The Increased Expression of Estrogen-Related Receptor α Correlates with Wnt5a and Poor Prognosis in Patients with Glioma

This article is featured in Highlights of This Issue, p. 1

Human glioma is the most common primary tumor of the central nervous system in adults, representing an unmet clinical challenge as nearly two thirds of them are high-grade lesions with poor outcome (1–3). Glioma tumors are histologically classified into grades I to IV in the order of increasing malignancy according to the World Health Organization (WHO) criteria (4). Grade IV glioma, accounting for 54% of all human glioma, is the most aggressive and lethal type of brain tumors because of its high invasiveness, robust neovascularization, high relapse after treatments, and recurrence rates (5). Despite recent advances in therapeutic modalities comprising maximal surgical resections, adjuvant chemotherapy, and postoperative radiotherapy, the median survival for patients with glioblastoma (GBM) tumors remains less than 16 months (6, 7). Thus, it is of significant scientific and clinical value for glioma patients to improve clinical prognosis by early diagnosis and new rational therapeutic targets.

The orphan nuclear receptor estrogen-related receptor alpha (ERRα) regulates a range of physiologic and biochemical events together with its certain transcriptional coactivators in a coordinated manner (8–10). The expression pattern of ERRα is characterized by high levels in tissues with increased metabolic demands, such as the heart, skeletal muscles, and adipose tissue (11). Interestingly, there is accumulating evidence supporting that ERRα contributes to tumorigenesis of several types of cancers (12). It has been reported that the high level of ERRα expression correlated with carcinogenesis and poor clinical outcomes in various human tumors, including hormone-dependent breast or ovary cancers and those hormone-independent tumors. Recently, the results from the early clinical trials revealed that a steroidal ERRα inhibitor, SR16388, was effective in patients who have relapsed after the treatment of tamoxifen or an aromatase inhibitor (13). ERRα has been considered to be among the most promising targets for cancer therapy.

It is demonstrated that activation of ERRα in cancer cells results in a significant increase of vascular endothelial growth factor (VEGF) and hypoxia-inducible factor-1 (HIF1) expression (14, 15). However, few studies have been conducted in regard to the role of ERRα in glioma. A research suggests that ERRα is differentially expressed in several human glioma and astrocytoma cell lines (16). Nevertheless, a detailed examination of ERRα expression level and biological function of ERRα in glioma remains unclear. Additionally, its relationship with the pathologic features and the overall survival in human glioma has not been addressed yet.

Wnt proteins are a large family of cysteine-rich secreted molecules (17). Wnt5a, a secreted glycoprotein signal transducer of the Wingless/Int1 (Wnt) family that activate the β-catenin independent pathway (18), has been considered mainly a tumor suppressor (19). Previous studies had revealed higher Wnt5a protein levels and mRNA expression in glioma compared with nonmalignant control brain tissue. It is also reported that Wnt5a was upregulated by ERRα in prostate cancer cells (20). Furthermore, ERRα regulates osteoblast differentiation via Wnt/β-catenin signaling (21). In glioma cells, Wnt5a stimulated the proliferation, motility, invasiveness, and aggressiveness of human glioma cell lines, suggesting the oncogenic property of Wnt5a in human glioma (22–24).

In this study, we show for the first time that ERRα is more broadly expressed in grade III and IV tumors than that in grade I and II brain tissues. Increased ERRα expression levels are correlated with a worse overall survival (OS) in glioma patients. Moreover, the inhibition of ERRα expression impairs proliferation and invasion of GBM cell lines, suggesting that ERRα may be required in the progression of glioma.

A total of 107 paraffin-embedded tissue samples of glioma were collected from patients with detailed clinical information in Huashan Hospital North and Huashan Hospital, Fudan University, Shanghai, China. And 11 normal brain (NB) tissues were obtained from patients who underwent posttrauma surgery for severe traumatic brain injury. The Cancer Genome Atlas (TCGA) mRNA-seq database was downloaded from public databases (https://cancergenome.nih.gov/). This study was approved by the Ethics Committee of Huashan Hospital (2016-229). Written informed consent was obtained from all patients and the examination of tissue samples used in this study.

The clinical pathologic information of glioma was listed in Supplementary Table S1. All specimens were divided into three groups: NB, low-grade glioma (LGG, grades I and II), and high-grade glioma (HGG, grades III and IV).

The human glioma cell lines SNB-19, SF-295, A172, T98G, LN229, and LN18 were purchased from American Type Culture Collection, without cell line authentication. U87MG and U251 were tested on May 22, 2013, and January 7, 2013, respectively (genomic DNA). A172, U87MG, HEK293T, and MDA-MB-231 were cultured in Dulbecco's modified Eagle's medium (DMEM, HyClone), while other cell lines were maintained at minimal essential medium (MEM, Invitrogen). Both media were supplemented with 10% fetal bovine serum (FBS), 1% l-glutamine, 1% sodium pyruvate, and 100 U/mL penicillin/streptomycin (Wisent) at 37°C in a 5% CO₂ humidified incubator. Primary GBM samples were derived from patients undergoing surgery at Huashan Hospital. Tumors were dissociated and grown as adherent cultures on laminin coated dishes (23). All patient-derived cells (GBM001, 002, 003) were cultured in NeuroCult NS-A basal medium, supplemented with NeuroCult NS-A proliferation supplements, 10 ng/mL human bFGF (Sigma) and 20 ng/mL human EGF (PeproTech). Human neurons (HN) were maintained in neuronal medium (ScienCell) supplemented with neuronal growth supplement (ScienCell). HMC3 and human astrocyte (HA) were cultured in EMEM and DMEM with 10% FBS.

The lentiviral shRNA expression vector targeting hERRα (NM_004451; pLKD-CMV-G&PR-U6-shRNAERRα, shERRα) and scrambled control (pLKD-CMV-G&PR-U6-shRNAscramble, shCtl) were produced by Obio Technology). Lentiviruses were obtained by transfecting the shRNA (Supplementary Fig. S1) expression vector targeting hERRα, reducing 85% expression compared with shCtl into HEK293T cells. For transduction, U87MG and U251 cells were seeded at 50% to 60% confluence and infected with viral suspension containing 5 μg/mL polybrene (Sigma-Aldrich) for 6 hours. Two days after infection, cells were selected for puromycin resistance for 48 hours (2 μg/mL, Sigma-Aldrich) and used for experiments within a week (22). For overexpressing, lentiviral particles were produced with the use of a lentivirus packaging mix (Supplementary Methods). U87MG and U251 cells were treated with lentivirus for 48 hours and harvested for examination of gene expression by qRT-PCR and Western blotting (Supplementary Fig. S1).

Total RNA was extracted from the U87MG and U251 cell lines, glioma tissues, and NB tissues using TRIzol kit (Takara), according to the manufacturer's instructions. Relative quantification of the ERRα mRNA level was calculated after normalization to GAPDH mRNA using the 2^−ΔΔCt method. The primer sequences are shown in Supplementary Table S3.

Cells were treated as presented in the figure legends, and cell lysates were prepared as previously described (25). The samples were blotted on PVDF membranes, and membranes were probed overnight at 4°C with anti-ERRα, anti-β-catenin, anti-Wnt5a, c-myc, p-GSK3β, GSK3β, Cyclin D1, Axin1, and anti-β-actin.

For IHC staining, after routine deparaffinization in xylene, rehydration, and inhibition of endogenous peroxidase activity with 3% hydrogen peroxide, sections were exposed to microwave-enhanced antigen retrieval procedure in 0.01 M sodium citrate buffer (pH 6.0). After blocking with 3% BSA, the sections were incubated with rabbit anti-human antibody directed against ERRα (1: 100; Abcam) for 1 hour, followed by biotin-labeled rabbit anti-goat antibody for 0.5 hours. Sections were subsequently incubated with avidin-conjugated horseradish peroxidase (HRP) for 0.5 hours at room temperature.

For IF staining, sections were deparaffinized and dehydrated in gradient alcohol, and then rinsed in phosphate-buffered saline (PBS). The sections were subjected to antigen retrieval by immersion in citric acid antigen retrieval solution and heated in a microwave oven. Then, the sections were costained with the patient serum, and a commercial rabbit polyclonal antibody was directed against ERRα (red) and Wnt5a (green) and treated with FITC-conjugated rabbit polyclonal antibodies (all purchased from Abcam). All images were captured under a fluorescence microscope (Nikon 80i; Nikon).

MTT assay and BrdUrd incorporation assay in U87MG and U251 cells was measured as previously described (26). Analysis of adenovirus mediated cell proliferation was measured by MTT (24 hours) and BrdUrd incorporation assay. HN, HA, HMC3, U87MG, and U251 cells were treated with PBS, shCtl, or shERRα. The inhibitory rates were calculated using the formula in triplicate: inhibitory rates (%) = [(1 − OD_experimental)/ OD_control] × 100%.

For colony formation assay, U87MG and U251 cells transfected with shCtl or shERRα were plated in 6-well culture plates at 100 cells per well. The cells were washed with PBS and stained with crystal violet staining. Colonies were counted only if they contained >50 cells under a microscope.

A total of 1 × 10⁶ U87MG and U251 cells were, respectively, plated in 60-mm culture plates, and the cells were pretreated with shCtl or shERRα. After 48 hours, the cells were trypsinized and then DNA was stained with propidium iodide (PI; Sigma-Aldrich) in the dark. Then, the samples were run on FACScanII flow cytometry (Biosciences). At least 10,000 cells were counted, and the percentages of cells within each phase of the cell cycle were analyzed using ModFit software.

The cell migration assay was performed using a transwell chamber (6.5 mm in diameter, 8-μm pore size; Costar), allowing migrating for 24 hours according to our previous study (26). Analysis of adenovirus-mediated cell was measured by transwell migration assay. HN, HA, HMC3, U87MG, and U251 cells were treated with PBS, shCtl o,r shERRα.

U87MG cells were transiently cotransfected with TCF/LEF1-Luc, ERRα, or PGC-1α plasmids and then treated with shERRα for the indicated times. The pSV-β-galactosidase plasmid (pCH110) was cotransfected and used for data normalization. Subsequently, luciferase activity was measured according to the manufacturer's protocol, and expressed as the ratio of luciferase activity to β-galactosidase. For measuring the transcriptional activity of Wnt5a, the pGL3-Basic-Wnt5a-luc, which contains Wnt5a promoter with the sequence −2869/+170, was transfected into U87MG cells.

Co-IP assays were performed as previously described (27). Briefly, whole-cell extracts were prepared using nondenaturing lysis buffer. The proteins were immunoprecipitated using antibodies to ERRα, β-catenin, mouse IgG, and protein-A/G PLUS-Agarose beads, then separated by 10% SDS-PAGE and detected by Western blotting.

ChIP assays were performed on U87MG as described previously using primer sequences listed in Supplementary Table S3. Briefly, soluble chromatin fragments were immunoprecipitated using antibody to ERRα and then treated with RNase A and proteinase K. Isolated DNA fragments were purified with QIAquick spin kit, and used as templates to detect the putative ERR response elements (ERRE; Supplementary Table S4) by quantitative qRT-PCR.

Briefly, male BALB/c nu/nu mice (4–6-week-old, 19 g) were anesthetized with an IP injection of ketamine (75 mg/kg) and medetomidine (0.5 mg/kg), placed in a stereotactic apparatus, and a hole was drilled in the skull for cell implantation. Then, 3 × 10⁵ U87MG cells in 2 μL were injected into the right striatum (+0.5 mm AP, +2.1 mm ML, −2.9, −3.2, −3.5, −3.8, −4.1 mm DV from bregma: 1 μL per site) using a 10 μL Hamilton syringe with a 33-gauge needle. The needle was left in place for 3 minutes prior to removal to allow tumor cells to settle at the injection site. Five days after tumor implantation, mice received an intratumoral injection of either 3 μL PBS, inactivated-adenovirus (Ad-Ctl) or Ad-ERRα-shRNA solution: 8 mice received PBS, 8 mice received Ad-Ctl at 1 × 10⁸ pfu/mouse, and 8 mice received Ad-ERRα-shRNA at 1 × 10⁸ pfu/mouse. Treatments were administered using the same hole (at −2.9, −3.2, −3.5, −3.8, −4.1 mm DV from bregma: 1 μL per site). Mice were euthanized when moribund or when survived 71 days after tumor implantation. Meanwhile, the mouse OS curves were recorded according to the Kaplan–Meier method. Tumor volume was calculated by using a modified ellipsoidal formula: tumor volume = (L × W²) × 0.5, where V is the tumor volume, L is the longitudinal diameter, and W is the transverse diameter (28, 29). The brains of the mice were extracted and fixed in 10% formalin. Next, they were embedded in paraffin for IHC of CD68 and myelin basic protein (MBP), as markers of macrophages/microglia and an index of oligodendrocyte integrity, respectively. Nissl staining was used to determine the histopathologic features of the brains. Brain sections (5 μm) were mounted on gelatinized glass slides and incubated in cresyl violet (0.1%; Sigma) and were passed through destain solution (70% ethanol, 10% acetic acid) and dehydrated (100% ethanol and xylene). Then, tissues were photographed. Human glioma xenografts were generated according to supplementary methods.

All experiments were performed according to the institutional ethical guidelines on animal care and approved by the Institute Animal Care and Use Committee of Fudan University (approval number: SOP-ICE-033-01.0-AF05).

Statistical significance of the observed differences was addressed by statistical tests using GraphPad Prism V7.0. Data are presented as mean ± SEM. Each experiment was repeated a minimum of 3 times. Survival curves were plotted using the Kaplan–Meier method, and statistical significance was determined by the log-rank (Mantel–Cox) test. The Cox proportional hazard model was used for calculating the impact of ERRα expression in the univariate and multivariate analysis of variables and patient's OS duration. P value of < 0.05 was considered to be statistically significant (*, P = 0.05; **, P = 0.01; ***, P = 0.001).

To assess the role of ERRα in glioma, we first measured the expression of ERRα mRNA transcripts using qRT-PCR in 34 freshly collected glioma tissues and 11 NB tissues. Compared with the NB group (1.017 ± 0.193), the LGG and HGG obtained a much higher ERRα mRNA expression level, which were 3.908 ± 0.668 (P < 0.01) and 6.46 ± 0.684 (P < 0.001), respectively (Fig. 1A). In addition, ERRα protein was also found to be upregulated in 6 cases of glioma compared with NB tissues (Fig. 1B).

We then examined the protein level of ERRα in 107 achieved paraffin-embedded glioma samples (Supplementary Tables S1 and S2) by IHC. A semiquantitative ERRα expression was compiled using intensity of staining as well as the number of stained cells. We observed that protein level of ERRα was markedly higher in glioma tissues than that in NB tissues (Fig. 1C). Meanwhile, ERRα expression was significantly high as the glioma pathologic grade increased. According to the ERRα staining score, 58 (54.21%) patients were classified as low ERRα group and 49 (45.79%) patients were classified as high ERRα group (Supplementary Table S2). This result was well validated in TCGA RNA-seq data (Fig. 1D). We analyzed the RNA-sequencing data of glioma from the TCGA database and found that ERRα was enriched in HGG.

To further investigate the potential of ERRα expression to predict patients’ clinical outcomes in glioma, we assessed the association between ERRα expression and patients’ survival using the Kaplan–Meier method. The primary endpoint was OS, defined as the time of completion surgery to the date of death or the last follow-up contact. We also studied progression-free survival (PFS), defined as the date of completion surgery to tumor recurrence, death from any cause, or last follow-up contact. The OS and PFS curves were summarized in Fig. 1E and F.

Highly statistically significant correlations were also observed between OS/PFS and the levels of ERRα immunoreactivity (Fig. 1E and F; P < 0.01). The increased ERRα immunoreactivity in the HGG group was markedly associated with shortened OS and PFS. We further analyzed the prognostic value of ERRα in a total of 551 glioma patients in the TCGA database. As shown in Fig. 1G, high expression of ERRα predicted significantly poor OS. Taken together, these results indicated that ERRα expression might affect both PFS and OS of glioma patients.

Because ERRα expression was associated with the pathologic grade of glioma, we investigated the prognostic significance of ERRα for glioma patients. After univariate analysis by Cox proportional hazards model, KPS score (P = 0.031, HR = 1.229), necrosis (P = 0.026, HR = 0.726), WHO grade (P = 0.002, HR = 4.298) and ERRα immunoreactivity (P = 0.018, HR = 3.573) were turned out to be significant prognostic clinical parameters for OS in 107 glioma patients (Table 1).

Additionally, we performed a stepwise multivariate analysis adjusted for the same parameters to determine whether ERRα immunoreactivity is an independent prognostic factor for glioma patients. The results reflected that, after the correction for patient KPS score (P = 0.095) and necrosis (P = 0.224), the elevated expression of ERRα (P = 0.021, HR = 1.135) and WHO grade (P = 0.031, HR = 3.652) is associated with a shortened OS and PFS, acting as an independent prognostic factor with a relative risk over 1.0 for patients.

To understand the significance of ERRα expression in glioma, we measured ERRα expression in a panel of human glioma cell lines using Western blotting (Fig. 2A). Among the glioma cell lines tested, protein levels of ERRα revealed higher in the grade IV–derived glioma cell lines U251, U87MG, and A172, which was in accordance with a previous study (16).

We therefore assessed whether the downregulation of ERRα by its specific inverse agonist XCT790 contributed to the reduction of proliferation in 5 glioma cell lines. Downregulation of ERRα with XCT790 selectively decreased the proliferation of glioma cell lines that have high level of ERRα expression (Fig. 2B). The estrogen receptor–negative MDA-MB-231 human breast cancer cell acted as a positive control. The LN229 cells did not respond to XCT790, consistent with their lack of ERRα expression. The regression analysis indicated that there was a statistically strong correlation between inhibition rates of XCT790 and ERRα expression (Supplementary Fig. S2).

We then performed knockdown assays by using shERRα in HNs, HA, human microglia cell (HMC3), and GBM patient-derived primary cells (GBM001, GBM002, and GBM003), respectively. Silencing of ERRα slightly reduced the HA and HMC3 cell growth compared with glioma cells (Fig. 2C) in the MTT assay. In contrast, as shown in Fig. 2D, ERRα depletion decreased proliferation of GBM patient-derived cells. We also found that knockdown of ERRα decreased proliferation of U87MG and U251 after transfection of shERRα at 48 and 72 hours (Fig. 2E). On the contrary, overexpression of ERRα remarkably promoted cell proliferation in U87MG and U251 cells at 24, 48, and 72 hours after transfection of lenti-ERRα (Fig. 2F; Supplementary Fig. S3). The colony formation assay showed that the colony formation rates of U87MG and U251cells transfected with shERRα were lower than in the same cells transfected with the scramble (Fig. 2G). In addition, we measured the effect of ERRα knockdown on the glioma cell cycle. The data presented in Fig. 2H showed that reduced ERRα expression delayed the progression of the cell cycle and inhibited cell proliferation by arresting cells at G₀–G₁ phase.

As shown in Fig. 2I, overexpression of ERRα after the transfection of lenti-ERRα increased the cell migration of the corresponding cells. Cell migration was decreased in the shERRα group (Fig. 2J, P < 0.05). These results indicated ERRα promoted tumor cell growth and metastasis, functioning as an unfavorable factor in glioma.

To explore the effects of ERRα on the transcriptional activity of β-catenin, U87MG cells were cotransfected with TCF/LEF1-Luc and GFP, PGC-1α, or ERRα vectors. It is noted that both PGC-1α and ERRα significantly enhanced transcriptional activity of β-catenin upon ERRα overexpression compared with GFP in U87MG cells (Fig. 3A and B, P < 0.05). As expected, when U87MG cells were treated with PGC-1α and shERRα, PGC-1α/ERRα-evoked transactivation of Wnt/β-catenin signaling pathway was markedly suppressed (Fig. 3C). These results were further ascertained by co-IP of endogenous ERRα and β-catenin from U87MG cell extracts followed by Western blotting (Fig. 3D). Following the observation of modulation, we detected the transcription of a panel of the Wnt pathway to gain a detailed understanding of which Wnt gene was changed by ERRα inactivation. After the transfection of shERRα or shCtl in U87MG cells, qRT-PCR was performed to quantify mRNA of Wnt. Interestingly, we found a significant downregulation of Wnt5a mRNA in U87MG cells compared with shCtl treatment (Fig. 3E). Thus, we have confirmed that Wnt/β-catenin and ERRα signaling pathway converge, a finding of high significance in carcinogenesis.

To detect the association of ERRα with Wnt5a expression in vivo, we checked the expression of ERRα and Wnt5a in 77 human glioma tissues by immunofluorescence assays (Fig. 3F). Of the 75 glioma cases, 47 (61.03%) and 49 (63.64%) had elevated ERRα and Wnt5a, respectively (Fig. 3G). The analysis showed that scores for ERRα and Wnt5a immunostaining were significantly correlated in tissues tested (P < 0.05). This result suggested ERRα expression is positively correlated with Wnt5a in clinical glioma.

To further decipher the mechanism involved in the role of ERRα, we examined the relative protein expression level of Wnt5a in U87MG cells treated by XCT790 or shERRα. After treatment of XCT790 or transfection with shERRα, the expression of ERRα and Wnt5a in U87MG was significantly decreased, respectively (Fig. 4A). To ascertain whether Wnt5a was directly regulated by ERRα, we analyzed the human Wnt5a promoter sequence in silico to detect potential ERRE composed of the extended ERE half-site AAGGT. Four potential ERREs were found located within the proximal promoter of Wnt5a (Fig. 4B). Then, we transfected the Wnt5a promoter reporter gene plasmid (pGL3-Basic-Wnt5a-luc) into U87MG to evaluate the effect of ERRα on Wnt5a promoter activity. The data showed that the activity of pGL3-Basic-Wnt5a-luc was significantly decreased by treatment with shERRα (Fig. 4C) or XCT790 (Fig. 4D).

A ChIP assay was performed to examine whether ERRα binds to Wnt5a promoter in the native chromatin environment of glioma cells. ChIP with anti-ERRα in U87MG cells showed that knockdown of ERRα resulted in a significant decrease in the genomic fragment that contained the ERRE-1 and ERRE-3 sites in the Wnt5a promoter (Fig. 4E). Together, these data suggested that ERRα can bind to the cis-regulatory domain of the endogenous Wnt5a promoter and thus regulates the transcription of Wnt5a.

To elucidate whether ERRα influences glioma growth and migration through regulating the expression of Wnt5a, we used Wnt5a protein (500 ng/mL) to perform the rescue experiments. The data showed that inhibitory effects on cell migration (Fig. 4F) and growth (Fig. 4G) caused by shERRα were both partly abolished by Wnt5a.

To investigate the possible involvement of ERRα/Wnt signaling in glioma, we used U87MG cells. shRNA were used to target Wnt5a and β-catenin, an essential component of the canonical Wnt signaling pathway. Depletion of Wnt5a and β-catenin reversed the effects of ERRα overexpression on cell proliferation of U87MG (P < 0.01; Fig. 4H and I). The results suggest a role for the Wnt pathway in regulating glioma cell proliferation.

To explore the mechanism of ERRα/Wnt in U87MG, β-catenin and target genes (c-myc and cyclin D1) expression in the Wnt/β-catenin pathway were examined by Western blotting in U87MG transfected with shERRα. The results showed that knockdown of ERRα and Wnt5a significantly decreased levels of these genes (Fig. 4J and K). We also found that Wnt5a suppression decreased the expression of Axin1, β-catenin, and Dvl2 as well as the ratio of p-GSK3β/GSK3β in U87MG (Fig. 4L). Moreover, overexpression of Wnt5a could rescue the decreased expression of above genes, which was induced by ERRα depletion (Fig. 4M). Our data indicate that ERRα suppression inhibits cell proliferation by suppressing the WNT/β-catenin pathway.

To determine whether Ad-ERRα-shRNA affects neuronal, glial, and immune cells as well as glioma cells in the brain, we performed the MTT assay. We found that Ad-ERRα-shRNA slightly reduced HA growth compared with U87MG and U251 (Fig. 5A), in which cell proliferation was obviously inhibited. The result in the MTT assay is validated by the BrdUrd Assay (Fig. 5B). Moreover, Ad-ERRα-shRNA led to tumor regression and long-term survival in over 71% of the animals and increased the median survival from 25 days and 28 days in the PBS and Ad-Ctl groups to 45 days (Fig. 5C). Ad-ERRα-shRNA can also inhibit the growth of glioma (Fig. 5D). Neuropathological analysis of the brains from long-term survivors showed complete tumor regression and did not reveal signs of neurotoxicity (Fig. 5E). We then verified that reduced expression of ERRα, Wnt5a, Axin1, and Dvl2 in the Ad-ERRα-shRNA group compared with Ad-Ctl (Fig. 5F). Here, we show that ERRa may function as a molecule that is associated with the Wnt5a signaling pathway, leading to the degradation of the canonical β-catenin signaling pathway and the consequent suppression of its downstream genes, such as Axin and GSK3β, both in vitro and in vivo, which promotes glioma cell proliferation, cell cycle, and migration (Fig. 5G). In summary, a single intratumoral injection of Ad-ERRα-shRNA resulted in significant improvement in survival and no sign of neurotoxicity in intratumoral glioma xenografts. These data suggest that ERRα may be a promising therapeutic agent for glioma. It is consistent with the results in the subcutaneous xenograft model (Supplementary Fig. S4).

To our knowledge, this is the first study that applied clinical glioma tissues to elucidate the clinical signature and biological function of ERRα and Wnt5a in glioma. We showed that ERRα promoted cell proliferation and migration in human glioma cells. More importantly, ERRα was identified for the first time as a critical marker not only for tumor WHO classification but also for clinical outcomes in glioma patients.

An abnormal Wnt pathway is associated with many human diseases, including cancer and degenerative diseases (30). Wnt signaling via β-catenin has a recognized role in controlling the proliferation of cells in human cancers (31, 32). The role of Wnt5a in cancer is dual and controversial in various cancer types. Previous reports indicate that Wnt5a inhibits cell proliferation, invasiveness, and aggressiveness in some cancer cells, acting as a tumor suppressor (33, 34). In contrast, it has been suggested that Wnt5a has oncogenic properties. For instance, Wnt5a promotes invasion and migration in breast, gastric cancer cells (35) and melanoma (36). A report indicated that Wnt5a is highly expressed in high grades of glioma, and its expression is correlated with invasive activities (23). Taken together, these findings demonstrate that the role of Wnt5a is dependent on the cell type.

Wnt5a worked as an essential player in maintaining the proliferative and migratory capacity of glioma cells. Our data are also consistent with previous findings that the protein expression of Wnt5a was higher in the HGG group compared with the LGG group (22, 23, 37). Before this study, direct evidence for the association between ERRα and Wnt5a was scant. We hypothesized that knockdown of ERRα decreased malignant behavior of glioma cells through the Wnt/β-catenin pathway. We detected expressions of β-catenin and downstream target genes (c-myc and cyclin D1) of the Wnt/β-catenin pathway (38), and our data showed that significant reductions in β-catenin, c-myc, and cyclin D1 were observed in ERRα-knockdown glioma cells. The data also demonstrated that inactivation of the Wnt/β-catenin pathway reversed the effects of ERRα overexpression on cell proliferation, showing that canonical and noncanoncial Wnt signaling pathway is required to maintain the proliferative capacity of glioma cells. These data suggested that knockdown of ERRα inhibited the proliferation and migration of glioma cells through suppressing the Wnt/β-catenin pathway.

We provided compelling data for the involvement of ERRα in the carcinogenesis via the regulation of Wnt5a signaling pathway in vitro and in vivo. Further, this study confirmed that Wnt5a acts as a key molecule in the biological functions of ERRα in human glioma. The versatile role of Wnt/β-catenin signaling in controlling tumor cell properties such as motility, invasion, and clonogenicity has attracted significant attention of researchers for therapeutic strategies targeting this pathway (39, 40).

In our work, ERRα, as an upstream molecule of Wnt5a, may provide powerful target candidate for the treatment of glioma. Therefore, the identification of Wnt5a as a direct transcriptional target of ERRα was of great significance as the expression of this protein has previously been correlated with increased cell proliferation and migration (37).

In conclusion, our study demonstrates on the first attempt that ERRα is a potent independent prognostic factor to predict patients’ cancer-specific outcomes and guide personalized treatment decisions. Increased ERRα expression indicated worse prognosis and correlated with short survival time. As a consequence, understanding the regulation of ERRα/Wnt5a activity in glioma could lead to the identification of a reliable therapeutic strategy for the treatment of glioma and warrants investigation to improve patient survival.

No potential conflicts of interest were disclosed.

Conception and design: L. Zhang, Y. Zhu, W. Hua, Q. Li

Development of methodology: L. Zhang, Y. Zhu, H. Chen, L. Chen, H. Qi, G. Ren, J. Tang, W. Hua

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Zhu, Y. Zhu, H. Chen, G. Ren, J. Tang, W. Hua

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Zhu, Y. Zhu, H. Chen, L. Chen, H. Qi, G. Ren, J. Tang, X. Shi

Writing, review, and/or revision of the manuscript: L. Zhang, Y. Zhu, H. Cheng, W. Hua, Q. Li

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Cheng, J. Zhang

Study supervision: M. Zhong, W. Hua, Q. Li

The authors thank Drs Yu Zhang, Jinsen Zhang, and Li Chen for valuable discussions.

This study was supported by a grant from the National Natural Science Foundation of China (No.81302853, Q. Li).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.