Sp1 induced gene TIMP1 is related to immune cell infiltration in glioblastoma

Tumor immune microenvironment exerts a profound effect on the population of infiltrating immune cells. Tissue inhibitor of matrix metalloproteinase 1 (TIMP1) is frequently overexpressed in a variety of cells, particularly during inflammation and tissue injury. However, its function in cancer and immunity remains enigmatic. In this study, we find that TIMP1 is substantially up-regulated during tumorigenesis through analyzing cancer bioinformatics databases, which is further confirmed by IHC tissue microarrays of clinical samples. The TIMP1 level is significantly increased in lymphocytes infiltrating the tumors and correlated with cancer progression, particularly in GBM. Notably, we find that the transcriptional factor Sp1 binds to the promoter of TIMP1 and triggers its expression in GBM. Together, our findings suggest that the Sp1-TIMP1 axis can be a potent biomarker for evaluating immune cell infiltration at the tumor sites and therefore, the malignant progression of GBM.


Tumor immune microenvironment exerts a profound effect on the population of infiltrating immune cells. Tissue inhibitor of matrix metalloproteinase 1 (TIMP1) is frequently overexpressed in a variety of cells, particularly during inflammation and tissue injury. However, its function in cancer and
immunity remains enigmatic. In this study, we find that TIMP1 is substantially up-regulated during tumorigenesis through analyzing cancer bioinformatics databases, which is further confirmed by IHC tissue microarrays of clinical samples. The TIMP1 level is significantly increased in lymphocytes infiltrating the tumors and correlated with cancer progression, particularly in GBM. Notably, we find that the transcriptional factor Sp1 binds to the promoter of TIMP1 and triggers its expression in GBM. Together, our findings suggest that the Sp1-TIMP1 axis can be a potent biomarker for evaluating immune cell infiltration at the tumor sites and therefore, the malignant progression of GBM.

Results
Identification of prognostic biomarkers in patients with glioblastoma. In order to identify the potential prognostic biomarkers, we retrieved the gene expression profiles from glioblastoma patients in TCGA. After excluding the patients who died within 30 days and protein-coding genes with less than ten counts in at least 75% of the total subjects, a total of 14,801 genes in 169 tumorous and five control subjects were analyzed using edgeR. We found that the expression of 2413 genes was upregulated, whereas 2435 genes were downregulated ( Fig. 1A and Table S1). The top 30 most upregulated or downregulated genes were identified (Fig. 1B). Further analysis using the univariate factor cox regression model revealed 33 differentially expressed genes, which derived from 2413 upregulated genes, as potential prognostic biomarkers (Fig. S1C). Notably, LASSO regression was used to investigate the prognosis value of these potential biomarkers and we identified a risk score model ( Fig.S1D-E). Indeed, when we separated these patients into high-risk and low-risk groups according to their risk score, a significantly different survival curve was observed ( Fig. 2A). The gene enrichment analysis revealed that these 33 genes significantly contribute to cellular growth and development (Fig. 2B).

Identification of prognostic biomarkers correlating with immune infiltration. TIMER was uti-
lized to investigate the relationships of the levels of these 33 genes in glioblastoma with the recruitment of immune cells, including B cells, CD8+ T cells, CD4+ T cells, macrophages, neutrophils, and dendritic cells. We found that TIMP1, ITGA5, FCGR2B, UPP1, ISG20, TSPAN4, and LOXL1 are potential biomarkers correlated with the immune infiltration events in patients with glioblastoma ( Fig. 3 and Table 1). In all cases, the gene expression is significantly negatively related to tumor purity. TIMP1, ITGA5, UPP1, ISG20, TSPAN4, and www.nature.com/scientificreports/ LOXL1 have a significant positive correlation with infiltrating levels of dendritic cells, and FCGR2B has a significant positive correlation with infiltrating levels of neutrophils and dendritic cells.
In addition, we used the SurvExpress analysis to further assess the prognostic value of TIMP1, ITGA5, FCGR2B, UPP1, ISG20, TSPAN4, and LOXL1 using other glioblastoma patient cohorts, including TCGA Glioblastoma and GSE4412. Most of these seven genes displayed significant correlations with the survival of the glioblastoma patients in other databases examined (Fig. 4).
Correlations of TIMP1 level with survival and immune cell infiltration in the pan-cancer analysis. Earlier, we found that seven genes are associated with both prognosis and immune cell infiltration in glioblastoma. Amongst them, TIMP1 is one of the closest genes correlated with tumor development and immune responses 18 . TIMP1 is secreted and functions as a soluble protein, but its oncogenic role is not well understood 18 . Hence, TIMP1 was selected for further characterization.
To determine whether TIMP1 is differentially expressed in cancer, we found that TIMP1 actually upregulated in the majority of tumor types in comparison to normal tissues in Oncomine database (Fig. S1A,B). We then evaluated the prognostic value of TIMP1 and used PROGgeneV2 to investigate the potential correlations between TIMP1 level and survival outcome of cancer patients. As a result, we found higher TIMP1 levels are associated with worse survival rates (Fig.S2, S3 and Table S2).
Furthermore, we assessed the association between TIMP1 and tumor-infiltrating cells among multiple forms of human cancer in TIMER database. Analysis through GENE module revealed that TIMP1 expression significantly correlates with tumor purity, the levels of B cells, CD8+ T cells, CD4+ T cells, macrophages, neutrophils, and dendritic cells (Table S3). Importantly, when cross-referencing the results shown in Figure 4, we discovered that TIMP1 serves as potential biomarker for both survival and immune infiltration in GBM, STAD, HNSC, and LGG. More specifically, TIMP1 positively correlates with the levels of dendritic cell infiltration in GBM, STAD, HNSC, and LGG and with the levels of macrophage infiltration in STAD and HNSC. Moreover, high levels of TIMP1 are associated with high neutrophil infiltration in LGG (Fig. S4). NKX2−5  HOXD13  HOXD9  HOXC10  ADAMDEC1  HP  HOXA2  HOXA3  HOXA5  HOXA10  HOXD11  SAA1  HOXA7  HOXA6  SHOX2  HOXA4  LTF  HOXB4  HOXD8  EN1  PI3  PITX1  HOXB7  MYBL2  SAA2  POSTN  UBE2C  MMP9  RBFOX1  SYT1  TMEM130  TESPA1  GDA  SLC17A7  OPALIN  SLC30A3  SV2C  SLC12A5  SULT4A1  CAMK2A  SLC6A17  CREG2  FRMPD4  CACNA1B  SLC8A2  ATP2B3  MPPED1  SST  PACSIN1  SVOP  ATP8A2  RASAL1  DDN  GABRA5  CHD5  SV2B  GRIN1  GRIN2B  RYR2  PRKCG   Type   Type GBM Normal  www.nature.com/scientificreports/ To investigate further the relationships between TIMP1 and the diverse tumor-infiltrating immune cells, we studied the correlations between TIMP1 expression levels and those of immune markers for various immune cells in GBM, STAD, HNSC, and LGG using the CORRELATION Module in the TIMER database. We found significant positive correlations between TIMP1 expression and 44 and 47 out of the total 57 immune markers in STAD and LGG, respectively. Moreover, 18 and 41 out of the total 56 immune markers in GBM and HNSC were significantly correlated with TIMP1 expression, respectively (Table S4).
Similarly, cross-examination of STAD and LGG patient cohorts in GEPIA database, revealed significant correlations between the expression of TIMP1 and that of immune markers in CD8+ T cells, T cells, B cells, monocytes, tumor-associated macrophages, and other macrophages, indicating that TIMP1 may promote the polarization of macrophage in STAD and LGG tumors ( Fig. S5 and Table 2). TIMP1 level associates with poor prognosis in GBM and STAD patients. Previously, we assessed the expression of TIMP1 in GBM and STAD patients through Oncomine database (Fig. S1A,B). Moreover, we confirmed by IHC staining of GBM and STAD tissue microarrays ( Fig. 5A and S6A,B). We evaluated the correlation of TIMP1 expression and clinicopathological characters among 180 glioma patients (Table 3) and 94 stomach adenocarcinoma patients ( Table 4). The outcomes indicated that TIMP1 expression remarkably correlated with tumor encapsulation and recurrence. Furthermore, highly expressed TIMP1 can be recognized as an independent prognostic biomarker, as the p-value of overall survival (OS) was 0.007 and the p-value of diseasE−free survival (DFS) was less than 0.0001 in glioma patients (Fig. 5B). Notably, similar observations were detected in stomach adenocarcinoma patients (Fig. S6C). Besides, TIMP1 levels were also positively associated with the degree of lymph node metastasis (Fig.S6D), with the deeper the infiltration of cancer cells, the higher the level of expression of TIMP1 (Fig. S6E). Notably, univariable Cox regression analysis revealed that TIMP1 is significantly associated with poorer outcomes (Fig. 5C and S6F). Altogether, our results demonstrated that high TIMP1 expression is tightly linked to the worse prognosis of GBM and STAD patients.  www.nature.com/scientificreports/  www.nature.com/scientificreports/ by using TIMP1 promoter DNA sequences (−2000 bp ~ + 1000 bp) in PROMO database (Fig. 6A). Next, we assessed the potential correlation of the expression levels of the transcriptional factors with TIMP1 expression level, the survival and the differential expression in GBM patients from the TCGA and CGGA databases. The results indicated Sp1 is simultaneously in line with the three conditions ( Fig. S7A-D). Besides, Sp1 is over- Table 1. TIMER analysis of the correlation between immune infiltration and prognostic biomarkers.   1000  1500  2000  2500  43  31  26  18  13  7  6  5  4  1  1  42  21  9  6  4  2  2  2  0  0  0   0  5 00  1000  1500  2000  2500  43  31  26  19  13  8  7  6  4  1  1  42  21  9  5  4  1  1  1  0  0  0   0  5 00  1000  1500  2000  2500  43  33  28  19  14  8  7  6  3  1  1  42  19  7  5  3  1  1  1  1  0  0   0  500  1000  1500  2000  2500  43  30  24  17  12  7  7  6  3  1  1  42  22  11  7  5  2  1    www.nature.com/scientificreports/ expressed in brain glioma cancer and can facilitate proliferation and invasion of glioma cells 47 . More importantly, Sp1 correlates with immune cell infiltration, which corresponds with the results of TIMP1 of our previous findings. As a consequence, Sp1 was selected for further investigation. Sp1 ChIP-seq data from HEK293T cell (GSE92217) in GEO database was examined, and the result indicated a binding signal exists within the promoter region of TIMP1 (Fig. 6B). We then used the PROMO database to predicate the Sp1 binding sites within the TIMP1 promoter and identified 3 putative binding sites, BS1 (−138 to −128), BS2 (−48 to −38), BS3 (+248 to +258) (Fig. 6C).

Sp1 binds to TIMP1 promoter and enhances TIMP1 expression.
To identify the potential role of Sp1 on TIMP1 expression, we constructed Sp1 shRNAs to study their effects on TIMP1 expression. Both the RNA and protein levels of TIMP1 were decreased in response to Sp1 knockdown in glioma cells (Fig. 7A-D). Next, we developed luciferase reporter plasmids, including three predictive binding sites, to detect the efficacy of TIMP1 transcription. The results indicated that the TIMP1 promoter region responds to Sp1 at the −282/ +756bp (Fig. 7E,F). To further validate this finding and map the direct binding site, chromatin precipitation (ChIP) was performed. The result showed that Sp1 could interact predominantly with the binding site −48/ −38 of the TIMP1 promoter (Fig. 7G,H). Altogether, our findings indicated that Sp1 binds to the promoter of TIMP1 to enhance its expression.
The ratio of Sp1 to TIMP1 is a better biomarker to predict the prognosis of GBM. Either Sp1 or TIMP1 was highly expressed in glioblastoma patients and correlated with poor prognosis. However, whether the signature of Sp1 and TIMP1 could be a better predictive biomarker is unclear. Therefore, we evaluated the relationship between OS and the group based on TIMP1 and Sp1 expression (Sp1 high /TIMP1 high , Sp1 high /TIM-P1 low , Sp1 low /TIMP1 high or Sp1 low /TIMP1 low groups) in GBM patients from CGGA database. The results indicated that high expression of TIMP1 and Sp1 correlate with poor prognosis (Fig. 8A). Relations between the levels of Sp1 andTIMP1 expression and clinicopathologic characteristics also confirm that the combination of Sp1 and TIMP1 could be better in predicting prognosis than the individual marker alone (Fig. 8B and S8A-G). In addition, the significant differences of overall survival between no chemo-/radiotheray and chemo-/radiotherapy indicates gliomal patients may benefit from the treatment based on the expression levels of TIMP and Sp1 ( Fig  S7E,F).

Discussion
TIME plays an essential role in the development of GBM. It has been reported that TIME has both positive and negative effects on cancer development, and these function correlates with poor prognosis and therapy response in GBM patients 19 . However, there is still a lack of reliable biomarkers to predict both the prognosis and immune infiltration in glioma patients. Through analyzing patients' clinical information and genetic profiles from online databases, TIMP1, ITGA5, FCGR2B, UPP1, ISG20, TSPAN4, and LOXL1 were identified as potential prognostic biomarkers for glioblastoma. TIMP1 is a specific in-hibitor of matrix metalloproteinase, and the aberrant upregulation of TIMP1 exists in different types of cancers 22 . TIMP1 increases proliferation and metastasis in www.nature.com/scientificreports/ pancreatic and colorectal cancer 26,30 . Besides, TIMP1 also has a close relationship with immunity. For example, the peripheral level of TIMP1 could stimulate the granulopoiesis in the bone marrow of mice 54 . In our research, TIMP1 has been consistently identified to be a predictive factor of both prognosis and immune infiltration in our differential gene expression analyses. Our results show that high TIMP1 levels are positively associated with increased immune infiltration levels of tumor-infiltrating lymphocytes, which could migrate towards the tumor tissues in glioma, including LGG and GBM.  www.nature.com/scientificreports/ Additionally, TIMP1 expression is also correlated with immune markers, especially those commonly seen in macrophages that are in M1 or M2 phase, suggesting that TIMP1 may have regulative effects on tumor-associated macrophage polarization. These polarized macrophages engage the tumor growth and progression by providing adaptive immunity and inflammatory circuits. There were several researchers also reported the prognostic value of TIMP1. TIMP1 was recognized as a serum biomarker of colorectal cancer through meta-analysis 31 . It also exerts the same role in breast cancer 32 . Moreover, a recent study investigated the correlation between the TIMP family and immune infiltration in glioblastoma; however, it only focused on the TIMP family and lacked further validation 55 . Compared with previous studies, we evaluated the prognosis of glioma patients in large sample data and selected the genes correlating with survival. Then, these targeted genes were used to identify the relationship with tumor immune infiltration, followed by IHC validation of clinical sample chips. Therefore, our screening process is more reasonable and reliable than others. Although this relationship of TIMP1 and glioma patients was confirmed by IHC, we will further validate this result in vivo.
Transcriptional factors regulate various gene expression to mediate the malignant progression of tumors. However, the transcription factors regulating TIMP1 expression are still not uncovered in GBM. In our research, we found Sp1 could trigger the transcription of TIMP1 in GBM. Through prediction from the PROMO database, we identified 26 candidates that can potentially regulate TIMP1. Furthermore, to narrow the number down, we integrated the information of these transcriptional factors with their expression, survival analysis and correlation with TIMP1 in glioma patients. Among thesescreened transcriptional factors, Sp1 was over-expressed in most tumor cells and tissues. In addition, further studies also reported Sp1 regulates the inflammation and the immune cell infiltration of cancers. For example, Mina is an Sp1 regulated gene that functions in inflammation and immunity 56 .
In our study, we investigated the relationship between Sp1 and TIMP1 through database analysis and in vitro experiments. The results indicated Sp1 binds to the TIMP1 promoter region and upregulates its expression in GBM. The transcriptional regulation mechanisms of TIMP1 were also investigated. TIMP1 could be decreased by OGG−1 in response to oxidative stress in human airway epithelial cells 57 . Similarly, TWIST1 could also downregulate TIMP1 mRNA levels in SCCBHY cells 58 . We discovered that immune response relative transcription factors Sp1 could regulate TIMP1 expression in GBM. Our finding suggests a strategy for targeting TIMP1, which is short of an applicable drug in glioma patients, through inhibiting Sp1. There are several small molecular inhibitors that could decrease the expression or activity of Sp1, including WP631, and Doxorubicin  22 . Currently, a number of studies have indicated TIMP1 is highly expressed in cancer cells and its high expression correlates with poor prognosis of patients. However, there are also some reports on TIMP1 expression and regulation in immune cells. For example, IL−19 has been shown to regulate TIMP1 expression through toll-like receptor 2 in macrophages 62 . Lipopolysaccharide (LPS) can also activate TIMP1 expression to inhibit macrophage' function during HCV infection 63 . Sp1 are also be expressed in many cell types due to its general transcriptional function. In addition to cancer cells, Sp1 can transcriptionally regulate T-BET expression, a main regulator of IFN-γ, in NK cells and T cells 64 . Therefore, the function and regulation of TIMP1 and Sp1 in immune cells can have a vital role in tumor pregression and aslo warrants further investigation.  www.nature.com/scientificreports/ Most research on biomarkers focuses on single molecule to predict the prognosis of cancers. Hitherto, due to heterogeneity of cancers in individuals single predictive biomarkers cannot accurately predict the clinical outcome. As the tight regulatory relationship between Sp1 and TIMP1 has been identified here, we aim at investigating both molecules simultaneously as a diagnostic signature in assessing the prognosis of glioma patients. Our preliminary result indicates that TIMP1 and Sp1 together could be a better and more reliable biomarker than either alone in terms of prognostic value of patient survival and clinical characterization.

Conclusion
We analyzed online datasets in Oncomine, TIMER, and some other database comprehensively to identify potential biomarkers for prognosis and immune infiltration in tumor patients, and found that TIMP1 associates not only with prognosis of GBM but also with the levels of immune infiltration seen in those patients. Moreover, database analyses and in vitro experiments demonstrated Sp1 binds to the TIMP1 promoter and enhances TIMP1 expression in GBM. Therefore, the Sp1-TIMP1 axis could be a potential biomarker in clinical applications. Future studies of TIMP1 in larger patient cohorts are necessary to evaluate its value as a reliable biomarker in GBM patients. The potential impact of TIMP1 on lymphocytes, especially regulation of immune genes and other factors, also remains to be determined.

Methods
Oncomine database. Oncomine database is a large tumor gene chip database (https:// www. oncom ine. org). It was used to study TIMP1 levels between tumor tissues and normal tissues. The P-value <0.001 and 2 fold difference was defined as statistical significant difference. TIMER database. TIMER is an integrated assets to assess immune infiltration generally from different tumor types (https:// cistr ome. shiny apps. io/ timer/). We explored TIMER based on deconvolution-based tool to investigate tumor-infiltrating cell changes through the gene expression profiles of data derived from TCGA. We evaluated TIMP1 levels among diverse tumors, and the relationship between TIMP1, ITGA5, FCGR2B, UPP1, ISG20, TSPAN4, and LOXL1 levels and the amount of immunE−infiltrating cells, consisting of CD4+ T cells, CD8+ T cells, B cells, neutrophils, macrophages, and dendritic cells, through different gene modules. We also investigated the correlations of TIMP1 level with gene markers of tumor-infiltrating immune cells. Establishment of risk score model. To identify the prognosis value of differentially expressed genes, least absolute shrinkage and selection operator (LASSO) regression was used to establish the risk score model to predict the survival status of glioma patients. The R package glmnet was used. After we got the differentially expressed genes, the correlation between gene expression levels and the overall survival time of patients was investigated through univariate Cox regression analysis. Thirty-three genes were obtained according to the P-value and HR. Notably, these genes were used to establish the risk score model and receiver operating characteristic (ROC) was used to evaluate the predictive value of the risk score model. Finally, the risk score model, included six genes, SPDYE21, MTHFS, TNFSF14, LBH, ITGB7 and TIMP1.
Gene correlation analysis with GEPIA tool. GEPIA is an online web to analysis the data obtained from TCGA database (http:// gepia. cancer-pku. cn/ index. html). We used this tool to detect the correlation of genes in TIMER. GEPIA was also used to obtain survival curves, through gene expression with the log-rank test and the Mantel-Cox test in different types of cancers. Correlation analysis between gene expression profiles from given sets of TCGA cancer types was performed and the coefficient was generated through Spearman analysis.
Immunohistochemistry. Tissue microassays were applied to explore the connections of TIMP1 with prognosis of GBM and STAD patients. These microassays were purchased from Shanghai Outdo Biotech (Shanghai, China), including 180 GBM tissues, 94 STAD, and 84 STAD corresponding adjacent stomach tissues. Informed consent was obtained from the ethics committee of Shanghai Outdo Biotech Company. All of the methods in this study were in accordance with the approved guidelines, and the experimental protocols were approved by the ethics committee of Shanghai Outdo Biotech Company. IHC was conducted using Human Brain/Gastric Cancer Tissue Chip (HBraG180Su02 and HStmA180Su19). IHC total score was generated according to this equation: IHC total score = staining intensity x staining positive rate. The staining intensity was divided into 4 grades, 0, 1, 2, 3. The staining positive rate was also divided into 4 grades equally. Finally, total score was separated into 2 groups, high expression and low expression based on the detailed data. This judgment process should be conducted by two proficient pathologists independently.
Cell culture. The human glioblastoma cell lines LN229, U251 and the human embryo kidney cell line HEK293T of our study were obtained from American Type Culture Collection (ATCC, Manassas, VA, USA). LN229 and HEK293T were cultured in DMEM (Gibico, Thermo, Inc) consisting of 5% fetal bovine serum (Gibico, Thermo, Inc) and 1% penicillin and streptomycin (Gibico, Thermo, Inc). U251 was grown in DMEM (high glucose) consisting of 10% fetal bovine serum and 1% penicillin and streptomycin. All of these cells were grown at 5% CO 2 and 37 ℃.
Dual luciferase reporter assay. PGL4.15 plasmid was used to construct different promoter segments consisting of TIMP1 predictive binding sites. This assay was conducted using Dual-Luciferase Reporter Assay System Kit (E1910; Promega, Madison, WI, USA). The constructed plasmids were transfected into glioma cells.
Statistical analysis. The outcomes derived from Oncomine were described as P-values, fold changes, and ranks. The HR and P or Cox P values generated from log-rank test methods were used to describe the analysis of Kaplan-Meier plots, PROGgeneV2, and TIMER. The analysis of correlations among different factors was evaluated through the Spearman's statistical methods. The strength of analysis was defined as below: 0.00-0.33 "weak", 0.33-0.67 "moderate", 0.67-1.0"strong". The assay results were analysised through Graphad Prism 8. The outcomes were described as mean ± S.D. Two-sided Student's t-test was used for comparison of two groups. OnE−way ANOVA plus two-sided Dunnett test was performed for analyzing multigroups (each group compared with a control group). The P-value <0.05 was defined as statistically significant.
Ethical approval. The studies involving human participants were reviewed and approved by Ethics committee of Shanghai Outdo Biotech Company, as described in the Methods in more detail. The patients/participants provided their written informed consent to participate in this study. All methods were performed in accordance with relevant guidelines and regulations.