.

"People are always saying that anatomic molecular pathology is the future of pathology," she says. "There are some real tests, but most are in generation right now and are not yet ready for prime time."

How does Dr. Berg advise dealing with the conflict between the promise of the future and the financial realities of the present? "When I talk to residents," she says, "I tell them to keep pushing the glass."

Related Links

• Two investigational inhibitors of PI3K (PDF, 960K)