The nucleoside analogue nelarabine, the prodrug of arabinosylguanine (AraG), is effective against T-cell acute lymphoblastic leukaemia (T-ALL) but not against B-cell ALL (B-ALL). The underlying mechanisms have remained elusive. Here, data from pharmacogenomics studies and a panel of ALL cell lines reveal an inverse correlation between nelarabine sensitivity and the expression of SAMHD1, which can hydrolyse and inactivate triphosphorylated nucleoside analogues. Lower SAMHD1 abundance is detected in T-ALL than in B-ALL in cell lines and patient-derived leukaemic blasts. Mechanistically, T-ALL cells display increased SAMHD1 promoter methylation without increased global DNA methylation. SAMHD1 depletion sensitises B-ALL cells to AraG, while ectopic SAMHD1 expression in SAMHD1-null T-ALL cells induces AraG resistance. SAMHD1 has a larger impact on nelarabine/AraG than on cytarabine in ALL cells. Opposite effects are observed in acute myeloid leukaemia cells, indicating entity-specific differences. In conclusion, SAMHD1 promoter methylation and, in turn, SAMHD1 expression levels determine ALL cell response to nelarabine. Rothenburger et al. combine the analysis of data from large pharmacogenomics screens with cell line experiments to elucidate mechanisms behind nelarabine resistance in acute lymphoblastic leukaemia (ALL). They identify low expression of the dNTP hydrolase SAMHD1 as a determinant of nelarabine sensitivity, suggesting the enzyme as a biomarker.