Leukemic Engraftment In NOD.SCID Mice Is Correlated With Clinical Parameters and Predicts Outcome In Human AML

The NOD.SCID xenotransplantation assay has emerged as a key model system for interrogating the biology of human acute myeloid leukemia (AML). Previous work has established that approximately 50% of AMLs can generate grafts resembling the patient’s leukemia in immunodeficient mice. An understanding of the biologic properties of AML that enable engraftment in xenotransplant assays may aid in the development of prognostic tools and individualized therapy.

To investigate these properties in a broad cross-section of human AML, we transplanted bone marrow (BM) or peripheral blood (PB) cells from 307 AML patients intra-femorally into sublethally irradiated NOD.SCID mice pre-treated with an anti-CD122 antibody. Human engraftment in the recipient BM was assessed 8-10 weeks post-transplant via flow cytometry. AML xenografts, defined as a human graft composed of >90% myeloid (CD33+CD19-CD45+) cells, were generated by 134 samples (44%). The remainder (hereafter referred to as non-engrafting samples) generated 3 patterns of non-leukemic engraftment: no human graft, defined as <0.5% CD45+ cells (85 samples, 28%), isolated CD3+CD45+ T-cell engraftment (27 samples, 9%), and multilineage human engraftment, comprised of CD33+CD45+ myeloid cells with at least 10% CD19+CD45+ B-cells (61 samples, 20%). Using this classification scheme, 105 of 264 diagnostic patient samples (40%) generated an AML graft, compared to 29 of 43 relapse samples (66%; P=0.0007). Within our cohort there were 11 pairs of presentation and relapse samples. In 2 patients, the diagnostic sample did not generate a leukemic graft while the relapse sample did; engraftment patterns were congruent in the remainder. Among the diagnostic samples, leukemic engraftment was associated with a higher PB white cell count (mean of 92 x10⁹/L in engrafters vs. 67 x10⁹/L in non-engrafters; P=0.01), but there was no correlation with BM blast percentage (mean of 75% vs. 72%; P=0.38). A strong relationship between cytogenetically defined prognostic groups and engraftment ability was observed: AML xenografts were generated by 4/30 samples (13%) with favorable karyotypes, 63/153 (41%) with intermediate karyotypes and 23/43 (53%) with adverse karyotypes (P=0.002). No diagnostic samples harboring PML-RARα translocations were capable of initiating leukemic engraftment (n=10). Among patients with normal karyotype (NK) AML, the presence of an NPM1 mutation was not related to engraftment ability; however, 17 of 34 diagnostic samples (50%) from FLT3-ITD-positive patients established leukemic xenografts, compared to 11 of 41 (27%) NK FLT3-ITD-negative samples (P=0.04).

Of note, AML engraftment ability in NOD.SCID mice was strongly associated with a poor response to standard induction therapy. Complete remission was achieved in only 30 of 59 (51%) patients whose diagnostic samples established AML xenografts, compared to 82 of 103 (80%) non-engrafting samples (P<0.0001). Consistent with this observation, both overall survival (OS) and event-free survival (EFS) post-induction were lower in patients whose diagnostic samples were capable of engraftment compared to non-engrafters (median OS of 9.8 vs. 28.5 months; P<0.0001; median EFS 3.0 vs. 12.5 months; P=0.0001). Similar findings were noted among individuals with NK AML, with AML engrafters having a median OS of 11.4 months vs. 54.1 months in non-engrafters (P=0.002) and a median EFS of 6.3 vs. 15.2 months (P=0.002).

Thus, the NOD.SCID xenotransplant assay identifies a subset of AML patients with aggressive disease that responds poorly to standard induction therapy. Further in depth genomic and transcriptional analysis of our dataset will provide insight into the underlying determinants that link xenograft repopulation characteristics with clinical properties and identify those that could be developed into novel prognostic and predictive signatures.