E3 ligase MAEA-mediated ubiquitination and degradation of PHD3 promotes glioblastoma progression

Glioblastoma (GBM) is the most common malignant glioma, with a high recurrence rate and a poor prognosis. However, the molecular mechanism behind the malignant progression of GBM is still unclear. In the present study, through the tandem mass tag (TMT)-based quantitative proteomic analysis of clinical primary and recurrent glioma samples, we identified that aberrant E3 ligase MAEA was expressed in recurrent samples. The results of bioinformatics analysis showed that the high expression of MAEA was related to the recurrence and poor prognosis of glioma and GBM. Functional studies showed that MAEA could promote proliferation, invasion, stemness and temozolomide (TMZ) resistance. Mechanistically, the data indicated that MAEA targeted prolyl hydroxylase domain 3 (PHD3) K159 to promote its K48-linked polyubiquitination and degradation, thus enhancing the stability of HIF-1α, thereby promoting the stemness and TMZ resistance of GBM cells through upregulating CD133. The in vivo experiments further confirmed that knocking down MAEA could inhibit the growth of GBM xenograft tumors. In summary, MAEA enhances the expression of HIF-1α/CD133 through the degradation of PHD3 and promotes the malignant progression of GBM.


INTRODUCTION
Glioma is a common primary brain tumor, according to their different histopathological characteristics, the World Health Organization (WHO) classifies gliomas into four grades, among which grade III and IV gliomas are high-grade gliomas [1]. Glioblastoma (GBM) is the most frequent high-grade glioma, which can develop rapidly within weeks or months [2]. At present, the primary treatment methods of GBM are surgery, radiotherapy, temozolomide (TMZ) and other drugs, but the average clinical survival time is only approximately 15 months [3]. Due to the heterogeneity and resistance to radiotherapy and chemotherapy, GBM has a very high recurrence rate [4]. The average survival time of recurrent GBM patients is only 6-10 months [5]. Therefore, studying the mechanism of GBM's therapeutic resistance and recurrence is an urgent task to be addressed in the clinic.
Stemness plays a crucial role in tumor progression, including tumorigenesis, metastasis, therapeutic resistance and recurrence [6]. It has been reported that MiR-7-5p inhibits the stemness of GBM cells and enhances the sensitivity of GBM cells to TMZ by targeting yinyang1 (YY1) [7]. Glioma cells with high expression of CD133, which is a cellular surface protein that has been reported to be a cancer stem cell marker, were more tolerant to TMZ [8]. Serpin family A member 3 (SERPINA3) promotes the invasion and stemness of glioma cells and is related to recurrence and poor prognosis [9]. Therefore, research on GBM treatment resistance and recurrence from the perspective of tumor stemness has been one of the main directions in recent years.
Ubiquitination modification of proteins is one of the most important mechanisms in controlling protein abundance and intracellular homeostasis [10]. The E3 ligase determines the specific recognition of target proteins [11] and plays an important role in various tumors by regulating the stability of tumor promoters or suppressors [12]. For example, E3 ligase HECW1 promotes the metastasis of non-small-cell lung cancer cells by mediating the ubiquitination degradation of Smad4 [13]. The ubiquitination degradation of IκBα by TRIM22 activates NF-κB to promote the progression of GBM [14]. Studies have shown that E3 ligase is involved in tumor cell stemness. E3 ligase STUB1 can target ubiquitination transglutaminase 2 (GLS2) to promote its degradation, thereby reducing the stemness and tumorigenicity of oral cancer cells [15]. UBR5 inhibits the proliferation and stemness of esophageal cancer via the ubiquitination degradation of SOX2 at lysine 115 [16]. OTUB1 promotes the stemness of glioma cells by stabilizing the SLC7A11 protein and inhibiting ferroptosis [17]. However, the functions of other E3 ligase and their relationships with tumors are unclear.
E3 ligase Macrophage-Erythroblast Attacher (MAEA) is one of the core components of the CTLH E3 ubiquitin protein complex [18]. MAEA is involved in the proliferation of normal cells; the knockout of MAEA in mammalian normal retinal pigmen epithelium (RPE) cells can significantly inhibit cell proliferation [19]. In addition, MAEA can promote the autophagy of hematopoietic stem cells and maintain stemness [20]. Recently, a study identified a series of differentially expressed proteins from the peripheral blood samples of pancreatic ductal adenocarcinoma (PDAC) patients in both early recurrence and later recurrence groups, and MAEA was one of the most highly expressed molecules in the early recurrence group [21]. These works suggest that the disorder of MAEA leads to abnormal cell proliferation and stemness, and it may be related to tumor recurrence. However, studies on the function and mechanism of MAEA in tumor progression have not been reported to date.
In the present study, MAEA was identified as a novel molecule associated with the recurrence and malignant progression of GBM. MAEA could target K159 of PHD3 by K48-linked polyubiquitination to promote its degradation, thus enhancing the stability of HIF-1α and the expression of CD133, resulting in an improvement in the stemness and TMZ resistance of GBM cells.

RESULTS
MAEA is highly expressed in glioma and is associated with glioma/GBM recurrence and poor prognosis To identify the key molecules involved in the malignant progression of GBM, tandem mass tag (TMT)-based quantitative proteomic analysis was performed on 6 pairs of clinical tissue samples from primary and corresponding recurrent gliomas. Differentially expressed proteins between two groups were determined and the expression levels of 84 proteins were significantly altered. Among them, 44 proteins were upregulated and 40 proteins were downregulated (recurrent/primary fold change >1.2 or <0.83, p < 0.05) (Fig. 1A). A heatmap of proteins was generated from proteomics data (Fig. 1B). To further elucidate the potential role of differentially expressed proteins in glioma carcinogenesis and recurrence, GO molecular functions analysis was performed and it demonstrated the differential proteins significantly participated in molecular function regulator and enzyme regulator activity (Fig. 1C). Subsequently, top 15 upregulated proteins in recurrent samples (recurrent/primary fold change >1.5, Fig. S1A) were selected to perform the association analysis with standard glioma prognostic marker (methylation status of MGMT locus) [22] by using CGGA database. The results showed that four molecules (NOLC1, FAM192A, RBM5 and MAEA) were related to methylated MGMT, suggesting their potential correlation with disease recurrence and resistance to TMZ in glioma (Fig. S1B). Further, the survival analysis of the four molecules indicated that only highly expressed MAEA was positively correlated with poor prognostic (Fig. 1D, Fig. S1C). However, the function and mechanism of MAEA in malignant tumors have not been reported up to now. Therefore, we choose MAEA as the main research object in present study. According to the clinical characteristics in the public database CGGA, 325 clinical glioma samples were divided into primary gliomas (233 cases) and recurrent gliomas (92 cases). The data also showed that there were 139 cases of GBM among these 325 samples, including 54 cases of recurrent GBM. The survival analysis was conducted according to the level of MAEA expression. The results showed that the high expression of MAEA in primary and recurrent gliomas led to a poor prognosis, and similar results were observed in primary and recurrent GBM (Fig. 1D). Moreover, Differential expression analysis was performed and the results showed that, compared with normal tissue, MAEA was also highly expressed in glioma and GBM, and the MAEA expression in recurrent gliomas and GBM was higher than that in primary tissue (Fig. 1E). MAEA expression was further validated using the above 6 pairs of clinical samples via immunohistochemistry (IHC). MAEA protein expression was significantly enhanced in recurrent gliomas compared with the corresponding primary tumor tissues (Fig. 1F). Western blot and RT-qPCR assays also showed that, compared with normal glial cell HEB, the protein (Fig. 1G) and mRNA (Fig. 1H) levels of MAEA were higher in glioma cells. The above results suggest that MAEA is highly expressed in glioma cells, and its expression is related to the recurrence and poor prognosis of glioma/GBM.

MAEA promotes the malignant progression of GBM cells
To explore the role of MAEA in the progression of GBM cells, SF295 and T98G stably knocking down MAEA cells (SF295-shMAEA #1, SF295-shMAEA #2, T98G-shMAEA #1, T98G-shMAEA #2) and control cells SF295-NC and T98G-NC, and LN229 stably overexpressing MAEA cells (LN229-MAEA-OE) and control cell LN229vec, were established (Fig. S2). CCK8 and colony formation assays were performed and the data showed that the knockdown of MAEA in SF295 and T98G cells could inhibit cell proliferation, while the overexpression of MAEA in LN229 cells could promote cell proliferation ( Fig. 2A, B). The results of scratch and Transwell assays showed that the knockdown of MAEA could suppress the migration and invasion of cells; on the contrary, the overexpression of MAEA could increase the migration and invasion of cells (Fig. 2C, D). Furthermore, the data indicated that knocking down MAEA suppressed the expression of the stemness marker CD133, and Nestin and Nanog and sphere formation, while the overexpression of MAEA yielded the opposite results (Fig. 2E, F). The flow cytometry results showed that MAEA could increase the rate of CD133(+) cells (Fig. 2G). Moreover, the results of the CCK8 assay and colony formation assays showed that knocking down MAEA decreased the TMZ resistance of GBM cells, and the overexpression of MAEA could promote TMZ resistance (Fig. 2H,  Fig. S3). These results indicated that MAEA could promote the proliferation, invasion, stemness and TMZ resistance of GBM cells.

MAEA targets PHD3 for ubiquitination and degradation at K159
To clarify the molecular mechanism of MAEA in regulating GBM progression, the proteins interacting with MAEA were detected by immunoprecipitation (IP) and mass spectrometry (Fig. 3A), and multiple molecules interacting with MAEA were identified. Among them, the most prevalent were KHDRBS2 and PHD3, with high values. After the co-transfection of His-KHDRBS2 and Flag-MAEA plasmids in 293 T cells, the IP results showed that there was no interaction between them (Fig. S4). On the contrary, the interaction between MAEA and PHD3 was confirmed by IP in 293 T cells with the co-transfection of Flag-MAEA and His-PHD3 plasmids (Fig. 3B). Subsequently, the endogenous interactions between MAEA and PHD3 in SF295, T98G and LN229 cells were verified by IP as well (Fig. 3C). Immunofluorescence staining also indicated that MAEA and PHD3 were co-located in LN229 and SF295 cells (Fig. 3D). The results of cell fractionation experiments showed that MAEA was mainly enriched in the cytoplasm and cell membrane, while PHD3 was mainly enriched in the cytoplasm (Fig. 3E). The above results demonstrate that MAEA interacts with PHD3 in the cytoplasm.
Further, the protein stability assay using cycloheximide (CHX), a de novo protein synthesis inhibitor, showed that knocking down MAEA could increase the half-life of the PHD3 protein, and the overexpression of MAEA could reduce the stability of PHD3 (Fig. 4A, B). In addition, the proteasomal inhibitor MG132 could suppress the decrease in PHD3 that was induced by MAEA overexpression, and the upregulation of PHD3 was dramatically reversed in MAEA knockdown GBM cells treated with MG132 (Fig. 4C). These results suggest that the stability of PHD3 may be mediated by the ubiquitin-proteasome pathway. To clarify whether MAEA regulates PHD3 expression through ubiquitination modification, the potential ubiquitin binding sites of PHD3 at positions 159 and 172 were analyzed using the smart database, and the conservatism of potential ubiquitin binding sites was analyzed via the UniProt database (Fig. 4D). Then, the results of an IP assay showed that MAEA could increase the PHD3 ubiquitination in GBM cell lines (Fig. 4E). In addition, the same phenomena were observed in 293 T cells after the exogenous expression of Flag-MAEA, His-PHD3 and ubiquitin plasmids (Fig. 4F). In order to identify the ubiquitin modification site of PHD3, we generated mutant plasmids of His-PHD3-K159R and His-PHD3-K172R, after co-transfection into 293 T cells, the results showed that PHD3 ubiquitination was downregulated after mutation lysine 159 of PHD3, but there was no significant change after mutation lysine 172 (Fig. 4G). Studies have shown that there are many types of ubiquitination involved in the regulation of target proteins, among which K48-linked polyubiquitination is mainly involved in protein degradation, while K63-linked polyubiquitination is predominantly associated with protein stability and signal transduction [23]. Thus, 293 T cells were, respectively, transfected with the HA-tagged K48R or K63R ubiquitin mutant plasmids to determine the ubiquitination type. The data showed that the MAEA-mediated ubiquitination of PHD3 disappeared after the K48 mutation on ubiquitin, while the K63 mutation did not change (Fig. 4H). The MAEA is in complex with the RMND5A E3 ligase that constitutes the CTLH E3 ligase complex, and both are needed to maintain the enzymatic activity of this complex [19]. Therefore, we explored whether RMND5A is involved in ubiquitination regulation of MAEA on PHD3 protein. The data demonstrated that the knockdown of RMND5A in SF295 and LN229 cells could inhibit the ubiquitination of PHD3 (Fig. 4I), and the knockdown of RMND5A in overexpressing MAEA cells could suppress the ubiquitination of PHD3 mediated by MAEA (Fig. 4J). These results suggest that the MAEA-mediated ubiquitination and degradation of PHD3 at K159 through K48-linked polyubiquitination, and RMND5A is involved in this process.
PHD3 is involved in MAEA-mediated proliferation, invasion, stemness and TMZ resistance of GBM cells In order to determine whether PHD3 was involved in the tumor progression regulated by MAEA, functional rescue experiments were employed, through the transfection of the shPHD3#1 or shPHD3#2 plasmids in stable SF295-shMAEA cells, or transfected His-PHD3 in LN229-MAEA-OE cells (Fig. S5). The results of CCK8 and colony formation assays showed that the overexpression of PHD3 could inhibit the MAEA-mediated proliferation of GBM cells, while knocking down PHD3 could compensate for the decrease in proliferation caused by MAEA knockdown (Fig. 5A, B). Further, the data of scratch and Transwell assays also indicated that MAEA affected cell migration and invasion through its regulation of PHD3 (Fig. 5C, D). Moreover, the results of a sphere formation assay and flow cytometry demonstrated that MAEA required PHD3 in its regulation of stemness (Fig. 5E, F). The results of the colony formation assay showed that PHD3 overexpression could reduce the MAEA-mediated TMZ resistance, while knocking down PHD3 could rescue the TMZ resistance of GBM cells inhibited by MAEA deletion (Fig. 5G). In addition, we also analyzed the biological function of RMND5A in GBM cells. CCK8, Transwell, sphere forming and plate cloning experiment results showed that knockdown of RMND5A could inhibit MAEA-mediated proliferation, invasion, stemness and TMZ resistance (Fig. S6A-D). These results suggest that MAEA promotes the proliferation, invasion, stemness and TMZ resistance of GBM cells mediated by PHD3.
MAEA regulates the expression of HIF-1α and CD133 by degrading PHD3 It has been reported that PHD3 can reduce the stability of transcription factor HIF-1α [24]. In addition, HIF-1α has been confirmed to be positively correlated with the mRNA and protein expression of CD133 in renal cell carcinoma [25] and hepatocellular carcinoma cells [26]. To verify that MAEA regulates HIF-1α and CD133 by degrading PHD3, western blot experiments were performed and the results indicated that knocking down MAEA could decrease the expression of HIF-1α and CD133 in GBM cells, while the overexpression of MAEA could increase both molecules' protein expression (Fig. 6A). Further results demonstrated that MAEA required PHD3 in its regulation of HIF-1α and CD133 expression (Fig. 6B). IHC analysis of 56 clinical glioma samples showed that the expression of HIF-1α and CD133 was lower and the expression of PHD3 was higher in the group with low expression of MAEA, while the opposite results were found in the group with high expression of MAEA (Fig. 6C). Pearson correlation analysis based on the IHC score indicated that there was a positive correlation between MAEA and HIF-1α or CD133, and a negative correlation between PHD3 and MAEA, HIF-1α or CD133 (Fig. 6D). These data suggest that MAEA can upregulate the expression of HIF-1α and CD133 via the degradation of PHD3.
MAEA promotes tumor growth and TMZ resistance in vivo, and there is a correlation among the expression of MAEA, PHD3, HIF-1α and CD133 in tumor tissue After SF295-shMAEA cells were implanted subcutaneously in nude mice, and combined with TMZ treatment, the results showed that knockdown MAEA or TMZ treatment individually could suppress tumor growth; the combination of knockdown MAEA and TMZ treatment could significantly inhibit tumor growth (Fig. 7A-C). The data on IHC staining showed that, compared with the SF295-NC group, the expression of MAEA, HIF-1α and CD133 was downregulated and the expression of PHD3 was upregulated in the SF295-shMAEA group (Fig. 7D). The above results suggest that knocking down MAEA can inhibit the malignant progression of GBM by increasing PHD3 and further reducing HIF-1α and CD133.

DISCUSSION
Tumor stemness is the main cause of radiotherapy and chemotherapy resistance and recurrence [27]. Studies have shown that CD133, Nestin, Nanog and SOX2 are the main stemnessrelated molecules in GBM cells [28]. Multiple mechanisms, including transcriptional regulation, epigenetic and posttranslational modification, contribute to the regulation of GBM cell stemness. For example, transcription factor FOXM1 promotes SOX2 transcription and then enhances stemness and radioresistance in GBM [29]. TRIM8 degrades PIAS3 through the proteasome pathway and promotes STAT3 phosphorylation to regulate the stemness of GBM [30]. Here, we identified E3 ligase MAEA as a novel molecule relative to the stemness of GBM. As is well known, this the first time that the function and mechanism of MAEA in tumors have been clarified. Our study shows that MAEA is highly expressed in GBM, and its high expression is associated with tumor recurrence and poor prognosis. Moreover, the overexpression of MAEA could promote the proliferation, invasion, stemness and TMZ resistance of GBM cells. Knocking down MAEA in vivo inhibits the growth and TMZ resistance of GBM xenografts. Although the role of MAEA in the recurrence of GBM remains to be further studied, the current work has fully confirmed the important role of MAEA in the malignancy of GBM. Fig. 1 MAEA is highly expressed in glioma and is associated with glioma/GBM recurrence and poor prognosis. A Volcano plot shows the differentially expressed proteins between primary and corresponding recurrent gliomas with TMT-based quantitative proteomic analysis. B The heatmap of differentially expressed genes in clinical samples of primary and recurrent gliomas was analyzed by proteomics. C GO analysis (Molecular Function) for the significantly changed proteins by recurrent gliomas compared to corresponding primary tumors. CGGA database was used to analyze (D) the overall survival of MAEA expression in glioma, rglioma, GBM and rGBM, (E) the expression of MAEA in normal tissue and glioma, primary glioma and rglioma, normal tissue and GBM, primary GBM and rGBM. F Immunohistochemistry (IHC) staining was used to detect the expression of MAEA in 6 pairs of primary and corresponding recurrent glioma tissue. The MAEA expression in normal glial cells HEB and glioma cells was detected by (G) western blot and (H) RT-qPCR. rglioma: recurrent glioma, rGBM: recurrent GBM. The mean ± SD of n = 3 independent experiments is shown. *p < 0.05, **p < 0.01, ***p < 0.001, respectively. E The expression of stemness marker Nanog, Nestin and CD133 protein were detected by western blot. F The capacity for spheroidization was analyzed by sphere formation assay. G The rate of CD133( + ) cells was detected by flow cytometry. H The proliferation ability of cells treated with TMZ was analyzed by colony formation assay. The mean ± SD of n = 3 independent experiments is shown. **p < 0.01, ***p < 0.001, respectively.
PHD3 is a member of the group of hypoxia-induced prolyl hydroxylase enzymes (PHD1, PHD2 and PHD3), and it is widely regarded as a tumor suppressor. Dopeso et al. found that the overexpression of PHD3 in lung cancer cells can inhibit EMT and metastasis and reduce resistance to Erlotinib [31]. PHD3 attenuates tumor progression by inhibiting cell growth through downregulating the EGFR expression in gliomas [32]. In the present study, we demonstrated that PHD3 is one of the critical substrate proteins of MAEA in GBM. To date, a variety of E3 ligases have been reported to be involved in the ubiquitination modification of PHD3. E3 ligase Siah2 could ubiquitinate and degrade PHD3, and the 143-239 peptide of PHD3 was necessary for the ubiquitination degradation of PHD3 by Siah2 [24]. E3 ligase APC targets the D-box (99-107 peptide) of PHD3 for ubiquitination degradation, thus promoting the tumorigenesis of hepatocellular carcinoma [33]. Our study showed that MAEA targets the K159 of PHD3 to promote its ubiquitination, mediated by K48-linked polyubiquitination, thereby degrading PHD3 and regulating the malignancy of GBM.
Many studies have shown that PHD3 is necessary for the ubiquitination degradation of HIF-1α [34]. PHD3 promotes the hydroxylation of HIF-1α, which led to the recognition and ubiquitination degradation of HIF-1α by E3 ligase VHL [35]. Gastric cancer cells can produce miR-301a-3p-rich exosomes in an anoxic TME, and the targeted inhibition of PHD3 by miR-301a-3p leads to the accumulation of HIF-1α and promotes the malignancy and metastasis of gastric cancer [36]. Tanaka et al. found that the upregulation of PHD3 expression suppressed the HIF-1α signaling pathway and led to poor angiogenesis in pancreatic ductal adenocarcinoma [37]. Moreover, the regulatory role of HIF-1α in the stemness of tumor cells has been fully proven. HIF-1α can promote the transcriptional level of CD133 and increase the rate of CD133( + ) cells, thus promoting the stemness of hepatocellular carcinoma cells [26]. Moreover, Ohnishi et al. found that HIF-1α activates the CD133 promoter through the transcription factor ETS family [38]. In GBM, our data fully show that MAEA upregulates the expression of HIF-1α and CD133 proteins through the ubiquitination degradation of PHD3 in vitro and in vivo.
In summary, we demonstrate that MAEA can upregulate HIF-1α and CD133 via the ubiquitination degradation of PHD3 and contribute to the proliferation, migration, stemness and TMZ resistance of GBM cells (Fig. 7E). This study provides a novel potential biomarker for the progression of GBM and a potential intervention target for TMZ resistance.

TMT-based quantitative proteomic analysis
Six pairs of primary and corresponding recurrent glioma tissue samples were collected from Xiangya Hospital of Central South University, cryopreserved and sent to PTM BIO (Hangzhou, China) for TMT-based quantitative proteomic analysis by high-performance liquid chromatography (EASY-nLC 1000, Thermo Fisher Scientific, Waltham, MA) and mass spectrometry (TMQ Exactive Plus, Thermo Fisher Scientific). The clinical patient information is shown in Table S3.

Bioinformatics analysis
The CGGA database (http://www.cgga.org.cn/) contains mRNA microarray data and clinical information from 325 Chinese glioma patients and 20 normal glial tissue samples. GraphPad (San Diego, California) software was used to analyze the survival curve. The smart database (http://smart.emblheidelberg.de/) was used to analyze potential ubiquitin sites in PHD3. The UniProt database (https://www.uniprot.org/) was used to analyze the conservation of potential ubiquitin sites in PHD3.

Cell viability assay
Cell viability was measured with a CCK8 kit (C0005, Targetmol, Boston, MA) according to the manufacturer's instructions. The cells of the control group and the treatment group were cultured in 96-well plates for an appropriate time, and the viability was measured at 450 nm after 2 h incubation with determination solution using a Microplate Reader (BioTek ELx800, Winooski, VT).

Colony formation assay
The same number of cells from the control group and treatment group were seeded in a 6-well plate; after incubation for 2 weeks, the colonies were fixed with 4% paraformaldehyde and stained with 0.125% crystal violet (V5265, Sigma-Aldrich, St. Louis, MO), and the number of viable colonies (defined as colonies with more than 50 cells) was calculated.

Scratch assay
Cells were seeded in 24-well plates and grown to 90% confluence; then, a sterile 200 μL pipette was used to make a single scratch. The migration of cells was measured at time 0 h and 24 h with serum-free medium. The reduced area was analyzed using ImageJ (NIH).

Transwell assay
The Transwell assay was performed with 24-well plates and Transwell inserts (BD Biosciences, SanDiego, CA) coated with Matrigel (356234, Corning, NY). First, 1 × 10 5 cells were added to the upper chamber in 0.2 mL of serum-free medium, and 0.8 mL of culture medium containing 10% FBS was added to the lower chamber; after incubation for 24 h, the invasive cells attached to the bottom surface of the filter were stained with crystal violet solution (V5265, Sigma-Aldrich) and were observed with a microscope (Leica DMI 3000B, Germany).

Mass spectrometry analysis
LN229-MAEA-OE cells were lysed in IP buffer (87787, Thermo Fisher Scientific) containing an inhibitor cocktail (4693116001, Roche). The protein concentration was determined by BCA reagent (AR0197, Boster Biological Technology). Subsequently, the cell lysates were incubated overnight with protein-A/G magnetic beads and Flag-tag antibody (80010-1-RR, Proteintech). The beads were then sent to PTM BIO for analysis by mass spectrometry (TMQ Exactive Plus, Thermo Fisher Scientific). Fig. 4 MAEA targets PHD3 for ubiquitination and degradation at K159. SF295-shMAEA#1, SF295-shMAEA#2, LN229-MAEA-OE cells and control cells SF295-NC and LN229-vec were used. A The expression of MAEA and PHD3 was detected by western blot. B After CHX (10 μg/mL) treatment, the protein expression of MAEA and PHD3 was detected by western blot, and the protein stability was analyzed by ImageJ. C After MG132 (20 μM) treatment, the protein expression of MAEA and PHD3 was detected by western blot. D The smart database was used to analyze the potential ubiquitin sites of PHD3. The species conservation of potential ubiquitin sites was analyzed by the UniProt database. The P4Hc domain represents the homolog of prolyl 4-hydroxylase α subunit. E After treatment with MG132 (20 μM) for 10 h, the ubiquitin level of PHD3 was detected by IP. F After 293 T cells were co-transfected with His-PHD3, Flag-MAEA and HA-Ub plasmids for 48 h, and further treated with MG132 (20 μM) for 10 h, the ubiquitin level of PHD3 was detected by IP. G After 293 T cells were co-transfected with Flag-MAEA, HA-Ub, and His-PHD3 or His-PHD3-K159R and His-PHD3-K172R plasmids for 48 h, and further treated with MG132 (20 μM) for 10 h, the ubiquitin level of PHD3 was detected by IP. H After 293 T cells were co-transfected Flag-MAEA, His-PHD3 and HA-Ub-WT or HA-Ub-K48R and HA-Ub-K63R plasmids for 48 h, and further treated with MG132 (20 μM) for 10 h, the ubiquitin level of PHD3 was detected by IP. I After transfected with shRMND5A plasmids in SF295 and LN229 cells for 48 h, and treated with MG132 (20 μM) for 10 h, the ubiquitin level of PHD3 was detected by IP. J After co-transfected Flag-MAEA and shRMND5A plasmids in SF295 and LN229 cells for 48 h, and treated with MG132 (20 μM) for 10 h, the ubiquitin level of PHD3 was detected by IP. Fig. 5 PHD3 is involved in MAEA-mediated proliferation, invasion, stemness and TMZ resistance of GBM cells. shPHD3 plasmid was transiently transfected into SF295-shMAEA cells or His-PHD3 plasmid was transiently transfected into LN229-MAEA-OE cells. A, B The cell proliferation ability was detected by CCK8 and colony formation assays. C, D The migration and invasive ability of cells was analyzed by scratch and Transwell assays. E The capacity for spheroidization was analyzed by sphere formation assay. F The rate of CD133( + ) cells was detected by flow cytometry. G The proliferation ability of cells treated with TMZ was detected by colony formation assay. The mean ± SD of n = 3 independent experiments is shown. *p < 0.05, **p < 0.01, ***p < 0.001.

Immunofluorescence
Flag-MAEA and His-PHD3 plasmids were co-transfected into the cells. After the cells attached to the poly-L-lysine (PLS) slides, they were then fixed with 4% paraformaldehyde, permeated with 5% Triton 100 and blocked with 10% donkey serum. The cells were incubated with primary antibodies Flag-tag (80010-1-RR, Proteintech) and His-tag (66005-1-Ig, Proteintech), and then incubated with Rabbit anti-Rat IgG (A-21211, Thermo Fisher Scientific) or Goat anti-Mouse IgG (A-21212, Thermo Fisher Scientific), followed by DAPI staining solution (E607303, BBI Life Sciences, Shanghai, China) to dye the nuclei. The images were collected with a confocal microscope (LSM 510 META, Carl Zeiss, Germany). Overlap rate analysis was performed using ImageJ.

Cell fractionation
A Cell Fractionation Kit (#9038) was purchased from Cell Signaling Technology. SF295 and LN229 cell fractions were separated into cytoplasm, organelles/membrane and nucleus/cytoskeleton according to the reagent manufacturer's instructions. Separated fractions were subsequently used for immunoblot analysis.

Animal experiment
Sixteen female nude mice (BALB/C, 5-week-old) were randomly divided into two groups for subcutaneously injection with SF295-NC or SF295-shMAEA cells (1 × 10 7 cells per mouse). When the tumor size reached 50 mm 3 , either the SF295-NC or SF295-shMAEA group was further randomly divided into two groups for treatment with DMSO or TMZ (60 mg/kg, intragastric administration, 5 consecutive days), respectively [39]. The tumor formation was examined and measured every 3 days. The volume of the tumor was calculated as volume (mm 3 ) = D 2 ×d/2, in which D and d were the shortest and longest diameters, respectively. After the animals were sacrificed at the indicated time, the tumor tissue was collected and weighed, and then fixed with 10% buffered formalin for immunohistochemical (IHC) analysis.

Immunohistochemistry (IHC)
Fifty-six paraffin-embedded tumor tissue sections with the clinical details of glioma patients obtained from the Department of Pathology of Xiangya Hospital (2019-2021) (Table S4). And the paraffin-embedded tumor tissue sections of 6 pairs of primary and corresponding recurrent gliomas also obtained from the Department of Pathology of Xiangya Hospital. The clinical tissue and animal tumor tissue sections were employed for IHC by using a two-step test kit (PV-9000, ZSGB-BIO, Beijing, China) and DAB kit (ZLI-9017, ZSGB-BIO). Then, all immunostained sections were re-stained with Hematoxylin (E607317, Sangon bio, Shanghai, China). The expression of each protein in IHC was semi-quantitatively evaluated using the method previously described [40].

Statistical analysis
The GraphPad Prism 9 (SanDiego, California) statistical software was used for statistical analysis. The data are expressed as average ±SD. A doubletailed Student t test was used to evaluate the differences among groups, and p < 0.05 was considered to be statistically significant.

DATA AVAILABILITY
Source data and reagents are available from the corresponding author upon reasonable request.