A stem and progenitor cell-derived gene expression signature is prognostic for survival in myelofibrosis

In this study, we hypothesized that transcriptomic features among disease-driving hematopoietic stem and progenitor cells (HSPC) could improve current paradigms for risk stratification in myelofibrosis (MF). Therefore, we performed bulk RNA sequencing (RNA-seq) of blood from 358 patients with MF, split into training and test cohorts. Single-cell RNA-seq data from Lin-CD34+ MF HSPCs were used to guide the development of a prognostic model trained on the bulk RNA-seq data. A NanoString assay was used to validate our final model in 2 independent cohorts of patients with MF. We identified a 24-gene weighted-sum expression score (MPN24) that was prognostic of overall survival (OS). In 2 independent cohorts of 170 and 100 patients with MF, MPN24-high scoring patients had worse 5-year OS compared with MPN24-low scoring patients (25.5% vs 86.9%, hazard ratio [HR] = 9.81, P = 4.0e-11 and 23.5% vs 75.9%, HR, 6.40, P = 1.8e-7, respectively). MPN24-high was also associated with clinical, mutation and karyotypic features known to confer adverse risk in MF. The MPN24 score retained independent prognostic significance in multivariable analysis incorporating these covariates and existing risk stratification models including Dynamic International Prognostic Scoring System (DIPSS), DIPSS plus, Mutation-Enhanced International Prognostic Score System (MIPSS70), and MIPSS70 plus version 2.0, adding significant prognostic value to these risk stratification models (P = 1.6e-5, 4.0e-6, 1.5e-6, and 5.4e-4, respectively). The MPN24 score was particularly useful in improving risk-stratification of DIPSS-intermediate, and MIPSS70-intermediate and high-risk patients. MPN24 is an HSPC-derived gene expression score that is associated with OS independent of clinical and genomic prognostic variables and improves risk stratification in MF.