Deregulated expression of miR-29a-3p, miR-494-3p and miR-660-5p affects sensitivity to tyrosine kinase inhibitors in CML leukemic stem cells

The development of Imatinib mesylate (IM), which targets the oncogenic BCR-ABL fusion protein, has greatly improved the outcome of Chronic Myeloid Leukemia (CML) patients. However, BCR-ABL–positive progenitors can be detected in CML patients in complete cytogenetic response. Several evidence suggests that CML stem cells are intrinsically resistant to Tyrosine Kinase Inhibitors (TKI), and therefore they represent the most likely candidate responsible for disease relapse. In this work, we investigated the microRNA (miRNA) expression profile of different subpopulations of CML Leukemic Stem Cells (LSCs): Lin-CD34+CD38- and Lin-CD34-CD38- cells. These cell fractions have been previously shown to be endowed with TKI intrinsic resistance. Our analysis identified 33 common deregulated miRNAs in CML LSCs. Among those, 8 miRNAs were deregulated in CML independently from BCR-ABL kinase activity and therefore are likely to be involved in the BCR-ABL-independent resistance to TKI that characterizes CML LSCs. In particular, the up-regulation of miR-29a-3p and miR-660-5p observed in CML LSCs, led to the down-regulation of their respective targets TET2 and EPAS1 and conferred TKI-resistance to CML LSCs in vitro. On the other hand, miR-494-3p down-regulation in CML LSCs, leading to c-MYC up-regulation, was able to decrease TKI-induced apoptosis. These results demonstrate that aberrant miRNA expression in CML LSCs could contribute to the intrinsic TKI-resistance observed in these cell populations, and support the development of novel therapies aimed at targeting aberrantly regulated miRNAs or their targets in order to effectively eradicate CML LSCs.


INTRODUCTION
Chronic Myeloid Leukemia (CML) is a stem cellderived malignant disorder characterized by the reciprocal translocation t (9;22) leading to the BCR-ABL fusion gene. The encoded 210 KDa protein has constitutively elevated tyrosine kinase activity and drives Hematopoietic Stem Cells (HSCs) transformation [1]. The BCR-ABL Tyrosine Kinase Inhibitors (TKIs), such as Imatinib mesylate, are the standard therapy for CML patients. Imatinib has been shown to induce complete molecular response in more than 80% of newly diagnosed patients in chronic phase [2]. However, it has been unable to completely eliminate BCR-ABL-expressing leukemic cells [3]. In order to achieve durable remission, CML patients require lifelong treatment with TKIs, at considerable cost and despite harmful side effects. There is an increasing body of evidence suggesting that TKIs are not effective in killing leukemic stem cells (LSCs) in CML [4,5]. In particular, Corbin et al. demonstrated that CML LSCs survive IM treatment independently of BCR-ABL kinase activity [4]. Recently, our group identified a novel quiescent LSCs subset (Lin-CD34-CD38-), endowed with stem cell properties and intrinsically resistant to TKI treatment both in vitro and in vivo [6]. It is therefore clear that a definitive treatment for CML requires the elimination of LSCs. Thus, gaining further understanding on the molecular and functional properties of the stem cell compartment in CML is mandatory for the development of more effective therapies that will eliminate TKI-resistant LSCs.
MicroRNAs (miRNAs) are small non-coding RNAs that control gene expression and play an important role in several biological processes such as differentiation [7], proliferation [8], and apoptosis [9]. In the last few years, increasing evidence shows that miRNAs expression is deregulated in both solid and hematological malignancies [10,11] and that deregulated miRNAs can induce and/or maintain a leukemogenic state.
In this study, we performed miRNA expression profiling (miEP) of Lin-CD34+CD38− and Lin-CD34-CD38-cells isolated from 5 CML patients and 4 healthy donors. This analysis identified a set of miRNAs aberrantly expressed in CML LSCs. In order to identify those miRNAs involved in the LSC-specific TKI escape, miRNAs whose expression is deregulated in CML independently from BCR-ABL kinase activity were selected. Our analysis allowed us to identify three novel miRNA/mRNA networks that confer BCR-ABL-independent TKI resistance to CML LSCs.

miRNA expression profiling of CML Lin-CD34-CD38-and Lin-CD34+CD38− cells
In order to shed light on the molecular properties of the CML stem cell compartment, we performed miEP on Lin-CD34-CD38-and Lin-CD34+CD38− cells from 5 CML patients and 4 healthy donors. To explore the relationships between samples, we performed a Principal Component Analysis (PCA). Figure 1A shows that the CML samples clustered together and were clearly separated from control samples. Of note, PCA revealed that CML Lin-CD34-CD38are closer to leukemic CD34+CD38+ and normal CD34+ subfractions whereas their normal counterparts cluster separately, in agreement with our previous findings on the gene expression profile [6]. Next, differentially expressed miRNAs (DEMs) in the comparison CML vs normal donors for each cell population were identified by two-tail unpaired t-Student test as described in Materials and Methods (Supplementary Table 2 and Supplementary Table 3). This analysis identified 134 DEMs in the comparison CML Lin-CD34-CD38-vs Normal Donor Lin-CD34-CD38− and 102 DEMs in the comparison CML Lin-CD34+CD38− vs Normal Donor Lin-CD34+CD38−. 33 miRNAs resulted concordantly deregulated in both cell populations ( Figure 1B, Table 1, Supplementary Tables 5 and 6).

Identification of BCR-ABL-independent deregulated miRNAs
miRNAs can be deregulated in CML through either a BCR-ABL kinase-dependent or a kinase-independent mechanism. To identify those miRNAs that were aberrantly expressed in CML LSCs through a BCR-ABL-independent mechanism, we selected miRNAs whose expression did not change after IM treatment. To this end, the BCR-ABL-positive K562 cells were treated with increasing doses of IM and the expression level of selected miRNAs was measured. For this analysis, 11 miRNAs, among the common 33 DEMs identified above, were selected based on their Fold Change (FC) in the comparison LSCs vs HSCs, and on their biological significance, evaluated as biological function, and expression level in other Myeloproliferative Neoplasms from our database [12]. The effectiveness of IM treatment on BCR-ABL kinase activity was assessed by measuring CrkL phosphorylation (p-CrkL) by flow cytometry. As shown in Figure 1C, the intracellular p-CrkL levels were significantly decreased in IM-treated samples compared to the untreated control. 8 miRNAs (miR-660-5p, miR-365-3p, miR-29a-3p, miR-19b2-3p, miR-494-3p, miR-204-5p, miR-486-5p and miR-486-3p) showed no statistically significant difference in their expression level after IM treatment ( Figure 1D, 1E). These miRNAs were therefore classified as deregulated in CML with a BCR-ABL independent mechanism ( Figure 1F).

miR-29a-3p protects CML cells from TKIsinduced apoptosis
In order to assess whether miR-29a-3p up-regulation observed in CML LSCs is involved in the escape from TKIs treatment, miR-29a-3p was overexpressed in K562 www.impactjournals.com/oncotarget cells and its effects on BCR-ABL kinase activity and cell survival in presence of TKIs treatment (1 µM IM  for 24 h or 0.4 µM IM for 48 h, or 10 nM Ponatinib for  24 h, or 100 nM Dasatinib for 24 h) were evaluated. A  significant overexpression of miR-29a-3p was detected by  qRT-PCR 24 hours after last nucleofection (RQ ± SEM, 42.75 ± 6.04, p < 0.01) (Figure 2A). Analysis of p-CRKL levels showed that IM treatment significantly inhibited BCR-ABL kinase activity in all samples tested, regardless of miR-29a-3p overexpression ( Figure 2B). Thus, miR-29a-3p does not directly affect BCR-ABL activity.
Interestingly, flow cytometric analysis of apoptosis performed by PI/Annexin V staining showed a significant decrease in the percentage of apoptotic cells in the sample transfected with miR-29a-3p mimic compared to the Neg CTR mimic, when incubated with IM ( Figure 2C-2E) as well as with second and third generation TKIs ( Figure 2F, Supplementary Figure 3).
Altogether, these data suggest that miR-29a-3p overexpression is able to protect K562 cells from TKIsinduced apoptosis without affecting BCR-ABL kinase activity.
Next, we investigated whether TET2 downregulation was able to reproduce the effects on IM sensitivity observed upon miR-29a-3p overexpression. To this end, TET2 expression was silenced in K562 cells by means of TET2 siRNAs. qRT-PCR analysis confirmed the downregulation of TET2 mRNA in TET2 siRNA sample compared to the control (RQ ± SEM, 0.507 ± 0.07, p < 0.01) ( Figure 3E). Twenty-four hours after last nucleofection, K562 cells were treated with either 1 µM IM for 24 h or 0.4 µM IM for 48 h, and p-CRKL intracellular levels and apoptosis were evaluated. In agreement with the results obtained upon miR-29a-3p overexpression, TET2 silencing did not affect IM inhibition on BCR-ABL kinase activity (Supplementary Figure 2A). Moreover, our data showed a statistically significant decrease in the percentage of apoptotic cells in TET2 siRNA sample compared to the NTsiRNA ( Figure 3F-3H). Altogether, these results demonstrate that TET2 silencing is able to protect cells from IM-induced apoptosis, thus phenocopying the effects mediated by miR-29a-3p overexpression.

miR-660-5p protects CML cells from TKIsinduced apoptosis
Among miRNAs up-regulated in CML LSCs with a BCR-ABL-independent mechanism, our analysis pointed out miR-660-5p. In order to test whether its up-regulation is able to affect TKI sensitivity, K562 cells were transfected with either miR-660-5p mimic or Neg CTR mimic and BCR-ABL kinase activity and cell survival were evaluated by flow cytometric analysis. A significant overexpression of miR-660-5p was detected by qRT-PCR (RQ ± SEM, 115.29 ± 6.28, p < 0.01) ( Figure 4A). Flow cytometric analysis of p-CRKL intracellular level showed no statistically significant difference between cells transfected with Neg CTR mimic and miR-660-5p mimic ( Figure 4B). Moreover, our data showed a statistically significant decrease in the percentage of apoptotic cells in the miR-660-5p mimic sample compared to the Neg CTR mimic after TKIs treatment ( Figure 4C-4F, Supplementary Figure 3). These data demonstrate that miR-660-5p overexpression is able to protect K562 cells from TKIs induced apoptosis without affecting BCR-ABL kinase activity.
Next, EPAS1 expression was silenced in K562 cells by means of transfection with EPAS1 siRNA (EPAS1 siRNA). www.impactjournals.com/oncotarget qRT-PCR analysis confirmed the downregulation of EPAS1 mRNA in EPAS1 siRNA sample compared to the NTsiRNA (RQ ± SEM, 0.337 ± 0.05, p < 0.01) ( Figure 5E). Twentyfour hours after last transfection, K562 cells were treated with either 1uM IM for 24 h or 0.4 uM IM for 48h and cells were analyzed for apoptosis and pCRKL levels. Our results showed that EPAS1 silencing does not affect IM inhibitory effect on BCR-ABL kinase activity (Supplementary Figure  2B). On the other hand, EPAS1 silencing was able to protect K562 cells from IM-induced apoptosis ( Figure 5F-5H). Thus, these data demonstrate that EPAS1 silencing is able to reproduce the effects exerted by miR-660-5p overexpression, suggesting that miR-660-5p is able to protect cells from IMinduced apoptosis through EPAS1 targeting.

miR-494-3p sensitizes CML cells to TKIsinduced apoptosis
Among miRNAs down-regulated in CML LSCs with a BCR-ABL-independent mechanism, our analysis highlighted miR-494-3p. To assess whether the downregulation of miR-494-3p observed in CML LSCs can control response to first, second and third generation of TKIs, we overexpressed miR-494-3p in K562 cells and evaluated apoptosis and p-CRKL levels after TKIs treatment. qRT-PCR showed a significant up-regulation of miR-494-3p after miR-494-3p mimic transfection (RQ ± SEM, 372.89 ± 86.92, p < 0.01) ( Figure 6A). Flow cytometric analysis of p-CRKL intracellular level showed no statistically significant difference between cells transfected with Neg CTR mimic and miR-494-3p mimic ( Figure 6B). These data suggest that miR-494-3p does not directly affect BCR-ABL kinase activity. Moreover, our data showed a statistically significant increase in the percentage of both apoptotic cells and late apoptotic/necrotic cells in miR-494-3p mimic sample compared to the Neg CTR mimic ( Figure  6C-6F, Supplementary Figure 3). These results demonstrate that miR-494-3p overexpression is able to sensitize K562 cells to TKIs-induced apoptosis without affecting BCR-ABL kinase activity.
Next, c-MYC expression was silenced in K562 cells by means of siRNAs targeting c-MYC (MYC siRNA). qRT-PCR analysis confirmed c-MYC downregulation in MYC siRNA sample compared to the control (RQ ± SEM, 0.38 ± 0.01, p < 0.01) ( Figure 7E). In agreement with the results obtained upon miR-494-3p overexpression, c-MYC silencing did not affect IM inhibitory ability on BCR-ABL kinase activity (Supplementary Figure 2C). Moreover, our data showed a statistically significant increase in the percentage of both apoptotic and late apoptotic/ necrotic cells in the MYC siRNA sample compared to the NTsiRNA ( Figure 7F-7H).

DISCUSSION
Treatment of CML has been notably improved in the past 15 years thanks to the development of TKIs blocking BCR-ABL kinase activity. However, the majority of patients achieving remission with imatinib continue to test positive for BCR-ABL transcript by RT-PCR.
Indeed, in recent years, several in vitro evidence suggested that CML LSCs are intrinsically resistant to Imatinib. Chu et al. [5] demonstrated that BCR-ABL+ stem cells persist in the bone marrow of CML patients in prolonged remission on Imatinib treatment. Corbin et al. [4] demonstrated that CML LSCs survive Imatinib with a mechanism independent from BCR-ABL kinase activity.
Notably, in vivo, only approximately 50% of patients who had stopped Imatinib therapy remained long-term disease free, at the molecular level, suggesting, in many cases, the persistence of LSCs [13]. Thus, treatment targeting BCR-ABL does not completely eliminate CML stem cells, and disease eradication will require additional therapeutic strategies. In contrast, Mustjoki et al. recently reported      that in vivo TKI therapy is able to clear most of Ph+ cells from the stem cell compartment [13]. These differences might be due to the different time points applied for LSCs analysis, usually at diagnosis for in vitro studies, at later time points in the work by Mustjoki et al. or to technical limitations that might have prevented the identification of few residual LSCs. Typically, CML stem cells have been defined as lineage-negative CD34-positive CD38negative (Lin-CD34+CD38−) cells. However, our group recently identified a new subset of LSCs negative for CD34 expression, therefore identified as Lin-CD34-CD38cells [6]. Our data demonstrate that approximately onethird of Lin-CD34-CD38-cells are leukemic and capable of engraftment in immunodeficient mice. Moreover, these cells showed intrinsic resistance to IM both in vitro and in vivo. Therefore, Lin-CD34-CD38-CML cells can be seen as a stem cell subset insensitive to targeted therapy that may contribute to disease persistence. In the last years, miRNAs expression has been reported to be deregulated in hematological malignancies [10], where it is responsible for disease development or maintenance.
Based on these premises, in order to gain further understanding on the molecular features of the CML stem cell compartment, we analyzed the miEP of both Lin-CD34-CD38-and Lin-CD34+CD38− LSCs, and miRNAs deregulated in both cell fractions were selected. Our analysis identified 33 differentially expressed miRNAs in the comparison LSCs vs HSCs ( Figure 1B). As Corbin et al. demonstrated that LSCs escape from IM treatment independently from BCR-ABL kinase activity, we sought to identify those miRNAs that were deregulated in CML LSCs with a mechanism that is independent from BCR-ABL. To this aim, BCR-ABLpositive K562 cells were treated with increasing doses of IM and miRNAs, whose expression did not change after IM treatment, were selected and classified as BCR-ABL-independent miRNAs. This analysis unraveled 8 miRNAs, whose expression is deregulated in CML LSCs independently from BCR-ABL ( Figure 1F). Next, in order to assess whether the deregulated expression of such miRNAs was able to affect TKIs-sensitivity, miRNA overexpression experiments were performed. As initial analysis, the function of selected miRNAs was studied in K562 cells, then, in order to validate our results, miRNA overexpression experiments were performed in Hematopoietic Stem/Progenitor Cells (HSPCs) isolated from CML patients collected in chronic phase before therapy. All eight BCR-ABL-independent miRNAs, except miR-486 whose role in CML has been recently described by Dr. Bhatia's group [14], were studied. Our results demonstrated that the overexpression of miR-204-5p, miR-19b-2-5p, and miR-365a-3p do not affect IM response in K562 cells (data not shown). On the other hand, miR-29a-3p, miR-660-5p and miR-494-3p showed a strong influence on TKIs sensitivity when overexpressed in BCR-ABL-positive cells. In particular, miR-29a-3p overexpression was able to protect cells from TKIs-induced apoptosis ( Figure 2C-2F). Of note, IM-induced BCR-ABL kinase inhibition was not affected by miR-29a-3p expression level ( Figure 2B), therefore demonstrating that miR-29a-3p up-regulation is able to confer IMresistance with a BCR-ABL-independent mechanism. By means of 3′UTR luciferase reporter assay, we were able to demonstrate for the first time in the hematopoietic context the direct interaction between miR-29a-3p and TET2 ( Figure 3C). Moreover, TET2 silencing mimicked the effects induced by miR-29a-3p overexpression, suggesting that miR-29a-3p protects cells from IMinduced apoptosis through TET2 targeting. Han YC et al. provided evidence that miR-29a-3p may serve as oncogene during Acute Myeloid Leukemia (AML) development: ectopic expression of miR-29a induce the acquisition of self-renewal capacity by myeloid progenitors, biased myeloid differentiation, and the development of a myeloproliferative disorder [15]. Somatic loss of function mutations in TET2 are frequently observed in myelodysplastic syndromes (MDS), myeloproliferative neoplasms (MPN), and AML [16,17]. Recently, decreased TET2 expression associated with t(4;6;11) rearrangement has been demonstrated to be involved in CML progression. Of note, the appearance of the TET2 deletion was not associated with an increase of the BCR-ABL transcript level, suggesting a BCR-ABL-independent role for this gene in CML progression [18]. Our data strongly suggest that the up-regulation of miR-29a-3p observed in CML LSCs is responsible for TET2 downregulation, which in turn is able to confer IM-resistance to CML LSCs. These findings imply that TET2 could be involved in the pathogenesis of myeloid malignancies through deregulated gene expression in addition to inactivating mutations and provide a rationale for targeted therapies that impact TET2 aberrant regulation.
Among miRNAs up-regulated in CML LSCs with a BCR-ABL-independent mechanism, our analysis pointed out miR-660-5p. miR-660-5p has been previously reported in the context of hematopoiesis as involved in polyploidization and megakaryocytes differentiation [19]. To our knowledge, this is the first report describing a role for miR-660-5p in myeloid malignancies. miR-660-5p overexpression was able to protect cells from TKIs-induced apoptosis ( Figure 4C-4F). IM-induced BCR-ABL kinase inhibition was not affected by miR-660-5p expression level ( Figure 4B), therefore demonstrating that miR-660-5p up-regulation is able to confer IM-resistance to K562 cells with a BCR-ABL-independent mechanism. Between miR-660-5p predicted targets, EPAS1 showed significant downregulation upon miR-660-5p overexpression both at the mRNA and protein level ( Figure 5A and 5D respectively). EPAS1 silencing was able to protect cells from IM-induced apoptosis thus phenocopying the effects induced by miR-660-5p overexpression (Figure  www.impactjournals.com/oncotarget 5F-5H). EPAS1, also known as hypoxia-inducible factor 2 alpha (HIF2A), is a transcription factor critical for the adaptive response to hypoxia. HIF2A expression has been associated with aggressive disease phenotype, metastasis and resistance to therapy in several solid tumors, such as neuroblastoma [20], non-small cell lung cancer [21], and breast cancer [22]. On the other hand, recent evidence showed that in HIF-2a inhibits high-grade soft tissue sarcomas cell growth in vivo [23], suggesting a cell context-dependent role for EPAS1 in tumor development. Our data strengthen this hypothesis, in fact, in the CML context, EPAS1 down-regulation is involved in the escape from IM treatment of BCR-ABL-positve cells.
On the other hand, between miRNAs down-regulated in CML LSCs independently from BCR-ABL activity, our miEP data highlighted miR-494-3p ( Figure 1E, 1F). miR-494 has been reported as an oncogenic miRNA in a variety of cancers including colorectal cancer [24], breast cancer [25], ovarian cancer [26], and prostate cancer [27]. However, in several solid tumors, i.e. oral and gastric cancer, miR-494-3p acts as an anti-oncogene by targeting HOXA10 and c-MYC respectively [28,29]. Our data demonstrated that miR-494-3p overexpression is able to sensitize K562 cells to TKIs-induced apoptosis without affecting BCR-ABL kinase acitvity ( Figure 6C-6F). Thus suggesting that miR-494-3p acts as tumor suppressor miRNA in CML, where its downregulation is necessary for the intrinsic resistance to TKIs that characterizes LSCs. In order to identify the molecular mechanisms governing miR-494-3p-induced IM sensitization, 3'UTR luciferase assay was performed and c-MYC was identified as miR-494-3p target ( Figure 7C). At the time this manuscript was in preparation, Tian C. et al. [30] reported c-MYC downregulation upon miR-494-3p overexpression in AML cell lines. To our knowledge, these are the first two reports demonstrating c-MYC targeting by miR-494-3p in hematological malignancies. Noteworthy, c-MYC silencing was able to sensitize cells to IM-induced apoptosis ( Figure 7F-7H), thus mimicking the effects induced by miR-494-3p overexpression. Recently an interesting dual role for c-MYC in CML has been reported: Reavie L. et al. showed that both depletion and overexpression of c-MYC can severely impact CML progression, suggesting that a specific level of c-Myc protein is essential for CML induction and progression [31]. In this view, our results suggest that c-MYC overexpression induced by miR-494-3p downregulation in LSCs is strongly connected to BCR-ABL-independent TKI-resistance. This hypothesis opens the way for novel therapeutic strategies combining TKI and c-Myc inhibitors to effectively target CML LSCs.
Worth of note, our data also demonstrated that ectopic expression of miR-29a-3p, miR-494-3p and miR-660-5p was able to affect the response to first, second and third generation TKIs of HSPCs isolated from CML patients ( Figure 8B, 8C), therefore suggesting that their deregulated expression observed in CML LSCs might be responsible for the escape from TKIs treatment in vivo.
In summary, our analysis uncovered several miRNAs aberrantly expressed in CML LSCs. In particular, we identified those miRNAs that are deregulated in LSCs independently from BCR-ABL kinase activity and therefore are likely to be involved in the BCR-ABLindependent resistance to TKI that characterizes CML LSCs. This work unraveled three mRNA/miRNA networks (miR-29a-3p/TET2, miR-660-5p/EPAS1 and miR-494-3p/ c-MYC) capable of affecting TKI-sensitivity in CML cells, thus leading to the design of novel targeted therapies aimed at eliminating TKI-resistant LSCs.

RNA extraction
Total cellular RNA was harvested from 1 × 10 5 cells from each sample using the miRNeasy Micro RNA isolation kit (QIAGEN), according to the manufacturer's instructions. RNA samples concentration and purity (assessed as 260/280 nm and 260/230 nm ratios) were evaluated by NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies; Wilmington, DE), while RNA integrity was assessed by using the Agilent 2100 Bioanalyzer (Agilent Technologies; Waldbrunn, Germany). miRNA expression profiling miEP was performed by using the miRCURY LNA TM Universal RT microRNA PCR system, Ready-to-use Human miRNome Panels I and II (Exiqon, Copenhagen, www.impactjournals.com/oncotarget Denmark). RT-qPCR data were analysed using the Exiqon GenEx software for the inter-plate calibration, the definition of a Ct=40 as detection cutoff, the labeling of not-amplified wells as missing data and the setting of a cutoff for miRNAs with a low call-rate (i.e. assays with a detection rate < 33,3%, i.e. assays not detected in 12 out of 18 samples, were excluded from analysis). For RT-qPCR data normalization, the comparative cycle threshold (CT) method was used by setting miR-23a-3p, miR-24-3p and miR-324-3p as reference miRNAs, as the best normalizers identified among the candidates by the NormFinder algorithm. The Relative Quantity (RQ) value was expressed as 2 −ΔΔCT . Statistical analyses on ΔCT values between CML Lin-CD34-CD38-and Normal donor Lin-CD34-CD38− (set as calibrator) and between CML Lin-CD34+CD38− and Normal donor Lin-CD34+CD38-(set as calibrator) were performed by a two-tail unpaired t-Student test using the GenEx software. Differentially expressed miRNAs (DEMs) in the Lin-CD34-CD38-cell population were selected as the miRNAs with a Relative Quantity (RQ) ≥ 2 or ≤ 0.5 (p < 0.05) in the comparison between CML Lin-CD34-CD38− and Normal donor Lin-CD34-CD38-samples. DEMs in the Lin-CD34+CD38− cell population were selected as the miRNAs with a Relative Quantity (RQ) ≥ 2 or ≤ 0.5 (p < 0.05) in the comparison between CML Lin-CD34+CD38− and Normal donor Lin-CD34+CD38− samples. miEP data have been deposited in the GEO public database [34] (http://www. ncbi.nlm.nih.gov/geo; series GSE90773).

Quantitative reverse transcription polymerase chain reaction (qRT-PCR)
Total RNA (100 ng) was reverse-transcribed to cDNA using a High Capacity cDNA Archive Kit (Life technologies; Carlsbad, CA, USA). TaqMan PCR was carried out using the TaqMan Fast Advanced PCR master mix and TaqMan gene expression assays (all reagents from Life Technologies), by means of a 7900HT Fast Real-Time PCR System (Applied Biosystems). Assays were performed in triplicate. Gene expression profiling was achieved using the comparative cycle threshold (CT) method of relative quantitation using Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as housekeeping genes. To normalize data, ΔΔCT was calculated for each sample using the mean of its ΔCT values subtracted from the mean ΔCT value measured in the control sample, set as a calibrator; relative quantitation (RQ) value was expressed as 2 −ΔΔCT .
Individual miRNA detection by RT-qPCR was performed using the TaqMan MicroRNA assays (Life technologies). Individual reverse transcription reactions (5 ng of total RNA each target) were performed using the Taqman microRNA Reverse Transcription Kit and the miRNA-specific looped-primers. TaqMan PCR was performed in triplicate by using the 7900HT Fast Real-Time PCR System (Applied Biosystems). miRNA expression RQ data were calculated as reported above, using U6 snRNA as housekeeping control.  Table 4) was included. After each transfection, K562 cells were transferred into prewarmed fresh medium in 12-well plates and maintained in the same culture conditions as described above. 24 h after last nucleofection, cells were treated with IM 1 μM or 0.4 μM (Sigma-Aldrich), or Dasatinb 100 nM or 25 nM (Bristol-Myers Squibb) or Ponatinib 10 nM or 2,5 nM (Ariad Pharmaceuticals) for additional 24 and 48 h respectively.

Electroporation of K562 cells
For 3′UTR luciferase reporter assays, cells were transiently co-nucleofected with either a miRNA mimic or miRNA mimic negative control (Neg CTR mimic) at a concentration of 3.6 μM and with either 3′UTRless luciferase or a full-length 3′UTR construct at a concentration of 200 ng/sample. For each electroporation, 10 6 cells were resuspended in 100 μL of SF Cell line Solution (Lonza) and pulsed with the program FF120.

CrkL phosphorylation assay
The CrkL Tyr207-phosphorylation assay was performed in untreated and IM-treated K562 cells after either 24h or 48h of IM treatment. For each datapoint, 10 5 cells per sample were labeled for PCrkL as already described [35]. Cells were analyzed by BD FACSCanto II (BD Biosciences; San Jose, CA USA). At least 10000 events were counted for each sample to ensure statistical relevance. The resulting flow cytometry data were normalized according to the mean fluorescent intensity (MIF) so that the untreated control K562 cell population www.impactjournals.com/oncotarget for each sample had pCrkl MIF = 1. Then, the results for each sample were calculated as a fraction of this.

Annexin V/PI staining
Apoptosis was evaluated by Annexin V assay (Annexin V-FITC Kit, Trevigen Inc.) as previously described [36]. Briefly, 5 × 10 5 cells were washed with cold PBS and incubated in 100 μL Annexin V Incubation Reagent for 15 min at room temperature in the dark. After staining, cells were analyzed by using a BD FACSCanto II (BD Biosciences; San Jose, CA USA). At least 100,000 events were counted for each sample to ensure statistical relevance. Apoptotic cells are Annexin V bright and Propidium Iodide (PI) low, late apoptotic cells or necrotic cells are Annexin V and PI bright [37].

Statistical analysis
The statistics used for data analysis in overexpression/ silencing experiments and 3′UTR luciferase reporter assays were based on one-way ANOVA and 2-tailed Student t-tests for average comparisons in paired samples (equal variance). Data were analyzed with Microsoft Excel (Microsoft Office, 2011 release) and StatPlus:mac LE (AnalystSoft) and are reported as mean ± standard error of the mean (SEM). A p-value < 0.05 was considered significant. Supplementary Table 7

Authorsʼ contributions
SS performed miEP analysis and miRNA overexpression experiments; VS performed CML LSCs isolation; CC performed 3′UTR Luciferase reporter assays, EG performed western blots; SR and GB performed