Distinct and non-overlapping epigenetic signatures differentiate MPN subtypes and reveal hydroxyurea-induced remodeling

Abstract

Abstract Background Myeloproliferative neoplasms (MPNs) including polycythemia vera (PV), essential thrombocythemia (ET), and myelofibrosis (MF), are clonal hematopoietic stem cell disorders driven by shared mutations in JAK2, CALR, and MPL genes. Despite overlapping genetic features, these diseases differ markedly in the clinical manifestations, disease prognosis, and treatment response. Epigenetic mechanisms including 5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC), are increasingly recognized as contributing to this heterogeneity. This study focuses on identifying regions of differential DNA methylation and hydroxymethylation between MPN subtypes and healthy individuals. Methods We developed a dual-color chemoenzymatic fluorescence labeling assay enabling simultaneous detection of 5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC) on the same DNA molecule. Peripheral blood samples were collected from patients treated in the department of Hematology at Shamir Medical Center, Be'er Ya'akov, Israel. Forty-two with PV (median age 60, range 20-86), 34 with ET (median age 66, range 36-91), 13 with MF (median age 73, range 57-85), and 36 age-matched healthy controls were included. Approximately 50% of patients and healthy controls were females. Within the PV and ET cohorts, patients were further classified based on Hydroxyurea (Hydrea) treatment status. Genomic DNA was extracted, and 5hmC and 5mC were fluorescently labeled using distinct fluorophores. The labeled samples were hybridized to custom-designed DNA microarrays, and fluorescence intensities were quantified and normalized for downstream comparative analysis. Results We identified distinct, non-overlapping epigenetic signatures for each MPN subtype. Both 5mC and 5hmC profiles revealed disease-specific alterations that enabled accurate classification from healthy controls. Each subtype; PV, ET, and MF displayed a unique epigenetic profile. Classification performance was high for both 5mC and 5hmC models. Using 100 differentiated probes, 5mC-based classifiers achieved 100% accuracy in distinguishing PV and MF from healthy controls, and 85% for ET. 5hmC-based models similarly reached 100% for PV and MF, and 88% for ET. Healthy controls were correctly classified with ≥94% accuracy in all comparisons. MF samples showed 100% classification accuracy across both 5mC and 5hmC, despite the smaller sample size. Hydroxyurea treatment induced consistent and distinguishable epigenetic remodeling in both modifications. In PV, 5hmC-based models achieved 100% accuracy in separating treated from untreated patients, and 5mC-based models reached 85%. For ET, the accuracy was 90% with 5mC and 90% with 5hmC. These Hydrea-associated profiles were independent of the disease-defining signatures, indicating that treatment induces a distinct epigenetic state rather than shifting patients along the disease axis. GO analysis of Hydrea-responsive loci in PV revealed significant enrichment in cell adhesion pathways, including cell-cell adhesion via plasma membrane adhesion molecules, suggesting that Hydrea may influence disease phenotype through modulation of bone marrow interactions. Conclusions Our dual-color epigenetic microarray platform reveals distinct and non-overlapping DNA methylation and hydroxymethylation signatures for each MPN subtype, enabling high-accuracy classification from peripheral blood. These findings highlight the epigenetic heterogeneity underlying PV, ET, and MF, beyond their shared mutational landscape. Moreover, we demonstrate that Hydroxyurea treatment induces a reproducible and orthogonal epigenetic shift, independent of disease subtype, suggesting therapy-driven remodeling of regulatory pathways. Enrichment of cell adhesion-related GO terms in Hydrea-treated PV samples further supports a potential link between treatment response and microenvironmental interactions. Together, these results underscore the utility of locus-specific 5mC/5hmC profiling for minimally invasive disease classification and therapeutic monitoring in MPN.