Paper Details

ABSTRACT

Background: Myelofibrosis (MF) is an advanced myeloproliferative neoplasm (MPN) associated with progressive bone marrow (BM) fibrosis, leukemic transformation, and shortened overall survival. Current models implicate reciprocal inflammatory signaling between mutant hematopoietic cells and a reprogrammed stromal-vascular niche in MF pathogenesis. Although emerging single-cell and spatial studies have revealed complex immune-stromal alterations in MF, how niches evolve across fibrosis grades and in other MPN sub-types, including polycythemia vera (PV), essential thrombocythemia (ET), and post-MPN acute myeloid leukemia (AML), remain incompletely defined.

Methods: We performed comprehensive spatial transcriptomic profiling of a large cohort of untreated MF patient BM spanning histologic fibrosis grade (N=8 MF-1, N=12 MF-2, N=10 MF-3), along with non-fibrotic MPN (N=8 PV, N=8 ET), post-MPN AML (N=6), and 8 healthy controls, using high-resolution Xenium technology (total N=60 pts). A custom 480-gene panel was used to capture various hematopoietic, mesenchymal stromal cell (MSC), collagen/fibroblastic, and other immune/inflammatory/stromal cell states. Mutant-specific probes targeting JAK2^VF, CALR^ins5, CALR^del52, and MPL^W515L were also available on a subset (N=27) of PV/ET and MF BM, enabling genotype-aware spatial analysis.

Results: Cell-composition profiling showed progressive reductions in erythroid and lymphoid populations and relative expansion of MSC and vascular/endothelial cells across control, PV/ET, and MF sub-groups. Thirty-five of 231 evaluable cell-pair relationships differed between MPN/MF and healthy BM after FDR correction, including increased endothelial, macrophage, and granulocytic-monocytic progenitor self-neighborhood enrichment and reduced vascular-myeloid, stromal-granulocyte, and erythroid-/megakaryocyte (MK)-progenitor associations. Ligand-receptor analysis showed stage-dependent reductions in multiple CXCL12 interactions, including CXCL12-CXCR4, which appeared secondary to reduced inferred signal strength/range rather than reductions in total ligand-emitting cells.

In parallel, multiple collagen-integrin and extracellular matrix (ECM)-sensing and MK-derived chemokine interactions were progressively increased with MF and post-MPN AML. This shift was accompanied by reductions in B-/T-lymphoid neighborhoods with emergence of myeloid-enriched clusters and, in particular, a heterogeneous hematopoietic-stromal neighborhood consisting of osteoclast/osteogenic stromal cells, HSCs, pericytes, MSCs, endothelial cells, macrophages, and monocytes. Expression analysis of this cluster revealed enrichment in multiple ECM/fibrosis (FN1, COL3/1A1, TGFB1) and myeloid-inflammatory (STAT1/3, CSF1R, SPP1) genes as well as a T-cell suppressive program (IDO1, TIGIT, CD276) suggesting coordinated fibrotic remodeling within this neighborhood. Dedicated analysis of pathologic MK “clusters”—a hallmark feature of MPNs—revealed MK hubs enriched in MSCs/osteoclasts and myeloid-/erythroid-progenitors over-expressing ECM (FN1, TGFB1, THBS1) and vascular reprogramming (PDGFB, VEGFA) genes in MF relative to non-clustered MKs within the same patients. Notably, these hub-associated transcriptional changes varied between MF and PV/ET suggesting that the processes contributing to MK clustering may differ based on disease phenotype.

In post-MPN AML, immature stem/progenitor and MK-biased leukemic blasts increasingly co-localized with collagen-expressing fibroblastic MSCs suggesting remodeled fibroblastic niches may persist with, or even support, AML evolution. Finally, exploratory genotype-aware spatial analyses revealed mutation-specific differences in both cell neighborhood organization and co-localization patterns, including shorter nearest-neighbor distances between macrophages and JAK2^VF cells compared to MPL^W515L cells, which co-localized more with MKs/myeloid progenitor cells and were relatively distant from macrophages.

Conclusions: MPN progression is characterized by coordinated spatial remodeling of the BM ecosystem, with expansion of disease-specific MK-centered and hematopoietic-stromal neighborhoods enriched in inflammatory and ECM/vascular remodeling programs. Exploratory genotype-aware analyses further suggest individual MPN driver mutations are associated with distinct cellular/spatial organizational patterns. Together, these findings may provide a framework for prioritizing niche-directed therapies for advanced MPN.