Skip to main content
Premium Trial:

Request an Annual Quote

Personalized Tumor Models Could Help Identify Combination Therapies for Hard-to-Treat Cancers

Premium

NEW YORK (GenomeWeb) – While there have been many advances in using next-generation sequencing to analyze cancer patients' tumors and identify treatment options, for pancreatic and some other cancers, this approach has been largely unsuccessful. Researchers are now developing different strategies for such cancers that still attempt to bring the promise of personalized medicine to patients, but in a way that does not rely only on matching specific genetic alterations with targeted therapies.

As described in a study published today in Cell Reports, patient-derived tumor models can instead enable researchers to both analyze the tumors' genetic make-up as well as to screen therapies and identify combinations that would not have been obvious by looking at the genome alone.

"No matter how comprehensively we define a tumor's genetics, it is still sometimes hard to predict how to treat it effectively," Erik Knudsen, a senior author of the study and associate director of basic research at the University of Arizona Cancer Center, told GenomeWeb.

In the study, Knudsen and his colleagues focused on pancreatic cancer, a notoriously difficult-to-treat cancer that has a poor prognosis even when diagnosed early, with a five-year survival rate of 27 percent for early stage and just 2 percent for late stage, according to the American Cancer Society.

Knudsen said that Agnieszka Witkiewicz, lead author of the study, became interested in the challenges of treating pancreatic cancer. Both he and Witkiewicz had previously studied breast cancer, "where precision medicine mostly works," he said. "In contrast, when we started working on pancreatic cancer, there was no sign that that would happen." The sequencing studies they performed, even of large numbers of cases, identified very few recurrently mutated genes. And, the genes mutated in most patients — KRAS and TP53 — are not drug targets. "At the end of the day, looking at the list of genes, it was very hard to understand how you could treat a patient" based on the mutations alone, he said.

The two decided to focus their efforts on building models of patients' tumors by establishing cell lines and growing tumors in mice.

In the Cell Reports study, the researchers sequenced the exomes of 28 pancreatic ductal adenocarcinoma tumors from patients who had consented to the genetic studies and development of the models. An analysis of the mutations identified by exome sequencing found "only a handful" of druggable targets, the authors wrote, and "many of the potentially actionable alleles represented variants of unknown significance."

Next, the researchers created cell lines and xenograft models from the patients' tumor material, and also used exome sequencing to characterize those models. The team confirmed that the mutational landscape was consistent across the primary tumor tissue, cell line model, and patient-derived xenograft. They then used the genetic data to inform treatment decisions in the cell line models.

Three cell line models had variations in genes involved in chromatin remodeling, a pathway thought to be altered in about 20 percent of pancreatic cancer cases and sensitive to EZH2 inhibition. However, when they treated the cell lines with those inhibitors, they found that the one that was most sensitive to the drug was wild type for chromatin remodeling genes. The three cell lines whose pathways were altered displayed "intermediate" sensitivity, "underscor[ing] the challenge of targeting specific genetic events occurring in complex cancer genomes and the need to functionally assess therapeutic sensitivities," the authors wrote.

However, while in some cases, the genetics could point directly to a specific drug, most tumors did not have any actionable alterations. Because of this, the researchers designed a drug sensitivity screen of 305 different agents, including compounds currently in clinical use and those in late stage development. They ran the drug screen on the cell line models, narrowing down the list to 76 candidate drugs that showed some activity in at least one of the cell lines.

Knudsen said that despite testing so many drugs, there was not one agent that was broadly effective in the majority of cell lines. "I thought we'd find one drug that worked well," he said, "but that didn't happen."

However, there were a few isolated cases of exception response to certain drugs. For instance, one cell line was very sensitive to MEK inhibition, which has so far had limited success in clinical trials in unselected pancreatic cancer patients. In the study, though, the researchers used RNA sequencing to show that this cell line had an "enrichment of genes involved in cell adhesion, epithelial versus mesenchymal differentiation, and KRAS dependence signature," the authors wrote, which was not present in the cell lines that did not respond to the MEK inhibitor. Knudsen noted that such findings could be used to guide future clinical trials of compounds.

But, he said, the most valuable aspect of the study was the ability to test combinations of drugs very rapidly and to evaluate how the drugs interacted with each other and with the targets in the tumors.

"We could see low-dose synergistic drug interactions that were [specific] to patient tumors," Knudsen said. "That changes the way we think about clinical trials."

In addition, many of the combinations would not have been predicted by analyzing the tumor genome alone.

After evaluating the drugs and their combinations in cell lines, the researchers tested promising candidates in the xenograft models. The MEK inhibitor in particular was especially promising in the cell lines when used in combination with several different drugs. When they tested it in combination with docetaxel in the mouse models, the drug combo suppressed tumor growth for more than 30 days in two, even though each drug alone could not prevent progression. Further analysis found that the MEK inhibitor "limited proliferation" while docetaxel "elicited cell death."

The findings on drug combinations could be especially useful, Knudsen noted, because making decisions about patient treatment becomes "exponentially more complicated when considering combination drugs." While cancer researchers have a good handle on how an MEK inhibitor works as a single agent, he said, "there is a lot to learn about dual pathways" when it is combined with another drug. "These models can be a good interim for addressing that," he said.

Knudsen said he will continue to conduct these types of studies in larger numbers of patient samples to better understand how different drug combinations interact and to search for different types of genetic signatures that can predict response to therapy.

The cell line/xenograft models could also be used in real time to stratify patients into specific clinical trials. In the study, the researchers were able to develop the models, screen cell lines, and select drugs and combinations to test in the mouse models within 10 to 12 months. Patients who first have surgery to remove their tumor and are treated with the standard of care have a median time-to-recurrence of 14 months, so 10 to 12 months could be quick enough to inform the next treatment option in patients who recur. Knudsen said he is looking to set up a clinical trial that would test the ability of xenograft models to guide therapy.

Aside from pancreatic cancer, he said, this type of model to choose therapies could be applicable to other hard-to-treat cancers, like ovarian cancer. A group at the Mayo Clinic, for example, started a clinical trial in 2014 using xenograft models of ovarian cancer patients' tumors to predict response to therapy.

Researchers at the Spanish National Cancer Research Center in Madrid have also tested whether tumor exome sequencing combined with drug testing in patient-derived xenografts can lead to treatments that improve patients' outcomes. In addition, a group at the Icahn School of Medicine at Mount Sinai has launched a mouse avatar study for patients with triple-negative breast cancer. However, the goal of the Mt. Sinai study is to evaluate how closely the characteristics of the mouse tumor match the patient's, rather than to use cell line and xenograft models to choose a personalized treatment plan.