Relapses after therapy-induced remissions remain a major challenge in cancer. 1, 2 Molecular detection of minimal residual disease (MRD) is valuable in relapse prediction but its cellular basis remains largely unknown. 3–5 Delineating therapy-resistant cells (Figure 1a) might inform new therapeutic strategies.‘Cancer stem cells’, defined by xenografting, have been suggested to preferentially evade therapy, but evidence for persistence of phenotypically and functionally distinct cells in patients undergoing treatment is largely lacking.

We investigated the cells responsible for MRD in childhood acute lymphoblastic leukaemia (cALL)(for patient information see Table 1), where molecular MRD monitoring has had a practisechanging impact. 6 At presentation, cALLs comprise phenotypically distinct B-cell stages, including ProB-like (CD34þ CD38þ CD19þ), and cells dubbed ‘Stem/B’coexpressing stem (CD34þ CD38À/low) and B-cell (CD19þ) markers, seen only in leukaemia and preleukaemia. 7, 8 Although recent studies suggest that cells capable of propagating cALL in xenografts are plastic and extend beyond the Stem/B compartment, 9 its leukaemia-specificity provides a readily trackable biomarker in patients.