ZEB1 directly inhibits GPX4 transcription contributing to ROS accumulation in breast cancer cells

Prior studies have noted that zinc finger E-box binding homeobox 1 (ZEB1) is a master transcription regulator, affecting the expression of nearly 2000 genes in breast cancer cells, especially in the epithelial–mesenchymal transition (EMT) process. We now tested the role of ZEB1 on the oxidative stress of cancer cells and explored its possible mechanisms. Two human breast cancer cell lines MDA-MB-231 and MCF7 were selected for the ROS test, PCR, immunofluorescence, Western blot, chromatin immunoprecipitation assay, luciferase assay, and enzyme assay. Mouse models experiments and bioinformatics analysis were conducted to test the indicated molecules. We observed ZEB1 could inhibit GPX4 transcription by binding to the E-box motifs and promote breast cancer progression by accumulating intracellular ROS. From the perspective of ROS clearance, Vitamin E enhanced GPX4 function to consume L-glutathione and eliminated excess intracellular ROS. ZEB1 could not only regulate EMT, but also inhibit GPX4 transcription by binding to the E-box motif. It was important to note that the ZEB1/GPX4 axis had a therapeutic effect on breast cancer metabolism.


Introduction
Zinc finger E-box binding homeobox 1 (ZEB1) is an important member of zinc finger homologous domain transcription factor family. It can bind the E-box and either promote or inhibit molecular transcription and regulate cancer cell differentiation and multidrug resistance [1][2][3]. Most of the previous studies have only concentrated on its role in epithelial-mesenchymal transition (EMT). Little is known about how ZEB1 affects reactive oxygen species (ROS) and oxidative stress metabolism.
Increased ROS is a typical malignancy feature in tumors [4]. The main endogenous sources of ROS are enzymes of the mitochondrial respiratory chain, NADPH oxidase (NOX), xanthine oxidase, and dysfunctional endothelial nitric oxide synthases [5]. ROS can be detrimental when produced in high amounts intracellularly. The cells generally respond to ROS by up-regulating antioxidants, such as superoxide dismutase (SOD) and glutathione peroxidase (GPx), which can initiate innate immune responses to tumors [6]. Vitamins E (Vit E) and C have also been shown to be ROS scavengers [7,8]. Abnormal cellular ROS levels may promote the growth of cancer cells or lead to cell death [9].
Glutathione peroxidase 4 (GPX4) is a member of the ROS elimination system that can reduce the level of hydroperoxides by catalyzing glutathione [10,11]. The GPX4 promoter region contains four E-box (CANNTG) motifs, which have been reported to be regulated by ZEB1 [12].
Thus, the aim of this study was to explore the influence of ZEB1 on cancer metabolism and oxidative stress cancer metabolism and to investigate its possible mechanisms of action.
After ZEB1 was knocked down by shZEB1 (231 sh) and three targeted sequences for siZEB1 (mcf7 #1, mcf7 #2, and mcf7 #3; Fig S2), flow cytometry results revealed a significant total ROS level reduction compared to the control (231 nc, mcf7 nc, and mcf7 siNC for short as mcf7 #nc) in MDA-MB-231 and MCF7 cells due to a decrease in DCFH-DA fluorescence intensity. The reduction in ROS levels was significant after ZEB1 knockdown, which was similar to the effect of ROS scavengers, such as SOD (1000 U/mL), NAC (5 mM), TEMPO (1 mM), and Vit E (10 µg/mL; Figs. 2A and S4, P < 0.05). These results raised the question of whether the total ROS reduction was due to less generation or increased scavenging pathways.
As mitochondrial ROS (mtROS) was the main source of ROS (approximately 90%) [27], ROS production associated with the mitochondrial electron transport chain was simply and systematically detected [5]. However, mitochondrial respiratory chain complex (MRCC) I/III activity, which is the main pathway of mtROS generation [5], remained unchanged when ZEB1 was knocked down (Fig. 2B, P > 0.05). The NADPH system, which is responsible for the remaining ROS sources (~ 10%), was also inspected. NOX1, NOX3, and NOX4 expression levels were determined by qPCR, where Vit E (10 µg/mL) served as the negative control and H 2 O 2 (80 mmol/L) was the positive control. According to these results, only NOX4 levels showed a statistically significant change in all of the ROS generation relative molecules (Fig. 2C, P < 0.05). It is reasonable to suggest that the ROS increase was not merely dependent on ROS production. Therefore, it is recommended that the scavenging system for ROS should be the primary focus in this study.

ZEB1 directly restrains GPX4 promoter transcriptional activity by directly binding to the E-box motif
GPX4 has been reported to be a key player in ZEB1-related tumor progression [28]. To determine whether ZEB1 regulates GPX4 expression, the present study demonstrated that GPX4 expression increased in both shZEB1 (231 sh) and siZEB1 (representative mcf7 #2) subgroups on mRNA levels ( Fig. 3A, P < 0.05). The mRNA expression level of ZEB1 in breast cancer cells had a significant negative correlation with GPX4, so it was necessary to determine whether GPX4 was regulated by ZEB1 transcriptional repression as a downstream target gene.
To address this question, further analysis of the GPX4 promoter sequence revealed four E-box regions, which could be the binding sites of ZEB1 (Fig. 3B). ChIP assay results found that both pairs of primers targeting the GPX4 promoter E-box region could be immunoprecipitated by anti-ZEB1. RT-PCR amplification analysis suggested that the region containing the ZEB1 recognition sequence was in the GPX4 DNA. Statistical analysis showed that anti-ZEB1 levels bound to the GPX4 promoter were higher compared to anti-IgG in 231 nc cells (Fig. 3C, P < 0.05). Luciferase assays further confirmed that ZEB1 directly inhibited GPX4 transcription. The pGL3 plasmids were constructed by loading GPX4 promoter regions (E-box, CAC CTG ) or its mutated (MUT) fragment regions (GPX4-MUT, AAA AAA; Fig. 3D). GV146-NC and GV146-ZEB1 were cotransfected in 231 nc cells for 48 h, then co-transfected with either the GPX4 promoter NC region plasmid with luciferase (luc) or the GPX4 promoter MUT region plasmid with luc for further testing. Significant decreases in luc activity were only monitored in over-expressed ZEB1 plasmid and transfected GPX4 promoter NC region plasmid with luc after 15and 45-min sessions (Fig. 3E, P < 0.05). However, despite the ZEB1 over-expression, GPX4 luciferase was stable in GPX4-MUT promoter-transfected cells, which showed that ZEB1 inhibited GPX4 promoter activity by binding to the E-box motifs (Fig. 3E).

ZEB1 restrains GPX4 expression contributing to ROS accumulation in vitro
Next, we examined the effect of GPX4 suppression by ZEB1 on the basic functions of breast cancer cells, such as ROSdriven oxidation, cell viability, and migration. First, the investigation in MDA-MB-231 cells treated with ShZEB1 ( Fig. 4A) and MCF7 cells treated with siZEB1 ( Fig. 4B) showed that both SOD and GPx activities increased after ZEB1 knockdown (P < 0.05). It was inferred that ZEB1 increased ROS accumulation by preventing the ROS elimination pathway, which enhanced SOD and GPx activities. In particular, GPx activity increased significantly after ZEB1 knockdown (P < 0.01), which strongly supported the finding that ZEB transitionally repressed GPX4.
It is already known that lipid alkoxyl (RO·) radicals (lipid ROS) are thought to contribute to ROS-dependent cell damage and death. It was further investigated how ZEB1 restrains GPX4 expression by testing lipid ROS levels. Lipid oxidation is one of the markers of GPX4 expression. C11-BODIPY, a lipid oxidation detection probe, was used to detect the change in fluorescence intensity after MCF7 promoter and E-box locations in GPX4 promoter region are shown. C, RT-PCR was used to detect gene abundance in the different groups, which were immunoprecipitated using anti-ZEB1 antibody in MDA-MB-231 cells. D, The GPX4 promoter region and its mutated fragment region (GPX4-MUT) were cloned into the pGL3 plasmid. E, MDA-MB-231 cells were first transfected with the GPX4-NC promoter or GPX4-MUT, then with GV146-NC or GV146-ZEB1 overexpression vectors. Relative luciferase activity was measured and normalized to promoter-transfected alone group. Data are represented as means ± S.E. in triplicate experiments, *P < 0.05 cells were treated with siZEB1. Cells (mcf7 #nc, mcf7 #1, mcf7 #2, and mcf7 #3) were incubated with C11-BODIPY and then the lipid peroxidation level was determined by flow cytometry. The mcf7 #nc group showed the highest fluorescence intensity compared with the siZEB1 group ( Fig. 4C, P < 0.05). The scavenging effect of lipid ROS on the siZEB1 groups might also curb cytotoxicity and migration ( Fig. 4D and 4E, P < 0.05). These results suggest that siZEB1 has a strong effect on cancer cell apoptosis and migration resistance, similar to GPX4 up-regulation-mediated removal of ROS.
In addition, it has been concluded that L-glutathione is the most important hydrophilic antioxidant that protects cells against exogenous and endogenous toxins, including ROS and reactive nitrogen species. In the GPx system, GPX4 catalyzes the H 2 O 2 reduction, leading to glutathione (GSH) oxidation into GSSG (Fig. 4H). Therefore, detection of changes in the concentration of L-glutathione and GSSG after siZEB1 or shZEB1 treatment in MCF7 or MDA-MB-231 cells was necessary to indirectly determine the expression level of GPX4. It was observed that the loss of ZEB1 caused a decrease in L-glutathione and an increase in GSSG in MCF7 (Fig. 4F, P < 0.05) and MDA-MB-231 ( Fig. 4I, P < 0.05) cells, while shZEB1 decreased L-glutathione and increased GSSG similar to the ROS scavengers SOD (1000 U/mL), NAC (5 mM), TEMPO (1 mM), and Vit E (10 µg/mL; Fig. 4G, P < 0.05). The cells were treated in the same manner as those in Fig. 2A, suggesting that ZEB1 directly inhibited GPX4 expression, contributing to ROS accumulation.
In addition, along with lipid ROS loss, L-glutathione consumption, and GSSG generation, Fe 2+ content was also decreased after siZEB1 treatment, and which was reduced further after Fer-1 addition (Fig. 4G, P < 0.05), suggesting that the effect of ZEB1 on lipid ROS might be related to ferroptotic events.

ZEB1 restrains GPX4 expression and Vit E treatment facilitates GPX4 function in vivo
Finally, a xenograft derived from cell lines was established using MDA-MB-231 cells to determine whether ZEB1 silencing influenced tumor ROS and oxidative stress metabolism via GPX4 (Fig. 5A). Vit E eliminates ROS and suppresses tumor growth [30]. It was observed that the tumor in the shZEB1 group was smaller and weighed less. This trend was more pronounced in groups that were maintained on diets supplemented with an antioxidant Vit E ( Fig. 5A-C, P < 0.05). Consistent with in vitro results, both SOD and GPx activities were increased in the shZEB1 group mice, especially in the Vit E feeding group (Fig. 5F, P < 0.05). Increased GPX4 expression was shown on the mRNA and protein levels after ZEB1 silencing, which was augmented in the Vit E feeding group (Fig. 5D, P < 0.05 and Fig. 5E).
Lung metastases were subsequently evaluated using India ink staining. The result was similar to that of the tumor growth and volume. The number of lung nodules in the shZEB1 group was lower than that in the 231 nc group. It was significantly lower in the shZEB1 combined with Vit E feeding group than in the shZEB1 alone group (Fig. 6A, P < 0.05), suggesting that the metastatic ability of shZEB1 breast cancer cells was affected in vivo.
Frozen fresh tumor tissues sections were stained with fluorescent probe O13 ROS. It was found that down-regulated expression of ZEB1 decreased ROS intensity, while Vit E diet further reduced ROS in the tumor tissues (Fig. 6B, P < 0.05), which was consistent with the GPX4 protein level in Fig. 5D. Animal experiments further showed that removing transcriptional inhibition of GPX4 via ZEB1 effectively scavenged ROS, which might be one of the key mechanisms associated with tumor growth inhibition.
Indeed, it was demonstrated that ZEB1 can inhibit GPX4 transcription by binding to the E-box motifs and promote breast cancer progression by accumulating intracellular ROS. From the perspective of ROS clearance, Vit E was able to enhance GPX4 function to consume L-glutathione and eliminate excess intracellular ROS (Fig. 6C), which might contribute to the progression of breast cancer. Finally, the role of ZEB1 and GPX4 in the survival of breast cancer patients was analyzed using PROGgenev2 [29]. Similar to the above conclusions, a higher ZEB1 mRNA expression level was associated with shorter relapse-free survival in breast cancer patients, while a higher GPX4 mRNA expression level did the opposite (Fig. 6D, P < 0.05 or P < 0.01). Therefore, the major achievement of this study was to demonstrate that ZEB1 is a promising therapeutic target for the oxidation system, antioxidant system, or lipid metabolism.

Discussion
The present study demonstrated that ZEB1 repressed the generation of scavenging ROS by directly repressing GPX4 transcription and thus promoting tumor progression. High expression of ZEB1 and low expression of GPX4 were the adaptive mechanisms used by malignant breast cancer in order to withstand environmental pressures. Vit E enhanced the expression and function of GPX4 in breast cancer.
A re-evaluation of ZEB1 has been warranted in order to show that it has pleiotropic functions in physiological and pathological conditions, which are not limited to cell invasion and dissemination [31]. The functions and mechanisms of ZEB1 are more complex than previously appreciated [32][33][34]. ROS has been implicated in ionizing radiationinduced EMT via activation of transcription factors, including ZEB1 [35], which is a favorable H 2 O 2 target [36]. Modulation of oxidative stress has been reported as an anticancer strategy [37,38]. In particular, research on the lipid GPX4 has become more widespread [39]. The present study aimed to investigate how ZEB1 regulates GPX4 expression and how it affects breast cancer cell malignancy.
GPX4 is an essential regulatory inhibitor of ferroptotic cancer cell death [40]. Ferroptosis is a mode of cell death involving the production of iron-dependent ROS [28]. However, GPX4 knockout mice are embryologically lethal [41]. Some small molecule inhibitors can prevent cell death in response to GPX4 deletion, which can trigger acute renal failure in mice [42]. Small intestinal epithelial cells in Crohn's disease exhibit impaired GPX4 function and show signs of lipid peroxidation [43]. The present data confirmed that GPX4 is of benefit to disease therapy. SOD and GPx activities were elevated after ZEB1 silencing, while ZEB1 transcriptionally suppressed the expression of GPX4. It is reasonable to expect that Vit E, which is a ROS scavenger, would show no additional effects on ZEB1 silencing. Vit E abolished docosahexaenoic acid effects on GPx inhibition, while it enhanced the activity of several anticancer drugs via an oxidative mechanism [30]. GPx is an important peroxidase that exists throughout the body. Selenium is a component of the GPx enzyme system, which can catalyze the transformation of GSH into GSSG, reduce toxic peroxide to non-toxic hydroxyl compounds, and promote the decomposition of H 2 O 2 in order to protect the structure and function of cell membrane from oxide interference and damage. Findings obtained by these in vitro and in vivo studies strongly suggest that selenoenzyme GPX4 is a critical factor for breast cancer cell survival, which is highly sensitive to Vit E status. This suggests that there is a specific molecular mechanism between Vit E and GPX4 circuit, which remains to be further explored. In the present study, ferroptosis markers, such as GPX4 expression, cell viability, membrane lipid peroxidation, Fe 2+ content, L-glutathione, and GSSG levels, were examined. Some results did not support the idea that ZEB/GPX4/ROS leads to GPX4-associated ferroptosis. For example, FSP1 acting in parallel with GPX4 to inhibit ferroptosis has been highlighted in recent years [44]. This suggests that ROS accumulation and oxidative stress contribute to breast cancer progression in a comprehensive manner. In addition, it was not surprising that ferroptosis was not definitely involved in the effect of ZEB1/ GPX4/ROS in breast cancer.
Given that cancer cells must be able to cope with high oxidative stress to promote metastasis, ROS exist as a carcinogenic factor [4]. When cellular ROS level reaches the "threshold" that can lead to cell death, ROS exert a cytotoxic effect to limit cancer progression [9]. The present investigation was mostly conducted on the basis of ZEB1 silencing, where intracellular ROS levels remained at a high level during its long-term oxidative adaptation, but did not reach the threshold. At the early stages of the experiment, the effect of ZEB1 on ROS production was detected, and siZEB1 did not interfere with MRCC I/III activity. In addition to MRCC, ROS production pathways also include NADPH oxidase and xanthine oxidase [5]. ROS can be detrimental (it is then referred to as "oxidative and nitrosative stress") when produced in high quantities in the intracellular compartments. Cells generally respond to ROS by up-regulating antioxidants, such as SOD, catalase, GPx, and GSH, which protect them by converting dangerous free radicals to harmless molecules (i.e., water). Therefore, the reason why ZEB1 up-regulated ROS levels in breast cancer not only via GPX4 inhibition of the clearance pathway, and whether there are generative pathways other than MRCC I/III or NOX1, NOX3, and NOX4 are involved, remains to be further studied.
In conclusion, ZEB1 is a universal transcription factor that can not only regulates EMT but also inhibits GPX4 transcription by binding to the E-box motif. It is important to note that the ZEB1/GPX4 axis has a therapeutic effect on breast cancer metabolism. . Western blot was used to detect GPX4 protein expression in fresh tissues from each tumor. F and G, SOD and GPx activities in fresh tumor tissue homogenate were detected. Data are represented as means ± S.E., *P < 0.05 and **P < 0.01

RNA sequencing (RNA-Seq)
The samples of MDA-MB-231 cells and shZEB1 MDA-MB-231 cells (n = 3) were analyzed for RNA-Seq by Beijing Genomics institution. The data analysis was based on Bioconductor packages "cluster Profiler" software.

RT-PCR and qPCR
Total RNA was isolated using TRIzol reagent (CWBIO, China). cDNA was obtained using EasyScript First-Strand cDNA Synthesis SuperMix (TransGen, China). 50 ng cDNA was used to perform qPCR with UltraSYBR Mixture (CWBIO, China). Gene expression was normalized to control using 2 −ΔΔCt method. Primer sequences are listed in Table 1.

Total ROS and lipid ROS measurement
Intracellular total ROS level was examined using fluorescent dyes 2,7-Dichlorodi-hydrofluoresce in diacetate (DCFH-DA) [14] as probes. The cells were trypsinized and incubated with DCFH-DA-containing serum-free medium, and the intensity of fluoresce was detected by flow cytometry. Lipid ROS level was evaluated by cells incubating with the lipid peroxidation sensor C11-BODIPY (5 μM) for 30 min, and the fluorescent intensity was also detected by flow cytometry study.

Measurements of MRCC I/III, SOD, GPx, L-Glutathione, GSSG, and Fe 2+
Activities of Mitochondrial respiratory chain complex (MRCC I/III), total SOD (T-SOD, SOD for short), and glutathione peroxidase (GSH-Px, GPx for short) were employed for biochemical detection of cell lysates or the supernatant of tumor tissue homogenate (Beyotime, Shanghai, China). Concentrations of total L-Glutathione, GSSG, and Fe 2+ were measured in cell lysates (Beyotime, Shanghai, China and BioVision, USA).

Cell viability and wound closure assay
Cells were seeded into 96-well plate (1*10 4 cells per well) and incubated for overnight at 37 °C. The cells were incubated with ROS scavengers at different concentrations for 24 h at 37 °C, and then analyzed via the standard CCK-8 assay (ab228554, Abcam). Different processing monolayer cells were scraped using 10-μl pipette tips for 12 h and 24 h. Image analyses were conducted of the wound closure area.

Chromatin immunoprecipitation (ChIP) assay
EZ-Magna ChIP™ A/G Chromatin Immunoprecipitation Kit (MERCK, Germany) was used. Antibodies used for immunoprecipitation were goat anti-rabbit ZEB1, isotype IgG, Fig. 6 ZEB1 knockdown decreases lung metastatic nodules and ROS level in tumor tissue, which is consistent with bioinformatics data on breast cancer patients. A, Metastatic lung nodules were represented and analyzed. B, Frozen sections stained for ROS intensity from fresh tumor tissue (representative sample number 3 among the 231 nc and 231 sh tumor groups and number 1 among the 231 sh with Vit E diet group). MFI for the region of interest (n = 4) was quantified, *P < 0.05. C, Work model. D, Bioinformatics data from PROGge-nev2 database showed relapse-free survival curves in breast cancer patients with high ZEB1 and low GPX4 mRNA expression, which was associated with poor prognosis, *P < 0.05 ◂ 1 3 and RNA polymerase II antibody (MERCK, Germany) as negative control and positive control, respectively. Primers for ChIP assay are listed in Table 2. End-point PCR was conducted to detect targeted DNA.

Luciferase assay
GPX4 promoter or mutant binding sites promoter sequences were cloned into pGL3 luciferase promoter vectors (General Biosystems) using the restriction sites of the enzymes KpnI and XhoI. MDA-MB-231 cells were co-transfected with the wild-type or mutant human GPX4 promoters and ZEB1 expression plasmid or control in 6-well plates. Cells were prepared after 48 h following transfection, and incubated with D-Luciferin, luciferase activity was measured according to the manufacturer's protocols.

Xenograft tumor growth and lung metastasis in nude mice
As approved by the ethics committee for animal use in Nankai University, the mice were randomized into 3 groups (n = 3), and MDA-MB-231 cells (2*10 6 ) transfected with shCtrl(231 nc) or shZEB1(231 sh) were implanted subcutaneously(SQ) into the mammary fat pads of female BALB/c nude mice, respectively. In the second day, 0.5 g/ kg Vit E was supplemented in the food and fed to the only one group until mice sacrificing, and the others were fed without vitamin E diet [13]. Tumors were monitored every 3 days by caliper, and tumor volume was calculated as 0.5 × length × width × width. The 9 tumors were removed and the fresh tissues were extracted for mRNA, protein, homogenate, and frozen sections. The 9 lungs were assayed by India ink perfusion.15% ink was injected into trachea to fill both lungs, then removed and washed them by Fekete's solution (75% alcohol, methanol, glacial acetic acid), and counted manually the pale nodules.

Frozen section immunofluorescence
Frozen sections were made from fresh mice implant tumors (within 60 min after surgical resection) (Thermo Cryotome E), spun dry, and circled around the tissues with a marking pen. Fluorescent probe O13 ROS of dye solution (BBoxiProbe® tissues active oxygen detection kit, BB-470512, Bestbio, Shanghai, China) was added to the tissue section within the circle and then the tissue was incubated with the probe for 30 min at 37 °C out of light. After the sections were dried, DAPI dye was added into the circle and incubated at room temperature in the dark for 10 min. Antifade Mounting Medium (Beyotime Shanghai, China) was used to seal the sections, which were imaged by confocal microscope (FV1000, Olympus) (O13 ROS probe is a red fluorescent ROS probe with a maximum excitation/emission wave length of 510/610 nm).

Statistical analysis
Data analyses were conducted in GraphPad5. All data from at least 3 experiments are averaged by ± SEM. The difference between two groups was analyzed using the Student's t-test. The difference among three or more groups was evaluated by one-way ANOVA, followed by the Dunnett' test.