MiR-139-5p is a potent tumor suppressor in adult acute myeloid leukemia

Hematopoiesis depends on a tightly controlled balance of selfrenewal, proliferation, cell death and differentiation. Although the disturbance of this equilibrium creates a predisposition to leukemogenesis, targeted manipulation or modulation can in turn lead to therapeutic advances. In addition to chromosomal aberrations and frequently mutated genes such as NPM1 and FLT3, dysregulation of microRNAs (miRNAs) is now recognized as having an important role in leukemogenesis. Distinct miRNA expression profiles can classify AML subgroups and miRNAs have recently emerged as novel therapeutic targets in hematopoietic malignancies. Building on our previous efforts to resolve potential involvement of miRNAs in hematopoiesis, we identified miR-139-5p as a myeloid-specific miRNA with expression being restricted to neutrophils and macrophages (Supplementary Figure S1A). This distinct expression profile during normal hematopoiesis pointed towards a possible deregulation in malignant hematopoiesis. In order to relate miR-139-5p expression to distinct genetic subgroups and clinical outcome in adult AML patients, we analyzed the published miRNA sequencing dataset from The Cancer Genome Atlas (TCGA). A trend for prolonged overall survival (OS) was observed for patients with miR-139-5p levels above the median independent of cytogenetic subgroup (P= 0.07, Figure 1a left panel), whereas AML patients with normal karyotype (CN-AML) exhibited a significantly better OS (P= 0.02, Figure 1a right panel). Our findings were further corroborated in CN-AML by comparing the expression quartiles, demonstrating that low miR-139-5p expressors (q3 and q4) are associated with an unfavorable OS (Supplementary Figure S1B). In addition, no effect could be demonstrated for the less abundant miR-139-3p (Supplementary Figure S1C), indicating that miR-139-5p is the active strand. Our findings are in line with Emmrich et al., who associated high miR-139-5p levels with a favorable outcome in pediatric AML. Further analysis of the TCGA dataset revealed a significant downregulation of miR-139-5p in CN-AML patients harboring mutated FLT3 (P= 0.03 and P= 0.001, respectively) compared with the total AML population. (Figure 1b left panel). Considering the poor prognosis of CN-AML patients with mutated FLT3, our findings reinforce the unfavorable outcome of CN-AML patients with low miR-139-5p levels. In contrast, miR-139-5p was upregulated in patients carrying an inv(16) (P= 0.001) translocation, a prognostically favorable AML subgroup (Figure 1b left panel). No differences were found for miR-139-5p levels in AML patients with t(15;17) and MLL rearrangements, indicating that miR-139-5p levels are associated with specific genetic subtypes rather than with the degree of differentiation. These findings were validated by qRT-PCR in an independent cohort consisting of 49 adult AML patients and 4 healthy donors of total bone marrow (bm) referred to as the Ulm cohort with uniform treatment procedure. All patient samples from the Ulm cohort were collected from adult patients enroled on German-Austrian AML Study Group (AMLSG) treatment protocols for younger adults (AMLSG-HD98A (NCT00146120) and AMLSG 07-04 (NCT00151242)) and comprised Ficoll gradient purified mononuclear cells mainly from diagnostic bm samples with blast counts of ~ 80% in all analyzed cases. In line with the TCGA dataset, patients harboring mutated FLT3 displayed the lowest miR-139-5p expression among the various cytogenetic subtypes in the Ulm cohort (P= 0.03, Figure 1b right panel). Decreased expression of miR-139 in NPM1-mutated CNAML patients, has been previously reported by Garzon et al. We observed a similar trend within the TCGA dataset (P= 0.06) (Figure 1b left panel), highlighting the robustness of both the datasets. Taken together, we have demonstrated in two independent AML patient cohorts that decreased miR-139-5p levels are associated with mutated FLT3 and with a lower OS in CN-AML in the TCGA dataset. However, it remains to be determined if miR-139-5p is directly or indirectly transcriptionally regulated through activated FLT3. The distinct expression patterns of miR-139-5p in CN-AML imply a tumor suppressor role for this miRNA. In order to further explore the therapeutic potential of miR-139-5p, we used a syngenic bm transplantation AML model based on the transformation of murine bm cells through the combined retroviral overexpression of Hoxa9 and Meis1 (Hoxa9/Meis1), which causes a rapid AML in vivo compared with the pre-leukemic Hoxa9 cells that lead to a long latency AML. Upregulation of HOXA9 and MEIS1 is frequently observed in NPM1-mutated CN-AML and enhances FLT3 signaling. Based on the expression pattern of miR-139-5p in AML patients, we further hypothesized that miR-139-5p would be downregulated in Hoxa9/Meis1 cells compared with untransformed wild-type bm cells. Indeed, miR-139-5p was significantly downregulated in Hoxa9/Meis1 cells compared with untransduced bm cells as measured by qRT-PCR (Supplementary Figure S1D), highlighting the relevance of this in vivo model. To validate the putative tumor suppressor function of miR-139-5p, its expression levels were restored through ectopic lentiviral coexpression of miR-139 and GFP as reported in Hoxa9/Meis1-transformed bm cells. Hoxa9/Meis1/miR-139-5p bm or Hoxa9/Meis1 cells overexpressing an empty control vector (miR-ctrl) were FACS-sorted for GFP, followed by transplantation of 2 × 10 cells per arm into lethally irradiated recipient mice. Restoration of miR-139 levels significantly delayed Hoxa9/Meis1-mediated leukemogenesis (P= 0.0003, Figure 2a) in two independent experiments, highlighting its tumor suppressor activity in a primary transplantation model. Bm cells of deceased mice retained overexpression of miR-139-5p compared with the control arm as quantified by qRT-PCR (Supplementary Figure S1E). Lentiviral expression of miR-139-5p in Hoxa9/Meis1 cells restored miR-139-5p at similar levels compared with untransduced bm cells (Supplementary Figures 1D and E), minimizing possible artefacts of ectopic expression. Notably, we detected no endogenous expression of miR-139-3p in the Hoxa9/Meis1/miR-ctrl bm and overexpression of miR-139 generated lower levels of miR-139-3p in comparison with miR-139-5p, highlighting miR-139-5p as the active strand (Supplementary Figure S1F). Recently, Gibbs et al. demonstrated that Hoxa9/Meis1 cells harbor three, immunophenotypically distinct compartments with varying tumor-initiating activity. Therefore, bm cells of deceased mice were analyzed by flow cytometry for their respective immunophenotype (Figure 2b). We found a significant reduction Citation: Blood Cancer Journal (2016) 6, e508; doi:10.1038/bcj.2016.110