The Novel Arsenical Darinaparsin Circumvents BRG1-Dependent, HO-1-Mediated Cytoprotection in Leukemic Cells

ORIGINAL ARTICLE The novel arsenical Darinaparsin circumvents BRG1-dependent, HO-1-mediated cytoprotection in leukemic cells

N Garnier1,2, LA Petruccelli1,2, MF Molina1,2, M Kourelis1,2, S Kwan1,2, Z Diaz1,2, HM Schipper2, A Gupta3, SV del Rincon1,2, KK Mann1,2 and WH Miller Jr1,2

Darinaparsin (Dar) is a more potent cytotoxic arsenical than arsenic trioxide (ATO). We hypothesized that the increased cytotoxicity of Dar may be because of a decreased cytoprotective response. We observed that, unlike ATO, Dar does not induce heme oxygenase-1 (HO-1), even though it induces expression of other nuclear factor (erythroid-derived 2)-like 2 (NRF2)-dependent detoxifying enzymes to a greater extent than ATO, in both cancer cell lines and patient-derived leukemic cells. This strengthens the emerging evidence, showing that response to reactive oxygen species (ROS) is stimuli speciﬁc. Dar treatment prevents recruitment of the transcriptional coregulator Brahma-related gene 1 (BRG1) to the HMOX1 promoter, which is required for HMOX1 expression. The inability of Dar to induce HO-1 correlates with arrest in G2/M cell cycle phase and BRG1 phosphorylation. Inhibition of HO-1 increases the toxicity of ATO, but has no effect on Dar-induced apoptosis. Accordingly, the lack of HO-1 induction is involved in Dar’s enhanced antileukemic properties. Our data highlight cytoprotective responses mediated by HO-1 and BRG1 as a novel target for enhancing the therapeutic range of arsenicals.

Leukemia (2013) 27, 2220–2228; doi:10.1038/leu.2013.54 Keywords: arsenic; leukemia; Darinaparsin; HO-1; BRG1; NRF2

INTRODUCTION we reviewed.1 We postulate that within the signaling differences Darinaparsin (Dar; ZIO-101, S-dimethylarsino-glutathione) is a lies the mechanism responsible for the enhanced antitumor effect promising new arsenic-based, anticancer drug candidate, of Dar. which is currently undergoing clinical studies in both hematolo- In response to oxidative stress, cells induce an arsenal of gical malignancies and solid tumors.1 This organic arsenical, protective proteins including the nuclear factor (erythroid-derived which consists of dimethylarsenic conjugated to glutathione, 2)-like 2 (NRF2) gene battery, which includes heme oxygenase-1 shows significant activity against multiple cancers in vitro.2 (HO-1, HMOX1 gene), NAD(P)H:quinone oxidoreductase (NQO1), Another arsenical, arsenic trioxide (ATO), is a proven thioredoxin-reductase-1 (TR1, TXNRD1 gene) and glutamate- 15 chemotherapy for acute promyelocytic leukemia (APL), but has cysteine ligase regulatory subunit (gGCS). NRF2 is a basic substantially reduced antitumor effects at clinically achievable leucine zipper transcription factor that mediates the expression of doses in other malignancies.3–5 We have shown previously that these key protective enzymes by binding antioxidant response Dar is a more potent inducer of apoptosis than ATO in various elements (ARE)16–18 in the regulatory regions of these genes. malignant cell lines and is highly active against APL cells.2 Under unstressed conditions, Kelch-like ECH-associated protein 1 Importantly, Dar has a maximum tolerated dose that is 50-fold confines NRF2 in the cytoplasm, where it is constitutively higher than ATO in mice and is active against xenograft tumors.6 degraded.19–21 In contrast, ROS liberate NRF2 from Kelch-like However, the mechanisms of its increased antitumor efficacy have ECH-associated protein 1, leading to NRF2 translocation and not been elucidated. accumulation in the nucleus.15 Transcriptional regulation of some Oxidative damage is postulated to be a key mechanism by of these detoxifying enzymes involves a competition between the which arsenicals initiate the apoptotic process.7–12 Consistent with activator NRF2 and the repressor BTB, and CNC homolog 1 its increased cytotoxicity, we have reported that Dar induces (BACH1) for interaction with their AREs.22 Of the NRF2 target - significantly more superoxide (O2 ), hydrogen peroxide and genes, regulation of HMOX1 expression is unique because it is oxidative damage than ATO in arsenic-sensitive NB4 and further regulated by the binding of Brahma-related gene 1 (BRG1, arsenic-resistant AR2 cells.2 Both ATO and Dar-induced ROS SMARCA4 gene), the catalytic subunit of SWI2/SNF2-like chromatin- 23 activate the c-Jun NH2-terminal kinase/stress-activated protein remodeling complex, to the ARE. kinase pathway, leading to apoptosis.2,13,14 However, we have also Here, we show that Dar selectively fails to induce HO-1 reported that Dar triggers apoptosis by inducing cellular expression in leukemic cell lines in vitro and in APL patient blasts responses that do not completely overlap with ATO.1,2 The ex vivo, even though it induces the expression of other NRF2 characterization of similarities versus differences in triggered target genes. We provide evidence that Dar prevents BRG1 pathways and intracellular ROS modulation between ATO and recruitment to the HMOX1 promoter. We also show that specific Dar have elicited interest among the labs working on arsenic, as inhibition of HO-1 enhances the toxicity of ATO, but not Dar, in

1Department of Oncology, Lady Davis Institute for Medical Research, Segal Cancer Center, Sir Mortimer B. Davis Jewish General Hospital, McGill University, Montreál, Quebec, Canada; 2Faculty of Medicine, Department of Experimental Medicine, McGill University, Montreál, Quebec, Canada and 3Osta Biotechnologies, Inc., Dollard-des-Ormeaux, Quebec, Canada. Correspondence: Dr WH Miller Jr, Department of Oncology, Lady Davis Institute for Medical Research, Segal Cancer Centre, Sir Mortimer B. Davis Jewish General Hospital, McGill University, 3755 Coˆte Ste-Catherine Road, Montreal, Quebec, Canada H3T 1E2. E-mail: [email protected] Received 9 November 2012; revised 4 February 2013; accepted 11 February 2013; accepted article preview online 21 February 2013; advance online publication, 15 March 2013 Darinaparsin prevents heme oxygenase-1 induction N Garnier et al 2221 leukemic cells. This suggests that the inability of Dar to induce have previously shown that Dar induces more ROS than ATO.2 HO-1 expression is partially responsible for the increased We hypothesized that the enhanced cytotoxicity of Dar is because pro-apoptotic potential we observe. of its inability to elicit a complete antioxidant enzyme response. Therefore, we investigated whether Dar induced significant levels of HO-1, the prototypic arsenic-induced phase II antioxidant MATERIALS AND METHODS enzyme. Figure 1b shows that Dar did not induce HO-1 protein Ethics statement expression in any of the tested cell lines, as opposed to ATO that The patient samples were obtained, collected and used in accordance to strongly induced expression in all cell lines. We were able to the ethical requirements and regulations of the Jewish General Hospital. reproduce these results in HeLa cells, suggesting that our results Informed written consent was obtained from all patients involved. are not limited to hematological malignancies (Figures 1a and b). To verify this observation, we treated both NB4 and U937 cells Cell culture with ATO or Dar and assessed HMOX1 mRNA levels by quantitative

All cells were grown in a humidified chamber at 37 1C with 5% CO2. NB4, PCR (qPCR). Treatment with Dar failed to induce significant HMOX1 U937, U266, Kasumi-1, LY8 and HeLa cells were obtained from the expression in either cell line. This result remained consistent in american type culture collection (ATCC) and cultured as recommended. NB4 (Figure 1c) and U937 (Figure 1d), regardless of treatment dose Blasts from APL patients were isolated from either peripheral blood or and/or duration. In addition, we tested all the cell lines in Figure 1a bone marrow (see Supplementary Materials). and found that Dar did not induce significant HMOX1 mRNA expression in any of them (Supplementary Figure 1). Cell viability and caspase-3/7 assay Cell viability and caspase-3/7 activity were assayed using CellTiter-Glo Dar fails to induce HO-1, but not other NRF2 target genes, in NB4 Luminescent Cell Viability Assay and Caspase-Glo 3/7 Assay (Promega, cells Madison, WI, USA) according to manufacturer’s instruction (see Supplementary Materials). We investigated whether Dar would induce significant levels of other phase II antioxidant enzymes, which are induced through activation of the NRF2-signaling pathway.15,17,18 NB4 cells were Propidium iodide staining treated for 3 or 6 h with increasing concentrations of ATO or Dar. As 2 Quantification of apoptotic cells was performed as previously described expected, HO-1 protein levels increased with ATO, but not with Dar (see Supplementary Materials). (Figure 2a). In contrast, Figures 2b, c and d show that both ATO and Dar increase, in a time- and dose-dependent manner, protein levels Protein quantification of TR1, gGCS and NQO1, with Dar consistently inducing expression Cells were lysed, whole-cell extracts were run on SDS-polyacrylamide gel at lower concentrations. To expand our observations, we also tested electrophoresis followed by western blot analysis (see Supplementary the ability of ATO and Dar to induce TR1 and gGCS protein levels in Table 1). Levels of phospho-DNA-PKcs (T2604) were measured by flow Kasumi-1 and U937, which are two non-APL cell lines. As expected, cytometry (see Supplementary Materials). NRF2 protein levels were both ATO and Dar increased protein levels of TR1 and gGCS assessed by densitometry using Image J (imagej.nih.gov). The intensity of each NRF2 band is normalized with the corresponding loading control (Supplementary Figure 2A and 2B). NQO1 protein expression is band (GAPDH or H3) and compared with the corresponding nuclear or shown as a time course because of the transient nature of its cytoplasmic control-treated bands, which were given the value of 1. expression. NQO1 protein levels increase quickly following exposure to arsenicals but decrease at 6 h (Supplementary Figure 2C). Our mRNA analysis data indicate that the activation of HO-1 expression in response to NRF2 signaling is different than for the other NRF2 target genes. Total mRNA was isolated by using TRIzol. Complementary DNA were amplified using Taqman or SYBR green technology (see Supplementary Materials and Supplementary Table 2). Dar is more toxic than ATO and does not induce HO-1 in APL cells from patients Chromatin immunoprecipitations Although we show that Dar fails to induce HO-1 expression in Chromatin immunoprecipitations (ChIP) were performed as adapted from several cell lines, we wanted to confirm these findings with a previously described method23 (see Supplementary Materials and primary APL blasts, the only malignancy in which ATO has proven Supplementary Table 1 and 2). to hold clinical benefit. Therefore, blasts were obtained from APL patients, and were treated ex vivo with ATO and Dar. Dar induces Short hairpin RNA knockdown more cell death as shown by cell viability assay and caspase-3 NB4 cells were transduced with TRC Mission lentiviral particles (SIGMA) activity assay (Figures 3a and b). We also quantified HMOX1 mRNA according to manufacturer’s instruction (see Supplementary Materials and levels in response to ATO and Dar. Figure 3c shows that even Supplementary Tables 1 and 2). though ATO treatment induced significant HMOX1 expression, Dar treatment was not followed by an increase in HMOX1 mRNA levels, Statistical analysis regardless of dose. As shown in Figure 3d, Dar treatment led to increased TXNRD1 mRNA levels in APL patient blasts treated Significance was determined by one-way analysis of variance followed ex vivo, confirming in patient cells the results obtained in cell lines by Newman–Keuls post-tests using Prism version 3.0 (GraphPad software, San Diego, CA, USA). *Po0.05; **Po0.01, ***Po0.001. wherein HO-1 is the only NRF2 target gene not induced by Dar. This suggests that this surprising observation is relevant in vivo. Figure 3a, b, c and d are summary results obtained from six APL RESULTS patients. Detailed results for individual patients are displayed in Dar is more toxic than ATO and fails to induce HO-1 in various Supplementary Figures 3, 4, 5 and 6. malignant cell lines We have previously shown that Dar induces more apoptosis than Dar does not induce the recruitment of BRG1, a necessary ATO in NB4 cells, a model for the study of human APL, and in U937 transcription coactivator, at the HMOX1 promoter cells. To expand these observations, we tested the effect of Dar We sought to investigate the mechanism for the selective failure and ATO on six cell lines isolated from various hematological of Dar to activate HMOX1. HMOX1 is a well-known target gene for malignancies, including NB4 and U937. Throughout the panel, Dar the transcription factor NRF2. Therefore, we investigated whether was more potent at inducing apoptosis, as shown in Figure 1a. We Dar failed to induce HMOX1 expression because of an altered

& 2013 Macmillan Publishers Limited Leukemia (2013) 2220 – 2228 Darinaparsin prevents heme oxygenase-1 induction N Garnier et al 2222

Figure 1. Dar is more toxic than ATO and fails to induce HO-1 in various malignant cell lines. Cells from various cell lines were treated with equimolar concentrations of ATO and Dar (2 or 5 mM) for 24 h. Cell death was then quantified by propidium iodide staining followed by FACS analysis (a), and HO-1 protein levels were quantified by western blot (b). NB4 cells were treated with 1 and 2 mM ATO or Dar for 6 h (c), or 2 mM ATO or Dar for 3 and 6 h (d). HMOX1 mRNA levels were quantified by reverse transcriptase-qPCR. HMOX1 mRNA levels were also quantified in U937 cells after ATO or Dar treatment (dose response at 6 h (c left and d left) and time course at 2 mM (c right and d right). *Po0.05, **Po0.001, ***Po0.001.

Leukemia (2013) 2220 – 2228 & 2013 Macmillan Publishers Limited Darinaparsin prevents heme oxygenase-1 induction N Garnier et al 2223 NRF2-signaling pathway. We first examined an early event in NRF2 signaling, the translocation of NRF2 from the cytoplasm to the nucleus. We treated NB4 cells with ATO and Dar, and then performed western blot analyses of the protein content of the cytoplasm versus nucleus. As shown in Figure 4a, both ATO and Dar induce NRF2 stabilization and translocation to the nucleus, in a time-dependent manner. This confirms that NRF2 is competent to signal after Dar treatment, as indicated by upregulation of TR1, NQO1 and gGCS. In addition, we examined whether ATO or Dar altered expression of a heterodimer partner of NRF2, the Small Maf protein G (sMafG). sMafs are required for NRF2 binding to AREs, and activation of target genes.24 Figure 4a shows that sMafG25 protein levels in the nucleus did not vary between ATO and Dar treatment. Next, we focused on whether NRF2 signaling at the HMOX1 promoter was impaired. We investigated the events at the HMOX1 AREs subsequent to treatment with ATO or Dar. In the presence of intracellular ROS, BACH1 is released from the AREs and NRF2 is recruited to the AREs.22,26 Therefore, we performed ChIP assays in NB4 cells treated with ATO or Dar and assessed binding of BACH1 and NRF2 to both the 9 kb and the 3 kb AREs.22 Both ATO and Dar caused BACH1 release from AREs in HMOX1 and TXNRD1’s promoters (Figure 4b). A lesser effect was observed at the TXNRD1 ARE, which is consistent with a minor role for BACH1 in the regulation of TXNRD1 transcription.26 Moreover, both ATO and Dar led to NRF2 binding to HMOX1 and TXNRD1 AREs (Figure 4c). However, NRF2 recruitment to HMOX1’s AREs is significantly less after Dar treatment compared with ATO (Figure 4c). In contrast, we observed NRF2 recruitment at the TXNRD1 ARE after treatment with either ATO or Dar (Figure 4c), indicating that the decreased NRF2 recruitment to the HMOX1 ARE is specific. HMOX1 transcriptional activation is distinct from other NRF2 Figure 2. Dar fails to induce HO-1, but not other NRF2 target genes, targets, because it further requires the recruitment of BRG1 to in NB4 cells. HO-1 (a), TR1 (b) and gGCS (c) protein levels were the promoter. BRG1 exhibits helicase activity and is able to fold quantified by western blot in NB4 cells after treatment with the chromatin, leading to a Z-DNA structure, in which BRG1 increasing doses of ATO and Dar for 6 h, or for 1 and 3 h with 2 mM interacts with AREs at À 3 kb and À 9 kb of the HMOX1 promoter ATO or Dar for NQO1 (d). region. Only then can transcription be initiated.23 We assessed

Figure 3. Dar is more toxic than ATO and does not induce HMOX1 in APL cells from patients. Blasts from six APL patients were isolated by Ficoll and treated ex vivo with 2 mM ATO or Dar for 48 h. Cell viability (a) and caspase-3/7 activity (b) were subsequently assayed. HMOX1 mRNA levels (c) and TXNRD1 mRNA levels (d) were quantiﬁed after 6 h treatment (c) by reverse transcriptase-qPCR, as well as TXNRD1 mRNA levels (d). Data are represented as mean±s.e.m. *Po0.05, ***Po0.001.

& 2013 Macmillan Publishers Limited Leukemia (2013) 2220 – 2228 Darinaparsin prevents heme oxygenase-1 induction N Garnier et al 2224

Figure 4. Dar does not induce the recruitment of BRG1, a necessary transcription coactivator, at the HMOX1 promoter. NB4 cells were treated for 1, 3 and 6 h with 2 mM ATO or Dar. Nuclear and cytoplasmic protein levels were quantiﬁed after cyto-nuclear fractionation followed by western blot (a). Intensities of NRF2 bands, measured by densitometry, are indicated as arbitrary units below the ﬁgure. NB4 cells were treated with 2 mM ATO or Dar for 3 h, ChIP for BACH1 (b), NRF2 (c) and BRG1 (d) at the À 3kbHMOX1 ARE, at the À 9kbHMOX1 ARE and at the TXNRD1 ARE, and at a chromosome 12 distal site followed by qPCR analysis, were then performed. *Po0.05, **Po0.01, ***Po0.001.

BRG1 binding at the HMOX1 AREs subsequent to treatment Dar causes G2/M arrest leading to BRG1 phosphorylation, which with ATO or Dar. As shown in Figure 4d, ATO recruits BRG1 to prevents it from being recruited to the HMOX1 promoter HMOX1 AREs, whereas Dar treatment fails to recruit BRG1. Neither Investigating the events leading to aborted HMOX1 activation in ATO nor Dar recruit BRG1 to the TXNRD1 ARE, which is expected as cells treated with Dar, we found that BRG1 is not recruited to the BRG1 is not required for TXNRD1 transcription, and confirms the HMOX1 promoter upon Dar treatment. However, ATO induces specificity of our findings at the HMOX1 AREs. Enrichment of BRG1 recruitment, allowing HMOX1 expression. Therefore, regula- BACH1, NRF2 and BRG1 was assessed at a distal site of the tion of BRG1 differs between ATO and Dar. Another difference chromosome 12 throughout these procedures, as a negative between ATO and Dar is their effects on the cell cycle. ATO can control. We showed no enrichment, confirming specificity of induce a G1 cell cycle arrest,27 whereas Dar induces a G2/M binding to AREs reported in this figure. Together, our results arrest.1,2 Figure 5a shows that ATO causes NB4 cells to accumulate suggest that NRF2 signaling that is induced by Dar lacks an in the G1 phase of the cell cycle, whereas Dar causes a G2/M essential step, involving BRG1, which is specifically required for accumulation. We therefore postulated that these two distinct the activation of HMOX1. features, in response to two arsenicals, might be related. In fact,

Leukemia (2013) 2220 – 2228 & 2013 Macmillan Publishers Limited Darinaparsin prevents heme oxygenase-1 induction N Garnier et al 2225

Figure 5. Dar causes G2/M arrest, leading to BRG1 phosphorylation and inactivation, which prevents HO-1 induction. NB4 cells were treated with 2 mM ATO or Dar for 18 h, or with 3 mM aphidicolin or 0.3 mM nocodazole for 12 h, with or without 2 mM ATO or Dar for 6 additional hours, then stained with propidium iodide. Cell cycle was analyzed by ﬂow cytometry (a). NB4 cells were treated with 2 mM ATO or Dar for 18 h or with 3 mM aphidicolin or 0.3 mM nocodazole for 12 h, with or without 2 mM ATO or Dar for 6 additional hours. HO-1 protein levels were assessed by western blot (b). NB4 cells were treated for 3 h with 2 mM ATO or Dar, and migration of BRG1 was assessed by western blot. Dar-treated extract was incubated for 1 h with phosphatase lambda at 37 1C (lane 4). Nocodazole (0.3 mM) was used as a positive control for phosphorylation of BRG1 (c). NB4 cells were treated with 2 mM ATO or Dar for 18 h, or with 2 mM Dar for 12 h, and 2 mM ATO for 6 additional hours, and then stained with propidium iodide. Cell cycle was analyzed by ﬂow cytometry (d). NB4 cells were treated with 2 mM ATO or Dar for 3 h, or for 3 h with Dar and 3 additional hours with ATO. ChIP for BRG1 at the À 3kbHMOX1 ARE and at the À 9kbHMOX1 ARE, followed by qPCR analysis, were then performed (e). NB4 cells were treated with 2 mM ATO or Dar for 6 h, or for 3 h with Dar and 3 additional hours with ATO. HMOX1 mRNA levels (f) and protein levels (g) were assessed. ***Po0.001. the literature supports the hypothesis that BRG1 activation can be polyacrylamide gel electrophoresis. The mobility of BRG1 was impaired in cells arrested in G2/M.28,29 If true, synchronizing the slower in Dar-treated cell extracts than in ATO or untreated cell cells in G2/M with nocodazole should hinder HO-1 induction. extracts (Figure 5c). The slower mobility of BRG1 in Dar-treated Conversely, synchronizing the cells in G1 with aphidicoline should extracts is similar to what was observed by other groups.28,29 favor HO-1 induction. Indeed, Figure 5b clearly shows that Lambda-phosphatase treatment shifted the migration of the nocodazole pre-treatment impairs HO-1 induction by ATO, and Dar-treated BRG1 proteins to that of untreated or ATO-treated cell that aphidicoline pre-treatment not only potentiates ATO-induced extracts (Figure 5c), supporting our hypothesis that the HO-1 protein expression, but also enables Dar to induce HO-1 slower migrating BRG1 band is a phosphorylated form of BRG1. protein. This experiment shows that cell cycle status is a key Nocodazole was used as a positive control for G2/M arrest- determinant for HO-1 induction. mediated BRG1 phosphorylation. Together, these results argue We next investigated the link between cell cycle and BRG1 strongly that the mobility shift we observed was due to changes in regulation. BRG1 is regulated via phosphorylation, which can the phosphorylation pattern of BRG1 upon Dar treatment. These occur either during the M phase of the cell cycle28,29 or upon DNA data support the hypothesis that Dar causes BRG1 phosphoryla- damage.30,31 Phosphorylation of BRG1 leads to its inactivation and tion through G2/M cell cycle arrest, and that BRG1 phosphoryla- exclusion from the chromatin.28,29 We therefore hypothesize that tion prevents an essential step for HMOX1 transcription. synchronizing cells in G2/M with nocodazole leads to BRG1 Next, we asked whether Dar could prevent HO-1 activation by hyperphosphorylation,28 rendering HO-1 induction impossible. ATO. Figure 5d shows that cells synchronized in G2/M by Dar Accordingly, Dar, but not ATO, induces BRG1 phosphorylation, as mostly remain in G2/M when subsequently treated with ATO. indicated by western blot analysis after separation by SDS- Moreover, in cells treated with Dar before ATO, BRG1 is neither

& 2013 Macmillan Publishers Limited Leukemia (2013) 2220 – 2228 Darinaparsin prevents heme oxygenase-1 induction N Garnier et al 2226

Figure 6. Dar is more toxic than ATO because it does not induce HO-1. NB4 (a) and U937 (b) cells were treated with 1 mM ATO or Dar, with or without 100 mM (NB4) or 20 mM (U937) OB-24 for 24 h. Cell death was then quantified by propidium iodide staining followed by FACS analysis. NB4 cells expressing non-targeted shRNA or shRNA against HMOX1 (c)orSMARCA4 (e) were treated with 2 mM ATO for 6 h. Protein levels were quantified by western blot. NB4 cells expressing non-targeted shRNA or shRNA against HMOX1 (d)orSMARCA4 (f) were treated with 2 mM ATO or Dar for 24 h. Cell death was then quantified by propidium iodide staining followed by FACS analysis. **Po0.01, ***Po0.001.

recruited to the HMOX1 AREs (Figure 5e), nor do they show transfected with HMOX1 shRNA are more sensitive to ATO than increased HMOX1 mRNA levels (Figure 5f). Consistent with this, NB4 cells transfected with non-target shRNA. Genetically silencing pre-treatment with Dar prevents ATO from inducing an increase in HMOX1 did not alter Dar-induced apoptosis. That was to be HO-1 protein levels (Figure 5g, lane 4, D-A). This suggests that expected as Dar does not lead to HMOX1 expression, and suggests impairment of HMOX1 expression by Dar is an active process. that the enhanced efﬁcacy of Dar is partially because of the absence of HO-1 induction. To test the role of BRG1 in HO-1 induction, we silenced BRG1 using lentiviral shRNA in NB4 cells. HO-1 is partially responsible for limiting the cytotoxicity of ATO Figure 6e shows by western blot that the knockdown of BRG1 The induction of HO-1 by ATO and other ROS inducers has been reduces ATO-induced HO-1 expression. In addition, NB4 cells reported to be cytoprotective in some malignant cell lines.16 We transfected with BRG1 shRNA are more sensitive to ATO-induced wanted to address whether or not inhibiting HO-1 could enhance apoptosis, whereas they remain sensitive to Dar. Taken together, ATO-induced apoptosis. To this end, we utilized a selective these results strongly suggest that BRG1 is required for the inhibitor of the HO-1 enzyme, OB-24.32 We treated NB4 (Figure 6a) cytoprotective HO-1 induction upon ATO treatment. or U937 (Figure 6b) cells with ATO or Dar with or without concomitant treatment with OB-24. We show that OB-24 can enhance signiﬁcantly the toxicity of ATO towards leukemic cells, DISCUSSION supporting a cytoprotective role of HO-1. In contrast, OB-24 We investigated the response of cancer cells to a new arsenic- induces no change in the level of Dar-induced apoptosis, where based, anticancer drug, Dar. We found an unexpected selective HO-1 is not induced. To validate these pharmacological results, we regulation of NRF2 target genes between ATO and Dar. HMOX1 silenced HMOX1 using lentiviral short hairpin RNA (shRNA) in NB4 stands out among the NRF2 targets tested as the sole gene not cells. Figure 6c shows by western blot analysis that the shRNA led induced by Dar. This is unexpected, considering the fact that HO-1 to lower HO-1 induction by ATO. Figure 6d shows that NB4 cells levels are commonly used as a marker for intracellular ROS. It is

Leukemia (2013) 2220 – 2228 & 2013 Macmillan Publishers Limited Darinaparsin prevents heme oxygenase-1 induction N Garnier et al 2227

Figure 7. Differential regulation of the NRF2 response battery by ATO and Dar. Under normal conditions, NRF2 is sequestered in the cytoplasm and its target genes are repressed by BACH1. ATO and Dar induce ROS, which causes the release of NRF2 from sequestration in the cytoplasm and subsequent translocation to the nucleus. NRF2 binds to the AREs of its targets to induce gene expression. HMOX1 activation requires an additional step that consists of the recruitment of BRG1. Dar induces BRG1 phosphorylation that prevents recruitment to the ARE. more surprising that ROS can lead to specific differential recruit NRF2 to the same levels as ATO. One possibility is that responses from the cell, considering that a given ROS does not BRG1 is required to stabilize NRF2 binding at HMOX1 AREs in our bear the mark of its origin. Previous work reported that indeed Dar model. It is also possible that Z-DNA formation, promoted by does not induce HMOX1 transcription in a multiple myeloma BRG1, would enhance NRF2 binding. However, data from Zhang model, where other NRF2 target genes regulated by AREs were et al.23 indicate that NRF2 binding is independent of BRG1 also not induced by Dar.33 In our system, Dar can induce NRF2 recruitment using shRNA knockdown of BRG1 in the SW480 target genes regulated via AREs other than HMOX1. In fact, Dar human colon adenocarcinoma cell line. The role of BRG1 in AML induces more expression of TR1, NQO1 and gGCS, but not enough cells may differ, and/or full knockout of BRG1 may destabilize to confer a cytoprotective response. This is consistent with the fact NRF2 at the HMOX1 promoter. Alternatively, maximal NRF2 that, for example, diminution of intracellular levels of glutathione binding at AREs may be necessary for BRG1 recruitment. (GSH) by inhibition of gGCS with buthionine sulfoximine does not Titrated knockdown of NRF2 in AML cells may define levels of increase Dar-induced apoptosis. Moreover, ATO-resistant cells NRF2 required for BRG1 recruitment and HMOX1 expression. We have higher levels of gGCS, hence of intracellular GSH than their have confirmed that ATO, but not Dar, induces HMOX1 in cell lines sensitive counterparts, but remain sensitive to Dar.2 NQO1 derived from multiple cell types. Importantly, both ATO and Dar induction can be transient. We detected induction of NQO1 by activate several other NRF2 target genes in APL and non-APL cell Dar 3 and 6 h after treatment, whereas previous work was lines. This suggests that multiple cell types respond to Dar- performed at 24 h.33 In addition, Matulis et al., showed that Dar did induced gene expression in a similar manner. However, it is not induce activation of a reporter gene downstream of a possible that alternative mechanisms can also lead to the lack of consensus ARE.33 Our data, as well as the work of others, suggest HMOX1 gene activation. Clearly, pathways other than NRF2 may that the chromatin context, the presence/absence of coregulators, be differentially regulated by ATO and Dar and lead to the the chromosomal localization and the genes regulated by the ARE antiapoptotic feedback induced by ATO. Activation of p38 studied are crucial determinants of the34—specificity of BACH1/ MAPK, ERK and the Akt-mTOR pathways have been proposed to NRF2 binding to a given ARE and1—specificity of the response counteract ATO-induced cell death.36 The impact of Dar treatment toward a given stress inducer.26 on these other negative feedback pathways induced by ATO Here, we propose a mechanistic model for the differential might bring new insights to the mechanisms underlying Dar’s induction of HO-1 by arsenicals. We hypothesize that the lack of enhanced efficacy. HO-1 activation is partially responsible for Dar-enhanced anti- HO-1 is widely considered as cytoprotective through handling tumor properties. As described in Figure 7, both ATO and Dar can of free heme. Intracellular accumulation of free heme occurs initiate the canonical NRF2-signaling cascade, causing BACH1 to when, in the presence of ROS, hemoproteins release it. The be released from the AREs and NRF2 to be recruited to these AREs. combination of ROS and free heme is cytotoxic, because the latter However, Dar also induces hyperphosphorylation of BRG1, promotes further ROS production through the Fenton reaction. preventing BRG1 from participating in the induction of HO-1. HO-1 not only inactivates free heme, but by doing so, also This correlates with the ability of Dar to induce G2/M cell cycle promotes the synthesis of antioxidant agents, including billirubin, arrest, during which BRG1 is phosphorylated and excluded from iron (Fe) stored in L-ferritin (Fe/FtL) and carbon monoxide.16,37–39 the chromatin.29 Accordingly, G2/M arrest related DNA-repair HO-1 is extensively studied for its cytoprotective effects,16 and we machinery, represented by DNApK and ATM, is activated by Dar, present evidence that HO-1 is cytoprotective in malignant cell as shown in Supplementary Figure 7. To date, no evidence lines. In contrast to our observations, it should be noted that shows that DNA damage induced by ATO or Dar differs. However, under some conditions, HO-1 has been reported to be cytotoxic.16 other reports suggest that rapid activation of ATM followed Iron released from free heme by the action of HO-1 can, if not by early induction of apoptosis can be associated with stored in Fe/FtL, lead to cytotoxicity through the Fenton reaction. chromatin decondensation.35 Future efforts will be directed HO-1 has been proposed to promote cancer progression and towards investigating these potential mechanistic differences. angiogenesis.32 This has led to the development of a selective Among all NRF2 target genes, HMOX1 regulation is unique and inhibitor of the HO-1 enzyme, the imidazole derivative OB-24.32 includes more levels of complexity. Although NRF2 binding to Alaoui-Jamali et al.32showed that OB-24 specifically inhibits the AREs is required, BRG1 is also needed.23 Dar treatment fails to activity of HO-1 in a human advanced prostate cancer cell line

& 2013 Macmillan Publishers Limited Leukemia (2013) 2220 – 2228 Darinaparsin prevents heme oxygenase-1 induction N Garnier et al 2228 in vitro and in intact mice. In our model, OB-24 enhances ATO- 15 Copple IM, Goldring CE, Kitteringham NR, Park BK. The Nrf2-Keap1 defence induced cytotoxicity. This is consistent with our hypothesis that pathway: role in protection against drug-induced toxicity. Toxicology 2008; 246: the lack of HO-1 expression subsequent to Dar treatment has a 24–33. role in Dar’s greater antitumor potential compared with ATO. 16 Gozzelino R, Jeney V, Soares MP. Mechanisms of cell protection by heme oxygenase-1. Annu Rev Pharmacol Toxicol 2010; 50: 323–354. 17 Moi P, Chan K, Asunis I, Cao A, Kan YW. Isolation of NF-E2-related factor 2 CONFLICT OF INTEREST (Nrf2), a NF-E2-like basic leucine zipper transcriptional activator that binds to the tandem NF-E2/AP1 repeat of the beta-globin locus control region. Proc Natl Acad Ziopharm Oncology Inc. has provided a research grant to WHM. HMS is on the Sci USA 1994; 91: 9926–9930. Advisory Board of Osta Biotechnologies. HMS and AG hold a joint patent. HMS has 18 Nioi P, McMahon M, Itoh K, Yamamoto M, Hayes JD. Identification of a novel Nrf2- served as a consultant in Osta Biotechnologies, Molecular Biometrics Inc., TEVA regulated antioxidant response element (ARE) in the mouse NAD(P)H:quinone Neurosciences and Caprion Pharmaceuticals. He holds stock options in Osta and oxidoreductase 1 gene: reassessment of the ARE consensus sequence. Biochem J equity in Molecular Biometrics Inc. The remaining authors declare no conflict of 2003; 374(Pt 2): 337–348. interest. 19 Itoh K, Wakabayashi N, Katoh Y, Ishii T, Igarashi K, Engel JD et al. Keap1 represses nuclear activation of antioxidant responsive elements by Nrf2 through binding to the amino-terminal Neh2 domain. Genes Dev 1999; 13: 76–86. ACKNOWLEDGEMENTS 20 Furukawa M, Xiong Y. BTB protein Keap1 targets antioxidant transcription We thank Ziopharm Oncology Inc. for providing Darinaparsin. We thank Osta factor Nrf2 for ubiquitination by the Cullin 3-Roc1 ligase. Mol Cell Biol 2005; 25: Biotechnology Inc. for providing OB-24. This work was supported by grants from the 162–171. Canadian Institutes of Health Research (CIHR) and the Samuel Waxman Cancer 21 McMahon M, Itoh K, Yamamoto M, Hayes JD. Keap1-dependent proteasomal Research Foundation (to WHM). WHM is a Chercheur National of Fonds de la degradation of transcription factor Nrf2 contributes to the negative regulation of Recherche en Sante´ du Quebec (FRSQ). KKM is a Junior 2 Chercheur Boursier of FRSQ. antioxidant response element-driven gene expression. J Biol Chem 2003; 278: ZD was supported by a Canadian Graduate Scholarship from the CIHR. The Quebec 21592–21600. Leukemia cell bank (www.bclq.org) is supported by a grant from the Cancer Research 22 Reichard JF, Sartor MA, Puga A. BACH1 is a specific repressor of HMOX1 that is Network of FRSQ and the Cole Foundation. inactivated by arsenite. J Biol Chem 2008; 283: 22363–22370. 23 Zhang J, Ohta T, Maruyama A, Hosoya T, Nishikawa K, Maher JM et al. BRG1 interacts with Nrf2 to selectively mediate HO-1 induction in response to oxidative REFERENCES stress. Mol Cell Biol 2006; 26: 7942–7952. 24 Itoh K, Chiba T, Takahashi S, Ishii T, Igarashi K, Katoh Y et al. An Nrf2/small Maf 1 Mann KK, Wallner B, Lossos IS, Miller Jr. WH. Darinaparsin: a novel organic heterodimer mediates the induction of phase II detoxifying enzyme genes arsenical with promising anticancer activity. Exp Opin Investig Drugs 2009; 18: through antioxidant response elements. Biochem Biophys Res Commun 1997; 236: 1727–1734. 313–322. 2 Diaz Z, Mann KK, Marcoux S, Kourelis M, Colombo M, Komarnitsky PB et al. A novel 25 Blank V, Kim MJ, Andrews NC. Human MafG is a functional partner for p45 NF-E2 arsenical has antitumor activity toward As2O3-resistant and MRP1/ABCC1-over- in activating globin gene expression. Blood 1997; 89: 3925–3935. expressing cell lines. Leukemia 2008; 22: 1853–1863. 26 Reichard JF, Motz GT, Puga A. Heme oxygenase-1 induction by NRF2 requires 3 Akao Y, Mizoguchi H, Kojima S, Naoe T, Ohishi N, Yagi K. Arsenic induces apop- inactivation of the transcriptional repressor BACH1. Nucleic Acids Res 2007; 35: tosis in B-cell leukaemic cell lines in vitro: activation of caspases and down- 7074–7086. regulation of Bcl-2 protein. Br J Haematol 1998; 102: 1055–1060. 27 Walker AM, Stevens JJ, Tchounwou PB. Arsenic trioxide modulates dna 4 Uslu R, Sanli UA, Sezgin C, Karabulut B, Terzioglu E, Omay SB et al. Arsenic trioxide- synthesis and apoptosis in lung carcinoma cells. Met Ions Biol Med 2010; 7: mediated cytotoxicity and apoptosis in prostate and ovarian carcinoma cell lines. 1996–2007. Clin Cancer Res 2000; 6: 4957–4964. 28 Sif S, Stukenberg PT, Kirschner MW, Kingston RE. Mitotic inactivation of a human 5 Hyun Park W, Hee Cho Y, Won Jung C, Oh Park J, Kim K, Hyuck ImY et al. Arsenic SWI/SNF chromatin remodeling complex. Genes Dev 1998; 12: 2842–2851. trioxide inhibits the growth of A498 renal cell carcinoma cells via cell cycle arrest 29 Muchardt C, Reyes JC, Bourachot B, Leguoy E, Yaniv M. The hbrm and BRG-1 or apoptosis. Biochem Biophys Res Commun 2003; 300: 230–235. proteins, components of the human SNF/SWI complex, are phosphorylated and 6 Camacho LHHD, Gutierrez C, Vertovsek S, Tannir N, Parker CA, Purdom MA et al. excluded from the condensed chromosomes during mitosis. EMBO J. 1996; 15: Phase-1 trial of ZIO-101, a novel organic arsenic in patients with advanced cancers 3394–3402. (abstract). J Clin Oncol 2006; 24: p 605S. 30 Dephoure N, Zhou C, Villen J, Beausoleil SA, Bakalarski CE, Elledge SJ et al. A 7 Dai J, Weinberg RS, Waxman S, Jing Y. Malignant cells can be sensitized to quantitative atlas of mitotic phosphorylation. Proc Natl Acad Sci USA 2008; 105: undergo growth inhibition and apoptosis by arsenic trioxide through modulation 10762–10767. of the glutathione redox system. Blood 1999; 93: 268–277. 31 Matsuoka S, Ballif BA, Smogorzewska A, McDonald 3rd ER, Hurov KE, Luo J et al. 8 Jing Y, Dai J, Chalmers-Redman RM, Tatton WG, Waxman S. Arsenic trioxide ATM and ATR substrate analysis reveals extensive protein networks responsive to DNA damage. Science 2007; 316: 1160–1166. selectively induces acute promyelocytic leukemia cell apoptosis via a hydrogen 32 Alaoui-Jamali MA, Bismar TA, Gupta A, Szarek WA, Su J, Song W et al. A novel peroxide-dependent pathway. Blood 1999; 94: 2102–2111. experimental heme oxygenase-1-targeted therapy for hormone-refractory pros- 9 Lee TC, Wei ML, Chang WJ, Ho IC, Lo JF, Jan KY et al. Elevation of glutathione tate cancer. Cancer Res 2009; 69: 8017–8024. levels and glutathione S-transferase activity in arsenic-resistant Chinese hamster 33 Matulis SM, Morales AA, Yehiayan L, Croutch C, Gutman D, Cai Y et al. Darinaparsin ovary cells. In Vitro Cell Dev Biol 1989; 25: 442–448. induces a unique cellular response and is active in an arsenic trioxide-resistant 10 Ochi T, Kaise T, Oya-Ohta Y. Glutathione plays different roles in the induction of myeloma cell line. Mol Cancer Ther 2009; 8: 1197–1206. the cytotoxic effects of inorganic and organic arsenic compounds in cultured 34 Tsimberidou AM, Camacho LH, Verstovsek S, Ng C, Hong DS, Uehara CK et al. BALB/c 3T3 cells. Experientia 1994; 50: 115–120. A phase I clinical trial of darinaparsin in patients with refractory solid tumors. 11 Davison K, Cote S, Mader S, Miller WH. Glutathione depletion overcomes Clin Cancer Res 2009; 15: 4769–4776. resistance to arsenic trioxide in arsenic-resistant cell lines. Leukemia 2003; 17: 35 Schou KB, Schneider L, Christensen ST, Hoffmann EK. Early-stage apoptosis is 931–940. associated with DNA-damage-independent ATM phosphorylation and chromatin 12 Ninomiya M, Kajiguchi T, Yamamoto K, Kinoshita T, Emi N, Naoe T. Increased decondensation in NIH3T3 fibroblasts. Cell Biol Int 2008; 32: 107–113. oxidative DNA products in patients with acute promyelocytic leukemia during 36 Platanias LC. Biological responses to arsenic compounds. J Biol Chem 2009; 284: arsenic therapy. Haematologica 2006; 91: 1571–1572. 18583–18587. 13 Kajiguchi T, Yamamoto K, Iida S, Ueda R, Emi N, Naoe T. Sustained activation of 37 Ryter SW, Alam J, Choi AM. Heme oxygenase-1/carbon monoxide: from basic c-jun-N-terminal kinase plays a critical role in arsenic trioxide-induced cell science to therapeutic applications. Physiol Rev 2006; 86: 583–650. apoptosis in multiple myeloma cell lines. Cancer Sci 2006; 97: 540–545. 38 Baranano DE, Rao M, Ferris CD, Snyder SH. Biliverdin reductase: a major physio- 14 Davison K, Mann KK, Waxman S, Miller Jr. WH. JNK activation is a mediator of logic cytoprotectant. Proc Natl Acad Sci USA 2002; 99: 16093–16098. arsenic trioxide-induced apoptosis in acute promyelocytic leukemia cells. Blood 39 D’Autreaux B, Toledano MB. ROS as signalling molecules: mechanisms that gen- 2004; 103: 3496–3502. erate specificity in ROS homeostasis. Nat Rev Mol Cell Biol 2007; 8: 813–824.

Supplementary Information accompanies this paper on the Leukemia website (http://www.nature.com/leu)

Leukemia (2013) 2220 – 2228 & 2013 Macmillan Publishers Limited