Role of TGF‐β1/miR‐382‐5p/SOD2 axis in the induction of oxidative stress in CD34+ cells from primary myelofibrosis

Primary myelofibrosis (PMF) is a myeloproliferative neoplasm characterized by an excessive production of pro‐inflammatory cytokines resulting in chronic inflammation and genomic instability. Besides the driver mutations in JAK2,MPL, and CALR genes, the deregulation of miRNA expression may also contribute to the pathogenesis of PMF. To this end, we recently reported the upregulation of miR‐382‐5p in PMF CD34+ cells. In order to unveil the mechanistic details of the role of miR‐382‐5p in pathogenesis of PMF, we performed gene expression profiling of CD34+ cells overexpressing miR‐382‐5p. Among the downregulated genes, we identified superoxide dismutase 2 (SOD2), which is a predicted target of miR‐382‐5p. Subsequently, we confirmed miR‐382‐5p/SOD2 interaction by luciferase assay and we showed that miR‐382‐5p overexpression in CD34+ cells causes the decrease in SOD2 activity leading to reactive oxygen species (ROS) accumulation and oxidative DNA damage. In addition, our data indicate that inhibition of miR‐382‐5p in PMF CD34+ cells restores SOD2 function, induces ROS disposal, and reduces DNA oxidation. Since the pro‐inflammatory cytokine transforming growth factor‐β1 (TGF‐β1) is a key player in PMF pathogenesis, we further investigated the effect of TGF‐β1 on ROS and miR‐382‐5p levels. Our data showed that TGF‐β1 treatment enhances miR‐382‐5p expression and reduces SOD2 activity leading to ROS accumulation. Finally, inhibition of TGF‐β1 signaling in PMF CD34+ cells by galunisertib significantly reduced miR‐382‐5p expression and ROS accumulation and restored SOD2 activity. As a whole, this study reports that TGF‐β1/miR‐382‐5p/SOD2 axis deregulation in PMF cells is linked to ROS overproduction that may contribute to enhanced oxidative stress and inflammation. Our results suggest that galunisertib may represent an effective drug reducing abnormal oxidative stress induced by TGF‐β1 in PMF patients. Database linking GEO: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE103464.


Introduction
Classical Philadelphia-negative chronic myeloproliferative neoplasms (MPNs) are clonal disorders of hematopoietic stem cell that include essential thrombocythemia (ET), polycythemia vera (PV), and primary myelofibrosis (PMF; Vardiman et al., 2009) (Arber et al., 2016). MPNs arise due to an acquired stem cell lesion with subsequent clonal expansion leading to proliferation and overproduction of mature myeloid cells. PMF carries the worst prognosis among MPNs with severe clinical hallmarks, including megakaryocyte and granulocyte hyperplasia, splenomegaly due to extramedullary hematopoiesis, and bone marrow (BM) fibrosis (Tefferi, 2005). The mutational landscape of MPNs has been widely studied identifying driver mutations in JAK2, MPL, and CALR genes in 90% of PMF patients (Klampfl et al., 2013;Nangalia et al., 2013;Pardanani et al., 2006). It is known that the mechanism of action of these mutations results in a constitutive activation of the JAK2/signal transducer and activator of transcription (STAT) signaling pathway (Araki et al., 2016;Chachoua et al., 2016;Marty et al., 2016). The JAK/STAT signaling is a central downstream pathway for the majority of the inflammatory cytokines linked to MPNs and their progression (Vainchenker and Kralovics, 2017). Indeed, PMF patients show increased serum level of pro-inflammatory cytokines (e.g., transforming growth factor-b1 (TGF-b1); interleukin (IL)-8, , and excessive production of reactive oxygen species (ROS), which play a central role in the disease evolution Tefferi et al., 2011;Vaidya et al., 2012). Several findings support the pathogenic model in which the oxidative stress contributes to a state of chronic inflammation and genomic instability that may give rise to DNA damages promoting disease development and progression (Yahata et al., 2011). Moreover, the state of chronic inflammation typical of PMF  is responsible for a self-perpetuating circle in which inflammation creates ROS that in turn create more inflammation (Gloire et al., 2006). Marty et al. demonstrated that the inhibition of ROS production could prevent the development of MPNs in JAK2V617F knock-in mouse model, giving further evidence of the significance of oxidative stress in the pathogenesis of these myeloid malignancies (Marty et al., 2013).
Even though extensive studies have provided unambiguous proof of the role of DNA mutations in the pathogenesis of MPNs, the genetic lesions alone do not solely contribute to the full spectrum of the clinical features of PMF, suggesting the existence of unknown genetic or epigenetic cofactors (Chen et al., 2010).
In the last decade, growing evidence highlighted that differentially expressed microRNAs (miRNAs) can regulate fundamental processes in hematopoietic stem and progenitor cells (HSPCs), such as proliferation, differentiation, and malignant transformation (Bartel, 2004;Zhan et al., 2013). Of note, miRNA profile studies in MPN patients showed differentially expressed miRNAs among different MPN subtypes (Bruchova et al., 2007;Zini et al., 2017). Recently, our group identified several differentially expressed genes and miRNAs potentially involved in PMF pathogenesis (Norfo et al., 2014). In particular, among the upregulated miRNAs, we detected miR-382-5p, known to be associated with other hematologic malignancies, such as acute promyelocytic myeloid leukemia (AML-M3; Jongen-Lavrencic et al., 2008); (Li et al., 2008). In our previous work, we elucidated the role of miR-382-5p in normal hematopoiesis and in HSPC fate demonstrating that its enforced expression favors the granulocyte commitment (Zini et al., 2016).
Here we focused on the mechanistic role of miR-382-5p in the pathogenesis of PMF, by showing its direct involvement in ROS overproduction and oxidative DNA damage induced by the inhibition of its target superoxide dismutase 2 (SOD2).
Then, as TGF-b1 has been already shown to induce miR-382-5p expression in renal cell lines and its role in PMF fibrosis and inflammation has been described (Agarwal et al., 2016;Leask and Abraham, 2004), we investigated whether TGF-b1 could take part in the overproduction of ROS. In this study, we investigated the link between TGF-b1 and the overproduction of ROS. The obtained data indicate that the axis TGF-b1/miR-382-5p/SOD2 is responsible at least in part for the enhanced oxidative stress in normal and PMF CD34+ cells, probably contributing to a state of chronic inflammation, that is a typical feature of PMF. Finally, our study demonstrates the efficacy of the inhibitor of TGF-beta receptor I kinase (TBRI) galunisertib in reducing oxidative stress in PMF CD34+ cells, unveiling a promising therapeutic application of this drug. Primary myelofibrosis CD34+ cells were isolated from the peripheral blood (PB) of three patients in a typical fibrotic phase of the disease according to World Health Organization (WHO) criteria updated in 2016 (Arber et al., 2016). The study was performed under the local institutional review board's approved protocol (Florence: approval date: April 22, 2011, approval file number # 2011/0014777; Pavia: approval date: February 24, 2011, file number #174). All subjects provided informed written consent, and the study was carried out in accordance with the Declaration of Helsinki.
Cells were analyzed 24 and 48 h after the last nucleofection for both cell viability and miR-382-5p and SOD2 expression.
2.5. RNA extraction and gene expression profile miRNeasy micro RNA isolation kit (Qiagen, Hilden, Germany) was used to isolate and purify total RNA containing small RNAs from CD34+ cells, following the manufacturer's instructions. The purity and integrity of RNA samples were determined by using disposable RNA chips (Agilent RNA 6000 Nano Lab-Chip kit) and the Agilent 2100 Bioanalyzer (Agilent Technologies, Waldbrunn, Germany). NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) was used to evaluate the RNA sample concentration, while 260/280 and 260/230 nm ratios were used to assess the RNA purity.
Gene expression profiling was performed in triplicate starting from 100 ng of total RNA obtained from three independent experiments. For microarray analysis, cDNA synthesis and biotin-labeled target synthesis were performed using the GeneAtlas 3 0 IVT Plus Reagent Kit according to the standard protocol supplied by Affymetrix (Santa Clara, CA, USA). The HG-U219 Array Strip (Affymetrix) hybridization, staining, and scanning were performed by using the GeneAtlas Platform.
Gene expression profile (GEP) data were analyzed by PARTEK GS 6.6 Software Package and normalized using the robust multi-array average (RMA) procedure (Irizarry et al., 2003). Differentially expressed genes were selected as the sequences showing a signal-log ratio average > 0.4 or < À0.4 in the pairwise comparisons between miR-382-5p mimic and negative control samples (mimic-Neg). Microarray data were analyzed with the two-tailed Student's t-test for comparison of signal averages in paired samples; P < 0.05 was considered significant. Analysis of variance (ANOVA) module included in PARTEK GS 6.6 Software Package was used for selection of differentially expressed genes shown in the heatmap. Ingenuity pathways analysis (IPA; www.ingenuity.com) was used to find the putative miRNA-mRNA interactions through IPA's MicroRNA Target Filter.

Quantitative RT-PCR (qRT-PCR)
Total RNA (50 ng per sample) was reverse-transcribed to cDNA using the High-Capacity cDNA Reverse Transcription Kit (Life Technologies), and real-time PCR was achieved in triplicate with TaqMan gene expression assays and Fast advanced master mix (all reagents from Life Technologies) by using the AB 7900HT Fast Real-Time PCR System (Applied Biosystems; Guglielmelli et al., 2007a,b). Relative quantity (RQ) was carried out using the comparative cycle threshold (CT) method using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as the housekeeping gene. To normalize the data, DDCT was calculated for each sample using the mean of its DCT values subtracted from the mean DCT value measured in the entire population of healthy subjects, considered as a calibrator; the RQ value was expressed as 2 ÀDDCT .
Individual miRNA detection was performed in triplicate using the TaqMan MicroRNA assays (Life technologies) by means of qRT-PCR. Briefly, 5 ng of total RNA was reverse-transcribed using the TaqMan MicroRNA Reverse Transcription Kit and miRNAspecific looped primers. miRNA expression RQ data were calculated as reported above, by using U6 snRNA as the housekeeping control (Norfo et al., 2014).

Luciferase reporter analysis
The 3 0 -UTR reporter analysis was performed to verify the real interaction between miR-382-5p and its target sequence on SOD2 3 0 UTR. Briefly, predicted target site of miR-382-5p was cloned into the Pme1 and Xbal sites of pmirGLO Dual-Luciferase miRNA Target Expression Vector (Promega Italia, Milano, Italy). Synthetic oligonucleotides (wild-type (WT) miRNAbinding site: 5 0 AAACTAGCGGCCGCTAGTTCAC TTATTTCATAAACAACTTAT3 0 , 3 0 TTTGATCGC CGGCGATCAAGTGAATAAAGTATTTGTTGAAT AGATC5 0 ) were designed based on TargetScanHuman Database 7.0 sequences and correspond to 16 nucleotides surrounding the miRNA-binding sites (seed region) sequence that were annealed before ligation into the pmirGlo plasmid. Mutations in the seed region and luciferase reporter assay in K562 cell line have been performed as previously described (Zini et al., 2016). Briefly, K562 cells were cotransfected by means of the Amaxa 4D-NucleofectorTM System with either a miR-382-5p or mimic-Neg at a concentration of 3.6 lM and with either empty vector or miRNAbinding site-containing construct (200 ngÁsample À1 ). Firefly and Renilla luciferase activities were measured at 36 and 48 h after electroporation using the Dual-Luciferase Reporter Assay System (Promega), and luminescence was recorded on a GloMaxÒ-Multi+ Detection System with Instinct Software (Promega), according to the manufacturer's protocol. Data normalization was performed as previously described (Zini et al., 2016).
Before flow cytometric analysis, the cells were removed from loading buffer and incubated in growth medium for 20 min at 37°C (Marty et al., 2013). Data acquisition and analysis were conducted using a BD FACSCanto II (BD Biosciences, San Jose, CA USA). At least 10 000 events were detected for each sample to guarantee a statistical relevance.

Determination of superoxide dismutase activity in CB and PMF CD34+ cells
Superoxide dismutase (SOD) activity was measured 24 h after the last nucleofection using Trevigen's Superoxide Dismutase Assay Kit according to the manufacturer's instructions (Trevigen, Gaithersburg, MD, USA). SOD activity was assayed on 20 lg of total cellular lysate, based on the reduction in nitroblue tetrazolium (NBT) to nitroblue tetrazolium-diformazan. Since SODs reduce superoxide ion concentration, the reduction in the nitroblue tetrazolium-diformazan production is a measure of total SOD activity (UÁmL À1 ), defined as the amount of enzyme needed to display 50% dismutation of the superoxide radicals. The reduction in nitroblue tetrazolium to nitroblue tetrazolium-diformazan induced by the superoxide radical (O À 2 ) was monitored with Beckman Coulter DUÒ730 Life Science UV/VIS spectrophotometer by reading the absorbance at 550 nm.

Measurement of 8-OH-dG level
Oxidative DNA damage was detected in CB and PMF CD34+ cells 24 h after the last nucleofection by measuring the formation of 8 0 hydroxy-2 0 deoxyguanosine (8-OH-dG), a ubiquitous marker of oxidative stress. Firstly, DNA was isolated using DNeasy Blood and Tissue Kit (Qiagen, Valencia, CA, USA) and the obtained RNA-free DNA was used to estimate 8-OH-dG levels using a competitive enzyme immunoassay according to the manufacturer's protocol (The OxiSelect TM Oxidative DNA Damage ELISA Kit, Cell Biolabs, San Diego, CA, USA). 8-OH-dG concentration was determined by measuring the absorbance at 450 nm with the Glomax Multi Detection System (Promega, Madison, WI, USA).

Measurement of CB and PMF CD34+ cell viability
Viability measurement was assessed by trypan blue exclusion assay 24 h after the last nucleofection (Humpe et al., 2005). In a Neubauer chamber, at least 100 cells were microscopically analyzed in triplicate for cell viability. The mean percentage of living cells of the three analyses was calculated. Furthermore, cell viability was evaluated using XTT Cell Proliferation Assay Kit (Trevigen) in accordance with the manufacturer's instruction. After 24 h from the last nucleofection, CD34+ cells were plated at a density of 50 000 cells/well in 96-well plates. The optical density at 450 nm was measured at 2, 4, 6, 8, and 24 h by Glomax Multi Detection System (Promega).

Ki-67 Flow cytometry analysis
Ki-67 expression was analyzed by flow cytometry. Ki-67 intracellular staining was performed using FIX & Briefly, 24 h after the last nucleofection, 1 9 10 5 cells were fixed for 15 min at RT in reagent A and subsequently fixed in 100% methanol for 10 min at 4°C. Then, fixed cells were washed with PBS and permeabilized with reagent B for 30 min at RT with rabbit polyclonal anti-human Ki-67 (#ab15580, Abcam) at 1 : 250 dilution. Finally, cells were washed with PBS and incubated with fluorescein isothiocyanate goat anti-rabbit IgG F(ab0)2 fragment (#F0382, Sigma-Aldrich, St. Louis, MO, USA) for 30 min at RT. Data acquisition and analysis were performed using a BD FACSCanto II (BD Biosciences). At least 10 000 events were detected for each sample to guarantee a statistical relevance.

Statistical analysis
The statistics used for data analysis were based on two-tailed Student's t-tests for averages comparison in paired samples. Data were analyzed by using Microsoft Excel (Microsoft Office, 2008 release) and are reported as mean AE standard error of the mean (SEM). P < 0.05 was considered significant.

Gene expression profile (GEP) of miR-382-5poverexpressing cells
In order to investigate the possible role of miR-382-5p in the pathogenesis of PMF, we performed a GEP in CD34+ cells overexpressing miR-382-5p. miRNA overexpression was achieved by nucleofection of miR-382-5p mimic. First, we assessed the transfection efficiency using BLOCK-IT fluorescent oligo and we showed that the proportion of transfected cells is more than 90% (Fig. S1). As shown in Fig. 1A, we obtained a significant overexpression of miR-382-5p upon miRNA mimic transfection at 24 and 48 h after the last nucleofection, which was maintained at least until 96 h (data not shown). The expression level of miR-382-5p in the three different control samples (i.e., untransfected cells, MOCKtransfected cells, and cells transfected with miRNA mimic negative control or mimic-Neg) was not changed (Fig. 1A). Based on these data, we selected the mimic-Neg sample as negative control and we included it in each experiment.
Microarray analysis was performed 24 h after the last nucleofection, the time point in which miR-382-5p expression level achieves the best upregulation in CD34+ cells, as depicted in Fig. 1A. Using the filtering procedure described in Methods, we generated a list of 97 genes significantly modulated in miR-382-5p-overexpressing cells (miR-382-5p) compared to mimic-Neg sample (Table S1). Fig. 1B shows the heatmap of differentially expressed genes further selected by ANOVA (P value < 0.05). Among the 75 downregulated genes, we identified SOD2, which is one of the most   favorable predicted targets of miR-382-5p (context score À0.59) according to TargetScanHuman Database (Table S2). Interestingly, GEP analysis of PMF CD34+ cells showed that SOD2 mRNA is significantly decreased in PMF patients compared to healthy donors ( Fig. S2; Norfo et al., 2014).
The SODs are a family of antioxidant enzymes that catalyze the dismutation of superoxide radical anions into hydrogen peroxide. In particular, manganese superoxide dismutase (MnSOD-SOD2) has emerged as a crucial tumor suppressor gene and a key mediator in the mitochondrial clearance of detrimental ROS (Van Remmen et al., 2003). Since ROS play a crucial role in disease progression of MPNs (Bjorn and , we decided to investigate the possible involvement of the miR-382-5p/SOD2 axis in the induction of oxidative stress in normal and PMF CD34+ cells.

SOD2 is a target of miR-382-5p
In order to validate the negative correlation between miR-382-5p and SOD2 expression in CD34+ cells, we measured the level of SOD2 in miR-382-5p-overexpressing cells. As shown in Fig. 2A, SOD2 mRNA is significantly downregulated upon miR-382-5p overexpression at 24 and 48 h after the last nucleofection. Then, to confirm that SOD2 is a real target of miR-382-5p, we performed a luciferase reporter assay, by cloning the predicted miRNA-binding site of SOD2 downstream of firefly luciferase gene in Dual Luciferase pmirGLO vector (Fig. 2B). Transient cotransfection of either miR-382-5p or mimic-Neg with WT or mutated miRNAbinding site was performed in K562 cells. As shown in Fig. 2C, luciferase activity is significantly decreased by miR-382-5p overexpression when K562 cells were transfected with the construct containing the WT miRNAbinding site of SOD2. Mutations in the seed regionbinding side abrogate the effect of miR-382-5p overexpression. Furthermore, we confirmed that miR-382-5p overexpression induces a reduction in SOD2 protein level at 24 and 48 h after the last nucleofection (Fig. 2D). Immunofluorescence for SOD2 showed a significant reduction in SOD2 in miR-382-5p-overexpressing cells compared to mimic-Neg (Fig. 2E,F).

miR-382-5p overexpression reduces SOD2 activity, and increases ROS production and oxidative DNA damage
Superoxide dismutase 2 is the major mitochondrial enzyme responsible for superoxide dismutation, preventing the production and accumulation of potentially damaging ROS. For this reason, we wondered whether miR-382-5p overexpression could affect SOD2 activity by assaying for superoxide dismutase reaction using Trevigen's Superoxide Dismutase Assay Kit. Fig. 3A shows that the SOD activity is significantly decreased in miR-382-5p-overexpressing cells compared to negative control. Then, by using CM-H 2 DCFDA staining, we detected an accumulation of ROS in miR-382-5p-overexpressing CD34+ cells (Fig. 3B,C). Since the enhanced production of ROS can lead to DNA damage in hematopoietic stem cells (Yahata et al., 2011), we measured DNA 8-OH-dG level, a known marker of oxidative stress-mediated DNA damage. As shown in Fig. 3D, we demonstrated a significant increase in 8-OH-dG level in cells overexpressing miR-382-5p compared to negative control.
These results suggest that miR-382-5p, through the inhibition of its target SOD2, causes an increase in intracellular ROS level, which in turn leads to the DNA damage.

miR-382-5p overexpression increases proliferation of CD34+ cells
Since SOD2 has been already described as a tumor suppressor gene and an inverse correlation between its expression level and proliferation rate has been already reported in HSPCs (Hurt et al., 2007), we investigated the role of miR-382-5p in CD34+ cell proliferation.
First of all, trypan blue exclusion assay highlighted a significant increase in cell count in miR-382-5p compared to mimic-Neg sample (Fig. 4A). Moreover, as shown in Fig. 4B, the effect of miR-382-5p on cell viability was further confirmed by XTT assay. Accordingly, flow cytometry analysis showed that cells overexpressing miR-382-5p display a high positivity for Ki-67 antigen, expressed in all proliferating cells Values are reported as mean AE standard error of the mean (SEM). Two-tailed Student's t-test: ***P < .001 versus anti-miR-Neg **P < .01 versus anti-miR-Neg; *P < .05 versus anti-miR-Neg.
Altogether, these results suggest a positive effect of miR-382-5p on CD34+ cell proliferation.

Inhibition of miR-382-5p reduces oxidative stress and decreases cell proliferation of PMF CD34+ cells
In order to clarify whether miR-382-5p could be a key player in PMF pathogenesis by promoting ROS overproduction in hematopoietic progenitors, we inhibited miR-382-5p in PMF CD34+ cells by transient transfection of mirVana miRNA inhibitor. As shown in Fig. 5A, we obtained a significant reduction in miR-382-5p expression level at 24 and 48 h after the last nucleofection. At the same time points, we evaluated SOD2 expression, demonstrating a relevant increase in SOD2 at both mRNA and protein level at 48 h after last nucleofection (Fig. 5B,C). These results were confirmed by SOD2 immunofluorescence analysis that shows a strong increase in protein expression in PMF CD34+ cells upon miR-382-5p downregulation ( Fig. 5D-E). Moreover, miR-382-5p inhibition restores SOD activity (Fig. 5F), causes a decrease in ROS accumulation (Fig. 5G), and leads to a significant reduction in oxidative stress-mediated DNA damage (Fig. 5H).
Interestingly, trypan blue exclusion assay highlighted a significant decrease in cell count upon miR-382-5p inhibition (Fig. 6A). Furthermore, flow cytometry analysis for Ki-67 revealed that miR-382-5p downregulation impairs the ability of PMF CD34+ cells to proliferate (Fig. 6B). These results were further confirmed by immunofluorescence analysis for the detection of Ki-67 antigen, which shows a reduction in proliferating Ki-67-positive cells after miR-382-5p inhibition (Fig. 6C,D). These data suggested that miR-382-5p inhibition could impair the proliferation capability of PMF CD34+ cells.
3.6. Transforming Growth Factor-b1 induces ROS production in CB CD34+ cells through the modulation of miR-382-5p/SOD2 axis Transforming growth factor beta 1 is a key mediator of inflammation and fibrosis, widely described as involved in PMF pathogenesis (Agarwal et al., 2016) and able to upregulate miR-382-5p in renal fibrosis (Kriegel et al., 2010). For this reason, we decided to study the axis TGF-b1/miR-382-5p/SOD2 in CB CD34+ cells in order to understand whether it could be involved in the overproduction of ROS.
To this end, we examined the expression level of both miR-382-5p and SOD2 at 24 and 48 h after TGF-b1 treatment. As shown in Fig. 7A, real-time PCR analysis reveals a strong induction of miR-382-5p in TGF-b1-treated CD34+ cells compared to untreated cells. Conversely, we observe a decrease in SOD2 mRNA at both 24 and 48 h after TGF-b1 treatment (Fig. 7B). The suppression of SOD2 protein level was further confirmed by western blot analysis and in situ immunofluorescence (Fig. 7C-E). Moreover, we demonstrated that TGF-b1 treatment induces a significant reduction in SOD activity and a remarkable induction of ROS production (Fig. 7F,G).
Overall, these data suggest a direct involvement of TGF-b1 in ROS accumulation through the modulation of miR-382-5p/SOD2 axis.

Galunisertib (LY-2157299) reduces ROS production in PMF CD34+ cells
Since we have shown the significant role of TGF-b1 in the induction of oxidative stress in normal CD34+ cells, we further tested the potential efficacy of galunisertib, an ALK5 inhibitor in PMF CD34+ cells. Firstly, we assessed the expression level of miR-382-5p and SOD2 after treatment with TGF-b1 or galunisertib alone or in combination in PMF CD34+ cells. According to results obtained in normal CD34+ cells, the expression level of miR-382-5p is significantly enhanced in TGF-b1-treated cells (Fig. 8A); conversely, SOD2 mRNA is reduced after TGF-b1 Fig. 6. Effect of miR-382-5p inhibition on PMF CD34+ cell proliferation. (A) Viability measurement was assessed by cell count (cellsÁmL À1 ) 48 h after the last nucleofection by trypan blue exclusion assay. (B) Percentage of Ki-67-positive cells measured by flow cytometry at 48 h after the last nucleofection. (C) Representative image of Ki-67 immunofluorescence analysis in PMF CD34+ cells transfected with miR-382-5p inhibitor or anti-miR-Neg at 48 h after the last nucleofection. Cells were labeled with anti-Ki-67 antibody (red fluorescence), and nuclear counterstaining was performed with DAPI (blue fluorescence). Scale bar: 20 lm. Original magnification = 40X. To ensure random sampling, 20 images/slide were captured and cells positive for SOD2 were scored. Fluorescence imaging was performed using the Zeiss LSM 510 Meta Confocal Microscope. (D) Quantification of Ki-67 positivity in PMF CD34+ cells transfected with miR-382-5p inhibitor compared to negative control. At least 200 cells for each experiment were scored. The results are representative of three independent experiments and are expressed as the mean AE SEM. Two-tailed Student's t-test: *P < .05 versus anti-miR-Neg. treatment (Fig. 8B). Galunisertib treatment seems to restore the expression of both miR-382-5p and SOD2 in TGF-b1-treated cells. In particular, in the presence of galunisertib alone, miR-382-5p expression is suppressed, whereas SOD2 mRNA is significantly increased compared to untreated cells (Fig. 8A,B). Western blot analysis and immunofluorescence further confirmed these results (Fig. 8C-E). Interestingly, SOD activity is restored in PMF CD34+ cells stimulated by TGF-b1 upon treatment with galunisertib. Moreover, we observed that galunisertib alone is able to restore the basal SOD activity in PMF CD34+ cells (Fig. 8F).
According to our previous results, flow cytometric analysis of ROS content shows an increase in ROS accumulation upon TGF-b1 treatment, while we observe a reduction in ROS level in TGF-b1 stimulated cells in the presence of galunisertib. Remarkably, the effect is also evident when PMF CD34+ cells are treated with galunisertib alone (Fig. 8G).
Altogether, these results demonstrate that galunisertib is able to abrogate the TGF-b1-driven induction of oxidative stress in PMF CD34+ cells, pointing out the importance of the axis TGF-b1/miR-382-5p/SOD2 in PMF pathogenesis.

Discussion
In the last decades, growing evidence has highlighted the role of deregulated miRNA expression in MPN pathogenesis. In order to identify a series of miRNAs and coding genes potentially involved in PMF pathogenesis, we recently obtained gene and miRNA expression profiles of CD34+ HSPCs from PMF patients. Integrative analysis of gene and miRNA expression profiles highlighted miR-382-5p among miRNAs significantly upregulated in PMF CD34+ cells (Norfo et al., 2014). Of note, miR-382-5p deregulated expression has been previously reported in other hematological malignancies, such as acute promyelocytic myeloid leukemia (AML-M3; Jongen-Lavrencic et al., 2008). In addition, we unveiled the key role of miR-382-5p in HSPC fate decision toward granulocyte lineage (Zini et al., 2016).
In order to shed light on the role of miR-382-5p in PMF pathogenesis, we assessed the gene expression changes induced by miR-382-5p overexpression in normal CD34+ cells. Among genes downregulated after miR-382-5p overexpression, we selected SOD2, which is associated with the most favorable prediction score according to TargetScanHuman prediction algorithm. Interestingly, SOD2 expression is decreased in PMF CD34+ cells compared to healthy subjects (Norfo et al., 2014).
The SOD antioxidant family is responsible for the dismutation of the superoxide radical anion in hydrogen peroxide and provides an essential defense against the harmful effect of ROS (Szeto, 2006). High levels of ROS are known to be involved in the pathogenesis of several solid cancers and hematological malignancies (Hurt et al., 2007;Waris and Ahsan, 2006;Zheng et al., 2013) and to play an important role in the initiation and progression of MPNs Vener et al., 2010). Indeed, as previously reported by Vener et al., MPN patients show significantly higher serum ROS level if compared to healthy donors (Vener et al., 2010). Among SOD family members, manganese superoxide dismutase 2 (MnSOD-SOD2) plays an essential role not only in cancer but also in a wide range of stress-induced diseases due to its localization in the mitochondria, where the production of ROS is higher compared to the other cellular compartments (Inoue et al., 2003). In particular, Fig. 7. Transforming Growth Factor-b1 induces ROS production in CB CD34+ cells through the modulation of miR-382-5p/SOD2 axis. (A,B) CB CD34+ cells were treated with 5 ngÁmL À1 of TGF-b1 for 48 h. qRT-PCR analysis was performed to evaluate the expression level of miR-382-5p and SOD2 upon TGF-b1 treatment at different time points: before treatment (T0), 24, and 48 h. U6 snRNA and GAPDH were used as internal controls, and untreated sample was used as calibrator. (C) SOD2 protein level was evaluated by western Blot analysis in lysates from TGF-b1-treated and untreated cells 24 and 48 h after treatment. b-Actin protein level is reported as loading control. (D) Immunofluorescence analysis of SOD2 (green fluorescence) in TGF-b1-treated and untreated cells at 24 h after treatment. Nuclear counterstaining was performed with DAPI (blue fluorescence). Scale bar, 20 lm. Original magnification = 40X. To ensure random sampling, 20 images/slide were captured and cells positive for SOD2 were scored. Fluorescence imaging was performed using the Zeiss LSM 510 Meta Confocal Microscope. (E) Quantification of SOD2 positivity in CD34+ treated with TGF-b1 compared to untreated cells. At least 200 cells for each experiment were scored. (F) TGF-b1 effect on SOD activity (UÁmL À1 ) at 24 h after treatment. (G) ROS level measured by DCF staining 24 and 48 h after TGF-b1 treatment. The results are representative of three independent experiments and are expressed as the mean AE SEM. Two-tailed Student's t-test: **P < .01 versus untreated cells; *P < .05 versus untreated cells.
SOD2 has been described as a tumor suppressor gene (Van Remmen et al., 2003) and its decreased expression level has been reported in a wide variety of diseases (Wang et al., 2016) (Hitchler et al., 2008, including hematological malignancies (Hurt et al., 2007).
For this reason, we decided to investigate the potential involvement of the axis miR-382-5p/SOD2 in the induction of oxidative stress in normal and PMF CD34+ cells. Firstly, we demonstrated that SOD2 is a real target of miR-382-5p by means of luciferase reporter assay. According to these data, we observed a significant downregulation of SOD2 at both mRNA and protein level upon miR-382-5p overexpression in CD34+ cells. Furthermore, we showed that enforced miR-382-5p expression leads to a reduction in SOD activity, which in turn determines a significant ROS accumulation and the consequent increase in DNA oxidation.
Moreover, we demonstrated that miR-382-5p overexpression stimulates CD34+ cell proliferation. Since elevated ROS level has been reported to promote mitosis in CD34+ cells (Hole et al., 2010), and enhanced SOD2 activity has been associated with proliferation inhibition in several cell lines (Oberley, 2005), we hypothesized that miR-382-5p could act on cell proliferation through the modulation of SOD2 expression and ROS overproduction.
As previously reported by Norfo et al., while miR-382-5p expression is increased, SOD2 mRNA is decreased in PMF CD34+ cells (Norfo et al., 2014). Therefore, in order to confirm that miR-382-5p is a Fig. 8. Galunisertib (LY-2157299) restores SOD2 activity and reduces ROS accumulation in PMF CD34+ cells. PMF CD34+ cells were treated with TGF-b1 or galunisertib (Inh) alone or TGF-b1 and galunisertib (TGF-b1+Inh) in combination. Untreated cells were used to normalize the data. (A,B) qRT-PCR analysis was performed to evaluate the expression level of miR-382-5p and SOD2 at 24 h after treatments. U6 snRNA and GAPDH were used as internal controls. (C) Western blot analysis was performed after 24 h of treatment on the whole-cell lysate from TGF-b1/TGF-b1+Inh/Inh-treated and untreated cells. b-actin protein level is reported as loading control. (D) Immunofluorescence analysis of SOD2 (green fluorescence) in TGF-b1/TGF-b1+Inh/Inh-treated and untreated cells at 24 h after treatments. Nuclear counterstaining was performed with DAPI (blue fluorescence). Scale bar, 20 lm. Original magnification = 40X. To ensure random sampling, 20 images/slide were captured and cells positive for SOD2 were scored. Fluorescence imaging was performed using the Zeiss LSM 510 Meta Confocal Microscope. (E) Quantification of SOD2 positivity. At least 200 cells for each experiment were scored. (F) SOD activity (UÁmL À1 ) has been determined after 24 h of TGF-b1/TGF-b1+Inh/Inh treatments and compared to untreated cells. (G) Level of ROS production measured by DCF staining 24 h after TGF-b1/TGF-b1+Inh/Inh treatment. The results are representative of three independent experiments and are expressed as the mean AE SEM. Two-tailed Student's t-test: **P < .01 versus untreated cells; *P < .05 versus untreated cells. # P < 0.05 versus TGF-b1-treated cells. Inh, inhibitor (galunisertib). Fig. 9. The axis TGF-b1/miR-382-5p/ SOD2. The axis TGF-b1/miR-382-5p/SOD2 as mechanism underlying the overproduction of ROS in CD34+ PMF cells. ROS level could increase as a consequence of: (a) the release of proinflammatory cytokines from granulocytes induced by miR-382-5p overexpression; (b) the enhanced expression of miR-382-5p, which reduces the expression of its target SOD2; (c) the upregulation of miR-382-5p by TGF-b1. The TGF-b-receptor I kinase inhibitor galunisertib could inhibit the production of ROS by reducing the expression level of miR-382-5p and restoring SOD2 activity. key player in PMF pathogenesis, we performed silencing experiments in PMF CD34+ cells. Our results show that miR-382-5p downregulation restores SOD2 expression and activity, reducing DNA damage and ROS production. Furthermore, we demonstrated that miR-382-5p inhibition impairs the proliferation of PMF CD34+ cells.
As a whole, miR-382-5p overexpression, leading to SOD2 downregulation in HSPCs, could explain the overproduction of ROS and therefore the accumulation of oxidative DNA damage observed in PMF patients. As a consequence, DNA damage generates genomic instability that leads to the acquisition of new mutations allowing the clonal evolution and the disease progression (Marty et al., 2013).
One of the main features underlying PMF pathogenesis is the deregulation of the serum levels of proinflammatory cytokines (Tefferi et al., 2011), which contribute to a state of chronic inflammation (Barosi, 2014;Skov et al., 2013). In particular, several findings have highlighted the key role of TGF-b signaling in inflammation through the modulation of cytokine production, tumor migration, and angiogenesis (Pickup et al., 2013;Siegel and Massague, 2003). Moreover, TGF-b1 has been reported to be a strong inducer of BM fibrosis (Agarwal et al., 2016), which is another important feature of PMF. Since it has been demonstrated that TGF-b1 stimulates miR-382-5p expression during renal fibrosis (Kriegel et al., 2010), we investigated whether it could be able to regulate the miR-382-5p expression in normal CD34+ cells. Here we demonstrated that TGF-b1 treatment enhances miR-382-5p expression and in turn suppresses SOD2 activity in normal CD34+ cells. According to our previous results, we highlighted an increase in ROS production after TGF-b1 treatment, probably due to a reduction in SOD2 activity.
Finally, in order to further confirm the involvement of the axis TGF-b1/miR-382-5p/SOD2 in PMF pathogenesis, we inhibited TGF-b1 pathway in PMF CD34+ cells using the TGF-b-receptor I kinase inhibitor galunisertib. Among the TGF-b inhibitors, galunisertib preclinical efficacy has been already tested in different mouse models (Yue et al., 2017), demonstrating a strong effect in reversing or reducing BM fibrosis. Moreover, it has been demonstrated that galunisertib is a strong inhibitor of TGF-b signaling in hematopoietic cells (Zhou et al., 2011). Here we showed that galunisertib is able to revert the effects induced by TGF-b1 in both normal and PMF CD34+ cells. However, galunisertib alone reduces miR-382-5p expression and increases SOD2 activity only in PMF samples. Interestingly, galunisertib is also able to reduce ROS accumulation in PMF CD34+ cells. These results are in line with the evidence that TGF-b1 signaling is a hyperactivated pathway in MPNs (Kota et al., 2008).
Our data support the idea that the chronic state of inflammation can itself promote ROS accumulation, which in turn generates more inflammation. Moreover, since we have already demonstrated that miR-382-5p stimulates the granulocyte lineage (Zini et al., 2016), we hypothesized that the increased release of proinflammatory cytokines (i.e., TGF-b1) from granulocytes themselves could activate the miR-382-5p/SOD2 axis in PMF CD34+ cells sustaining the high oxidative stress state (Fig. 9).
In summary, we propose a new role for TGF-b1 as a key mediator in ROS overproduction and we bring further evidence of its contribution in establishing a state of chronic inflammation. Strikingly, we identified a negative correlation between the expression level of SOD2 and the grade of BM fibrosis in PMF, suggesting that ROS accumulation caused by SOD2 reduction could favor the TGF-b1-driven progression of fibrosis (Fig. S4).
Moreover, the reduction in ROS content induced by galunisertib in PMF CD34+ cells indicates that it should be an effective therapy to decrease the oxidative stress produced and sustained by the increased level of TGF-b1 in PMF patients.
As a whole, our data suggest that the combination therapy of galunisertib with JAK2 inhibitors, which is currently the gold standard for PMF therapy, should represent a novel therapeutic option.

Conclusions
Here we describe a new role of TGF-b1 as ROS inducer. Alongside its well-known function in PMF fibrotic process, we demonstrated that TGF-b1 increases the production of ROS through the modulation of the axis miR-382-5p/SOD2. Moreover, we shed light on the novel therapeutic potential of galunisertib, a TGF-b1 inhibitor, in reducing ROS accumulation and oxidative stress in PMF CD34+ cells. Author contributions CR, RZ, and RM designed the study and wrote the manuscript; RZ, ET, and SeRo performed microarray experiments; CR, SaRu, and EB performed miRNA overexpression and silencing experiments; CR and GB carried out luciferase reporter assays and qRT-PCR; CR and ZP performed SOD2 activity assay, ROS detection, and DNA oxidation assays; CR, EG, and SS performed immunostaining analysis; AMV, PG, NB, GB, and VR enrolled patients.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. miRNA transfection efficiency. Fig. S2. SOD2 expression in CD34+ cells from PMF patients and healthy donors. Fig. S3. Galunisertib (LY-2157299) treatment in normal CD34+ cells. Fig. S4. Fibrosis grade according to SOD2 expression level in PMF CD34+ cells. Table S1. Deregulated genes upon miR-382-5p overexpression. Table S2. List of miR-382-5p predicted targets.