Targeting iron homeostasis induces cellular differentiation and synergizes with differentiating agents in acute myeloid leukemia

To gain new insights in the molecular mechanisms involved in the anti-tumor activity of iron-chelating therapeutic approaches, we sought to identify genes that are affected both by iron chelators (deferoxamine [DFO] and deferasirox [DFX]) and by impairment of Fe-Tf uptake (using the anti-TfR1 mAb A24 [Moura et al., 2004; Lepelletier et al., 2007; Callens et al., 2008]). Indeed, A24 targets TfR1 (the major receptor involved in the iron entry in the cells), whereas iron chelators can also capture other bivalent ions (e.g. Mn²⁺, Cu²⁺, or Zn²⁺) which could modulate genes irrelevant to iron homeostasis. Thus, identification of overlapping genes allows us to avoid selecting off-target genes. We focused our first studies in a cell line (HL60) to avoid heterogeneity in the data related to the multiple genetic events that have been implicated in AML oncogenesis. An analysis of the collected data revealed that 105 genes were similarly modulated by the three iron deprivation agents (Fig. 1 a and Table S1). Among these genes, several were related to cell differentiation. Markers of monocytes or activated macrophages were up-regulated, whereas granulocyte markers were down-regulated (Fig. 1 b). These results were confirmed by real-time quantitative PCR and extended to other AML cell lines from different AML subtypes (Fig. S1 a).

We further tested the ability of iron deprivation to override the block of cell differentiation observed in AML cells. We followed the expression of CD14 and CD11b monocyte cell surface markers (Kansas et al., 1990) after iron deprivation induced by treatment with A24, DFO, or DFX. Expression of both markers was induced by iron deprivation, suggesting that cells underwent differentiation (Fig. 1 c) similar to VD-treated cells (Fig. S1 b). Up-regulation of monocyte cell surface markers, initially detected in myeloblastic cells (HL60), was also observed in cell lines from myelo/monoblastic (OCI AML-3), monoblastic (U937 and THP1), and promyelocytic (NB4) origin (Fig. 1 d).

Induction of cell surface markers in treated cells was accompanied by the characteristic cytological modifications of monocytes, i.e., cytoplasm enlargement and loss of both cytoplasmic basophilia and azurophilic granules (Fig. 1 e and Fig. S1 c). Moreover, HL60 differentiated cells acquired functional properties of monocytes such as esterase activity (Fig. S1 d).

Supplementation of cultures with an excess of soluble iron abrogated expression of differentiation markers induced by iron deprivation (Fig. 1 f) but not by ATRA and VD (Fig. S2), confirming the role of iron availability in AML cell differentiation. We next investigated whether primary cells from AML patients could also be sensitive to iron deprivation therapy. Fresh AML blasts from different AML subtypes (categorized according to the French-American-British [FAB] classification system [Bennett et al., 1976]) were isolated at the time of diagnosis (Table S2, complete list of AML subtypes and biological parameters for the patients used in this study) and were cultured in the presence of A24 and DFO. An arrest of cell proliferation (Fig. 2 a) and induction of apoptosis (Fig. 2 b) were observed concomitantly to CD14 and CD11b expression (Fig. 2 c), indicating that, similar to cell lines, blasts from different AML subtypes, even if heterogeneous in the oncogenic events leading to their arrest in differentiation (Löwenberg et al., 1999), are susceptible to differentiation therapy based on iron deprivation.

To verify if iron homeostasis plays a role in the differentiation of primary granulocytes and monocytes, we tested the effect of iron deprivation on hematopoietic cord blood progenitors. Iron deprivation agents did not change the total number of colonies derived from CD34⁺ cells in methylcellulose plates (unpublished data). However, treatment with the iron chelators resulted in an increased number of bipotent CFU-GM colonies (Fig. 3 a) and enhanced the ratio of CFU-M to the detriment of CFU-G colonies (Fig. 3 b), suggesting that chelating iron from hematopoietic precursors induces commitment toward the monocyte over the granulocyte lineage. As expected, because iron is necessary for hemoglobin synthesis (Adamson, 1994) the number of CFU-E colonies was reduced by all the iron deprivation agents (Fig. 3 c).

When differentiation of CD34⁺ cord blood progenitors into granulocytes was induced in liquid cultures (Fig. S3), >65% of cells were polymorphonuclear cells after 18 d. We therefore added A24 or iron chelators to these granulocyte-oriented cell cultures every 3 d, at concentrations which did not induce cell toxicity (unpublished data). Under these conditions, iron deprivation did not affect cord blood cell proliferation (Fig. 3 d) but rather enhanced their commitment toward the monocyte lineage because there was a consistent up-regulation of genes coding for transcription factors specific for the monocyte lineage, such as HOXA10, EGR1, and MafB (Sieweke et al., 1996; Kelly and Gilliland, 2002; Taghon et al., 2002; Friedman, 2007; Fig. 3 e). Cytological examination confirmed that the monocyte cell number was higher in iron-poor cultures (Fig. 3 f). Therefore, in hematopoiesis, iron availability could be an important factor in determining whether a hematopoietic progenitor cell differentiates toward a monocyte or a granulocyte.

ROS production is highly dependent on the intracellular labile iron pool (LIP). Therefore, to investigate whether iron-chelating agents modulate ROS levels in AML cells, we used a metallo-sensor fluorescent probe (calcein) for LIP measurements (Espósito et al., 2002) and dihydrorhodamine 123 (DHR123) as an indicator of the degree of general oxidative stress or dihydroethidium (DHE) to monitor cellular superoxide production (Owusu-Ansah et al., 2008). As expected, previous addition of iron deprivation agents decreased LIP levels in AML cells and DFX was the fastest and most efficient compound (Fig. S4, a–c). Iron deprivation of cells by DFX (the most permeant chelators) was readily accompanied by ROS formation (Fig. 4 a), which was found to be concentration (Fig. 4 b) and time (Fig. 4 c) dependent. We further confirmed the specificity of ROS detection by preincubating iron-deprived cells with the anti-oxidant N-acetyl-l-cysteine (NAC; Fig. 4 d).

Iron chelators had opposite effects depending on whether AML cells were treated during short (1–3 h; Fig. 4, a–d) or long (16–18 h; Fig. S4, d–f) periods of time. In addition, in HL60-cultured cells treated for 16–18 h with iron chelators, the H₂O₂-mediated generation of ROS was abrogated in a dose-dependent manner (Fig. S4, d and e), whereas ATO induced a sustained production of ROS in NB4 cells (Fig. S4 f).

To further clarify which of the pro- or anti-oxidant effects of iron chelators was implicated in monocyte differentiation, iron-deprived AML cells were treated with or without the anti-oxidants NAC or pyrrolidine dithiocarbamate (PDTC). Anti-oxidant prevented the increase in CD14 mediated by iron deprivation differentiation, suggesting that cell differentiation is dependent on the increased ROS levels induced during the early course of cellular iron deprivation (Fig. 4, e and f). Thus, iron-depriving agents control cell differentiation through modulation of ROS levels.

MAPKs are serine/threonine protein kinases crucial to several biological responses including stress responses to ROS (Whitmarsh and Davis, 1999). The addition of the specific inhibitors of ERK, JNK, and p38 MAPK signaling pathways (which characterize the three MAPK families) abrogated cell differentiation induced by iron deprivation agents (Fig. 5 a). The iron-chelating agents induced the phosphorylation of ERK, p38, and JNK (Fig. 5 b). However, only inhibitors of JNK and p38, but not of ERK, prevented iron deprivation–induced apoptosis (Fig. 5 c).

VD can induce monocyte/macrophage differentiation of both primary cells and myeloid leukemia cell lines (Abe et al., 1981; Koeffler et al., 1984; Hmama et al., 1999; Bastie et al., 2004; Ji and Studzinski, 2004; Gemelli et al., 2008; Hughes et al., 2009). By comparing the pattern of genes induced by VD and by iron-chelating agents in a nonsupervised transcriptome analysis, we found a high similarity between the differentially expressed genes; of the 105 genes that were modulated by iron deprivation, 30 were also modulated by treatment with VD (Fig. 6 a and Table S3). Both VD and the iron-chelating agents up-regulated monocyte markers and down-regulated neutrophil markers (Fig. 6 b). Moreover, in gene chips (Fig. 6 c) and real-time quantitative PCR (Fig. 6 d) analysis, transcription of the c-Jun gene was induced by both VD and iron deprivation. Transcripts coding for c-Fos were also up-regulated by both treatments (Fig. 6 d), suggesting the involvement of the JNK pathway. Similar results were observed in cell lines from different AML subtypes (Fig. 6 e). To further address the role of JNK, cells were transfected with micro RNA (miRNA) constructs specific for different members of the JNK pathway (namely JNK1, JNK2, and c-Jun; Fig. S5, a–c). miRNA–mediated silencing of c-Jun, JNK1, and JNK2 markedly abrogated cell differentiation induced by iron deprivation agents (Fig. 6 f). Overexpression of each one of the two JNK proteins revealed that JNK1 was the most effective to induce cell differentiation of AML cells (Fig. 6 g).

Given the high similarities shared between VD and iron deprivation agents to induce cell differentiation, we investigated whether iron deprivation of cells could induce VDR signaling (Wang et al., 2003; Himes et al., 2006). As observed for its cognate ligand VD, VDR expression was induced in cells treated by iron-chelating agents (Fig. 6 h). The addition of iron chelators to cells starved of growth factors readily induced the VDR phosphorylation (Fig. 6 i) involved in the control of VDR activity (Jurutka et al., 1996). Altogether, these results indicate that iron-chelating agents are able to induce cell differentiation through the activation of JNK pathway. Iron deprivation also potentiates VDR signaling pathway through VDR expression and VDR phosphorylation.

Treatment of cells with both iron chelators and VD drastically increased the up-regulation of monocyte markers CD11b and CD14 (Fig. 7 a). We consequently investigated whether the combination of iron-chelating agents (as the iron chelator DFO and the anti-TfR1 A24) and VD had a synergistic effect. Applying the combination index (CI) isobologram method (Chou and Talalay, 1984), we evaluated interactions between DFO and VD, or A24 and VD (Figs. S6, a–d). At the concentration inducing 50% of CD14⁺ cells, CI values for the combination of DFO and VD, or A24 and VD, were <1 (0.87 and 0.89, respectively), indicating a synergistic interaction. Both combinations were accompanied by morphological changes in the treated cells (Fig. 7 b) and by higher expression of VDR (Fig. 7 c). The physiological relevance of increased VDR gene expression was demonstrated by the robust increase in the transcription of its downstream target genes, cathelicidin (1,000-fold increase) and CYP24A (6,000-fold increase; Fig. 7 d), and c-Fos and c-Jun (Fig. 7 e). Moreover, the phosphorylation of JNK increased in response to treatment with both VD and iron deprivation (Fig. 7 f). The combined effect of VD and iron chelators was specific to VDR signaling because iron chelators did not potentiate the ability of ATRA to promote cell differentiation in HL60 cells (Fig. S7). Finally, the efficacy of the combined therapy was validated ex vivo in primary cells from AML patients (Fig. 7 g).

Based on these observations, we postulated that the combination of iron deprivation and VD could intensify their anti-leukemic properties. Therefore, the efficacy of differentiation therapy was further evaluated in vivo in a mouse tumor xenograft model. A single dose of A24 completely inhibited the growth of HL60 transplanted tumors, whereas treatment with DFO increased survival and decreased the tumor size (Fig. 8, a and b). Using this model, we did not observe a significant reduction in tumor growth for mice treated with VD alone (Fig. 8 c). However, the combination of VD with DFO significantly reduced tumor growth (Fig. 8 d). The parameters of serum iron homeostasis, such as serum transferrin and ferritin levels, were significantly modified in the treated mice, which is an indication of effective iron deprivation (Fig. 8 e). The reduced tumor growth was associated with tumor cell apoptosis, which are features of cellular differentiation and MAPK phosphorylation (ERK and JNK; Fig. 8 f). In mice xenografted with OCI-AML3 cells, DFO greatly improved event-free survival (Fig. S8 a). There was no significant difference in event-free survival between mice treated with DFO and DFO + VD. However, a significant reduction in tumor growth was observed only in mice that received the DFO + VD association, although the DFO-treated mice presented reduction in tumor growth compared with control which did not reach significance (Fig. S8 b).

Because both iron chelators and VD analogues are safe drugs currently used as therapies in the treatment of diseases other than leukemia, their effectiveness in one AML patient was evaluated. A 69-yr-old man had a recent transformation of MDS into AML. After high-dose chemotherapy, the patient remained in non-blastic aplasia. 7 mo after the diagnosis, blasts reappeared in the blood. High-dose chemotherapy was denied, and treatment with an iron chelator (1 g/d DFX) and 25-hydroxycholecalciferol (4,000 IU/d) was initiated.

Blast counts, although progressing before therapy, were decreased upon iron chelation and vitamin D therapy, and pancytopenia was partially reversed, accompanied by an increase in monocyte numbers (Fig. 9 a) without inducing hypercalcemia (Fig. 9 b). Importantly, May Grünwald Giemsa (MGG) blood smears revealed an increased number of differentiating cells toward the monocyte lineage, suggesting that they were derived from the blast pool (Fig. 9 c). Cell sorting (Fig. 9 d), followed by fluorescent in situ hybridization analysis, showed that the original trisomy 8 found in the blasts of the patient (Fig. 9 e) was segregated with the sorted mature monocyte population but not with lymphocytes or NK cells (Fig. 9 f), confirming that the circulating monocytes were derived from the blast pool. In addition, the patient’s quality of life was improved because red cell and platelet transfusion needs were decreased. Furthermore, the patient could control several microbial infections that otherwise would have been life threatening, suggesting that the patient’s immune status was partially restored. These data underscore the differentiating efficiency of combined therapy in human AML.

We show in this paper that iron homeostasis plays a key role in the control of myeloid differentiation in both normal and pathological situations. Iron deprivation of undifferentiated myeloid cells, either through iron chelation or through blocking of iron uptake, induced transcription of several markers of the monocyte/macrophage lineage. At the same time, iron deprivation shunted progenitors away from the granulocytic lineage by dampening the transcription of granulocyte-specific markers. Morphological, phenotypic, and functional analyses confirmed and extended these observations. Not only did iron deprivation induce primary progenitor commitment toward the monocyte lineage, it also reestablished differentiation in AML blasts/cell lines.

Recent cumulated data in the literature have shown that ROS can modulate the activation of signal transduction pathways involved in several cellular functions such as cell proliferation and differentiation (Sattler et al., 1999). Specifically, in the mammalian hematopoietic system, it has been recently reported that hematopoietic stem cells produce low levels of ROS, which are increased in common myeloid progenitors. Modulation in ROS levels seems to be important in the control of myeloid precursor differentiation (Tothova et al., 2007). This process seems to be conserved throughout evolution because Drosophila melanogaster multipotent hematopoietic progenitors display increased levels of ROS which are down-regulated during differentiation (Owusu-Ansah and Banerjee, 2009). However, factors controlling ROS production in myeloid progenitors are not well known. We show in this paper that intracellular LIP could be one of these factors playing a critical role in controlling ROS levels and, thus, cell differentiation in the hematopoietic system.

Scavenging ROS from Drosophila melanogaster hematopoietic progenitors delays mature blood cell differentiation, which is dependent on the MAPK JNK (Owusu-Ansah and Banerjee, 2009). In this paper, we show that both genetic and pharmacological inactivation of the MAPK–JNK pathway inhibited iron deprivation-induced cell differentiation. Furthermore, pharmacological blocking of ROS abrogated iron deprivation–induced cell differentiation, suggesting that ROS and MAPK pathways are linked, most probably by the direct effect of ROS on MAPK activation (Martindale and Holbrook, 2002). Signal transduction cascades of MAPK pathways frequently converge to amplify signaling (Johnson and Lapadat, 2002; Lopez-Bergami et al., 2007). Inhibition of ERK and p38 pathways prevented cell differentiation induced by iron deprivation. Blocking of the ERK pathway did not block apoptosis induced by iron chelators, suggesting that redundant and nonredundant MAPK signaling pathways are involved in the control of AML cell differentiation and apoptosis. In agreement, it has been previously shown that the MAPK–ERK pathway is constitutively active in AML cells and in G-CSF–mobilized CD34⁺ precursors and inhibition of ERK pathway promotes growth arrest rather than apoptosis (Ricciardi et al., 2005). Combining iron chelation therapy with other MAPK/ERK activators, such as G-CSF, remains to be evaluated. Altogether those data suggest that cross talk between MAPK pathways could cooperate to regulate AML cell differentiation and its sensitivity to cytotoxic agents.

In the majority of the experiments presented in this paper, TfR1 targeting by A24 mimicked the effect of iron chelators at least in in vitro and ex vivo settings. Because each of the iron-chelating agents has its own chemical/biochemical characteristics (such as the structure or cell-permeant properties of the different iron chelators and cell targeting specificity of the monoclonal antibody) and pharmacological properties (such as the kinetics and half-life), the intensity of the response varied among them. For example, in tumor xenograft experiments, A24 mAb totally inhibited tumor engraftment, whereas iron chelators delayed the tumor growth, a difference probably involving additional effects of A24 such as antibody-dependent cellular cytotoxicity or complement-dependent cytotoxicity. In addition, A24 could exert a specific target effect by inducing TfR1 down-regulation specifically on TfR^high tumor cells (Moura et al., 2004; Lepelletier et al., 2007; Callens et al., 2008), whereas iron chelators are more likely to induce TfR1 messenger RNA (mRNA) stability by posttranscriptional mechanisms.

The principal molecular mechanism drawn from the experiments reported in this paper links the signaling pathways activated by iron deprivation and by VD. Gene expression studies showed a high similarity (30% common genes) between genes modulated by VD and the set of 105 genes targeted by iron deprivation. In addition, iron deprivation induced both VDR gene expression and VDR phosphorylation. The phosphorylation of VDR on serine 208 by casein kinase II has been shown to control activation of VDR gene transcription and VDR target genes expression (Jurutka et al., 1996). Therefore, iron deprivation could induce, by itself, the activation of VDR signaling pathway. Furthermore, the combination of iron deprivation agents and differentiating therapies acted synergistically, increasing AML cell differentiation.

Altogether, our data identified a mechanism involving sequentially iron-deprivation, ROS production, and activation of MAPK pathways, further amplified by induction of VDR expression and phosphorylation, culminating in monocyte/macrophage differentiation. Recent studies showed an unexpected constitutive activation of Syk tyrosine kinase in AML blasts (Hahn et al., 2009). Gefitinib has been identified as a Syk tyrosine kinase inhibitor (Hahn et al., 2009) and reported to be a potent inducer of AML cell differentiation in preclinical settings (Boehrer et al., 2008). These studies unveiled Syk as a previously unsuspected regulator of AML differentiation. Although the molecular pathways regulating the activation of Syk in AML cells remain unknown, a study by Agramonte-Hevia et al. (2003) reported that VD-induced cell differentiation down-regulated the high Syk activation observed in undifferentiated THP-1 cells. Therefore, down-modulation of Syk activity downstream VDR activation could participate in cell differentiation induced by iron deprivation agents and it remains to be evaluated.

VD is known to induce myeloid progenitors to differentiate into monocytes. In vitro and ex vivo experiments were promising, but in the first clinical trials in AML and MDS, only partial differentiation was obtained and clinical improvements were modest (Deeb et al., 2007). In addition, symptomatic hypercalcemia limited VD clinical application. Clinical trials using VD analogues, which do not promote hypercalcemia, did not improve patients’ prognosis (Koeffler et al., 2005), suggesting that VDR stimulation alone is not sufficient to promote clinical benefit. We show in this paper that iron deprivation could be a new therapeutic approach in AML, and likely in MDS, because it synergizes with VD (low dose). The combination of iron chelators and VD induced both pro-differentiating and antiproliferative effects in leukemia cells from one AML patient, as described in this paper in the form of a case report. This association reduced the rate of blast proliferation and promoted blast differentiation toward monocytes without emergence of hypercalcemia. The patient developed multiple infectious episodes, which were greatly reduced after the initiation of VD and iron chelation combined therapy. In vitro experiments showed that, in addition to its role in monocyte differentiation, the associated therapy promoted expression of genes related to antimicrobial defense such as cathelicidin. Therefore, the early use of iron chelators in AML therapy, in addition to preventing relapses and enhancing time to progression, could act at multiple levels: by preventing red cell transfusion-derived secondary iron overload, by avoiding defective bone marrow reconstitution after allogeneic HSCT, and also by improving patients’ immune status (Pullarkat et al., 2008; Pullarkat, 2009).

Because AML incidence increases with age and because the world population is growing older, we propose that the combined therapy would be particularly useful in elderly patients who are not eligible for high dose chemotherapy and bone marrow transplantation. This combined therapy could be also relevant in other diseases involving a deregulation of BM differentiation, such as MDS. Additional clinical studies are needed to validate its efficacy and determine in which patients it will be more beneficial.

Peripheral blood cells were obtained from AML patients and healthy donors after obtaining their written informed consent (Table S2, additional information) and approval obtained by the Necker and Saint-Louis Hospitals Institutional ethic committees. Peripheral blood was collected at the initial diagnosis before the administration of treatment. Mononuclear cells were separated by Ficoll-hypaque (PAA laboratories) density centrifugation and resuspended in IMDM (Invitrogen) supplemented with 15% FCS (Hyclone), 100 ng/ml stem cell factor (SCF), 10 ng/ml IL-3, and 25 ng/ml FLT3-L (all purchased from R&D Systems). Myelocytic (HL60 and OCI-AML3), monoblastic (THP1 and U937), and promyelocytic (NB4) leukemia cell lines used in the present study were provided by J.M. Cayuela (Hôpital Saint-Louis, Paris, France) and P. Villarese (Hôpital Necker, Paris, France). Cells were cultured in RPMI-1640 medium (Invitrogen) supplemented with 5% FCS and antibiotics.

DFX was provided by Novartis. Etoposide (VP16), ATRA, DFO, 25-OH-VD, H₂O₂, FeCl_3, NAC, and PDTC were purchased from Sigma-Aldrich. Calcein-AM, DHR123, and DHE were purchased from Invitrogen. The mAb A24 was produced and purified as described previously (Moura et al., 2001). The inhibitors of JNK (SP600125), ERK (PD98059), and p38 (SB203580) were obtained from Santa Cruz Biotechnology, Inc.

For proliferation assays, the HL60 cell line was resuspended in RPMI-1640 medium with 5% FCS, whereas the patient blasts were resuspended in IMDM with 15% BIT 9500 (STEMCELL Technologies Inc.). The cells were plated in triplicate at 2.5 × 10⁴ cells per well in 96-well tissue culture plates (Falcon) in 200 µl of culture medium. Proliferation was measured as described previously (Lepelletier et al., 2007).

LIP was evaluated according to a previously described method (Espósito et al., 2002) with some modifications. In brief, cells were treated with 0.5 µM calcein-AM for 30 min at 37°C. Before being subjected to flow cytometry analysis, cells were treated or not with an excess of cell-permeant iron chelators. The fluorescein channel was used to detect calcein fluorescence. The difference between calcein fluorescence before and after adding cell-permeant iron chelators indicated the relative amount of LIP (Espósito et al., 2002).

Fluorescence analysis was used to determine the relative levels of ROS in response to various stimuli. ROS levels were detected by FACS analysis. Labeling with cell-permeant redox-sensitive fluorochromes was performed for 30 min at 37°C in the presence of 1 µM DHE or 0.5 µM DHR123, both diluted in 0.5% BSA in HBSS solution (Invitrogen).

For overexpression, pEGFP-C3 (Takara Bio Inc.) expression vectors containing cDNA encoding JNK1 or JNK2 between the HindIII/BamHI sites (provided by T. Herdegen, Institute of Pharmacology, University Hospital Schleswig-Holstein, Kiel, Germany) were used. These proteins were expressed as a GFP fusion protein under the control of human CMV immediate early promoter.

For C-Jun, JNK1, and JNK2 gene silencing, artificial miRNA constructs were generated. Four sets of predesigned miR RNAi sequence oligonucleotides for each gene were obtained using BLOCK-iT miR RNAi Select (Invitrogen) and checked to ensure the designed sequences would not target other gene transcripts to avoid any off-target effects. These oligonucleotides were annealed and ligated into the pcDNA6.2-GW/EmGFP-miR vector according to the manufacturer’s instructions (Invitrogen) to express the pre-miRNA as an EmGFP fusion under the control of human CMV immediate early promoter. The pcDNA 6.2-GW/EmGFP-miR–negative control plasmid contains a scrambled miRNA insert, which is predicted not to target any known vertebrate gene.

HL60 cells were transfected with the plasmids using a nucleofector (Lonza) as described according to the manufacturer’s instructions (usually 40% cells were transfected). 5 µM DFO was added in culture medium 48 h after transfection and flow cytometry analysis was realized 24 h later on the basis of the EmGFP (for miRNA) or EGFP (for protein constructs) expression. Validation of miRNA constructs was performed on CHO cells electroporated with each miRNA-containing plasmid using the nucleofector and analyzed in Western blotting after 24 h to validate the constructs.

FISH was performed using standard protocols with specific centromeric probes for chromosome 8 (pZ8.4) and chromosome 12 (pBR12; provided by M. Rocchi, University of Bari, Bari, Italy) used as control.

Total RNA from time-matched controls and treated samples was extracted using NucleoSpin RNA II (Macherey-Nagel). After DNase treatment, first-strand cDNA was synthesized using SuperScript II reverse transcription (Invitrogen). Quantitative real-time PCR was performed using a Chromo-IV PCR System (MJ Research), and the PCR products were quantified using Sybr Green Technology (Jumpstart MasterMix; Sigma-Aldrich). The results of the real-time quantitative PCR were analyzed using the delta-delta Ct method (Livak and Schmittgen, 2001). The complete list of primer sequences is available in Table S4.

For the microarrays, the Agilent 44 K Whole Human Genome (G4112A) long (60 bp) oligonucleotide microarray and the dual-color analysis method were used with probes from the treated samples and from reference RNA that were differentially labeled with cyanine 5 and cyanine 3. These microarrays have 44,290 features with 41,000 distinct oligonucleotides belonging to 33,715 sequences defined by their accession number. cRNA from each treated sample was labeled with Cy3-cytidine triphosphate (CTP) and the untreated HL60 RNA reference pool with Cy5-CTP for direct comparison. Reverse transcription, linear amplification, cRNA labeling, and cRNA purification were performed with the linear amplification kit (Agilent Technologies). The microarray data were mainly analyzed with the Resolver software (Merck). All data were filtered to eliminate low-intensity values <50 arbitrary units for both colors. The microarray data analyzed in this paper have been submitted to the Array Express data repository at the European Bioinformatics Institute (http://www.ebi.ac.uk/arrayexpress/) under accession no. E-TABM-925.

For tumor establishment, 5 × 10⁶ HL60 or 2.5 × 10⁶ OCI-AML3 cells were mixed with Matrigel (1:1, vol/vol) and injected subcutaneously into 8–10-wk-old female athymic Nude mice. The mice were then injected i.p. with A24 (40 mg/kg on day 1), DFO (20 mg/day, 5–7×/wk), VD (1 µg, 2×/wk), and the combination of DFO and VD or PBS as a vehicle control. The tumor growth was measured as previously described (Lepelletier et al., 2007).

Serial sections (3 µm) in paraffin blocks were deparaffinized in xylene and hydrated in a graded series of alcohol. Staining was performed using the Immunohistochemical Autostainer (Laboratory Vision Corporation) with primary antibodies against CD11b (Sigma-Aldrich, 1:200), p-ERK, and p-JNK (Cell Signaling Technology, 1:200), biotinylated anti-Ig, and streptavidin-peroxidase. The antibody binding was developed in PBS with 0.05% DAB and 0.003% H₂O₂. Hematoxylin-eosin (HE)–stained sections were used to evaluate cytomorphology and to detect apoptotic cells with condensed nuclei. Apoptosis was assessed by the TUNEL technique according to the manufacturer’s instructions (Roche). The images were obtained using a microscope (DM-2000; Leica) equipped with a digital camera (DFC320; Leica) and IM50 software. Data were quantified using Image J software (National Institutes of Health).

CD34⁺ cord blood cells were isolated by immunomagnetic separation (Miltenyi Biotec) and cultured in IMDM supplemented with 15% BIT 9500 (Stem Cell Technologies), 100 ng/ml SCF, 10 ng/ml IL-3, 25 ng/ml G-CSF, and 10 ng/ml GM-CSF for myelomonocytic differentiation (all cytokines were purchased from R&D Systems). 10 µg/ml A24, 5 µM DFO, or 3 µM DFX was added iteratively every 3 d in the culture medium.

For colony-forming assays, 10³ CD34⁺ cells were plated in quadruplicate in semisolid methylcellulose medium (H4434; Stem Cell Technologies) in the presence or absence of 10 µg/ml A24, 5 µM DFO, or 3 µM DFX. At day 14, colonies containing >50 cells were counted and classified as granulocyte colony (CFU-G), monocyte-macrophage colony (CFU-M), mixed granulocyte-monocyte colony (CFU-GM), or erythroid colony (CFU-E). The colonies were counted by standard criteria in 60-mm gridded scoring dishes with cross marks (Stem Cell Technologies) under an inverted microscope. To verify if iron deprivation leads to an altered formation of myelomonocytic colonies, we represented the ratio between CFU-G and CFU-M in both nontreated and treated groups.

Cell differentiation was determined by the following criteria: cell morphology was evaluated after MGG staining. Esterase activity was detected using the LEUCOGNOST EST kit (Merck). The expression of CD11b (IM2581U; Beckman Coulter) and CD14 (A07765; Beckman Coulter) was evaluated by flow cytometry (FACSCalibur; BD). Cell death was evaluated using the annexin-V/propidium iodide (PI) double staining kit (BD) according to the manufacturer’s instructions.

All experiments were made with serum-starved cells (12–16 h) to avoid detection of constitutive phosphorylation induced by growth factors (Wang and Proud, 2007). Whole cell extracts were prepared as described previously (Amir-Moazami et al., 2008). In brief, cell pellets (10⁷ cells) were washed twice with PBS and then resuspended in 1 ml of lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM h-glycerophosphate, 1 mM Na₃VO₄, 1 mM phenylmethylsulfonyl fluoride, 1 µg/ml leupeptin, and 1 µg/ml aprotinin). The protein concentration was determined using the BCA method (Thermo Fisher Scientific) and an equal amount of SDS sample buffer containing 150 mM Tris-HCl, pH 6.8, 30% glycerol, 3% SDS, 1.5 mg/ml bromophenol blue, and 100 mM DTT was then added to each sample. For immunoblotting, 40 µg of protein was separated in 4–12% or 7% SDS-PAGE and transferred onto a PVDF membrane. The blots were first probed with an anti-phospho protein, p-ERK (Thr202/Tyr204), p-JNK, p-p38 (all from Cell Signaling Technology), or p-VDR (Ab63572; Abcam) antibody. The membranes were stripped, dried, and probed with an antibody against total ERK (Cell Signaling Technology), VDR (Ab39990; Abcam), or HSC70 (sc-71270; Sants Cruz Biotechnology, Inc.). The protein bands were visualized using a chemiluminescence system.

CD14 expression was used to evaluate the combined efficacy of DFO, vitamin D, and mAb A24. In each experiment, HL60 cells were treated with serial dilutions of each agent individually and with a fixed ratio of two agents simultaneously. The results were analyzed using the median effect/CI isobologram method of Chou and Talalay (1984). F(a) is the fraction of cells affected (i.e., CD14⁺) and f(u) is the fraction of cells unaffected (i.e., CD14⁻). F is the ratio f(a)/f(u) and D is the drug concentration. Log(F) was plotted against log(D). The dose response is displayed as a linear regression line and the x-intercept correspond to the log(IC50). The CI was defined as follows: CI = (D)₁/(D_x)₁ + (D)₂/(D_x)₂, where (D_x)₁ and (D_x)₂ are the doses required to achieve a given effect level for each drug, i.e., a specified value of f(a). (D)₁ and (D)₂ are the doses of each drug in a given mixture which gives the same f(a). When CI values are 1, <1, or >1, it may be concluded that summation, synergism, or antagonism, respectively have been observed at the particular level examined. The CI was calculated based on the conservation assumption of mutually nonexclusive drug interactions.

The data are expressed as the mean ± SEM. Statistical analyses were performed using Prism 5 software (GraphPad Software, Inc.). We used the Student’s t test or the Mann-Whitney test to compare two groups, and multigroup comparisons were made using a one-way ANOVA followed by a post-hoc Bonferroni test. We used the Kruskal-Wallis test followed by a post-hoc Dunn test for nonparametric comparisons, where indicated. To compare tumor-free animal curves, the Log-rank test was used. The results were considered statistically significant at a p-value <0.05 (*), <0.01 (**), or <0.001 (***).

Fig. S1 shows that iron deprivation induces morphological and phenotypical modifications of monocytic cells related to cellular differentiation. Fig S2 shows that iron deprivation inhibits proliferation and induces differentiation of AML blasts. Fig. S3 shows the in vitro granulocytic cord blood progenitor differentiation system. Fig. S4 shows detection of intracellular LIP and ROS in AML cells. Fig. S5 shows the efficacy of miRNA constructs targeting c-JUN, JNK1, and JNK2. Fig. S6 shows determination of the synergic versus additive effect of the combined DFO/VD and A24/VD therapies on cell differentiation. Fig. S7 shows that A24 and iron chelators do not synergize with ATRA to induce differentiation. Fig. S8 shows the anti-tumor effect of iron deprivation and iron deprivation combined with VD in Nude mice xenografted with AML cell lines.

The authors would like to thank the members of the hematology and obstetric units (Necker Hospital and St. Louis Hospital, Paris, France) for collecting patient samples and Dr. T. Herdegen (Institute of Pharmacology, University Hospital Schleswig-Holstein, Kiel, Germany) for JNK1 and JNK2 expression plasmids.

This work is supported by the Fondation pour la Recherche Médicale, Fondation de France, Association Laurette Fugain, Association de la recherche contre le cancer, Institut National de la Santé et de la Recherche Médicale (INSERM)/PNRNU2007, and Novartis grants. This work was also supported by a donation from Cyril Mangin’s family and Friends. C. Callens was a recipient of an INSERM Poste d’accueil fellowship.

The authors have no conflicting financial interests.