The Pyroptosis-Related Risk Genes APOBEC3D, TNFRSF14, and RAC2 Were Used to Evaluate Prognosis and as Tumor Suppressor Genes in Breast Cancer

Background Pyroptosis is a type of cell death that plays an important role in predicting prognosis and immunoregulation in cancers. However, the pyroptosis-related gene signature for prognosis and immune infiltration prediction has not been studied in breast cancer (BC). Methods The Gene Expression Omnibus (GEO) and Cancer Genome Atlas (TCGA) databases were used to obtain the expression and clinical data of genes. 52 pyroptosis-related genes were obtained from TCGA-BC and estimated differentially expressed genes by the limma program. To categorize the molecular subtypes of pyroptosis-related genes, the ConsensusClusterPlus tool was utilized. Cox and Lasso regression analyses were used to create a signature. TCGA-BC dataset as the training set and the GSE37751 test set for risk research. Gene set enrichment analysis (GSEA) was used to conduct KEGG and GO studies of subtype groups. We also used the ssGSEA approach in the GSVA package to calculate the risk score of immune cells. Finally, pyroptosis-related genes in BC were validated using qPCR and immunohistochemical assays. Clone formation and EDU assays were used to explore the ability of signature genes to regulate the proliferation of BC cells. Results Based on pyroptosis-related genes, the C1 and C2 subtypes were obtained. Survival analysis results showed that the C2 group had a better prognosis. Then, a three-gene signature (APOBEC3D, TNFRSF14, and RAC2) were created by Lasso regression analysis, which had a good prediction effect in the TCGA-BC and GSE37751 datasets. Our nomogram has a fair degree of accuracy in predicting the survival rates of BC patients. The pyroptosis-related signature has a good predictive effect in evaluating the tumour microenvironment score, 28 types of immune cells and response to immune checkpoint therapy. Finally, qPCR and immunohistochemistry staining results indicated that APOBEC3D, TNFRSF14, and RAC2 expression in BC tissues was low. The results of clone formation and EdU assays showed that high expression of signature genes inhibited the proliferation ability of BC cells. Conclusions Based on pyroptosis-related genes (APOBEC3D, TNFRSF14, and RAC2), we built a novel prognostic molecular model for BC that might be used to assess prognostic risk and immune infiltration in BC patients. These signature genes are also tumor suppressor genes and may serve as potential targets for BC.


Introduction
Breast cancer (BC) is the main health concern and is the most prevalent tumour among females worldwide. It is estimated that in 2020, 4.57 million new BC cases will be detected, and approximately 680,000 people are expected to die from BC [1]. e NCCN guidelines recommend that BC is mainly treated with surgery, chemotherapy and antioestrogen therapy [2]. However, the value of treatment is not very good among advanced BC patients. BC is a diverse tumor with four major molecular subgroups; therefore, finding new biomarkers is still vital for early diagnosis and treatment methods. Pyroptosis is the process of gasderminmediated programmed cell death (PCD), which is known to involve extracellular responses and has been widely studied in many cancers [3]. Pyroptosis has been shown to successfully remove malignant cells and provide novel cancer treatment strategies [4]. Surprisingly, inflammasome-mediated pyroptosis has been linked to tumor formation and immunology in recent research [5]. As a result, finding a pyroptosis-related signature to predict BC prognosis and treatment methods is extremely important. e TCGA project, which provides a comprehensive genetic examination of various malignancies and demonstrates links with clinical outcomes. In addition, the tumour project of TCGA includes mutations, genomic copy number changes, transcriptome, and methylation profiles [6]. To characterize molecular profiles, researchers combined information from transcriptome RNA sequencing with applied genomic characterizations, which revealed potential druggable targets for female tumors including BC [7][8][9]. In addition to identifying nearly all genes previously linked to BC, the researchers discovered numerous new and severely altered genes, including BRCA1 and BRCA2, which could be used as therapeutic targets.
In this study, we identified pyroptosis-related genes from TCGA-BC and used them to construct a novel predictive molecular model for BC. In addition, the model has the potential to be a useful tool for assessing prognostic risk and immune infiltration in BC patients. In conclusion, our findings imply that the signature might be utilized to assess prognosis and immune infiltration in BC and that the signature genes could be employed as possible targets for the disease.

Molecular Subtype Identification.
Limma software was used to analyse the differentially expressed genes (DEGs) based on the threshold false discovery rate (FDR) < 0.05 after the 52 pyroptosis-related genes expression data were matched with the TCGA-BC dataset. Next, Consensu-sClusterPlus was used to find new molecular subclasses of BC, which provides quantitative evidence for determining the number and membership of possible clusters within the TCGA dataset.

Multivariate Analyses and Molecular Risk Model
Construction. For the TCGA-BC dataset, we used Cox regression analysis. A p value of 0.05 was judged survival linked based on the results of multivariate analysis. Furthermore, the R software package glmnet for lasso Cox regression was used to compress the screened genes and used to build the risk model. We also employed the TCGA-BC dataset as the study's training set and the GSE37751 test set.

Analysis of Immune Scores between Clusters.
e immunological score among the clusters in TCGA-BC dataset was determined using the GSVA package's single-sample gene set enrichment analysis (ssGSEA) approach. We used ESTIMATE software to estimate the tumor microenvironment score for tumor purity, StromalScore, ImmuneScore, and ESTIMATEScore. Twenty-eight different types of immune cells were evaluated using the GSVA program ssgsea. e differences in immune ratings between the molecular subtypes were then compared.
Furthermore, we analysed the correlation of the molecular risk model with immune-inhibitory markers. We collected 6 immune-inhibitory markers, including CD274, PDCD1, PDCD1LG2, CTLA4, HAVCR2, and IDO1, from the published literature. Using the chi-square test, the response to immune checkpoint therapy was estimated and compared.

Tissue
Samples. Ten BC tissues were collected and kept at 80°C. Preoperative antitumor treatments were not given to any of the patients. Informed consent papers were signed by patients. is study was approved by the Ethics Committee of Shanghai Tongren Hospital (2021-088-02).

2.7.
Immunohistochemistry. Paraffin sections of breast cancer tissue were used for immunohistochemistry. e slides were dewaxed with methanol and rehydrated with alcohol after being dried at 60°C. e slides were then submerged in 3% hydrogen peroxide overnight and labelled with antibodies. e experiment was carried out with the manufacturer's instructions. e antibodies purchased from Abcam as follows: APOBEC3D antibody (ab105869), anti-TNFRSF14 antibody (ab47677) and anti-RAC2 antibody (ab2244). e immunohistochemistry results were evaluated under a microscope at 20 × 10. e IHC findings were analysed by Image-Pro Plus 6.0 Software.

Cell Lines and Transfection.
e human normal mammary epithelial cell line MCF10 A, and BC cell lines (MDA-MB-231 and MCF-7) were purchased from the National Collection Authenticated Cell Cultures (Shanghai, China). All cells were incubated at 37°C and 5% CO 2 in a incubator. Transfection was carried out by Lipofectamine 3000 reagent (Invitrogen, China, No. L3000015) according to the instructions.
e coding sequences of human APOBEC3D, TNFRSF14, and RAC2 were cloned into the pEZ-M03 vector.

Ethynyl Deoxyuridine (EdU) Assays.
e experiment was carried out exactly as instructed. Cells were cultivated at a density of 10000 cells in 96-well plates per well. e 96-well plates were then incubated for 3 hours at 37°C with 10 M EdU labelling medium (Beyotime Biotechnology, Shanghai, China). After fluorescence microscopy inspection, the percentage of EdU-positive cells was determined.

Colony Formation Assay.
A total of 1000 cells were placed in six-well plates for the colony formation test. e cells were mixed together and grown for one week in culture media containing 10% FBS. A single colony was defined as a cluster of 30 cells or less.

Statistical Analysis.
e SPSS 13.0 statistical software program was used to analyse the data (IBM Corporation, Armonk, NY, USA). GraphPad Prism 8.0 was used to create the graphs (GraphPad Software, Inc., San Diego, CA). Statistical significance was defined as a p value < 0.05.

Identification and Molecular Pyroptosis-Related Type.
e TCGA-BC dataset was used to calculate 52 pyroptosisrelated genes expression, and 21 genes was high expression and 17 genes was low expression (Figure 1(a)) in BC. To further investigate the interrelationship among the DESs, a PPI network and correlation analysis were constructed. GSDMD and CHMP6 were shown to be linked to the risk of BC in the study (Figures 1(b) and 1(c)).
e Consensu-sClusterPlus tool was also used to perform clustering analysis. e 1096 BC samples were classified into C1 and C2 clusters (Figure 1(d)). As shown in Figure 2(a) (p � 0.006), C1 had the worst prognosis, and C2 had the best prognosis in BC. In addition, we counted the differentially expressed genes based on the clusters. A total of 1190 DES (padj <0.05 and |log2FC|>1) were found to be common between the two groups ( Figure 2(b)). Between the 1190 candidate DESs mentioned above and the survival data, we ran multivariate Cox regression analyses. APO-BEC3D, TNFRSF14, and RAC2 were all found to be risk variables in a forest plot of HRs. To minimize the genes number for the risk model, Lasso regression was utilized (Figure 2(d)). As shown in Figure 2(e), we then utilized a 10-fold cross test to build the model and confidence interval for each lambda. e following is the final 3-gene signature formula: RiskScore

Risk Model Analysis and Comparison.
We used the TCGA-BC dataset as the training set and the GSE37751 test set for risk research to determine whether our signature was feasible. To validate the prognostic relevance of the risk score, the Kaplan-Meier survival curves, ROC curves, and risk score distributions for OS prediction were examined. In both the training and test sets, the risk model was highly connected to the prognosis of BC patients, as shown in Figures 3(a) and 3(b). ROC curve results showed that the prognostic prediction for 1, 3, and 5 years had good classification efficiency (Figures 3(c) and 3(d)). ree prognostic risk models (PMID 34589498) were chosen for comparison with our risk model. e 1-, 3-, and 5-year AUC values for the 3-gene signature model were lower than those for our model. is finding demonstrates that our model produces better results (Figure 3(e)). In the TCGA-BC dataset, as the risk score increased, the expression levels of APOBEC3D, TNFRSF14 or RAC2 were downregulated, and the number of surviving patients decreased ( Figure 3(f )).
ese findings in the GSE37751 external test set, which were from different data sources, indicate that the risk signature performs well in predicting the survival of BC (Figure 3(g)).

Cox Regression Analysis and Nomogram Construction.
In data mining, PCA and t-SNE are commonly utilized. In both the training (4(a) and 4(b)) and test sets, we discovered that risk models can effectively discriminate risk patients (Figures 4(c) and 4(d)). Between the survival data and the risk model, univariate and multivariate Cox regression analyses were performed. e forest plot revealed that separate survival time parameters in the training (Figures 4(e) and 4(g)) and test sets influenced the risk model (Figures 4(f ) and 4(h)). Furthermore, we analysed the DEGs involved in pyroptosis using KEGG pathway enrichment analysis and GO analysis. e enriched biological process (BP) term was linked to the humoral immune response, the enriched molecular function (MF) term to T-cell activation (Figure 4(i)), and the enriched KEGG pathways to the NF kappa B signaling network and T-cell receptor signaling pathway (Figure 4(j)). Risk signatures may be applied intuitively and successfully with nomograms, and outcomes can be predicted with ease. Our nomogram, as shown in Figure 5, has a fair degree of accuracy in predicting the survival rates of BC patients.

Analysis of Immune Scores among Molecular Subtypes.
We used ESTIMATE software, which can predict the tumour microenvironment score. Our model can distinguish the estimate score ( Figure 6(a)), stromal score ( Figure 6 (Figure 7(p)). Furthermore, a heatmap was used to evaluate the tumour microenvironment score (Figure 8(a)) and immune cells (Figure 8(b)). Patients with lower risk scores had a better response to ICI therapy, indicating that the pyroptosis-
In the present study, we identified pyroptosis-related genes from TCGA-BC, which used them to construct a novel predictive molecular model for BC (APOBEC3D, TNFRSF14, and RAC2). For the risk analysis, we used the training set (TCGA-BC dataset) and test set (GSE37751 dataset) to determine whether our signature was feasible. Our signature had good classification efficiency of the Kaplan-Meier survival curves, ROC curves, and risk score distributions for OS prediction. Furthermore, we compared our model with other risk models and our model has a more effective result. e risk model is the influence of survival time and accuracy for forecasting the survival rates of BC patients in Cox regression analysis and the nomogram. Our model can identify the estimate score, stromal score, immunological score, and purity of tumors well, which is another key point of the risk model. Additionally, the pyroptosis-related signature has a good predictive effect in evaluating immune cells and checkpoint therapy. e qPCR and immunohistochemistry results showed that APOBEC3D, TNFRSF14, and RAC2 were expressed at lower levels in stages III and IV (high-risk group) in BC tissues. Furthermore, the results of biological functions revealed that overexpression of APOBEC3D, TNFRSF14, and RAC2 greatly suppressed MCF-7-cell proliferation. e findings of the current study provide more effective tools for predicting prognosis and immune infiltration in BC, and the signature genes may serve as potential targets for BC, which have not been found in previous studies.
Here, pyroptosis-related APOBEC3D, TNFRSF14, and RAC2 genes were considered risk genes for BC. It has been reported that the expression of APOBEC3D [19,20], TNFRSF14 [21,22], and RAC2 [23,24] is dysregulated and is a potential target for therapy in cancer. In BC, TNFRSF14 and RAC2 are prognostic markers, which is consistent with our findings. For APOBEC3D, we report for the first time that APOBEC3D could be used as a new molecular marker in BC. However, our study also has some limitations: 1. A small number of clinical samples were used to test the pyroptosis-related APOBEC3D, TNFRSF14, and RAC2 genes. In future studies, we will expand the number of samples for research. 2. e function of APOBEC3D in vitro experiments will be analysed in the future.
Finally, our research study presents a unique pyroptosisrelated genes prognostic molecular model (APOBEC3D, TNFRSF14, and RAC2) that could be used to assess prognostic risk and immune infiltration in BC. e overexpression of APOBEC3D, TNFRSF14, and RAC2 significantly reduced MCF-7 cell proliferation, according to the results of biological functions. Our pyroptosis-related signature could be utilized to assess prognosis and immune infiltration and it could be used to identify potential targets for BC.

Data Availability
e datasets used or analysed during the current study are available from the corresponding author on reasonable request.

Conflicts of Interest
e authors declare that they have no conflicts of interest.

Authors' Contributions
e study was planned and designed by YJ, X, and FY. e experiments were carried out by CQ. e data were analysed and the manuscript was written by JH and CGY. e final manuscript was approved by all authors.