Being able to accurately predict how a given cancer will respond to chemotherapy would spare patients with non-responsive tumors the burden of undergoing toxic and ultimately unhelpful treatment.
Just as important, knowing which of a patient's cancer-causing genetic lesions are contributing to drug resistance might help doctors redesign therapy for maximum benefit.
Researchers led by Professor Scott Lowe, Ph.D., of Cold Spring Harbor Laboratory (CSHL), have come closer to achieving these critical goals for human cancer therapy by developing new mouse models for human acute myeloid leukemia (AML), an aggressive and devastating cancer of white blood cells.
As Lowe and colleagues report in the April 1st issue of the journal Genes and Development , their models, which include treatment protocols that closely mimic current AML therapy in people, precisely recapitulate genetic associations that have been linked to favorable or adverse treatment responses in patients. These findings provide compelling evidence for the notion that such models can predict how human cancers will respond to therapy and help to identify genes promoting resistance or sensitivity to any cancer drug. The mouse models, the CSHL team notes, also serve as an effective test system for new drugs and treatment strategies in AML.
Need for a reliable prediction model
"By giving us a better understanding of how a cancer's genetic makeup, or 'genotype,' influences the outcomes of treatment, these models might help improve how existing drugs are used in people, and spur the design of more effective therapies," Lowe observes. "The new models are an important preclinical tool that will allow knowledge gained from cancer genetics to be put to effective use in the clinic."
The Lowe team's mouse models also provide valuable insights into how leukemia develops and progresses. Their study traces the intracellular network controlled by the p53 gene, a linchpin of the cell's anti-tumor defense response, as the key determinant of AML's aggressiveness and response to therapy.
Most patients with AML receive the same standardized treatment – an initial phase of intense chemotherapy followed by additional chemotherapy cycles or bone marrow transplantation. Yet only a quarter of patients are cured and most die within a few months. This diversity in treatment response is due to AML's genetic heterogeneity, meaning that the hundred or so mutations associated with this form of cancer occur in different combinations in each patient and influence therapeutic outcomes in different ways.
Some gene mutations in cancer have been correlated with clinical outcome. But AML has proved to be genetically too complex, and the current experimental systems for predicting treatment response too unreliable for the information to be used in a standard way in the clinic.
These standard systems, in which anticancer drugs were tested in human cancer-cell lines, do not factor in the effects of a real tumor's environment on its growth. At the same time, placing these cells into animals to create a so-called "xenotransplant" model has not been an effective solution, according to Johannes Zuber, M.D., a Clinical Fellow in the Lowe lab who played a key role in the research and is first author on the team's paper. The reason, explains Zuber, is that "the cells are so poorly defined at the genetic level and tend to defy analysis of the molecular factors that influence drug response."
Making mice with human-like AML
Lowe's group surmounted these problems by first identifying the most commonly occurring mutations in a group of 111 children with AML and then engineering these mutations into mice, which soon developed leukemia. Among the participating AML patients, the two most common mutations were observed to occur when chromosomes broke apart and reattached in new places. This incorrect fusion is thought to create cancer-causing genes, or "oncogenes," that encode so-called fusion proteins. It's the proteins that are the most immediate "cause" of cancer. They profoundly disturb the developmental program of cells.