Identification of prognostic biomarkers and correlations with immune infiltrates among cGAS-STING in hepatocellular carcinoma

Abstract The cyclic GMP-AMP synthase-stimulator of interferon genes (cGAS-STING) pathway induces innate immunity by activating the production of inflammatory cytokines and type I interferons. Recently, studies revealed that self-DNA from by-products of chromosome instability and tumors could activate the cGAS-STING pathway, and subsequently promote or inhibit tumor development. However, the prognostic value and correlations with immune infiltrates of the cGAS-STING pathway in hepatocellular carcinoma (HCC) have not been clarified. In the present study, we used the Molecular Signatures Database, Oncomine, UALCAN, Human Protein Atlas, Kaplan–Meier plotter, LinkedOmics, and Tumor Immune Estimation Resource databases. Overexpression of XRCC5, IRF3, TRIM21, STAT6, DDX41, TBK1, XRCC6, TREX1, PRKDC, and TMEM173 was markedly correlated with clinical stages and pathological grades in HCC. Moreover, higher mRNA expression of XRCC5, XRCC6, and PRKDC was significantly related with shorter overall survival. However, higher mRNA expression of IFI16, STAT6, NLRC3, and TMEM173 was associated with favorable overall survival. Our results suggested that the kinase targets of the cGAS-STING pathway included the SRC family of tyrosine kinases (LCK and LYN), phosphoinositide 3-kinase-related protein kinase (PIKK) family kinases (ATM and ATR), and mitogen-activated protein kinase 1 (MAPK1). Furthermore, we identified significant correlations among the expression of cGAS-STING pathway and infiltration of B cells, CD4+T cells, CD8+ T cells, macrophages, neutrophils, and dendritic cells in HCC. The expression of the cGAS-STING pathway also exhibited strong relationships with diverse immune marker sets in HCC. These findings suggest that cGAS-STING pathway members may be used as prognostic biomarkers and immunotherapeutic targets HCC patients.


Introduction
Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related mortality worldwide [1]. The 5-year survival rate of patients with advanced liver cancer is poor due to high tumor recurrence, metastasis, and the lack of early diagnostic biomarkers with high sensitivity and specificity [2]. Owing to poor liver function correlated with cirrhosis and extrahepatic metastasis, most patients with HCC are resistant to common cytotoxic therapies [3]. Although doxorubicin was initially viewed as a first-choice drug for advanced HCC, a controlled trial showed that it was related with a low survival rate [4]. The overall life expectancy of patients with advanced HCC does not exceed 1 year even under treatment with sorafenib or regorafenib [5].Therefore, there is an urgent need to identify novel potential prognostic and therapeutic targets that are related with tumor formation and progression in patients with HCC. between 20 different cancer samples and their normal adjacent tissues from the Oncomine database. The threshold was determined based on the following values: P=0.01, fold change = 1.5, gene rank = 10%, and data type for mRNA.

UALCAN database analysis
UALCAN (http://ualcan.path.uab.edu) is an interactive web server for the analysis of 31 types of cancer with transcriptome data from The Cancer Genome Atlas (TCGA). This tool is based on clinical data and level 3 RNA-sequence. UALCAN can be used to assess the association of transcriptional expression with relative clinicopathologic features [26]. In our study, UALCAN was utilized to analyze the mRNA expression of cGAS-STING pathway genes in HCC tissues, and estimate their association with clinicopathologic features. A P<0.05 denoted statistically significant difference.

Human protein atlas
The Human Protein Atlas (https://www.proteinatlas.org) is an online site containing immunohistochemistry-based expression data for approximately 20 common types of tumors. It is used to map the human proteins in organs, tissues, and cells through integration of different omics technologies, such as mass spectrometry-based proteomics and transcriptomics [27]. In the present study, the protein expression of different cGAS-STING pathway genes between human normal and HCC tissues was compared using immunohistochemistry images.

K-M plotter database analysis
The K-M plotter (www.kmplot.com) is used to estimate the effect of 54,000 genes for prognostic analysis in 21 common types of cancer. The database contains gene chip and RNA-sequence data-sources obtained from databases, such as the Gene Expression Omnibus [28,29]. We evaluated the prognostic value of the mRNA expression of cGAS-STING pathway genes in liver cancer via the K-M plotter database. We divided patients according to the median and accepted other default settings prior to the analysis. The hazard ratio (HR) with 95% confidence intervals and log-rank P-values was computed. A P<0.05 denoted statistically significant differences.

LinkedOmics
LinkedOmics (http://www.linkedomics.org/) is a publicly available portal containing all 32 types of cancer included in TCGA. It serves as a unique tool for biologists and clinicians to access, compare, and analyze cancer multi-omics data [30]. In this study, we determined the kinase target enrichment of cGAS-STING pathway genes using the "Link-Interpreter" module. We conducted analyses with a number of size of 3 and simulations of 500 in LIHC dataset. We considered 0.05 as the P-value cutoff and used the Spearman correlation test to analyze the results statistically.

TIMER database analysis
TIMER (https://cistrome.shinyapps.io/timer/) is a reliable database for estimating immune cell infiltration using data from TCGA, including 10,897 samples from 32 types of cancer [31]. In the present study, we initially analyzed the expression of cGAS-STING pathway genes in HCC. Subsequently, we examined the correlation between the expression of cGAS-STING pathway members and immune cell infiltrates, including B cells, CD8+ T cells, CD4+ T cells, macrophages, DCs, and neutrophils with gene modules. Highly expressed genes in the tumor microenvironment tend to have negative correlations with tumor purity [32]. The generated scatterplots suggest statistical significance and provide the purity-corrected partial Spearman's rho value. Further, correlations between the expression of cGAS-STING pathway members and gene markers of immune cells were further evaluated with correlation modules. The cGAS-STING pathway genes were represented on the X-axis, and related marker genes were used for the y-axis. Options for partial correlation under the condition of tumor purity were conducted.

Statistical analysis
The results generated from Oncomine were displayed with fold changes, P-values, and t-test results. Survival curves were generated via K-M plots. The results of KM plots were presented by P-values and HR obtained from a log-rank test. A P<0.05 denoted statistically significant differences.

mRNA expression levels of cGAS-STING pathway members in HCC
We analyzed the differences in the transcriptional levels of cGAS-STING pathway members in patients with HCC using the Oncomine database. The results revealed that the mRNA expression of TRIM21, IFI16, NLRC3, DDX41,  XRCC6, TREX1, PRKDC, and TMEM173 was significantly higher in HCC tissues in multiple datasets ( Figure 1 and Table 1). In the Wurmbach Liver dataset [33], higher expression of TRIM21 was revealed in HCC tissues compared with normal adjacent tissues (fold change = 1.607 and P = 4.78E-04). Up-regulation of IFI16 was also found in HCC tissues compared with normal adjacent tissues. The data from the Mas Liver dataset [34] showed increase of 3.474-fold (P = 4.23E-14) and 4.715-fold (P = 1.91E-16), respectively. Wurmbach [33] observed a 3.649-fold (P = 1.02E-06) and 2.250-fold (P = 4.06E-04) increase, respectively, in IFI16 mRNA expression in HCC samples. Wurmbach [33] also found a 2.197-fold (P = 1.35E-05) increase in NLRC3 mRNA expression in HCC tissues. Similarly, in the Roessler Liver dataset [35], 2.012-fold (P = 7.60E-8) and 1.701-fold (P = 2.95E-41) increases in DDX41 mRNA expression were found in HCC samples compared with normal tissues. Wurmbach [33] observed a 1.719-fold (P = 5.25E-7) increase in DDX41 mRNA expression in HCC samples. In the Roessler Liver 2 dataset [35], high expression of XRCC6 was observed in liver cancer tissues compared with normal adjacent tissues (fold change = 1.828 and P = 1.47E-70, respectively). In addition, TREX1 showed higher expression levels in HCC tissues versus normal tissues in Wurmbach Liver (fold change = 1.552 and P=0.002) [33]. Similarly, results from three datasets suggested that PRKDC expression was higher in HCC tissue versus normal adjacent tissue [33,35,36]. The mRNA expression of cGAS was not found on Oncomine.
After confirming the up-regulation of mRNA expression levels of cGAS-STING pathway members in HCC tissues, we subsequently assessed their protein expression levels based on the Human Protein Atlas website. As shown in Figure 3, STAT6, TBK1, and TREX1 proteins were not expressed in normal liver tissues, whereas they exhibited medium or high expressions in HCC samples ( Figure 3D,F,H). In addition, low or medium protein expression of XRCC5, IRF3, TRIM21, DDX41, XRCC6, and cGAS was observed in normal liver tissues, while high protein expression was detected in HCC tissues ( Figure 3A-C,E,G,I). However, the immunohistochemical images of IFI16, NLRC3, PRKDC, and TMEM173 expression were not found in this website. In general, these results revealed that the transcription and protein levels of most cGAS-STING pathway members were upregulated in patients with HCC.
We next analyzed the correlation between the mRNA expression of cGAS-STING pathway members and the pathological stage of patients with liver cancer according to UALCAN. The results are presented in Figures 4 and 5. The mRNA expression of cGAS-STING pathway members, except for IFI16 and NLRC3, was significantly related with individual cancer stages and tumor grades in HCC. Furthermore, patients in advanced stages and tumor grades of HCC had higher mRNA expression levels of cGAS-STING pathway members compared with those in early stages of the disease. In Figure 4, the highest mRNA expression of DDX41 was found in stage 4 ( Figure 4G), while the highest mRNA expression of XRCC5, IRF3, TRIM21, STAT6, TBK1, XRCC6, TREX1, and PRKDC was found in stage 3 ( Figure  4A-C,E,H-K), which may be attributed to the small sample numbers (only six patients with stage 4 HCC). Figure 4L shows that the highest mRNA expression of TMEM173 was found in stage 1. Similarly, the highest mRNA expression of XRCC5, IRF3, DDX41, TBK1, XRCC6, and PRKDC was found in grade 4 tumors (Figure 5A,B,G-I,K), while the highest mRNA expression of TRIM21, STAT6, and TMEM173 was found in grade 3 ( Figure 5C,E,L). However, the highest mRNA expression of IFI16 and TREX1 was found in grade 1 or 2; as the tumor grade increased, their mRNA expression tended to be lower ( Figure 5D,J). These results suggested that XRCC5, IRF3, TRIM21, STAT6, DDX41, TBK1, XRCC6, TREX1, PRKDC, and TMEM173 were strongly correlated with tumorigenesis and tumor progression in patients with HCC.

Prognostic value of mRNA expression of cGAS-STING pathway members among patients with HCC
We further analyzed the prognostic potential of mRNA expression of cGAS-STING pathway members in patients with HCC using the K-M plotter. Overall survival (OS) curves are illustrated in  The mRNA expression of XRCC5, IRF3, TRIM21, STAT6, DDX41, TBK1, XRCC6, TREX1, PRKDC, and TMEM173 was found to be up-regulated in HCC tissues versus normal tissues (A-C,E,G-L). The transcriptional levels of IFI16 and NLRC3 did not show significant differences (D and F); ***P<0.001.

Kinase targets of cGAS-STING pathway members in HCC
Based on the significantly different expression of cGAS-STING pathway members in HCC tissues versus normal tissues, we estimated possible kinase targets of the differentially expressed cGAS-STING pathway members. In the present study, we investigated the top two kinase targets of cGAS-STING pathway members via the LinkedOmics database. As shown in Table 2, PLK1 and ATR were the top two kinase targets of XRCC5. DYRK1A and EGFR were noted as the targets for the IRF3 kinase-target network. LYN and LCK were mainly correlated with TRIM21. Constituents of the IFI16 kinase-target network were primarily correlated with SYK and LCK. ATR and PLK1, and LCK and SYK were the top two kinase targets of STAT6 and NLRC3, respectively. Mitogen-activated protein kinase 1 (MAPK1) and DYRK1A were regarded as the kinase targets of DDX41. ATM and DAPK1 were mainly related with TBK1. AURKB and PLK1 were shown as kinase targets of XRCC6. ATM and CDK2, as well as ATM and CDK1,  were the top two targets of TREX1 and PRKDC, respectively. BUB1 and HCK were regarded as the kinase targets of TMEM173.The kinase targets of cGAS were not found in the LinkedOmics database.

Immune cell infiltration of cGAS-STING pathway members in HCC
Tumor-infiltrating lymphocytes can be used as independent predictors for sentinel lymph node status and survival in cancer [37]. Therefore, we investigated the correlations between differentially expressed cGAS-STING pathway

Correlation analysis between the expression of the cGAS-STING pathway and immune marker sets
We investigated the relationships between cGAS-STING pathway members and different marker genes of immune cells of HCC using the TIMER database. The immune marker sets of various immune cells, such as monocytes, M1 and M2 macrophages, tumor-associated macrophages (TAMs), and DCs were analyzed. In addition, we investigated various functional T cells, such as T helper 1 (Th1), Th2, and regulatory T (Treg) cells. The results are presented in Supplementary Tables S1-3. After correlation adjustment by purity, we found that the expression levels of cGAS-STING pathway members were significantly related with most immune marker sets of different immune cells and various functional T cells in liver cancer. Importantly, the expression of immune marker sets of TAMs, monocytes, M1 and M2 macrophages was associated with the expression levels of most cGAS-STING pathway members, including XRCC5, TRIM21, IFI16, STAT6, NLRC3, TBK1, XRCC6, and PRKDC (Supplementary Tables S1-3).
Furthermore, the results also suggested that the expression of most marker sets of DCs, such as HLA-DPB1, CD1C, NRP1, and ITGAX had strong correlations with XRCC5, TRIM21, IFI16, STAT6, NLRC3, TBK1, XRCC6, and PRKDC expression in HCC (Supplementary Tables S1-3). These results suggested that there was a strong correlation between cGAS-STING pathway members and DC infiltration. In addition, there was a significant connection between FOXP3 and TGFβ1 for Treg cells and cGAS-STING pathway members in HCC.

Discussion
The cGAS-STING pathway has emerged as a potential mechanism to induce inflammation-mediated tumorigenesis. Actually, persistent activation of this pathway and its downstream effectors, such as TBK1, has been connected with chronic inflammation and cancer progression [38,39]. The development of HCC is a multistep process that involves continuous inflammatory damage, such as hepatocyte necrosis [40]. Some studies have observed correlations between cGAS-STING pathway members, the tumor microenvironment, and cancer immunotherapy. However, the prognostic value and potential therapeutic targets of the cGAS-STING pathway in HCC are poorly characterized. In the present study, we investigated the expression, prognostic values, and correlations with immune infiltrates of different cGAS-STING pathway members in HCC.
We first investigated the expression of cGAS-STING pathway members and its correlations with the pathological stage in HCC. Based on the Oncomine database, the mRNA expression of TRIM21, IFI16, NLRC3, DDX41, XRCC6, TREX1, PRKDC, and TMEM173 was significantly higher in HCC tissues in multiple datasets. Moreover, in patients with HCC, high expression of mRNA and protein was found in cGAS-STING pathway member genes, including XRCC5, IRF3, TRIM21, STAT6, DDX41, TBK1, XRCC6, TREX1, PRKDC, and TMEM173. Furthermore, the mRNA expression of these 10 genes was strongly associated with cancer stages and tumor grades in patients with HCC. These data demonstrate that differentially expressed cGAS-STING pathway members may play a significant role in HCC. Recently, high expression of XRCC5 has been found in breast and gastric cancer [17,41]. A strong correlation was found between the variable number tandem repeat polymorphism in the XRCC5 gene and the risk of breast cancer [17]. Overexpression of XRCC5 was also detected in gastric cancer, in which XRCC5 regulated the overexpression of chloride channel 3 (CLC3) [41]. Additionally, Liu [42] found that higher expression of XRCC5 was related with metastasis through the Wnt/β-catenin signaling pathway in patients with HCC, which was consistent with our results. In conclusion, XRCC5 may participate in the tumorigenesis of HCC.
Recently, Shi found that pharmacological targeting or knockdown of IRF3 using amlexanox. This drug is used for anti-inflammatory treatment, and can inhibit gastric tumor growth in a Yes-associated protein-dependent manner. The expression of IRF3 is up-regulated and prognosticates patient survival in gastric cancer [43]. A study showed that the expression of TBK1 was increased in mesenchymal small cell lung cancer cell lines [44]. Research found that TRIM21 modulated epithelial-mesenchymal transition (EMT) by regulating the stability of Snail in breast cancer [45]. To our knowledge, this was the first study to investigate the correlation among IRF3, TBK1, and TRIM21 with HCC. Our results suggested that all three may participate in the tumorigenesis of HCC.
The potential prognostic value of the mRNA expression levels of cGAS-STING pathway members in patients with HCC was subsequently investigated. The findings revealed that higher mRNA expression of XRCC5, XRCC6, and PRKDC was significantly associated with shorter OS. In contrast, higher mRNA expression of IFI16, STAT6, and NLRC3 was significantly related with better OS. Liu [42] observed that the high mRNA expression of XRCC5 predicted poor prognosis in patients with HCC, which was in accordance with our results. Studies have shown that NLRC3 was a negative regulator in innate immune signaling activated by STING. NLRC3 could inhibit STING-TBK1 interaction and the production of downstream type I IFN [46]. IFI16 enhanced STING activation by interacting with STING via the PYRIN domain. Moreover, the IFI16-induced inflammasome was found to inhibit HCC growth and metastasis, and was a tumor suppressor during the development [23]. In our study, we found that NLRC3 and IFI16 were independent prognostic factors for favorable OS in patients with liver cancer.
We also investigated the kinase targets of the cGAS-STING pathway members. We discovered that the SRC family of tyrosine kinases (LCK and LYN), phosphoinositide 3-kinase-related protein kinase (PIKK) family kinases (ATM and ATR), and MAPK1 were potential kinase targets of the cGAS-STING pathway members. These kinase targets affect the progression of cell cycle, DNA damage, and EMT [47][48][49][50][51]. Furthermore, they are involved in tumor progression by mediating tumor cell invasion, apoptosis, and migration [51][52][53]. As a result, differentially expressed cGAS-STING pathway members may modulate DNA repair, the EMT, and cell cycle progression by regulating these kinases in patients with HCC.
Another important finding in the present study is that the expression of cGAS-STING pathway members is associated with various immune cell infiltration levels in liver cancer. The cGAS-STING pathway can mediate protective immune defense toward infection with a great number of DNA-containing pathogens, as well as generate intrinsic antitumor immunity [54]. The accumulation of tumor DNA could activate STING-IRF3-induced IFN signaling to enforce tumor-antigen presentation on DCs and cross-prime CD8 + T cells for antitumor immunity [55]. There is an increasing body of evidence supporting that immune cell infiltration could affect cancer recurrence and progression, and also play an important role in clinical outcome and response to immunotherapy [56,57]. CD4 + T cells may be involved in the recognition of tumor antigens, and the activation of M1 macrophages may mediate the inhibition of tumor growth [58]. Our results demonstrate that there is a significant correlation between the expression of cGAS-STING pathway members and the infiltration of immune cells, such as T cells, macrophages, and DCs. This implies that cGAS-STING pathway members may act as potential prognostic indicators, as well as reflect the immune status. STING activation in hepatic macrophages could induce the production of proinflammatory cytokines, leading to nonalcoholic steatohepatitis that is characterized by hepatic steatosis [59].
Moreover, the correlations between the expression of cGAS-STING pathway members and the marker sets of immune cells suggest the role of cGAS-STING pathway members in regulating tumor immunology in HCC. Firstly, the M1 macrophage markers (e.g., IRF5 and PTGS2) and the marker genes of M2 macrophages (e.g., VSIG4, and MS4A4A) were significantly associated with the expression of XRCC5, TRIM21, IFI16, STAT6, NLRC3, TBK1, XRCC6, and PRKDC. Collectively, the results suggest the regulatory role of these cGAS-STING pathway genes in the polarization of TAMs. Moreover, our results reveal a strong relationship between the expression of XRCC5, TRIM21, IFI16, STAT6, NLRC3, TBK1, XRCC6, and PRKDC and DC infiltration. These genes also play a potential role in activating Treg cells and inducing T-cell exhaustion. DCs promote tumor metastasis by reducing the cytotoxicity of CD8 + T cells and increasing Treg cells [60]. Further studies are warranted to investigate whether these genes are important factors in inducing the DCs and tumor metastasis.
In addition, the increase in the expression of XRCC5, TRIM21, IFI16, STAT6, NLRC3, TBK1, XRCC6, and PRKDC was significantly related with the expression of Treg and T-cell exhaustion gene markers, such as FOXP3, STAT5B, TGFβ1, PDCD1, CTLA4, HAVCR2, and LAG3 in patients with HCC. HAVCR2 is an important surface protein from exhausted T cells [61]. In human squamous cell carcinomas, STING signaling abrogated tumor immunogenicity by recruiting Treg cells [62]. This is highly correlated with the expression of XRCC5, TRIM21, IFI16, STAT6, NLRC3, TBK1, XRCC6, and PRKDC in HCC. Furthermore, significant associations were also found between the expression of XRCC5, TRIM21, IFI16, STAT6, NLRC3, TBK1, XRCC6, and PRKDC and the regulation of some marker genes of T helper cells (e.g., Th1, Th2, and Th17). These associations suggest the potential mechanism through which these cGAS-STING pathway members may regulate T-cell functions in HCC. Collectively, the results suggest that cGAS-STING pathway members (e.g., XRCC5, TRIM21, IFI16, STAT6, NLRC3, TBK1, XRCC6, and PRKDC) may play a crucial role in the regulation and recruitment of immune infiltrating cells in HCC.
The present study had some limitations. First, all the data in the present study were obtained from online databases; hence, further studies involving larger sample sizes, and in vitro and in vivo experiments are warranted to confirm our results. Second, analysis based on the transcriptional levels may indicate a few aspects of the immune status; however, its ability to detect global changes is limited. Finally, we did not analyze the potential mechanisms of cGAS-STING pathway members in HCC. Future studies should be performed to investigate the detailed mechanism among cGAS-STING pathway members and HCC.

Conclusions
XRCC5, IRF3, TRIM21, STAT6, DDX41, TBK1, XRCC6, TREX1, PRKDC, and TMEM173 were found to be significantly positive correlated with clinical cancer stages and tumor grades in patients with liver cancer. In addition, high mRNA expression of XRCC5, XRCC6, and PRKDC was significantly related with poor OS. In contrast, high mRNA expression of IFI16, STAT6, NLRC3, and TMEM173 was strongly correlated with favorable OS in HCC. Increased expression levels of cGAS-STING pathway members are correlated with increased infiltration levels of immune cells, including B cells, CD8 + T cells, CD4 + T cells, neutrophils, macrophages, and DCs in HCC. Therefore, cGAS-STING pathway members, especially XRCC5, IFI16, STAT6, NLRC3, XRCC6, and PRKDC, are likely involved in immune infiltration and can be used as potential prognostic biomarkers for patients with HCC. We hope our results provide a new perspective for the design of new immunotherapeutic drugs against HCC.