Background Classic Ph-negative myeloproliferative neoplasms (MPN) include essential thrombocythemia (ET), polycythemia vera (PV) and primary myelofibrosis (PMF). Clonal evolution can lead MPN patients in chronic phase (CP) to developacute myeloid leukemia (AML), called blast phase (BP), a rare and late event in the natural history of MPN.Aim To evaluate differences in clinical, cytogenetic and molecular features and outcome in 92 patients with BP MPN, according to MPN diagnosis and mutational profile.Methods We identified in our database all patients affected with ET, PV and PMF who developed AML according to 2016 WHO criteria and for whom at least one DNA sample was available to define the mutational status of the 3 MPN driver genes (JAK2, CALR, MPL). Target sequencing was performed using the Illumina TruSight Myeloid Sequencing Panel kit on the HiSeq 2000 platform. For each patient circulating granulocytes or CD34+ cells were analyzed as neoplastic tissue and CD3+ T cells as control tissue; data were analyzed with the pipeline for somatic variant calling implemented at CeMM. Treatment used in BP was classified as palliation (supportive care or low intensity chemotherapy) or induction chemotherapy (de novo AML-like therapy).Results We retrospectively identified 92 patients who progressed to BP, with a known molecular profile: 29 ET (18 JAK2 V617F , 9 CALR, 1 MPL mutated and 1 triple-negative), 39 PV (all JAK2 V617F mutated) and 24 PMF patients (18 JAK2 V617F, 1 CALR, 2 MPL mutated and 3 triple-negative).Median age at BP was 70.8 years (range 46.3-86); age at evolution was not significantly different (P 0.182) in the 3 diagnostic categories (ET, PV, PMF).The complete blood count at leukemic evolution was not influenced by the initial diagnosis.Median time to leukemic evolution was 110 months (range 3.6-367.6) and it was shorter in PMF compared to ET (29.8 vs 153.4 months, P < 0.001) and PV (29.8 vs 129.7 months, P < 0.001). Moreover, time to leukemic evolution was shorter in JAK2 V617F positive ET compared to CALR mutated ET (117.2 vs 210.9 months, P 0.027), but not statistically different in JAK2 mutated compared to CALR mutated PMF (P 0.312).At the time of BP, 35 out of 49 patients (71%) for whom cytogenetic analysis was available showed an abnormal karyotype (22 patients with complex karyotype or high risk aberrations).JAK2 mutated MPN can evolve into JAK2 wild type AML (8 of 32 patients with blast DNA available), whereas CALR and MPL mutations were identified also in AML blasts in all patients for which DNA was available (3 for CALR and 3 for MPL mutated AML).We performed targeted re-sequencing in 53 patients, in 23 studying both the CP and the BP sample. Overall, we identified 176 variants in 39 different genes (68% were missense mutations); the most commonly involved pathways were DNA methylation (22%), transcriptional regulation (20%) and chromatine regulation (18%). The most frequently involved genes were TET2, TP53, RUNX1 and ASXL1 . In the 23 paired cases, we observed a statistically significant increase in the median number of mutations in the BP, compared to CP [4 (range 1-10) vs 2 (range 0-9); P 0.002]. This was observed in JAK2 mutated patients (P 0.006) and in those with PV (P 0.008), but not in subjects with TE or PMF. In our paired-sample cohort, BP evolution was associated with the acquisition of mutations (e.g. IDH1) and/or an increase in mutation occurrence or variant allele frequency (VAF) of genes implicated in leukemogenesis including TET2 (P 0.026), RUNX1 (P 0.003), ETV6 (P 0.068) and TP53 (P 0.003), the latter known to be involved also in JAK2 mutated MPN evolving into JAK2 wild type AML. In addition, VAFs suggested an heterogeneous clonal architecture and evolution of somatic mutations in these patients.Outcome was dismal, with a median OS of 3.8 months (range 3.1-5.3), independently from MPN diagnosis (P 0.603) and treatment during BP (5.6 months with induction chemotherapy vs 3.5 months with palliation; P 0.171).Conclusion Clinical phenotype and outcome of BP is not influenced neither by the diagnosis in CP nor by the type of driver mutation; likewise the outcome is poor irrespective of treatment. PMF patients have a shorter time to BP than ET and PV patients; in ET JAK2 V617F mutation is associated with a shorter time to BP compared to CALR mutation. TP53, RUNX1, TET2 and ETV6 are more frequently involved in leukemic evolution and the number of mutations increases from CP to BP. DisclosuresNo relevant conflicts of interest to declare.