Using UCSC Xena (https://Xena.UCSC.edu/), an online tool for exploring gene expression and clinical and phenotypic data, we downloaded the RNA sequence, somatic mutation, and related clinical data from the TCGA (comprising 10,201 samples from 33 types of cancer). Gene expression data from 31 different normal tissues were downloaded from the GTEx database (https://commonfund.nih.gov/GTEx). All gene expression data were normalised using log2 transformation. Normal and cancer tissues were compared using the t-test. The Kaplan-Meier curve, logarithmic rank test, and Cox proportional hazard model were used in all survival analyses. The correlation between variables were calculated using the Spearman or Wilcoxon tests. All statistical analyses were performed using R software (version 4.0.2 or 3.6.3, https://www.r-project.org). Statistical significance was set at p < 0.05.

We entered “CASP3” in the tumour immune estimation resource (TIMER2) “Gene DE” module (version 2, http://timer.cistrome.org/) to investigate the differences in CASP3 expression among different tumours or specific tumour subtypes in the TCGA data using tumour and adjacent normal tissues. Certain tumours did not have the corresponding normal tissue samples or displayed lower expression in normal tissues. Under these conditions, we entered “CASP3” in the gene expression profiling interactive analysis (GEPIA2; http://gepia2.cancer-pku.cn/#index), an online tool for gene expression and co-expression analysis, in the “DIY Expression” module’s “Box plot”, with the following parameters: p-value cut-off = 0.01, log2FC (fold change) cut-off = 1, log2 (TPM + 1) for log-scale, and “Match TCGA Normal and GTEx”, and obtained the differential scatter plot and the expression bodymap of CASP3 between tumours and corresponding normal tissues in the GTEx database. The UALCAN portal (http://ualcan.path.uab.edu/analysis-prot.html) is an online tool for the proteomic analysis of several types of cancer. We used the Clinical Proteomic Tumour Analysis Consortium (CPTAC) dataset for protein expression analysis. We analysed the expression of CASP3 in tumour and normal tissues by inputting “CASP3”. Six available tumour datasets were selected, namely breast, ovarian, colon, clear cell renal cell carcinoma (RCC), UCEC, and lung adenocarcinoma (LUAD). Subsequently, to evaluate the difference in CASP3 expression at the protein level, we downloaded and compared the immunohistochemistry (IHC) images for CASP3 protein in normal tissues and six tumour tissues from the Human Protein Atlas (HPA; http://www.proteinatlas.org/).

By combining gene expression and clinical data from each sample extracted from the TCGA database, the relationship between the CASP3 expression and patient prognosis was studied using four indices: overall survival (OS), disease-specific survival (DSS), disease-free interval (DFI), and progression-free interval (PFI). Kaplan-Meier and log-rank tests were used for the survival analysis (p < 0.05). The prognostic data were visualised using R software (version 3.6.3), with the R packages “survival” and “survminer”. In addition, the Cox analysis used the R packages “survival” and “forestplot” to determine the pan-cancer relationship between CASP3 expression and survival. We then obtained the stage differences of CASP3 in different tumours using the GEPIA2 “Stage plot” module and constructed the relevant violin plots using log2[TPM + 1].

We entered the gene “CASP3” in the UALCAN portal and analysed the differences in methylation expression in tumour and normal tissues. Thereafter, we used SurvivalMeth (http://bio-bigdata.hrbmu.edu.cn/survivalmeth/), an online tool for studying the effect of gene methylation on prognosis, with the parameters set as Method “T-test” and “Threshold Value” = 0.01, the “Maxstat” grouping strategy, and the remaining settings at “Without Restriction” to obtain the effect of CASP3 methylation levels on overall survival in TCGA database and the Kaplan-Meier survival curve with a statistically significant p-value (p < 0.05).

On the cBioPortal website (https://www.cbioportal.org/), an interactive exploration dataset for multiple cancer genomics, we selected “TCGA PanCancer Atlas study” in the “Quick By Gene” and entered “CASP3” to query for genetic mutation-related characteristics of CASP3. Mutation-related results were observed in all the TCGA tumours in the “Cancer Type Summary” module. Information regarding the CASP3 mutation site can be displayed in a protein structure sketch map or three-dimensional (3D) structure using the “Mutation” module. We also used the “Comparison/Survival” module to obtain OS, DSS, DFI, and PFI data for TCGA cancer cases with more CASP3 mutations and generated the Kaplan-Meier survival curve. TMB is defined as the total number of somatic gene coding, base insertion, replacement, or deletion errors detected per million bases (Yarchoan et al., 2017), which is an important biological indicator of the extent of mutation in tumours; a higher TMB often indicates better outcomes for tumour immunotherapy (Le et al., 2015). The MSI results were obtained as a result of functional defects in DNA mismatch repair in tumour tissues, and MSI-H tumours often imply better treatment outcomes (Lin et al., 2020). We then analysed the association between CASP3 expression and TMB using mutation data from the TCGA database, derived from the 2018 study by Thorsson et al. (Thorsson et al., 2018) using R software (version 4.0.3); p < 0.05 was considered statistically significant. We simultaneously analysed the correlation between CASP3 expression and MSI data obtained from the study by Bonneville et al. (Bonneville et al., 2017).

Before analysing the tumour microenvironment, we first analysed the correlation between the expression of CASP3 and eight immunologic checkpoints (SIGLEC15, ID O 1, CD274, Havcr2, PDCD1, CTLA4, LAG3, and PDCD1LG2). The expression values of these eight genes and CASP3 in all tumours were extracted, and a correlation heatmap was visualised using R software (version 4.0.3). We used “the cancer-associated fibroblast” module in TIMER2 to explore the relationship between CASP3 expression and fibroblast infiltration, and used the EPIC, MCP-counter, and TIDE algorithms to evaluate the data, which were visualised as a heatmap and scatter plot. Thereafter, we used the “Outcome” module of TIMER2 to analyse the prognosis and obtain the Kaplan-Meier survival curve with p < 0.05. We then used the previously downloaded mRNA data from TCGA database and the latest algorithms from TIMER, xCell, MCP-counter, CIBERSORT, EPIC, and quanTIseq with the R package “Immunedeconv” (R software, version 4.0.3) (Sturm et al., 2020) to perform immune infiltration analysis and evaluate the correlation between CASP3 expression and infiltration of immune cells in different tumours, which was then visualised as a heatmap. The Kaplan-Meier curve was obtained using the same methods.

Gene Ontology (GO) and Kyoto Encyclopaedia of Genes and Genomes (KEGG) gene sets were downloaded from the GSEA website (https://www.GSEA-msigdb.org/GSEA/downloads.jsp). Functional analysis was performed using the R software (version 3.6.3) packages “Limma”, “Org.Hs.eg.db”, “DOSE”, “ClusterProfiler”, and “Enrich plot” to visualise the five most significant CASP3 enrichment pathways in different tumours. p < 0.05 was considered statistically significant.

We used the gene name “CASP3” and the “Homo sapiens” condition to search the STRING database (https://string-db.org/) with the parameter minimum required interaction score (“Low confidence (0.150)”), meaning of network edges (“evidence”), and max number of interactors (“no more than 50 interactors” in 1st shell), to obtain the CASP3-associated protein network map. Next, we used the GEPIA2 “Similar gene detection” module based on TCGA and GTEx datasets to obtain the first 100 genes closely related to CASP3. In addition, the co-expression of CASP3 and related genes in different tumours was plotted using the R package “Limma” (R software, version 3.6.3). Thereafter, CASP3 and the selected genes, namely DDX46, GNAI3, PDS5A, SCYL2, and TMPO, were analysed using “Correlation analysis” in GEPIA2 and a scatter plot was obtained. We used two sets of data to perform the GO and KEGG enrichment analysis and used R software (version 4.0.3) and the packages “ClusterProfiler”, “Org.Hs.eg.db”, “Enrichplot”, and “ggplot2” to visualise the data as bubble plots. The first five correlation pathways were selected to draw the loop plot of the related genes.

We first analysed the differential expression of CASP3 among different tumours in the TCGA database using the TIMER2 method (Figure 1A) We observed a statistically significant overexpression in bladder urothelial carcinoma (BLCA), breast invasive carcinoma (BRCA), cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), cholangiocarcinoma (CHOL), oesophageal carcinoma (ESCA), kidney renal clear cell carcinoma (KIRC), kidney renal papillary cell carcinoma (KIRP), lung squamous cell carcinoma (LUSC), glioblastoma multiforme (GBM), head and neck squamous cell carcinoma (HNSC), liver hepatocellular carcinoma (LIHC), lung adenocarcinoma (LUAD), stomach adenocarcinoma (STAD), thyroid carcinoma (THCA), and uterine corpus endometrial carcinoma (UCEC). For tumours with no or few corresponding normal tissue samples, such as lymphoid neoplasm, diffuse large B-cell lymphoma (DLBC), brain lower grade glioma (LGG), CASP3 as a negative expression of colon adenocarcinoma (COAD), and rectum adenocarcinoma (READ), we used normal tissue expression from the GTEx database added to GEPIA2 to analyse the results (Figures 1B,C). In previously negative and non-normal tumours, CASP3 expression was significantly higher in DLBC, LGG, COAD, and READ tissues compared with that in normal tissues (p < 0.001). CASP3 expression was high in almost all other human tumours (Figure 1D).

Subsequently, to evaluate CASP3 expression at the protein level, we analysed the results of IHC staining in the HPA database and compared them with the levels of protein expression in six tumours obtained from the CPTAC database. The expression of CASP3 protein was significantly higher in all tumours, except in the colon (p < 0.001, Figures 2A–F). The remaining images indicate CASP3 IHC staining in normal tissues (Figures 2G–K) and in five tumours (Figures 2L–P). Unfortunately, no staining results have been reported for clear cell RCC. CASP3 IHC staining was low in normal tissues of the breast, uterus, ovary, and lung, and was moderate in normal tissues of the colon. Breast cancer and ovarian cancer tissues showed strong staining whereas colon cancer, UCEC, and LUAD tissues all showed moderate staining; with the exception of the colon, CASP3 protein expression in tumour tissues was significantly higher than that in normal tissues, which was consistent with the results obtained from the CPTAC database.

Concerning prognosis, we performed survival association analysis for each tumour type, including OS, DSS, DFI, and PFI data. The Cox model of OS rate showed that CASP3 expression was correlated with adrenocortical carcinoma (ACC, p < 0.001), KIRC (p < 0.001), acute myeloid leukaemia (LAML, p = 0.019), LGG (p < 0.001), SKCM (p < 0.001), thymoma (THYM, p = 0.008), and uveal melanoma (UVM, p < 0.042, Figure 3A). CASP3 was identified as a high-risk gene in ACC, KIRC, LGG, and UVM, particularly in ACC, with a hazard ratio of 4.274, whereas it was a low-risk gene in the remaining tumours. Kaplan-Meier survival analysis also showed that high CASP3 expression in ACC (p = 0.005, Figure 3B) and LGG (p < 0.001, Figure 3C) was associated with low OS, whereas high CASP3 expression in SKCM (p = 0.016, Figure 3D), THYM (p = 0.013, Figure 3E), and COAD (p = 0.036, Figure 3F) was associated with better prognosis.

Furthermore, in DSS, high CASP3 expression in ACC (p < 0.001), KIRC (p < 0.001), LGG (p < 0.001), mesothelioma (MESO, p = 0.039), and UVM (p = 0.05) was associated with poor prognosis (Figure 3G). However, we observed opposing results in LUSC (p = 0.034), ovarian serous cystadenocarcinoma (OV, p = 0.028), and SKCM (p < 0.001). Kaplan-Meier survival analysis showed that high CASP3 expression was not only associated with poor prognosis in ACC (p = 0.004, Figure 3H), KIRC (p = 0.019, Figure 3I), and LGG (p < 0.001, Figure 3J) but also with the prognosis of LUSC (p = 0.022, Figure 3K), SKCM (p = 0.014, Figure 3L), BLCA (p = 0.045, Figure 3M), UCEC (p = 0.025, Figure 3N), and THYM (p = 0.041, Figure 3O). In DFI (Figure 4A), in contrast to the CASP3-related good prognosis in OV (p = 0.013), CASP3 overexpression was associated with poor prognosis in ACC (p = 0.008), KIRP (p = 0.048), LUAD (p = 0.005), and THCA (p < 0.001). The Kaplan-Meier survival analysis showed the same results for OV (p = 0.012, Figure 4B), with a good prognosis, whereas the prognoses of ACC (p = 0.012, Figure 4C), BLCA (p = 0.009, Figure 4D), KIRP (p = 0.012, Figure 4E), and LUAD (p = 0.005, Figure 4F) were poor. Finally, high CASP3 expression was associated with low PFI in ACC (p < 0.001), KIRC (p = 0.018), LGG (p < 0.001), LUAD (p = 0.044), prostate adenocarcinoma (PRAD; p = 0.002), and UVM (p = 0.002, Figure 4G). Kaplan-Meier survival analysis showed that the high CASP3 expression was associated with poor prognosis in ACC (p < 0.001, Figure 4H), LUAD (p = 0.029, Figure 4I), LGG (p = 0.001, Figure 4J), PRAD (p = 0.017, Figure 4K), and KIRP (p = 0.036, Figure 4L); high CASP3 expression was also associated with poor prognosis in sarcoma (SARC; p = 0.016, Figure 4M) and good prognosis in READ (p = 0.006, Figure 4N), COAD (p = 0.028, Figure 4O). Further, we used the “Stage Plot” module of GEPIA2 to observe the relationship between CASP3 expression and the pathological staging of tumours. We observed correlations between the expression of CASP3 and COAD (Figure 4P), SKCM (Figure 4Q), OV (Figure 4R), and THCA (Figure 4S, p < 0.05). The remaining tumours are shown in Supplementary Figure S1.

First, we analysed the differences in CASP3 methylation levels in TCGA using the UALCAN database. In ESCA (Figure 5B), KIRC (Figure 5C), LUSC (Figure 5D), and SARC (Figure 5G), CASP3 methylation levels were high, whereas in BLCA (Figure 5A), PRAD (Figure 5E), KIRP (Figure 5F), testicular germ cell tumours (TGCT; Figure 5H), and UCEC (Figure 5I), the reverse was true. We then used the “Single Case” plate of SurvivalMeth to analyse the relationship between CASP3 methylation and tumour prognosis in TCGA database and obtain the Kaplan-Meier survival curve. Despite tumours that could not be analysed effectively because of lack of data, we observed that CASP3 methylation levels only in ESCA (p = 0.047, Figure 5K) indicated a good prognosis, whereas in BLCA (p = 0.027, Figure 5J), KIRC (p = 0.010, Figure 5L), and LUSC (p < 0.001, Figure 5M), CASP3 methylation levels were correlated with poor prognosis (p < 0.05). The levels of methylation in the other tumours and the associated prognoses are shown in Supplementary Figure S2.

We observed genetic changes in CASP3 in different TCGA tumours using cBioPortal. The highest frequency of CASP3 mutations was 3% in UCEC (Figure 6A). In DLBC, “deep deletion” highlighted the full range of mutations (>8%). Notably, almost all tumours with mutations had a missing copy number of CASP3. All genetic loci of CASP3 and the number of cases are shown in Figure 6B. The missense mutation of CASP3 was the main type of genetic change, accounting for more than 80% of cases. In all three cases, mutations at R147C were detected in SKCM. A 3D map of CASP3 mutations at this site is shown in Figure 6C. In addition, we continued to explore the relationship between CASP3 mutations and clinical survival. In DLBC, CASP3 mutations resulted in poor OS (p = 0.005, Figure 6D), DSS (p = 0.006, Figure 6E), and PFI (p = 0.04, Figure 6F), therefore, it was chosen for the analysis as it had the most mutations. However, for DFI (p = 0.7, Figure 6G), which was not statistically significant because there was only one case of mutation data.

We also analysed the association of CASP3 expression with TMB and MSI in all tumours in TCGA database. In STAD (p < 0.001), UCEC (p < 0.001), BRCA (p < 0.001), ACC (p = 0.002), COAD (p = 0.002), LGG (p = 0.002), PRAD (p = 0.015), and pancreatic adenocarcinoma (PAAD, p = 0.025), CASP3 expression was positively correlated with TMB, whereas was negatively correlated in UVM (p = 0.003), THCA (p < 0.001), and LIHC (p = 0.044, Figure 7A). In READ (p = 0.001), UCEC (p < 0.001), STAD (p < 0.001), and KIRC (p = 0.019), CASP3 expression was positively correlated with MSI, but was negatively correlated with MSI in DLBC (p = 0.008) and LUAD (p = 0.03, Figure 7B).

We first analysed the correlation between CASP3 and eight immunologic checkpoint-related genes in all tumours. CASP3 expression was negatively correlated (p < 0.05) with these immunologic checkpoint-related genes in all tumours, except in ACC, GBM, THYM, and TGCT (Figure 7C). Thereafter, the tumour microenvironment was analysed. Tide, MCP-counter, and EPIC algorithms were selected using TIMER2’s “Associated cancer fibroblast” module to analyse the relationship between CASP3 expression and fibroblast infiltration in different TCGA tumours. CASP3 expression was positively correlated with the invasion of fibroblasts in ACC, BRCA-lumA, HNSC-HPV, GBM, KIRC, KIRP, LGG, LIHC, LUAD, PAAD, SARC, TGCT, and THCA, but was negatively correlated with fibroblast infiltration in READ (Figure 8A); the relevant scatter plots are shown in Figures 8D–I. We further analysed the effects of CASP3 expression and fibroblast infiltration on tumour prognosis. We chose the algorithm with obvious differences in prognosis; in turn, in GBM, KIRP, and LGG, poor prognosis was observed when fibroblasts were highly infiltrating (Figures 7J,K). The remaining relevant scatter plots are shown in Supplementary Figure S3.

Next, TIMER, xCell, MCP-counter, CIBERSORT, EPIC, and QUANTISEQ algorithms were used to investigate the potential relationship between the invasion level of different immune cells and CASP3 expression in different cancer types in TCGA. Based on MCP-counter and TIMER algorithm data, we found that, almost all other tumours in the presence of CASP3 expression was positively correlated with relevant immune cell infiltration in almost all tumours, with exception of ACC, GBM, and TGCT (Figures 9A,B). Based on several algorithms, we found that CASP3 expression was closely related to the immune invasion of many cell types. We selected the TIMER algorithm and analysed the prognostic correlation. The scatter plots and Kaplan-Meier survival curve are shown in Figures 9C–J. Increased numbers of CD4⁺ T cells, CD8⁺ T cells, neutrophils, and macrophages in LGG were associated with poorer prognosis. High neutrophil infiltration in ESCA and high macrophage infiltration in LIHC were also associated with poor prognosis (Figures 10A–N). In BLCA, increased numbers of CD4⁺ T cells and B cells indicated an improved prognosis, and increased neutrophil infiltration in SKCM and increased CD4⁺ T cell infiltration in PAAD and SARC also predicted improved prognosis. The correlation scatter plots of the remaining tumours are shown in Supplementary Figure S4.

We performed GSEA to investigate the biological significance of CASP3 expression in different tumours from TCGA. The results based on KEGG and GO are shown in Figures 11A–Q. In KEGG enrichment analysis, CASP3 expression in CHOL, LGG, OV, PRAD, THYM, and THCA was related to autophagy regulation, cytosolic DNA sensing pathway, cytokine receptor interaction, ascorbate and aldarate metabolism, neuroactive ligand receptor interaction, and porphyrin and chlorophyll metabolism, and mostly occurred through positive regulation, although it occurred through a negative regulation in STAD. These included related pathways that were positively regulated in most tumours, such as autophagy regulation and cytosolic DNA sensing pathway. Further, CASP3 expression in OV and TGCT was positively regulated by antigen presentation and processing. In GO analysis, CASP3 was most strongly associated with the detection of chemical stimulus pathways in almost all tumours, including the sensory perception of smell, olfactory receptor activity, and mRNA binding. These pathways were positively regulated in BLCA, ESCA, HNSC, and DLBC, and negatively regulated in STAD, LUSC, PRAD, and READ. However, the result of CASP3 expression in OV in GO was the same as that in KEGG, indicating the involvement in B cell activation and immune response-regulating signalling pathways. CASP3 also participates in the CCR chemokine receptor binding pathway and produces positive regulation in DLBC. The remaining tumour-associated GO enrichment analysis is shown in Supplementary Figure S5.

To further investigate the mechanism of CASP3 in tumorigenesis and cancer progression, the first 100 genes related to CASP3 expression were obtained from all tumour expression data in TCGA using GEPIA2. A heatmap of the correlation is shown in Figure 12A. In addition, we screened 50 CASP3-binding proteins from the STRING website, supported by experimental data, as shown in Figure 12B. The expression of DDX46 (R = 0.46), GNAI3 (R = 0.56), PDS5A (R = 0.57), SCYL2 (R = 0.51), and TMPO (R = 0.63) were positively correlated with the CASP3 levels (p < 0.001, Figures 12C–G). We then combined the two datasets for KEGG and GO enrichment analyses. According to the KEGG results (Figures 12H,I), in addition to apoptosis, CASP3 may play a role in the tumour through infection and as a spliceosome. GO enrichment analysis data (Figures 12J,K) further showed that most of these genes affect cell biology by regulating the activities of apoptosis-related proteases, including peptidase and endopeptidase.