Precision Oncology

Precision Oncology at Penn: Combining 'Omics Data with Functional Assays to Improve Care

While initial cancer treatments were one size fits all, it has become clear that not only are each of us individuals but so are our tumors. Stratification of tumors both by clinical assays and molecular sequencing has revolutionized cancer treatments over the last few decades. Deep molecular profiling of the mutational landscape of tumors has revealed new targets for therapeutic interventions. However, it has also become increasingly clear that defining this landscape is inadequate for most patients as every tumor carries a mosaic of genetic and epigenetic alterations with largely unpredictable effects on phenotype. Predicting which drug will work for any given patient based on genomic information alone is not straightforward.

While there are clear examples of the success of genome-guided therapy including imatinib for BCR-ABL and crizotinib for ALK translocations, these are not applicable to most cancer patients. Indeed, there are no FDA-approved therapeutics that act on MYC, TP53 or RAS family mutations which are the most commonly mutated cancer genes. And simply harboring a mutation in a gene does not equate to responsiveness as only subsets of patients harboring the same mutation respond suggesting that additional alleles in the tumor act as modifiers of responsiveness. There are additional difficulties in implementing sequencing-guided precision approaches since there are few targeted therapeutics available. Functional precision medicine has the potential to be a powerful ally to current genomic approaches. The unifying principle is that we cannot accurately predict sensitivities, and there are approved drugs that target diverse biologies, thus exposure to a larger panel of drugs will reveal new sensitivities and biomarkers that coupled with genotype and other clinical data will inform new treatments.

Acute leukemias are malignant neoplastic diseases that are characterized by the proliferation of immature, non-functional cells in the bone marrow that are subsequently released into the bloodstream. In AML, malignant myeloid precursor cells impair hematopoiesis and, in spite of clinical efforts, the overall 5-year survival rate remains around 30%. On average, patient tumors contain clear genetic mutations in less than 5 of ~70 identified recurrently mutated genes; however, no single gene is mutated in more than 30% of all patients and many mutations are loss-of-function in transcription factors and epigenetic modifiers making predictions on drug sensitivities difficult. Moreover, few of these lesions are actionable, and even for those that are, trial have shown that targeted therapeutics are only active in a subset of patients that harbor the mutant allele. Therefore, much work remains to be done to uncover the mechanistic basis of disease and identify markers or determinants of effective therapy. In addition, our incomplete knowledge of the mutational landscape coupled with our incomplete understanding of drug sensitivities limits our predictive capabilities for the identification of drugs that are active against individual patient tumors in AML.

Since the leukemic cells are readily accessible, in partnership with Penn Center for Precision Medicine, we created a new Program for Chemogenomic Discovery to fill this gap. We developed a functional precision medicine pipeline in AML to experimentally determine active drugs against individual patients’ tumor cells. To accomplish these goals we utilized the UPENN HTS Core, and patient cells from the leukemia tumor bank with Dr. Martin Carroll. We screened ~50 AML patients for their sensitivities to a panel of ~3000 ‘actionable’ drugs (~1100 FDA approved, ~1100 clinical trials) and discovered that no two patients displayed the same sensitivity, but that there were drugs that are active in subsets of patients suggesting a path to the clinic.

Working with Dr. Selina Luger, we narrowed down the panel to 30 drugs that could be rapidly deployed in clinical trials including: drugs FDA-approved in AML, drugs FDA-approved in other cancers, and targeted therapeutics that have been used in clinical trials with literature supporting use in AML. We created a diagnostic plate of these drugs in dose-response and optimized the assay for use in routine clinical implementation. Again, we observed clear heterogeneity across patients. In addition to matching patients with drugs that their cells respond to, we have also found that responses to some drugs can predict whether patients will respond to intensive chemotherapy making this a novel biomarker.

Since we have clear patient-specific signatures, and the potential to use this assay as a biomarker for chemotherapy responsiveness, we have partnered with Dr. Roth to validate the assay in the CLIA lab, and Dr. Luger to begin clinical trials to determine if drug sensitivity profiling can direct therapeutic interventions. There are clear directions. First, we can help to assign salvage therapies as there are three regimens that are used routinely with little guidance. One therapy uses etoposide, which we found shows variable responsiveness with a large fraction of patient tumor cells non-responsive. Second, to assign treatment arms based on sensitivity data. For example, the MEK inhibitor trametinib, FDA-approved in solid tumors, and which failed in small-scale clinical trials in AML, shows patient-specific activity in ~10% of patients. Third, we found that we could predict which patients would respond to intensive chemotherapy based upon their responses to a targeted therapeutic, and this information, if indeed predictive, would inform patient care for a subset of AML patients.  We are on the cusp of full-scale implementation of this new functional precision medicine approach in AML.