AUTHORS
Lucas Wadley, PhD, BS, Madison Fitzsimmons2, Angela Fleischman, MD, PhD3
AFFILIATIONS
1 University of California Irvine, Irvine, California, United States
2 University of California Irvine, Irvine, California, United States
3 University of California Irvine, Irvine, California, United States
ABSTRACT
Introduction: The JAK2V617F mutation most commonly drives myeloproliferative neoplasms (MPN), yet clinical heterogeneity persists, particularly in response to interferon-alpha (IFNα) therapy. We hypothesized that the mechanistic basis of JAK2 activation, canonical receptor-dependent signaling versus non-canonical JAK2 transactivation, creates distinct gene regulatory landscapes that influence therapeutic vulnerability. This study integrates RNA sequencing and chromatin accessibility profiling to elucidate how JAK2 activation mechanism shapes malignant gene regulation.
Methods: We utilized Ba/F3 cell models to dissect JAK2 activation mechanisms. Ba/F3 cells co-expressing JAK2V617F with JAK2 canonical homodimeric receptor, MPL, or a non-canonical heterodimeric receptor,IL-10R, were selected to IL-3 independence; parental and JAK2-WT cells were controls. Paired RNA-seq and Omni-ATAC-seq were analyzed by DESeq2, GSEA and GO, HOMER motif analysis, and TOBIAS footprinting. Ruxolitinib-resistant derivatives were generated by dose escalation and profiled in parallel. Integration of RNA-seq and ATAC-seq data identified concordant gene regulatory changes linking chromatin accessibility to gene expression.
Results: PCA from both RNA and ATAC data revealed distinct clustering of JAK2V617F-transformed clones by activation mechanism, demonstrating that non canonical JAK2V617F signaling (IL-10R) and canonical receptor-dependent activation (MPL) create unique transcriptional states despite shared malignant features. Core transformation signatures were conserved across both contexts, including upregulation of MYC and cell cycle-related genes, confirming shared malignancy programs. Non-canonical cells were selectively enriched for type I and II IFN response programs at expression and accessibility levels, with TNFα/NF-κB (NES +2.10), TGF-β (NES +1.95), and epithelial-mesenchymal transition (NES +1.63; all padj<0.05), linking this mode to the inflammatory and profibrotic axis of MPN progression. Motif and footprinting analyses identified the mechanism: the two STAT1 response elements segregated to opposite sides of the comparison. ISRE, the ISGF3 element, was among the most enriched motifs in non-canonical-gain regions (4.67-fold, p=1e-99), whereas the GAF element, GAS, was enriched in canonical-gain regions (q=1e-19) with AP-1 motifs. At occupancy, 77.7% of 3,061 ISGF3 sites were preferentially bound in non-canonical cells (OR 3.74 vs GAF reference, p<0.001), while 6,398 GAF sites behaved neutrally (48.3%). ISGF3 sites overlapped non-canonical-accessible regions at 2.40:1 and GAF sites at 0.32:1, a 7.5-fold inversion. NF-κB and C/EBP motifs were preferentially bound in non-canonical cells. To test whether this state is reachable under therapeutic pressure, we drove both models into ruxolitinib resistance. Canonical resistant cells shifted toward the non-canonical transcriptome: of 3,101 genes significant in both the resistance and baseline contrasts, 73.6% moved concordantly (Fisher OR 7.75, p=9.5e-158). Non-canonical cells, already in that state, changed minimally. GSEA showed strong IFNa and IFNg enrichment in canonical resistant cells (padj<1e-10) versus modest enrichment in non-canonical (padj=0.005). The IFN response did not converge uniformly: ISGF3-dependent type I ISGs (Ifit1, Oas3, Isg15, Irf7) converged upward, whereas GAF-dependent type II targets (Stat1, Irf1, Tap1) moved oppositely (panel r=+0.13, p=0.44). Ruxolitinib pressure thus reproduces, by an orthogonal perturbation, the type I versus type II partition found by footprinting.
Conclusions: Receptor context determines the regulatory identity of JAK2V617F-transformed cells. Non-canonical transactivation establishes a coherent inflammatory-interferon chromatin state (ISRE accessibility, NF-kB and C/EBP binding, EMT priming); canonical signaling drives a GAS/AP-1 program. The same STAT1 protein is partitioned into opposing effector complexes by receptor scaffold, not expression level. This defines an axis of MPN heterogeneity independent of allele burden, co-mutations, and mutation order, which ruxolitinib pressure selects for, and predicts that receptor context shapes IFNα responsiveness and ropeginterferon efficacy.
At current time, in vivo models are being designed for further investigation in a physiological model. Additonally, public datasets from previous MPN studies are being analyzed for interferon and STAT1 complex signatures across disease stage and driver context.
DISCLOSURES
Lucas Wadley, PhD, BS: No relevant conflicts of interest to declare.