The TCGA database (http://gdc-portal.nci.nih.gov/) and GTEx database (https://gtexportal.org/) were used to investigate the expression of AJUBA in pan-cancers and normal samples. Among them, 33 cancers were comprised in the TCGA and GTEx databases. In addition to, we also analyzed the expression of AJUBA in colon adenocarcinoma (COAD), ESCA, LIHC, and stomach adenocarcinoma (STAD) with violin plots (abbreviation referring to Supplementary Table S1). Herein, the results of gene expression analysis were further validated by the use of TIMER (https://cistrome.shinyapps.io/timer/). Because there is no normal sample in certain tumors, the data were supplemented by TCGA database and GTEx databases with a dot plot. Moreover, we analyzed the AJUBA expression at different clinical and pathological stages (stages I, II, III, and IV) in diverse cancers through the “Stage Plot of Expression Analysis” module from GEPIA (http://gepia2.cancer-pku.cn/). In our study, gene expression levels were shown using a log2 (TPM + 1) scale, where TPM stands for transcripts per million. Meanwhile, we input AJUBA in the “Clinical Proteomic Tumor Analysis Consortium (CPTAC),” and phosphoprotein analysis was performed using the UALCAN database (http://ualcan.path.uab.edu/). The protein expression for breast cancer, ovarian cancer, colon cancer, lung adenocarcinoma (LUAD), clear cell renal cell carcinoma (RCC), and uterine corpus endometrial carcinoma (UCEC) is available with phosphorylation at the S79, S119, S137, S230, Y231, S237, S263, and Y267 sites.

The HPA (https://www.proteinatlas.org/) contains an atlas of human protein expression models in normal and tumor cells, tissues, and organs. We used the HPA to recognize the pattern of special protein expression that are differentially expressed in specific tumors. Herein, the expression pattern of AJUBA in normal and tumor tissues [colon and COAD, breast and breast invasive carcinoma (BRCA), lung and LUAD, liver and LIHC] was accomplished by using immunohistochemistry images (Supplementary Table S2).

The prognostic survival value of AJUBA in different tumors [lower-grade glioma (LGG), LIHC, sarcoma (SARC), kidney renal clear cell carcinoma (KIRC), skin cutaneous melanoma (SKCM), and adrenocortical carcinoma (ACC)] was analyzed using GEPIA2 (http://gepia.cancer-pku.cn/). In GEPIA2, TCGA tumor patients were divided into the high- and low-expression cohorts based on the high and low cut-off values (50% and 50%), respectively, to obtain the overall survival (OS) and disease-free survival (DFS) analyzed using the log-rank test.

The receiver operating characteristic (ROC) curves were prepared depending on TCGA database (https://portal.gdc.cancer.gov/) to assess the diagnostic performance of AJUBA. From the ROC curves, the area under the ROC curve (AUC) values were calculated using the vertical coordinate of false-positive rate and the horizontal coordinate of true-positive rate among different tumors [PRAD, colon and rectal adenocarcinomas (COADREAD), STAD, cholangiocarcinoma (CHOL), COAD, ESCA, HNSC, kidney chromophobe (KICH), rectum adenocarcinoma (READ), glioblastoma multiforme (GBM), acute myeloid leukemia (LAML), pancreatic adenocarcinoma (PAAD), SKCM, thymoma (THYM), oral squamous cell carcinoma (OSCC), and LGG).

cBioPortal (http://www.cbioportal.org/) is a powerful website for gene analysis, providing visualization tools for research and analysis of cancer genetic data (Gao et al., 2013). We input AJUBA in the “Quick Search Beta!” module to obtain the characteristic mutations of AJUBA in TCGA tumors. The outcomes of alteration frequency, type of mutation, and copy number alteration of the AJUBA gene were processed using the “Cancer Types Summary” in the cBioPortal database. Meanwhile, the information of mutation sites of AJUBA can be manifested using the protein structure graph or the 3D structural maps through the “Mutations” section. Furthermore, AJUBA gene alteration was associated with clinical prognosis [progression-free survival (PFS), disease-specific survival (DSS), DFS, and OS] in UCEC, which were obtained in the “Comparison” module. The survival curves were shown using the log-rank p-value.

TIMER is an online open website resource to evaluate the correlation of AJUBA expression with immune infiltration level across diverse TCGA cancers (http://timer.cistrome.org/). In the “Immune-Gene” module, the results of immune infiltration were calculated automatically using the TIMER, EPIC, MCPCOUNTER, XCELL, and TIDE algorithms. The purity-adjusted Spearman’s rho across several cancer types was displayed using the heatmap with numbers. In addition, the scatterplot was popped out to present the relationship between immune infiltration in different cancers of the TCGA database via clicking the heatmap in the “Gene-Correlation” module of TIMER2. The correlations between AJUBA expression and immune infiltration were calculated using the purity-adjusted partial Spearman’s rank correlation test.

The gene expression data of LIHC, ESCA, and PAAD in HTSeq-FPKM were downloaded from the TCGA database (https://portal.gdc.cancer.gov/) for analysis. In this study, the number of samples in LIHC, ESCA, and PAAD was 424, 173, and 182, respectively. In addition, the co-expressed genes of AJUBA in LIHC, ESCA, and PAAD were screened using Pearson correlation coefficients (r > 0.4 or r < −0.3 and p < 0.001). Meanwhile, we performed an intersection analysis using a Venn diagram to estimate the genes that interact and bind with AJUBA. Gene ontology (GO) and Kyoto encyclopedia of genes and genome (KEGG) enrichment analyses were achieved using co-expressed genes to investigate probable biological functions and signaling pathways affected by AJUBA in LIHC, ESCA, and PAAD. Gene set enrichment analysis (GSEA) is a knowledge-based method for interpreting genome-wide expression profiles (Subramanian et al., 2005). GSEA was performed via MSigDB collections (https://www.gsea-msigdb.org/gsea/ msigdb/index.jsp). The enrichment analysis of AJUBA was visualized using the “clusterProfiler” and “ggplot2” R packages (Yu et al., 2012). Statically significant difference was regarded when p-value < 0.05.

Data are presented as means ± standard error (SD). Differences between groups were analyzed using a Student’s t-test. Genes correlation analysis was performed using Spearman correlation coefficients. Statistical analyses were performed using R 3.6.3. p < 0.05 (two-tailed) was considered statistically significant (*p < 0.05, **p < 0.01, ***p < 0.001).

Initially, AJUBA expression levels in pan-cancer were examined in the TCGA and GTEx databases. The analysis revealed that AJUBA expression was significantly higher in various tumors than in the respective normal tissues, including bladder urothelial carcinoma (BLCA), cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), CHOL, COAD, lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), ESCA, GBM, HNSC, brain LGG, LIHC, lung squamous cell carcinoma (LUSC), PAAD, READ, and STAD as well as thyroid carcinoma and THYM. In contrast, AJUBA expression was low in KICH, LAML, LUAD, pheochromocytoma and paraganglioma (PCPG), PRAD, SKCM, testicular germ cell tumors (TGCT), UCEC and uterine carcinosarcoma (UCS) (p < 0.05) (Figure 1A). In a similar manner, we also assessed expression among various cancer types using the TIMER2 database. As shown in Figure 1B, AJUBA expressions in human tumors of BLCA, CESC, CHOL, COAD, ESCA, GBM, HNSC, LIHC, LUSC, READ, and STAD were dramatically higher than expressions in paired normal groups (p < 0.05). In addition, as we know that the increased methylation level leads to the decreased expression of target gene, the AJUBA promoter methylation level was analyzed using the UALCAN database. The results demonstrated that the methylation levels of the AJUBA promoter in most tumors were significantly compared with the normal tissues (Supplementary Figure S1).

Specifically, we further analyzed AJUBA expression in digestive system tumors (COAD, ESCA, LIHC, and STAD) (p < 0.001) (Figure 1C). In addition, as shown in Figure 1D, the normal tissue of the GTEx dataset was provided as supplementary to further evaluate the expression difference of AJUBA between the normal groups and cancer groups in DLBC, LGG, THYM, and UCS. We also explored the correlation between AJUBA expression level and the pathological stages of tumors, including in KIRC, PAAD, SKCM and UCEC (p < 0.01) using the “Pathological Stage Plot” module of GEPIA2 (Figure 1E).

Meanwhile, the HPA database was used to assess the AJUBA expression levels among different tumors. The results displayed that higher AJUBA expression levels were observed in different tumors (COAD, BRCA, LUAD, and LIHC). In contrast, its expression is low in normal tissue (colon, breast, lung, and liver) (Figure 2). These results demonstrated that AJUBA may play a significant role in various cancers.

Further, to evaluate the value of AJUBA in clinical prediction, the correlation between AJUBA expression and OS and DFS was analyzed via the GEPIA2 database. For OS, higher expression of AJUBA was dramatically associated with decreased OS in LGG (p = 0.00016), LIHC (p = 0.0054), SARC (p = 0.0014), and SKCM (p = 0.042). Higher AJUBA expression was significantly associated with reduced DFS in ACC (p = 0.023) and LGG (p = 2.9e−05). In contrast, it was shown in KIRC that as the expression of AJUBA increases, its OS (p = 0.0062) and DFS (p = 0.0043) also increase (Figure 3).

ROC curve analysis was used to assess the diagnostic value of AJUBA in various cancers patients. Specifically, the closer the AUC value is to 1, the better the prediction performance. As shown in Figure 4, results reveal the high diagnostic value of AJUBA in OADREAD, COAD, KICH, READ, GBM, LAML, SKCM, and THYM (AUC > 0.9). Moreover, the analysis shows that AJUBA has a definite value for diagnosis among PRAD, STAD, CHOL, ESCA, HNSC, and PAAD, as well as OSCC and LGG (0.8 < AUC < 0.9). Moreover, the Nomogram plot were made for the prognosis predicting of cancer patients at 1, 3, and 5 years in different tumors (Supplementary Figure S2). These results demonstrate that AJUBA could be a potential diagnostic marker in pan-cancer.

To investigate the correlation of the AJUBA gene mutation in different cancer types, we analyzed the mutation status of AJUBA through the cBioPortal platform based on TCGA data. AJUBA has the highest mutation frequency in patients with HNSC (6%) in Figure 5A. Strikingly, all patients with CHOL had a structural variant of the AJUBA gene, which displayed an alteration frequency of ∼3%. The 3D structure of the AJUBA protein is presented in Figure 4B. As shown in Figure 4C, we can intuitively understand the types, sites, and case number of the AJUBA genetic alteration, including the missense, truncating, and fusion mutations. In addition to, missense and truncating of AJUBA were the predominant mutation (Figure 5C). Furthermore, we found that R371*/Q alteration in the LIM domain was detected in two cases of CESC and one case of HNSC. In addition to, we analyzed potential links between genetic alteration of AJUBA and the survival prognosis of patients in various cancers. These results indicated that UCEC patients with altered AJUBA displayed better prognosis in PFS (p = 0.038), but not in DFS (p = 0.143), DFS (p = 0.174) and OS (p = 0.523), compared with groups without AJUBA alteration (Figure 5D).

We made use of the CPTAC database to explore the phosphorylation of AJUBA between cancer tissues and normal tissues. Figure 6A displays the sites of AJUBA phosphorylation. Meanwhile, we found that three types of tumors (ovarian cancer, colon cancer, and renal clear cell carcinoma) show a higher phosphorylation level in all primary cancer tissues than do normal tissues at the S263 locus of AJUBA (p < 0.05), followed by an increased phosphorylation level of the S119 locus for breast cancer (p = 7.8e−04) and the S137 locus for ovarian cancer (p = 0.0039) and colon cancer (p = 7.2e−22) (Figure 6B). In contrast, lower protein phosphorylation of AJUBA was detected in LUAD and UCEC at the S263 locus of AJUBA (p < 0.001) (Figure 6B). Therefore, these results reveal at least the importance of the phosphorylation of S263, S119, and S137 in tumor development.

Immune infiltration plays a significant role in the process of various tumor initiation, progression, and metastasis. As an important part of the immune microenvironment, cancer-associated fibroblasts (CAFs) have various functions in the occurrence and development of tumors; herein, they hold promise as potential targets for cancer therapy (Liu et al., 2019; Sahai et al., 2020). Here, we used the TIMER, EPIC, MCPCOUNTER, XCELL, and TIDE algorithms to perform a comprehensive exploration of correlation between the infiltration level of CAFs and AJUBA gene expression in a variety of tumor types of TCGA (Figure 7A). Through a series of analyses, a statistical positive correlation was observed between the immune infiltration of CAFs and AJUBA expression in BRCA, BRCA-LumA, BRCA-LumB, COAD, LGG, LUAD, ovarian serous cystadenocarcinoma (OV), and TGCT based on all or most algorithms. As shown in Figure 7B, the scatterplot data of the above cancers were displayed via only one algorithm (p < 0.001). Furthermore, the correlation between AJUBA and immune checkpoints was also investigated. And the results demonstrated that AJUBA was positively associated with a few immunosuppressive checkpoints in the vast majority of tumors, including PD-L1, PD-L2, PD-1, CTLA4, LAG3, TGFB1/TGFBR1, TIGT, TIM-3, and IL10/IL10RB (Supplementary Figure S3). Above all, these findings strongly suggest that AJUBA promotes tumor development via enhancing the level of CAF immune infiltration and being positively associated with immunosuppressive checkpoints.

To further study the molecular mechanism of the AJUBA during tumor progression, we tried to screen out the AJUBA-related genes via a series of functional enrichment analyses. Initially, in the TCGA database, we screened 2,828 positively correlated genes and 17 negatively correlated genes in LIHC, 1,781 positively correlated genes and 2,017 negatively correlated genes in ESCA, and 3,452 positive and 1,018 negative genes in PAAD through Spearman correlation coefficients (r > 0.4 or r < −0.3 and p < 0.001). Moreover, the top 10 co-expressed genes with positive and negative correlation are presented in the form of a heatmap (Figure 8A). Meanwhile, the Venn diagram revealed that 189 genes were discovered to be co-expressed with AJUBA in LIHC, ESCA, and PAAD (Figure 8C). As demonstrated in Figure 8B, GSEA provides insight into novel reactome pathways, which indicated reactome_extracellular_matrix_organization, reactome_class_ a_1_rhodopsin_like_receptors, and reactome_G_alpha_i_signalling_events were separately enriched in LIHC, ESCA, and PAAD. However, reactome_gpcr_ligand_binding was enriched in LIHC, ESCA, and PAAD (Figure 8B). We used the TCGA database to perform GO and KEGG enrichment analyses. The GO data of Figure 8D indicate that “protein serine/threonine kinase activity,” “cell–cell junction,” and “covalent chromatin modification” might play an important role in tumorigenesis and development. As demonstrated in Figure 8B, the KEGG enrichment analysis data further reveal that a number of these genes exhibit a certain level of correlation with Salmonella infection and the Hippo signaling pathway.