Oncomine (www.oncomine.org), a practical online platform containing 715 datasets and 86,733 samples for expression analyses of target genes (15, 16), was used to compare the expression of ADAM12 mRNA between CRC and matched normal tissues. First, we entered the ADAM12 gene into the search box and obtained the expression profiles of ADAM12 across cancers. Second, “cancer vs. normal analysis” was selected, and the “cancer type” was limited to CRC. Meanwhile, the threshold was set to p=0.05 and fold change=1.5 by default. Finally, statistical significance was obtained directly from the corresponding web analysis outcomes.

UALCAN (http://ualcan.path.uab.edu/index.html), a comprehensive and user-friendly website for the analysis of cancer data (17), was used to examine ADAM12 expression in human cancers and to evaluate the promoter methylation level in COAD samples in TCGA. Additionally, we also analyzed the effects of different clinical characteristics on ADAM12 expression, such as age, sex, weight, stage, race, TP53 mutation status, histological subtype, and nodal metastasis status. Moreover, we supplemented the effect of the status of four DNA mismatch repair proteins (MLH1, PMS2, MSH2, and MSH6) on ADAM12 expression via TIMER2.0 (http://timer.cistrome.org/) (18, 19).

Immunohistochemical staining for the ADAM12 protein in colorectal tissue was obtained from the Human Protein Atlas (https://www.proteinatlas.org/). In addition, its protein model and 3D structure were explored using the SWISS MODEL (https://swissmodel.expasy.org/) (20).

An analysis of the association of specific genes with the survival of patients with COAD was conducted using Gene Expression Profiling Interactive Analysis (GEPIA), PrognoScan, and Kaplan–Meier Plotter. GEPIA (http://gepia.cancer-pku.cn/index.html), a comprehensive and interactive web server (21, 22), provided us with associations between ADAM12 expression and prognosis. PrognoScan (http://www.prognoscan.org/), a new database for meta-analyses of the relation between gene expression and prognosis (23), was used for the survival analysis in our study. The Kaplan–Meier Plotter database (https://kmplot.com/analysis/), an open website designed to evaluate the effect of genes on the survival prognosis of patients with different cancers (24), was applied to further validate our survival analysis results. Patients with COAD were divided into two groups based on the median expression of the specific gene (high vs. low expression) for further prognostic analysis. Hazard ratios (HRs; 95% confidence intervals) and P values were determined for survival, and P < 0.05 served as the significance threshold.

The cBioPortal database (www.cbioportal.org), an open web resource for analysis (25), was used to analyze alterations in the ADAM12 gene. We chose the “Query” module, selected “DFCI, Cell Reports 2016,” “TCGA, Firehose Legacy,” and “TCGA PanCancer Atlas” in the “bowel” section, clicked the “Query by gene” button, and entered the “ADAM12” gene. The results of structural variant data, mutation data, and CNA data are shown in the “Cancer Types Summary” module. The results of the analysis of ADAM12 mutations are shown in a schematic diagram, and its 3D structure was obtained by clicking the “View 3D Structure.” In addition, “Comparison/Survival” was selected to assess the prognostic value of somatic mutation frequency and microsatellite instability.

The associations between the ADAM12 expression level and the abundances and corresponding specific markers of immune cells were investigated using the TIMER2.0 and GEPIA databases. Next, TISIDB (http://cis.hku.hk/TISID B/index.php), a web portal for tumor and immune system interactions (26), was applied to determine the correlation between the ADAM12 expression level and immune components, including lymphocytes, immunomodulators, and chemokines.

STRING (http://string-db.org) was applied to construct the protein–protein interaction network of ADAM12 and explore the interacting proteins (27). The minimum required interaction score was set to a medium confidence of 0.400, and the maximum number of interactors was ≤50. The 50 interactors that were most strongly associated with ADAM12 were collected and set as target ADAM12-binding proteins. Second, the ADAM12 gene was entered into the “Query Search” module of GEPIA, and the top 100 ADAM12-related target genes were collected (genes correlated with ADAM12 expression).

Metascape (https://metascape.org), an effective tool for experimental biologists to analyze big data (28), was applied to further analyze the functional enrichment of related genes. Then, an interactive Venn diagram was created to compare the genes between target ADAM12-binding proteins and genes correlated with ADAM12 expression using Bioinformatics, an online drawing software package (http://www.bioinformatics.com.cn/). Pearson’s correlation analysis was conducted between ADAM12 and the correlated genes using the “Gene_Corr” module of TIMER2.0, and an expression heatmap was created to assess the expression patterns of each gene in different cancer types using bioinformatics. In addition, we evaluated the effect of the expression of the correlated genes on the OS/disease-free survival (DFS) of patients with CRC using GEPIA. If the P value was less than 0.01, it was considered significant.

Differences in ADAM12 expression between tumor and normal tissues from patients with different cancers were analyzed using the Oncomine database. ADAM12 mRNA was overexpressed in tumor tissues, such as breast cancer, cervical cancer, CRC, and esophageal cancer. However, its expression in leukemia and myeloma was low ( Figure 1A ). UALCAN was used to further analyze ADAM12 expression in human cancers. As a result, an upregulation of BLCA, BRCA, CESC, CHOL, COAD, ESCA, GBM, HNSC, KIRC, LUAD, LUSC, PCPG, PCPG, READ, SARC, SKCM, and STAD expression and a reduced expression of PAAD, THYM, and UCEC were observed. Nevertheless, no significant differences were observed between tumor and normal tissues from patients with KICH, LIHC, PRAD, or THCA ( Figure 1B ). The ADAM12 protein model and 3D structure are shown in Figures 1C, D , respectively. The abbreviations for each cancer are summarized in Supplementary Material 1 (S1) .

Moreover, we investigated ADAM12 expression levels in COAD using Oncomine. Compared with the corresponding normal tissue samples, ADAM12 overexpression was observed in colon cancer tissues, including COAD, colon mucinous adenocarcinoma, rectal adenocarcinoma, and cecum adenocarcinoma ( Figures 2A–G ). The promoter methylation level in COAD was also explored in the TCGA analysis module of the UALCAN database. Compared with the normal tissues, the expression and methylation level of ADAM12 in colon adenocarcinoma were significantly upregulated ( Figures 2H, I ). Images of immunohistochemical staining showing ADAM12 protein expression in colorectal tissues are shown in Figure 2J .

In our study, we analyzed the effect of different clinical characteristics on ADAM12 expression using the UALCAN database. In terms of age, the statistical analysis revealed significantly lower ADAM12 expression in the healthy group than in patients with COAD in the 21–40, 41–60, 61–80, and 81–100 years age groups ( Figure 3A ). Regarding the individual cancer stages, significant differences were observed between the healthy group and the stage 1–4 group, and the stage 1 group presented with a lower expression than the stage 2–3 group ( Figure 3B ). Patients with CRC carrying a TP53 mutation tended to express ADAM12 at a higher level ( Figure 3C ). In addition, mucinous adenocarcinoma showed higher ADAM12 expression than adenocarcinoma ( Figure 3D ). ADAM12 expression appeared to be higher in Caucasians than in Asians ( Figure 3E ). Different body weights and sexes did not affect ADAM12 expression in patients with CRC ( Figures 3F, G ), while different nodal metastasis statuses affected ADAM12 expression ( Figure 3H ). Differences in ADAM12 expression based on different clinical indicators are summarized in Table 1 . In addition, high ADAM12 expression was often associated with mutant PMS2 (P=0.033) and MSH6 (P=0.0088) but was unrelated to the mutation status of MLH1 (P=0.3) or MSH2 (P=0.051) ( Figures 3I–L ).

The correlation between ADAM12 expression and prognosis in patients with CRC was subsequently investigated using different databases. Using the GEPIA method, low ADAM12 expression significantly correlated with a longer OS (n = 136, HR = 2.1, p = 0.027) ( Figure 4A ); in contrast, a significant difference was not observed for DFS (n = 136, HR = 1.7, p = 0.12) ( Figure 4B ). Interestingly, ADAM12 was not significantly associated with prolonged OS (n = 165, HR = 0.55, p = 0.15) or DFS (n = 47, HR = 0.33, p = 0.3) ( Figures 4C, D ) as determined using the Kaplan–Meier Plotter.

The PrognoScan database was applied to further evaluate the relationships between ADAM12 expression and prognosis. Unexpectedly, we newly identified some studies (GSE17536-Smith and GSE14333-Jorissen) that showed a correlation between low ADAM12 expression and a shorter OS (n = 177, HR = 0.034, COX-P = 0.042) and DFS (n = 145, HR = 0.41, COX-P = 0.0027; n = 226, HR = 0.27, COX-P = 0.010; n = 226, HR = 0.23, COX-P = 0.017) for patients with COAD ( Figures 4E–H ). In conclusion, based on the aforementioned databases, we discussed the effect of ADAM12 expression on patient prognosis and survival and proposed that ADAM12 is a factor associated with an adverse prognosis in patients with CRC, suggesting that it is a potential target worthy of further study.

Genetic alteration is proven to be an essential factor affecting the occurrence and progression of different cancers. In our paper, we investigated the relationship between the genetic alteration status and ADAM12 expression in patients with CRC in TCGA cohorts. As a result, approximately 3.2% of COAD samples had somatic mutations in ADAM12. Mutation type copy number variation was the primary type identified in CRC. Amplification and deep deletion types were detected, and the proportions of these mutations are shown in Figures 5A, B . Information on the mutation sites, mutation types, and the number of cases is shown in the mutation diagram, which is colored with respect to the corresponding mutation types ( Figure 5C ). In patients with CRC, 61 mutations in the ADAM12 gene were identified, including 8 duplicate mutations in patients with multiple samples. Moreover, we intuitively identified that the R258Q/W alteration was in the reprolysin domain, and the 3D structure of the ADAM12 protein at the R258Q/W site is shown in Figure 5D . Additionally, we further evaluated whether alterations in the ADAM12 gene in CRC were associated with patient survival. However, our results indicated that more genetic alterations were not linked to a poor OS (P = 0.221), DFS (P = 0.355), progression-free survival (P = 0.924), or disease-specific survival (P = 0.231) for patients with CRC ( Figures 5E–H ).

Target ADAM12-binding proteins and genes correlated with ADAM12 expression were identified using the STRING and GEPIA databases to functionally analyze the potential mechanism of ADAM12 in tumor progression. Fifty ADAM12-binding proteins and the top 100 genes correlated with ADAM12 expression were searched. All genes that were screened are summarized in Supplementary Material 2 (S2) , and the interaction network of ADAM12-binding proteins and genes correlated with ADAM12 expression is shown in Figure 6A . Using the interactive Venn diagram to analyze the difference between the two datasets, our results indicated that ADAM12 expression in COAD was strongly correlated with the Periostin (POSTN), Matrix metallopeptidase 14 (MMP14), Integrin subunit beta 1 (ITGB1), TIMP metallopeptidase inhibitor 2 (TIMP2), and Matrix metallopeptidase 2 (MMP2) genes ( Figure 6B ). According to the Human Protein Atlas database, the main functions of these interacting genes are tissue development and regeneration, embryo development, tissue repair, and angiogenesis, among others. In addition, the expression heatmap also showed that the ADAM12 gene was positively correlated with the five interacting genes listed above in 30 different cancers ( Figure 6C ).

We evaluated the relationship between the expression of these interacting genes and the target gene ADAM12 in COAD using the TIMER database and OS and PFS with the GEPIA database to obtain a deeper understanding of the relationships and their effects on prognosis. On the one hand, the analysis results showed that ADAM12 expression was closely associated with the expression levels of POSTN (Cor=0.913, P=3.52e-180), MMP14 (Cor=0.805, P=1.40e-105), MMP2 (Cor=0.876, P=7.5e-147), ITGB1 (Cor=0.572, P=3.32e-41), and TIMP2 (Cor=0.866, P=1.30e-139) ( Figures 6D–H ). On the other hand, high POSTN expression was significantly associated with shorter OS (n=270, HR=2, P=0.0064) and DFS (n=270, HR=1.7, P=0.036). Regarding the MMP14, MMP2, ITGB1, and TIMP2 genes, higher expression of these genes led to shorter DFS (n=270, HR=1.9, P=0.012; n=270, HR=1.8, P=0.017; n=270, HR=1.8, P=0.019; n=270, HR=1.8, P=0.018); in contrast, no significant difference was observed in OS (n=270, HR=1.4, P=0.13; n=270, HR=1.4, P=0.15; n=270, HR=1.1, P=0.28; n=270, HR=1.4, P=0.19) ( Figure 7 ).

We combined the ADAM12-binding proteins and the top 100 genes correlated with ADAM12 expression to conduct Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses in Metascape and further predicted the functions and pathways in which ADAM12 was enriched. The GO enrichment analysis data indicated that these related genes were mainly enriched in extracellular matrix (ECM) organization, NABA CORE MATRISOME, ECM proteoglycans, NANA MATRISOME ASSOCIATED, and skeletal system development. Moreover, ossification, collagen metabolic process, syndecan interactions, platelet activation, signaling and aggregation, cell-substrate adhesion, endodermal cell differentiation, response to growth factor, blood vessel development, proteoglycans in cancer, regulation of insulin-like growth factor (IGF) transport and uptake, and other pathways were also involved in the regulation of ADAM12-interacting genes ( Figure 8A ). Furthermore, the results of the protein–protein interaction network and KEGG pathway analysis also showed enrichment in extracellular matrix organization, NABA CORE MATRISOME, ECM proteoglycans, NANA MATRISOME ASSOCIATED, and skeletal system development ( Figures 8B–E ).

The correlation between ADAM12 expression and immune cell infiltration in COAD was analyzed using the TIMER database. In our results, ADAM12 expression was significantly negatively correlated with tumor purity (Cor=-0.404, P=2.03e-17) and was significantly positively correlated with CD8+ T cells (partial cor=0.285, P=4.73e-09), CD4+ T cells (partial cor=0.386, P=9.43e-16), macrophages (partial cor=0.597, P=2.64e-40), neutrophils (partial cor=0.557, P=5.51e-37), and dendritic cells (partial cor=0.555, P=7.53e-34) ( Figure 9A ). However, the correlation between ADAM12 expression and B cells was not significant (partial cor=0.065, P=1.92e-01). Nevertheless, we further generated Kaplan–Meier plots using the TIMER database to explore whether a high level of immune cell infiltration was linked to a longer survival time of patients with COAD. The results indicated no significant associations [B cells (P=0.478), CD8+ T cells (P=0.239), CD4+ T cells (P=0.162), macrophages (P=0.19), neutrophils (P=0.789), dendritic cells (P=0.785), or ADAM12 (P=0.261)] ( Figure 9B ).

Additionally, correlations between ADAM12 expression and the expression of immune cell-specific markers were also investigated using the TIMER and GEPIA databases, including CD8+ T cells (CD8A and CD8B), T cells (CD2 and CD3D), B cells (CD79A and CD19), monocytes (CD86 and CSF1R), tumor-associated macrophages (TAMs) (CD68 and IL10), M1 macrophages (NOS2 and PTGS2), M2 macrophages (CD163 and VSIG4), natural killer cells (KIR2DL1, KIR2DL3, KIR3DL1, and KIR3DL2), dendritic cells (HLA-DPB1 and CD1C), Th2 cells (GATA3 and STAT6), Tfh cells (BCL6 and IL21), Th1 cells (STAT4 and STAT1), Th17 cells (STAT3 and IL17A), neutrophils (CCR7 and CEACAM8), Treg cells (FOXP3, CCR8), and T cell exhaustion (CTLA4 and LAG3). Based on the TIMER database, we observed a significant positive correlation between ADAM12 and most markers, including CD8A, CD8B, CD2, CD3D, CD79A, CD19, CD86, CSF1R, CD68, IL10, PTGS2, CD163, VSIG4, CCR7, KIR2DL1, KIR2DL3, KIR3DL1, KIR3DL2, HLA-DPB1, CD1C, STAT4, STAT1, GATA3, STAT6, BCL6, IL21, STAT3, FOXP3, CCR8, CTLA4, and LAG3, and a significant negative correlation between ADAM12 and the markers NOS2, CEACAM8, and IL17A. All results obtained from the TIMER database are shown Figure 9C . In addition, we obtained similar results using the GEPIA database, and our analysis results are shown in Table 2 . Therefore, these results strongly suggested that the ADAM12 gene mediated the regulation of immune cell infiltration in CRC, thereby promoting immune escape and the chemotherapy resistance of tumor cells.

We explored the relationships between the ADAM12 expression level and immune components, such as lymphocytes, immunomodulators, and chemokines, in patients with COAD using the TISIDB tool to obtain a deeper understanding of the association of ADAM12 with immune cell infiltration.

First, the relationships between the abundance of tumor-infiltrating lymphocytes and ADAM12 expression levels were investigated to examine which types of TILs might be regulated by the ADAM12 gene. The outcomes indicated that the ADAM12 expression level was positively correlated with Tfh cells (rho=0.594, P<2.2e-16), Tcm_CD8 cells (rho=0.572, P<2.2e-16), Treg cells (rho=0.7, P<2.2e-16), NK cells (rho=0.728, P<2.2e-16), and MDSCs (rho=0.608, P<2.2e-16) ( Figure 10A ). However, the abundance of monocytes was not regulated by the ADAM12 gene (rho=-0.002, P=0.96) ( Figure 10A ).

Next, we evaluated the relations between three different types of immunomodulators and ADAM12 expression, and these immunomodulators included immunoinhibitors, immunostimulators, and major histocompatibility complex molecules. The immunomodulators were collected from the study by Charoentong. Correlations between immunoinhibitors, including IL10_exp, CD96_exp, CTLA4_exp, CD160_exp, TGFBR1_exp, and PDCD1_exp, and ADAM12 are shown in Figure 10B . Correlations between immunostimulators, including CD28_exp, CD80_exp, CD86_exp, CXCL12_exp, IL6_exp, and TNFRSF17_exp, and ADAM12 are shown in Figure 10C . Correlations between major histocompatibility complex molecules, including HLA-A_exp, HLA-B_exp, HLA-DOA_exp, HLA-DRA_exp, HLA-E_exp, and TAP1_exp, with ADAM12 are shown in Figure 10D .

Finally, we explored the relationships between chemokines, receptors, and ADAM12 expression. Correlations between chemokines, including CCL3_exp, CCL5_exp, CCCL11_exp, CCL17_exp, CXCL9_exp, and XCL1_exp, and ADAM12 are shown in Figure 10E . Correlations between receptors, including CCR2_exp, CCR4_exp, CCR8_exp, CXCR1_exp, CXCR4_exp, and CXCR5_exp, and ADAM12 are shown in Figure 10F .

The results strongly suggested that ADAM12 regulated various immune molecules in the tumor microenvironment of CRC through many pathways, thereby affecting immune cell infiltration.