Abstract Background: Acute myeloid leukemia (AML) is sustained by a rare population of leukemic stem cells (LSCs) that reside within protective bone marrow (BM) niches. These LSCs are responsible for disease propagation, therapy resistance, and relapse. While the endosteal and sinusoidal niches have been well characterized, the longitudinal spatial organization of the BM—particularly the distinction between metaphysis and central marrow—has not been systematically investigated in the context of AML. A better understanding of how spatial cues influence LSC behavior is critical for developing effective niche-targeted therapies. Methods: We utilized murine AML models (MLL-AF9 and AML1-ETO9a) with genetic manipulation of Dpp4, Cxcl12, and Gpc3, specifically in leukemic cells or niche-resident N-cadherin⁺ mesenchymal stromal cells (MSCs). Using flow cytometry, immunofluorescence microscopy, cytokine profiling, and functional transplantation assays, we evaluated the distribution, proliferation, apoptosis, and stemness of LSCs across different BM regions—proximal/distal metaphysis and central marrow. Single-cell RNA sequencing (scRNA-seq) was performed on non-hematopoietic stromal compartments from the metaphysis to identify Cxcl12-producing cells and niche-specific regulatory factors. Results: We found that LSCs preferentially localize to the proximal and distal metaphysis (PM/DM), where they exhibit quiescence and high leukemogenic potential. In contrast, LSCs in the central marrow (CM) display increased cell cycle activity and reduced self-renewal. Genetic deletion of Dpp4 in AML cells disrupted a previously unrecognized CXCL12 gradient across the BM, causing LSCs to redistribute from the metaphysis to the CM. This redistribution triggered increased proliferation and apoptosis, indicative of functional exhaustion. Notably, Dpp4 deletion also reversed the CXCL12 gradient between BM and peripheral blood, confining AML cells within the marrow and dramatically reducing peripheral dissemination. scRNA-seq analysis identified N-cadherin⁺ MSCs as the primary Cxcl12-expressing niche population within the metaphysis. These cells also express high levels of Glypican-3 (GPC3), a known DPP4 inhibitor. Mechanistically, GPC3 preserved CXCL12 activity locally by inhibiting DPP4 enzymatic degradation, establishing a localized CXCL12-rich environment that supports LSC maintenance. Conditional deletion of either Cxcl12 or Gpc3 in N-cadherin⁺ MSCs phenocopied the effects of Dpp4 loss, leading to LSC displacement, exhaustion, and impaired leukemogenesis. Transcriptomic profiling confirmed convergence of Dpp4-deficient and Cxcl12/Gpc3-depleted models, with shared downregulation of stemness-related pathways and upregulation of metabolic stress and cell death signatures. Conclusions: This study defines a novel spatial regulatory mechanism in AML, where the CXCL12–DPP4–GPC3 axis governs LSC localization between the metaphysis and central marrow. By disrupting this axis, LSCs are mobilized out of protective niches and driven toward exhaustion. Furthermore, reversing the CXCL12 gradient restricts AML cells to the marrow, reducing dissemination and prolonging survival. These findings not only uncover a previously underappreciated dimension of BM niche organization but also highlight DPP4 inhibition as a rational therapeutic approach to target the AML stem cell–niche interaction. Given that DPP4 inhibitors are already FDA-approved for other indications, these insights hold strong translational potential for rapid clinical application in AML.
Zhou et al. (Mon,) studied this question.