Determinants of clinical phenotype in myeloproliferative neoplasms

Background: Myeloproliferative neoplasms, (MPNs) such as polycythemia vera, essential thrombocythemia, and myelofibrosis, are chronic hematologic cancers with varied progression rates. The genomic characterization of patients with myeloproliferative neoplasms offers the potential for personalized diagnosis, risk stratification, and treatment. Methods: We sequenced coding exons from 69 myeloid cancer genes, common to two separate bait sets, in patients with myeloproliferative neoplasms, comprehensively annotating driver mutations and copy-number changes or copy. We developed a genomic classification for myeloproliferative neoplasms and multistage prognostic models for predicting outcomes in individual patients. Classification and prognostic models were validated in an external cohort. Cytokine profiles of over 400 patients were also analysed to determine the contribution of the inflammatory microenvironment to phenotype and progression risk. Results: A total of 2035 patients were included in the analysis. 33 genes had driver mutations in at least 5 patients, with mutations in JAK2, CALR, or MPL being the sole abnormality in 45% of the patients. The numbers of driver mutations increased with age and advanced disease. Driver mutations, germline polymorphisms, and demographic variables independently predicted whether patients received a diagnosis of essential thrombocythemia as compared with polycythemia vera or a diagnosis of chronic-phase disease as compared with myelofibrosis. In particular a set of mutations that included ASXL1, SRSF2, U2AF1 and EZH2 was enriched in myelofibrosis and associated with poor outcomes. The JAK2 46/1 haplotype strongly correlated with the presence of 9pUPD and independently with a PV phenotype, demonstrating that the underlying germline background can play a role in determining somatic events and can affect the patient’s phenotype in its own right. We defined eight subgroups based solely on clustering of genomic data that showed distinct clinical phenotypes, including blood counts, risk of leukemic transformation, and overall survival. These included a sub-group defined by mutations the same set of chromatin and splicesome component genes described above, and a subgroup enriched for TP53 mutations and chromosomal changes, which carried a significant risk of AML transformation. Patients with no detectable mutations had very low rates of progression or death. By integrating 63 clinical, demographic and genomic variables, we created prognostic models capable of generating personally tailored predictions of clinical outcomes in patients with chronic-phase myeloproliferative neoplasms and myelofibrosis. The predicted and observed outcomes correlated well in internal cross-validation of a training cohort and in an independent external cohort. The prognostic model performed as well as or better than a number of existing risk scores including the high molecular risk genetic score and international prognostic scoring systems and even within individual categories of existing prognostic schemas, our models substantially improved predictive accuracy. Cytokine profiles varied significantly across MPN subtypes, with high levels of TNFalpha and IP-10 seen in myelofibrosis, and to a lesser extent in polycythemia vera. Patients with essential thrombocytosis however, were found to have high levels of GROalpha and EGF, and levels of these at single time points or when measured longitudinally were predictive for the risk of progression to myelofibrosis. Conclusions: Comprehensive genomic characterization identified distinct genetic subgroups and provided a classification of myeloproliferative neoplasms on the basis of causal biologic mechanisms. Integration of genomic data with clinical variables enabled the personalized predictions of patients’ outcomes and may support the treatment of patients with myeloproliferative neoplasms.This research was funded by grants from Bloodwise and Kay Kendall Foundatio