Identification of Existing Bioactive Compounds That Target Acute Myeloid Leukemia Stem Cells

Current therapies for acute myeloid leukemia (AML) are able to achieve complete remission in 80-85% of patients; however large portions of these cases relapse. The 5-year survival rate is estimated at 30%. Thus, the development of more efficient approaches to treat AML is absolutely needed. Similarly to normal hematopoiesis, leukemias are organized as cellular hierarchies supported by leukemic stem cells (LSCs). LSCs are relatively resistant to chemotherapy and may act as a cancer cell reservoir that leads to relapse. The discovery of therapeutics that can eradicate these cells is essential to achieve complete patient recovery.

We hypothesize that by using an in silico strategy and stem cell gene expression data we can identify bioactive compounds that target LSCs with minimal impact on normal hematopoietic stem cell (HSC) function. We had previously generated human LSC and HSC expression signatures from 16 primary AML and 4 independent pooled cord blood samples that were prognostic for patient survival (Eppert et al Nat Med 2011). By interrogating existing gene expression profiles of drug response with our LSC and HSC signatures, as well as additional novel signatures, we identified 92 molecules predicted to inhibit human LSC properties without harming HSCs. Nearly 60% of the positive hits were anti-infectives, psychotropics, hormones and steroids, anticancer or cardiovascular therapies. In support of the efficacy of this bioinformatic approach, we identified known chemotherapeutic agents such as azacitidine, daunorubicin and doxorubicin, although these were predicted to harm HSCs. We assessed the anti-LSC activity of one sub-group known to target a class of G-protein-coupled receptors (GPCR) to further validate the list of candidate compounds. Using a primary human AML sample (8227) with known LSC phenotype (CD34+), we tested 2 candidate anti-GPCR compounds (hits A and B) and observed that they were highly effective in targeting both bulk cells and LSCs. The IC50 for bulk cells was 3.4 µM and 4.2 µM for compounds A and B, respectively (A - 6 day treatment, B - 2 day treatment). The IC50 of the CD34+ cell fraction, which included stem and progenitor cells, was 5.9 µM and 4.1 µM, respectively. Thus, these compounds are effective at targeting both bulk and LSC-enriched AML cell populations. The efficacy of these predicted compounds against the primitive cells within the CD34+ population was assessed using colony formation assays on human AML. 8227 cells treated with compound B resulted in significantly fewer colonies compared to the DMSO control (3.8 ± 3.0 vs 13.3 ± 2.8; at 4.5 µM). To test whether our candidate compounds target leukemic cells and not normal blood cells, we examined 3 additional primary human AML samples and a cord blood sample. There was only a minor effect on normal cells, including stem and progenitor cells, but significant cell death in all leukemia samples (70% and more at ≥ 3 µM). Next, we will determine the effectiveness of our compounds in vivo using established xenografts of primary human AML samples and cord blood samples.

In conclusion, using our in silico approach we successfully identified novel anti-LSC compounds that have efficacy in vitro with minimal toxicity on primary HSCs. Furthermore, as each predicted compound is associated with specific molecular pathways, our approach has improved our understanding of LSC biology by acting as a screen for LSC-related pathways. In the near future, we will continue investigating the 92 candidate compounds to identify, develop, and transition a novel anti-LSC compound into clinical use.