Engraftment Patterns in NOD. SCID Mice Predict Outcome in Human AML

The NOD. SCID xenotransplantation assay is a key model system for interrogating the biology of leukemic stem cells (LSCs) in human acute myeloid leukemia (AML). Approximately 50% of AMLs can generate human grafts in immunodeficient mice that recapitulate the features of the parent sample. However, some AML samples generate non-leukemic grafts upon xenotransplantation. We recently reported that multilineage (ML) grafts, comprised of both B-lymphoid and myeloid cells, are generated by preleukemic hematopoietic stem cells (preL-HSCs; Shlush et al. Nature 2014). PreL-HSCs contain a subset of the mutations present in leukemic blasts and have a competitive growth advantage over wildtype HSCs leading to clonal expansion in vivo , yet retain multilineage differentiation capacity. To investigate the relationship between patient outcomes and the biological properties of LSCs and preL-HSCs as reflected by different engraftment patterns, we transplanted 272 diagnostic patient samples, representing a broad cross-section of adult AML, into sublethally irradiated NOD. SCID mice by intrafemoral injection. Human chimerism was assessed 8-10 weeks post-transplant by flow cytometry. 41% of samples generated AML xenografts, defined as a human graft containing >90% myeloid (CD33+CD19-CD45+) cells. Three patterns of engraftment were seen with the remaining samples: no human graft, defined as 10% CD19+CD33-CD45+ B-cells (22%). Among patients whose samples generated ML grafts compared to AML grafts or TC/no graft, secondary AML was less common (10% vs. 27% vs. 26%, respectively; P=0.03). Associations between engraftment pattern and other baseline clinical characteristics, including age, white blood cell (WBC) count and cytogenetics, did not reach statistical significance in this cohort. However, there was a strong correlation between AML engraftment capacity and response to standard induction chemotherapy. AML engrafters had lower complete remission (CR) rates compared to all other patients as a group (51% vs. 81%; P 16.6%) vs. low (≤16.6%) levels of human chimerism (median OS of 8.2 vs. 14.5 mos; P=0.001). The ability of a diagnostic AML sample to generate ML xenografts likely reflects a high frequency of preL-HSCs, postulated to be a source for relapse. Indeed, despite comparable CR rates in patients whose samples generated ML vs. TC/no graft (85% vs. 77%; P=0.28), the relapse-free survival of ML engrafting patients was shorter (median of 19.3 vs. 50.2 months; P=0.04). The cumulative incidence time-to-relapse (TTR) of ML engrafters (median TTR 19.7 mos) was also less than that of patients whose samples generated TC/no graft (median TTR not reached; P=0.01) and comparable to that of AML engrafters (9.6 mos; P=0.11). Thus, the presence of significant numbers of preL-HSCs in diagnostic AML samples, as evidenced by the generation of ML xenografts, is associated with earlier relapse, consistent with our prior finding that preL-HSCs persist in remission and thus may serve as a reservoir for clonal evolution. Interestingly, despite the shorter time-to-relapse, the OS of patients whose samples generated ML grafts was comparable to those who generated TC/no grafts (median 30.4 vs. 27.1 mos, P=0.95), and significantly longer than those who generated AML grafts (10.6 mos; P de novo disease and responds well to re-induction. Thus, the NOD. SCID xenotransplantation model captures functional properties of LSCs and preL-HSCs that are clinically relevant, validating its use to study the biology of these disease-sustaining cell populations and evaluate novel AML therapies. Disclosures No relevant conflicts of interest to declare.