Hypoxia Enhanced Glioblastoma Resistance to Erastin-Induced Ferroptosis by Up-Regulating GPX4 via PI3K/AKT/HIF-1α Axis


 Background

Glioblastoma is the deadliest type of primary brain tumor with a high rate of recurrence and treatment resistance. Hypoxia contributed much to radiotherapy resistance and chemoresistance of cancer. Ferroptosis is a nonapoptotic, oxidative cell death and identified as a potential anticancer mechanism in recent years. Erastin acts as a ferroptosis activator and shows a potential role in tumor treatment but the relationship between hypoxia and erastin resistance in glioblastoma has not been explained. This study aimed to investigate the role and underlying mechanism of hypoxia in erastin-induced ferroptosis in glioblastoma.
Methods

Cell proliferation and viability were determined by Cell Counting Kit-8 (CCK-8) assay, flow cytometry, TUNEL assays, and clone formation assay. Lipid peroxides level was analyzed by Malondialdehyde (MDA) assay and flow cytometry using C11-BODIPY dye. The correlation between HIF-1α and GPX4 expression was detected in data from the TCGA database and was determined by ChIP-qPCR assay and luciferase reporter assay. Subcutaneous xenograft and orthotopic xenograft models were established to test our findings in vivo.
Results

Hypoxia for at least 16 hours significantly suppressed erastin-induced ferroptosis by up-regulating glutathione peroxidase 4 (GPX4) expression in U87 and U251 cells. Hypoxia promotes GPX4 expression via enhancing the PI3K/AKT/HIF-1α pathway. Mechanistically, HIF-1α directly bound to the GPX4 gene promoter region and promoted GPX4 transcription. AKT inhibitor MK-2206 and HIF-1α inhibitor PX-478 could significantly reverse the effect. Besides, under normoxia, PX-478 could induce a higher lipid peroxidation level by decreasing GPX4 expression in U87 and U251 cells but cannot induced cell death directly, and it could significantly enhance the tumor cell killing effect of erastin. In vivo, combination of PX-478 and erastin had a coordinated intensification effect on anticancer activity uncovered by subcutaneous xenograft and orthotopic xenograft mouse models.
Conclusions

Hypoxia enhanced glioblastoma resistance to erastin-induced ferroptosis by activating PI3K/AKT/HIF-1α pathway and promoting GPX4 expression in a transcriptional regulation way. Combination therapy of PX-478 and erastin may be a potential strategy against glioblastoma.


Page 3/19
Background Glioma is most common primary malignant brain tumor in adults, in which glioblastoma is the most malignant and aggressive form [1]. Glioblastoma is an anaplastic, poorly differentiated malignant tumor with a peak of incidence between 45 and 70 years [2]. Despite recent advances in diverse therapy for glioblastoma incorporating surgery, radiotherapy, chemotherapy, the median survival is only about 15 months [3][4]. Thus, more e cient therapeutic strategies should be developed.
The tumor microenvironment (TME) refers to the local biological environment in which solid tumors are located [5]. The TME often displays at least some degree of hypoxia [6]. Coincidentally, glioblastoma undergoes malignant progression under hypoxia condition [7]. Tumor hypoxia is generated by irregular and tortuous vasculature formed within solid tumors, hence causing a poor delivery of oxygen to cells.
Hypoxia is associated with malignancy and tumor aggressiveness, by increasing tumor cell proliferation, metastasis [8]. Besides, hypoxia has been associated with resistance to radiotherapy and several chemotherapies. For one thing, hypoxia contributes to radioresistance by controlling several cellular processes including cell cycle [9], apoptosis and senescence [10], creation of reactive oxygen species [11], invasion and cancer cell stemness [12]. For another, hypoxia enhanced the resistance to temozolomide by different mechanisms including inducing ferritin light chain [13], regulating glioma stem cells properties [14], mediating ATP-binding cassette proteins [15]. Hypoxia-inducible factor (HIF) is upregulated in the tumor hypoxia microenvironment, in which HIF-1α and HIF-2α are considered the main regulators of the hypoxia response [16]. Under hypoxia, HIF-1α is stabilized and translocated into the nucleus to play roles in transcription [17]. As a selective HIF-1α inhibitor, PX-478 interferes with HIF-1α and induces cell cycle arrest in cancer cells [18][19][20].
Ferroptosis is a form of nonapoptotic, oxidative cell death that plays an essential role in various diseases, including cranial trauma, neuronal diseases, and brain tumors [21]. Lipid peroxidation underlies the mechanism of ferroptosis [22]. GPX4 and ferroptosis suppressor protein 1 constitute two major ferroptosis defense systems through the inhibition of lipid peroxidation [23]. Erastin as the ferroptosis activator can not only cause tumor cells death, but also increase the sensitivity of tumor cells to chemotherapeutic drugs by coordinating with chemotherapy drugs [24]. For example, erastin has been proved to enhance the sensitivity of glioma cells to temozolomide, which showed the potential role of erastin in tumor treatment [25].
To date, the relationship between hypoxia and erastin-induced ferroptosis has not been reported in glioblastoma. In the present study, we investigated the role of hypoxia in erastin-induced ferroptosis in glioblastoma cell lines and determined the involved molecular mechanisms. Our study not only identi ed a novel mechanism of erastin resistance but also suggested a novel treatment thought by combination of ferroptosis activator and HIF-1α inhibitor.

Page 4/19
Cell lines Human glioblastoma cell lines U87 and U251 were directly purchased from Cell Bank of Type Culture Collection of Chinese Academy of Sciences (Shanghai, China). These cells were cultured in high glucose Dulbecco's modi ed Eagle's medium (DMEM, Gibco, USA) with 10% fetal bovine serum (FBS, Biological, Industries) and 1% penicillin-streptomycin (10378016, Invitrogen, USA) at 37°C with 5% CO2.
Cell growth and cell viability assay Cell growth was determined using a Cell Counting Kit-8 (CCK-8) assay kit (HY-K0301, MCE, Shanghai, China). U87 and U251 cells were seeded into 96-well plates for 0, 24, 48, and 72 h at a density of 3000 cells per well. Then, 10 µl CCK-8 solution was added to each well and incubated with the cells for 2 h. Absorbance (Abs) was detected at 450 nm using a microplate reader (Bio-Rad, Hercules, CA, USA).
For cell viability assay, cells were seeded at 1000 cells per well in 96-well plates with fresh medium and analyzed by using the CCK-8 assay kit according to the above instructions. The experimental group contained erastin (HY-15763, MCE, Shanghai, China), and the cells treated with DMSO were used as the control group. Herein, cell viability = (Abs of the experimental group-Abs of the blank group)/ (Abs of the control group-Abs of the blank group) × 100%.

Vector construction and transduction
Stable overexpression of GPX4 and the control genes in cells were achieved by transfection with lentivirus synthesized by Genomeditech (Shanghai, China). Cells were infected with Lv-GPX4 or negative control lentivirus vectors. GPX4 siRNA was used to silence GPX4 gene expression. The target sequence of GPX4 was 5′-CCGATACGCTGAGTGGGT-3′. A scrambled siRNA targeting 5′TTCTCCGAACGTGTCACGT-3′ was used as a negative control.
Half-maximal inhibitory concentration assay (IC50) The cells were planted in 96-well plates at 1.0 × 104 cells per well. The corresponding concentrations of erastin were given to cells for 24 h. After 24 h, CCK-8 was used to measure erastin sensitivity at 450 nm after incubating for 2 h at 37°C.

RNA extraction and quantitative real-time PCR array
Total RNA was extracted from cultured cells using a Total RNA extraction kit (RC112-01, Vazyme Biotech, Nanjing, China). The concentration and purity of RNA were measured by the absorbance at 260 nm and the ratio of 260/280nm in NanoDrop ND-1000 (NanoDrop, Wilmington, DE, USA). Total RNA from each sample was reversely transcribed using an all-in-one cDNA synthesis superMix (R333-01, Vazyme Biotech, Nanjing, China). SYBR Green PCR kit (R311-02, Vazyme Biotech, Nanjing, China) was used for real-time PCR. The primers of related genes are listed in Table S1. Quantitative PCR arrays are designed to analyze a panel of ferroptosis-related genes in human glioblastoma cell line U87 and U251 following the instructions of the manufacturer (Wcgene Biotechnology Corporation, China).
Immuno uorescence (IF) and Immunohistochemistry (IHC) U87 and U251 cells were grown on cover slides, xed, blocked with 3% BSA, and permeabilized with PBS containing 0.1% w/v Triton X-100. The mouse brain of the orthotopic intracranial mouse model was cut in the thickness of 20 µm by a frozen section and blocked with 3% BSA. For protein detection, the primary antibodies used have been already described in reagents and antibodies section. For IF, Secondary antibodies were goat IgG anti-rabbit conjugated to a uorochrome. Nuclei were stained by 4,6-diamidino2phenylindole (DAPI Cell apoptosis analysis by ow cytometry Cell apoptosis was measured by the Annexin V-uorescein isothiocyanate (FITC)/propidium iodide (PI) kit (BD556547, BD Bioscience, USA). Brie y, U87 and U251 cells were seeded into 6-well plates treated with saline, PX-478, erastin or PX-478 combined with erastin for 6 hours, and then the cells were collected and washed twice with cold PBS. Next, the cells were resuspended in 1×binding buffer at a concentration of 1×10 6 cells per ml and transferred 100 µl of the solution (1 × 10 5 cells) to a 5 ml culture tube. Incubated with 5 µL Annexin V-FITC and 10 µL PI for 15 min at (RT) 25°C in the dark and then add 400 µl of 1 × Binding Buffer to each tube. The results were analyzed using a FACS Calibur or an EPICS XL ow cytometer (BD Biosciences).

ChIP-qPCR assay
To determine whether HIF-1α can bind to the GPX4 promoter, a ChIP-qPCR assay was used. Brie y, cells were cross-linked using 1% formaldehyde, followed by scraping into ice-cold PBS with protease inhibitors. Next, the cells were collected and resuspended in lysis buffer (20 mM HEPES, pH = 7.9, 420 mM NaCl, 0.2 mM EDTA, 0.5% NP-40, 25% glycerol, 1.5 mM MgCl2), followed by several periods of sonication. One-third of the cell extract was put as input sample and two-thirds of the cell extract was used as substrate for immunoprecipitation with anti-HIF-1α antibody. Next, 5MNaCl was used to reverse the cross-linking after which the eluted DNA was extracted for PCR analysis. The primer sequences of GPX4 promoter were listed in Table S1.
TCGA data analysis mRNA expression data from the TCGA database was used to assess correlation between HIF-1α and GPX4 expression. The analyses were performed using the online analysis software named cBioPortal (http://www.cbioportal.org). We used Whole-exome and/or whole-genome sequencing of 257 tumor/normal pairs in glioblastoma (TCGA, Cell 2013) study [26].

Luciferase reporter assay
To investigate the role of interaction between HIF-1α and GPX4, we constructed GPX4 promoter-driven luciferase reporter plasmids and transfected into U87 and U251 cells. Cells were treated with PX-478 (HY-10231, MCE, Shanghai, China), or DMSO and exposed to 1% O2 for 16 h. The ratios of re y and Renilla luciferase activities were determined 48 h post-transfection using a Dual-Luciferase Assay kit (E1910, Promega, USA) according to the instructions of the manufacturer.

Malondialdehyde (MDA) assay
The relative MDA concentration in cell lysate was assessed using an MDA Assay kit (BC0025, Servicebio, China) according to the manufacturer's instructions [27]. Measure the absorbance at 532 nm using a microplate reader.

Subcutaneous xenograft and orthotopic xenograft mouse model
All experimental animal procedures were conducted strictly by the Guide for the Care and Use of Laboratory Animals and approved by the Animal Care and Use Committee of the Shandong provincial hospital a liated to Shandong University. The male BALB/c nude mice were randomized divide into 5 groups in a blinded manner, each group including six 4-weeks-old nude mice. In the subcutaneous xenograft model, 1 × 10 6 U87 cells were subcutaneously injected in the right anks of nude mice. At the second week after injection, control group was intraperitoneally administered DMSO meanwhile oral given saline twice a day; erastin group was intraperitoneally administered erastin (5 mg/kg) meanwhile oral given saline twice a day; PX-478 group was intraperitoneally administered DMSO meanwhile oral given PX-478 (100 mg/kg) twice a day; PX478+erastin group was intraperitoneally administered erastin (5 mg/kg) meanwhile oral given PX-478 (100 mg/kg) twice a day; PX-478+Lv-GPX4+erastin group was subcutaneously implanted 1 × 10 6 Lv-GPX4 transfected U87 cells and was intraperitoneally administered erastin (5 mg/kg) meanwhile oral given PX-478 (100 mg/kg) twice a day. At day 23 after implantation, subcutaneous tumors were collected, xed with 4% paraformaldehyde, and sectioned for measurement.
In the orthotopic xenograft mouse model, each mouse was intracranially injected with 1 × 10 5 luciferase transfected U87 cells in 10µL PBS solution using a stereotactic head frame at a depth of 3 mm through a bur hole placed 2 mm lateral and 2 mm Posterior to the bregma. At the second week after injection, mice received erastin (5 mg/kg) and PX-478 (100 mg/kg) treatment the same as subcutaneous xenograft model for 2 weeks. In vivo tumor growth was monitored with a Xenogen IVIS Spectrum system (PerkinElmer) every week after implantation.

Lipid peroxides assay
Lipid peroxides level was analyzed by ow cytometry using C11-BODIPY dye. U87 and U251 cells were seeded into 6-well plates at 3 × 10 5 cells per well and treated with ferroptosis inducer erastin as well as ferroptosis inhibitor ferrostatin-1 (HY-100579, MCE, Shanghai, China). The culture medium was replaced with 1ml medium containing 10 µM of C11-BODIPY (D3861, Thermo Fisher Scienti c, USA), and then cells were incubated for 30 min in a humidi ed incubator (at 37°C, 5% CO2). Media then were removed and washed cells with PBS three times. For measurement by using uorescence microscope, Images were acquired with a ImageXpress Micro Confocal (Molecular Devices, USA). For measurement by using ow cytometry, Cells were harvested in 15 ml tubes and resuspended in 500 µl of PBS. The cell suspension was ltered through cell strainer (0.4 µm nylon mesh). The results were analyzed using a FACS Calibur or an EPICS XL ow cytometer (BD Biosciences).

Transmission electron microscopy assay
Transmission electron microscopy analyses were conducted [28]. Brie y, U87 and U251 cells were seeded into 6-well plates at 3 × 10 5 cells per well and exposed to saline, PX-478 or lv-GPX4 combined with PX-478 for 6 h. After that, cells were collected, washed three times with PBS, and xed with 2.5% glutaraldehyde. Samples were then pretreated according to standard procedures including staining, dehydration, embedding, and slicing to obtain ultra-thin sections. During the analysis, images were acquired using a HITACHIH-7000 transmission electron microscope (TEM; Hitachi, Tokyo, Japan).

Statistical analysis
Data were presented as mean values ± standard deviation (SD) from at least three experiments. Student's t-test used to analyze the differences between two groups and One-way analysis of variance (ANOVA) used for the comparison among three or more groups were conducted using GraphPad Prism 8 (GraphPad Software, CA, USA), SPSS (IBM, NY, USA). p-values less than 0.05 were considered statistically signi cant (*p< 0.05, **p< 0.01).

Hypoxia protected glioblastoma cells against erastininduced ferroptosis
Hypoxia condition (1% O 2 , 5%CO 2 , 94%N 2 ) was a crucial factor that contributed to the aggressive behavior of glioblastoma [29]. We rst evaluated whether hypoxia could affect the viability of U87 and U251 cells by CCK-8 cell viability assay. As depicted in Fig. 1A, Hypoxia condition promoted U87 and U251 cell proliferation signi cantly. Then we explored the relationship between hypoxia and erastin resistance over a series of time points in U87 and U251 cells. Cell resistance was measured by the ratio (cells exposed to erastin (10 µM, 6 hours) / cells exposed to DMSO). We found hypoxia for at least 16 hours could induce a stronger resistance to erastin (Fig. 1B). Then we explored the cell viability at different concentrations of erastin for 6 h. We found glioblastoma cells under hypoxic environment became less sensitive to erastin (Fig. 1C). The half-maximal inhibitory concentrations (IC50) of hypoxic U87 and U251 cells to erastin were obviously high than that of normoxic cells (Fig. 1D). Furthermore, the lipid peroxidation level was measured by ow immunolabeling using C11-BODIPY probe. Results showed that high lipid peroxidation level induced by erastin (10 µM, 6 hours) was decreased signi cantly by hypoxic incubation (1% O 2 , 16 hours) and ferroptosis inhibitor ferrostatin-1 (fer-1, 0.5 µM, 6 hours) (Fig. 1E). Moreover, we explored the effect of hypoxia and erastin on cell colony formation ability. We found hypoxia condition and Fer-1 apparently enhanced the cell colony formation ability when treating cells with 10 µM erastin for 2 weeks (Fig. 1F). Together, these results suggested that hypoxic treatment for at least 16 hours decreased the sensitivity of glioblastoma cells to erastin. GPX4 acted as a potential effector molecular of hypoxia To de ne potential genes that played vital roles in erastin-resistance under hypoxic condition, an ferroptosis pro ling qPCR array was employed by comparing cells under hypoxic condition for 16 hours and normoxic condition. Each group was repeated four times ( Fig. 2A). Heterogeneity of tumor cells resulted in diverse qPCR array results between U87 and U251 cells. Considering that both cell lines showed similar decreased sensitivity to erastin, we identi ed genes that changed signi cantly in both cell lines as our potential targets. Three genes were picked out, namely GPX4, STEAP3 metalloreductase (STEAP3), carbonic anhydrase 9 (CA9) (Fig. 2B). Next, we detected the protein expression level of the three genes over a series of time points by western blot. We found the time course of GPX4 protein expression con rmed the trends of enhanced erastin-resistance, but not STEAP3 and CA9. At the 16th hour under hypoxia, GPX4 protein level increased signi cantly with the enhancement of resistance to erastin (Fig. 2C). To further verify this observation, we assessed GPX4 protein location and expression level via cellular immuno uorescence. The results showed higher GPX4 protein level after at least 16 hours hypoxic incubation and no apparent location changes at several indicated time points (Fig. 2D).
Taken all together, GPX4 acted as a potential effector molecular of hypoxia and we identi ed GPX4 as a target for further investigation.
Hypoxia enhanced the resistance to erastin via increasing GPX4 expression To investigate the role of GPX4 in hypoxia-induced resistance to erastin, we used GPX4 siRNA to inhibit GPX4 gene expression in U87 and U251 cells. Firstly, we analyzed the e cacy of GPX4 knockout (GPX4 KO). Western blot showed siRNA can signi cantly inhibit the protein expression level of GPX4 (Fig. 3A). Quantitative analysis results were showed in gure S1. CCK-8 cell viability assay showed GPX4 knockout did not in uence the cell proliferation rate of glioblastoma cells (Fig. 3B). Cytotoxicity experiments showed GPX4 knockout could reverse the enhanced erastin-resistance induced by hypoxia, meanwhile, GPX4 knockout under normoxia could cause cells to be more sensitive to erastin (Fig. 3C).
The IC50 of GPX4 knockout cells was obviously lower than that of scramble cells under hypoxic condition (Fig. 3D). Next, we explored whether GPX4 knockout could in uence the cell colony formation ability. We found GPX4 knockout could reverse the enhanced cell colony formation ability induced by hypoxia when treating with erastin and GPX4 knockout under normoxia led to less colony formation when treating with erastin (Fig. 3E). In short, we identi ed GPX4 as one of the most vital gene which regulated the ferroptosis defense ability of glioblastoma cells under hypoxic environment. PI3K/AKT/HIF-1α pathway involved in the up-regulation of GPX4 induced by hypoxia PI3K/AKT pathway had been reported to participate in malignant progression of glioblastoma and considered as an important regulator of ferroptosis [30][31]. HIF-1α was a well-known transcription factors that play an important role in the early stage of hypoxia [16]. Thereby, we hypothesized that the increased GPX4 expression might owing to the activation of the PI3K/AKT/HIF-1α pathway. To prove our hypothesis, we explored the protein changes in the PI3K/AKT/HIF-1α pathway by western blot. Results indicated that hypoxia incubation for 16 hours obviously activated PI3K, phosphorylated AKT, HIF-1α and GPX4. Treatment with the AKT inhibitor MK-2206 along with hypoxia incubation signi cantly attenuated the expression level of p-AKT, HIF-1α and GPX4 (Fig. 4A). Quantitative analysis results were showed in gure S2. Then we employed HIF-1α inhibitor PX-478 to further test the function of HIF-1α. Results showed PX-478 can markedly decreased HIF-1α and GPX4 protein levels (Fig. 4B). Then we analyzed public data [26] and found the expression of GPX4 was positively correlated with HIF-1α in 257 glioblastoma tumor samples (Fig. 4C). To understand the molecular mechanism underlying the regulation of GPX4 by HIF-1α, we performed chromatin immunoprecipitation (ChIP) with an anti-HIF-1α antibody to test whether HIF-1α can bind to the GPX4 promoter region. We designed three pairs of primer to detect different regions of GPX4 promoter respectively and uncovered that HIF-1α can indeed bind to the GPX4 promoter region −131 bp to 1 bp (Fig. 4D). Therefore, we transfected GPX4 promoter-driven luciferase reporter plasmid into U87 and U251 cells and tested relative luciferase activity under normoxia condition, hypoxia condition, and hypoxia with PX-478 condition. Hypoxia group exhibited increased luciferase activity and hypoxia with PX-478 group resulted in the decrease of luciferase activity (Fig. 4E). Finally, we explored whether the erastin-resistance ability induced by hypoxia was related to HIF-1α. We found inhibition of HIF-1α by PX-478 can reverse the enhanced erastin-resistance (Fig. 4F) and elevated IC50 (Fig. 4G) induced by hypoxia in both cell lines.

HIF-1α inhibitor PX-478 induced lipid peroxidation in glioblastoma cells
Given that ferroptosis resistance can be induced by hypoxia mentioned above and lipid peroxidation underlay the mechanism of ferroptosis [22]. we determined whether suppression of HIF-1α induced lipid peroxidation in glioblastoma cells. Indeed, we utilized C11-BODIPY probe and observed that HIF-1α inhibitor PX-478 (20µM) signi cantly increased lipid peroxidation level in the glioblastoma cells. To determine whether GPX4, as a downstream of HIF-1α, could reverse lipid peroxidation level induced by PX-478, the glioblastoma cells were transfected with lentivirus which overexpressed human GPX4 (PX-478+ Lv-GPX4 group). As expected, the lipid peroxidation level was decreased in the PX-478+ lv-GPX4 group (Fig. 5A). As depicted in Figure. 5B, the ratios of C11-BODIPY+ cells in PX-478 treatment group were 12.7% and 21.4% in U87 and U251 cell lines respectively, and those in control group were 3.55% and 4.43%, indicating the lipid peroxidation level after treating with PX-478 was signi cantly increased. Importantly, combination of PX-478 and lv-GPX4 resulted in a low lipid peroxidation level, the ratios of C11-BODIPY+ cells were 4.62% and 5.15% in U87 and U251 cells respectively. Furthermore, we con rmed it by measuring malondialdehyde (MDA), a product of lipid oxidized damage. Results showed that the lipid oxidized damage level was increased after treating with PX-478 and reversed in the presence of Lv-GPX4 (Fig. 5C). Transmission electron microscopy revealed that PX-478 treated cancer cells exhibited shrunken mitochondria with enhanced membrane density, which were morphologic features of ferroptosis. Black arrays represented normal mitochondrial. White arrays represented shrunken and high membrane density mitochondrial. GPX4 overexpression could reverse it (Fig. 5D). Then, we studied the potential role of PX-478 in erastin-induced ferroptosis in normoxic condition. We found that treatment with PX-478 alone could not cause cell death, which suggested that a modest degree of lipid peroxidation induced by PX-478 will not result in cell death directly. We examined the combined effect of erastin and PX-478. Results indicated that the combined medication had a signi cant higher cell death rate than erastin alone uncovered by ow cytometry analysis and TUNEL staining, meanwhile, GPX4 overexpression blocked the cytotoxic effect (Fig. 5E, F).

PX-478 enhanced the anticancer activity of erastin in vivo
To further investigate the role of PX-478 in promoting erastin sensitivity in vivo, we established subcutaneous xenograft and orthotopic xenograft mouse model of glioblastoma as described in methods. In both subcutaneous xenograft and orthotopic xenograft model, erastin (5 mg/kg) was intraperitoneal injected twice a day and PX-478 (100 mg/kg) was oral given twice a day for 2 weeks. Detailed experimental process was put in Materials and Methods. Compared with control group (DMSO intraperitoneal injection and saline oral given), erastin group (erastin intraperitoneal injection and saline oral given) exhibited a certain degree of anticancer effect but not PX-478 group (DMSO intraperitoneal injection and PX-478 oral given). Combination group (erastin intraperitoneal injection and PX-478 oral given) had an obviously better anticancer activity than erastin group. However, Lv-GPX4 group (injection with GPX4 overexpression cells, erastin intraperitoneal injection and PX-478 oral given) showed no anticancer effect measured by tumor volume and tumor weight in subcutaneous xenograft model ( Fig. 6A-C). Similar results were obtained in orthotopic xenograft model measured by luciferase activity (Fig. 6D-E). The overall survival of combination group was also higher than erastin group (Fig. 6F). Furthermore, the histological analysis indicated that PX-478 group and combination group exhibited weaker immunoreactivity for GPX4 and Lv-GPX4 could e ciently increase GPX4 expression (Fig. 6G).
Collectively, PX-478 had very little therapeutic effect on glioblastoma, but it can greatly enhance the anticancer effect of erastin in both subcutaneous xenograft and orthotopic xenograft mouse model.

Discussion
The incidence of glioblastoma continues to increase globally with the aging of the world's population and glioblastoma remains to have a very low 5-year overall survival rate of only 6.8% [32]. Tumor cells can develop resistant to various drugs during therapy process including drugs based on ferroptosis [33]. Therapies underlying ferroptosis attracted much attention with high expectations in recent years [34]. Our work has rst documented that hypoxia for at least 16 hours can enhance glioblastoma resistance to erastin-induced ferroptosis. Therefore, a better understanding of the mechanism underlying erastinresistance may bene t ferroptosis-targeted treatment. We found one of the most important underlying mechanism of erastin resistance was the up regulation of GPX4, which functioned as the primary enzyme that prevented ferroptosis [35]. Other genes such as STEAP3 and CA9 uncovered by PCR array may also participated in the erastin resistance induced by hypoxia, but GPX4 expressed stable in different types of cells and was considered as the most crucial gene.
PI3K/AKT signaling pathway plays a central role in cell survival [30]. The allosteric AKT inhibitor MK2206 has been reported to decrease tumor growth and enhance the antitumor e cacy of chemotherapeutic agents [36][37]. We found that activated PI3K/AKT/HIF-1α axis by hypoxia was essential for the upregulation of GPX4. Both AKT inhibitor MK-2206 and HIF-1α inhibitor PX-478 could signi cantly inhibit GPX4 expression under hypoxic environment. These results suggested the up-regulation of GPX4 induced by hypoxia was dependent on the activation of PI3K/AKT/HIF-1α axis.
HIF-1α was a famous transcription factor and regulated multiple environmental adaption genes, but there was no passage about the relationship between HIF-1α and any ferroptosis-related genes. Here, we rst found HIF-1α directly regulated GPX4 transcription by binding to its promoter region -131 bp to 1 bp. HIF-1α inhibitor PX-478 had antitumor effect with fewer side effect, but the effect of PX-478 alone was limited [38]. Scientists paid much attention on the synergetic effect of PX-478 including combination with radiotherapy or immunotherapy [39][40]. Radiotherapy and immunotherapy were related to ferroptosis [41][42]. Given that HIF-1α regulated ferroptosis core gene GPX4, we hypothesized HIF-1α inhibitor PX-478 played a vital role in promoting ferroptosis by inhibiting GPX4 expression. We found cells treated with PX-478 exhibited higher lipid peroxidation level by C11-BODIPY staining. We also found deeply stained, shrunken mitochondria by transmission electron microscopy.
Tumor microenvironment was featured with regional hypoxia [43]. So, we hypothesized the regional hypoxia of tumor microenvironment in vivo strengthened the erastin-resistance by HIF-1α/GPX4 axis. We employed subcutaneous xenograft and orthotopic xenograft mouse model to test the function of PX-478 in erastin therapy. We obtained ideal experimental results, which showed PX-478 can effectively decreased the GPX4 expression level in tumor tissue and promoted the anticancer effect of erastin.
To our best knowledge, this is the rst study to explore the role of hypoxia in erastin resistance in glioblastoma. And this is the rst study to uncover the relationship between PI3K/AKT/HIF-1α pathway and ferroptosis core gene GPX4. We rst put forward that combination of erastin and PX-478 had a good effect on anticancer effect. However, we did not test other ferroptosis-related therapy. Besides, we had not focused on patients, because erastin had limitations when used in human because of its low solubility in water and side effects.

Conclusion
In conclusion, we demonstrated that hypoxia enhanced glioblastoma resistance to erastin-induced ferroptosis by activating the PI3K/AKT/HIF-1α pathway and promoting GPX4 expression (FIG. 7). We

Declarations
Ethics approval and consent to participate The protocol of this study was approved by the Ethics and Human Subject Committee of Shandong Provincial Hospital, and all experiments and methods met the standards of the relevant guidelines and regulations.

Consent for publication
Not applicable.
Availability of data and materials The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests  GPX4 was screened out as a potential effector molecular of hypoxia A. Ferroptosis-related qPCR array of normoxia and hypoxia identi ed several differentially expressed genes in U87 and U251 cells. B. The volcano map of differentially expressed genes depicted by difference degree and p-value. GPX4, STEAP3 and CA9 were picked out. C. Protein expression level of GPX4, STEAP3 and CA9 over a series of time points measured by western blot. GPX4 protein expression increased signi cantly at the 16th hour under hypoxia. D. GPX4 protein location and expression level uncovered by Immuno uorescence. Higher GPX4 protein level after at least 16 hours hypoxic incubation but no apparent location changes. The bar graph showed mean ± SD of 3 independent experiments. *P < 0.05, **P < 0.01.    graph showed mean ± SD of 3 independent experiments. *P < 0.05, **P < 0.01.