Frederick (Fritz) Eilber is a cancer surgeon and researcher at the University of California, Los Angeles, where he specializes in soft tissue sarcoma (STS). Treating these frequently fast-growing malignancies is often a race against time, he says, as doctors try one drug after another to find what might work. Patients often endure months of toxicity without any benefits.
In 2016, Eilber and his colleagues started testing a different strategy. Their approach was to surgically implant fragments of a patient’s tumour into immunocompromised mice, which could then be used as personalized avatars for drug efficacy testing. By testing multiple therapies in groups of animals simultaneously, they reasoned, doctors could rapidly identify effective treatments while ruling out likely failures.
These patient-derived xenografts (PDX) already have a history in translational research – investigators use them as tools for understanding tumour biology in preclinical studies.1 Usually the tumour fragments are implanted under the skin, but Eilber’s team implant them orthotopically, meaning into corresponding sites in the animals’ bodies. Published results2 show that aggressive STS tumours implanted in this way sprout vasculature, metastasize and “grow quickly enough to be clinically relevant to patients”, Eilber says.
With new cancer diagnoses hovering at 20 million a year worldwide, interest in PDX models is reviving. But not all these models are created equal – researchers have found that for some cancers, subcutaneous models fail to spread or to imitate human treatment responses.3 This is where orthotopic-PDX (O-PDX) models offer an advantage.4 By mimicking the cancer’s natural microenvironment, they give rise to tumour behaviours and responses that more closely predict clinical outcomes, says Noah Federman, a paediatric oncologist at UCLA. “The idea is that you allow the tumour to grow in an animal under very similar conditions to how it would grow in the body,” Federman says. “There’s an important benefit to putting tumours where they belong: they get exposed to cancer-specific growth factors that are not found in subcutaneous settings.”
O-PDX models also address a key limitation with current definitions of precision oncology, which focuses on trying to match treatments with a tumour’s genetic profile. Cancer cells can possess dozens or even hundreds of mutations,5 and tests that identify genomic alterations do not delineate between the genuine ‘drivers’ that promote tumour growth and incidental ‘passenger’ mutations that do not. Just because a patient has a tumour-associated mutation, there is no guarantee that a treatment directed against it will work.5 Targeted therapies might fail for many reasons that cannot be predicted by either next-generation sequencing (such as the chaotic mutational environment surrounding the tumour) or studies that expose a patient’s cancer cells to drugs in vitro. “You can kill anything in a petri dish,” says Eilber. “But with O-PDX models, you get a lot more insight into whether a drug will actually be effective.”
Typical O-PDX MRI results
O-PDX in practice
The process of evaluating therapeutic options entails what Eilber describes as a “mini clinical trial”. First, isolated tumour fragments are implanted into parts of the animals’ bodies where they are likely to grow fastest.
“Once we establish actively growing tumour in the first generation of mice, we engraft the tissue orthotopically to create the number of avatars needed for the patient’s study,” says Jonathan Nakashima, chief scientist at Certis Oncology, a San Diego, CA-based provider of O-PDX pharmacology testing services.
Groups of five mouse avatars each are then treated with one of the therapeutic options being considered for the patient. Researchers assess tumour response using magnetic resonance imaging (MRI) and digital tumour volume measurement. A control group of five tumour-engrafted animals receives no treatment (see 'Typical O-PDX MRI results').
“The results guide our choice of treatment regimens,” says Federman. Doctors and patients alike are better prepared to answer the question ‘what next?’ should a first-line treatment fail (see 'Typical O-PDX results graph'). “Finding out when a tumour is resistant to therapy is also very important,” Federman adds. “Then you can avoid giving drugs that will only cause toxicity.”
According to preliminary Certis data, O-PDX models and human responses are “highly concordant”, especially for treatment failures, says Nakashima.
Typical O-PDX results graph
Eilber, who co-founded Certis and is now a scientific adviser to the company, says that the best candidates for O-PDX testing are patients with more-aggressive (high-grade) tumours — meaning tumours that are invading nearby tissues and/or metastasizing. Less aggressive (low-grade) cancers that have minimal risk of spreading will not engraft well in mice. O-PDX tests are particularly helpful for patients with metastatic solid tumours with limited treatment options. Doctors can test drugs they might not ordinarily turn to when treating such patients. “I’ve seen them work, and it can be a real surprise,” Eilber says.
O-PDX models do have some limitations. One is that from start to finish it can take four to six months to generate useful results. Another is cost. The mice used lack immune systems that would otherwise reject implanted tumours and so must be handled with extra care and housed in sterile environments. To reduce costs, Certis has a two-step payment process, and charges for pharmacology testing only if tumours engraft successfully.
Despite these issues, O-PDX models are a potentially valuable expansion of the precision oncology toolbox. Patients with recurrent, drug-resistant, metastatic or rare cancers can benefit, or where no effective standards of care exist. What really matters to these patients is time. “When they waste months on treatments that ultimately don’t work, it can be devastating to the patient and treating oncologist,” says Eilber.