Differentiation therapy with all-trans retinoic acid (atRA) has markedly improved outcome in acute promyelocytic leukemia (APL) but has had little clinical impact in other AML sub-types. Cell intrinsic mechanisms of resistance have been previously reported, yet the majority of AML blasts are sensitive to atRA in vitro. Even in APL, single agent atRA induces remission without cure. The microenvironment expression of cytochrome P450 (CYP)26, a retinoid-metabolizing enzyme was shown to determine normal hematopoietic stem cell fate. Accordingly, we hypothesized that the bone marrow (BM) microenvironment is responsible for difference between in vitro sensitivity and in vivo resistance of AML to atRA-induced differentiation. We observed that the pro-differentiation effects of atRA on APL and non-APL AML cells as well as on leukemia stem cells from clinical specimens were blocked by BM stroma. In addition, BM stroma produced a precipitous drop in atRA levels. Inhibition of CYP26 rescued atRA levels and AML cell sensitivity in the presence of stroma. Our data suggest that stromal CYP26 activity creates retinoid low sanctuaries in the BM that protect AML cells from systemic atRA therapy. Inhibition of CYP26 provides new opportunities to expand the clinical activity of atRA in both APL and non-APL AML.