Modulation of Oxidative Stress by Twist Oncoproteins

Expression of developmental genes Twist1 and Twist2 is reactivated in many human tumors. Among their oncogenic activities, induction of epithelial to mesenchymal transition is believed to increase cell motility and invasiveness and may be related to acquisition of cancer stem cell phenotype. In addition, Twist proteins promote malignant conversion by overriding two oncogene-induced failsafe programs: senescence and apoptosis. Reactive oxygen species (ROS) are also important mediators of apoptosis, senescence and motility and are tightly linked to disease, notably to cancer. We report here that Twist factors and ROS are functionally linked. In wild type cells both Twist1 and Twist2 exhibit antioxidant properties. We show that Twist-driven modulation of oncogene-induced apoptosis is linked to its effects on oxidative stress. Finally, we identify several targets that mediate Twist antioxidant activity. These findings unveil a new function of Twist factors that could be important in explaining their pleiotropic role during carcinogenesis.


Introduction
The two Twist genes (Twist1 and Twist2), related in sequence and function, belong to the super-family of basic Helix-Loop-Helix (bHLH) transcription factors. They are essential for embryonic development as they are involved in mesoderm patterning and the differentiation of multiple cell lineages, including muscle, cartilage and osteogenic cells [1]. Their embryonic functions appear to be related to their ability to induce cell migration by promoting the epithelial to mesenchymal transition (EMT) [2], to protect cells from apoptosis [3] and to control inflammation [4]. Twist expression was reported in several types of precursors and stem cells [5,6], where it was suggested to control the maintenance of their pluripotency. In contrast, Twist proteins are not detectable in most differentiated tissues [7], but their expression is reactivated in many human cancers, including carcinomas (breast, liver, ovarian, prostate), sarcomas, melanomas, gliomas and neuroblastomas [8]. Their oncogenic potential is thought to arise from a combination of multiple properties. First, by promoting the epithelial to mesenchymal transition (EMT), Twist proteins favour invasiveness and may confer to cells self-renewal properties [9,10,11,12]. Second, by disrupting both Rb-and p53-dependent pathways, they override the two major oncogene-induced cellular failsafe programs, which are senescence and apoptosis [13,14,15], thereby promoting malignant conversion. Third, several reports describe a role for Twist1 in the acquisition of resistance to therapeutic drugs [16,17], linking Twist to tumor recurrence.
Reactive oxygen species (ROS), which include superoxide anions, H 2 O 2 , organic peroxides, and hydroxyl radicals, are byproducts of oxidative phosphorylation and as such are constantly generated in all aerobic cells. ROS are eliminated by antioxidant enzymes: scavenger enzymes, superoxide dismutases (SOD), convert superoxide anions to H 2 O 2 , which is subsequently detoxified by catalase, glutathione peroxidase or thioredoxin peroxidase systems [18]. ROS, although essential for processes such as immune response, angiogenesis and stem cell renewal [19], can also be harmful for cells. An uncontrolled production of these molecules provokes damage, including DNA lesion, protein oxidation, and lipid peroxidation [19]. Cellular response to major unrepaired damage includes apoptosis or senescence. Because Twist and ROS overproduction both impact on activities related to the regulation of several common cellular processes [13,15], we explored the possibility that Twist is involved in the control of ROS and, more generally, in cellular response to oxidative stress.
We report here that Twist lowers the level of intracellular ROS, displaying an antioxidant activity in several cell types. Twist-driven inhibition of oxidative stress participates in the protection of cells against c-Myc induced apoptosis. Analysis of molecular mechanisms involved in this new function reveals the existence of an oxidative genetic program mediated by several new targets. Our findings unveil a novel property of Twist proteins that could have important implications for their role during tumor progression.

Twist proteins display antioxidant properties
To investigate the effect of Twist proteins on ROS production, we first generated foreskin-derived human diploid fibroblasts (HDF) ectopically expressing Twist1 or Twist2. The cells displayed no noticeable morphological changes and grew at a similar rate to both parental and control cells (data not shown). As assessed by labeling cells with DiHydroEthidium (DHE) ( Figure 1A-C) or with DiChlorodihydroFluorescein DiAcetate (DCFDA) ( Figure S1A-B), two ROS fluorescent indicators, ectopic expression of either Twist1 or Twist2 was associated with a significant reduction of intracellular ROS. Expression of Twist was as efficient in inhibiting ROS as was the treatment of cells with 10 mM Nacetyl-cysteine (NAC), a widely used antioxidant compound ( Figure S1C-D). The antioxidant property of both Twist proteins was next confirmed using additional cellular models, including primary Mouse Embryonic Fibroblasts (MEFs) ( Figure 1D-F), immortalized Rat Embryonic Fibroblasts (REF52) ( Figure 1G-I) and immortalized human mammary epithelial cells (HMEC-hTert) ( Figure S1E). These results indicate that the antioxidant activity of Twist is neither cell type-nor species-dependent.
We reasoned that since ectopic expression of Twist diminished ROS levels, inhibition of endogenous Twist expression should give rise to the opposite effect. Since HDFs mainly express Twist2 (data not shown), we explored the consequences of its depletion in these cells. This was achieved through the use of either siRNA or shRNA that recognized unrelated sequences of Twist2 mRNA. Both gave rise to inhibition of Twist2 expression and, in accord with our hypothesis, both led to a significant increase of endogenous ROS ( Figure 2). Of note, this increase was reverted by NAC ( Figure S2). Overall, our results strongly suggest that Twist proteins exhibit antioxidant properties in several cellular models.

Twist, ROS and c-Myc-induced apoptosis
As ROS drive induction of the same cellular failsafe programs [20,21] that are overridden by Twist proteins, we assessed whether the newly identified antioxidant property of Twist had an impact on its oncogenic potential. In light of the role of ROS in c-Myc induced cell death and transformation [21,22,23], we focused our attention on Twist's anti-apoptotic activity, which was originally discovered through a genetic screen aimed at identifying genes that counteracted c-Myc induced apoptosis [14]. Interestingly, several subsequent reports showed that generation of ROS accompanied c-Myc overexpression and was linked to the cell death phenotype, as well as to DNA damage, genetic instability and cellular transformation [21,22,23].
To investigate whether the anti-apoptotic function of Twist is functionally related to its antioxidant properties, we ectopically expressed a tamoxifen inducible c-Myc-ER fusion construct in rat embryonic fibroblasts (REF52), the cell type that was used to unveil the apoptotic activity of c-Myc [24]. We first confirmed that c-Myc activation sensitized REF52 cells to apoptosis upon growth factor withdrawal ( Figure 3A). Interestingly, these conditions are associated with a significant increase in endogenous ROS production ( Figure 3B-C) provoked mainly by the removal of serum ( Figure S3A). To evaluate whether ROS production is instrumental in c-Myc induced cell death in this model, cells were cultured in the presence of Tiron, an efficient antioxidant compound [25]. Of note, unlike NAC, Tiron did not display any cytotoxic effects in REF52. As shown in Figure 3D-F, reduction of ROS by this antioxidant compound was associated with improved cell survival, confirming the reported role played by ROS in c-Myc-induced apoptosis [26]. As expected, we found that ROS induction was necessary but not sufficient to induce a strong apoptotic response in this model ( Figure S3B).
As previously described, ectopic expression of Twist proteins protected cells from c-Myc-induced apoptosis [14] upon growth factor withdrawal ( Figure 3G). Strikingly, the increased survival was associated with a significant reduction in ROS ( Figure 3H-I, Figure S3A), suggesting that the antioxidant property of Twist is indeed related to its oncogenic potential. Comparable results were obtained in experiments performed using mouse embryonic fibroblasts (MEFs) (Figure 4).
We conclude that Twist-mediated inhibition of apoptosis is accompanied by the repression of either the production or the accumulation of reactive oxygen species.

Molecular mechanisms regulating the antioxidant activity of Twist
We next investigated the molecular mechanisms involved in the antioxidant activity of Twist. Because Twist proteins are transcription factors, we first tested whether they modulate the expression of the main cellular detoxifying enzymes. We quantified mRNA levels of Catalase, MnSOD, Cu/ZnSOD and Gpx1 by RT-qPCR in MEF cells transduced with Twist1, Twist2 or a control empty retroviral vector. As shown in Figure S4, none of the genes tested showed a significant differential expression in the presence of Twist. Similar results were obtained in HDF and REF52 (data not shown).
In search of Twist target genes involved in this new activity, we next performed a microarray analysis comparing MEF cells expressing Twist1 with MEF transduced with an empty vector. 188 genes were found to be up-regulated and 246 repressed by more than 50% following Twist1 expression. The results confirmed that none of the main antioxidant enzymes were deregulated (data not shown). However, the analysis identified several genes, previously described as being involved in the control of oxidative stress, whose expression is regulated by Twist1 ( Figure  5A). More precisely, we found four up-regulated genes: Cdo1, an enzyme required for the synthesis of taurine, a well known antioxidant compound [27], Mgst3, a microsomal gluthation S transferase carrying a peroxidase activity [28], ApoD, a protein involved in the inhibition of lipid peroxidation [29] and Dhcr24, described recently as a ROS scavenger [30]. Two genes were down-regulated: Aox1, an enzyme that produces ROS by oxidizing aldehydes [31] and Osgin1, a mitochondrial oxidative stress induced gene that was recently shown to increase intracellular ROS levels when overexpressed [32].
Because both Twist factors display an antioxidant activity, we next investigated whether the genes we identified were also modulated by Twist2. We generated MEF-empty vector, Twist1 and 2 lines and mRNA level of each target was quantified by RT-qPCR. As shown in Figure 5B, similarly to Twist1, 2 regulates the expression of the oxidative stress factors revealed by the microarray analysis. This result suggests that both Twist proteins modulate intracellular ROS level in a similar manner by controlling a common set of factors.
Interestingly, none of the genes we identified was highly deregulated, suggesting that none by itself would account for Twist's antioxidant properties. We reasoned that an alternative mechanism of Twist's action might be a more global modulation of a genetic program of oxidative stress regulators. A similar situation has been reported for Nrf2, a master gene of intracellular  [33]. Upon induction of oxidative stress, this transcription factor is activated and triggers expression of numerous genes encoding enzymes involved in ROS detoxification [33]. It is the joint action of these proteins that efficiently removes excess of ROS from the cell. If such was indeed the mode of action of Twist, one would expect that modulation of individual target gene expression would only partially mimic Twist antioxidant activity. To test this hypothesis we overexpressed Mgst3, which is one of the genes up-regulated by Twist. In parallel, we silenced Osgin1, for which Twist is a repressor. Twist1 overexpressing MEF cells were used as a positive control. Ectopic expression and silencing were verified by western-blot and RT-qPCR, respectively ( Figure 6A-C). Intracellular ROS levels were decreased similarly by up-regulation of Mgst3 or down-regulation of Osgin1 ( Figure 6D). As expected, their antioxidant activity was lower than that displayed by Twist1 ( Figure 6E), suggesting that each gene participates but does not fully substitute for Twist in decreasing ROS levels. Next, we tested the impact of the modulation of the same genes on c-Myc-induced apoptosis. As shown in Figure 6F, perturbation of the expression of Mgst3 and Osgin1 partially protects cells from cell death. Once again, this effect was not as marked as that produced by Twist1 ( Figure 6G). We conclude that both Twist target genes are indeed involved in, but are not sufficient for, the Twist anti-apoptotic activity. Thus, our data indicate that Twist antioxidant and anti-apoptotic activities are mediated by a joint action of several of its downstream effectors.
In addition to their anti-apoptotic activity, Twist proteins have been very well described in terms of inducing EMT and enhancing cell motility [7]. We thus asked if the antioxidant and pro-   Figure S5A and B). Whereas Twist expression accelerated wound closure, as expected, Mgst3 had no effect on cell migration in this assay. It would thus appear that lowering the intracellular ROS level has no significant influence on cell migration in this experimental model, suggesting that these two activities of Twist are uncoupled.
presence of 10% FBS and absence of OHT (white bars) or in 0.1% serum with OHT (dark bars) in the presence or absence of 2 mM Tiron, as indicated. Apoptosis was quantified as in A. (E) Cells were cultured as in D and ROS levels were monitored as in B. (F) Quantification of E (*: n = 6, compared to black bar. # : n = 6, compared to green bar). (G) Quantification of apoptosis in REF52-MycER cells expressing Twist1 or Twist2. Black bars: cells are grown under pro-apoptotic conditions (active MycER, 0.1% FBS), white bars: control cultures. Cell death was quantified by Cy3-Annexin-V staining as in A. (H) Cells expressing Twist1 or Twist2 were treated as in G and ROS levels were determined as in D. (I), quantification of ROS levels from H (*: n = 6, compared to black bar. # : n = 6, compared to blue bar). doi:10.1371/journal.pone.0072490.g003

Discussion
Twist transcription factors regulate several distinct cellular functions, [7]. Our study uncovered a novel activity carried by both Twist1 and Twist2, namely modulation of intracellular levels of reactive oxygen species (ROS). This finding might have important implications for understanding consequences of Twist activation in physiological and in physiopathological settings.
We concentrated our analysis on the impact of ROS on Twist's control of apoptosis. Twist displays anti-apoptotic activities in many cellular contexts and in response to multiple stimuli [7] and whether they are all governed by the regulation of ROS remains to be tested. For example, several reports have suggested that Twist expression in cancer cells is correlated with therapeutic drug resistance [16,17] and it was shown that ROS are induced by therapeutic drugs and might participate in their cytotoxic action  [34]. It is thus possible that the effect of Twist on drug resistance results, at least in part, from its antioxidant activity.
We demonstrated that the inhibition by Twist of c-Myc oncogene -induced apoptosis is correlated with its ability to lower ROS levels. This observation is especially interesting in the context of reports showing that mice invalidated for Twist1 display an increased apoptotic index during development [3]. Moreover, MEFs invalidated for Twist2 show an increased sensitivity to stimuli such as TNF-alpha [4], a cytokine that induces and requires ROS to exert its apoptotic activity. Thus, the antioxidant effect could participate in Twist regulation of cell death during development.
Modulation of ROS by Twist may also be important in adult tissues. It is well established that stem cells, in particular hematopoietic stem cells (HSC), display low levels of ROS [34]. This is essential for their maintenance and renewal, whereas an increase of ROS is required for differentiation [34]. The pattern of Twist expression is coherent with its role in regulation of this process. Specifically, Twist factors have been shown to control the maintenance of mesenchymal stem cells [5] and Twist2 inhibits differentiation of the myeloid lineage [6]. It will be of great interest to evaluate whether the antioxidant activity of Twist is conserved, and relevant, in these cellular settings.
Twist plays a major role in mediating metastasis [12]. In our experimental model Twist pro-migratory activity does not appear to be linked to its effect on the intracellular ROS levels. However, a recent study showed that Snail, another epithelial to mesenchymal transition inducing factor, triggers EMT and gives rise to a cancer stem cell phenotype by inhibiting ROS [35]. It is thus possible that Twist antioxidant activity, while not involved in the migration of mesenchymal cells, might act upstream during the transition process. These results also suggest that the control of intracellular ROS levels might be a common mechanism employed by different EMT-inducing factors. It will be of interest to test this hypothesis, notably in the context of the welldocumented role of ROS in physiology and pathology and particularly in the process of cancer initiation and progression.

Ethics Statement
Experiments were performed according to national regulations and this study was specifically approved (ID approval Nu D34-172-16) by the regional ethics committee of Languedoc-Roussillon (Comité Régional d'Ethique sur l'Expérimentation Animale-Languedoc-Roussillon), France.

Constructs
N-terminal Myc-tagged Twist1 and Twist2 and HA-tagged Mgst3 murine cDNAs were cloned in pMSCV-puromycin retroviral vectors (Clontech). pBabe-MycER was a kind gift from Dr A. Gandarillas. ShRNA sequences against Twist2 were cloned into the previously described pRetro-SUPER vector [13]. An irrelevant sequence, with no significant hits when blasted against the NCBI nucleotide sequence NR database (5'-GTACGTAC-GATCACAGTCA-3') was used as an shRNA control.

Cell culture
Human diploid fibroblasts (HDF) were a kind gift from Dr J. Piette [36]. Mouse embryonic fibroblasts (MEF), derived from C57BL/6 mice, were prepared from 13.5 DPC embryos. All cell lines including 293T cells [37] were cultured in Dulbecco's modified Eagle medium containing 10% fetal bovine serum, 100 ng/mL streptomycin, and 100 U/mL penicillin at 37 uC with 5% CO2. HMEC were grown as recommended by the American Type Culture Collection (ATCC).

Retroviral infection
1.5610 6 293T cells were transfected twice at 24 hours intervals, with 3 mg of retroviral vector using jetPEI TM (Polyplus). Fortyeight hours after the last transfection, supernatants were collected, filtered (0.45 mm, Millipore), supplemented with 8 mg/ml of polybrene and added to cells to be transduced for 16 hr. M.O.I was always # 1 as we noted that high viral titers induced a significant amount of non-specific oxidative stress. Infected cells were selected after 48 hr. with the appropriate antibiotics.

Immunoblot Analysis
Cells were lysed in boiling Laemmli buffer supplemented with protease inhibitors, then sonicated and complemented with DTT. Protein concentration was determined by BCA (Thermo Scientific) assay. Fifteen to twenty micrograms of total protein was loaded onto SDS-PAGE gels and transferred onto nitrocellulose membranes. The membrane was blocked with TBST (16 TBS with 0.1% Tween 20) + 5% milk at room temperature for 1 hr. and incubated with primary antibody and then with horseradish peroxidase (HRP)-coupled secondary antibody (Amersham). Activity was visualized by electrochemiluminescence. Antibodies used in this study are anti-Myc (9E10), anti-Actin (Sigma Aldrich) and anti-HA High Affinity (Roche).

Reverse Transcription and Real-Time PCR
Total mRNA was isolated using a RNeasy mini kit (Qiagen). Reverse transcription was performed with random hexamers and . Empty vector was used as a control. Expression of the transgenes was verified by western-blotting (1: Empty vector; 2: Transgene). (C) MEF cells were transfected with a siRNA control or a siRNA directed against Osgin1 and the silencing efficiency was measured by RT-qPCR normalized to HPRT. (D) Quantitative analysis of ROS levels of cell lines from A, B and C (*: n = 6, compared to control). (E) MEF-MycER cells were infected with either control (pMSCV), Twist1, or Mgst3 or were transfected with corresponding siRNA as in C (n = 3). Apoptosis was induced and determined as in Figure 4. (F) Alternative presentation of data from D and E, allowing better visualization of partial effects on ROS (*: n = 6, compared to red bar) and apoptosis (*: n = 3, compared to red bar) exercised by Mgst3 and Osgin-1. These graphs show the percentage of inhibition compared to control (pMSCV of si-CT). doi:10.1371/journal.pone.0072490.g006 M-MLV Reverse Transcriptase (Invitrogen). Real-time PCR was performed in triplicates with LC FastStart DNA Master SYBR Green I on a LightCycler rapid thermal cycler system (Roche Diagnostics), according to the manufacturer's instructions. Housekeeping genes GAPDH or HPRT were used for normalization. Primers sequences are listed in supplementary table (Table S1).

ROS detection
Cells were incubated with 30 mM DHE (Invitrogen D11347) for 25 min or with 5 mM of CM-H 2 DCFDA (Invitrogen C6827) for 30 min, at 37uC in PBS. The cells were collected and fluorescence was quantified on a FACSCalibur flow cytometer using CellQuestPro software (BD Biosciences).

Apoptosis analysis.
Cells were plated at low density (10 5 cells/well) in 6 well plates 24 hours before induction of apoptosis. c-Myc-

Statistical analysis
Experiments were repeated at least three times. Data are presented as mean 6 SEM. An Independent Student's t test was performed to analyze the assay results; a two-tailed Student's t test was used to compare the intergroup differences. Significance was accepted for values where P#0.05 (* /# ), P#0.01 (** /## ), P#0.0005 (*** /### ).  Text S1

Supporting Information
(DOC)