BMP and ERK signaling modulate AML-stroma interactions and contextually alter cytarabine sensitivity in AML cells.

AIMS: Relapse in acute myeloid leukemia (AML) frequently arises from therapy-resistant cells persisting within the bone marrow microenvironment. This study investigated stromal interactions and signaling pathway dependencies regulating cytarabine (AraC) response during AML progression and resistance acquisition.

MATERIALS AND

METHODS

AML cell lines representing FAB-M1 (KG1a), FAB-M5 (THP1), and an Ara-C-resistant THP1 derivative (CR-THP1) were cultured alone or with bone marrow-derived mesenchymal stromal cells (BMSCs). Cell viability, clonogenicity, signaling pathway activation, and gene expression were analyzed following Ara-C treatment with or without BMP and ERK pathway inhibition. Primary AML mononuclear cells and publicly available transcriptomic datasets were used for validation.

KEY FINDINGS: Stromal coculture protected THP1 from Ara-C-induced apoptosis but sensitized KG1a to AraC. Despite reduced stromal adhesion, CR-THP1 cells exhibited increased Ara-C sensitivity under coculture, indicating altered microenvironmental signaling during resistance acquisition. BMP pathway inhibition enhanced Ara-C cytotoxicity in parental THP1 through suppression of stromal-induced β-catenin upregulation but failed to sensitize CR-THP1 cells. In contrast, ERK inhibition restored Ara-C sensitivity in CR-THP1, downregulated ABCB1 and ABCG2 expression, and prevented leukemia regrowth in an in vitro relapse model. Transcriptomic analyses identified alterations in drug efflux, apoptosis imbalance, BMP, ERK, and Wnt/β-catenin networks during resistance acquisition.

SIGNIFICANCE: These findings identify BMP and ERK signaling as context-dependent regulators of Ara-C response in AML and demonstrate that stromal interactions evolve during resistance acquisition. Targeting BMP and ERK pathways, together with modulators of DNA repair or drug efflux, may provide therapeutic strategies to reduce relapse and improve AML outcomes.