Developing anti-cancer drugs has always depended on testing new drug candidates in mice. Unfortunately, only about 5% of drug candidates that look promising in mice ever prove to be successful in humans. This emphasizes the need to develop better mouse models of cancer—ones that more accurately reflect the properties of human cancers, so that the effects of drugs can be predicted more successfully.

According to Garth Powis, D.Phil, director of SBP’s NCI-designated Cancer Center, “Mouse models with tumors derived from cancer cell lines are largely responsible for the cancer drugs that we currently have. However, the tumors produced in these models are genetically too simple to adequately mimic the complexity of human tumors, which spells trouble for useful drug testing.”

In trying to solve this problem, Powis recently co-authored a report in Molecular Cancer Therapeutics that evaluated the feasibility of a patient-derived xenograft (PDX) approach, in which surgical samples of human colorectal cancers were implanted to initiate tumors (i.e. xenografted) in mice. Powis explains that, “Our thinking was that these patient-derived tumor samples would have all the genetic complexity of the original human tumors, and could therefore be more accurate predictors of the effects of experimental drugs to treat cancer.”

The report establishes several reasons for optimism about PDX-based studies.

• First, human tumors can be successfully established in mice using samples from both primary tumors and tumor metastases. Since tumors may change properties when they metastasize to other organs, it’s important to test drugs on both types of tumors.
• Second, specimens from patients undergoing cancer therapy can still produce tumors in mice. This is important because cancer therapy may change the properties of a tumor, including its sensitivity to the new drugs.
• Third, for successful tumor initiation in mice, tumor samples used 24 hours after surgery were as effective as samples used 2 hours after surgery. This is important to researchers who may have limited access to tumor specimens or who lack immediate access to surgically removed tumors.

Powis emphasizes that, “These results show that PDX models can be established from a wider range of tumors than we initially expected, which really expands our horizons for future drug testing. In addition, PDX strategies are in line with the goals of personalized medicine, since discovering biomarkers in susceptible tumors will allow us to predict which patients are most likely to respond to a drug therapy.