Interferon-Stimulated Exonuclease Gene 20 is A Prognosis and Predictive Biomarker and Correlated with Macrophages M2 Polarization in Glioblastoma: A Pan-Cancer Analysis


 Background: Since the mutation of isocitrate dehydrogenase 1 was confirmed to be different in the tumor microenvironment of multiple cancer types, several researchers have included it in the study of tumor-infiltrating immune cells. Interferon-stimulated exonuclease gene 20 (ISG20) plays a role in the modulation of immunity and inflammation, and its abnormally high expression is conducive for the progression of tumor malignancy. However, whether ISG20 is associated with isocitrate dehydrogenase 1 mutation during tumorigenesis and cancer progression remains unknown to date. Methods: TIMER2.0, ONCOMINE, GEPIA2, TCGA and CGGA were applied to assess the clinical significance of ISG20 and its correlation with tumor-infiltrating immune cells in glioma. cBioPortal and MethSurv databases were used to observe the genetic and DNA methylation changes of ISG20, respectively. Visualization of data was mostly achieved by R language. Quantitative real-time PCR (qRT-PCR) and Immunohistochemistry (IHC) was performed to evaluate the mRNA and protein expression.Results: ISG20 expression was significantly different in most cancers. However, when we combined ISG20 with isocitrate dehydrogenase 1 mutation, we found significant differences only in glioblastoma (GBM). The clinical values of ISG20 in glioblastoma showed that the ISG20 overexpression was strongly associated with a worse overall survival (OS). Additionally, ISG20 was altered in 9% of samples of patients with GBM, and ISG20 expression was negatively correlated with its DNA methylation level. More importantly, ISG20 expression was associated with macrophage alternatively activated (M2) polarization in glioblastoma. Conclusions: ISG20 overexpression is conducive to malignant phenotype but adverse to OS, suggesting that ISG20 is a potential therapeutic target and prognosis and predictive biomarker in patients with GBM.

It has been documented that several regulators such as extracellular vesicles, chemokines, growth factors and cytokines can contribute to the communication between cancer cells and the TME [11,12]. IFNs are cytokines that activate the Janus kinase (JAK)-signal transducer and activator of transcription (STAT) pathway, which regulates the expression of many IFN-stimulated genes (ISGs) [13]. One such ISG, IFNstimulated exonuclease gene 20 (ISG20), also known as HeLa estrogen-modulated, band 45 (HEM45), is a member of RNA exonuclease superfamily. Expression of ISG20 is regulated by double-stranded RNA or IFN types I and II [14] and it is involved in tumorigenesis, progression, angiogenesis, and immune modulation [15,16].
In the past, the central nervous system (CNS) is thought to be an immune-privileged organ, but recently it has been found that the CNS is only immunocompromised, but can still produce an immune response when diseases occur. Microglia and macrophages are critical tumor-associated immune cells and account for 20%-30% in some highly aggressive tumors [17]. As one of the most abundant tumorin ltrating immune cells (TIICs) in the TME, macrophages play a key role in tumor pathogenesis and have two major polarization phenotypes (M1 and M2). M2 macrophages, contrary to the M1 cells that are proin ammatory and possess anti-tumor functions, are mainly involved in anti-in ammatory responses and promote tumor growth [18,19].
However, whether the gene ISG20 is correlated with IDH1 mutation and macrophage M2 polarization in aggressive tumors is unknown. In the present study, our results demonstrated that ISG20 expression was signi cantly different in most cancers. However, when we combined ISG20 with IDH1 mutation, we found signi cant differences only in GBM. Therefore, it is important to assess the correlation between ISG20 and clinical prognosis factors and identify the potential molecular mechanisms in GBM. Additionally, we examined the correlation between ISG20 and survival status, genetic alterations, DNA methylation, immune in ltration, and identi ed the relevant cellular pathways.

Data collection
Non-tumor brain tissues (n = 5), low-grade glioma (LGG; n = 7), and GBM (n =9) samples were collected from the First A liated Hospital of Harbin Medical University. This study was approved by the Clinical Research Ethics Committee of Harbin Medical University and was conducted in accordance with the Declaration of Helsinki. All the participants provided written informed consent. Transcriptome sequencing data of 325 glioma samples and 20 normal brain tissues and clinical data were downloaded from the Chinese Glioma Genome Atlas (CGGA) database. From the Cancer Genome Atlas (TCGA) database (https:// www.cbioportal.org/study/summary?id=GBM_ tcga) and public databases (https:// portal.gdc.cancer.gov/ and http: //www. cbioportal. org/), the following were downloaded: mRNA-seq data of 698 glioma samples and ve normal brain tissues. Clinical data were also downloaded from a public database (http: //www. cbioportal. org/). Data on ISG20 methylation and its copy number alterations (CNAs) were also downloaded from TCGA database (http: //www. cbioportal. org/study/summary? id=GBM_tcga).

TIMER2.0, GEPIA2, and ONCOMINE analyses
The TIMER2.0 website (http://timer.cistrome.org/) was used to identify differences in ISG20 expression between tumors and adjacent normal tissues for the different tumors. It was used to further analyze the expression of ISG20 according to IDH1mut status. The GEPIA2 (Gene Expression Pro ling Interactive Analysis, version 2) web server (http://gepia2.cancer-pku.cn/#analysis) and ONCOMINE database (http://www. oncomine.org) were used to verify the expression.

Genetic alteration analysis
First, we logged into the cBioPortal website (https://www.cbioportal.org/) and selected the Glioblastoma Multforme "TCGA Pan Cancer Atlas" under the mRNA expression z-score setting (RNA Seq V2 RSEM) with a z-score threshold of ±1.8. Next, we examined the mutation type, alteration frequency, and mutated site of the ISG20 gene; then we analyzed the survival outcome of GBM patients with or without ISG20 genetic alteration.

DNA methylation analysis
Using the MethSurv database (https://biit.cs.ut.ee/methsurv/) to screen the CpG island methylation sites of ISG20 in GBM, we validated the relationship between the key methylation sites and overall survival (OS) time of GBM patients. DNA methylation data were obtained from TCGA Genome Data Analysis Center Firehose (http://gdac.broadin-stitute.org/) using the HM450K array. The methylation status of DNA is represented as β-values (0-1).

Functional correlation analysis
Pearson correlation analysis was performed to con rm the correlations between ISG20 and other genes in whole-genome gene pro ling of 169 GBM patients in the TCGA RNA-seq set. Signi cantly related genes (|R| > 0.5, P < 0.05) were mapped to Gene Ontology (GO) functional annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways analyses using the "clusterPro ler" package in R language and Gene Set Variation Analysis (GSVA R package http://www.bioconductor.org/) was used to verify the relationship between ISG20 and the candidate pathways.
2.6. Analysis of Immune Cell Characteristics "CIBERSORT" (R package) was used to con rm the proportions of the 22 TIICs in each glioma patients' sample. Moreover, we validated the correlation of ISG20 expression with macrophages and tumor purity in GBM using the TIMER2.0 database. 2.7. Quantitative real-time polymerase chain reaction (qRT-PCR) Total RNA was extracted from glioma samples using TRIzol reagent (Sigma, USA). cDNA synthesis was carried out using a ReverTra Ace qRT-PCR kit (Toyobo, Osaka, Japan). Quantitative RT-PCR was implemented on triplicate samples in a reaction mixture of SYBR Green (Roche, Basel, Switzerland) with a Gene Amp PCR System 9700 (Thermo Fisher Scienti c, Waltham, MA, USA). The data were normalized by glyceraldehyde 3-phosphate dehydrogenase (GAPDH) using the ∆∆Ct method. The primer sequences for GAPDH and ISG20 were as follows: ISG20, F-5'TGGACTGCGAGATGGTGG3, R-5'GGGTTCTGTAATCGGTGAT3; GAPDH, F-5'CACCCACTCCTCCACCTTTGA3, R-5'ACCACCCTGTTGCTGTAGCCA3′.
2.8. Immunohistochemistry (IHC) for ISG20 expression For the IHC assay, the tumor tissues were immediately placed in 10% formalin for xation, followed by dehydration, para n embedding, and sectioning. Next, the sections were immunostained with primary antibodies and an anti-mouse/rabbit secondary antibody. Anti-ISG20 antibody (BIOGOT, BS72481) was used at a dilution of 1:100.

Statistical analysis
Differences between two groups and multiple groups were estimated using the Student's t-test and oneway analysis of variance, respectively. Pearson's correlation coe cients were used to analyze the relationships between variables, and the statistical signi cance of the prognostic value between different groups was evaluated using the log-rank test. All statistical analyses were performed using GraphPad Prism 7.0 and R 3.6.2. P < 0.05, was considered statistically signi cant. *, P <0.05; **, P <0.01; ***, P <0.001; ****, P <0.0001.

Differential expression of ISG20 in various tumors and in different IDH1mut states
Based on the TCGA dataset, we used the TIMER2.0 website to investigate ISG20 mRNA expression in various cancer types and found that ISG20 expression was signi cantly different when most tumors were compared with normal tissues. With breast invasive carcinoma (BRCA), cholangiocarcinoma (CHOL), esophageal carcinoma (ESCA), GBM, head and neck squamous cell carcinoma (HNSC), kidney renal clear cell carcinoma (KIRC), kidney renal papillary cell carcinoma (KIRP), liver hepatocellular carcinoma (LIHC), lung adenocarcinoma (LUAD), and uterine corpus endometrial carcinoma (UCEC), ISG20 expression was signi cantly upregulated. ISG20 expression was downregulated in colon adenocarcinoma (COAD), kidney chromophobe (KICH), lung squamous cell carcinoma (LUSC), rectum adenocarcinoma (READ), skin cutaneous melanoma (SKCM), and thyroid carcinoma (THCA). These results suggest that ISG20 may have different regulatory mechanisms in different tumors (Fig. 1A). Next, by combining ISG20 with IDH1mut, we found that there were signi cant differences only in GBM, lower grade glioma (LGG), and prostate adenocarcinoma (PRAD) (Fig. 1B-E, Table. S1). To further verify ISG20 expression between normal samples and tumor tissues of GBM, LGG, and PRAD, we examined the GEPIA2 website and found that the expression level of ISG20 was statistically different only in GBM. These results indicate that ISG20 could play crucial roles in the tumor progression of GBM.

Transcription levels of ISG20 in GBM Patients
To verify ISG20 mRNA expression between normal human brain tissues and GBM using ONCOMINE databases, our results showed that the transcription levels of ISG20 were signi cantly elevated in patients with GBM in four datasets ( Fig. 2A-E). Next, a meta-analysis of the mRNA expression levels of ISG20 between GBM and non-tumor brain tissues was performed using the ONCOMINE database. ISG20 mRNA expression was markedly upregulated in GBM tissues (P=1.31E-5) compared to those in non-tumor brain tissues, and the median rank of ISG20 was 945.5 in GBM tissues (Fig. 2F).
3.3 ISG20 expression was upregulated in GBM tissues and IDH1wt group, and its overexpression was unfavorable for the prognosis of GBM patients LGG, GBM, and normal brain tissues with ISG20 mRNA expression pro les and corresponding clinical information were obtained from the TCGA database. First, we found that ISG20 mRNA expression in GBM tissues was upregulated compared to that in non-tumor and LGG tissues (Fig. 3A), but there was no difference between LGG tissues and non-tumor tissues. Moreover, low ISG20 expression was more favorable for IDH1mutation (Fig. 3B). More importantly, our results revealed that high ISG20 mRNA expression has a shorter OS time in patients with GBM (Fig. 3C). All the above results were veri ed using the CGGA data ( Fig. 3D-F). Subsequently, we plotted the receiver operating characteristic curve (ROC) for ISG20 mRNA expression and IDH1wt subtype of GBM to recon rm the above nding. Interestingly, the areas under the curve (AUC) were 0.795 and 0.799 in the TCGA RNA-seq set and CGGA RNA-seq set, respectively (Fig. 3G, H). We also collected non-tumor human brain tissues and fresh glioma tissues to detect ISG20 mRNA and protein expression using qRT-PCR (Fig. 3I) and IHC (Fig. 3J) analyses. The results demonstrated that ISG20 mRNA and protein expression were both increased in GBM tissues compared to those in non-tumor human brain and LGG tissues. Moreover, IHC analysis showed that compared with the IDH1mut group, ISG20 protein expression was higher in the IDH1wt group (Fig. 3K). Collectively, our results demonstrate that ISG20 expression is upregulated in human IDH1wt GBM tissues, and high ISG20 expression predicts a poor prognosis in GBM patients.

Genetic alteration of ISG20 in GBM
Genetic and epigenetic changes play a crucial role in regulating cancer development and TIICs. First, we used cBioPortal to observe the genetic alteration status of ISG20 in GBM and found that ISG20 was altered in 14/155 (9%) GBM patient samples (Fig. 4A). The types of genetic alterations included mutation, ampli cation, and high and low mRNA levels. Simultaneously, we found that ISG20 had one missense mutation, R27C, in its RNase-T domain (Fig. 4B), and observed the R27C site in the 3D structure of ISG20 protein (Fig. 4C). Additionally, we found that compared with the unaltered group, the altered group of ISG20 had favorable OS, disease-free survival (DFS), and progression-free survival (PFS) in GBM patients (Fig. 4D, E, and F). Furthermore, we explored somatic mutations between cases with low and high ISG20 mRNA expressions and found that mutations in TP53, epidermal growth factor receptor (EGFR), and SPTA1 were signi cantly enriched in low ISG20 mRNA expression cases (Fig. 4G). However, gliomas with high ISG20 mRNA expression possessed higher frequency of PTEN and TTN mutations (Fig. 4G). Subsequently, we also explored CNAs in glioma according to ISG20 mRNA expression, and observed that chromosome (Chr) 7 was signi cantly ampli ed in gliomas, on the contrary, Chr 10 was deleted (Fig.   4H); simultaneously, 7p11.2 and 15q24.3 peaks were signi cantly ampli ed in gliomas with high ISG20 mRNA expression (Fig. 4I). Thus, the anomalous increase of ISG20 mRNA expression in GBM is likely caused by genetic alterations. Next, we demonstrated whether epigenetic mechanisms are involved in the regulation of ISG20 mRNA expression.

ISG20 DNA methylation and their prognostic values in GBM
DNA methylation is an epigenetic form of regulating gene expression without altering the sequence of the genome [20]. First, we performed the Pearson correlation analysis to investigate the correlation between ISG20 mRNA expression and its DNA methylation levels in GBM and found a signi cantly negative correlation between them (Fig. 5A). Subsequently, the β values of all eleven ISG20 CpG sites in GBM according to HM450K arrays were analyzed and displayed using MethSurv tool (Fig. 5B). Among them, the CpG sites cg19951424 (1stExon; 5'UTR), cg24181188 (1stExon; 5'UTR), and cg04515913 (1stExon; 5'UTR) showed low DNA methylation levels. In addition, clustering analysis in the form of heat maps showed that methylation sites cg1995142, cg24181188, and cg04515913 in the '1stExon; 5'UTR' sub-region had differential hypo-methylation levels in GBM patients (139 samples) (Fig. 5C), which was consistent with the β values presented in Fig. 5B. Furthermore, to evaluate the prognostic value of each CpG site, our results revealed that only the DNA methylation levels of cg24181188 were negatively correlated with OS in GBM patients, using the median of DNA methylation level (Fig. 5D, Fig. S1A, B). Finally, we observe that the methylation level of cg24181188 (R2 = 0.1337) was also decreased, accompanied by an increase in ISG20 expression (P < 0.05) (Fig. 5C). Thus, these ndings demonstrated that increased ISG20 mRNA expression in GBM samples could be associated with hypomethylated CpG sites.

Functional enrichment analysis of ISG20 in patients with GBM
To investigate the biological processes of ISG20 in GBM, Pearson correlation analysis between ISG20 and other genes was performed using the TCGA RNA-seq set. We then chose 301 genes that were most relevant to ISG20(|R| > 0.5, P < 0.05) to draw the heat maps (Fig. 6A) and performed enriched GO and KEGG analyses. GO results showed that the major biological processes included response to IFN-γ, neutrophil activation involved in immune response, neutrophil activation and degranulation, antigen processing and presentation, response to tumor necrosis factor, and response to virus (Fig. 6B). The major cellular components (CCs) were collagen-containing extracellular matrix, secretory granule lumen, vesicle lumen, focal adhesion, major histocompatibility complex protein, and lysosomes (Fig. 6C). The major molecular functions (MFs) were enzyme inhibitor activity, protease binding, peptide antigen binding, 1 phosphatidylinositol 3 kinase regulator activity, immunoglobulin binding, and cysteine-type endopeptidase activity involved in the apoptotic process (Fig. 6D). KEGG pathway analysis revealed that ISG20 is mainly involved in Salmonella infection, HIV 1 infection, phagosomes, antigen processing and presentation, natural killer cell-mediated cytotoxicity, and complement and coagulation cascades (Fig.  6E). We also performed gene set enrichment analysis (GSEA) between ISG20 expression and other genes in the TCGA database. The GSEA data showed that allograft rejection, in ammatory response, E2F targets, G2M checkpoint, mammalian Target of rapamycin complex 1 (mTORC1) signaling, hypoxia, mitotic spindle, p53, and KRAS signaling up, and early estrogen response were enriched in the high ISG20 mRNA expression group (Fig. 6F). Therefore, our results indicate that ISG20 is involved in immune regulation in glioma, and we next explored its relationship with TIICs.

ISG20 expression is correlated with immune in ltration levels and macrophages M2 polarization in glioma patients
TIICs, as crucial components of the TME, induced consequences for the initiation, progression, and outcome of cancer [21,22]. Therefore, we studied the potential immunological correlation between ISG20 and TIICs. First, we used the CIBERSORT algorithm to determine the levels of the 22 types of immune cells (Fig. 7A, B). Macrophages M2 and CD4 memory resting T cells accounted for a large proportion of glioma in ltrating immune cells, consistent with previous studies [17]. By comparing the high ISG20 and low ISG20 expression groups; LGG and GBM groups; and IDH1wt and IDH1mut groups, we found signi cant differences in the in ltrated levels of macrophage M2 between them but not in the immune cell abundance of CD4 memory resting T cells (Fig. 7C, Fig. S2A, B). Subsequently, we used TIMER2.0 database to explore the relationship of ISG20 mRNA expression with tumor purity and macrophage M2 in GBM, and found that ISG20 mRNA expression was negatively correlated with tumor purity (P < 0.05) ( Figure 7E), indirectly indicating that high ISG20 mRNA expression may predict an unfavorable prognosis, since previous studies showed that tumor purity re ects the characteristics of TME and low tumor purity is associated with poor outcome of glioma [23]. Simultaneously, our results further veri ed that ISG20 mRNA expression was positively correlated with M2 macrophage in ltration levels ( Fig. 7F-H). Finally, we analyzed the correlation between macrophage M2 markers and ISG20 expression, and the grade of glioma and IDH1 mutation status to explore the possible molecular mechanism of macrophage M2 polarization. Our results showed that the expression of CD163 and CD204 were all up-regulated in high ISG20 expression, GBM, and IDH1wt groups compared to low ISG20 expression, LGG, and IDH1mut groups, respectively (Fig. 7D, Fig. S2C, D). On the contrary, CD206, another marker of M2 macrophages, was decreased in high ISG20 expression and IDH1wt groups compared to low ISG20 expression and IDH1mut groups, and not different between LGG and GBM groups (Fig. 7D, Fig. S2C, D). These further demonstrate that macrophage M2 polarization is regulated by multiple molecular mechanisms. Moreover, IHC was performed; our results revealed that compared to the LGG and low ISG20 groups, the protein expression of CD163 and CD204 were higher in GBM and high ISG20 expression groups (Fig. 7I, J); however, CD163 and CD204 protein expression was lower in the IDH1mut group than in the IDH1wt group (Fig. 7K). Collectively, our data demonstrated that ISG20 high expression was correlated with macrophages M2 cells in ltration levels in GBM. The high expressions of CD163 and CD204 are likely the results of ISG20 high expression, and macrophages M2 polarization predicted malignant biological phenotypes in glioma.

Discussion
Cytokines are low molecular weight soluble proteins released by cancer cells or stromal cells and required for the formation of TME, because cytokines can regulate innate and adaptive immunity, hematopoiesis, cell growth, and repair of damaged tissues through endocrine, autocrine, and paracrine modes, and enable cells in TME to communicate [24,25]. IFNs are cytokines that modulate resistance to viral infections, inhibit cell proliferation, adjust immunity and anti-tumor effects, and induce the production of hundreds of ISGs [26,27]. ISG20, a human cDNA encoding promyelocytic leukemia nuclear bodies (PML-NBs)-associated protein [28], is an ISG that is directly induced by IFN [29]. It has been reported that promyelocytic leukemia (PML) is a tumor suppressor and essential for the proper assembly of PML-NB (which contains over 70 proteins) and its anti-tumor functions, according to its binding protein [30][31][32][33]. In addition, ISG20 expression has been shown to be up-regulated during infection and in several types of cancers, including glioma[16, [34][35][36][37][38]. In these cancers, ISG20 high expression indicates a poor OS. However, systematic and comprehensive analyses to further elucidate ISG20 underlying role in tumors are lacking. In the current study, our results revealed that ISG20 was highly expressed in most cancers using a pan-cancer analysis.
In addition, IDH1 is an NADP+-dependent enzyme located in the cytoplasm, and its mutant form can convert α-KG to the oncometabolite 2-HG [3]. The enrichment of 2-HG produced by IDH1mut at residue R132 may be conducive to cancer development by inhibiting histone demethylases and collagen maturation, and upregulating HIF1α, leading to increased histone methylation and altered gene expression [39,40]. To further explore the correlation between ISG20 expression and IDH1mut status in different tumors, we combined ISG20 with IDH1 mutation and found that there were signi cant differences only in GBM. In addition, our results showed that ISG20 high expression seems to be linked to malignant entities, such as GBM and IDH1wt status in gliomas. Simultaneously, the OS curve indicated that ISG20 high expression was detrimental to the survival of patients with GBM. To further explain this nding, ROC for ISG20 expression and IDH1wt subtype of GBM were generated, and the AUC were 0.795 and 0.799 in the TCGA RNA-seq set and CGGA RNA-seq set, respectively. More importantly, we also collected non-tumor brain tissues and fresh glioma tissues to verify ISG20 mRNA and protein expression and found that the expression of ISG20 was higher in GBM and in its IDH1wt status compared to that in non-tumor tissues, LGG tissues and IDH1mut status, respectively. Summarily, our results indicate that ISG20 expression is upregulated in human GBM tissues and its IDH1wt status, and high ISG20 expression predict a poor prognosis in GBM patients.
Subsequently, we analyzed the possible mechanisms of ISG20 high expression from genetic and epigenetic changes. First, we observed that ISG20 was altered in 14/155 (9%) GBM patient samples and found that its alterations were conducive to OS, DSS, and PFS; simultaneously, our results revealed that the frequency of PTEN and TTN mutations increased in gliomas with high ISG20 mRNA expression, and the 7p11.2 and 15q24.3 peaks were signi cantly ampli ed. Thus, the abnormal increase of ISG20 mRNA expression in GBM may be caused by genetic change and affect genomic alterations. Additionally, DNA hypomethylation is common in cancer and often leads to upregulation of oncogenes [41,42]. Comparing ISG20 mRNA expression with its DNA methylation level, we con rmed that ISG20 mRNA expression was signi cantly inversely correlation with its DNA methylation level, indicating that methylation alteration may be another mechanism for the increase of ISG20 mRNA expression. Because not all methylation sites are responsible for gene expression, the most signi cant methylation sites required further investigation. Eleven CpG sites were grouped into the ISG20 gene sub-regions. Among them, cg19951424, cg24181188, and cg04515913 showed a low DNA methylation level, and their prognostic value in CpG site was explored. The results indicated that only DNA methylation levels of cg24181188 were statistically correlated with the prognosis of GBM patients. Simultaneously, the methylation level of cg24181188 (R2 = 0.1337) also decreased, accompanied by an increase in the expression levels of ISG20 (P < 0.05), which indicated that the change of cg24181188 may be one of the crucial epigenetic mechanisms for the increase of ISG20 mRNA expression.
Regarding the biological functions of ISG20, we revealed that ISG20 played a crucial role in the in ammatory response and immune regulation in GBM. To further investigate the association of ISG20 expression with TIICs in glioma, we explored the abundance of TIICs, which are related to tumor progression, metastasis, or prognosis [21,22]. The results showed that macrophages M2 and CD4 memory resting T cells accounted for a large proportion of glioma in ltration immune cells, and ISG20 high expression contributed to macrophages M2 in ltration and was negatively correlated with GBM purity (percentage of malignant cells in a tumor tissue). However, ISG20 expression exhibited no statistical correlation with the abundance of CD4 memory resting T cells in gliomas. In the TME, tumorassociated macrophages (TAMs) are considered to be of the polarized M2 phenotype, which enhances tumor progression and indicates a poor prognosis [43]. Therefore, we focused on the correlation between ISG20 expression and markers of macrophages M2. These ndings revealed that the expression of CD163 and CD204 were all up-regulated in high ISG20 expression, GBM, and IDH1wt groups compared to low ISG20 expression, LGG, and IDH1mut groups, respectively. However, CD206 was not different between LGG and GBM, and was negatively correlated with the high ISG20 and IDH1mut groups, which can be explained by the fact that macrophage polarization is a complex process regulated by many factors. In the current study, we did not thoroughly explore the mechanism of CD206 abnormal expression, which will be veri ed in future studies. Therefore, our results showed that the polarization of macrophages M2 predicted malignant biological phenotypes in glioma and was likely the result of the abnormal high expression of ISG20.

Conclusion
In summary, our study is the rst to evaluate ISG20 combined with IDH1mut in tumors, by pan-cancer analysis. This study provides comprehensive evidence for ISG20 as a potential therapeutic target and prognosis and predictive biomarker for GBM. However, the exact mechanisms by which ISG20 regulates macrophage M2 polarization are largely unknown and need to be explored in future studies. LGG and PRAD in the TCGA project, the corresponding normal tissues of the GTEx database were included as controls. Box plots derived from gene expression data for GEPIA2 were supplied. P<0.05; * P<0.05; ** P<0.01; *** P<0.001.

Figure 2
Transcription levels of ISG20 in GBM Patients. (A) The heatmap represents data with statistically signi cant upregulation (red) or downregulation (blue) of ISG20 transcription levels in various human cancer tissues as compared with their normal control. The numbers in the heatmap indicate the published independent datasets of mRNA microarray experiments. (B-E) Box plots from gene expression data in TCGA database comparing the expression of ISG20 in normal and GBM tissue. (F) Meta-analysis on the mRNA expression levels of ISG20 in GBM tissues vs. non-cancerous brain tissues using the four ONCOMINE datasets. The colored squares represent the median rank of these genes (vs. normal tissue) across the four datasets. The signi cance level for the median rank analysis was set at P<0.05. ISG20 expression was upregulated in human GBM tissues and IDH1wt group, and its overexpression was unfavorable for the prognosis of GBM patients. (A, D) Differential expression of ISG20 mRNA in normal brain tissues, LGG tissues and GBM tissues in TCGA and CGGA, respectively. (B, E) The correlation between the ISG20 mRNA expression and IDH1 mutation status in GBM were analyzed by TCGA database and CGGA database, respectively. Kaplan-Meier survival analysis of ISG20 in GBM based on the TCGA cohort (C) and CGGA cohort (F). ROC curves of ISG20 expression to predict IDH1wt subtype in TCGA (G) and CGGA (H) datasets. (I) Differential expression of ISG20 mRNA in human normal brain tissues (Normal), human LGG tissues (LGG) and human GBM tissues (GBM). (J) Representative photographs of immunohistochemical staining of ISG20 in IDH1 mutation status of GBM. (K) Representative photographs of immunohistochemical staining of ISG20 in human normal brain tissues (Normal), human LGG tissues (LGG) and human GBM tissues (GBM). Positive cells are stained brown. The data are presented as the mean ± SD, P < 0.05. * P<0.05; ** P<0.01; *** P<0.001.

Figure 4
Genetic alteration of ISG20 in GBM. The alteration frequency with mutation type (A) and mutation site (B) are displayed. We display the mutation site (R27C) in the 3D structure of ISG20 (C). We also analyzed the potential correlation between genetic alterations and overall survival (D), disease-speci c survival (E) and progress-free survival(F) of GBM using the cBioPortal tool. Differential somatic mutations were detected in gliomas with low and high ISG20 expression (G). The overall CNAs pro le in order of increasing ISG20 expression (H). GISTIC 2.0 ampli cations and deletions in gliomas with low and high ISG20 expression (I). Chromosomal locations of peaks of signi cantly recurring focal ampli cation (red) and deletions (blue) were presented. P < 0.05. correlation scatterplot of ISG20 CpG sites. Y-axis represents mRNA expression fold change (Log10). Xaxis represents cg24181188 methylation beta value. P < 0.05.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.