In many patients with acute lymphoblastic leukaemia (ALL), most cells respond to chemotherapy but a minority show resistance and cause relapse with poor outcome. Using patient-derived xenograft models, Ebinger et al. identified a subpopulation of resistant and relapse-inducing cells with long-term dormancy and stemness properties similar to those in primary ALL cells from patients at minimal residual disease (MRD). When dissociated from the in vivo environment, these cells started proliferating and became sensitive to treatment, which suggests that patients with ALL might benefit from therapeutic strategies that release MRD cells from the niche.