An integrated pan‐cancer analysis of TFAP4 aberrations and the potential clinical implications for cancer immunity

Abstract Studies have shown that transcription factor activating enhancer binding protein 4 (TFAP4) plays a vital role in multiple types of cancer; however, the TFAP4 expression profile is still unknown, as is its value within the human pan‐cancer analysis. The present study comprehensively analysed TFAP4 expression patterns from 33 types of malignancies, along with the significance of TFAP4 for prognosis prediction and cancer immunity. TFAP4 displayed inconsistent levels of gene expression across the diverse cancer cell lines, and displayed abnormal expression within most malignant tumours, which closely corresponded to overall survival. More importantly, the TFAP4 level was also significantly related to the degree of tumour infiltration. TFAP4 was correlated using gene markers in tumour‐infiltrating immune cells and immune scores. TFAP4 expression was correlated with tumour mutation burden and microsatellite instability in different cancer types, and enrichment analyses identified TFAP4‐associated terms and pathways. The present study comprehensively analysed the expression of TFAP4 across 33 distinct types of cancers, which revealed that TFAP4 may possibly play a vital role during cancer formation and development. TFAP4 is related to differing degrees of immune infiltration within cancers, which suggests the potential of TFAP4 as an immunotherapy target in cancers. Our study demonstrated that TFAP4 plays an important role in tumorigenesis as a prognostic biomarker, which highlights the possibility of developing new targeted treatments.

human malignancies. 2 Pan-cancer analysis is the analysis of the molecular abnormalities of various types of cancer, which can identify any common features and heterogeneities during vital biological processes that are under dysregulation as the result of diverse cancer cell lineages. Pan-cancer analysis projects, such as the Cancer Cell Line Encyclopedia (CCLE) and The Cancer Genome Atlas (TCGA), have been created based on the assessment of different human cancer cell lines and tissues at epigenomic, genomic, proteomic and transcriptomic levels. [3][4][5] Recently, pan-cancer analysis has been used to identify certain functional and pathway genes, which allows for a comprehensive and thorough understanding of human cancers.
For example, tumour hypoxia-associated multiomic molecular characteristics have been investigated, and it has been suggested that some molecular alterations can be correlated with drug sensitivity or resistance to antitumour agents. This helps to comprehensively understand tumour hypoxia at the molecular level and has certain implications for cancer treatment in clinical practice. 6 New data on FOXM1 up-regulation frequency, aetiology and outcomes in human cancers have been defined from 33 TCGA-derived cancers. 7 The information obtained from these cancers has revealed lncRNA-mediated dysregulation within the cancer at a system level, and provides a valuable approach and resources to investigate lncRNA functions in the context of cancer. 8 Characterizing immune phenotype occurrence frequency and variability in a variety of types of cancer helps to understand the immune status of untreated cancers, and this approach has been used in more than 9000 TCGA-derived cancer gene expression data sets. 9 Therefore, pan-cancer analysis can illustrate patterns beneficial for developing combination and individualized therapies for the treatment of various cancers.
Transcription factor activating enhancer binding protein 4 (TFAP4) is involved in cancer proliferation, metastasis, differentiation, angiogenesis and other biological functions. 10 In recent years, it has been suggested that the overexpression of TFAP4 may indicate a poor prognosis for various cancers, including hepatocellular carcinoma (HCC), non-small cell lung carcinoma (NSCLC), prostate cancer (PCa), colorectal cancer (CRC) and gastric cancer (GC). [11][12][13][14][15] According to our prior research, TFAP4 plays a role as an efficient prognostic biomarker, which also activates the PI3K/AKT signal transduction pathway to enhance the metastasis and invasion of HCC. 16 Other studies have been carried out to examine the proliferation, overexpression or mutation of TFAP4 in specific types of cancer, but those studies had low sample sizes and diverse methods. Additionally, research on TFAP4 has mainly focused on an individual or limited number of types of cancers, and no available studies have comprehensively examined several types of cancers simultaneously to identify their similarities and differences. This information is of great importance for understanding the roles of TFAP4 in various cancers, so a comprehensive analysis is urgently needed.
To that end, and taking advantage of the large data sets from TCGA, the present study aimed to examine TFAP4 expression profiles and their prognostic significance among human cancers. Additionally, the associations between TFAP4 and the levels of tumour infiltration, tumour mutational burden (TMB) and microsatellite instability (MSI) were analysed for different types of tumour using correlation analysis.
Gene set enrichment analysis (GSEA) was conducted to investigate any possible underlying mechanisms. The results of the present study can help to understand vital parts of TFAP4 in the context of tumours, reveal the possible association of TFAP4 with tumour-immune interactions and illustrate the potential mechanism.

| Patient data sets and processing
TCGA, a cornerstone of the cancer genomics projects, had characterized more than 20,000 primary cancer samples and corresponding non-carcinoma samples from 33 types of cancers. In the present study, the TCGA-processed level 3 RNA-sequencing data sets, along with the corresponding clinical annotations, were obtained using the University of California Santa Cruz (UCSC) cancer genome browser (https://tcga.xenah ubs.net, accessed April 2020). The CCLE public project has comprehensively characterized a tremendous number of human tumour models both genetically and pharmacologically (https://porta ls.broad insti tute.org/ccle). To examine differential gene expression in cancers on a larger scale, the CCLE database, which contains RNA-sequencing data sets for over 1,000 cell lines, was used. For this research, only open-access data were used, which precluded the requirement of approval of the Ethics Committee.

| Screening of TFAP4 differential expression and its survival-associated cancers
To compare gene expression levels between the cancerous and adjacent normal samples, data regarding TFAP4 gene expression were extracted from the 33 TCGA cancer types to form an expression matrix, as shown in Table S1. Thereafter, the expression matrix and clinical information were matched by patient ID. Afterwards, a univariate Cox model was used to calculate any association between gene expression levels and patient survival, where a difference of P < .05 for TFAP4 in a specific cancer was deemed statistically significant. The survival-associated forest plot was further drawn, and a Kaplan-Meier (KM) analysis was conducted to compare the overall survival (OS) for TCGA cancer patients stratified according to the median TFAP4 expression level, using the log-rank test.

| TFAP4 and tumour immunity
The Tumour Immune Estimation Resource (TIMER, https://cistr ome. shiny apps.io/timer/) represents the integrated approach to systemically analysing the immune infiltrates of different types of cancers. 17 In TIMER, the deconvolution statistical approach is used for inferring tumour-infiltrating immunocyte levels based on gene expression data. 18 Using the TIMER algorithm, we examined the associations between TFAP4 levels and six different immune infiltrate levels   (Table S2). For the present study, the association of TFAP4 expression with each leucocyte phenotype across 33 cancer types was computed.
Additionally, we examined the associations of TFAP4 levels with tumour-infiltrating immunocyte gene markers selected based on previous research. [20][21][22] The correlation analysis generated the estimated statistical significance and Spearman's correlation coefficient.
Then, an expression heat map was plotted for gene pair within the specific type of cancer.
The estimation of stromal and immune cells in malignant tumour tissues using expression data (refer to ESTIMATE for short) represents an approach that uses gene expression profiles to predict the purity of both tumours and the infiltrating stromal cells/immunocytes within tumour tissues. 23 The ESTIMATE algorithm produces three scores on the basis of single sample Gene Set Enrichment Analysis (ssGSEA), including 1) stromal score, which determines stromal cells within the tumour tissues, 2) immune score, which assesses immunocyte infiltration within the tumour tissues, and 3) estimate score, which can infer the purity of tumour. In this study, we used the ESTIMATE algorithm to estimate both immune and stromal scores (Table S3) for tumour tissues according to the corresponding transcription data. Then, we calculated the correlations between these scores and TFAP4 expression.
TMB measures the mutation number in a specific cancer genome. Numerous studies have explored the significance of using TMB as a biomarker for predicting which patients would be most responsive to checkpoint inhibitors. 24 We downloaded the somatic mutation data for all TCGA patients (https://tcga.xenah ubs.net),

F I G U R E 2
The box plot shows the association of TFAP4 expression with pathological stages for 21 types of cancers calculated their TMB scores (Table S4) and then determined the correlation between TMB and TFAP4. MSI is characterized by the widespread length polymorphisms of microsatellite sequences due to DNA polymerase slippage. Recently, it has been suggested that patients with high-MSI cancers gain benefits from immunotherapy, and MSI has been utilized as an indicator of genetic instability for the cancer detection index. 25 We computed the MSI score for each patient, as shown in Table S5, and subsequently performed a correlation analysis between MSI and TFAP4.

| Pan-cancer expression landscape of TFAP4
According to CCLE analysis results, TFAP4 displayed inconsistent gene expression levels among various cancer cell lines (P = 1.3e-11,

| TFAP4 level was related to the level of immune infiltration
Tumour-infiltrating lymphocytes (TILs) can serve as independent predictors of sentinel lymph node status and cancer survival. As a result, the present study examined the correlation between TFAP4  Figure 7A, and pan-cancer associations of TFAP4 levels with the levels of immune infiltration are presented in Figure S1 and Table S6.
Using CIBERSORT, detailed immunocyte compositions of all TCGA patients were calculated, after which the correlations between 22 immunocytes and TFAP4 expression were determined for 33 types of cancer, as seen in Table S7. We found that many immunocytes were significantly correlated with TFAP4 levels. As seen in Figure 7B, in CHOL, OV, UCS and UVM, only one type of immunocyte was correlated with TFAP4 level, while at least two immunocytes were correlated with TFAP4 levels in other cancers.

| Correlations of TFAP4 level with immune markers
To investigate the association of TFAP4 expression with different immune infiltrating cells, the relationships between TFAP4 expression might regulate the immune response in these cancer types.

| Correlation analysis with ESTIMATE score, TMB and MSI
The ESTIMATE method was developed to calculate the immune and stromal scores of cancer tissues. Using the ESTIMATE method, we calculated the immune, stromal and estimate scores, after which we evaluated the relationship between immune/stromal scores and TFAP4 expression. Figure 7D shows the typical results for HCC, in which TFAP4 expression is significantly correlated with both stromal and estimate scores. The detailed correlation results are summarized in Table 2.
Moreover, the association between TMB/MSI and TFAP4 expression was also evaluated, as seen in Table 3. We found that TFAP4 ex-

| Functional analysis
The biological effect of TFAP4 expression was assessed using GSEA. In HCC, TFAP4 showed significant enrichment in the following GO terms: METABOLISM. These can be seen in Figure 8C and 8D, respectively.
The pan-cancer functional GO and KEGG lists of TFAP4 are available in Tables S9 and S10. ysis. TFAP4 was also found to be correlated with TIL gene markers, as seen in Figure 7C. ESTIMATE was reported as a metric for evaluating cancer patient prognosis. 29 Recently, numerous studies have used the ESTIMATE method to assess various tumours, and it has been successfully applied to genomic data. For instance, ESTIMATE is used to predict prognoses in glioblastoma and cutaneous melanoma patients. 30,31 Using the TCGA cohort, the ESTIMATE approach was utilized to generate immune and stromal scores. We found that TFAP4 was negatively correlated with the ESTIMATE scores.

| D ISCUSS I ON
Gene mutations are the primary cause of cancer formation. 32 Specific gene mutations may predict patient prognosis and treatment response. 33 However, further studies are required to determine whether TFAP4 can serve as a predictor for the efficacy of immunotherapy in these types of cancers. Taken together, the findings of the present study provide clues for the association between TFAP4 and cancer immunity.
Collectively, our comprehensive pan-cancer analysis has illustrated the characterization of TFAP4 within cancer cell lines and tissues. Moreover, we have found that TFAP4 can serve as a valuable prognostic biomarker for some types of cancer. Based on the results of the present study, the TFAP4 level is related to cancer immunity.
Moreover, our new integrative omics-based workflow may be adopted to generate hypotheses about novel targets for cancers.

CO N FLI C T O F I NTE R E S T
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Publicly available data sets were analysed in this study. These data can be found at https://tcga.xenah ubs.net and https://porta ls.broad insti tute.org/ccle.