CCNB2/SASP/Cathepsin B & PGE2 Axis Induce Cell Senescence Mediated Malignant Transformation

Glioma is the most frequent and aggressive adult brain tumor with maximum mortality. However, the gene alteration and mechanism underlying malignant transformation of glioma remain largely unknown. We aimed to find key factors regulating tumor progression and malignant transformation of glioma. Here we compared the gene expression profiles of 693 glioma patients by HGG vs. LGG model, and identified a key factor CCNB2 for malignant transformation in glioma. CCNB2 induced a senescence-associated secretory phenotype (SASP) of glioma cells, and the malignant progression, such as invasion and excessive proliferation was mediated by secreting SASP cytokines, Cathepsin B and PGE2. These findings demonstrated a previously undiscovered link between senescence, CCNB2/SASP/Cathepsin B & PGE2 axis and malignant transformation in glioma. This might provide novel insights on developing new therapeutic regimens for abrogating aggressiveness of glioma.


Introduction
Glioma is a most prevalent primary tumor of central nervous in adults [1,2]. According to the malignancy grade WHO I-IV and histological subtypes, glioma was categorized to Low-grade glioma (LGG) and High-grade glioma (HGG). HGG is highly malignant, easily recurrent, and the most lethal and difficult to treat, especially for glioblastoma (GBM). The 5 year overall survival of malignant glioma is no greater than 35% [2]. High malignancy deteriorates the prognosis of patients and enhances the financial burden of society. How to find the key in malignant progression and restrain the aggressiveness of HGG is very important. Therefore, researches regarding the credible therapeutic targets and precise molecular mechanisms of malignant transformation in glioma have drawn extensive attention.
To solve this problem, we compared the gene expression profiles of 693 glioma cases by HGG vs.
LGG model and chose 45 core genes in Top 1 module. In PPI network, a crucial core factor-CCNB2 and related pathway-cellular senescence were more attractive considering its high MCODE_score and intensive interactions in signaling pathway. Cellular senescence is a general phenomenon to prevent limitless cell proliferation. Its notable feature is the sustaining metabolic activity in growth retardation [3]. Irreversible cell cycle arrest, cell swelling, Senescence-associated β-Galactosidase (SA-β-Gal) activation and less response to mitogen are hallmarks Ivyspring International Publisher of senescence, which can be triggered by environmental stimuli [4,5]. Thus, induction of senescence is a potent protection against tumorigenesis and tumor progression originally. However, the defense mechanism of cellular senescence is a double-edged sword. Aged fibroblasts with senescence-associated secretory phenotype (SASP) are capable of remodeling a microenvironment for tumor development via secreting various cytokines [6,7]. Moreover, senescent tumor cells also present the pro-tumor potential by paracrine effects on surrounding non-senescent cells [8]. Therefore, we further studied the role of CCNB2 on secreting of SASP cytokines and cell senescence mediated tumor aggressiveness. CCNB2, Cyclin B2, is a crucial component in cell cycle checking. It was reported promoting tumor malignancy such as radio-resistent in HGG (glioblastoma) [9], which was somewhat consistent with our findings.
To the best of our knowledge, this study is the first one to screen key factors of malignant transformation in glioma, and proved crucial modulating roles of CCNB2/SASP/Cathepsin B & PGE2 axis on cell senescence mediated tumor aggressiveness ( Figure 1). Revealing the complicated mechanisms within malignant progression of glioma will provide novel ideas on development of individualized therapeutic regimens.

differential expression genes (DEGs) were screened by HGG vs. LGG model
Gene expression profile, clinical characteristics and demographics of 693 glioma patients were screened from The Cancer Genome Atlas (TCGA) (https://www.cancer.gov/about-nci/organization/ ccg/research/structural-genomics/tcga); (https:// xenabrowser.net/datapages/) and The Chinese Glioma Genome Atlas (CGGA) (http://www.cgga.   [10]. To assess the differences between HGG and LGG patients, the gene expression profiles of 693 glioma cases were compared using HGG vs. LGG model. There were 510 DEGs (FC>2, P<0.05) chosen out. The expression data for top 20 up/down-regulated genes in HGG and LGG groups were shown by hierarchical cluster heatmap (Figure 2A). Among all the DEGs, 470 genes containing DLK1, MYBPH, HOXD11, IGFBP2, VEGFA exhibited significantly up-regulation and 40 genes such as GLI1, GCGR, PTGER1showed notably down-regulation. We further performed GO and pathway analyses to illustrate the underlying function of HGG-related DEGs. The enrichment factor (EF) indicated the significance of particular function. In  Figure 2B, C). GSEA enrichment also suggested that cell cycle, DNA damage and G1 to S phase control are crucial in malignant transformation of glioma ( Figure 2D).
To better understand the advanced function of the DEGs, we conducted the protein-protein interaction (PPI) network analysis by STRING online tools (https://string-db.org/) and Cytoscape 3.8.0 ( Figure 3A, B). HGG-related PPI network was composed of 4 major modules, containing 187 nodes and 1486 edges (PPI enrichment P<1.0e-16). Average node degree and average local clustering coefficient were 15.9 and 0.61, respectively. The top 1 module had 968 edges and 45 nodes, including CCNB2, CCNB1, BIRC5, MELK, TPX2 and CDCA8 ( Figure  3C). The MCODE_scores of nodes in module 1 were the highest among all the modules (all score>25). They had the most intensive interactions in PPI network for HGG-related DEGs, suggesting the crucial cell-function-regulating roles of these nodes. Among them, combined_score (0.821) and MCODE_score (37) of CCNB2 were the biggest. Its FC was 3.55 ( Figure 3D, E). Therefore, we chose CCNB2 for the further study. The potential impact of these malignant phenotype related genes for glioma deserved advanced investigations.

Expression correlation and survival analyses for key gene CCNB2 in glioma malignant transformation
To depict the potential function of CCNB2 in malignant transformation and malignant phenotype maintaining of glioma, we evaluated the expression correlation and prognostic value of CCNB2 and its related genes in PPI network. According to the KEGG pathway analysis, CCNB2 belonged to cellular senescence pathway (KO: 04218). Other PPI node genes in this pathway were CCNB1, CDK1, CXCL8, IGFBP3, MYBL2, SERPINE1 (as listed and indicated with arrows in PPI network, Figure 4A). Interestingly, IGFBP3 was just the cell marker for senescence. This finding suggested that CCNB2 was closely related with modulation of cell senescence. Expression correlation analyses also confirmed this point. CCNB1, CDK1, IGFBP3 and MYBL2 was significantly correlated with CCNB2 in expression level, with correlation coefficient 0.869, 0.849, 0.422 and 0.792 (r>0.3, P<0.05), respectively ( Figure 4B).
Moreover, we evaluated the clinical predicting value of these genes. 693 Glioma patients with both gene expression and survival data were categorized into high or low expression group. Kaplan-Meier curves showed that high expression of CCNB2 significantly predicted poor prognosis for overall survival and recurrent-free survival (RFS) by log-rank test (HR>1, P=0.000<0.05) ( Figure 4C). The upregulation of CCNB1, CDK1, CXCL8, IGFBP3, MYBL2, SERPINE1 also notably indicated poor overall survival of glioma patients (HR>1, P=0.000<0.05).

CCNB2 facilitates G0/G1-phase arrest and cell swelling of glioma cells
Pathway analysis suggested that CCNB2 was a pivotal factor in cell senescence. Thus, we focused on this cellular biological process to perform deep explorations. Senescence is a general phenomenon to prevent limitless cell proliferation. Its notable feature is the sustaining metabolic activity in growth retardation. Cell cycle arrest, swelling and highactivity for SA-β-gal are important manifestations of senescent cells [3].  To confirm the role of key factor CCNB2 in senescence of glioma cells, we firstly detected the effect of CCNB2 on cell cycle and morphology. After ectopic expression of CCNB2 and cultivated for 7 days, glioma cells were stained with PI and examined on a flow cytometer. As shown in Figure 5A and 5B, CCNB2 induced an obvious G0/G1-phase arrest in contrast to Lenti-Vector group (79.9±1.19 vs. 66.1±1.08%, P=0.000), while the difference was not significant for G2/M-phase (6.5±0.51 vs. 5.6±1.45%, P=0.368). Furthermore, cell morphology was observed under an upright fluorescence microscope. Cells were significantly presenting swelling status in Lenti-CCNB2 group, with dilatated nucleus compared to Lenti-Vector group (23.7±2.28 vs. 11.8±2.05μm, P=0.003) ( Figure 5D, E). These data suggested that CCNB2 notably facilitated G0/G1phase arrest and cell dilatation, which contributed to cellular senescence. To assess whether senescence exactly occurred in the cultures, we examined the expression of IGFBP3, the typical cell marker for senescence. Consequently, IGFBP3 + cells were obviously increased in Lenti-CCNB2 group relative to control (41.9±3.56 vs. 8.1±1.66%, P=0.000) ( Figure 5C, F). This finding confirmed the aging phenotype of cell cycle arrest and morphological changes.

CCNB2 induces senescence and a SASP of glioma cells
To firmly establish the senescence signatures of CCNB2-overexpressed glioma cells, we tested two chromatin markers and a metabolic marker for senescence.
The chromatin markers are phosphorylated histone H2A.X (Ser139) and heterochromatin protein 1β (HP1β) [11,12]. The accumulation of pH2A.X (Ser139) reveals high frequency of DNA double strand breaks (DSBs), an irreversible DNA damage [13]. HP1β is a crucial component of heterochromatin foci. Using Immunostaining, we found that the pH2A.X (Ser139) foci were dramatically increased in CCNB2 ectopically expressed cells compared to control (369±57 vs. 152±24cells/1000 cells, P=0.004), indicating these cells becoming senescent ( Figure 6A, B). Simultaneously, a similar phenomenon was observed in HP1β heterochromatin foci test (Lenti-CCNB2 vs. Lenti-Vector: 172±18 vs. 74±11 cells/1000 cells, P=0.001) ( Figure 6C, D). We also examined the activity of the metabolic marker SA-β-Gal for senescence in different glioma cells. Cytochemical reaction results revealed that CCNB2 overexpression significantly led to an increase of SA-β-Gal activity (Lenti-CCNB2 vs Lenti-Vector: 3.7±0.25 vs. 0.8±0.18folds, P=0.000) ( Figure 6E, F). According to the above, high level of CCNB2 definitely induced senescence of glioma cells. These findings were in consistent with the bioinformatics analyses for glioma TCGA data. However, we were facing with a crucial question: how did CCNB2-related senescent cells work in malignant transformation for glioma?  Recently, increasing evidences showed that senescent tumor cells were capable of stimulating tumorigenesis and tumor progression via secreting various cytokines, which was called SASP. These cytokines would present paracrine effects on surrounding tumor cells [8,14]. To assess whether CCNB2 induced a SASP to promote the malignant transformation of glioma cells, we compared the cytokines concentration of medium from different groups by ELISA. These included cytokines are essential components of senescence secretome [8]. As a result, Cathepsin B, bFGF, TMP2, PGE2 were significantly upregulated at 36 or 48h after senescence occurring (i.e., the 7 th day of CCNB2 overexpression was recorded as 0 hour in Figure 7). For instance, Cathepsin B at 48 h, the values of CCNB2 group in contrast to control was 230.0±16.70 vs. 144.7 ±9.29 pg/ml/10 6 cells, P=0.002. These data corroborated that CCNB2 exactly induced a SASP of glioma cells, which might transmit messages for malignant transformation.

CCNB2/SASP/Cathepsin B axis facilitates cell migration and invasion of glioma
To further explore the role of these SASP cytokines in glioma aggressiveness, we examined their effects on cell malignant transformation. Invasiveness and excessive proliferation underlie aggressiveness and are hallmarks of cancer malignancy. With screening, among these cytokines, Cathepsin B was considered as an invasion and metastasis promoter for tumor in generous studies [15][16][17][18]. It was a lysosomal cysteine cathepsin. High level of Cathepsin B was correlated with dissociation from extracellular matrix and enhanced movement of tumor cells. Combined with our studies, Cathepsin B was significantly upregulated in recurrent glioma and HGG, compared to primary glioma and LGG, respectively ( Figure 8A, B). These findings suggested that Cathepsin B might play pivotal roles in glioma metastasis, invasion and thus tumor malignant progression. Then we constructed a coculture cell model to measure the impacts of secreted Cathepsin B on migration and invasion of glioma cells ( Figure 8C, Supplemental Figure 1). In scratch-wound healing assay, horizontal migration detection, coculture with Lenti-CCNB2 cells markedly facilitated the migration in contrast to vector group (69.5±3.91 vs. 42.3±3.57%, P=0.001). However, the migration potential was significantly inhibited by CA074, the inhibitor for Cathepsin B (Lenti-CCNB2+CA074 vs. Lenti-CCNB2: 47.9±5.57 vs. 69.5±3.91%, P=0.005), which suggested that Cathepsin B secreted from CCNB2-related senescent cells promoted tumor migration ( Figure 8D, E). In transwell assay, the increase for vertical migration ability induced with SASP cytokines was remarkably abrogated by CA074 (Lenti-CCNB2+ CA074 vs. Lenti-CCNB2: 160±22 vs. 344±44cells, P=0.003) ( Figure 8F, G). Interestingly, the similar effect of CA074 was observed in invasion potential tests (Lenti-CCNB2+CA074 vs. Lenti-CCNB2: 143±25 vs. 308±61cells, P=0.012) ( Figure 8H, I). Summarily, these senescent cells exhibited high metabolic activity and had the potential to influence the behavior of surrounding cells, which were nonsenescent.
Moreover, we constructed the animal models to investigate the effect of CCNB2/SASP/PGE2 axis on tumor growth in vivo. 24 Male nude mice were divided into 3 groups, bearing U251 xenografts intratumor injected with medium from Lenti-Vector, Lenti-CCNB2 or Lenti-CCNB2+CAY10650 cells. According to the growth kinetic curves, the propagation of tumors was dramatically accelerated by SASP cytokines in medium of Lenti-CCNB2 cells ( (Figure 9F, H). Body weight data of mice in different groups manifested that there was no significant change between them ( Figure 9G). Our study authenticated the regulation role of CCNB2 on SASP cytokines secretion and tumor malignant transformation.

Enzyme-linked immunosorbent assay (ELISA) of SASP cytokines
The concentration of SASP cytokines in medium conditioned by Parental, Lenti-Vector or Lenti-CCNB2 cells was measured applying Human Enzyme-linked Immunosorbent Assay Kits specific for the cytokine (R&D Systems, Minneapolis, MN, USA). All operations were down according to the manufacturer's instructions and the quantification of cytokines was performed on Synergy H1 Hybrid Multi-Mode Microplate Reader (BioTek Instruments, Winooski, VT, USA). The analysis software was Gen 5 TM 3.0 All-In-One Microplate Reader Software (BioTek Instruments, Winooski, VT, USA).

SA-β-Gal activity tests
The activity of SA-β-Gal in Lenti-CCNB2 cells was examined using SA-β-Gal staining kit (Beyotime, Shanghai, China) according to the manufacturer's recommendation. In brief, cells in 6-well plates were fixed, washed and treated by chromogenic substrate X-Gal overnight in 37 °C. After washing, the blue metabolic product of X-Gal, chloro-bromoindigo, indicating the activity of SA-β-Gal, was assessed using brightfield mode of Upright Fluorescence Microscope (Olympus BX53, Tokyo, Japan). Quantification was performed by calculating average integral optical density of the blue product with Image J (NIH, Bethesda, MA, USA).

Transwell migration and invasion assay
Transwell migration and invasion assay was carried out using 24-well Transwell® units (8.0μm pore; Corning Inc., NY, USA). In brief, a coculture cell model for detection of migration and invasion potential was established. 5×10 3 U251 cells in 150μl DMEM+10% FBS were added in upper inserts. To assess the influence of cytokines secreted by senescent or nonsenescent cells on neighboring glioma cells, 1×10 4 Lenti-CCNB2 or Lenti-Vector cells in 500μl DMEM+10% FBS were filled in lower chambers. Different from migration inserts, invasion inserts were pre-coated with Corning Matrigel Basement Membrane Matrix (Corning Inc., NY, USA) (Supplemental Figure 1). After Incubation in 37℃ for 24 h, non-traversed cells in the inserts were wiped away. Migrated and invaded cells were fixed and stained in Crystal Violet Solution (Beyotime, Shanghai, China). Eventually, graphs were captured on an Inverted Fluorescence Microscope (Olympus IX71, Tokyo, Japan). Cells were counted using Image J (NIH, Bethesda, MA, USA). Each test was conducted in triplicate.

3D tumor spheroid assay
After 7 days of Lenti-Vector or Lenti-CCNB2 virus transduction (senescence occurring), the medium was collected and incubated with glioma cells in different groups. 3 days later, for lower layer, 100 μl pre-cooled Corning Matrigel Basement Membrane Matrix (5 mg/ml, Corning Inc., NY, USA) were added in 1well/24 well plate. To solidify the 3D culture Matrigel, the 24 well plates were maintained in 37 °C for 30 min. For upper layer, 1×10 4 cells (100 μl) mixed with 200 μl pre-cooled Matrigel were filled in the well (37 °C, 30 min). At last, 500 μl conditioned medium of different groups was injected. After incubation for 10 days, 3D tumor spheroids were photographed by an Inverted Fluorescence Microscope (Olympus IX71, Tokyo, Japan) and calculated.

Xenograft models construction
BALB/c nude mice (8 weeks old, male) was obtained from Shanghai Animal Center of Chinese Academy (Shanghai, China). Firstly, 1×10 7 U251 cells were flank injected to 3 nude mice. When tumor size was ~100 mm 3 , the tumors was cut into 8mm 3 globules and inoculated to the right flank of nude mice. Secondly, Animals were randomly categorized into 3 groups (n=8/group) while xenografts reached ~50 mm 3 . The medium conditioned by Lenti-Vector, Lenti-CCNB2 and Lenti-CCNB2+CAY10650 cells was intratumorally injected to mice in corresponding group every 2 days. Tumor volume and body weight of these mice were monitored every 5 days. The animal study protocol was approved by the ethics committee of Shanghai Deji Hospital.

Statistical Analysis
Statistical analyses were two-sided and performed by GraphPad Prism 8.0 (GraphPad Software, La Jolla, CA, USA) or Statistical Program for Social Science (SPSS 18.0, Chicago, IL, USA). All data were written as mean ± SD. Variance between different groups was calculated using Student's t test (two groups) or one-way ANOVA (multi groups). Log-rank tests was applied to assess the differences of survival analyses. P<0.05 was identified as statistically significant in the current study.

Discussion
In this study, we identified a key factor CCNB2 and CCNB2/SASP/Cathepsin B & PGE2 axis inducing cell senescence mediated malignant transformation in glioma.
Firstly, with TCGA and CGGA data mining, global gene expression profiles of 693 glioma cases were compared by HGG vs. LGG model. There were 510 malignant transformation-related DEGs extracted out. We further performed GO and pathway analyses to illustrate the underlying function of DEGs. GO analysis indicated Cell division, Cell cycle and so on. For KEGG pathway, malignant transformationrelated DEGs were evidently enriched in protein digestion and absorption, AGE-RAGE signaling pathway, Cellular senescence etc. PPI network analysis generated 4 major modules. The top 1 module included CCNB2, CCNB1, BIRC5, MELK, TPX2 and CDCA8. Interestingly, these core genes are all participating in modulation of cell cycle, cell division or cell senescence. Simultaneously, they were also play roles in malignant progression of tumor [22][23][24][25][26][27]. Among them, CCNB2 was chosen for the functional study due to its most intensive interactions in signaling pathway. The potential impact of this malignant phenotype related factor in glioma deserved advanced investigations.
Subsequently, we evaluated the expression correlation and prognostic value of CCNB2 and its related genes in cellular senescence pathway. CCNB1, CDK1, IGFBP3 and MYBL2 was significantly correlated with CCNB2 in expression level. Survival analysis indicated that high expression of CCNB2 significantly predicted poor prognosis for overall survival and recurrent-free survival (RFS) of glioma. That is to say, CCNB2 was a pivotal regulator of cell senescence, and implicated in malignant transformation.
Additionally, the phenotypes of senescence, such as G0/G1-phase arrest and cell swelling were authenticated in CCNB2 glioma cells. To assess whether senescence exactly occurred, we confirmed the expression of IGFBP3, the typical cell marker for senescence. The chromatin markers, pH2A.X (Ser139) and HP1β, and the metabolic marker SA-β-Gal were also identified in CCNB2-overexpression cells. These evidences were in consistent with the bioinformatics analyses for glioma patients.
Senescence is a double-edged sword. Originally, induction of senescence by environmental stimuli is a potent protection against tumorigenesis and tumor progression [11]. However, aged cells are capable of remodeling a microenvironment for tumor development via secreting various cytokines [6,7]. ELISA data revealed that CCNB2 indeed induced a SASP including various cytokines such as growth factor and proteinases. Cathepsin B, bFGF, TMP2 and PGE2 were all significantly upregulated. To test whether these cytokines transmitted messages for malignant transformation, we examined their effects on cell invasiveness and excessive proliferation. Those were hallmarks of cancer malignancy. With clinical information and expression profiles analysis, we found that Cathepsin B and PGE2 synthase-Cox-2 was significantly upregulated in recurrent glioma and HGG, compared to primary glioma and LGG, respectively. These data indicated that Cathepsin B and PGE2 might play pivotal roles in glioma malignant progression. Scratch-wound healing and transwell assays demonstrated that secreted Cathepsin B from CCNB2-related senescent cells obviously facilitated migration and invasion of glioma cells. Using inhibitor, PGE2 in CCNB2 medium was verified to enhance proliferation potential of glioma in vitro and in vivo, which worked through CCNB2/SASP/PGE2 axis in glioma cells. Conclusively, these senescent cells exhibited high metabolic activity and had the potential to influence the behavior of surrounding cells, which were nonsenescent. Similarly, for epithelial tumor cells, researchers found that senescent fibroblasts could drive tumor progression, such as cell proliferation and vascularization through SASP [28]. This phenomenon was called maladaptive senescence [28][29][30][31].
Summarily, this study demonstrated, for the first time, that CCNB2 functioned as a malignant transformation regulator and a tumor promoter of glioma. We extracted a series of malignant phenotype-related genes from gene expression profiles of 693 glioma patients. Then we identified the key factor-CCNB2 which significantly promoted invasion and excessive proliferation of glioma through CCNB2/SASP/Cathepsin B and CCNB2/ SASP/PGE2 axis. In the development from Low to High-grade glioma, CCNB2 enhanced SASP cytokines production and the deterioration of tumor microenvironment, which significantly accelerated cell malignant transformation and progression of glioma ( Figure 10).
Our findings might arouse novel insights on decoding the complicated mechanism of malignant transformation and developing new therapeutic regimens for glioma. This investigation would also be benefit for the prediction of prognosis, along with target therapy development for the malignant transformation of glioma. In conclusion, the correlation of CCNB2/SASP/Cathepsin B & PGE2 axis and senescence mediated malignant transformation in glioma is valuable enough to warrant advanced explorations.

SASP:
Senescence-associated secretory phenotype; SA-β-Gal: Senescence-associated β-Galactosidase; TCGA: The Cancer Genome Atlas; LGG: Low-grade glioma; HGG: High-grade glioma; DEG: Differential expression gene; FC: Fold change; GO: Gene ontology; KEGG: Kyoto Encyclopedia of Genes and Genomes; GSEA: Gene set enrichment analysis; PPI: Protein-protein interaction; EF: Enrichment factor; CCNB2: Cyclin B2; PGE2: Type 2 prostaglandin endoperoxide. Figure 10. Schematic diagram depicting the potential role of CCNB2 in modulating cell senescence and progression of glioma. In the progression from low to high grade glioma, the malignant phenotype of tumor cells is induced by CCNB2 mediated SASP. CCNB2 might modulate cell senescence via initiating the mitogenic/ stress signals (A-C), thus facilitate the activating of p16/p21 complex, and γH2A.X phosphorylation. Subsequently, it further enhances relative SASP factors production and the deterioration of tumor microenvironment, which significantly accelerates cell malignant transformation and progression of glioma.