We first used the online TIMER2 (http://timer.cistrome.org/) webserver tool to explore the differential BATF expression between 33 TCGA tumor types and corresponding adjacent normal tissues (Li et al., 2020). Next, we validated the expression profile of BATF between the tumor tissues and corresponding normal tissues based on tumor data from the TCGA database and Genotype-tissue Expression (GTEx) database, respectively. The “Pathological Stage Plot” module of GEPIA2 tool (http://gepia2.cancer-pku.cn/#about) was used to obtain plots of the BATF expression in various pathological stages of TCGA tumors (Tang et al., 2019). When p < 0.05, it was considered to be statistically significant.

The “Gene_Outcome” module of TIMER2 was applied to assess the clinical relevance of BATF expression across 33 type tumors. A Cox proportional hazards model was conducted to explore the outcome significance of BATF expression and a heatmap was drawn to show the normalized coefficient of BATF in the Cox model. According to median expression levels of BATF, each type of tumors was divided into high-expression and low-expression groups. After that, survival analysis of patients with tumors was conducted. A Kaplan–Meier plotter curve of BATF was created to evaluate the relationship between overall survival and BATF expression in TCGA tumors (Li et al., 2020). When p < 0.05, it was considered to be statistically significant.

We investigated the methylation of BATF in the 33 type tumors and corresponding adjacent normal tissues using the UALCAN portal (http://ualcan.path.uab.edu/). UALCAN is a comprehensive, user-friendly, and interactive web resource for analyzing cancer omics data (Chandrashekar et al., 2017). We applied the University of California, Santa Cruz’s Xena (https://xenabrowser.net/heatmap/) tool to analyze the relationship between survival time and BATF methylation levels in the corresponding tumors. The Xena is a platform, which provides huge amounts of processed cancer omics data from large cancer research projects (e.g., TCGA, CCLE, and PCAWG) (Wang et al., 2021). When p < 0.05, it was considered to be statistically significant.

We used the cBioPortal portal (https://www.cbioportal.org/) to investigate genetic alterations of the BATF gene in the pan-cancer cohort. We identified genetic alterations, such as structural variants, mutation type, and copy number alterations, in the 33 TCGA tumor types via the “Cancer Types Summary” module. We also applied the “Comparison” module to gain insights into the effects of BATF genetic alterations on the overall, progression-free, disease-specific, and disease-free survival rates. When p < 0.05, it was considered to be statistically significant.

We applied the “Similar Gene Detection” module of the GEPIA2 to obtain the top-100 BATF-correlated targeting genes across all the TCGA tumor samples (Tang et al., 2019). Using these 100 correlated genes, we performed GO enrichment and KEGG analysis using the R software program (R Foundation for Statistical Computing, Vienna, Austria). A p-value < 0.05 was considered to be statistically significant.

Data on the 33 tumor patients in the TCGA database, tumor RNA-seq data (TCGA) and BATF data can be downloaded from the Genomic Data Commons data (https://portal.gdc.cancer.gov/) portal website. Each tumor had messenger RNA expression data available from matched normal tissue samples. For reliable immune score evaluation, we used immuneeconv, an R software package that integrates six newer algorithms, including TIMER, xCell, MCP-counter, CIBERSORT, EPIC, and quanTIseq (Li et al., 2016; Aran et al., 2017; Sturm et al., 2019). When p < 0.05 and R-values between −1 and one were obtained, viz. the Spearman’s rank correlation test, and were considered to be statistically significant.

We applied the “Drug” module of the Gene Set Cancer Analysis (GSCA) to obtain the correlation between BATF expression and drug sensitivity (IC₅₀) of updated new drugs. GSCA integrates more than 10,000 multi-dimensional genomic data across 33 cancer types from TCGA and more than 750 small molecule drugs from Genomics of Drug Sensitivity in Cancer (GDSC) and Cancer Therapeutics Response Portal (CTRP) (Liu et al., 2018).

Glioma U251 and gastric cancer SGC7901 cells were purchased from ATCC. Cells were cultured with DMEM medium (Gibco) containing 10% fetal bovine serum (Gibco) and 1% penicillin and streptomycin (Sigma). These cells were cultivated in cell incubator with 5% CO₂ at 37°C. 2000 cells/well were seeded in 96-well plates and cultured for 24 h before transfection. After that, BATF siRNA were transfected with Lip3000 (Thermo Fisher Scientific) according to the manufacturer’s instructions.

Trizol reagent (Invitrogen) was applied to extract total RNA from tumor cells. Reverse transcription and Real-time PCR was conducted by using a PrimeScript RT reagent kit (Takara) and SYBR Premix Ex Taq II (Takara), respectively. The primers of BATF was were synthesized from Sangon Biotech (Forward: 5′-CCC​TGG​CAA​ACA​GGA​CTC​AT-3′; Reverse: 5′- GAT​CTC​CTT​GCG​TAG​AGC​CG-3′). The relative expression levels were normalized by using the 2^−ΔΔCt method.

48 h after transfection, cell culture medium were replaced with fresh medium and cultured for indicated times. 10 μl of CCK-8 solution was added to each well and incubated for 1 h at the same incubator conditions. After that, the absorbance at 450 nm was detected with a micro-plate analyzer.

SPSS software (ver. 21.0; SPSS, Chicago, IL) were applied for statistical analyses and the significance was determined by one-way ANOVA or t-test. A p-value < 0.05 was considered to be statistically significant.

First, we explored expression levels of BATF in normal tissues and single cells. As shown in Supplementary Figure S1A, BATF expression in areas of the central nervous system, such as the cerebral cortex and cerebellum, was low, while it was more highly expressed in organs of the immune system, like the spleen. In most cells, such as neuronal cells and glial cells, BATF is retained at very low expression levels. In contrast, immune cells—for instance, B-cells, T-cells, and granulocytes—present high expression levels, which may explain the roles of BATF in normal immune cells (Supplementary Figure S1B). Next, we compared the expression levels of BATF in 33 tumor types and adjacent normal tissues. As shown in Figure 1A, we found that BATF was upregulated in bladder urothelial carcinoma (BLCA), lung adenocarcinoma (LUAD), stomach adenocarcinoma (STAD), esophageal carcinoma (ESCA), breast invasive carcinoma (BRCA), cholangiocarcinoma, kidney renal papillary cell carcinoma (KIRP), glioblastoma multiforme, prostate adenocarcinoma (PRAD), kidney renal clear cell carcinoma (KIRC), colon adenocarcinoma (COAD), head and neck squamous cell carcinoma (HNSC), lung squamous cell carcinoma (LUSC), liver hepatocellular carcinoma (LIHC), rectum adenocarcinoma (READ), and uterine corpus endometrial carcinoma (UCEC). Compared with the corresponding normal tissues, BATF was abnormally expressed in 27 TCGA tumor types, including cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), ower-grade glioma (LGG), adrenocortical carcinoma (ACC), thymoma, UCEC, l pancreatic adenocarcinoma (PAAD), cutaneous melanoma (SKCM), uterine carcinosarcoma (UCS), lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), kidney chromophobe, ovarian serous cystadenocarcinoma (OV), STAD, testicular germ cell tumors (TGCT), and thyroid carcinoma (THCA). We also explored mRNA expression levels of BATF in cancer cell lines of different tumors via the CCLE database (https://portals.broadinstitute.org/ccle/home). The CCLE database is a public database that contains expression data of more than 30 types of tumors (including glioma and gastric cancer), 1457 tumor cell lines. Similarly, BATF was also highly expressed in several tumor cells, and BATF protein was dominantly distributed in the nucleus of these cells (Supplementary Figure S2). Interestingly, we found that BATF expression was only correlated with BLCA and KIRC clinicopathological stages; in other tumors, BATF expression levels showed no correlation with tumor stages (Figure 2). These results indicated that BATF was indeed upregulated in most tumors, and the forced BATF expression was unlikely associated with tumor progression, except for BLCA and KIRC.

The prognostic role of BATF in tumors is presented in Figure 3. The Cox regression result showed that BATF was a high-risk indicator in LAML, ACC, LGG, COAD, LIHC, KIRC, and UVM. On the contrary, upregulated BATF was a favorable factor in HNSC, CESC, BRCA, SARC, BLCA, and UCEC (Figure 3A and Table 1). We next divided the tumor species into high- and low-expression groups according to median expression levels of BATF to evaluate the correlation between BATF expression and survival times of patients with tumors. We found that patients with KIRC, LGG, LAML, ACC, and UVM and higher BATF expression levels had poorer overall survival. In contrast, a higher expression of BATF was associated with a significantly better prognosis in HNSC, BLCA, BRCA, CESC, and UCEC. The results of Kaplan–Meier plotter analysis were in line with the results of Cox regression analysis (Figures 3B–K). These results indicated that the prognostic role of BATF in tumors is tissue-dependent.

DNA methylation is the most common way to regulate gene transcription, and we speculated that the dysregulated expression of BATF may be modulated in this manner. As shown in Figure 4, patients with BLCA, READ, LUAD, CESC, KIRC, PRAD, LUSC, PCPG, KIRP, TGCT, COAD, THCA, HNSC, and UCEC demonstrated lower methylation levels with respect to corresponding normal tissues, which may explain the higher BATF expression levels in these tumors. Notice that, although BATF was upregulated in PRAD samples, methylation levels of PRAD are higher than those in normal tissues (Figure 4J). Apart from DNA methylation, post-transcriptional modification like N6-methyladenosine (m6A) is a vital way to affect gene expression (Jiang et al., 2021). Abnormal expression of BATF in PRAD may be controlled in m6A-dependent manner. In ACC, BLCA, BRCA, CESC, HNSC, and UCEC, patients with higher levels of methylation of the BATF gene had shorter survival times. In contrast, lower methylation levels of the BATF gene expression predicted poorer overall survival in LGG and UVM (Supplementary Figure S3). These results suggest that increased BATF expression in tumors is probably due to the low methylation of its promoter in cancers other than PRAD. In addition, methylation levels predicting overall survival in tumor patients are also tissue-dependent.

Gene alteration plays crucial roles in tumor development and progression. Hence, we investigated genetic alterations of BATF in a pan-cancer cohort containing 10,969 tumor samples. The top three genetic alterations with rates over 2% were SKCM, DLBL, and UCEC, respectively. Among the types of genetic alteration, mutation, amplification, and deep deletion account for almost all types in the majority of tumors; structural variations were only found in ACC (Figure 5A). There were 34 missense and one splice mutations, and R37c was the most common mutated site (Figure 5B). We found that there was no difference in overall survival (OS) rates between BATF-altered and -unaltered groups across different cancers (Figure 5C). It seems that genetic alterations of BATF play a limited role in carcinomas.

We obtained the top 100 genes similar to BATF across 33 tumor types from the GEPIA2 “Similar genes” module and performed GO enrichment analysis to gain insights into the roles of BATF in tumors. As shown in Figure 6A, BATF expression was mainly involved in regulating leukocyte cell–cell adhesion, leukocyte proliferation, lymphocyte proliferation, mononuclear cell proliferation, and lymphocyte activation. To validate this result, we performed an immune-cell infiltration analysis. As shown in Figure 6B, there is a positive correlation between BATF expression and infiltration levels of most immune cells. However, BATF expression was negatively correlated with common lymphoid progenitor infiltration in most tumors. In DLBC, BATF expression was only correlated with several immune cell types, such as natural killer T-cells, CD4+ Th1 cells, hematopoietic stem cells, class-switched memory B-cells, and naive B-cells. These results demonstrate that BATF is involved in modulating immune infiltration in tumors.

To gain insights into the potential mechanisms of BATF in cancers, we performed KEGG analysis. As shown in Supplementary Figure S4, BATF was positively correlated with the natural killer cell–mediated cytotoxicity, B-cell receptor signaling pathway, programmed death ligand one expression and the programmed death 1 (PD-1) checkpoint pathway in cancer, antigen processing and presentation, T-cell receptor signaling pathway, cytokine–cytokine receptor interaction, Th1 and Th2 cell differentiation, and cell adhesion. To validate these findings, we also explored correlations between BATF expression and immunomodulators and chemokines. As shown in Figures 7, 8, BATF was positively associated with most immunostimulators, immunoinhibitors, chemokines, and related receptors. In addition, BATF expression was also highly correlated with almost all MHC molecule genes (Supplementary Figure S5A). These results indicate that BATF might regulate immune cell infiltration via an immune-related pathway.

We have previously contended that BATF is a transcriptional factor, and we then questioned how BATF performs its roles in tumors. The STRING tool was applied to draw a protein–protein interaction network for BATF, setting a minimum required interaction score [“highest confidence (0.900)”] for obtaining BATF-binding proteins (Figure 9A). We also obtained an interactive functional association network for BATF via the GeneMANIA tool (Figure 9B). There were seven genes (BATF3, IRF4, JUNB, FOS, FOSB, FOSL1, and FOSL2) that overlapped (Figure 9C). Among these genes, only BATF3 and IRF4 were highly co-expressed with BATF in the majority of cancers, which suggested that BATF might carry out its role by cooperating with BATF3 and IRF4 (Figure 9D).

After knowing the roles of BATF in immune infiltration and immune checkpoint pathway in cancer, we wondered whether there is a correlation between BATF expression and immunotherapeutic response. As shown in Supplementary Figure S5B, BATF expression was highly associated with most immune checkpoint molecules in the majority of tumors. A low correlation between BATF and immune checkpoint molecules was observed in PAAD, LAML, and DLBC. Tumor mutation burden (TMB) and microsatellite instability (MSI) are vital immune-response biomarkers (Choucair et al., 2020; Kok et al., 2019). We found that BATF was positively associated with TMB in ESCA, CESC, UCS, PAAD, SARC, LGG, COAD, UCEC, OV, KIRC, and DLBC, but was negatively correlated with other tumors (Supplementary Figure S6A). In addition, a positive association between BATF expression and MSI was found in COAD, DLBC, READ, SKCM, BRCA, UVM, ESCA, PAAD, LAML, LUAD, LIHC, and KIRC as well as a negative association in other tumors (Supplementary Figure S6B). In five independent clinical cohorts, patients with melanoma, urothelial cancer, or clear cell renal cell carcinoma received anti–PD-1 therapy (Hugo et al., 2016; Riaz et al., 2017; Snyder et al., 2017; Mariathasan et al., 2018; Miao et al., 2018). Nevertheless, we did not found there is evident difference in BATF expression between the responder and non-responder groups (Table 2).

Although immunotherapy is a promising option for tumor treatment, chemotherapy still occupies an important position. Patients with tumors may benefit from targeting BATF with chemotherapy drugs. Hence, we investigated the correlation between BATF expression and chemotherapy sensitivity. As shown in Figure 10, there was a negative correlation between CTRP or GDSC drug sensitivity and BATF expression. In other words, patients with higher BATF expression show a lower chemotherapy response. The high expression of BATF may partially account for drug resistance in tumor chemotherapy.

We chose two malignant cancers glioma and gastric cancer for further analysis and experiments. Firstly, we also performed a GO and KEGG analysis of BATF in these two types of tumors. We found that BATF was not only associated with immune-cell infiltration, but also associated with cell activation in both gliomas and gastric cancer (Supplementary Figures S7, S8). Previous studies reported that BATF was crucial for T cell proliferation and exhaustion; we wondered whether BATF could also regulate tumor cell viability. Glioma U251 and gastric cancer SGC7901 cell lines were used for in vitro assay. As shown in Figures 11A,B, tumor cells transfected with BATF siRNA significantly downregulated BATF mRNA expression. We found that both U251 and SGC7901 cell viability were attenuated following BATF knockdown (Figures 11C,D). These results indicated that BATF not only regulated immune-cell infiltration, but also modulated tumor cell proliferation.