SVILUPPO E CARATTERIZZAZIONE DI MODELLI PRECLINICI IN VIVO PER L'IDENTIFICAZIONE DI NUOVI APPROCCI TERAPEUTICI PER LA LEUCEMIA MIELOIDE ACUTA PEDIATRICA

In pediatric acute myeloid leukemia (AML) chemotherapy is the standard of care, but more than 25% of patients still relapse and after a disease recurrence the survival probability is extremely low (<50%). There is an urgent need to discover new treatments to ameliorate patients’ outcome. However, pediatric drug development is extremely reduced by the need of a better understanding of the adverse event profile of adult cancer indications in children, by the lack of pediatric-specific formulations, and by the reduced number of pediatric AML patients that can be included in clinical trials. Therefore, robust preclinical AML models to faithfully predict new drug efficacy is urgently needed to advance new drugs in clinical setting. This PhD thesis aims at the development of more reliable preclinical models to improve treatments for children suffering from life threatening AML. During my PhD I established patient derived xenograft models (PDX) of pediatric AML. Primary AML samples have been inoculated in NSG mice and, when engrafted, in 3 consequent mice recipients (namely P0, P1 and P2-PDXs). These models have been characterized for AML immunophenotypic profile and by RNA and whole-exome sequencing (WES). According to somatic mutations we determined AML clonal composition. These characterization allowed to select new drugs to be screened in vitro and in vivo. We finally generated 22 AML-PDXs mostly representing high-risk AML genetic markers such as t(5;11)NUP98-NSD1, t(3;5)NPM1-MLF1, t(16;16)CBFA2T3-GLIS2, t(16;21)FUS-ERG, KMT2A somatic translocations, or FLT3ITD mutation. We monitored the AML associated immunophenotype in PDXs finding it similar to that of the original AML. By WES we detected a consistent number of variants in each patient’s AML (ranging from 28 to 69), confirming a high intra-tumoral AML heterogeneity. Furthermore, we did not find any mutation recurrence among models, underlining a high inter-tumoral heterogeneity. In all models we tracked clonal evolution from patients’ AML to P2-PDXs highlighting that most variants were maintained, with very few variants acquired or lost during model development. Monitoring clonal dynamics we recognized a specific "founder" clone characterized by an average of 30 variants which are maintained up to P2 at the same allelic frequency, other small clones with an average of 10 variants increasing the allelic frequencies in P2 and, in a restricted number of models some lost clones. By WES and transcriptome analysis we highlighted druggable mutations and pathways allowing the selection of novel targeted drugs. We screened their efficacy in vitro alone or combined with Venetoclax, using AML ex vivo cells and mesenchymal stromal cells in a 3D co-culture system for exploring their synergy in reducing AML proliferation. This refinement permitted to select two drugs currently under evaluation in AML-PDX models. Overall, in this work we created a large series of paired AML and xenograft models for advancing pediatric AML therapeutics. Our models represent a concrete perspective for both the identification of new variants and pathways involved in AML progression to deepen into AML biology, as well as the possibility to perform novel drug screenings that will be further used to increase AML drug portfolio, with the final aim to translate new drugs into human clinical trials, contributing to cure pediatric AML.