Distinct Properties of Leukemia Stem Cells in Primary Refractory Acute Myeloid Leukemia

Introduction: Acute Myeloid Leukemia (AML) patients who are refractory to induction chemotherapy (RF) have a dismal prognosis. Properties of AML cell populations that cause the refractory phenotype are poorly understood. It is postulated that the leukemia stem cell (LSC) pool promotes chemotherapy resistance in AML; yet whether differences exist in this pool as a function of eventual remission status, and thus may account for induction failure, is unknown. Our group has recently found that in AML patients who initially achieve complete remission (CR) and then relapse, the LSC pool evolves to become larger and more immunophenotypically diverse at the time of relapse [Blood 2013; 122(21): abstract #883]. Since relapsed AML patients are often refractory to salvage chemotherapy, we postulated that diagnostic specimens from RF AML patients would have enlarged, immunophenotypically-diverse LSC pools relative to AML patients achieving CR, implicating these properties in the refractory phenotype. Defining unique properties of LSCs in RF AML might provide greater insight into the mechanisms responsible for the treatment-resistant phenotype.

Methods: RF AML patients were defined by having ≥ 6% leukemic blasts in their bone marrow or peripheral blood after 1 or 2 cycles of anthracycline/cytarabine-based induction chemotherapy. All studies were conducted with pre-treatment AML specimens on approved IRB protocols at the University of Rochester Medical Center. LSC frequency was determined by limiting dilution analysis and xenotransplantation into sublethally irradiated NOD scid gamma (NSG) mice. To determine immunophenotypically-defined cell populations harboring functional LSC activity, AML specimens were sorted into four distinct populations defined by CD34 and CD38 staining. The presence of LSC activity in the four resultant populations (CD34+/CD38-, CD34+/CD38+, CD34-/CD38+, and CD34-/CD38-) was determined by the ability of each sorted population to engraft NSG mice in primary and secondary transplantation experiments. RNA-seq analysis was conducted on LSC-enriched cell populations.

Results: Specimens from RF AML patients were more likely to engraft NSG mice relative to those from AML patients achieving CR (6/6 vs. 13/29 in RF vs. CR patients, respectively; P=0.02). The LSC pool from six RF AML patients was studied in detail. LSC frequency ranged from 1/19 to 1/326,123. Four of the 6 RF specimens harbored LSC frequencies 6- to 7,700-fold higher than those previously reported in primary AML [Sarry et al. J Clin Invest 2011; 121(1384-395]. In 5/6 RF specimens, LSCs were present in more than one immunophenotypically-defined cell population. Relative to specimens from AML patients achieving CR, specimens from RF patients were more likely to harbor LSCs in both CD34+ and CD34- populations (P=0.04). Treatment of NSG mice xenografted with specimens from RF patients with anthracycline/cytarabine-based chemotherapy failed to eradicate disease from engrafted mice; thus, therapeutic outcome of NSG mice engrafted with specimens from RF patients resembles that seen in patients. To identify genes potentially driving the refractory LSC phenotype, we compared gene expression profiles of CD34+ cells from RF patients to those from AML patients achieving long-term remission and to those from normal donors. Ninety-nine genes were uniquely deregulated in RF LSCs, including numerous Homeobox transcription factors and components of the Wnt and Hedgehog signaling pathways, all implicated in the maintenance of stemness. Pathway analysis revealed that amino acid metabolic pathways critical to other treatment-resistant cancers and molecules involved in IL-1 signaling and implicated in metastasis of solid tumors are dysregulated in primary refractory LSCs.

Conclusions: This study is the first to systematically analyze the LSC pool in primary refractory AML. Similar to our findings in relapsed AML, RF AML patients harbor enlarged, phenotypically diverse LSC pools relative to AML patients achieving CR. Challenge of leukemic mice with an anthracycline/cytarabine treatment regimen mimics the effect seen in patients, facilitating the development of patient-derived xenograft models of RF AML. We identify a refractory LSC gene signature, opening the door to future mechanistic investigations and novel therapeutic approaches for this AML patient population in great need.