三結構域蛋白家族 5（TRIM5）在自噬中起重要作用，并參與免疫和腫瘤進程，然而 TRIM5 在神經膠質瘤中的功能尚不清楚。本研究旨在通過生物信息學分析來評估 TRIM5 在膠質瘤中的作用。本研究神經膠質瘤數據庫臨床樣本包括低級別神經膠質瘤（LGG）與多形性成膠質細胞瘤（GBM）。通過 Oncomine、基因表達譜交互分析（GEPIA）和癌癥基因組圖譜（TCGA）數據庫探尋了 TRIM5 在膠質瘤組織中的表達。基于 TCGA 數據庫，我們利用生存分析和多因素 Cox 回歸分析評價 TRIM5 的預后作用。利用 STRING 數據庫預測 TRIM5 相關蛋白網絡，并通過 KEGG 富集分析預測 TRIM5 在膠質瘤中的潛在分子通路。此外，采用 CIBERSORT 和 TIMER 數據庫進行免疫浸潤分析。結果表明，與 Oncomine、GEPIA 和 TCGA 數據庫中的正常樣本相比，神經膠質瘤樣本中的 TRIM5 表達明顯上調。生存分析結果顯示，較高的 TRIM5 表達與 LGG+GBM 患者以及 LGG 患者較差的總體生存（OS）有關，但與 GBM 患者 OS 無關。臨床相關性分析結果顯示，TRIM5 表達與年齡（χ2=44.31，P<0.001）、病理學分級（χ2=130.10，P<0.001）以及組織學類型（χ2=125.50，P<0.001）具有相關性。多因素 Cox 風險分析結果顯示 TRIM5 表達（HR=1.48，95% CI=1.20～1.80，P<0.001）、年齡（HR=1.05，95% CI=1.03～1.10，P<0.001）以及病理學分級（HR=3.11，95% CI=2.30～4.20，P<0.001）是膠質瘤患者（LGG+GBM）預后的獨立危險因素；TRIM5 表達（HR=1.82，95% CI=1.42～2.32，P<0.001）、年齡（HR=1.06，95% CI=1.05～1.08，P<0.001）、病理學分級（HR=1.92，95% CI=1.22～3.01，P=0.005）以及組織學類型（HR=0.71，95% CI=0.57～0.89，P=0.003）是 LGG 患者的獨立預后因素。相互作用網絡分析發現，IRF3、IRF7、OAS1、OAS2、OAS3、OASL、GBP1、PML、BTBD1 以及 BTBD2 蛋白與 TRIM5 具有相互作用。此外，KEGG 分析還發現細胞凋亡、腫瘤以及免疫相關通路在 TRIM5 升高時顯著富集。免疫浸潤分析顯示，TRIM5 表達可以影響膠質瘤中活化 NK 細胞、單核細胞、活化肥大細胞、巨噬細胞等免疫細胞浸潤水平。以上結果提示，TRIM5 在膠質瘤組織中顯著上調，并與預后不良和免疫浸潤相關。TRIM5 可能作為神經膠質瘤預后與指導免疫治療的生物標志物。
引用本文： 陳悅, 李琴, 張潔, 顧睿, 李凱, 趙崗, 袁航, 豐天宇, 歐德瓊, 林蘋. TRIM5 高表達與膠質瘤患者不良預后和免疫浸潤的相關性研究. 生物醫學工程學雜志, 2020, 37(3): 469-479. doi: 10.7507/1001-5515.202004064 復制
Introduction
Glioma is the most common brain tumor with high morbidity of 6 per 100 000 people each year[1-2]. Based on histopathology, gliomas can be classified into gradesⅠ-Ⅳ. Of which, grade IV is regarded as glioblastoma multiforme (GBM), with shorter overall survival (OS) even with routine treatment[3]. Currently, postoperative radiotherapy and chemotherapy are the main treatment for glioma, but the prediction of clinical outcome is still inaccurate due to the histological grading variability, especially for low grade glioma (LGG) patients (gradesⅠ-Ⅲ, including pilocytic astrocytoma, anaplastic astrocytoma, oligodendroglioma, oligodendrocytoma, anaplastic oligodendrocytoma and anaplastic oligodendroglioma and etc.)[4-5]. Therefore, efforts are needed to identify more biomarkers to evaluate the prognosis for glioma.
The tripartite motif (TRIM) family, constituted as immune-regulated proteins, played a general role in autophagy and involved in immune and tumor processes[6]. Tripartite motif 5 (TRIM5), well-known as a retrovirus limiting factor, has the function of autophagic degradation, protecting immune cells from HIV-1 infection[7]. Bioinformatics analysis by Srihari et al. found that TRIM5 was associated with BRCA1 and BRCA2 breast cancer[8]. A paper from Leal et al., revealed that the restriction factor of TRIM5 may act as a significant role in defense against neuroinflammation[9]. Unfortunately, the function of TRIM5 remains much less explored in glioma.
In current study, we aimed to explore the prognostic role and potential mechanisms of TRIM5 in glioma. Firstly, we obtained the TRIM5 expression based on Oncomine, Gene Expression Profiling Interactive Analysis (GEPIA) and The Cancer Genome Atlas (TCGA) databases and analyzed the prognostics value of TRIM5 by TCGA database. Then, Oncomine, STRING databases and Kyoto Encyclopedia Genes and Genomes (KEGG) analysis were used to predict the possible pathways of TRIM5. Finally, immune infiltration analysis was performed between TRIM5 and glioma by using CIBERSORT and TIMER analyses.
1 Materials and Methods
1.1 Database
The TRIM5 mRNA expression in glioma was compared by the Oncomine (https://www.oncomine.org/) database[10] including Murat Brain and Sun Brain glioma studies[11-12]. TCGA (https://cancergenome.nih.gov/) database[13] was also searched to obtain TRIM5 expression and clinical data of glioma including 5 normal brain tissues, 529 LGG tissues and 169 GBM tissues. GEPIA (http://gepia.cancer-pku.cn/detail.php) database[14], an online webserver including 8587 normal and 9736 tumor samples, was utilized to analyze the TRIM5 expression (Tumor or Normal) from TCGA. Besides, data of 4 clinicopathological features including age, gender, grade and histological type were also extracted from TCGA.
1.2 Survival analysis
To explore associations between TRIM5 expression and glioma patient OS, TRIM5 expression level and clinical data of glioma were downloaded from TCGA. Then, package edgeR and R Limma package were used to transform the information. After excluding patients with incomplete information, we performed OS analysis by using R package.
1.3 Clinical correlation and Cox risk analyses
The relevant clinical and survival information of LGG and GBM patients was obtained from TCGA database. We eliminated patients without complete clinical data and acquired 668 patients (508 LGG and 160 GBM) for analyses. As for clinical correlation analysis, LGG and GBM patients were divided into high and low TRIM5 groups (grouped according to the median value), with 334 cases in each group.
1.4 Network analysis
To identify the possible interaction networks of TRIM5, we conducted co-expression analysis via Oncomine database with tumor type limited to “brain”. Protein-protein interaction (PPI) analysis of TRIM5 was performed using STRING (https://string-db.org/) database[15]. In addition, KEGG pathway enrichment analysis was conducted by GSEA 4.0.3 software with 1000 permutations[16]. Normal P value < 0.05 and false discovery rate (FDR) < 0.05 was the filter.
1.5 Immune infiltration analysis
CIBERSORT (https://cibersort.stanford.edu/) online tool is a deconvolution algorithm utilized to explore the correlation between gene expression and tumor-infiltrating immune cells (TIICs)[17]. We uploaded the standard TRIM5 expression data of 703 samples from TCGA to CIBERSORT algorithm running, and then selected samples with P < 0.05 to the final study cohort. Subsequently, the CIBERSORT obtained the proportion of TIICs in the high- and low-TRIM5 groups. In addition, TIMER (http://timer.cistrome.org/)[18] was utilized to report the correlation between immune infiltration and TRIM5 level in glioma. The analyzed immune cells include B cells, CD4+, Dendritic cells, CD8+ T cells, Macrophages and Neutrophils. Kaplan-Meier curves were downloaded from TIMER to analyze the relationship between immune infiltration or TRIM5 level and OS of LGG and GBM patients respectively.
1.6 Statistical analysis
The statistical analyses were conducted using R software 3.6.3 and GraphPad Prism 8.0.1. Student’s t-test or ANOVA were utilized to assess the TRIM5 difference between glioma tissues and normal tissues. The relationship between TRIM5 level and clinicopathological features was compared by chi-square tests. Survive analyses were evaluated by the Kaplan-Meier with the log-rank test and Cox regression analysis. P < 0.05 was considered significant.
2 Results
2.1 High expression of TRIM5 in glioma
We found TRIM5 was elevated in central nervous system (CNS) and brain tumors compared with normal tissues via Oncomine (Fig. 1a). For verification, we conducted meta-analysis of TRIM5 expression in 4 analyses in Oncomine database (Fig. 1b). Compared with normal brain tissues, the TRIM5 level was statistically increased in GBM or Astrocytoma tissues (P < 0.001, Fig. 1c and 1d). Compared with LGG tissues, the TRIM5 expression was significantly higher in GBM tissues based on Sun Brain and TCGA Brain (P < 0.001, Fig. 1e and 1f). In order to further verify the above results, we observed that TRIM5 was increased in numerous cancers including LGG and GBM by using GEPIA tool (Fig. 2a). As expected, TRIM5 expression in LGG and GBM was notably higher than that in normal control (Fig. 2b). These results demonstrated that TRIM5 was upregulated in glioma and might be an indicator to predict the malignancy of glioma.
圖1
Elevated TRIM5 expression in glioma
a. TRIM5 expression between cancer and normal tissues of Brain and CNS cancer in Oncomine; b. meta-analysis of TRIM5 expression in 4 analyses; c-f. TRIM5 expression in Murat Brain, Sun Brain and TCGA Brain
a. TRIM5 expression between cancer and normal tissues of Brain and CNS cancer in Oncomine; b. meta-analysis of TRIM5 expression in 4 analyses; c-f. TRIM5 expression in Murat Brain, Sun Brain and TCGA Brain
圖2
TRIM5 expression based on GEPIA database
a. TRIM5 was notably increased in variety tumors; b. TRIM5 expression in LGG and GBM compared with normal control. *,
a. TRIM5 was notably increased in variety tumors; b. TRIM5 expression in LGG and GBM compared with normal control. *,
2.2 Relationship between TRIM5 level and prognosis in glioma
We performed Kaplan-Meier to analyze the influence of TRIM5 in OS of glioma patients (LGG+GBM), which showed that high TRIM5 expression would lead to a worse outcome (P < 0.001, Fig. 3a). Besides, subgroup analysis indicated that higher TIRM5 was risk factor for 1-year, 3-year and 5-year OS in glioma patients (all P < 0.001, Fig. 3b–3d). Subsequently, we performed prognostic analysis in LGG and GBM patients respectively, the results revealed that increased TRIM5 predicted poor OS in patients with LGG (all P < 0.001, Fig. 3e–3h), while was not correlated to prognosis of GBM patients (all P > 0.05, Fig. 3i–3l).
圖3
TRIM5 upregulation was associated with poor prognosis in glioma on TCGA database
a–d. OS of LGG+GBM patients; e–h. OS of LGG patients; i–l. OS of GBM patients
a–d. OS of LGG+GBM patients; e–h. OS of LGG patients; i–l. OS of GBM patients
2.3 Clinical correlation and Cox analyses based on TCGA
We performed the correlation analyses between TRIM5 level and clinicopathological characteristics in glioma patients by TCGA (LGG+GBM). As shown in Tab. 1, high TRIM5 expression was significantly correlated to older age (χ2 = 44.310, P < 0.001), higher grade (χ2 = 130.100, P < 0.001) and more malignant histological type (χ2 = 125.500, P < 0.001). Above results illustrated that TRIM5 contributed to predict glioma progression.
表1
Correlations between TRIM5 expression and clinicopathological characteristics in glioma patients (LGG+GBM)
To further confirm the prognostic role of TRIM5, univariate and multivariate Cox analyses were employed (LGG and GBM patients). As shown in Tab. 2, TRIM5 expression was a risk factor (HR = 1.481, 95% CI = 1.201–1.824, P = 0.000) for glioma OS. Besides, age (HR = 1.047, 95% CI = 1.035–1.060, P < 0.000) and grade (HR = 3.110, 95% CI = 2.300–4.206, P < 0.000) were both risk factors. Next, we continued to perform analyses in LGG patients and the results showed that TRIM5 expression (HR = 1.816, 95% CI = 1.419–2.322, P < 0.000) could be acted as an independent prognostic factor. In addition, age (HR = 1.064, 95% CI = 1.047–1.082, P < 0.000), grade (HR = 1.915, 95% CI = 1.219–3.009, P =0.005) and histological type (HR = 0.709, 95% CI = 0.566–0.890, P = 0.003) were all risk factors for LGG patients OS (Tab. 3).
表2
Univariate and multivariate Cox analyses of OS in LGG+GBM patients
表3
Univariate and multivariate Cox analyses of OS in LGG patients
2.4 The network interactions with TRIM5
To explore the possible molecular mechanisms of TRIM5, we predicted some co-expressed genes with TRIM5 in glioma by Oncomine database. We found strong correlation between NMI, TRIM22 and DDX60 etc. with TRIM5 with the correlation score > 0.7 (Fig. 4a). STRING indicated that IRF3, IRF7, OAS1, OAS2, OAS3, OASL, GBP1, PML, BTBD1 and BTBD2 were contacted with TRIM5 (Fig. 4b). Lastly, KEGG pathway enrichment of TRIM5 showed that highly expressed TRIM5 was mainly enriched in “pathways in cancer”, “apoptosis”, and some immune-related pathways (Fig. 5). The specific information of the enrichment pathways is shown in Tab. 4.
圖4
The potential interaction networks prediction of TRIM5
a. transcriptome prediction of TRIM5 in glioma in Oncomine database; b. TRIM5 protein interaction networks in STRING
a. transcriptome prediction of TRIM5 in glioma in Oncomine database; b. TRIM5 protein interaction networks in STRING
表4
Results of part KEGG enrichment analyses
圖5
Enrichment plots from GSEA analysis
2.5 Correlation analyses between TRIM5 and TIICs
According to the above results, TRIM5 may play an immune-related role in glioma. And tumor-infiltrating lymphocytes were widely believed to be the predictors of OS in cancer patients[19]. Therefore, we then used CIBERSORT algorithm to calculate the proportion of 22 immune cells in the low and high TRIM5 expression groups. As shown in Fig. 6a, gamma delta T cells, activated NK cells, Monocytes, activated mast cells and Eosinophils were decreased (all P < 0.05) in high TRIM5 level group, while M0 and M1 Macrophages and resting mast cells were increased (all P < 0.05) compared with low TRIM5 level group. Moreover, we compared the correlations between 22 immune cell types (Fig. 6b). The heat map showed a strong negative correlation between M0 Macrophages and Monocytes (r = ? 0.76), and moderate negative correlation between M2 Macrophages and activated mast cells (r = ? 0.53), as well as resting NK cells and activated NK cells (r = ? 0.48). We also observed that TRIM5 was positively correlated with B cells (r = 0.205, P = 6.50e?06), CD4 + T cells (r = 0.312, P = 2.81e?12), Dendritic cells (r = 0.433, P = 2.55e?23), Macrophages (r = 0.421, P = 5.86e?22) and Neutrophils (r = 0.418, P = 1.15e?21) in LGG, and CD4 + T cells (r = 0.297, P = 4.22e?04), Dendritic cells (r = 0.353, P = 2.31e?05) and Macrophages (r = 0.19, P = 2.60e?02) in GBM by using TIMER (Fig. 7). These results showed that TRIM5 might influence the immune infiltration of glioma patients.
圖6
The proportion of 22 immune cells effected by TRIM5 expression
a. the ratio of 22 immune cells in glioma in low (blue) and high (red) TRIM5 expression groups, *
a. the ratio of 22 immune cells in glioma in low (blue) and high (red) TRIM5 expression groups, *
圖7
Correlation analyses between TRIM5 and immune infiltration
3 Discussion
Glioma is one of the most common tumors of brain. Surgery is the general treatment for glioma, but postoperative chemoradiotherapy is often needed because of the invasive growth characteristics of glial cells[20-21]. GBM is the most malignant glioma with a median lifetime of 12-16 months and a poor prognosis because of tumor recurrence and resistance[22-23]. Therefore, biomarkers for early diagnosis of glioma can provide guidance for treatment and prognosis of patients.
Over the past few years, there have been many reports on the functions of TIRM5 in autophagy[24-26]. As an autophagic receptor, TRIM5 protects cells from HIV-1 infection by autophagically degrading the retrovirus capsid[7]. Previous study has shown that autophagy was a regulator of homeostasis and survival and had a dual role in cancers[27]. One side, autophagy deficiency could lead to oxidative stress and genome instability which can cause tumorigenesis and progression[28-30]. On the other hand, the upregulation of autophagy can help the tumor cells growth and survival[31-33]. Activation of autophagy has been found in many malignant brain tumors, while autophagy induction in GBM is controversial[34]. In addition, numerous genes involved in autophagy, such as SH3GLB1, LAMP2, ULK1, SQSTM1/p62 and MAPK8IP1, have been found to participate in the diagnosis and prognosis of glioma[34-35]. However, the expression and prognostic role of TRIM5 in glioma are still unknown. In this study, we elucidated the potential function of TRIM5 in glioma and found that TRIM5 was upregulated in glioma tissues by Oncomine, GEPIA and TCGA databases. Meanwhile, TRIM5 was higher in GBM samples than that in LGG. Therefore, we speculate that TRIM5 may be a useful biomarker to distinguish the malignancy degree of glioma.
We further explored the prognostic value of TRIM5 in glioma. Obviously, patients with high TRIM5 expression level had shortened OS. Further, the increased expression of TRIM5 decreased the 1-year, 3-year and 5-year OS in glioma patients (LGG+GBM) and LGG patients without influencing GBM patients. In line with the previous study, diagnosis age, histological type and grade were risk factors for OS of LGG patients[36]. In our study, we also found that TRIM5 was an independent prognostic factor for glioma patients. Besides, the clinical correlation analysis also showed that TRIM5 was positively correlated with stage and histological type, which again confirmed the idea that TRIM5 could act as an indicator biomarker to distinguish degree of glioma malignancy. Nonetheless, more research is needed to determine whether TRIM5 prediction is appropriate for GBM patients. And, the relationship between TRIM5 and existing glioma prognostic biomarkers (such as IDH 1/2 and 1p/19q co-deletion) will be further explored.
And then we look for the molecular pathways that TRIM5 might be involved in. The STRING database revealed that TRIM5 may interact with IRF3, IRF7, OAS1, OAS2, OAS3, OASL, GBP1, PML, BTBD1 and BTBD2 proteins. Previous study illustrated that IRF3 and IRF7 were two transcription factors participating in virus induction and IFN production[37]. In addition, IRF3 and IRF7 deficiency can lead to the lack of innate immune response caused by IFNα and IFNβ absence after viral infection[38]. Human 2'-5'- oligoadenylate synthetase (OAS) proteins have been found to contribute to the antiviral property via inducing apoptosis of infection cells[39]. In the KEGG analysis, TRIM5 upregulation was concentrated in apoptosis pathway, cancer-related molecular pathways and immune-related signaling pathways. In view of the characteristics of TRIM5 and the above predicted results, we speculated that TRIM5 could play a role in glioma by immune-related mechanisms.
To verify the above assumption, we used the CIBERSORT and TIMER database for analysis. As expected, CIBERSORT analysis revealed that activated NK cells, activated mast cells and Monocytes (all P < 0.05) were notably reduced in the group with high TRIM5 level, whereas M0 and M1 Macrophages (all P < 0.001) were remarked increased. And, strongly negative correlation existed between M0 macrophages and monocytes (r = ? 0.76). Moreover, our finding showed that TRIM5 expression was associated with B cells, CD4+ T cells, Dendritic cells, Macrophages and Neutrophils in LGG, and CD4+ T cells, Dendritic cells and Macrophages in GBM. NK cells are innate lymphocytes that can rapidly secrete cytokines such as IFNγ or TNFα to cancer cells, thereby inhibiting angiogenesis and carcinogenesis[40-41]. In recurrent malignant glioma patients, autologous NK cells injection therapy was partially effective with no severe neurological toxicity[42]. From our results, we can infer that increased TRIM5 inhibits the infiltration and immune response of NK cells in glioma. Macrophages are recognized as myeloid-derived suppressor cells (MDSCs), which have the ability to induce strong immunosuppressive[43]. Previous study has shown that blocking monocytes transformation into MDSCs can produce more robust anti-tumor response in patients with non-immunosuppressed GBM[44]. Consistent with above results, we revealed that the high expression of TRIM5 could reduce the proportion of Monocytes and increase the proportion of Macrophages, resulting in a poor prognosis in glioma patients, which may be related to immunosuppression. Mast cells participated in innate and specific immunity, which can suppress the stemness of glioma cells, providing a feasibility for anti-tumor immune response[45]. Our study also showed the reduction of activated mast cells proportion in high TRIM5 expression glioma. These data confirm that TRIM5 may be an auxiliary biomarker in immunotherapy of glioma. However, the molecular mechanism by which TRIM5 affects immune infiltration in glioma remains poor understood and need to be further explored.
In conclusion, we demonstrated that elevated TRIM5 was correlated with the increase of malignancy in glioma and could act as an effective prognostic biomarker for glioma patients. Moreover, this is the first report to identify the correlation between TRIM5 expression and immune cells infiltration. These data may provide a theoretical basis for the prognosis and treatment of patients with glioma, and further exploration of the deep mechanism of TRIM5 in glioma immune infiltration and clinical validation are of great significance.
Acknowledgments
We sincerely thanked the public database: Oncomine, GEPIA, TCGA, CIBERSORT and TIMER.
Conflict of interest
The authors declare no potential conflicts of interest.
Introduction
Glioma is the most common brain tumor with high morbidity of 6 per 100 000 people each year[1-2]. Based on histopathology, gliomas can be classified into gradesⅠ-Ⅳ. Of which, grade IV is regarded as glioblastoma multiforme (GBM), with shorter overall survival (OS) even with routine treatment[3]. Currently, postoperative radiotherapy and chemotherapy are the main treatment for glioma, but the prediction of clinical outcome is still inaccurate due to the histological grading variability, especially for low grade glioma (LGG) patients (gradesⅠ-Ⅲ, including pilocytic astrocytoma, anaplastic astrocytoma, oligodendroglioma, oligodendrocytoma, anaplastic oligodendrocytoma and anaplastic oligodendroglioma and etc.)[4-5]. Therefore, efforts are needed to identify more biomarkers to evaluate the prognosis for glioma.
The tripartite motif (TRIM) family, constituted as immune-regulated proteins, played a general role in autophagy and involved in immune and tumor processes[6]. Tripartite motif 5 (TRIM5), well-known as a retrovirus limiting factor, has the function of autophagic degradation, protecting immune cells from HIV-1 infection[7]. Bioinformatics analysis by Srihari et al. found that TRIM5 was associated with BRCA1 and BRCA2 breast cancer[8]. A paper from Leal et al., revealed that the restriction factor of TRIM5 may act as a significant role in defense against neuroinflammation[9]. Unfortunately, the function of TRIM5 remains much less explored in glioma.
In current study, we aimed to explore the prognostic role and potential mechanisms of TRIM5 in glioma. Firstly, we obtained the TRIM5 expression based on Oncomine, Gene Expression Profiling Interactive Analysis (GEPIA) and The Cancer Genome Atlas (TCGA) databases and analyzed the prognostics value of TRIM5 by TCGA database. Then, Oncomine, STRING databases and Kyoto Encyclopedia Genes and Genomes (KEGG) analysis were used to predict the possible pathways of TRIM5. Finally, immune infiltration analysis was performed between TRIM5 and glioma by using CIBERSORT and TIMER analyses.
1 Materials and Methods
1.1 Database
The TRIM5 mRNA expression in glioma was compared by the Oncomine (https://www.oncomine.org/) database[10] including Murat Brain and Sun Brain glioma studies[11-12]. TCGA (https://cancergenome.nih.gov/) database[13] was also searched to obtain TRIM5 expression and clinical data of glioma including 5 normal brain tissues, 529 LGG tissues and 169 GBM tissues. GEPIA (http://gepia.cancer-pku.cn/detail.php) database[14], an online webserver including 8587 normal and 9736 tumor samples, was utilized to analyze the TRIM5 expression (Tumor or Normal) from TCGA. Besides, data of 4 clinicopathological features including age, gender, grade and histological type were also extracted from TCGA.
1.2 Survival analysis
To explore associations between TRIM5 expression and glioma patient OS, TRIM5 expression level and clinical data of glioma were downloaded from TCGA. Then, package edgeR and R Limma package were used to transform the information. After excluding patients with incomplete information, we performed OS analysis by using R package.
1.3 Clinical correlation and Cox risk analyses
The relevant clinical and survival information of LGG and GBM patients was obtained from TCGA database. We eliminated patients without complete clinical data and acquired 668 patients (508 LGG and 160 GBM) for analyses. As for clinical correlation analysis, LGG and GBM patients were divided into high and low TRIM5 groups (grouped according to the median value), with 334 cases in each group.
1.4 Network analysis
To identify the possible interaction networks of TRIM5, we conducted co-expression analysis via Oncomine database with tumor type limited to “brain”. Protein-protein interaction (PPI) analysis of TRIM5 was performed using STRING (https://string-db.org/) database[15]. In addition, KEGG pathway enrichment analysis was conducted by GSEA 4.0.3 software with 1000 permutations[16]. Normal P value < 0.05 and false discovery rate (FDR) < 0.05 was the filter.
1.5 Immune infiltration analysis
CIBERSORT (https://cibersort.stanford.edu/) online tool is a deconvolution algorithm utilized to explore the correlation between gene expression and tumor-infiltrating immune cells (TIICs)[17]. We uploaded the standard TRIM5 expression data of 703 samples from TCGA to CIBERSORT algorithm running, and then selected samples with P < 0.05 to the final study cohort. Subsequently, the CIBERSORT obtained the proportion of TIICs in the high- and low-TRIM5 groups. In addition, TIMER (http://timer.cistrome.org/)[18] was utilized to report the correlation between immune infiltration and TRIM5 level in glioma. The analyzed immune cells include B cells, CD4+, Dendritic cells, CD8+ T cells, Macrophages and Neutrophils. Kaplan-Meier curves were downloaded from TIMER to analyze the relationship between immune infiltration or TRIM5 level and OS of LGG and GBM patients respectively.
1.6 Statistical analysis
The statistical analyses were conducted using R software 3.6.3 and GraphPad Prism 8.0.1. Student’s t-test or ANOVA were utilized to assess the TRIM5 difference between glioma tissues and normal tissues. The relationship between TRIM5 level and clinicopathological features was compared by chi-square tests. Survive analyses were evaluated by the Kaplan-Meier with the log-rank test and Cox regression analysis. P < 0.05 was considered significant.
2 Results
2.1 High expression of TRIM5 in glioma
We found TRIM5 was elevated in central nervous system (CNS) and brain tumors compared with normal tissues via Oncomine (Fig. 1a). For verification, we conducted meta-analysis of TRIM5 expression in 4 analyses in Oncomine database (Fig. 1b). Compared with normal brain tissues, the TRIM5 level was statistically increased in GBM or Astrocytoma tissues (P < 0.001, Fig. 1c and 1d). Compared with LGG tissues, the TRIM5 expression was significantly higher in GBM tissues based on Sun Brain and TCGA Brain (P < 0.001, Fig. 1e and 1f). In order to further verify the above results, we observed that TRIM5 was increased in numerous cancers including LGG and GBM by using GEPIA tool (Fig. 2a). As expected, TRIM5 expression in LGG and GBM was notably higher than that in normal control (Fig. 2b). These results demonstrated that TRIM5 was upregulated in glioma and might be an indicator to predict the malignancy of glioma.
圖1
Elevated TRIM5 expression in glioma
a. TRIM5 expression between cancer and normal tissues of Brain and CNS cancer in Oncomine; b. meta-analysis of TRIM5 expression in 4 analyses; c-f. TRIM5 expression in Murat Brain, Sun Brain and TCGA Brain
a. TRIM5 expression between cancer and normal tissues of Brain and CNS cancer in Oncomine; b. meta-analysis of TRIM5 expression in 4 analyses; c-f. TRIM5 expression in Murat Brain, Sun Brain and TCGA Brain
圖2
TRIM5 expression based on GEPIA database
a. TRIM5 was notably increased in variety tumors; b. TRIM5 expression in LGG and GBM compared with normal control. *,
a. TRIM5 was notably increased in variety tumors; b. TRIM5 expression in LGG and GBM compared with normal control. *,
2.2 Relationship between TRIM5 level and prognosis in glioma
We performed Kaplan-Meier to analyze the influence of TRIM5 in OS of glioma patients (LGG+GBM), which showed that high TRIM5 expression would lead to a worse outcome (P < 0.001, Fig. 3a). Besides, subgroup analysis indicated that higher TIRM5 was risk factor for 1-year, 3-year and 5-year OS in glioma patients (all P < 0.001, Fig. 3b–3d). Subsequently, we performed prognostic analysis in LGG and GBM patients respectively, the results revealed that increased TRIM5 predicted poor OS in patients with LGG (all P < 0.001, Fig. 3e–3h), while was not correlated to prognosis of GBM patients (all P > 0.05, Fig. 3i–3l).
圖3
TRIM5 upregulation was associated with poor prognosis in glioma on TCGA database
a–d. OS of LGG+GBM patients; e–h. OS of LGG patients; i–l. OS of GBM patients
a–d. OS of LGG+GBM patients; e–h. OS of LGG patients; i–l. OS of GBM patients
2.3 Clinical correlation and Cox analyses based on TCGA
We performed the correlation analyses between TRIM5 level and clinicopathological characteristics in glioma patients by TCGA (LGG+GBM). As shown in Tab. 1, high TRIM5 expression was significantly correlated to older age (χ2 = 44.310, P < 0.001), higher grade (χ2 = 130.100, P < 0.001) and more malignant histological type (χ2 = 125.500, P < 0.001). Above results illustrated that TRIM5 contributed to predict glioma progression.
表1
Correlations between TRIM5 expression and clinicopathological characteristics in glioma patients (LGG+GBM)
To further confirm the prognostic role of TRIM5, univariate and multivariate Cox analyses were employed (LGG and GBM patients). As shown in Tab. 2, TRIM5 expression was a risk factor (HR = 1.481, 95% CI = 1.201–1.824, P = 0.000) for glioma OS. Besides, age (HR = 1.047, 95% CI = 1.035–1.060, P < 0.000) and grade (HR = 3.110, 95% CI = 2.300–4.206, P < 0.000) were both risk factors. Next, we continued to perform analyses in LGG patients and the results showed that TRIM5 expression (HR = 1.816, 95% CI = 1.419–2.322, P < 0.000) could be acted as an independent prognostic factor. In addition, age (HR = 1.064, 95% CI = 1.047–1.082, P < 0.000), grade (HR = 1.915, 95% CI = 1.219–3.009, P =0.005) and histological type (HR = 0.709, 95% CI = 0.566–0.890, P = 0.003) were all risk factors for LGG patients OS (Tab. 3).
表2
Univariate and multivariate Cox analyses of OS in LGG+GBM patients
表3
Univariate and multivariate Cox analyses of OS in LGG patients
2.4 The network interactions with TRIM5
To explore the possible molecular mechanisms of TRIM5, we predicted some co-expressed genes with TRIM5 in glioma by Oncomine database. We found strong correlation between NMI, TRIM22 and DDX60 etc. with TRIM5 with the correlation score > 0.7 (Fig. 4a). STRING indicated that IRF3, IRF7, OAS1, OAS2, OAS3, OASL, GBP1, PML, BTBD1 and BTBD2 were contacted with TRIM5 (Fig. 4b). Lastly, KEGG pathway enrichment of TRIM5 showed that highly expressed TRIM5 was mainly enriched in “pathways in cancer”, “apoptosis”, and some immune-related pathways (Fig. 5). The specific information of the enrichment pathways is shown in Tab. 4.
圖4
The potential interaction networks prediction of TRIM5
a. transcriptome prediction of TRIM5 in glioma in Oncomine database; b. TRIM5 protein interaction networks in STRING
a. transcriptome prediction of TRIM5 in glioma in Oncomine database; b. TRIM5 protein interaction networks in STRING
表4
Results of part KEGG enrichment analyses
圖5
Enrichment plots from GSEA analysis
2.5 Correlation analyses between TRIM5 and TIICs
According to the above results, TRIM5 may play an immune-related role in glioma. And tumor-infiltrating lymphocytes were widely believed to be the predictors of OS in cancer patients[19]. Therefore, we then used CIBERSORT algorithm to calculate the proportion of 22 immune cells in the low and high TRIM5 expression groups. As shown in Fig. 6a, gamma delta T cells, activated NK cells, Monocytes, activated mast cells and Eosinophils were decreased (all P < 0.05) in high TRIM5 level group, while M0 and M1 Macrophages and resting mast cells were increased (all P < 0.05) compared with low TRIM5 level group. Moreover, we compared the correlations between 22 immune cell types (Fig. 6b). The heat map showed a strong negative correlation between M0 Macrophages and Monocytes (r = ? 0.76), and moderate negative correlation between M2 Macrophages and activated mast cells (r = ? 0.53), as well as resting NK cells and activated NK cells (r = ? 0.48). We also observed that TRIM5 was positively correlated with B cells (r = 0.205, P = 6.50e?06), CD4 + T cells (r = 0.312, P = 2.81e?12), Dendritic cells (r = 0.433, P = 2.55e?23), Macrophages (r = 0.421, P = 5.86e?22) and Neutrophils (r = 0.418, P = 1.15e?21) in LGG, and CD4 + T cells (r = 0.297, P = 4.22e?04), Dendritic cells (r = 0.353, P = 2.31e?05) and Macrophages (r = 0.19, P = 2.60e?02) in GBM by using TIMER (Fig. 7). These results showed that TRIM5 might influence the immune infiltration of glioma patients.
圖6
The proportion of 22 immune cells effected by TRIM5 expression
a. the ratio of 22 immune cells in glioma in low (blue) and high (red) TRIM5 expression groups, *
a. the ratio of 22 immune cells in glioma in low (blue) and high (red) TRIM5 expression groups, *
圖7
Correlation analyses between TRIM5 and immune infiltration
3 Discussion
Glioma is one of the most common tumors of brain. Surgery is the general treatment for glioma, but postoperative chemoradiotherapy is often needed because of the invasive growth characteristics of glial cells[20-21]. GBM is the most malignant glioma with a median lifetime of 12-16 months and a poor prognosis because of tumor recurrence and resistance[22-23]. Therefore, biomarkers for early diagnosis of glioma can provide guidance for treatment and prognosis of patients.
Over the past few years, there have been many reports on the functions of TIRM5 in autophagy[24-26]. As an autophagic receptor, TRIM5 protects cells from HIV-1 infection by autophagically degrading the retrovirus capsid[7]. Previous study has shown that autophagy was a regulator of homeostasis and survival and had a dual role in cancers[27]. One side, autophagy deficiency could lead to oxidative stress and genome instability which can cause tumorigenesis and progression[28-30]. On the other hand, the upregulation of autophagy can help the tumor cells growth and survival[31-33]. Activation of autophagy has been found in many malignant brain tumors, while autophagy induction in GBM is controversial[34]. In addition, numerous genes involved in autophagy, such as SH3GLB1, LAMP2, ULK1, SQSTM1/p62 and MAPK8IP1, have been found to participate in the diagnosis and prognosis of glioma[34-35]. However, the expression and prognostic role of TRIM5 in glioma are still unknown. In this study, we elucidated the potential function of TRIM5 in glioma and found that TRIM5 was upregulated in glioma tissues by Oncomine, GEPIA and TCGA databases. Meanwhile, TRIM5 was higher in GBM samples than that in LGG. Therefore, we speculate that TRIM5 may be a useful biomarker to distinguish the malignancy degree of glioma.
We further explored the prognostic value of TRIM5 in glioma. Obviously, patients with high TRIM5 expression level had shortened OS. Further, the increased expression of TRIM5 decreased the 1-year, 3-year and 5-year OS in glioma patients (LGG+GBM) and LGG patients without influencing GBM patients. In line with the previous study, diagnosis age, histological type and grade were risk factors for OS of LGG patients[36]. In our study, we also found that TRIM5 was an independent prognostic factor for glioma patients. Besides, the clinical correlation analysis also showed that TRIM5 was positively correlated with stage and histological type, which again confirmed the idea that TRIM5 could act as an indicator biomarker to distinguish degree of glioma malignancy. Nonetheless, more research is needed to determine whether TRIM5 prediction is appropriate for GBM patients. And, the relationship between TRIM5 and existing glioma prognostic biomarkers (such as IDH 1/2 and 1p/19q co-deletion) will be further explored.
And then we look for the molecular pathways that TRIM5 might be involved in. The STRING database revealed that TRIM5 may interact with IRF3, IRF7, OAS1, OAS2, OAS3, OASL, GBP1, PML, BTBD1 and BTBD2 proteins. Previous study illustrated that IRF3 and IRF7 were two transcription factors participating in virus induction and IFN production[37]. In addition, IRF3 and IRF7 deficiency can lead to the lack of innate immune response caused by IFNα and IFNβ absence after viral infection[38]. Human 2'-5'- oligoadenylate synthetase (OAS) proteins have been found to contribute to the antiviral property via inducing apoptosis of infection cells[39]. In the KEGG analysis, TRIM5 upregulation was concentrated in apoptosis pathway, cancer-related molecular pathways and immune-related signaling pathways. In view of the characteristics of TRIM5 and the above predicted results, we speculated that TRIM5 could play a role in glioma by immune-related mechanisms.
To verify the above assumption, we used the CIBERSORT and TIMER database for analysis. As expected, CIBERSORT analysis revealed that activated NK cells, activated mast cells and Monocytes (all P < 0.05) were notably reduced in the group with high TRIM5 level, whereas M0 and M1 Macrophages (all P < 0.001) were remarked increased. And, strongly negative correlation existed between M0 macrophages and monocytes (r = ? 0.76). Moreover, our finding showed that TRIM5 expression was associated with B cells, CD4+ T cells, Dendritic cells, Macrophages and Neutrophils in LGG, and CD4+ T cells, Dendritic cells and Macrophages in GBM. NK cells are innate lymphocytes that can rapidly secrete cytokines such as IFNγ or TNFα to cancer cells, thereby inhibiting angiogenesis and carcinogenesis[40-41]. In recurrent malignant glioma patients, autologous NK cells injection therapy was partially effective with no severe neurological toxicity[42]. From our results, we can infer that increased TRIM5 inhibits the infiltration and immune response of NK cells in glioma. Macrophages are recognized as myeloid-derived suppressor cells (MDSCs), which have the ability to induce strong immunosuppressive[43]. Previous study has shown that blocking monocytes transformation into MDSCs can produce more robust anti-tumor response in patients with non-immunosuppressed GBM[44]. Consistent with above results, we revealed that the high expression of TRIM5 could reduce the proportion of Monocytes and increase the proportion of Macrophages, resulting in a poor prognosis in glioma patients, which may be related to immunosuppression. Mast cells participated in innate and specific immunity, which can suppress the stemness of glioma cells, providing a feasibility for anti-tumor immune response[45]. Our study also showed the reduction of activated mast cells proportion in high TRIM5 expression glioma. These data confirm that TRIM5 may be an auxiliary biomarker in immunotherapy of glioma. However, the molecular mechanism by which TRIM5 affects immune infiltration in glioma remains poor understood and need to be further explored.
In conclusion, we demonstrated that elevated TRIM5 was correlated with the increase of malignancy in glioma and could act as an effective prognostic biomarker for glioma patients. Moreover, this is the first report to identify the correlation between TRIM5 expression and immune cells infiltration. These data may provide a theoretical basis for the prognosis and treatment of patients with glioma, and further exploration of the deep mechanism of TRIM5 in glioma immune infiltration and clinical validation are of great significance.
Acknowledgments
We sincerely thanked the public database: Oncomine, GEPIA, TCGA, CIBERSORT and TIMER.
Conflict of interest
The authors declare no potential conflicts of interest.
