NEW YORK (GenomeWeb) − A team of researchers led by the Spanish National Cancer Research Center, CNIO, in Madrid has been exploring whether a combination of tumor exome sequencing with drug testing in patient-derived xenografts can lead to personalized cancer treatments that improve the outcome of patients with advanced cancer.
Following a proof-of-concept study that was presented at the American Society of Clinical Oncology meeting last year and recently published in Clinical Cancer Research, the researchers, led by CNIO Director Manuel Hidalgo, are currently recruiting patients for a follow-on trial that will compare the new approach to the standard of care.
Tumor profiling by next-gen sequencing, in particular gene panels, is increasingly used to help decide on the best treatment strategy for advanced cancer patients, but "the reality is that all these diagnostic precision medicine tests are being used with really no trial to show that they benefit anybody," Hidalgo said, which he hopes the new trial will help answer. "We assume that having [the mutational] information is useful but we don't think it's that clear."
His team's approach involves exome sequencing of a patient's tumor and matched control to identify genomic alterations that could be targeted by drugs. In parallel, they engraft the tumor into mice to generate patient-derived xenografts, or avatar mouse models, in order to test treatments suggested by the tumor's mutational profile prior to giving them to the patient. "The hypothesis is that those patients where we were able to do all that will have a better outcome" than patients receiving the standard of care, Hidalgo said.
The research team conducting the proof-of-concept study included members of Champions Oncology, a New Jersey-based company Hidalgo co-founded that provides patient-derived xenograft mice as a service, and Personal Genome Diagnostics, a Johns Hopkins University spinout focusing on cancer genome analysis. For the study, conducted over four years, the researchers, sequenced the exomes of 25 advanced cancer patients with a variety of primary tumor types, including colorectal cancer, glioblastoma, pancreatic cancer, non-small cell lung cancer, and melanoma.
Patients' tumors or metastases were sequenced around the time of diagnosis, and they received standard-of-care treatment while the sequencing and avatar testing proceeded.
When they developed resistance and their disease progressed, the experimental results were used to guide further treatment of some of the patients. "Of course it would be better to do [the analysis] at the time of progression, because then the information is closer to the patient's current situation, but we need time to do this," Hidalgo explained.
Twenty-three of the exome analyses were successful. From the mutations and copy number variations, the researchers manually selected potentially actionable mutations that could be targeted with drugs. For 20 patients, they found up to five putative drug targets, while for three, they did not identify any, though all 23 tumors had cancer-relevant somatic mutations.
Finding several targets in a single patient is both a blessing and a curse, according to Hidalgo: "It's good because you have things to work with but it's bad because you don't know which one to begin with," he said.
For testing several treatment options, the avatar mouse models came into play. Avatar mice were generated successfully for 10 patients out of 14 for which they were attempted − the remaining patients either refused or there were other technical issues. In seven of the avatars, the researchers tested various drug regimens, in part based on the exome results. The process of generating the mice and testing the drugs took around six months, at a cost of about $15,000 to $20,000 per patient, Hidalgo said.
Some treatments to be tested in the avatars are obvious, Hidalgo said – for example, an amplification of the ERBB2 gene, also known as HER2, suggests trying Herceptin. Others are less straightforward: for example, a tumor with a mutation in the same pathway as ERBB2 may or may not respond to Herceptin. "Those are the ones where the avatars are very helpful," he said.
But even if the exome results strongly point to a targeted treatment, this does not mean it will work in the avatar, or in the patient. In one case, for example, the researchers found a mutation in the catalytic domain of the PI3K kinase, but a PI3K inhibitor did not work in the avatar model, while a drug cocktail that combined PI3K and MEK inhibitors did.
In cases where exome sequencing did not yield any drug targets, the avatar models were still helpful in selecting drug therapies. In a patient with small cell lung cancer, for example, no druggable mutations were found but based on tests of various drugs in the avatar, the patient received two consecutive treatments that initially had favorable outcomes.
However, experimental results are only one factor that determines what treatment a patient eventually receives. For example, Hidalgo said, if a drug is not approved for a patient's cancer type, it may be difficult to get insurance to pay for it, or the patient may not be eligible for a clinical trial of a new drug that the avatar results suggest might work.
A total of 13 patients received at least one treatment that was based on either their genomic profile or data from the avatar model, or both. Of these, nine patients had either a partial response or stable disease.
Thirteen treatments were directed by avatar results, and the treatment response in the mice mimicked the patient response for 11 of these.
While the study illustrates that patients' tumors can be analyzed and potential treatments be tested in avatar mice in a timeframe that can help guide their treatment when they develop resistance to standard drugs, "the sample size remains at the time too small and heterogeneous to conclude if this approach will be better than the standard-of-care approach to select therapy," the authors noted.
"This is a nice demonstration of where the field is going," said Matthew Goetz, a professor of oncology at the Mayo Clinic, who with his colleague Judy Boughey has been conducting a research study of high-risk early-stage breast cancer patients that also involves exome sequencing and avatar mouse models but does not currently use the information to guide the treatment of those patients.
While the Spanish study is limited by its size and the heterogeneity of cancer types, it demonstrates that exome sequencing can identify drug targets, and that avatar mice are useful for validating those findings. "Overall, I think this is a great paper," he said.
Others disagreed. "If done correctly in the future, this may be a useful approach," said Andrea Califano, a professor of chemical and systems biology at Columbia University who has been conducting "N-of-1" trials that involve tumor sequencing and treatment testing in patient-derived xenograft models. However, "this is a descriptive study that is at best anecdotal and tells us very little about whether exome sequencing and PDXs are doing something for the patient or not," he told Clinical Sequencing News.
In their randomized follow-up trial, which started about a month ago and has enrolled a handful of patients, Hidalgo and his colleagues plan to answer that question. The trial will compare 50 patients receiving the standard of care to 100 patients who will have their tumor exomes sequenced and avatar models tested to guide their treatment. The trial is expected to be completed in two to three years, he said.
In the meantime, several technical challenges remain that hold back the use of exome sequencing and avatar mice to help personalize cancer treatment. One is the time required to make the patient-derived xenografts, Hidalgo said, the other the availability of tumor tissue suitable for sequencing and generating avatar mouse models. His team is pursuing an approach to generate avatars from circulating tumor cells, a project that he said is at an early stage of research.