PDIA3P1 promotes Temozolomide resistance in glioblastoma by inhibiting C/EBPβ degradation to facilitate proneural-to-mesenchymal transition

TMZ sensitivity data from GBM cell lines were obtained from the GDSC (www.cancerRxgene.org↗) database which is the largest public database on molecular markers of cancer drug sensitivity and drug response [28]. Corresponding cell lines expression data were available from the CCLE (https://portals.broadinstitute.org/ccle/↗) [29]. Transcript level data, somatic mutation and associated clinical information of TCGA GBM were extracted from GDC Data Portal (https://portal.gdc.cancer.gov/↗). The RNA-seq transcriptome data and clinical traits of the CGGA GBM were downloaded from CGGA database (http://www.cgga.org.cn/↗). The microarray information of GSC expression was available in the Gene Expression Omnibus (GEO) database (GSE68029 at www.ncbi.nlm.nih.gov/geo↗).

The limma R package was leveraged to identify differentially expressed genes (DEGs) between TMZ resistance and sensitive cell lines. The top 30 upregulated genes sorted according to p-value in TMZ resistant group were visualized using the pheatmap R package.

To determine the abundance of GBM immune infiltration levels, immune gene signatures were obtained from data of Bindea et al. [30] to perform ssGSEA. The immune cell infiltration levels were estimated using “GSVA” R package based on deconvolution algorithm.

The gene sets of “c2.cp.kegg.v7.4”, “c5.go.bp.v7.4”, “verhaak glioblastoma mesenchymal” and “verhaak glioblastoma proneural” were obtained from The Molecular Signatures Database (MSigDB; http://www.gsea-msigdb.org/gsea/login.jsp↗) database for running GSVA. P-value < 0.05 indicates statistical significance.

Human glioma cell lines U118MG, U87MG, LN229 and U251MG were purchased from the Chinese Academy of Sciences Cell Bank and cultured in DMEM medium (Gibco, USA) with 10% fetal bovine serum (FBS). The neural progenitor cell (NPC) and GBM patient-derived GSC cell lines and were kindly donated by Dr. Krishna P.L. Bhat (The University of Texas, M.D. Anderson Cancer Center, Houston, TX). GSC lines (GSC20, GSC267, GSC8–11, GSC11) have been used extensively in previous studies and the subtypes of GSCs have been clarified according to the Verhaak or Philips gene signatures, respectively. GSCs and NPC were cultured in DMEM/F12 (Gibco, USA) with 2% B-27 no serum supplement (Gibco, USA), 20 ng/mL human recombinant bFGF (R&D Systems) as well as 20 ng/mL human recombinant EGF (R&D Systems, USA). The GSC or NPC spheres were digested using accutase solution (Sigma-Aldrich, USA). All cell lines were cultured in a humid chamber at 37 °C and containing 5% carbon dioxide and 5% oxygen.

TRIzol (Invitrogen, USA) was used to extract total RNA according to manufacturer’s instruction. The high capacity cDNA Reverse Transcription Kit (Toyobo, China) was leveraged for reverse transcription in accordance with the manufacturer’s protocol. An Mx-3000P Quantitative PCR System (Applied Biosystems, USA) was used for qRT-PCR. The primers used for RT-qPCR were: 5′-GGAAAACCACTGGGGAGGAC-3′ (forward) and 5′-CAGTGCAGCTAAGAAATGGCT-3′ (reverse) for PDIA3P1; 5′-GCACCGTCAAGGCTGAGAAC-3′ (forward) and 5′-TGGTGAAGACGCCAGTGGA-3′ (reverse) for GAPDH; 5′-TTTGTCCAAACCAACCGCAC-3′ (forward) and 5′-GCATCAACTTCGAAACCGGC-3′ (reverse) for CEBPB.

Human full-length PDIA3P1 as well as sh-PDIA3P1 plasmids were used in the current study for stable overexpression and knockdown, respectively, whereas empty plasmid was used as a control. Lentiviral particles were constructed by transfecting 293 T cells with the packaging vectors psPAX2 and pMD2G. Lentiviral particles were collected 24 and 48 hours after transfection of 293 T cells, filtered through a 0.45 μm filter (Corning), and then used to treat cells in culture. After 48 hours, cells were selected by Puromycin (2 μg/mL). All small interfering RNAs (siRNA) and overexpression plasmids were purchased from Genepharma (shanghai, China). For short-term knockdown and overexpression of GBM cells, cells were transfected of siRNAs and plasmids using the Lipofectamine 3000 kit (Invitrogen, USA) according to the manufacturer’s instruction.

TMZ and NEF (Synonyms: VX-745) were purchased from MedChemExpress (MCE, https://www.medchemexpress.cn/↗). TMZ and NEF were dissolved in dimethyl sulfoxide (DMSO) at concentrations of 100 mM and 10 mM, respectively. TMZ and NEF in solvent are stored at − 20 °C and used up within 1 month. The primary antibodies used in this study are listed as follows: β-actin (Cell Signaling Technology, 8480), CD44 (Cell Signaling Technology, 3570), C/EBPβ (Abcam, ab32358), YKL-40 (Cell Signaling Technology, 47,066), SOX2 (Cell Signaling Technology, 3579), γ-H2AX (Cell Signaling Technology, 7631), MDM2 (Abcam, ab259265), JUN (Cell Signaling Technology, 9165), p-JUN (Cell Signaling Technology, 3270), ubiquitin (Cell Signaling Technology, 3933), P38 (Cell Signaling Technology, 8690), p-P38 (Cell Signaling Technology, 4511).

CCK-8 reagent (RiboBio, China) was used to assess GBM cells viability. We seeded GBM cells in 96-well plates at a density of 2 × 10³ cells per well in 100 μl of Gibco DMEM containing 10% FBS. The cells were incubated at 37 °C 12 h for cells adhesion and then treated with different concentrations of TMZ or NEF. After incubation for 48 h, 10 μl of CCK-8 solution was added to each well for 1 h before measurement. Absorbance (OD value) at 450 nm was measured using a microplate.

The alkaline comet assay was used to detect the damaged DNA with high sensitivity [31]. GBM cells in different groups were harvested in PBS at a 1–3*10⁵ cells/ml density. Cells were mixed with molten LM agarose at a ratio of 1:10 (V/V) and 50 μl of the mixture was immediately pipetted onto a CometSlide. Then cells were lysed in alkaline lysis solution at 4 °C for 12 h for lysis. After that, the slides were soaked with alkaline electrophoresis buffer for 20 minutes away from light and electrophoresis for 30 min at 25 V. After precipitation and washing, the slides were stained with Green-DNA Dye and images were captured by fluorescence microscopy.

GBM cells were fixed in 4% paraformaldehyde for 15 min and washed three times in PBS. Then cells were permeabilized in 0.3% Triton X-100 for 10 min and blocked with 5% Goat serum for 1 h. Then the cells were incubated with indicated primary antibodies overnight at 4 °C. Cells were then incubated with fuorescent second antibody at room temperature for 1 h. DAPI was used to counterstain nuclei for 15 min. Images were captured using a LeicaSP8 confocal microscope.

EdU cell proliferation assay kit (RiboBio, China) was used to determine cell proliferation. Cells were incubated with 200 μl of 5-ethynyl-20-deoxyuridine at 37 °C for 2 h. After fixed and permeabilized, the cells were incubated with Apollo reagent for 30 min and the Hoechst were used to stain nuclei. The images were viewed and obtained using fluorescence microscope.

Both suspended and adherent GBM cells were obtained for apoptosis analysis after treating with TMZ or DMSO (solvent control of TMZ) for 48 h. Annexin VFITC and PI staining (BD Biosciences, USA) was leveraged for apoptosis analysis according to the instruction. The number of cells were counted by BD Accuri C6 flow cytometer.

We seeded about 2000 GBM cells in 6-well plates per well in 1.5 ml of Gibco DMEM containing 10% FBS. The cells were incubated in a humidified chamber containing 5% CO2 and 5% O2 at 37 °C for 2 weeks. After that, colonies were fixed and stained with crystal violet (Solarbio, China) for 20 min. The colonies were washed with PBS for at least three times and the number of colonies were counted.

We seeded about 1000 GSCs per well in 6-well plates with 1.5 ml DMEM/F12 containing 2% B-27. After 7 days incubation at 37 °C, the images were acquired and the relative diameters of neurospheres were calculated.

We implanted GSCs into ultralow-attachment 96-well plates at densities of 0, 2, 4, 8, 16, 32, 64 and 128 cells per well in 10 replicates. The number of wells with neurospheres formation was counted after 7 days incubation. Collected data was analyzed using (http://bioinf.wehi.edu.au/software/elda/↗).

CHX was used to inhibit new proteins synthesis. Cells were treated with 100 μg/ml CHX for 0 h, 2 h, 4 h, 6 h and 8 h prior to protein collection. The proteins levels were detected by western blot assay.

Biotinylated PDIA3P1 and its anti-sense sequence were synthesized by RiboBio (GenePharma, China). Pierce™ Magnetic RNA-protein pull-down kit (Thermo Fisher Scientific, SA) was used for RNA pull-down assay. Cell lysates of GSC267 were firstly incubated with a biotin-labelled PDIA3P1 probe. Then the conjugated magnetic beads were added to cell lysates and the interacting proteins were separated by western blot and then the silver staining was used for visualization.

Magna RIP kit (Millipore, USA) was leveraged for RIP assay according to manufacturer’s instruction. RT-qPCR was used for detecting the relative expression of immunoprecipitated RNA. The IgG antibody (from Magna RIP kit) was used for negative control.

The IP assay was performed using Pierce Classic Magnetic immunoprecipitation (IP)/Co-IP Kit (Thermo Fisher, USA) according to the manufacturer’s instruction. Firstly, the different antibodies were incubated with protein A/G magnetic beads. Then the cell lysates from GSCs were collected and incubated with antibody coupled beads. The beads interacting proteins were washed and denatured and the proteins were examined using western blotting.

To assess the combination of effect of TMZ and NEF, GBM cells were treated with different concentrations of TMZ and NEF for 48 h in 3 replicates. CompuSyn software (Biosoft, Ferguson, MO, USA) was leveraged to evaluate drug synergism. The combination index (CI) values were calculated using non-constant ratios drug combination analysis according to instruction of the software. CI < 0.75, CI = 0.75–1.25, and CI > 1.25 were defined as synergistic, additive, and antagonistic effects, respectively.

Luciferase labeled and stably transfected sh-PDIA3P1-U118MG cells or sh-Control-U118MG, or ov-PDIA3P1-U251 or ov-vector-U251 were injected into the brains of randomly grouped 4-week male BALB/c nude mice (5 × 10⁵ cells/mouse). On the fifth postoperative day, the mice were randomly divided into TMZ treatment or control groups. Mice were treated with or without TMZ by oral gavage per week (5 mg/kg, p.o., 5 times per week). For evaluating the anti-tumor effect of TMZ in combination with NEF in vivo, luciferase-labeled GSC267 cells (1 × 10⁶ cells/mouse) were implanted into the brains of 4-week male BALB/c nude mice. After 7 days post-operative, the mice were randomly divided into four groups, control, TMZ only (5 mg/kg, p.o.,5 times per week), NEF only (5 mg/kg, p.o.,5 times per week) and combination group. To evaluate the intracranial tumor, bioluminescence imaging was used to quantify tumor burden using an IVIS Lumina Series III (PerkinElmer). All procedures used for animal treatments and experiments were approved by and under the requirements of the Animal Care and Use Committee of the Qilu Hospital of Shandong University.

All statistical analysis was conducted by R 4.1.1 and GraphPad Prism 8.0 software. Acquired data were certified as normal distribution through Shapiro-Wilk Normality test and homogeneity of variances through Bartlett test. Then t-tests and one-way ANOVA were used for comparisons between two independent samples and comparisons among multiple samples, respectively. The Wilcoxon test was used for non-parametric data. P-value < 0.05 was considered statistically significant (*p-value < 0.05; **p-value < 0.01; ***p-value < 0.001). The receiver operating characteristic (ROC) curve was used to evaluate the diagnostic value of PDIA3P1, and the area under the curve (AUC) was quantified using the pROC R package. Pearson correlation was used to calculate the correlation between two or more groups. Kaplan-Meier curve and log-rank test were used to evaluate survival between different groups.

To investigate the functional role of PDIA3P1 in promoting TMZ resistance, PDIA3P1 was knocked down using two independent shRNAs in U118MG and U87MG cells and overexpressed in U251 and LN229 cells. The expression of PDIA3P1 was detected using qRT–PCR (Fig. S1D). Knockdown of PDIA3P1 in resistant cell lines (U118MG and U87MG) resulted in a notable reduction in IC₅₀ and further inhibition of the tumor cell growth rate upon TMZ treatment (Fig. 1D and Fig. S1E). In contrast, overexpression of PDIA3P1 in sensitive cell lines (U251 and LN229) resulted in a significant increase in IC₅₀ values and counteracted the inhibitory effect of TMZ on tumor cell growth (Fig. 1E and Fig. S1E).

To evaluate the effect of PDIA3P1 on the TMZ-resistant phenotype in vivo, 5 × 10⁵ luciferase-labeled U118MG-shPDIA3P1 or U118-shNC and U251-PDIA3P1 or U251-Vector cells were injected into nude mice. We tracked tumor proliferation using in vivo bioluminescence imaging. Despite the initial tumor size being similar (Fig. S2A, B), xenografts bearing U118MG-shPDIA3P1 cells displayed significant tumor growth inhibition, whereas xenografts bearing U251-PDIA3P1 cells exhibited tumor growth promotion. As expected, TMZ treatment (5 mg/kg, p.o., 5 times per week) reduced tumor burden. Tumor size in the U118MG-shPDIA3P1 group was reduced compared to that in the control group (Fig. 1F, G), while tumor size in the U251-PDIA3P1 group was relatively increased compared to that in the control group (Fig. 1H, I). Consistently, Kaplan–Meier curves demonstrated that the overall survival time of mice was prolonged in the PDIA3P1 knockdown group with and without TMZ treatment (Fig. S2C). Although TMZ treatment significantly prolonged the survival time of mice in the U251-Vector group, PDIA3P1 overexpression decreased the survival time of mice in both the treatment and control groups (Fig. S2D). H&E-stained mouse brain sections showed that knockdown of PDIA3P1 greatly reduced tumor invasiveness, with or without TMZ treatment, whereas overexpression of PDIA3P1 promoted tumor invasiveness (Fig. S2E, F). Taken together, these findings indicate that PDIA3P1 promotes glioma cell resistance to TMZ both in vitro and in vivo.

We further investigated the interaction of PDIA3P1-C/EBPβ. PDIA3P1 knockdown decreased protein expression of C/EBPβ (Fig. 5D, K), but not mRNA levels of C/EBPβ (Fig. S7E), suggesting that PDIA3P1 regulates protein levels of C/EBPβ by affecting translational or posttranslational modification. The ubiquitin–proteasome system (UPS) is the primary pathway of protein degradation, and it participates in the degradation of more than 80% of proteins in cells [34]. To confirm the possibility that PDIA3P1 regulates C/EBPβ through the proteasome, GSCs were treated with the proteasome inhibitor MG132. Knockdown of PDIA3P1 significantly decreased the expression of C/EBPβ, whereas MG132-treated GSCs with silenced PDIA3P1 displayed minimal changes in C/EBPβ levels (Fig. 5K). Then, we blocked protein synthesis using cycloheximide (CHX) and found that PDIA3P1 knockdown significantly shortened the half-life of C/EBPβ protein in GSC267 cells (Fig. 5L). Consistently, the half-life of the C/EBPβ protein in GSC8–11 cells stably overexpressing PDIA3P1 was significantly longer than that in the corresponding control cells (Fig. S7F). Immunoprecipitation (IP) results demonstrated that knockdown of PDIA3P1 significantly increased the ubiquitylation of C/EBPβ in GSC267 cells, whereas overexpression of PDIA3P1 significantly decreased the ubiquitylation of C/EBPβ in GSC8–11 cells (Fig. 5M). Taken together, our data suggest that PDIA3P1 is involved in the posttranslational modification of C/EBPβ.

E3 ubiquitin ligases are a family of more than 700 proteins that bind ubiquitin to target proteins and play a major role in protein degradation [34]. To further investigate the E3 ubiquitin ligases involved in the posttranslational modification of C/EBPβ, we reviewed numerous references and found that mouse double minute 2 homolog (MDM2) targets C/EBPβ for degradation [35]. MDM2 is an E3 ubiquitin ligase of the RING finger family that is involved in the degradation of p53 [36]. Since we confirmed that PDIA3P1 directly binds to C/EBPβ to affect the ubiquitination levels of C/EBPβ, we hypothesized that PDIA3P1 may impact C/EBPβ-MDM2 complex formation. To test this hypothesis, the interaction between C/EBPβ and MDM2 was investigated using co-IP assays. The results demonstrated that overexpression of PDIA3P1 hampered the interaction between C/EBPβ and MDM2 in GSC267 cells. In addition, knockdown of PDIA3P1 resulted in increased MDM2 binding to C/EBPβ in GSC267 cells (Fig. 5N). Collectively, our data suggest that PDIA3P1 stabilizes C/EBPβ by disrupting the C/EBPβ-MDM2 complex.

Targeting p38α may block the stress response of tumor cells, preventing TMZ-induced upregulation of PDIA3P1. Therefore, we reviewed small molecule inhibitors specifically targeting p38 from DRUGBANK and MCE. We screened NEF as a potential drug for the treatment of Alzheimer’s disease (AD), which has been preliminarily confirmed to be safe for human use [39]. NEF has excellent BBB permeability, suggesting its value for CNS disorders, and some studies have demonstrated antitumor activity of NEF [40–42]. CCK-8 cell proliferation was performed to determine the IC₅₀ of NEF in four cell lines (Fig. 6D). GBM cells treated with NEF inhibited TMZ-induced upregulation of PDIA3P1 (Fig. 6E). Then, we further explored whether TMZ combined with NEF could synergistically inhibit GBM cell growth. GBM cells were treated with the indicated concentrations of TMZ and NEF, and cell growth inhibition was assessed using the CCK-8 assay (Fig. 6F). Based on the results of proliferation inhibition, we calculated the combination index (CI) score to evaluate the combined effect of TMZ and NEF (Fig. 6G and Fig. S8A). CI > 1.25, CI = 0.75–1.25, and CI < 0.75 were defined as antagonistic, additive and synergistic effects, respectively. For instance, in GSC267 cells, a relatively low concentration of TMZ (50 μM) and NEF (20 μM) may exhibit a better synergistic effect (CI = 0.44), despite their relatively low growth inhibitory effects of approximately 23%. When GSC20 cells were treated with moderate concentrations of TMZ (800 μM) and NEF (80 μM), they exhibited an additive effect despite their relatively high growth inhibition of approximately 81%. Collectively, these data reveal that TMZ in combination with NEF exhibits synergistic effects at the indicated concentrations.

Activation of the p38-MAPK signaling pathway could further activate certain transcription factors, such as JUN. We observed a significantly positive relationship between expression of JUN and PDIA3P1 in the TCGA and CGGA datasets (Fig. 6H). Knockdown of JUN not only reduced PDIA3P1 expression but also counteracted TMZ-induced upregulation of PDIA3P1 (Fig. 6I), preliminarily suggesting that JUN is responsible for PDIA3P1 transcription. We constructed four fragments of different lengths located upstream of the TSS based on the JUN binding motif (Fig. 6J). The four luciferase reporter plasmids were transfected into GSC267 cells individually to verify their JUN binding sites. The luciferase activity of fragment 4 was statistically unchanged after knockdown of JUN, demonstrating that JUN does not bind to fragment 4 (Fig. 6K). To further determine the binding sites in more detail, we designed three pairs of PCR primers and performed a ChIP assay. The qRT–PCR assay yielded approximately 10-fold enrichment and 5-fold enrichment for site #1 and site #2, respectively, while there was no significant enrichment at site #3 (Fig. 6L and Fig. S8B). In conclusion, our data suggest that JUN is involved in TMZ-induced upregulation of PDIA3P1 and directly binds and initiates PDIA3P1 transcription.

Additional file 1: Supplementary Figure 1. A The IC50 for TMZ and expression of PDIA3P1 of 10 glioma cell lines (TMZ sensitive and TMZ resistant cell lines). B Patients with high PDIA3P1 expression exhibited shorten progress disease survival time. C The PDIA3P1 expression in 4 glioblastoma cell lines. D PDIA31P was knocked down in U118MG and U87MG cells by two different shRNA and overexpressed in LN229 and U251 cells. E Cell proliferation was assessed by CCK-8 assay. Supplementary Figure 2. A B Bioluminescence imaging of tumor growth on day five in U118MG (A) and U251 (B) xenograft nude mice. C D Kaplan–Meier visualized survival time for animals in different groups for U118MG (C) and U251 (D). E F Representative images of hematoxylin and eosin (H&E) staining in sections from U118MG (E) and U251 (F) xenografts. Scale bar, 400 μm. Supplementary Figure 3. A Patients were divided into high and low PDIA3P1 expression groups according to PDIA3P1 expression, and GSVA analysis was performed, and the results were presented using heatmap. B Heatmap visualized immune infiltration using ssGSEA, and the correlation between PDIA3P1 expression and immune infiltration was analyzed using the chi-square test (lower panel). C Waterfall plot showing tumor somatic cell mutations in low (left panel) and high (right panel) PDIA3P1 expression groups. Supplementary Figure 4. A B DNA damage was assessed by comet (A. Scale bar, 100 μm) and γ-H2AX IF staining (B. Scale bar, 10 μm) assays. Knockdown of PDIA3P1 significantly promoted TMZ treatment-induced DNA damage. C D Cell proliferation was assessed by EdU (C. Scale bar, 200 μm) and colony formation (D. Scale bar, 200 μm) assays. Knockdown of PDIA3P1 further significantly increased the proliferation inhibitory effect caused by TMZ. The lower panel exhibited the quantification of EdU and colony formation assays. E F Representative images of comet (E. Scale bar, 100 μm) and γ-H2AX IF staining (F. Scale bar, 10 μm) assays for LN229 cells. Overexpression of PDIA3P1 remarkedly reduced TMZ treatment-induced DNA damage. Supplementary Figure 5. A B DNA damage was assessed by γ-H2AX IF staining assay in GSC20. Knockdown of PDIA3P1 significantly promoted TMZ treatment-induced DNA damage, whereas overexpression of PDIA3P1 counteracted the DNA damage induced by TMZ treatment (Scale bar, 10 μm). C D DNA damage was assessed by comet assay in GSC20 (Scale bar, 100 μm). E F Cell proliferation was assessed by EdU assay in GSC20. Knockdown of PDIA3P1 promoted proliferation inhibition caused by TMZ treatment, whereas overexpression of PDIA3P1 partially counteracted TMZ-mediated cell growth inhibition (Scale bar, 200 μm). Supplementary Figure 6. A The expression of PDIA3P1 in classical, proneural (PN) and mesenchymal (MES) tissues in the CCGA dataset. B ROC curves of PDIA3P1 for MES-GBM subtype prediction in CCGA. C Knockdown of PDIA3P1 in GSC20 and GSC267, overexpression of PDIA3P1 in GSC8–11. D Neurospheres formation assay revealed knockdown of PDIA3P1 reduced self-renewal capacity of GSC20 and GSC267. Scale bar, 200 μm. The right panels were the quantification of sphere diameters. E The quantification of IF staining for SOX2 and CD44 in GSC20 (left panel) and GSC267 (right panel). Knockdown of PDIA3P1 resulted downregulation of CD44 and upregulation of SOX2 in GSC20 and GSC267. Supplementary Figure 7. A B Representative images of γ-H2AX IF staining (A. Scale bar, 10 μm) and comet (B. Scale bar, 200 μm) assays revealed DNA damage in GSC267. C D Representative images of γ-H2AX IF staining (C. Scale bar, 10 μm) and comet (D. Scale bar, 200 μm) assays revealed DNA damage in GSC8–11. E The expression of C/EBPβ mRNA detected by qPCR. F Western blotting analysis of C/EBPβ in GSC8–11 PDIA3P1 stable overexpressed and control cells after treatment with CHX (100 μg/ml) for indicated times. G Western blotting analysis of p-P38 and P38 expression with increasing concentrations of TMZ treatment. Supplementary Figure 8. A Visualization of the Fa-CI (fraction afected Combination Index) results obtained form CompuSyn for U118MG, GSC20, GSC267, and U251. B The detailed sequence of PDIA3P1 promoter (610–1310) and three predicted binding sites for JUN. Supplementary Figure 9. A B EdU assay (A. Scale bar, 200 μm) and apoptosis assay (B) showed that TMZ combined with NEF exhibited excellent anti-tumor cells effects, respectively. C Bioluminescence imaging of tumor growth on day five in GSC267 xenograft nude mice. D Kaplan–Meier visualized survival time for mice in different treatment groups. E Representative images of hematoxylin and eosin (H&E) staining in sections from GSC267 xenografts. Scale bar, 400 μm.Additional file 2: Table S1. The information of DEGs between TMZ resistant cell lines and TMZ sensitive cell lines. Table S2. The Results of GSVA. Table S3. The Results of ssGSEA for Tumor Immune Infiltration Analysis. Table S4. Mass spectrometry result of proteins related to PDIA3P1.