Personalized Therapy Design for MPN Using Predictive Simulation Methodology with in Vitro, Ex Vivo, and in Vivo Validatio

Background: Currently approved therapies for myeloproliferative neoplasms (MPNs) are limited to cytoreductive agents such as hydroxyurea and the Jak1/2 inhibitor ruxolitinib. While these agents alleviate constitutional symptoms, they are unable to provide significant histopathologic, cytogenetic, or molecular remissions. Furthermore, adverse events and drug resistance are associated with these frontline therapies. This therefore warrants the design of novel therapies that can provide efficacy, either alone or as adjuvants, while minimizing adverse events and resistance. To address this, we developed predictive simulation software to create patient simulation avatars based on signaling networks impacted by hyper-kinetic JAK2 signaling. The avatars can then be simulated with FDA approved drugs from across indications, individually or in combinations, with the intent of identifying novel therapies for MPN.

Methods: A bone marrow aspirate from a PV patient was analyzed for chromosomal alterations using array Comparative Genomic Hybridization (aCGH) and peripheral blood from the same patient was analyzed for cytokine expression levels. The bone marrow cells were positive for the JAK2-V617F mutation in 44.7% of cells. aCGH analysis of bone marrow and peripheral blood cells did not identify other genomic aberrations. Cytokine profiling indicated an increase in MMP9, RANTES (CCL5) and VEGF, and a decrease in MCP1, relative to a non-diseased control sample. Using this information, a predictive simulation patient avatar was created. A library of over 75 FDA approved agents with unique mechanism of actions was then simulated against the patient avatar. A list of predicted drugs was then generated and subsequently validated using i) BaF3/Jak2-V617F cells, ii) bone marrow mononuclear cells derived from the patient, and iii) an animal model of Jak2-mediated MPN.

Results: The predictive simulation indicated synergistic efficacy of two drugs: roflumilast and chloroquine. The combination of inhibiting PDE4 and autophagy using roflumilast and choloroquine, respectively, was predicted to reduce proliferation and viability of JAK2-V617F cells which have hyper-activated JAK2, STAT3, and STAT5. Roflumilast was predicted to reduce the dominance of ERK, NF-kB, SHH pathways, and induce CDKN1A and CDKN1B, cell cycle inhibitors. Choloroquine was predicted to induce ROS, ER Stress, TP53, ceramide biosynthesis, and cell cycle inhibition. The combination of the two, through different mechanisms, was predicted to reduce proliferation and viability of mutant JAK2 cells. Furthermore, they were predicted to synergize with the action of low dose ruxolitinib. To validate the simulation predictions, we first treated BaF3/Jak2-V617F cells with increasing concentrations of roflumilast (0 – 30 μM) and chloroquine (0 – 30 μM). We found that both agents significantly reduced the viability of the cells in a dose-dependent manner, both alone and in combination. When the patient’s bone marrow cells were similarly treated with roflumilast (0 – 20 μM) and chloroquine (0 – 20 μM), the numbers of BFU-E were reduced dose-dependently. Furthermore, when roflumilast (10 – 20 μM) and chloroquine (10 - 20 μM) were used in combination with low dose ruxolitinib (100 nM), there was a synergistic effect leading to the near elimination of mutant clonogenic growth potentials of the patient’s cells (44±4.24 BFU-E for vehicle vs. 3±2.83 BFU-E for 20 μM Rof/20 μM Chloro/100 nM Rux). Lastly, when BaF3/Jak2-V617F cells were xenografted into nude mice, roflumilast (8 mg/kg/day) and chloroquine (50 mg/kg/day) were highly effective when used in combination as measured by an improved red blood cell count, reduced splenomegaly, and a decreased tumor burden within the liver.

Conclusions: This study demonstrates how predictive simulation technology can be used to identify novel therapeutic strategies for the treatment of MPN. Specifically, genomic and cytokine profiling were used to map dysregulated signaling pathways and then accurately predict novel therapeutic agents that can act either alone, in combination, or as an adjuvant to an existing therapy. The results from this study are significant as they i) serve as the basis for an early phase clinical trial using roflumilast and chloroquine for the treatment of PV and ii) have validated a novel means to rapidly re-purpose existing FDA approved drugs for the treatment of MPN.