E2F2 inhibition induces autophagy to impair metastasis via the PI3K/Akt/mTOR pathway in gastric cancer

Background: E2F2 is a member of the E2F family of transcription factors with important yet incompletely understood biological functions in cancer. In some cancer types, controversial tumor-promoting and tumor-suppressive roles of E2F2 have been reported. However, the biological role of E2F2 in gastric cancer (GC) remains to be determined. Methods: We analyzed E2F2 expression via multiple gene expression databases. The prognostic value of the E2F2 was determined by Kaplan-Meier Plotter and Cox regression. The correlations between E2F2 and cancer immune inltrates were investigated via Tumor Immune Estimation Resource (TIMER). The functions and pathways of E2F2 and its 50 frequently changed genes closely associated with the family members were analyzed using Database for Annotation, Visualization, and Integrated Discovery (DAVID) software. We used immunohistochemistry (IHC), quantitative real-time PCR (qPCR) and western blot to verify the expression level of E2F2 in GC and further studied the effects of E2F2 on PI3K/Akt/mTOR activity; GC cell autophagy, migration, and invasion through wound healing assays, transwell assays, Western blotting, and transmission electron microscopy. Results: We observed that compared with normal gastric tissues/cells, E2F2 is highly expressed in gastric cancer tissues and cells in both the public datasets and in our experimental verication. High E2F2 expression was associated with poorer overall survival (OS). Moreover, E2F2 expression showed strong correlations with diverse immune marker sets in GC. Moreover, E2F2 overexpression promoted GC cell migration and invasion in vitro by inactivating PI3K/Akt/mTOR-mediated autophagy. Conversely, E2F2 inhibition suppressed GC cell migration and invasion in vitro by activating PI3K/Akt/mTOR-mediated autophagy. Conclusions: In conclusion, this study provides multi-level evidence for the importance of E2F2 in gastric carcinogenesis and its potential as a biomarker in GC. We demonstrated that E2F2 is overexpressed in GC and that high E2F2 expression is associated with aggressive tumor features and poorer patient prognosis. Further, our results suggest a potential novel immune regulatory role of E2F2 in tumor immunity. Functionally, we discovered a new role of E2F2 in regulating PI3K/Akt/mTOR-mediated autophagy and the downstream processes of cell migration and invasion. Our results could potentially reveal new targets and strategies for GC diagnosis and treatment. Genome Atlas (TCGA) and various public databases. Using multi-dimensional analysis, we evaluated genomic alterations and functional networks related to E2F2 in GC and explored its role in tumor immunity. Finally, we used GC cell lines to verify the high expression of E2F2 in gastric cancer, and examined the functional role(s) of E2F2 in GC cell invasion, migration and autophagy. Our data highlight the novel role of E2F2 in mediating autophagy via the PI3K/Akt/mTOR pathway, with resulting effects on GC cell metastasis. This study provides multi-level evidence for the importance of E2F2 in gastric carcinogenesis and its potential as a biomarker in GC. We demonstrated that E2F2 is overexpressed in GC and that high E2F2 expression is associated with aggressive tumor features and poorer patient prognosis. In addition, our analyses revealed that the levels of immune inltration and a list of immune markers were signicantly correlated with E2F2 expression in GC. Our analyses provide novel insights into the prognostic role of E2F2 and potential role of E2F2 in the tumor immunology of GC. Functionally, we rstly uncovered a novel mechanism of E2F2 in regulating autophagy at least in part via the PI3K/Akt/mTOR pathway, which in turn regulated GC cell migration and invasion. These ndings call for large-scale clinical research and deeper mechanism studies. Nevertheless, the current results are encouraging, and noteworthy in the eld of identifying promising prognostic biomarkers for GC.

addition, the lack of markers that are speci c for tumor type or disease stage represents a critical gap in the current understanding and treatment of GC.
E2Fs is a family of transcription factor proteins that are considered to be the main regulators of cell growth and proliferation. The main functions of E2F proteins are to regulate the cell cycle, cell differentiation, the DNA damage response and cell death (5). Whereas certain members of E2Fs, such as E2F1, has been extensively studied (6)(7)(8)(9), the biological effects of E2F2 in GC remain largely unknown.
Previous studies have shown that E2F2 plays contradictory roles in the development of tumors. On the one hand, E2F2 can inhibit tumorigenesis by inhibiting cell cycle regulators (10). On the other hand, E2F2 can act as an "activator" to increase target expression and cause cancer (11). In gastric cancer, whether E2F2 exists as an oncogene or a tumor suppressor gene is worthy of our further research. This study aims to ll in the gaps in understanding the contribution of E2F2 to GC development and/or progression, which may form the basis for developing new therapeutic strategies for GC.
Tumor metastasis causes > 90% of cancer deaths, and there is increasing evidence that autophagy is intricately involved in tumor metastasis (12,13). Autophagy is a highly conserved catabolic process, which maintains cellular homeostasis by targeting damaged proteins or organelles to lysosomal compartments for degradation (14,15); this process is important in both normal human physiology and diseases, including cancer (16). In GC, the inhibition of autophagy was associated with enhanced cell proliferation, migration, and invasion, likely involving activation of the phosphoinositide-3 kinase (PI3K)/Akt/mammalian target of rapamycin (mTOR) signaling pathway (17). While autophagy and the PI3K/Akt/mTOR pathway appear to play critical roles in GC metastasis, it is unknown whether E2F2 may be a mediator in these processes.
Here, we investigated E2F2 expression and mutations in data from patients with GC in The Cancer Genome Atlas (TCGA) and various public databases. Using multi-dimensional analysis, we evaluated genomic alterations and functional networks related to E2F2 in GC and explored its role in tumor immunity. Finally, we used GC cell lines to verify the high expression of E2F2 in gastric cancer, and examined the functional role(s) of E2F2 in GC cell invasion, migration and autophagy. Our data highlight the novel role of E2F2 in mediating autophagy via the PI3K/Akt/mTOR pathway, with resulting effects on GC cell metastasis.

ONCOMINE database
The ONCOMINE database (www.oncomine.org) is an integrated online cancer microarray database containing data from DNA-and RNA-sequencing (DNA-seq and RNA-seq, respectively) analyses used to classify the differential expression between common cancer types and the corresponding normal tissues, as well as clinical and pathological data (18). In our study, the transcriptional expression data for E2F2 in different cancer tissue samples and their corresponding adjacent normal samples were obtained from the ONCOMINE database. Differences in transcriptional expression were compared by Student's t test. The pvalue cutoff and fold change threshold were as follows: p-value, 0.01; fold change, 1.5; gene grade, 10%; and data type, mRNA.
2. UALCAN database UALCAN (http://ualcan.path.uab.edu) is an interactive web resource developed based on the grade 3 RNA sequences and clinical data of 31 cancer types in TCGA database (19). In this study, UALCAN was used to analyze the mRNA expression of E2F2 in primary gastric cancer tissue samples and the relationships of these members with clinicopathological parameters. Differences in transcriptional expression were compared by Student's t test (p <0.01).

GEPIA database
The Gene Expression Pro ling Interactive Analysis (GEPIA) database (http://gepia.cancer-pku.cn/) is an interactive web that includes 9,736 tumors and 8,587 normal samples from TCGA and the GTEx projects (20). In this study, GEPIA was applied to verify the E2F2 expression in GC.

LinkedOmics Database
The LinkedOmics database (http://www.linkedomics.org/login.php) is a web-based platform for analyzing 32 TCGA cancer-associated multi-dimensional datasets (21). E2F2 co-expression was analyzed statistically using Pearson's correlation coe cient, presenting in volcano plots, heat maps, or scatter plots. The rank criterion was FDR < 0.05 and 1000 simulations were performed.
6. TIMER database TIMER is a comprehensive resource for systematic analysis of immune in ltrates across diverse cancer types from TCGA (https://cistrome.shinyapps.io/timer/), which includes 10,897 samples across 32 cancer types (23). TIMER applies a deconvolution method (24) to infer the abundance of tumorin ltrating immune cells (TIICs) from gene expression pro les. We analyzed E2F2 expression in STAD and the correlation of E2F2 expression with the abundance of immune in ltrates.

Protein-protein interaction (PPI) network construction and gene enrichment analyses
The STRING database (http://string-db.org/) provides information regarding the signi cant associations of PPIs (25). In this study, the STRING database was used to analyze E2F2 and 50 frequently changed genes closely associated with the family members. We used the Database for Annotation, Visualization, and Integrated Discovery (DAVID) (http://www.DAVID.org) (26) to conduct agonistic gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses of 51 genes, including E2F2 enrichment analysis. GO enrichment analysis can predict gene function based on biological processes (BPs), cell composition (CC) and molecular function (MF), and KEGG can be used to analyze gene enrichment pathways.

Tissue microarrays (TMAs)
We collected 60 fresh GC tissues and adjacent tissues for IHC staining and evaluation. The staining intensity was given a score from 0 to 3 based on the following criteria: 0, no staining; 1, weak staining; 2, moderate staining; or 3, strong staining. All individual patients understood the objectives of this study and provided written informed consent. The Medical Ethics Committee of Qingdao University and the A liated Hospital of Qingdao University approved the collection of clinical materials for research purposes.
9. Cells and culture conditions AGS and HGC27 cell lines were purchased from the cell bank of the Chinese Academy of Sciences and were cultured in RPMI-1640 medium containing 10% fetal bovine serum (FBS) (Gibco, NY, USA). The cells were placed in an incubator at 37 °C with a CO2 concentration of 5%.

RNA extraction and qPCR
Total RNA from cultured cells and frozen tissues was extracted with TRIzol (Invitrogen, Carlsbad, CA). cDNA synthesis was performed using the PrimeScript™ RT Kit (TaKaRa, Otsu, Japan). SYBR Premix EX Taq™ (TaKaRa, Otsu, Japan) was used for qPCR on an FTC-3000p Real-Time PCR system (Funglyn Biotech, Shanghai, China). Relative gene expression was determined by the comparative 2−ΔΔCT method. The PCR primers used are listed in Supplementary Table 1.

Western blotting analysis
Western blotting analysis of protein expression was performed as described previously (27). Brie y, protein lysates (20 µg) were separated using sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and target proteins were detected by western blotting with antibodies against E2F2 (1:1000) and GAPDH (1:50,000). Other antibodies used in this study are listed in Supplementary Table 1. 13. Electron microscopy Cells were xed with 2.5% glutaraldehyde (Solarbio, Beijing, China) for 4 h at 4 °C, washed with phosphate-buffered saline (PBS), and post-xed with 1% OsO4 buffer for 2 h at 4 °C. The cells were then washed and dehydrated in a graded series of ethanol solutions and embedded in Epon812 epoxy resin. Ultrathin (90 nm) sections were collected on copper grids, double-stained with 1% uranyl acetate and 0.2% lead citrate, and examined by a JEOL-1200EX transmission electron microscope (Beijing, China).

Cell migration and invasion assays
Cell migration and invasion were assessed using 8-μm-pore transwell compartments (Corning, NY). For migration assays, 5 × 10 4 cells were suspended in serum-free medium in the upper compartment. After cells were incubated at 37 °C for 24 h, the translocated cells were stained with 0.5% crystal violet for 20 min at room temperature. For invasion assays, Matrigel (BD Biosciences, San Jose, CA) was added to each well according to the manufacturer's instructions before 2 × 10 5 cells were suspended in the upper compartment. After cells were incubated at 37 °C for 24 h, the translocated cells were incubated with 0.5% crystal violet for 20 min at room temperature. For quanti cation, cells were counted under a light microscope (Nikon, Tokyo, Japan) in ve elds (upper, lower, middle, left, right; at ×40 magni cation).

Wound-healing assay
Cell migration was also assessed using the wound-healing assay. Brie y, a wound was generated in a 6well plate by scratching the surface with a 200 μL pipette tip. The wounded areas were photographed under a light microscope (Nikon, Tokyo, Japan) at the time the wound was created (0 h) and at 24 h after. The percentage of wound healing was calculated using the following formula: [1 − (empty area 24 h/empty area 0 h)] × 100.

Statistical analysis
All statistical analyses were carried out using the Graphpad Prism 7.0 Software (GraphPad, La Jolla, CA). Categorical data were analyzed using chi-square (χ2) test. Cox proportional hazard model was used for univariate and multivariate survival analysis. Survival curves were plotted using the Kaplan-Meier method and compared using log-rank test. Analysis of variance and Student's t test were used for comparison among groups. P < 0.05 was considered statistically signi cant. Results:

Elevated expression of E2F2 in GC
We rst evaluated the expression of E2F2 in multiple cancer types. After analyzing more than a dozen cancers including gastric cancer and their corresponding normal tissues, it was found that the mRNA expression of E2F2 in gastric cancer tissues was signi cantly higher than that of adjacent normal tissues ( Figure 1A). Subsequently, we used Oncomine and GEPIA tools to analyze the expression of E2F2 in gastric cancer, and the results showed that compared with normal gastric tissue, the expression level of E2F2 in gastric cancer was signi cantly increased ( Figure B and C). K-M Plotter is used to analyze the effect of E2F2 on OS in patients with gastric cancer. We found that high expression of E2F2 is associated with poor prognosis of patients ( Figure 1D). Further sub-group analysis of multiple clinic-pathological features of TCGA-STAD samples in UALCAN database consistently showed elevated transcription level of E2F2. The expression of E2F2 was signi cantly higher in GC patients than normal controls in subgroup analysis based on gender, age, ethnicity, disease stages, and tumor grade ( Figure 2). Thus, E2F2 expression may serve as a potential diagnostic indicator in GC.
2. Independent prognostic value of E2F mRNA expression in terms of OS in gastric cancer patients we downloaded the clinical and E2F mRNA expression data of 407 gastric cancer patients from the TCGA database (Table 1). We used survival, glmnet and other software packages to integrate and standardize the downloaded raw data and then performed univariate and multivariate Cox survival regression analyses. In the univariate analysis, we found that age, regional lymph node involvement, distant metastasis, pathological stage, and E2F2 are closely related to patient survival ( Table 2). In the multivariate analysis, we found that a higher pathological stage (p= 0.036) and higher E2F2 mRNA expression (p = 0.005) were related to shorter OS of gastric cancer patients (Table 3).

Co-expression analysis and Genomic alterations of E2F2 in GC
To gain the insight of E2F2 biological meaning in GC, the function module of LinkedOmics was used to examine E2F2 co-expression mode in STAD cohort. As shown in Figure 3A, 9963 genes (dark red dots) were shown signi cant positive correlations with E2F2, whereas 10262 genes (dark green dots) were shown signi cant negative correlations (false discovery rate, FDR < 0.01). The top 50 signi cant genes positively and negatively correlated with E2F2 were shown in the heat map ( Figure 3B). We then used the cBioPortal tool to determine the types and frequency of E2F2 alterations in GC based on DNA sequencing data from STAD patients. E2F2 was altered in 16 of 473 (3%) STAD patients ( Figure 3C). These alterations include Missense Mutation in 5 cases, mRNA High in 5 cases, Ampli cation in 4 cases and Deep Deletion in 1cases. As shown in Figure 3D, compared with the diploid group, the E2F2 expression level in the gain or ampli cation group was higher. Next, the frequency distribution of E2F2 CNV patients in different stage and grade groups was presented in Figure 3E and 3F, suggesting the high occurrence and an early-event of E2F2 CNV alteration in GC.

Relationships of E2F2 with immune cells and PD-1/PD-L1
Search the TIMER database to estimate the correlation between E2F2 mRNA expression and immune cell in ltration. As shown in Figure 4A, after adjusting the purity, the expression of E2F2 in GC is compared with M1 cells (Figure 4A), M2 cells ( Figure 4B), Th1 cells ( Figure 4C), Th2 cells ( Figure 4D) and Treg cells ( Figure 4E) Signi cant correlation. Moreover, due to the promising prospect of immunotherapy, we further determined the association between E2F2 expression and PD-1 and PD-L1 ( Figure 4F). We also noticed a positive correlation of E2F2 with PD-1 (r = 0.185, P = 1.49e-4) and PD-L1 (r = 0.279, P = 8.67e-9). Using the data of GISTIC2.0, the effect of different copy status of E2F2 on immune in ltration compared with normal tissues was tested. We observed that E2F2 CNV has a signi cant correlation with the in ltration level of B cells, CD8+ T cells CD4+ T cells, macrophages, neutrophils and dendritic cells ( Figure 4G).

Protein-protein interaction (PPI) network construction and gene enrichment analyses
In order to understand the function of E2F2 in gastric cancer, we used the STRING website to explore the proteins that can interact with E2F2, selected the top 50 genes with high correlation, and used the cytoscape tool to construct an integrated network ( Figure 5A). We found that genes related to the cell cycle, including CCNE1, CCNE2, CDK2, CDK4, CDKN1 and CDKN2A, are closely related to the changes in E2F2. Figure 5B showed 10 hub genes closely related to E2F2. Next, The functions of E2F2 and its 50 neighboring genes were analyzed by GO and KEGG analyses in DAVID. GO enrichment analysis predicts gene function from three aspects, namely, BPs, CCs and MFs. As shown in Figure 5C, we found that in BPs, target genes were mainly enriched in biological regulation metabolic process and response to stimulus; in CCs, target genes were enriched in membrance,nucleus and membrance-enclosed lumen; and the MFs were mainly protein binding,ion binding and nucleic acid binding. KEGG enrichment analysis ( Figure 4D) shows that these genes are most enriched in cell cycle related pathways, which is consistent with the PPI analysis results. It is worth noting that PI3K-Akt signaling pathway also showed a high degree of enrichment, which attracted our attention.
6. E2F2 is a regulator of the PI3K/Akt/mTOR pathway Therefore, we want to further verify the relationship between E2F2 and PI3K-Akt signaling pathway through in vitro experiments. To verify the high expression of E2F2 in GC, we subjected 60 GC tissues and their adjacent tissues to immunohistochemical staining (IHC). Representative images are shown in Figure  6A . The level of E2F2 expression in normal tissues adjacent to cancer tissues was signi cantly lower than that in the cancer tissues. We used qPCR and western blotting of the matched GC and adjacent nontumor frozen tissues and similarly observed signi cant overexpression of E2F2 in GC ( Figure 6B and 6C). We use PCR and Western blot experiments to explore the expression of E2F2 in GC cells. We found that the mRNA and protein expression levels of E2F2 were signi cantly increased in AGS and HGC27 compared to GES-1 cells (Figure 6D and 6E). In the next experiment, we transfected HGC27 cells and AGS cells with GV141-E2F2 and siE2F2, respectively, to better understand the effect of E2F2 on the biological behavior of gastric cancer. High overexpression e ciency and knockdown e ciency were observed in the treated cells ( Figure 6F and G). Subsequently, we evaluated the effect of E2F2 on the PI3K/Akt/mTOR pathway in AGS cells. Western blotting showed that the expression levels of PI3K 110β, phospho-Akt (p-Akt), and phospho-mTOR (p-mTOR) levels increased in HGC27 cells at 24 h after E2F2 overexpression ( Figure 6H). Conversely, following E2F2 knockdown with siE2F2 for 24 h, the expression levels of PI3K 110β, p-Akt, and p-mTOR were markedly decreased in AGS cells ( Figure 6I).

Effects of E2F2 expression levels on GC cell migration and invasion via autophagy mediation
Since E2F2 can regulate the PI3K/Akt/mTOR pathway, which play an important role in affecting autophagy (28,29), we further examined the relationship between E2F2 and autophagy-related protein expression in GC cells that were transfected with GV141-EE2F2 or siE2F2. We found that the expression levels of LC3II decreased most markedly after E2F2 was overexpressed, while P62 protein expression was signi cantly increased ( Figure 7A). Conversely, the expression levels of LC3II increased most markedly after E2F2 was inhibited, while P62 protein expression was signi cantly reduced ( Figure 7B). Interestingly, the number of autophagic vesicles was also signi cantly decreased in HGC27 cells transfected with GV141-E2F2 , and the number of autophagic vesicles was observed to be greatest in AGS cells transfected with siE2F2 ( Figure 7C). Collectively, our data demonstrated that E2F2 overexpression inhibited autophagy and E2F2 inhibition induced autophagy in GC cells. We observed E2F2 overexpression signi cantly increased cell migration and invasion (Figure7D and 7E) in GC cells. Moreover, E2F2 suppression-mediated anti-metastasis effects were reversed when cells were co-treated with siE2F2 and the autophagy inhibitor 3-MA ( Figure 7F and 7G). We also con rmed that co-treatment with 3-MA signi cantly decreased E2F2 inhibition-dependent LC3-II and Beclin1 protein expression and increasd P62 and MMP9 protein expression( Figure 7H). Thus our results demonstrated that E2F2 overexpression promoted GC cell migration and invasion via autophagy inhibition. However, E2F2 suppression induced autophagy, resulting in inhibition of GC cell migration and invasion.

Discussion:
E2F2 plays contradictory roles in the development of tumors. On the one hand, E2F2 can inhibit tumorigenesis by inhibiting cell cycle regulators. On the other hand, E2F2 can act as an "activator" to increase target expression and cause cancer (30). Previous studies have shown that changes in E2F2 protein expression are closely related to the occurrence of different cancers. It has been reported that knocking down E2F2 signi cantly reduces the metastatic ability of breast cancer cells, and mutations in E2F2 are related to tumor proliferation and survival in breast cancer patients (11). E2F2 functions as an oncogene in liver cancer (10), while in prostate cancer, E2F2 inhibits tumor cell proliferation by targeting miRNAs (31). However, The role of E2F2 in gastric cancer has not been fully con rmed. To gain more detailed insights into the potential functions of E2F2 in GC and its regulatory network, We conducted bioinformatics analysis of public data and experimental veri cation to guide the future research of gastric cancer.
Using large independent datasets, we con rmed the signi cantly enhanced expression of E2F2 in GC tissues compared to non-tumor gastric tissues. GC patients with higher E2F2 expression have poorer prognosis than those with lower E2F2 expression; Cox regression models consistently established E2F2 expression level in GC is an independent predictor of the OS of GC patients. Further sub-group analysis of multiple clinic-pathological features of TCGA-STAD samples showed that the expression of E2F2 was signi cantly higher in GC patients than normal controls in subgroup analysis based on gender, age, ethnicity, disease stages, and tumor grade. Thus, our results suggest that E2F2 up-regulation occurs in many cases of GC and deserves further clinical validation as a potential diagnostic and prognostic marker.
A recent study found that genomic alteration, such as somatic mutations in the genes encoding components of the spliceosome, occurs frequently in human neoplasms (32). CNVs can have major genomic implications, such as disrupting genes, altering genetic content, and lead to phenotypic differences (33). Our study found that the copy number of E2F2 was increased in GC and that the major type of E2F2 alteration was Missense Mutation and Ampli cation, which was associated with shorter survival.
The tumor microenvironment is the non-cancerous cells present in and around a tumor, having a strong in uence on the genomic analysis of tumor samples (34).Tumor-in ltrating immune cells are part of the complex microenvironment and are associated with the biological behavior and patient survival in GC (35). In this study, we observed that E2F2 expression was positively correlated with M1/M2 cells, Th1/Th2 cells, Tregs and Macrophages in GC tissues. According to a previous research, E2F2 are differentially required for homeostasis-driven and antigen-induced T cell proliferation in vivo (9).Azkargorta et al. found that E2F2 modulates cellular sensitivity to xenobiotic signals through the negative regulation of the Ahr pathway in T Lymphocytes (36). Our results support the ndings that E2F2 is speci cally correlated with immune in ltrating cells in GC, which suggests that E2F2 plays a vital role in immune in ltration in the tumor microenvironment. Further studies need to be done to elucidate whether E2F2 is a crucial factor in mediating immunity therapy.
Our observed correlation of E2F2 expression with the mTOR pathway (from KEGG analysis) led us to investigate the novel role of E2F2 in autophagy, a process which is modulated by the PI3K/Akt/mTOR pathway. Speci cally, our data demonstrated that E2F2 overexpression upregulated the PI3K/Akt/mTOR pathway, resulting in inhibition of autophagy and promotion of GC cell migration and invasion in vitro. And E2F2 inhibition downregulated the PI3K/Akt/mTOR pathway, resulting in induction of autophagy and inhibition of GC cell migration and invasion. The autophagy inhibitor, 3-MA, partially reversed the effects of E2F2 inhibition on GC cell migration and invasion, further con rming the contribution of autophagy toward these processes. Previous studies have reported that E2F1 can regulate the effect of autophagy (37), and E2F2, as a member of the E2Fs transcription factor family, its role in autophagy has not been con rmed. Our research innovatively discovered that E2F2 may play an important role in autophagy. This may provide new insights into the dual role of E2F2 in cancer cells. It is possible that E2F2 regulates GC growth and progression in a complex manner, perhaps through the modulation of autophagy, which has been shown to be a double-edged sword in cancers, including GC.

Conclusion
This study provides multi-level evidence for the importance of E2F2 in gastric carcinogenesis and its potential as a biomarker in GC. We demonstrated that E2F2 is overexpressed in GC and that high E2F2 expression is associated with aggressive tumor features and poorer patient prognosis. In addition, our analyses revealed that the levels of immune in ltration and a list of immune markers were signi cantly correlated with E2F2 expression in GC. Our analyses provide novel insights into the prognostic role of E2F2 and potential role of E2F2 in the tumor immunology of GC. Functionally, we rstly uncovered a novel mechanism of E2F2 in regulating autophagy at least in part via the PI3K/Akt/mTOR pathway, which in turn regulated GC cell migration and invasion. These ndings call for large-scale clinical research and deeper mechanism studies. Nevertheless, the current results are encouraging, and noteworthy in the eld of identifying promising prognostic biomarkers for GC.

Availability of data and materials
All data generated or analysed during this study are included in this published article. Further details are available from the corresponding author upon reasonable request.

Competing interests
The authors declare that they have no con ict of interest.
Authors' contributions HL and SZ analyzed the data and wrote the manuscript. LS, PW and SL assisted in editing the manuscript. YM, ZL and GW contributed to the design of the study. WQ and JL are the corresponding authors of the paper. All authors read and approved the nal manuscript.