The promise of personalized medicine in cancer will only be met once physicians have both the arsenal of therapeutics they need and a more complete understanding of genomic tumor profiles. At The Wistar Institute’s Melanoma Research Center, we are using Patient-Derived Xenograft (PDX) models to create laboratory avatars for individual melanoma patients. With these avatars, we can explore and test therapeutic schemes that we can then translate back into patient care.

One of our goals is to characterize an entire PDX bank, a database that one day will allow oncologists to select the individual drug or combination therapy based on biomarkers exhibited by patients’ tumors. PDX mice also allow us to test novel drugs in the tumor grafts of patients who have relapsed during treatment, and to screen for novel drug combinations. These three projects allow us to study melanoma as it is found in patients and focus on a personalized approach to the treatment of this highly heterogeneous disease. We expect that these studies will lead to novel biomarker signatures of response for subgroups of melanoma patients and improvements in therapeutic combinations.

To create an avatar, we must first collect a sample from a patient. We graft our tumor sample underneath the skin of a NOD scid gamma (NSG) mouse. PDX yields tumors that more faithfully recapitulate the natural growth of tumors in humans, and with a higher rate of spontaneous metastases and fewer artifacts than we see in cultured cell lines.

To create a set of avatars, we need to amplify the available amount of patient tissue. To do this, we remove the growing tumor from our avatar, dice it up, and implant the diced tissue into a set of new NSG mice. We can repeat this process multiple times to create an array of avatars for a study.

This is going on right now at Wistar, with melanoma patient samples from our clinical collaborators. Wistar enjoys a historic close relationship with the University of Pennsylvania’s Abramson Cancer Center in this regard, but we have recently also established partnerships with the Helen F. Graham Cancer Center at Christiana Care, which has provided some rare tumor subtypes; the Wills Eye Institute, which has provided ocular melanoma tissue; and with Massachusetts General Hospital and MD Anderson Cancer Center, both of which are leaders in melanoma treatment.

In one example, we received a lymph node metastasis biopsy from a patient just before he was starting therapy with a BRAF inhibitor for his previously untreated stage IV BRAFV600E melanoma. Unfortunately, the patient’s best response was stable disease and he progressed after only 4 months of therapy.

After successfully establishing the patient’s tumor as xenografts, we began treating the mice with the same therapy the patient had received and saw that the tumor, again, was resistant to the BRAF inhibitor alone. We then treated another cohort of mice with the combination of a BRAF and a MEK inhibitor, a combination that since then has been shown to be superior to BRAF inhibitor single-agent therapy in clinical trials. This time, we saw tumor regression in the mouse avatars of the patient (

Figuring out the molecular mechanisms behind these differing responses is the focus of ongoing investigation, with the goal of informing future treatment decisions. Although the information derived from this experiment could not be used for the benefit of this particular patient (he never received this combination therapy), we hope to streamline the process in the future to make this kind of personalized medicine possible in melanoma.

PDX models are powerful tools for predicting a patient’s response to therapies. In time, we hope to establish a database cataloging cancer biomarkers and drug responses to upwards of 1000 PDX avatars, which will allow us to work with all subgroups of the disease and draw statistically relevant conclusions regarding their phenotypes and responses to therapy.

Of course, it can take almost a year to grow the amount of tissue we need to establish a large cohort of avatars, which limits the utility of using PDX avatars to determine individual patient therapy. On the scale of weeks, however, such avatars could help oncologists rule out proposed therapies or help choose the better of two options. The potential of creating patient avatars is immense, yet underexplored in the clinic.

		Figure. Tumor growth curve of a BRAFV600E melanoma PDX derived from a treatmentnaïve patient.

		Each treatment group consisted of 10 mice; error bars are standard error of the mean. The tumors were implanted into mice after several rounds of serial implantation to expand the original biopsy. Once tumors were established, it took 20 days to reach the maximal volume in the control group (dark blue line). When treated with a BRAF inhibitor (light blue line) no treatment effect could be observed. Interestingly, when the BRAF inhibitor was combined with a MEK inhibitor, tumors regressed in all animals (yellow line).

Articles

The promise of personalized medicine in cancer will only be met once physicians have both the arsenal of therapeutics they need and a more complete understanding of genomic tumor profiles. 

Clemens Krepler, MD

Articles

Patient-Derived Xenograft (PDX) Models as Avatars for Patient Response to Therapy