The laboratory of Thea Tlsty, Ph.D., professor at the University of California, San Francisco, studies the regulation of cancer initiation in human tissues. At the conference, she will be discussing the use of biomarkers to evaluate the risk of future breast tumor formation in women diagnosed with ductal carcinoma in situ (DCIS).
Fifteen to thirty percent of women with DCIS go on to develop tumors within ten years after a lumpectomy. Unfortunately, many patients end up being treated too aggressively (needless mastectomies) or not aggressively enough.
Nuclear grade, surgical margins, and tumor size are all measured in the clinic, but none have been found to have strong enough predictive value to affect therapeutic decisions.
Interestingly, Dr. Tlsty found that cells expressing high levels of p16 and/or COX-2, when coupled with proliferation, go on to become basal-like invasive tumors. These particular biomarkers indicate an abrogated response to cellular stress; cells overexpressing them that continue to proliferate have bypassed pRb-mediated signals to senesce.
In contrast, cells with high p16 and/or COX-2, but low proliferation, have an intact Rb checkpoint and senescent program and do not go on to become tumorigenic. These biomarkers can be measured years before tumors actually arise, and thus can be used clinically to help dictate individualized treatment options.
Richard Bender, M.D., the chief medical director of Agendia, will be talking about the company’s development of genomic biomarkers for prognosis and prediction in early-stage breast cancer. Agendia is marketing MammaPrint, which measures the expression profiles of 70 different genes in early-stage breast cancers.
MammaPrint was designed to specifically predict which 30% of patients are at a high risk of recurrence, which it does remarkably well, according to Dr. Bender; it has a hazard ratio of 10, whereas most prognostics hover around 3–4. As a side benefit, these are the same 30% that are sensitive to chemotherapy. Dr. Bender notes that this type of gene-expression profiling is a “way of personalizing treatment for patients.”
Agendia also has TargetPrint, which quantitates ER, PR, and HER2 expression in breast cancer. In contrast to MammaPrint, this test was designed to determine which drugs should be given to which patients.
Drug Response and Resistance
The human xenograft tumor model has been the standard preclinical model for testing the efficacy of new cancer drugs. However, many promising drugs, including the latest antibodies and kinase inhibitors that were so effective in these models, have failed in the clinic.
Murray Robinson, Ph.D., svp of oncology at AVEO Pharmaceuticals, notes that his whole company is predicated on the premise that we “can do better figuring out which patients are going to respond to which drugs. The biggest problem in oncology development is that we have little understanding of who will respond to a drug and how to test it.”
AVEO has developed the Human Response Prediction Platform to tackle this. It genetically engineers various tumor-initiating mutations (for example HER2 and EGFR) into human cells, then introduces them into mice and allows secondary mutations to arise spontaneously. These tumors end up with 100 different causative mutations, rather than just the two or three seen in xenografts, and, thus, more closely mimic the range of genetic heterogeneity observed in human tumors.
This population-based approach should help specify the important causative mutations that account for the variability seen in drug responses. Each tumor is extensively characterized by microarray analysis, so researchers can correlate sensitivity or resistance to drugs with expression information and other changes in biomarkers.
This is, in fact, what the company has done with AV-951, its highly potent triple VEGFR inhibitor. The molecule is in Phase II trials for treatment of metastatic renal cell carcinoma because all xenograft models responded to it. Less than half of patients did, however. Using its HRP platform, AVEO scientists correlated a panel of 33 genes with lack of response—and found all of them are involved in a very specific biology, non-VEGF driven angiogenesis. “The hope and promise,” Dr. Robinson comments, “is that using more accurate models will help us gain insight into responsive populations and promote cheaper and faster drug approval.”