The Positive Feedback Loop Between RLIP76 and HIF-1α Promotes Glycolysis and Tumorigenesis in Glioma Cells Under Hypoxic Conditions

Background Hypoxia is intimately associated with increased glycolysis in gliomas, and HIF-1α plays a critical role in this process. Here, we aim to show that RLIP76 is a novel target of HIF-1α and is involved in hypoxia-enhanced glycolysis in glioma cells. The human glioma cell lines U87 and U251 were used to explore the interactions of RLIP76-HIF-1α and RLIP76-VHL using Western blot, a Biotin pull-down assay, Immunoprecipitation, and a Chromatin immunoprecipitation assay under hypoxia. U251 cells pretreated with hypoxia were used for tumor xenografts. Twenty-four hours after transfection, the cells were cultured under normoxic or hypoxic conditions for 24 h. Lactate levels (H), ECAR values (I) and intracellular glucose levels (J) were then examined. (K) U251 cells treated with either control or RLIP76 siRNA were transfected with or without Lenti-HIF-1α. Twenty-four hours after transfection, the cells were cultured under normoxic or hypoxic conditions for 24 h. Real-time RT-PCR and Western blotting were used to measure LDHA, GLUT1, and HIF-1α mRNA and protein expression levels, respectively.


Abstract Background
Hypoxia is intimately associated with increased glycolysis in gliomas, and HIF-1α plays a critical role in this process. Here, we aim to show that RLIP76 is a novel target of HIF-1α and is involved in hypoxiaenhanced glycolysis in glioma cells.

Methods
The human glioma cell lines U87 and U251 were used to explore the interactions of RLIP76-HIF-1α and RLIP76-VHL using Western blot, a Biotin pull-down assay, Immunoprecipitation, and a Chromatin immunoprecipitation assay under hypoxia. U251 cells pretreated with hypoxia were used for tumor xenografts.

Results
RLIP76 is a novel target of HIF-1α and contributes to hypoxia-enhanced glycolysis in glioma cells. HIF-1αinduced RLIP76 can critically regulate the stability of HIF-1α by alleviating VHL-mediated HIF-1α ubiquitination under hypoxia. RLIP76 interacts with HIF-1α and VHL through an RLIP76 GAP-dependent mechanism. The GAP function of RLIP76 regulates its complex formation because of the dependence of RLIP76 GAP on interactions between RLIP76 and these proteins. RLIP76 knockdown results in decreased U251 cell growth after hypoxia pretreatment in vivo.

Conclusions
This study reveals a positive feedback loop between HIF-1α and RLIP76, suggesting that RLIP76 may undergo a complex series of GAP-dependent interactions with HIF-1α and VHL to protect HIF-1α from degradation while promoting glycolysis under hypoxic conditions. We indicate that RLIP76 is crucial for the regulation of hypoxia-enhanced glycolysis and may provide a new gene therapy approach for glioma patients.

Background
Glycolysis is a key regulatory pathway for glucose metabolism in aerobic and anaerobic organisms [1].
Glycolysis not only supplies pyruvate to the TCA cycle through oxidative phosphorylation to generate ATP but also helps to produce precursor molecules for protein, nucleic acid and lipid synthesis [2]. Glycolysis is regarded as the main energy source for the rapid proliferation of normal cells. However, enhanced glycolysis and decreased oxidative phosphorylation, known as the Warburg effect, are often observed in various cancerous cells [3]. Emerging evidence suggests that highly proliferating tumor cells can shift their metabolism from mitochondrial glucose oxidation to cytoplasmic glycolysis to meet the energy demands of biosynthesis and bioenergy [4]. In light of this fact, glycolysis inhibition may provide a rational strategy for the long-term management and prevention of cancers.
Hypoxia is a major characteristic of solid tumors, especially in rapidly growing gliomas, which contain incomplete blood vessel networks [5]. Hypoxic conditions induce a molecular response that results in the activation of hypoxia-inducible factor 1α (HIF-1α), a key transcription factor that regulates various genes involved in glioma metabolism and escape from a nutrient-deprived environment [6]. It has been reported that the expression level of HIF-1α increases gradually with increasing pathological grade of glioma, highlighting its importance in glioblastoma (GBM) [7]. In addition, the HIF-1α pathway is a positive regulator of glycolysis, as its suppression can lead to the inhibition of several glycolytic enzymes, which catalyze the irreversible reactions of the glycolytic pathway [8]. Emerging evidence indicates that a high glycolytic rate is the de ning characteristic of gliomas and that inhibition of the HIF-1α pathway causes a signi cant decrease in glycolysis and contributes to the enhancement of oxidative phosphorylation in several glioma cell lines [9][10][11], reinforcing the important role of HIF-1α in regulating the glycolytic pathway.
RLIP76 is an important member of the GTP enzyme activating protein family [12]. It is thought that RLIP76 participates in the oncogenesis and progression of several cancers by activating a wide array of signaling pathways, such as Ras signaling, which is a key pathway responsible for modulating HIF-1α expression and activity [13][14][15]. In glioma, RLIP76 is highly expressed and acts as an independent prognostic marker [16]. Several oxygen-dependent chemotherapy drugs (cyclophosphamide, platinum, doxorubicin, etc.) can induce the chemoresistance of various tumor cells by affecting HIF-1α-induced glycolysis [17][18][19][20], while RLIP76 knockdown can dramatically enhance the sensitivity of tumor cells to the above chemotherapy drugs [21][22][23]. Therefore, these results imply that RLIP76 is closely associated with HIF-1α and glycolysis. However, little attention has been paid to the role of RLIP76 in the HIF-1α signaling pathway, and the mechanism of RLIP76-mediated glycolysis is still unclear.
In this study, we identi ed RLIP76 as an important regulator of glioma progression through suppression of glucose metabolism under hypoxic conditions. Mechanistically, RLIP76 increased the accumulation and stabilization of HIF-1α in a GTPase-activating protein (GAP)-dependent manner by decreasing HIF-1α degradation through von Hippel-Lindau (VHL); in turn, the upregulated HIF-1α could directly interact with RLIP76 in hypoxia to increase glucose consumption and lactate excretion. These results establish RLIP76 as a novel regulator of glucose metabolism that enhances hypoxia-induced glycolysis to facilitate tumor progression.

Materials And Methods
Tissue samples and cell culture The study protocol complied with the National Regulations on the Use of Clinical Samples in China. The use of human specimens in this study was approved by the Specialty Committee on Ethics of Biomedicine Research, Shanghai Tongji University (PJ2015-011-08). Written informed consent was obtained from all participants prior to their participation in the study. The glioma cancer specimens were obtained from patients who underwent surgery at Shanghai Tongji Hospital between July 2016 and July Page 4/28 2018. The malignant glioma cell lines U87 and U251 were obtained from the Chinese Academy of Sciences (Shanghai, China). All glioma cell lines were grown in Dulbecco's Modi ed Eagle's Medium supplemented with 10% fetal bovine serum (FBS) in a humidi ed atmosphere containing 5% CO 2 at 37 °C.

Biotin pull-down assay
Biotin pull-down assays were performed as previously described [24]. In vitro-synthesized biotin-labeled RLIP76 (3 mg) was incubated with puri ed GST, GST-VHL protein, or HIF-1α or VHL deletion mutants for 3 h for the in vitro RNA pull-down assay. The RNA-protein complex was isolated by streptavidin-coupled Dynabeads (Invitrogen).

Immunoprecipitation assays
Immunoprecipitation was performed using lysates from 1 × 10 7 cells treated with RNase A inhibitor and DNase I before centrifugation at 4 °C for 15 min. Cell lysates were incubated with protein A/G beads coated with the indicated antibodies at 4 °C for 3 h and then washed with wash buffer. Total protein was diluted to 1 µg/µl with PBS to decrease the concentrations of detergents. The supernatant was centrifuged and then loaded on 10% SDS-PAGE gels and transferred onto nitrocellulose membranes (Millipore) for Western blot analysis.
Luciferase assay U251 cells treated with either control or RLIP76 siRNA were transfected with hypoxia response element (HRE)-WT or HRE-MT reporter plus Renilla luciferase plasmid to investigate the effect of RLIP76 on the activity of the HRE reporter construct. Twelve hours after transfection, cells were cultured under normoxia or hypoxia for another 24 h. A luciferase assay kit (Promega) was used to measure the reporter activity.
To explore the effect of HIF-1α on the transcriptional regulation of RLIP76, U251 cells treated with either control or HIF-1α siRNA were cotransfected with a pGL3-based construct containing the RLIP76 promoter plus the Renilla luciferase plasmid. Twenty-four hours after transfection, cells were cultured under normoxia or hypoxia for an additional 24 h. A luciferase assay kit (Promega) was used to measure the reporter activity, which was normalized to Renilla luciferase activity.

Bioinformatic analysis
LinkedOmics (www.linkedomics.org) helps researchers understand complex correlations among the vast amounts of clinical and molecular attributes of cancers [25]. A LinkedOmics analysis was performed for The Cancer Genome Atlas (TCGA) GBM database to examine the expression of RLIP76 in GBM. DAVID Bioinformatics Resources 6.8 (http:// david.ncifcrf.gov) was used to perform gene ontology (GO) analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis on the differentially expressed genes.
After analyses were performed to determine the signi cance and the false discovery rate (FDR), GO terms were selected from the signi cantly enriched gene sets (P < 0.05 and FDR < 0.05) as previously described [26]. These analyses were conducted using Gene Set Enrichment Analysis (GSEA) with a minimum number of 5 genes (size), a simulation of 1000 and a P-value of 0.05. PROMO (http://alggen.lsi.upc.es/cgi-bin/promo_v3/promo) is a virtual laboratory for the study of transcription factor binding sites within DNA sequences [27]. We used a PROMO analysis to determine the potential binding sites in the hypoxia response element (HRE) within the RLIP76 promoter region.

Xenograft analysis
Four-week-old male nude mice were randomly divided into two groups (n = 6 mice per group) and then injected subcutaneously with control or RLIP76 siRNA-transfected U251 cells, which were pretreated with hypoxia for 12 h before injection. Twenty-eight days after injection, the mice were sacri ced, and the tumors were excised and analyzed. Animal studies were approved by the Ethics Committee of the School of Shanghai Tongji Hospital of Tongji University.

Extracellular acidi cation rate (ECAR)
The ECAR measurement was performed as previously described [28]. Brie y, the experiment was conducted by an XF96 Extracellular Flux Analyzer (Seahorse Bioscience). Cells were incubated in nonbuffered DMEM containing 10 mM glucose. After the addition of 100 µM oligomycin and 500 mM 2deoxy-D-glucose, the ECAR was measured under basal conditions. Glucose uptake and lactate production Cellular glucose uptake was evaluated by a uorescence-based glucose assay kit (BioVision) according to the manufacturer's instructions. To measure lactate production, U87 and U251 cells transfected with either control or RLIP76 siRNA were cultured in normoxia or hypoxia for 24 h. Lactate production was measured using a uorescence-based lactate assay kit (BioVision).
Chromatin immunoprecipitation assay (ChIP assay) U251 cells treated with either control or RLIP76 siRNA were cultured in normoxic or hypoxic conditions for 24 h for ChIP assays. The ChIP assay was conducted according to the protocol of the ChIP assay kit (Upstate Biotechnology, Lake Placid, NY).

Real-Time RT-PCR
Real-time RT-PCR was performed as previously described [24]. The primers used are listed in Table S1.

Statistics
All data are expressed as the mean ± SD values from triplicate experiments. Paired groups were compared by Student's t-test, whereas three or more groups were compared by analysis of variance followed by Tukey's test for multiple pairwise comparisons. The correlation between RLIP76 and HIF-1α expression was assessed using Spearman's correlation coe cient. The statistical analysis was performed with SPSS 13.0 software, and P < 0.05 was considered statistically signi cant.

Results
Hypoxia-Induced RLIP76 is Required for Hypoxia-Enhanced Glycolysis We rst searched for potential hypoxia response elements (HREs) in the RLIP76 promoter to explore whether RLIP76 acts as a hypoxia-responsive gene that participates in hypoxia-enhanced glycolysis in hypoxic conditions. Two potential HREs in the promoter region of RLIP76 were identi ed by bioinformatics analysis ( Figure S1A), which suggests that RLIP76 is a potential hypoxia-responsive gene. Real-time RT-PCR and Western blotting analysis were utilized to explore the expression levels of RLIP76 in U87, U251, SW1088, and A172 glioma cell lines under normoxic or hypoxic conditions. We found that RLIP76 mRNA and protein expression levels were signi cantly induced by hypoxia in all glioma cells (Figs. 1A and 1B). In addition, RLIP76 was increased under hypoxic conditions in a dose-dependent manner in U251 cells (Figs. 1C and 1D). All these results strongly suggest that RLIP76 is a hypoxiaresponsive gene in glioma cells.
To investigate whether RLIP76 is required for the malignant progression of glioma under hypoxic conditions, we exposed U87 and U251 to normoxia or hypoxia (1% O2) for 24 h. As shown in Figure S1B, glioma cells treated with RLIP76 siRNA had a signi cant decrease in the RLIP76 protein. RLIP76 knockdown suppressed cell proliferation, promoted apoptosis, and inhibited migration, both in normoxic and in hypoxic conditions (Figures S1C-S1E). Moreover, RLIP76 knockdown critically reversed the effect of hypoxia on cell proliferation, apoptosis, and migration (Figures S1C-S1E). These results indicate that RLIP76 critically regulates hypoxia-induced malignant cell progression of gliomas.
We next performed bioinformatics analysis on RLIP76 coexpressed genes to predict the function of RLIP76 in GBM. We investigated 17,084 genes associated with RLIP76 expression by LinkedOmics analysis and displayed them in a heat map (Spearman correlation analysis, p < 0.05) that showed genes that are positively (8104 genes) and negatively (8980 genes) correlated with RLIP76 expression in 150 samples from the TCGA GBM database (Figures S2A and S2B). GO analysis and KEGG analysis of RLIP76 coexpressed genes were performed to explore the potential function of RLIP76 in GBM. GO analysis showed that the RLIP76 coexpressed genes were mainly involved in biological processes such as metabolic process, response to stimulus and cell communication. (Figure S2C). KEGG analysis showed that RLIP76 coexpressed genes were mainly involved in signaling pathways such as oxidative phosphorylation and ubiquitin-mediated proteolysis, which are signi cant processes in the progression of hypoxia-enhanced glycolysis ( Figure S2D). These ndings prompted us to investigate the relationship between RLIP76 and glucose metabolism. To determine whether RLIP76 modulates glucose metabolism, we explored the effects of RLIP76 knockdown on glucose uptake, lactate production and the extracellular acidi cation rate (ECAR) in glioma cell lines (U87 and U251) with RLIP76 knockdown by RLIP76 siRNA. Hypoxia treatment resulted in a dramatic enhancement in glucose uptake and lactate production (Figs. 1E and 1F). To con rm these results, we determined the glycolytic capacity by measuring the ECAR, which is mainly caused by the excretion of lactic and carbonic acids during glucose metabolism, in hypoxic U251 cells in hypoxic conditions and revealed substantially enhanced glycolytic capacity (Figs. 1G and 1H). Conversely, all these hypoxia-induced effects were signi cantly rescued by downregulation of RLIP76 (Figs. 1E-1H), indicating that RLIP76 serves as an important regulator of hypoxia-enhanced glycolysis. RLIP76 Regulates Hypoxia-Enhanced Glycolysis via a HIF-1α-Dependent Mechanism Given that the downregulation of RLIP76 signi cantly inhibited hypoxia-induced glucose metabolism (Figs. 1E-1H), we then investigated LDHA enzymatic activity and the expression of key glycolysis enzymes. RLIP76 knockdown led to a signi cant suppression of the enhancing effect of hypoxia on LDHA enzymatic activity and LDHA and GLUT1 expression (Figs. 2A-2C). Mounting evidence has revealed that increases in LDHA and GLUT1 induced by hypoxia are regulated by HIF-1α. We next investigated whether RLIP76 regulates hypoxia-enhanced glycolysis by modulating HIF-1α transcriptional activity. Considering that HIF-1α functions by binding to the HRE in hypoxic conditions, we explored the effects of RLIP76 on HIF-1α-regulated luciferase expression using an HRE luciferase reporter system. Hypoxia signi cantly promoted the luciferase expression of the WT HRE reporter but not the Mut HRE reporter, and this effect was signi cantly inhibited by downregulation of RLIP76 (Fig. 2D). Furthermore, when HIF-1α activity was suppressed by the HIF-1α inhibitor digoxin or when HIF-1α expression was decreased via HIF-1α-siRNA, the downregulation of RLIP76 had no effects on WT reporter activity under hypoxic conditions (Figs. 2E, 2F, S3A, and S3B). This nding suggests that hypoxia-induced transcription is in uenced by RLIP76 through HIF-1α. Notably, we also revealed that the expression levels of hypoxiainduced HIF-1α and HIF-1α target genes (LDHA and GLUT1) were strongly attenuated by RLIP76 inhibition (Fig. 2G), which indicates that RLIP76 plays a signi cant role in hypoxia-induced HIF-1α signaling pathway. Moreover, overexpression of RLIP76 by Lenti-RLIP76 (an RLIP76 overexpression plasmid) reversed the HIF-1α-mediated reduction in hypoxia induced by RLIP76 knockdown ( Figure S3C). These results suggest that RLIP76 modulates HIF-1α transcriptional activity by regulating its expression under hypoxic conditions. To further explore whether RLIP76 regulates hypoxia-induced glycolysis via HIF-1α, RLIP76 knockdown cells were transfected with exogenous HIF-1α. The effect of RLIP76 inhibition on lactate production (Fig. 2H), ECAR (Fig. 2I), glucose uptake (Fig. 2J), and GLUT1 and LDHA expression (Fig. 2K) was abrogated by exogenous HIF-1α transfection. Together, these results indicate that the acceleration of hypoxia-induced glycolysis caused by RLIP76 is mediated through HIF-1α signaling via the promotion of GLUT1 and LDHA expression. Hypoxia-Induced RLIP76 Increases HIF-1α Stabilization by Attenuating the VHL-HIF-1α Interaction We then investigated the mechanism by which hypoxia-induced HIF-1α expression is regulated by RLIP76. Real-time RT-PCR analysis revealed that HIF-1α mRNA levels did not change under hypoxic conditions after RLIP76 knockdown (Fig. 3A). However, RLIP76 inhibition resulted in a signi cant reduction in the HIF-1α half-life under hypoxic conditions (Fig. 3B). This effect of RLIP76 on HIF-1α was correlated with proteasome-dependent degradation, as the proteasome inhibitor MG132 could restore the RLIP76 knockdown-induced reduction of HIF-1α expression under hypoxic conditions (Fig. 3C). In addition, the RLIP76 knockdown-induced HIF-1α reduction could be reversed by VHL knockdown (Fig. 3D), implying that the effect of RLIP76 on HIF-1α expression is dependent on VHL. The binding of hydroxylated HIF-1α at proline 564 (hyp564) is necessary for VHL-mediated HIF-1α ubiquitination and subsequent degradation. Therefore, we examined the effect of RLIP76 on hyp564-HIF-1α expression levels. RLIP76 knockdown signi cantly decreased hyp564-HIF-1α expression in hypoxic conditions (Fig. 3B). This effect of RLIP76 on hyp564-HIF-1α could be restored by MG132 treatment (Fig. 3E), indicating that RLIP76 knockdown could promote hyp564-HIF-1α degradation. Moreover, we demonstrated that the stimulatory effect of RLIP76 on hyp564-HIF-1α is depended on VHL without directly in uencing HIF-1α hydroxylation status, as VHL inhibition recovered RLIP76 knockdown-caused hyp564-HIF-1α decrease under both normoxic and hypoxic conditions (Fig. 3D), while no signi cant difference was observed on hyp564-HIF-1α expression in VHL-knockdown cells upon downregulation of RLIP76 under hypoxia ( Figure S3D). As predicted, RLIP76 inhibition signi cantly enhanced HIF-1α polyubiquitination under hypoxia (Fig. 3F). Consistent with these ndings, the polyubiquitination of hypoxic HIF-1α was signi cantly increased in RLIP76-siRNA-transfected cells (RLIP76 knockdown cells) and returned to that of the controls upon restoration of RLIP76 by transfecting the Lenti-RLIP76 plasmid (RLIP76-overexpressing cells) (Fig. 3G). We further demonstrated that RLIP76 knockdown promoted HIF-1α-VHL and hyp465-HIF-1α-VHL associations under hypoxic conditions (Figs. 3H and 3I). As illustrated in Fig. 3J, RLIP76 was able to suppress the associations between VHL and HIF-1α. Collectively, these data imply that hypoxia-induced RLIP76 attenuates the interaction of VHL and HIF-1α to inhibit the VHLmediated degradation of HIF-1α by the ubiquitin-proteolytic pathway.
It has been shown that RLIP76 can function as a regulator of heat shock factor by interfering with its ability to form complexes with its target molecules [13], and we hypothesized that RLIP76 could attenuate the VHL-HIF-1α interaction in a similar manner. As shown in Fig. 4A and 4B, RLIP76 could bind to VHL and HIF-1α according to immunoprecipitation assays. The complexes formed between RLIP76-HIF-1α and RLIP76-VHL were also con rmed by GST pull-down of endogenous RLIP76 (Figs. 4C, and 4D). A subsequent co-IP assay further demonstrated that RLIP76 was not capable of forming a trimeric complex with VHL and HIF-1α (Fig. 4E), implying that there is a competitive interaction among RLIP76, HIF-1α and VHL.
To determine the structural factors for RLIP76-VHL and RLIP76-HIF-1α complex formation, we conducted deletion mapping experiments. The mutant VHL (aa 54-213, aa 54-213), which is known to be able to interact with HIF-1α, was found in a complex with RLIP76, while the other three VHL deletion mutants (aa 72-213, 1-156, and 1-186) without HIF-1α-binding ability showed no interaction with RLIP76 (Figs. 4F and 4G), which implies that RLIP76 and HIF-1α might competitively interact with VHL. In addition, all HIF-1α deletion mutants could interact with RLIP76 (Figs. 4H and 4I), which suggests that RLIP76 could be associated with HIF-1α via multiple domains. All these data suggest that RLIP76 may affect the HIF-1α-VHL complex by directly interacting with each protein.
Signi cance of the GAP Function of RLIP76 in the Interaction with HIF-1α and VHL The RLIP76 protein serves as an ATP-binding plasma membrane transporter with two ATP-binding sites at residues 69-74 and 418-425 [16]. RLIP76 ATPase activity has been well documented to be involved in the regulation of multiple pathways, including the ubiquitin-proteolytic pathway. Using RLIP76 ATPbinding mutants (K74M and K425M), we then explored whether the ATPase activity of RLIP76 is necessary for its interactions with HIF-1α and VHL. ATPase RLIP76 mutants appeared to form complexes with HIF-1α or VHL (Figs. 4J-4N), which indicates an ATP-independent role for RLIP76 in its interactions with HIF-1α and VHL.
RLIP76 was rst cloned as a Ral-binding GTPase-activating protein (GAP) and negative regulator of heat shock factor-1 (Hsf-1) because of its ability to form the RLIP76-Hsf-1 complex, which is due to the GAP function of RLIP76. We then utilized GAP-de cient R208L and K224R RLIP76 mutants to explore the effect of RLIP76 on HIF-1α or VHL complex formation. In contrast to ATP-de cient mutants, the R208L and K244R RLIP76 mutants failed to integrate with HIF-1α or VHL (Figs. 4J-4N), suggesting the involvement of the GAP function of RLIP76 in binding to HIF-1α and VHL. Overall, the GAP function of RLIP76, but not the ATPase function, is required for formation of the RLIP76-HIF-1α and RLIP76-VHL complexes. RLIP76 is a Novel Transcriptional Target of HIF-1α To investigate the molecular mechanism of aberrant RLIP76 expression induced by hypoxia, we inhibited HIF-1α, HIF-2α, and p53, known as classical hypoxia-responsible transcriptional factors, in U251 cells. Downregulation of HIF-1α, but not HIF-2α or p53, strongly alleviated RLIP76 mRNA and protein upregulation induced by hypoxia (Figs. 5A and 5B). In addition, overexpression of HIF-1α greatly increased RLIP76 expression in a dose-dependent manner (Fig. 5C). U251 cells treated with CoCl2 or DMOG (known as HIF-1α inducers) showed a signi cant increase in RLIP76 mRNA expression (Fig. 5D). These ndings indicate that HIF-1α is associated with hypoxia-induced RLIP76 expression. We then explored the effect of HIF-1α on the regulation of RLIP76 expression at the transcriptional level. We found two putative HREs in the promoter of the RLIP76 gene by using PROMO analysis ( Figure S1A). Subsequent ChIP assays con rmed the interaction of HIF-1α and HRE-containing genomic DNA fragments (P1 and P2) in RLIP76 under hypoxic conditions (Fig. 5E). By constructing a re y luciferase reporter plasmid containing wild-type (WT) or mutant (Mut) HREs in the promoter region, we further explored whether the RLIP76 HREs are necessary for HIF-1α-dependent transcriptional activity. As predicted, hypoxia induced dramatically higher luciferase expression in the WT reporter than in the Mut reporter (Figs. 5F and 5G). Moreover, hypoxia-induced luciferase expression was signi cantly suppressed by HIF-1α inhibition (Figs. 5H and 5I). These results suggest that RLIP76 is a direct transcriptional target of HIF-1α.
Hypoxia-Induced RLIP76 is Critical for the Regulation of Tumorigenesis To investigate the possible correlation between hypoxia-induced RLIP76 and tumorigenesis in vivo, we established a xenograft glioma model. U251 cells were pretreated with hypoxia to activate the HIF-1α/RLIP76 pathway before injection into the anks of nude mice. Downregulation of RLIP76 inhibited the tumorigenicity of U251 cells, as demonstrated by the tumor volume results (Figs. 6A and 6B). BrdUrd incorporation revealed that cell proliferation was dramatically suppressed in RLIP76 knockdown tumors compared with control tumors (Fig. 6C). The apoptotic cells in xenograft gliomas were observed by TUNEL staining, and RLIP76 downregulation resulted in a signi cant increase in RLIP76 knockdown tumors (Fig. 6D). Immunohistochemical (IHC) analysis showed that the expression level of CD31 was signi cantly decreased in tumors derived from RLIP76 knockdown cells compared with those derived from control transfectants (Fig. 6E). In addition, real-time RT-PCR analysis indicated that the expression levels of HIF-1α, LDHA, and GLUT1 were effectively downregulated, whereas the VHL expression level was dramatically upregulated in RLIP76 knockdown xenografts compared with the control (Fig. 6F). These data strongly indicate that RLIP76 enhances cell proliferation and angiogenesis in vivo and suggest that the hypoxia-induced RLIP76 axis plays a critical role in promoting tumorigenesis. We further investigated RLIP76, HIF-1α, and VHL expression in 96 human GBM samples by IHC staining (Fig. 6G). RLIP76-high GBM exhibited considerably higher immunoreactivity for HIF-1α and lower immunoreactivity for VHL expression (Fig. 6H). Spearman's correlation analysis revealed that RLIP76 expression was positively associated with HIF-1α by immunostaining ( Figure S3E). As we expected, a signi cant reverse association was also found between VHL and RLIP76 expression ( Figure S3F). Together, these data demonstrate that hypoxia-induced RLIP76 plays a key role in cellular mechanisms related to glucose metabolism and glioma progression in vivo.

Discussion
RLIP76 is an ATP-dependent transporter that has been implicated in the transportation of anionic metabolites, such as glutathione conjugates of electrophile compounds [29]. In addition to its function as a transporter, RLIP76 also acts as a regulator of the cell response to oxidant injury [30]. Previous studies have reported that Hsf-1 binds to RLIP76, which is required for the response to oxidative stress [31].
Oxidative stress-induced injury may not only induce apoptosis and autophagy, but it may also affect energy metabolism in the cell [32,33]. Energy metabolism is primarily manifested as the tricarboxylic acid cycle, the mitochondrial oxidative respiratory chain, inhibition of oxidative phosphorylation and enhancement of glycolysis. Increased glycolysis effectively enhances the energy supply under conditions of hypoxia-induced oxidative stress injury, which is regarded as a typical tumor adaptive response [34,35]. Here, we identi ed RLIP76 as a hypoxia-inducible molecule that is necessary for hypoxia-enhanced glycolysis, which suggests that non-ABC transporters might contribute to the Warburg effect. [36]. Although there is new interest in the energy metabolism of gliomas, it is generally believed that the metabolic defects of tumor cells and the Warburg effect are mainly due to select genomic mutability during tumor development [3]. Here, we performed bioinformatic analysis on RLIP76 coexpressed genes from the TCGA GBM database and found that RLIP76 plays a crucial role in the reduction of oxidative phosphorylation and ubiquitinmediated proteolysis in GBM. It has been well documented that the ubiquitin-and proteasome-mediated degradation of HIF-1α plays a pivotal role in enhanced glycolysis in GBM [37]. Under hypoxic conditions, HIF-1α can escape ubiquitin-mediated proteolysis by disassociating with VHL, a component of an E3 ubiquitin-protein ligase complex [38]. VHL has been regarded as an important member of the ubiquitinmediated proteolysis pathway during HIF-1α degradation. In the absence of VHL, HIF-1α accumulates, leading to better adaptation to low-nutrient environments and survival in hypoxia [39]. We found that the level of HIF-1α was greatly reduced after RLIP76 knockdown in a xenograft mouse model, whereas the level of VHL was increased. In addition, a positive association was found between HIF-1α and RLIP76, while a negative correlation was found between VHL and RLIP76 in GBM specimens. These results strongly imply that RLIP76 has a close functional relationship with HIF-1α and VHL in vivo and may facilitate glycolysis by regulating them in GBM.

Enhanced glycolysis and decreased oxidative phosphorylation have been viewed as identical features of aberrant metabolism that maintain the viability of tumor cells in gliomas
Hypoxia leads to enhanced expression of HIF-1α, HIF-2α and p53, which are classical oxygen-dependent transcriptional activators [40]. Here, we showed that only HIF-1α downregulation, not HIF-2α or p53 downregulation, signi cantly suppressed hypoxia-induced RLIP76 overexpression in GBM. These results are consistent with previous studies showing that knockdown of RLIP76 can inhibit tumor progression in different human neoplasms of the central nervous system, such as GBMs and neuroblastomas, independent of p53 status [16,41]. These studies, together with ours, suggest that RLIP76 may be fundamental to the HIF-1α-induced signaling pathway and that its relationship to tumor cell energy is pivotal. In addition to the interactions between RLIP76-HIF-1α and RLIP76-VHL, several other hypoxiainduced molecules, such as Hsf-1 and HSP90, have recently been shown to form complexes with RLIP76 to control the hypoxia-induced oxidative stress response [42,43]. Speci cally, a GAP-dependent role for RLIP76 in interacting with Hsf-1 has been reported, as RLIP76 promotes Hsf-1 translocation into the nucleus and therefore functions to suppress Hsf-1 by increasing heat shock protein transcription [42]. In the present study, we further reveal the critical genomic structure of RLIP76 in complex formation under hypoxic conditions in which the GAP function but not the ATPase function of RLIP76 is essential for the RLIP76-HIF-1α and RLIP76-VHL interactions in glioma cells. Intriguingly, we previously found that the ATPase function of RLIP76 is required for the degradation of Rac1 by the ubiquitin-proteolytic pathway under normoxic conditions in glioma cells [16]. All these ndings con rm that RLIP76 is a multifunctional protein that contributes to gene regulation by catalyzing modi cations of various transcription factors to facilitate tumor progression.
The enhancement of glycolysis is a principal characteristic of energy metabolism during glioma development in hypoxic conditions [10]. It has been demonstrated that HIF-1α activation in response to hypoxic stimuli can constitutively promote the expression of glycolytic genes, such as LDHA, GLUT1, HK2 and PKM2, in GBM cells to regulate tumor glycolysis [11,44]. Here, we demonstrated that RLIP76 drives a positive feedback loop to promote the HIF-1α-related signaling pathway by stabilizing HIF-1α, leading to the promotion of hypoxia-induced glycolysis in hypoxic conditions. It has also been shown that the binding of RLIP76 to VHL has an inhibitory effect on the binding of VHL and HIF-1α, which is necessary for the VHL-induced ubiquitin degradation of HIF-1α. This suppression is attributed to the fact that after competitively binding to VHL, which means that RLIP76-VHL complex formation will further prevent HIF-1α from interacting with VHL, RLIP76 could induce HIF-1α activation and ultimately promote glycolysis. Lee et al. previously found that RLIP76 may regulate the function of HIF-1 in tumor cells without affecting HIF-1 expression [45]. Here, we con rmed that upregulation of RLIP76 is capable of restoring RLIP76 knockdown-induced decrease in HIF-1α under hypoxia, while RLIP76 overexpression failed to enhance HIF-1α expression under normoxia, which is consistent with a previous study. Considering the above results, we believe that the suppressive effect of RLIP76 on the HIF-1α-VHL interaction might be part of the mechanism for RLIP76-induced HIF-1α activation in hypoxia.
The HIF-1α signaling pathway is a positive regulator of cancer proliferation, as its suppression often results in tumor suppression [6,38]. It has been reported that HIF-1α is aberrantly expressed and inversely associated with prognosis in GBM patients [46]. Moreover, either suppressing HK2 activity by its inhibitor 3-bromopyruvate or downregulating LDHA by siRNA inhibits the malignant phenotype of glioma cells, suggesting that HIF-1α-responsive genes related to glycolysis could be potential targets for glioma treatment [47,48]. It has been well established that RLIP76 has a close association with several HIF-1αrelated genes, such as VEGFA and AKT, which leads to malignant development of gliomas [45,49]. Here, we reveal that knockdown of hypoxia-induced RLIP76 results in the suppression of U251 cell tumorigenicity by using a xenograft mouse model and demonstrate that RLIP76 may serve as a vital oncogene in GBM by participating in the HIF-1α signaling pathway. These ndings suggest that inhibition of RLIP76 may provide a novel and promising method for glioma therapy.

Conclusions
In conclusion, we describe the interaction of RLIP76, HIF-1α and VHL in the regulation of hypoxiaenhanced glycolysis under hypoxic conditions (Fig. 6I). RLIP76 functions as a hypoxia-responsive gene that is rapidly induced by HIF-1α. Then, RLIP76 promotes the stability of HIF-1α by disrupting the HIF-1α-VHL complex, and by doing so, it leads to increased expression levels of HIF-1α target genes (LDHA and GLUT1) and ultimately enhances glycolysis. Therefore, a better understanding of RLIP76-mediated HIF-1α signaling could contribute to the development of combination therapies and become a potential strategy for GBM treatment in the future. Ethics approval and consent to participate All the patients were informed of sample collection and usage. The tissue samples were collected and used in accordance with approval by the Specialty Committee on Ethics of Biomedicine Research, Shanghai Tongji University (PJ2015-011-08). Use of animal was approved by the Shanghai Tongji hospital animal ethic committee.

Consent for publication
Not applicable.

Availability of data and materials
The data that support the ndings of this study are available from the corresponding author upon reasonable request.

Competing interests
The authors declare that they have no competing interests. Contributions CZ and QW were responsible for designing the study and writing the manuscript. LZ and HRC were responsible for data analysis. JQ was responsible for editing the manuscript. QKB and CL were responsible for re-editing the manuscript. All authors read and approved the nal manuscript.    RLIP76 Interacts with HIF-1α and VHL in a GAP-Dependent Mechanism (A) U251 cells were cultured under hypoxic conditions for 24 h before co-IP assays were conducted. Control IgG and anti-HIF-1α antibody were used in co-IP assays. **, P<0.01. (B) U251 cells were cultured under hypoxic conditions for 24 h before co-IP assays were conducted. Control IgG and anti-VHL antibodies were used in co-IP assays. **, P<0.01. (C) U251 cells were cultured under normoxic or hypoxic conditions for 24 h before GST pull-down assays were conducted. GST-HIF-1α can readily pull-down RLIP76. (D) U251 cells were cultured under normoxic or hypoxic conditions for 24 h. Protein was puri ed by GST antibody-conjugated columns and incubated with U251 cell lysates and then repuri ed through immunoprecipitation and subjected to western blotting. GST-VHL can readily pull-down RLIP76. (E) p3×Flag-HIF-1α and pEGFP-VHL were transfected into U251 cells. Twelve hours after transfection, cells were treated with 5 μM MG132 under hypoxia for another 24 h. Anti-Flag antibody or an isotype-matched control IgG were employed in the rst immunoprecipitation. 3× Flag peptides were used to elute the proteins in the immunoprecipitates. Ten percent of the eluent was used for Western blotting, and another 10% was analyzed by real-time RT-PCR.
The remaining 80% was immunoprecipitated with either control IgG or anti-GFP antibody, and then the immunoprecipitates were washed. Ten percent of this eluent was subjected to Western blotting, and the remaining 90% was analyzed by real-time RT-PCR. ND indicates that RLIP76 was not detectable in the second immunoprecipitates. **, P<0.01. (F) Schematic representation of full-length VHL and its deletion mutants. (G) U251 cells were transfected with full-length VHL or its deletion mutants. Twelve hours after transfection, cells were cultured under hypoxic conditions for an additional 24 h. The puri ed GST or GSTtagged VHL fragments were mixed with the cell lysates of U251 cells expressing Flag-RLIP76 and then subjected to GST-pulldown. (H) Schematic representation of full-length HIF-1α and its deletion mutants.
(I) U251 cells were transfected with full-length HIF-1α or its deletion mutants. Twelve hours after transfection, cells were cultured under hypoxic conditions for an additional 24 h. The puri ed GST or GSTtagged HIF-1α fragments were mixed with the cell lysates of U251 cells expressing Flag-RLIP76 and then subjected to GST-pulldown. (J) Schematic representation of full-length RLIP76. (K) U251 cells were transfected with full-length RLIP76 or its GAP-de cient or ATP-de cient mutants. VHL in the GST pulldown assay was analyzed by western blotting. (L) U251 cells were transfected with full-length RLIP76 or its GAP-de cient/ATP-de cient mutants. HIF-1α in the GST pulldown assay was analyzed by western blotting. (M) U251 cells were transfected with Flag-full-length RLIP76 or its GAP-de cient/ATP-de cient mutants. Twelve hours after transfection, cells were cultured under hypoxia for an additional 24 h. An anti-Flag antibody or an isotype-matched IgG was used to immunoprecipitate with the cell lysates. VHL in the immunoprecipitates was analyzed by western blotting. (N) U251 cells were transfected with Flag-fulllength RLIP76 or its GAP-de cient/ATP-de cient mutants. Twelve hours after transfection, cells were cultured under hypoxia for an additional 24 h. An anti-Flag antibody or an isotype-matched IgG was used to immunoprecipitate with the cell lysates. HIF-1α in the immunoprecipitates was analyzed by western blotting.