Normal human brain specimens (n = 6) and fresh glioma specimens (n = 19) were obtained from the Department of Neurosurgery, First Affiliated Hospital of Harbin Medical University. The study protocol was approved by the Clinical Research Ethics Committee of Harbin Medical University. All patients provided written informed consent, and the study was conducted in accordance with the Declaration of Helsinki.

The U251, U87, and HUVEC cell lines were provided by the China Infrastructure of Cell Line Resource (National Science and Technology Infrastructure) and cultured in Dulbecco’s Modified Eagle medium (RNBK0465, Sigma-Aldrich, St. Louis, MO, USA) supplemented with 10% fetal bovine serum (AllBio Science, Taichung, Taiwan). MG132, chloroquine (CQ), and PECT were from MedChemExpress (Monmouth Junction, NJ, USA).

The cells were added to a 96-well plate at a density of 5,000 cells/well. At different time points, 10 μL of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide (MTT; cat# HY-15924, MedChemExpress) dye was added to each well and incubated for 4 h. Finally, we replaced the medium with 150 μL of dimethyl sulfoxide and measured the cell viability (Infinite M200 PRO; Tecan Trading AG, Männedorf, Switzerland). The GBM cells were centrifuged and evenly inoculated into a six-well plate at a density of 800 cells/well. After 14 days of culture at distinct concentrations of PECT, the colonies were measured using the ChemiDoc™ MP imaging software (Bio-Rad Laboratories, Inc., CA, USA).

After fixation, permeabilization, and sealing, the treated cells and paraffin-embedded glioma tissue sections were incubated with primary antibody and fluorescent secondary antibody, and images were observed with a fluorescence microscope (Lionheart FX; BioTek, Beijing, China). The following primary antibodies were used: CDK1 (BF0091, 1:500, Affinity, USA), p62 (A19700, 1:100, Abclonal, Wuhan, China), LAMP2 (66301-1-lg, 1:100, Proteintech, Hubei, China), Ki67 (A2094, 1:100, Abclonal, Wuhan, China).

GBM cells were interfered with siRNA, plasmid, and lentiviruses. The siRNA-NC/RRM2 and RRM2 overexpression plasmid pEnter-NC/RRM2 were from Miaolingbio (Wuhan, China) and lentiviruses were from GeneChem (Shanghai, China). The sequences of siRNA-NC/RRM2 and LV-shNC/LV-shRRM2 are shown in Table 1 .

U251 cells were first fixed with 2.5% glutaraldehyde at 4°C, and post-fixed with 1% osmium tetroxide. The immobilized cells were then dehydrated in increasing concentrations of ethanol and acetone. Finally, Transmission Electron Microscopy (TEM) was used to observe autophagic vesicles.

For Western Blot (WB) analysis, the tissues and treated cells were first extracted and lysed. After separating the proteins using 12.5% sodium dodecyl sulfate-polyacrylamide gel electrophoresis, they were transferred to a polyvinylidene fluoride membrane, which was blocked and then incubated with primary antibodies and fluorescent-dye conjugated secondary antibodies. Protein bands were imaged using a ChemiDoc™ MP Imaging System (Bio-Rad, Hercules, CA, USA). The following primary antibodies were used: RRM2 (BS7520, 1:1,000, BioWorld, Nanjing, China); LAMP2 (66301-1-lg, 1:1,000, Proteintech, Hubei, China); LC3B (83506S, 1:1,000, Cell Signaling Technology, USA); CCNA2 (A7632, 1:1,000), CCNB1 (A2056, 1:1,000), CDK1 (A0220, 1:1,000) and p62 (A19700, 1:1,000) were purchased from Abclonal (Wuhan, China).

The treated cells were first digested with 0.25% trysin, centrifuged, and then fixed with 70% ethyl alcohol at 4°C. The next day, the digested cells were stained using a cell cycle analysis kit (P0010, Beyotime, Shanghai, China). Finally, a flow cytometer (Agilent NovoCyte, China) was used to measure the cell cycle distribution.

After treatment under different conditions, U251 cells transfected with RFP-GFP-LC3 lentivirus were fixed and used to measure autophagy flux. The fluorescence intensity was analyzed using a FluoView FV300 confocal microscope (Olympus Corporation, Tokyo, Japan).

Total RNA was extracted from the clinical samples or treated cells using TRIzol (Sigma-Aldrich), and then one microgram of total RNA was reverse-transcribed into cDNA using a ReverTra Ace qPCR RT Kit (Toyobo, Osaka, Japan). Quantitative Real-Time Polymerase Chain Reaction (qRT-PCR) was performed on triplicate samples in a reaction mixture of SYBR Green (Roche, Basel, Switzerland) with a Gene Amp PCR System 9700 (Thermo Fisher Scientific, Waltham, MA, USA). The data were normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) using the ΔΔCt method. The primer sequences for GAPDH, RRM2 and CDK1 are shown in Table 1 .

First, the primary antibody was added to Protein A/G Plus-Agarose and incubated at 4°C for 2 h, and then the cell protein lysates were added and incubated at 4°C overnight. We next separated the magnetic beads, collected the supernatant, and performed WB.

Each tumor bearing mice brain tissue sample was mixed with methanol (containing 5 μg/mL 2-chloro-L-phenylalanine as an internal standard) and mixed with a vortex mixer for homogenization. The samples were centrifuged and transferred to sampler vial. An in-house quality control (QC) sample was prepared by mixing equal amounts of each sample. We performed Liquid Chromatography-Mass Spectrometry (LC-MS) analysis of the QC samples. The raw data was converted to a common format, and the degree of aggregation of QC samples was assessed using the principal component analysis modeling method and PCA modeling method was used to check the aggregation degree of QC simples.

BALB/c nude mice were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China). For each mouse, 5 × 10⁶ U251 cells were subcutaneously implanted into the right flank, or 1 × 10⁶ U251 cells were implanted into the brain. The mice were randomly divided into control and PECT groups (six mices per group). PECT (25mg/kg every other day, prepared as a 150 mM stock solution in DMSO and stored at ‐20°C) diluted with 100 μL PBS was administered intraperitoneally for 3 weeks, whereas the control group treated with the same amount of DMSO diluted with 100 μL PBS. Additionally, 1 × 10⁶ U251 cells transfected with LV-shNC or LV-shRRM2 per mice were implanted into the brain. All animal study protocols were approved by the Animal Experiments Ethics Committee of Harbin Medical University and the study was conducted in accordance with the Declaration of Helsinki.

After dehydration and paraffin embedding, the mice brain tumor and organ sections were stained with Hematoxylin and Eosin (H&E). For Immunohistochemical (IHC) assay, tumors were first formalin-fixed, paraffin-embedded, and sliced, and then the sample sections were immunostained with primary antibodies and an anti-mouse/rabbit secondary antibody.

We used DESeq2 or edgeR to analyze differences in expression and identify enriched functional terms and pathways (via Gene Ontology [GO] and Kyoto Encyclopedia of Genes and Genomes [KEGG]). The clinical information and mRNA sequencing data of the 1,038 patients with glioma were obtained from the Chinese Glioma Genome Atlas (CGGA) database http://www.cgga.org.cn/index.jsp, which included data from 625 low-grade glioma (LGG), 388 GBM, and 20 non-glioma patients. OS was determined among 592 LGG and 374 GBM cases, and the median was used as the demarcation point.

Differences between two groups and multiple groups were estimated using Student t-test and one-way analysis of variance, respectively. The statistical significance of OS between different groups was evaluated by log-rank test. All statistical analyses were carried using GraphPad Prism version 7.0 software (GraphPad, Inc., San Diego, CA, USA), and a value of p < 0.05 was considered to indicate statistically significant results.

The 2D structure of PECT is presented in Figure 1A . Firstly, human GBM cells U87 and U251, and human normal cell HUVEC were used to verify the anti-proliferation effect of PECT. MTT results showed that PECT inhibited GBM cells proliferation in dose- and time-dependent manners, but not that of HUVECs. The half-maximal inhibitory concentration (IC₅₀) values of PECT for 48 h among U87 and U251 cells were 21.17 and 21.00 μM, respectively ( Figure 1B ). Therefore, the drug concentration closest to IC₅₀ of 20 μM was selected for follow-up experiments. Subsequently, we used clonogenic assays to verify the sensitivity of GBM cells to PECT and founded that PECT can prevent the tumor sphere formation in a dose-dependent manner ( Figure 1C ). Additionally, PECT inhibited GBM cells proliferation as shown using ki67 staining ( Figure 1D ). Therefore, PECT inhibited GBM cells proliferation and can be administered safely in vitro.

The mice accepted the treatment of PECT showed a much smaller brain tumor size than another ( Figure 1E ) and had a significantly prolonged lifespan ( Figure 1F ). In addition, we performed LC-MS assays to detect the concentration of PECT in gliomas and found that PECT can penetrate the blood-brain barrier (BBB) ( Figures 1G, H , Figure S1 ), which was consistent with the outcomes described in the traditional Chinese medicine systems pharmacology database and analysis platform (TCMSP) (http://lsp.nwu.edu.cn/tcmsp.php) ( Table S1 ) (15). These data indicate that PECT had an anti-tumor effect on GBM in vivo.

To determine the mechanism of the anti-tumor effect of PECT on GBM, we screened differentially expressed genes (DEGs) using RNA-Sequencing (RNA-Seq) analysis after PECT treatment, and detected 1,279 DEGs (662 upregulated genes and 617 downregulated genes, |log2 fold-change|>1.0, p<0.05), the top 50 of which are shown in Figure 1I . The functional annotation analysis of the DEGs using GO. DNA replication and cell cycle regulation were the mainly GO terms (p.adjust <0.05) ( Figure 1J ). In addition, KEGG enrichment pathway (p.adjust <0.05) analysis revealed that the DNA replication, cell cycle, lysosomal, and proteasome pathways were the major signs pathways influenced by PECT ( Figure 1K ). Among the top 50 of DEGs, RRM2 was closely correlated with DNA replication, cell cycle, lysosomal and proteasome pathways, so we hypothesized that RRM2 may be a key target gene for the anti-tumor effect of PECT on GBM. To verify the changes of RRM2 expression, we exposed GBM cells to PECT for 48 h and founded that RRM2 mRNA and protein expression gradually decreased with increasing PECT concentrations ( Figures 2A, B ).

Next, we investigated whether PECT regulates autophagy in GBM cells. TEM results revealed that autophagic vacuoles gradually increased after adding PECT to U251 cells from 0 h to 48 h ( Figure 2C ); simultaneously, LC3B-II and LAMP2 protein expression increased and p62 protein expression decreased ( Figure 2E ), indicating that PECT increased autophagic vacuoles in GBM cells by increasing autophagic flux. Further, we performed a RFP-GFP-LC3 transfection assay on U251 cells to re-confirm the above results, and found that both free red puncta (autolysosomes) and yellow puncta (autophagosomes) increased in the merged sections of U251 cells treated with PECT in a time-dependent manner, and the red puncta increased more ( Figure 2D ), validating the increase of autophagic flux.

Subsequently, we investigated whether the autolysosome was involved in RRM2 degradation. chloroquine (CQ), a late-stage autophagy inhibitor, of which low dose that can partially inhibit pH value increase and block protein degradation in the autolysosome, but has no effect on cell activity (16), was used. The results showed that 2 μM CQ did not affect the cell viability of GBM ( Figure S2 ), but increased p62 and LC3B-II protein expression, indicating inhibition of autophagic flux ( Figure 2F ). Therefore, 2 μM CQ was selected. An increased number of yellow puncta but a decreased number of free red puncta was observed in the merged sections of the PECT+CQ-treated groups compared to the PECT-treated groups ( Figure 2G ); simultaneously, the reduction of the p62 and RRM2 protein expression was partially recovered, but the LC3B-II protein expression further increased ( Figure 2F ), indicating that CQ blocked the autophagic flux induced by PECT and inhibited RRM2 protein degradation. As expected, immunofluorescence assay of LAMP2 [a lysosomal membrane marker that is involved in autophagy and critical for some proteins degradation in lysosomes (17)] and RRM2 showed that PECT treatment resulted in more LAMP2 protein expression but less RRM2 protein expression in U251 cells, and increased the cytoplasmic co-localization of LAMP2 and RRM2 ( Figure 2H ), further indicating that RRM2 depends on autolysosome pathway degradation.

Additionally, we verified that the proteasome is also involed in RRM2 protein degradation, which can be partly reversed by the proteasome inhibitor MG132 ( Figure 2F ).

Overall, our in vitro data confirmed that PECT treatment can increase autophagy flux in GBM cells, and RRM2 protein reduction after PECT treatment is not only dependent on transcriptional inhibition but also proteasomal and autolysosomal degradation.

To assess the effect of RRM2 on GBM, we used siRNA to interfere with RRM2 expression in GBM cells. The siRNA efficiency was shown in Figure 3A ; similarly, RRM2 knockdown not only decreased RRM2 protein expression ( Figure 3B ) but also suppressed GBM cells viability ( Figure 3C ). Moreover, we founded that the proportion of G2/M phase cells increased ( Figures 3D, E ) and CDK1 protein expression was decrease with RRM2 knockdown ( Figure 3F ), but CCNA2 and CCNB1 protein expression no change ( Figure S3 ). In addition, we founded that the expression of CDK1 mRNA was also not change ( Figure S4 ). These results demonstrated that G2/M cell cycle arrest is caused by the decrease of CDK1 protein, which is consistent with previous reports (18, 19).

Subsequently, RRM2 knockdown was induced by RRM2 lentivirus in U251 cells. After testing the transfection efficiency ( Figure S5 ), lentivirus-treated cells were implanted into the brains of 5-week-old nude mice. Mice treated with RRM2 lentivirus showed much smaller tumor volumes compared to controls ( Figure 3G ) and exhibited significantly prolonged survival time ( Figure 3I ). IHC analysis showed that compared with the LV-shNC group, the protein expression of CDK1 and RRM2 decreased in LV-shRRM2 group ( Figure 3H ).

Taken together, our results revealed that RRM2 knockdown could induce G2/M phase cell cycle arrest by reducing the level of CDK1 protein and inhibit GBM cells proliferation in vitro and in vivo.

Firstly, we founded that RRM2 knockdown can promote the autophagic vacuoles accumulated in U251 cells ( Figure 4A ). Simultaneously, a RFP-GFP-LC3 transfection assay revealed that both free red puncta and yellow puncta increased in the merged sections of si-RRM2#1 U251 cells compared with that in si-NC U251 cells, but the free red puncta increased more ( Figure 4B ), indicating the autophagic flux increased.

Subsequently, we sought to determine whether autophagy is required for RRM2-mediated cell cycle progression. The data showed that RRM2 knockdown promoted G2/M phase cell cycle arrest, while this effect was reversed by CQ, an inhibitor of autophagic flux that prevents autophagosome-lysosome fusion and lysosomal protein degradation ( Figures 4C, D ). Interestingly, the change in CDK1 protein level was consistent with the cell cycle distribution ( Figure 4E ). These data indicate that the blockage of G2/M phase cell cycle induced by RRM2 knockdown is dependent on the function of RRM2 in autophagy.

To further investigate the mechanism regarding whether the degradation of CDK1 protein depends on the autolysosomal pathway during RRM2 knockdown, we performed Co-Immunoprecipitation (Co-IP) assays to analyze the interaction between CDK1 and p62, an adapter protein, is required for some protein degradation through the autolysosome pathway (20). As shown in Figure 4F , p62 was able to interact with CDK1. Additionally, immunofluorescence analysis showed that CDK1 could co-localize with p62 in U251 cell, but the co-localization of p62 with CDK1 decreased after RRM2 knockdown ( Figure 4G ). Thus, our results demonstrated that RRM2 knockdown induced cell cycle arrest in GBM via promoting CDK1 protein degradation by increasing autophagic flux in vitro.

To verify whether RRM2 mediates the anti-tumor effect of PECT on GBM, we overexpressed RRM2 and exposed them to PECT, and founded that GBM cells viability ( Figure 5A ) were partially rescued with RRM2 expression upregulation ( Figures 5B, C ). Similarly, the increased proportion of G2/M phase GBM cells ( Figures 5D, E ) and the downregulation of p62 and CDK1 protein and upregulation of LAMP2 and LC3B-II protein caused by PECT were partially recovered with RRM2 expression upregulation ( Figure 5F ). Overall, our results showd that PECT can inhibit GBM proliferation and promote G2/M phase cell cycle arrest as well as increase autophagic flux by decreasing RRM2 expression in vitro.

In vivo, our results showed that PECT suppressed tumor proliferation, accompanied by reducing tumor weight ( Figures 6A, B ) and tumor weight/mouse weight ( Figure 6C ), and downregulating RRM2 mRNA ( Figure 6D ) and protein ( Figure 6E ) expression compared to that in controls. However, the body weight of mice in each group did not differ significantly ( Figure 6F ). Moreover, we founded that PECT itself did not induce additional toxicities by evaluating morphological changes in tissues in all groups ( Figure 6G ). IHC results showed that compared with the control group, the protein expression of LC3B and LAMP2 increased while that of Ki67, RRM2, p62, and CDK1 decreased in xenograft tumors administered with PECT ( Figure 6H ). Taken together, PECT exerts an anti-tumor effect in GBM and can be safely administered in vivo.

To verify the clinical significance of RRM2 in glioma patients, firstly, we performed IHC staining analysis, and founded that compared with normal human brain tissues, RRM2, CDK1, and p62 protein expression had increase in distinct grades of glioma, especially in GBM tissues ( Figure 7A ). Additionally, according to the CGGA data and qRT-PCR results, compared with non-tumor brain tissues, RRM2 mRNA expressions gradually increased with increasing the degree of glioma malignancy ( Figure 7B ). WB analysis revealed that RRM2 and CDK1 protein expression in human GBM tissue was higher than that in LGG and normal human brain tissues, but no significant difference between LGG and normal human brain tissues, indicating that RRM2 may be a predictive marker of GBM ( Figures 7C, D ).

We confirmed that p62 can directly interact with CDK1 and is an adapter for the degradation of CDK1 protein in autolysosome in vitro. Therefore, we performed immunofluorescence staining, and founded that p62 and CDK1 protein are co-located in the lesional cells from human GBM cases, which indicated that p62 and CDK1 protein could interact directly in vivo ( Figure 7E ). More importantly, we examined the clinical data of patients with glioma in the CGGA database and revealed that RRM2 expression was inversely correlated with the OS time, indicating that RRM2 may be a prognosis marker of GBM ( Figure 7F ).