MiR-10a and HOXB4 are overexpressed in atypical myeloproliferative neoplasms

DNA and RNA were obtained from total peripheral blood leucocytes (PBL) for aMPN, CML, RHL and healthy donors, or bone marrow mononuclear cells (BMMC) for AML.

Cyanine-3 (Cy3) labeled miRNA was prepared from 0.2 μg RNA using the miRNA complete labeling kit version 2.2 (Agilent Technologies). For each sample, the labeled miRNAs were hybridized overnight at 55 °C onto Human miRNA Microarray V2 (Agilent Technologies). Normalized data for all samples have been deposited in NCBI Gene Expression Omnibus and are accessible through GEO Series accession number GSE75666.

10 ng DNA were used for highly multiplex amplification of DNMT3A (exons 15–23), IDH1/2 (exon 4), ASXL1 (exon 12), TET2 and EZH2 (whole coding region) with an AmpliSeq™ panel (Thermo Fischer Scientific) before sequencing on an Ion Torrent PGM (Life Technologies) with 314 chips.

For pharmacological experiments, AML cell lines were used as model of myeloid diseases. U937 (monocytic cell line), KG1a (promyeloblast cell line) and OCI-AML3 (myelomonocytic cell line) (DSMZ) were treated by 5-aza-2’deoxycytidine (DEOX) (Sigma-Aldrich) and/or valproic acid (VPA, Calbiochem) and/or retinoic acid (RA). K562 cells (erythroleukemia cell line) were used for proliferation assays. CD34⁺ cells were cultured in Stem Span SFEM (Stemcell Technologies), MethoCult H4534 Classic without EPO (Stemcell technologies) or on MS5 cells in Myelocult H5100 (Stemcell Technologies). When indicated, transduced cells were selected by puromycin treatment (Sigma-Aldrich).

For miR-10a over-expression, the pri-miRNA precursor was cloned into plasmids carrying a puromycin resistance or a GFP gene, under the control of a MND promoter. Wild Type and R882H mutated DNMT3A cDNA were obtained from pMYs vectors previously reported [25] and re-cloned into lentiviral vectors carrying a puromycin resistance gene, under the control of a EF1alpha promoter. Lentiviral vectors were produced by the vectorology platform of the University of Bordeaux.

Cellular suspensions were incubated with different panels of fluorochrome-labelled antibodies and analyzed on a FACS CANTO II (Beckton Dickinson). Data were analyzed with FacsDiva software.

Three sets of experiments were conducted, each with 10 female NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) mice for each condition. Mice were bred under specific pathogen-free conditions and experiments were performed in the animal housing facility of the University of Bordeaux in conformity with the rules of the Institutional Animal Care and Use committee (Ministerial approval number 00048.2). CD34⁺ cell engraftment protocol is detailed in the Additional file 1: Supplementary Material.

Student’s t-test and Mann and Whitney test have been performed using GraphPad Prism. More details about these methods are available in the Additional file 1: Supplementary Material.

In order to discover micro-RNAs specifically dysregulated in aMPN compared to CML with similar blood leukocyte and differential counts, a total of 18 aMPN (Patients 1 to 18 in Additional file 2: Table S1) and 10 CML samples were hybridized on Agilent miRNA microarrays. An univariate analysis identified genes that were differentially expressed in the two groups. Using a stringent significance level for each univariate test of 0.001, 3 miRNAs were characterized as differentially expressed in aMPN versus CML (miR-155: p = 7.3 × 10^− 6, FDR = 0.006; miR-10a: p = 4.2 × 10^− 4, FDR = 0.132 miR-26a: p = 4.8 × 10^− 4, FDR = 0.132, Fig 1a). These 3 miRNAs were also found as differentially expressed by a SAM (significance analysis of microarrays) using a target proportion of false discovery of 0.05 (not shown). Among them, miR-10a had the highest fold-change (4.8). We then focused our analyses on the overexpression of this microRNA. In CML patients, miR-10a expression has been reported lower than in healthy donors [11]. We confirmed microarrays results by comparing leukocytes from aMPN (n = 18) to 10 patients with CML at diagnosis, 14 reactive hyperleukocytosis (RHL) and 7 healthy donors (HD) by specific RT-qPCR. The significant overexpression of miR-10a in the aMPN group compared to the three others was confirmed (Fig. 1b).

Since mutations in epigenetic regulators have been described in aMPN, and because demethylating agents have proved efficient in treating myelodysplastic syndromes, we used myeloid cell lines to investigate whether miR-10a expression could be regulated by epigenetic mechanisms in hematopoietic cells. MiR-10a sitting immediately upstream of HOXB4, we studied the expression of both RNAs. In KG1a cells, treatment by valproic acid (VPA) at dose of 1 mM for 48 h induced an overexpression of miR-10a with a stronger effect when VPA was combined with 2 μM 5-aza-2’deoxycitidine (DEOX). On the contrary, 2 μM retinoic acid (RA) alone or in combination was inefficient at increasing miR-10a expression (Fig. 2a). Regarding HOXB4, DEOX alone, and to a larger extent in combination with VPA, also increased its expression. In contrast to miR-10a, the addition of RA had an additional effect, suggesting that the optimal increase in HOXB4 expression due to RA requires an optimal chromatin opening (Fig. 2b). This epigenetic regulation was variable depending on the AML cell line: in NPM1- and DNMT3A-wild type U937 cells, treatment by 1 mM VPA alone did not induce any modification in HOXB4 or miR-10a levels whereas 2 μM DEOX alone did induce overexpression of both transcripts. This effect was potentiated by VPA for HOXB4, but not for miR-10a, and combination of the three drugs induced a 1.5 to 3-fold overexpression compared to DEOX alone, for both transcripts (Fig. 2c-d). Finally, in the NPM1- and DNMT3A-mutated cell line OCI-AML3, treatment with 2 μM DEOX did not induce HOXB4 overexpression, whereas 1 mM VPA induced a 3 to 4-fold overexpression. This effect was not potentiated by 2 μM RA or 2 μM DEOX, but the association of the three drugs induced an 8-fold overexpression of HOXB4, significantly superior to that observed with VPA alone (Additional file 3: Figure S1A). These data suggest that the level of basal expression of miR-10a and HOXB4 depends on epigenetic regulation in these myeloid cell lines (Additional file 3: Figure S1B) and their drug sensitivity profile seems to be variable depending on their mutational status.

Having established the epigenetic regulation of HOXB4/MiR-10a expression, we wondered whether the expression level of these genes may correlate with the mutational status of epigenetic modifiers frequently mutated in hematological malignancies. Six epigenetic regulators were sequenced in 39 patients with myeloid malignancies (Additional file 2: Table S2) and levels of HOXB4 and miR-10a were measured. We observed a significantly higher HOXB4 and miR-10a expression in patients with DNMT3A mutations while TET2, ASXL1, IDH1/2, and EZH2 mutational status did not influence HOXB4 nor miR-10a expression level (Fig. 3a and data not shown). Although this validation cohort did not include aMPN patients, taken together our data strengthen the evidence of an epigenetic regulation of miR-10a and HOXB4. In order to confirm the impact of DNMT3A mutational status on miR-10a expression, we transduced primary hematopoietic CD34⁺ cells with lentiviruses carrying the wild type or the R882H mutant DNMT3A cDNA. As previously reported for HOXB4 [25], the expression of miR-10a was significantly higher in the R882H DNMT3A compared to the wild type DNMT3A condition (Fig. 3b).

The role of HOXB4 in stem cell expansion in hematopoiesis being well documented [26‐28], we explored the role of miR-10a overexpression in the main functions of hematopoietic cells relevant to leukemogenesis. Firstly, we transduced K562 and U937 cell lines with a vector carrying the pri-miR-10a precursor and evaluated the impact of miR-10a overexpression on cell proliferation. Despite an overexpression by a mean factor of 5 and 45 for K562 and U937 respectively, we did not observe any effect on cell proliferation (Additional file 3: Figure S2). To evaluate cellular properties more related to leukemogenesis such as differentiation or self-renewal, we used the same construction to force overexpression of miR-10a in human primary CD34⁺ cells. Impact on late hematopoiesis was analyzed by culturing the cells for 3 weeks in liquid medium. Despite a mean 13.8-fold increased expression as assessed by qRT-PCR (Fig. 4a), similar to that observed in aMPN patients, miR-10a overexpression did not alter cell proliferation (Fig. 4b). At days 7, 14 and 21, differentiation was evaluated by flow cytometry and clonogenic capacities. We did not observe any impact of miR-10a overexpression on the level of granulocytic, erythroblastic or megakaryocytic differentiation markers (Fig. 4c), nor on the clonogenic capacities of primary cells (Fig. 4d). Similar results were observed when CD34⁺ cells were transfected with a synthetic miR-10a precluding any abnormal processing of the pri-miR-10a precursor (Additional file 3: Figure S3). Effects of miR-10a overexpression on more immature compartments of hematopoiesis were assessed in vitro by cultures on a stromal cell layer (LTC-IC assay). After 5 weeks, clonogenic capacities of transduced cells were evaluated. As in liquid culture experiments, miR-10a overexpression did not alter differentiation nor self-renewal of primary cells (Fig. 4e). Finally, the roles of miR-10a in hematopoiesis were analyzed in xenograft experiments. CD34⁺ cells were transduced with the pri-miR-10a or the empty vector both co-expressing GFP and injected into the retro-orbital sinus of NSG mice. Twelve weeks after engraftment, bone marrow cells were analyzed by flow cytometry. MiR-10a was efficiently over-expressed with a mean 5.6-fold increased expression compared to controls. Meanwhile, we did not observe any impact on engraftment efficiency, self-renewal (determined by the proportion of human CD34⁺ cells among GFP⁺ cells), cell proliferation (as assessed by the proportion of GFP⁺ cells among human CD45⁺ cells) (Fig. 5a) nor cell differentiation (Fig. 5b). Effects on self-renewal were further analyzed by secondary engraftment experiments. Once again, we did not observe any effect of miR-10a over-expression on the level of the different markers compared to controls (Additional file 3: Figure S4).