In a study of 1,110 patient-derived xenografts (PDX) from 24 cancer types, researchers found that the genomic landscape of these tools changes dramatically over time. Some of the genetic changes observed have been associated with drug sensitivity in previous studies, raising questions about the use of PDXs to predict drug response in personalized medicine.

Patient-derived xenografts (PDX) are generally considered to be an accurate experimental method for determining how human tumor cells respond to treatment. However, in a recent study, researchers found that their genomic landscapes can change dramatically over time, suggesting that PDXs may be a less reliable way to gauge the response of tumor cells to drugs than previously thought (Nat Genet 2017;49:1567–75).

The researchers analyzed 1,110 PDX samples from 24 cancer types and found that by the end of the fourth passage, 88% of PDXs had acquired at least one large chromosomal aberration. Also, a median of 12% of the genome was affected by copy-number alterations. These genomic changes differed from those observed in evolving tumors in patients, and some of the mutations involved have been associated with drug sensitivity in previous studies.

“PDXs can rapidly become quite different from the specific tumors from which they were derived,” says Uri Ben-David, PhD, of the Broad Institute in Cambridge, MA, and first author of the study.

Juliet Williams, PhD, an executive director of Oncology Drug Discovery at Novartis, who was not involved in the study, says that this is the latest in a series of papers showing that PDXs exhibit genomic changes as they are passaged. In light of this body of work, “avatar experiments, in which a PDX is generated from a patient for personalized medicine, could very well be a misleading approach, especially if a tumor is heterogeneous and only a few clones contribute to the PDX.”

Although these results raise questions about using PDXs to determine the best therapy for a particular patient, Ben-David believes these models will remain a valuable tool in cancer research. “Our findings are consistent with previous studies that suggest PDXs can be useful for large-scale studies aimed at identifying genotype–phenotype associations.” However, because of the costs and labor required to produce PDXs, he says it may be worthwhile for scientists to consider whether it makes more sense to generate multiple cell lines from a given tumor rather than a single PDX, depending on the biological question at hand.

To aid such decisions, Williams would like to see a rigorous evaluation of how faithfully PDXs recapitulate primary tumors, relative to newly created cell lines. “The new wave of cell-line generation may represent patient populations as well as PDXs, unlike historically available cohorts of cell lines, but that has yet to be comprehensively understood,” she notes.

In addition, “a key outstanding question is what drives the observed genomic evolution in PDXs, and whether it can be attenuated,” says Ben-David. If selection pressures imposed by the mouse immune system or the microenvironment of the tumor site are primarily responsible, for example, then generating PDXs in humanized mice or performing orthotopic transplantation of tumors may reduce PDX divergence from parent tumors.

Ben-David adds that it could be fruitful to study genomic events that are highly recurrent in human tumors but tend to disappear in PDXs. “Understanding what makes them disappear,” he says, “could help us understand what underlies their recurrence in the first place, and thus how these events could be targeted therapeutically.”

Finally, Williams would like to see more robust data on the functional consequences of the genomic changes in PDXs to better determine how likely they are to affect drug discovery. –Kristin Harper

For more news on cancer research, visit Cancer Discovery online at http://cancerdiscovery.aacrjournals.org/content/early/by/section.