Samples were harvested from 15 fresh matched pairs of tissues samples isolated from GBM patients who were hospitalized in The Second Affiliated Hospital of Dalian Medical University between September 2014 and December 2016 (Supplementary Table S2). Tumoral tissues in addition to their adjacent non-cancerous tissues (ANCTs) were obtained from all patients during surgery under the conventions of the Ethics Committee. ANCT was defined as the normal breast tissue diagnosed by the pathologists through H&E staining. Written informed consent was obtained from all patients. The Ethics Committee of The Second Affiliated Hospital of Dalian Medical University approved the study protocol, and all experiments were performed in accordance with the approved guidelines.

Human U87MG (U87) and U251 cell lines were obtained from American Type Culture Collection (Manassas, VA, United States). Cells were maintained in DMEM supplemented with 10% fetal bovine serum (FBS). All cell cultures were maintained at 37°C in a humidified atmosphere containing 5% CO₂.

The lentiviral vectors were purchased from Shanghai Genechem Company Ltd., China. A non-silencing shRNA (5′-GCCTAACTGTGTCAGAAGGAA-3′) was used as the negative control (shCtrl). The siRNA sequence targeting the NUDT21 gene was 5′-ACCTCCTCAGTATCCATAT-3′. Cells were seeded into a six-well plate (∼5 × 10⁴ cells per well) and incubated at 37°C with 5% CO₂ until they reached ∼30% confluence before transfection.

Cell proliferation was determined by MTT (Roche Diagnosis, Indianapolis, IN, United States) assay. Cells were plated at a certain density (1 × 10³ cells per well) in 96-well plates. Cells were then incubated in complete media for another 5 days. After 5 days incubation, cell growth was measured. The effect of NUDT21 knockdown on cell viability was assessed as the percent cell viability compared with the untreated control group, which was assigned 100% viability. The optical density (OD) values were determined (wavelength: 490 nm). All experiments were performed in triplicate.

After achieving the logarithmic growth phase, U87MG or U251 cells were digested with trypsin, resuspended in standard medium, and then seeded into 96-well plates at a density of 2,000 cells/well. The number of green fluorescent protein fluorescence-positive cells was counted using a Cellomics Array Scan High Contents Screening Reader on five consecutive days.

Cells were plated in six-well plates for 24 h. After the G1/GO synchronization by FBS deprivation for 48 h, cells were incubated in DMEM supplemented with 10% FBS again for another 48 h. Then cells were collected and washed once with cold PBS, then stained with a propidium iodide (PI) solution (50 mg/mL PI and 0.5 mg/mL RNase in TBST buffer) for 15 min at 4°C. Cell cycle distribution was measured using a FACS Accuri C6 flow cytometer (Genetimes Technology Inc.).

Cell total proteins were collected using RIPA lysis buffer containing phosphatase and protease inhibitors (Roche, Switzerland) in accordance with the manufacturer’s instructions. Glioma and normal tissues were lysed in lysis buffer with shaking at 4°C for 30 min. Proteins were quantified by a bicinchoninic acid assay protein assay kit. Proteins were separated by SDS-PAGE, and then transferred to PVDF membranes, blocked with 5% fat-free dry milk or 5% BSA in TBST and immunoblotted with primary antibodies at 4°C overnight. The following day, the membranes were incubated with second antibodies at room temperature for 2 h. The protein bands were detected by enhanced chemiluminescence. The NFKBIZ antibody was purchased from Proteintech. The fibronectin-1 (FN1) and chitinase-3-like protein 1 (CHI3L1) antibodies were purchased from Abcam.

Total RNA was obtained using the TRIzol one-step RNA isolation kit (TaKaRa Bio, Dalian, China). Specific cDNA were synthesized from total RNA using PrimeScript RT Reagent Kit (TaKaRa Bio, Dalian, China) according to the manufacturer’s instructions. The relative mRNA expression of each gene was normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) RNA levels and was analyzed using the 2^-ΔΔC_T method. The primers were synthesized by Invitrogen (Shanghai, China). The primers were sense 5′-GGTCACTCAGTTCGGCAACAA-3′, antisense 5′-CTCATGCGCTGAAATCTGGC-3′ for NUDT21; sense 5′-CAAAGGCTTACAATGGCAAC-3′, antisense 5′-CATCGGGAACCAAATGCAC-3′ for NFKBIZ; and sense 5′-TGACTTCAACAGCGACACCCA-3′, antisense 5′-CACCCTGTTGCTGTAGCCAAA-3′ for GAPDH.

Total RNA was extracted from U87MG cells after transfection with shNUDT21, and 50–500 ng of RNA was used to generate biotin-modified amplified RNA (aRNA) using a GeneChip 3′ IVT Express Kit (Affymetrix, United States). Reverse transcription was performed using a T7 oligo (dT) primer, and a first-strand IVT Labeling Master Mix was used to produce multiple copies of biotin-modified aRNA. The aRNA was then purified and quantified. After fragmentation, the aRNA was hybridized to the GeneChip PrimeView Human gene expression array cartridge (Affymetrix, United States). After hybridization, the chips were stained with phycoerythrin and washed in a GeneChip Fluidics Station 450. The microarray signals were scanned and analyzed using a GeneChip Array Scanner 3000 7G.

Gene signatures were compared between the shCtrl and shNUDT21 cells. To identify pathways commonly deregulated in the shNUDT21 U87MG cells, enrichment of the differentially expressed gene signatures in human pathways was evaluated by gene set enrichment analysis (GSEA; Subramanian et al., 2005) using pathways collected in the hallmark and c2 curated gene sets¹ with 1,000 gene label permutations (gene sets). Significantly enriched gene sets, defined by nominal p < 0.05, were compared between shCtrl and shNUDT21 gene signatures.

All the animals were maintained in the SPF Laboratory Animal Center at Dalian Medical University, which was also the site at which all the animal experiments were performed. Female nu/nu mice (4–6 weeks old) were used for these experiments. To evaluate the effect of NUDT21 knockdown in a human U87MG xenograft glioma mouse model, we subcutaneously injected U87MG cells (2 × 10⁶ in 100 μl PBS) near the axillary fossae of the nude mice using a 27-gauge needle. The tumor cell-inoculated mice were then randomly divided into two groups, each of which contained five mice. The tumors were measured with a caliper every day, and tumor volumes were calculated using the following formula: V = 1/2 (width 2 × length). Body weights were also recorded. On day 5 after tumor cell inoculation, all the experimental mice were euthanized with ether anesthesia, and their total tumor weights were measured.

All animal maintenance and procedures were carried out in strict accordance with the recommendations established by the Animal Care and Ethics Committee of Dalian Medical University as well as the guidelines by the United States National Institutes of Health Guide for the Care and Use of Laboratory Animals. The protocol was approved by the Animal Care and Ethics Committee of Dalian Medical University. In the animal study, all efforts were made to minimize the suffering of the mice. All mice were humanely sacrificed under ether anesthesia inhalation before death.

Student’s t-test (two-tailed), t-test with Welch’s correction, F-test were performed to analyze the data using GraphPad Prism 6.0 software. The concrete methods of t-test analysis in current study are as follows: the data of two groups for comparison were analyzed by F-test firstly (homogeneity test of variance): If the value of F-test >0.05, the value of t-test was obtained according to heteroscedasticity double sample test. If the value of F-test <0.05, the value of t-test was obtained according to the heteroscedasticity double sample test; the value of t-test <0.05 indicated that there was significant difference between the two experimental groups, and the value of t-test >0.05 indicated that there was no significant difference between the two experimental groups. Two-way analysis of variance (ANOVA), followed by a Bonferroni’s test for multiple comparisons, were performed to analyze the data for Figures 2C,D,F, 3A and Supplementary Figures S1B,E, S2A. While one-way ANOVA, followed by a Tukey’s post-test for multiple comparisons, were performed to analyze the data for Figure 5B. For most of the in vitro and animal experiments, Student’s t-tests were used to calculate the p-value. p-Values less than 0.05 were considered statistically significant. ns, significant; ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; ^∗∗∗∗p < 0.0001.

To examine the NUDT21 expression level, we analyzed TCGA Level-3 expression levels in GBM tissues compared with normal tissues and found that NUDT21 was significantly upregulated in GBM tissues (Figure 1A). Then, we re-analyzed NUDT21 expression in GBM or normal tissue based on public data on the Oncomine website² and found that NUDT21 expression was also significantly increased in GBMs of two different datasets (Figures 1B,C). To confirm this result, we examined NUDT21 expression using quantitative real-time reverse transcription polymerase chain reaction (RT-qPCR) in 15 pairs of GBM tissues and corresponding non-tumor tissues (Figure 1D). All these indicated that upregulation of NUDT21 may be important for GBM pathogenesis.

The results of RT-qPCR indicated that all four glioblastoma cell lines could be of high-expression of NUDT21. To investigate the biological consequence of the loss of NUDT21 in GBM cells, the U87MG and U251 cell lines were selected for further study (Figure 2A). First, NUDT21 knockdown was performed using shRNAs (Figure 2B). Next, the MTT assay indicated that, compared with the cell proliferation in control cells, NUDT21 knockdown significantly decreased cell proliferation in U87MG cells. The OD values of three accessory holes were measured, followed by statistically analysis (t-test) between shNUDT21 group and control group at indicated days (Figure 2C). Next, the Celigo assay was performed to detect cell viability, and t-test was exerted to confirm the statistical significance of differences between shNUDT21 group and control group at indicated days. The colony numbers of U87MG cells transfected with shCtrl were significantly higher than those transfected with shNUDT21 (Figure 2D). These results indicated that the functional loss of NUDT21 inhibits GBM cell proliferation. U87MG cells were transfected with shNUDT21 to further elucidate the physiological role of NUDT21 in glioma cell growth. After 48 h, cell apoptosis was analyzed using flow cytometry. Our experiments confirmed that transfection of shNUDT21 enhanced apoptosis (Figure 2E). Besides, flow cytometry analysis was performed to further examine whether NUDT21 could affect the proliferation of GBM cells by altering cell cycle progression. Above all, shNUDT21 transfected into U87MG cells could block cell cycle progression (Figure 2F). The bar chart represented the percentage of cells in G0/G1, S, or G2/M phase, as indicated. And the results indicated that cells were arrested in the G2/M phase, the percentage of cells in the S phase decreased, while the G1 phase did not change significantly. Then we exerted t-test to further confirm the significance of differences between groups. The results of other GBM cell line, U251, confirmed those observed in U87MG cells (Supplementary Figure S1). Above all, these results suggested that the functional loss of NUDT21 may inhibit cell proliferation, block cell cycle progression and promote cell apoptosis.

To examine whether the NUDT21 expression level could affect tumorigenesis, shCtrl/shNUDT21-transfected U87MG cells were inoculated into nude mice. The average tumor volume was measured in each group, followed by statistically analysis (t-test) between shNUDT21 group and control group at indicated days. It could be indicated that the tumor growth rate was significantly slower in the shNUDT21 group than that in the control group (Figure 3A). Up to 16 days after injection, the average tumor weight in the shNUDT21 group was markedly lower than that in the shCtrl group (Figure 3B). Then we exerted RT-qPCR analysis, to examine the NUDT21 expression level in tumor tissues (Figure 3C). Furthermore, the tumors developed from shNUDT21-transfected U87MG cells were also found to display lower proliferating cell nuclear antigen (PCNA) staining compared to the tumors formed by shCtrl-transfected U87MG cells, as detected by immunohistochemistry analysis (Figure 3D). Besides, the data of U251 cells further determined the role of NUDT21 expression in GBM tumorigenesis (Supplementary Figure S2).

Abnormal activation of the MES phenotype in GBM is associated with an increase in cellular motility, glycolysis and genes related to inflammatory properties (Zhong et al., 2010). In order to elucidate the molecular mechanisms underlying the role of NUDT21 in GBM cell phenotypes, we performed gene microarray analysis to compare gene expression of shCtrl-transfected versus shNUDT21-transfected U87MG cells. Differentially expressed genes with more than 1.5-fold change in could be identified. The results demonstrated that 293 and 457 genes were upregulated and downregulated in the shNUDT21-transfected cells, respectively (Supplementary Table S1).

Next, we conducted GSEA (Subramanian et al., 2005) to determine whether particular gene sets were enriched in differentially expressed genes regulated by NUDT21. Several gene set classifications, as annotated by the collections of GSEA, were significantly enriched. Interestingly, we found that the knockdown NUDT21 signature was significantly reduced in cells with the MES signature (Figure 4A). In addition, other microarray data (Kim et al., 2016) also showed that NUDT21 expression was significant upregulated with 1.5-fold change in the GBM cells classified as MES (Figure 4B).

Recently, studies have demonstrated that NF-κB is a master regulator of MES transdifferentiation. Our GSEA results showed that genes affected by NUDT21 silencing in GBM cells had an enriched NF-κB pathway signature (Figure 4C). Thus, we decided to focus on the differential expression of the NF-κB signaling pathway. NFKBIZ, as an inhibitor of NF-κB activity, was identified as being upregulated by 1.5-fold after NUDT21 silencing. To further determine targets regulated by NFKBIZ, we examined the knowledge-based interactome surrounding the regulation of NFKBIZ using the Ingenuity Pathway Analysis, and then we overlaid the results with the microarray data that exhibited a 1.5-fold change cut-off. Several targets reported to be involved in the MES identity in GBM cells were differentially regulated, including CHI3L1 and FN1 (Figure 4D). These data indicate NUDT21 as an upstream regulator of NF-κB signaling that could affect MES identity in GBM cells.

To validate the results of the microarray analysis, RT-qPCR and western blotting were conducted, and both showed that, compared with NFKBIZ in shCtrl-transfected cells, NFKBIZ was upregulated in NUDT21-silenced U87MG cells. And the result of t-test further confirmed the statistical significance of difference between Ctrl group and shNUDT21 group or between shCtrl group and shNUDT21 group (Figures 5A,B). To confirm whether NUDT21 regulates CHI3L1 and FN1 expression, we knocked down the expression of NUDT21 in U87MG using shNUDT21. The results indicated that knockdown of NUDT21 in U87MG cells decreased the expression of MES identity genes CHI3L1 and FN1 (Figures 5C,D). In addition, since NFKBIZ regulates NF-κB signaling by acting as an inhibitor of NF-κB (Yamazaki et al., 2001), we further examined the downstream NF-κB expression (Figure 5E). NFKBIZ could bind to the promoter region of secondary response genes by forming a complex with either p50 homodimers or p50–p65 heterodimers of NF-κB to exert transcriptional regulation. Although NFKBIZ has demonstrated an inhibitory role (Yamazaki et al., 2001), its function as a transcriptional activator dominates. Besides, the role of NFKBIZ and NF-κB in the shNUDT21-mediated downregulation of cell proliferation have also been determined (Figure 5F). Cell proliferation assay indicated that NFKBIZ silencing could reverse the effect of NUDT21 in human glioma, and NF-κB nuclear translocation induced by LPS could also reverse the anticancer effect of NUDT21 knockdown. All these indicated that both NFKBIZ and NF-κB played a role in the shNUDT21-mediated reduced proliferation. Furthermore, to further determine the role of NUDT21 in the NF-κB signaling, the regulating effect of NFKBIZ on the NUDT21-mediated downstream factor CHI3L1 was detected. Western blotting showed that siCHI3L1 reversed the decreased expression by NUDT21 knockdown (Figure 5G).