A novel signature integrated endoplasmic reticulum stress and apoptosis related genes to predict prognosis for breast cancer

Abstract

Background

Breast cancer (BC) is the primary cause of cancer mortality. Herein, we aimed to establish and verify a prognostic model consisting of endoplasmic reticulum stress and apoptosis related genes (ERAGs) to predict patient survival.

Methods

The Cancer Genome Atlas (TCGA) database was used to download gene expression and clinical data to identify the differentially expressed genes (DEGs). Using univariate Cox regression analysis and the Least Absolute Shrinkage and Selection Operator (LASSO)-penalized Cox proportional hazards regression analysis, the prognostic ERAGs were screened. The predictive performance was evaluated using Kaplan-Meier (KM) survival and receiver operating characteristic (ROC) curve analysis. Furthermore, a nomogram model incorporating clinical parameters and risk scores was constructed and subsequently evaluated using ROC and KM analysis. The correlation analysis, mutation analysis, functional enrichment analysis, and immune infiltration analysis were employed to investigate the specific mechanism of ERAGs. We also used Quantitative Real-Time PCR (RT-qPCR) to verify the differential expression of DE-ERAGs between the breast cancer cell line and mammary epithelial cell line.

Results

We constructed a prognostic signature comprising 16 ERAGs. ROC, KM analysis and the nomogram model demonstrated high effectiveness in accurately predicting the overall survival (OS) of BRCA patients. The results of these analysis could provide reference for further mechanism exploration.

Conclusion

We developed and assessed a novel molecular predictive model for breast cancer that focuses on endoplasmic reticulum stress and apoptosis in this study. It is a valuable complement to the existing prognostic prediction models for breast cancer.

1. Introduction

Breast cancer has emerged as the most often diagnosed cancer globally, ranking first for incidence and mortality in most countries in 2020 [1]. The prevalence and fatality rates of breast cancer have notably increased in developing countries, especially in China [2]. The primary cause of the elevated death rate is the occurrence of metastasis in critical organs, such as the bone, lung, and brain [3]. Breast cancer treatment encompasses a range of therapeutic modalities based on the stage of the disease. These interventions include surgical procedures, radiation therapy, endocrine therapy, immunotherapy, chemotherapy, and molecular-targeted medication therapy [4]. Despite significant advancements in medical research and treatment, breast cancer is still the leading cause of death among women. Hence, additional comprehensive research is required to investigate molecular mechanisms, find prognostic indicators, and develop new therapeutic targets for breast cancer patients.

The endoplasmic reticulum (ER) is an organelle that plays a vital function in protein synthesis and maturation, as well as serving as a calcium reservoir to regulate intracellular calcium homeostasis. When several circumstances, such as glucose depletion, hypoxia, and others, hinder the ER function of tumor cells, it leads to ER stress, triggered by the accumulation of unfolded or misfolded proteins and the disturbance of calcium homeostasis [5]. Consequently, the unfolded protein response (UPR) is activated to rectify misfolded and unfolded proteins and alleviate the stress [6]. Inositol acquisition enzyme 1α (IRE1α), the protein kinase RNA-like ER kinase (PERK), and activating transcription factor 6 (ATF6) are the three primary stress sensors regulating the UPR [7]. The cell will undergo apoptosis if these correction processes prove inadequate [8]. Apoptosis, a programmed cell death process, is crucial in cancer development and is induced by internal and external pathways. Tumor cells commonly avoid the process of apoptosis by regulating anti-apoptotic and pro-apoptotic gene expression [9]. The UPR is stimulated to activate apoptotic processes via ATF6, PERK, and IRE1 signaling pathways to transduce apoptotic downstream pathways, when ER stress persists beyond the capacity to handle misfolded proteins [10].

Cancer cells are unable to restore ER homeostasis via the UPR when ER stress worsens, and ER stress thus become a factor to promote apoptosis [8,11]. Several studies have provided evidence supporting that ER stress might participate in promoting the apoptosis of tumor cells. Additionally, ER stress has been found to impact various aspects of breast cancer, including multidrug resistance, metastasis, and immunotherapy [[12], [13], [14], [15]]. However, there is still insufficient comprehensive research about ER stress and apoptosis in breast cancer through bioinformatics analysis. Evaluating prognostic factors has significant value in accurately identifying BC patients with poor prognoses, enabling individualized precision treatment strategies. There is currently an absence of a practical prognostic model utilizing ERAGs to predict outcomes in patients diagnosed with BC. Further investigation is required to fully explore the prognostic value and mechanism of ERAGs in BC patients.

In this study, we comprehensively analyzed ERAGs in BC patients. We built and subsequently validated a risk model to forecast the prognostic outcomes of BC patients from TCGA databases. Moreover, we explored the functional enrichment analysis, correlation, function and mutation information of ERAGs to provide evidence for the underlying interaction mechanism study. Additionally, we investigate the composition of immune cells in high- and low-risk groups. The findings of our study suggest that the signature of 16 ERAGs could serve as a reliable prognostic indicator for OS in BRCA patients. Fig. 1 shows the analysis procedure of this study, including specific bioinformatics methods, data processing tools, and research results.

Fig. 1.

Analysis procedure of this study, including specific bioinformatics methods, data processing tools, and research results.

2. Materials and methods

2.1. Data acquisition and processing

From TCGA data portal (TCGA-BRCA cohort-FPKM) (http://portal.gdc.cancer.gov/), we obtained the gene expression and clinical data of 1113 BRCA samples and 113 control samples. Excluding samples with RNA-seq data that lacked corresponding clinical data or retested data, a total of 1176 samples were obtained, including 1065 BRCA samples and 111 control samples. The genes related to endoplasmic reticulum stress and apoptosis were searched with the keywords “endoplasmic reticulum stress” and “apoptosis” in the GeneCards database (https://www.genecards.org/). Genes with a relevance score >2 were screened in this study. The lists of endoplasmic reticulum stress related genes and apoptosis related genes are shown in Table S1 and Table S2. The gene expression validation data set GSE42568 [16] based on the GPL570 platform was downloaded from the GEO database (https://www.ncbi.nlm.nih.gov/geo/), containing 104 breast cancer samples and 17 normal breast biopsies.

The DESeq2 algorithm was used to process gene expression data. The Variance Stabilizing Transformation (VST) function from the DESeq2 package (version 1.36.0) was used to standardize the count data [17,18]. Data were visualized using the “ggplot2” package (version 3.3.6) in R (version 4.2.1). As the data resources were entirely collected from internet databases, no ethical committee permission was necessary.

2.2. Screening and identification of differentially expressed ERAGs

Initially, the DEGs between TCGA-BRCA samples and normal breast samples were analyzed using the "DESeq2″ package. An adjusted P value < 0.05 and |log2-fold change| > 2 were confirmed as screened criterion. The volcano plot was made by the “ggplot2” package. An intersection of the DEGs, ER stress-related genes, and apoptosis-related genes were performed by the Venn diagram. Differentially expressed ERAGs (DE-ERAGs) were identified for further analysis. The gene ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis and their visualization presented in bubble chart were performed by the “clusterProfiler” (version 1.36.0), “GOplot” (version 1.0.2), and “ggplot2” packages [19,20]. The Benjamin–Hochberg adjusted P < 0.05 was considered significant screen criteria.

2.3. Construction and assessment of the risk scoring system

To further screen the prognostic value of DE-ERAGs, the Univariate Cox regression analysis with a P value < 0.05 was performed in SPSS Statistics. The LASSO tenfold cross-validation was performed on the DE-ERAGs using the “glmnet” packages (version 4.1.7) [21]. Through the previous analysis, the most significant genes were identified. After that, the acquired genes were incorporated into the risk model, and the standardized gene expression levels and coefficients served as the foundation for the risk scoring system. The risk scoring system was constructed by utilizing the following formula:

$$\text{Risk}\text{score}={\sum }_{i=1}^n{\text{expression}}_{\text{genei}}\times {\text{coefficient}}_{\text{genei}}[21]\text{.}$$

The TMM algorithm was utilized to calculate normalized gene expression levels by the "edgeR" package (version 3.40.2) [18]. Using the "ggplot2″ package, a risk factor plot was generated. The ROC curves analysis was performed and visualized using the “timeROC” package (version 0.4) and “ggplot2” package. The patients were divided into high- and low-risk groups based on the median risk score. Survival analysis was performed and visualized by the “survival” (version 3.3.1), “survminer” (version 0.4.9), and “ggplot2” packages. The different levels of risk scores in the various clinical groups were then visualized.

2.4. Construction and assessment of the nomogram

To assess the potential of the risk scoring system, we conducted univariate and multivariate Cox regression analyses on risk group and various clinicopathological parameters, involving age, pathologic T stage, pathologic N stage, pathologic M stage, PR stage, ER status, and HER2 status. The result of Cox regression analyses was presented by the forest plot using the “ggplot2” package. The "rms" package (version 6.3–0) was used to develop a nomogram that predicted OS probability at 1-, 3- and 5-year using all independent prognostic factors. The nomogram was assessed by ROC, KM survival analysis, and calibration analyses.

2.5. Correlation and verification of the DE-ERAGs

To show the Spearman correlation among DE-ERAGs, the correlation heatmap was generated using the “ggplot2” package. The protein-protein interactions (PPI) network with interaction score >0.15 was constructed in the STRING database (https://string-db.org/) [22]. The differential expression of DE-ERAGs between the normal and BRCA groups in GSE42568 was drawn by the “ggplot2” package. Furthermore, we used RT-qPCR to verify the differential expression of DE-ERAGs between the breast cancer cell line (MDA-MB-231) and the mammary epithelial cell line (MCF-10A) both purchased from Wuhan Procell Corporation (China). MDA-MB-231 cells were cultured at 37 °C, 5% CO2 with Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% foetal bovine serum and 1% penicillin-streptomycin. MCF-10A cells were maintained in complete MCF-10A culture medium (Fenghui Biotechnology, China), which comprised a 1:1 mixture of DMEM and Ham's F12 medium supplemented with 10 μg ml−1 insulin, 0.5 μg ml−1 hydrocortisol, and 20 ng ml−1 epidermal growth factor. The extraction of total RNA was performed using M5 Total RNA Extraction Reagent (Mei5 Biotechnology Co. Ltd, Beijing). Then, the total RNA was reverse-transcribed into cDNA with PrimeScript RT Master Mix (Takara Bio, Shiga, Japan). Finally, the target cDNA samples were amplified in a real-time quantitative PCR system with TB Green® Premix Ex Taq™ II (Takara). The relative expression level was analyzed using the 2−ΔΔCt method, with β-actin mRNA selected as an internal control. The primer sequences of DE-ERAGs are shown in Table S4. Immunohistochemical analysis of DE-ERAGs between BC and normal breast tissue was shown by The Human Protein Atlas database (http://www.proteinatlas.org/).

2.6. Mutation analysis in BRCA

Data on somatic mutations were accessed from the TCGA database using the GDC data portal (https://portal.gdc.cancer.gov/). We downloaded the “Masked Somatic Mutation” data from the four subtypes of data files. For visualization, we prepared the Mutation Annotation Format (MAF) of somatic variants and analyzed using the "maftools" package (Version 2.6.05), which contains several analytic modules [23].

2.7. Immune cell infiltration level analysis

We used CIBERSORTx (https://cibersortx.stanford.edu/) to investigate the composition of infiltrating immune cells between the high- and low-risk groups [24,25]. The gene expression data of BRCA samples was uploaded to CIBERSORTx's online platform. The validated leukocyte gene signature matrix (LM22) was used to identify 22 human hematopoietic cell subsets. We utilized 1000 permutations for the analysis in relative mode. Quantile normalization was disabled as the dataset was generated using RNA-seq. The filter criteria are set as the CIBERSORT calculation of P < 0.05 [26]. The overview of the infiltrated immune cells is shown by the “ggplot2” package.

2.8. Statistical analysis

The study's statistical analyses were all completed by R software (version 4.2.1). KM survival analysis was performed by log-rank tests. The hazard ratios (HRs) and 95% confidence intervals were calculated using Cox proportional hazards regression models. Kruskal-Wallis Test was used for multiple group comparisons and Student's t-test was used for two group comparisons. A two-tailed P value of <0.05 was considered statistically significant. Significance was tested by the Wilcoxon rank sum test.

3. Results

3.1. Identification of DE-ERAGs in BRCA patients

We utilized the DESeq2 method to detect 1813 DEGs between 1113 TCGA-BRCA samples and 113 normal breast samples using the DEGs criterion (Fig. 2A). 1813 identified DEGs, 4767 ER stress-related genes, and 2184 apoptosis-related genes were analyzed by Venn diagram. In this way, we obtained 137 DE-ERAGs in BRCA (Fig. 2B). Additional enrichment analysis was performed to investigate the functions and mechanism of DE-ERAGs. The genes were significantly enriched in response to peptide, mitotic cell cycle phase transition, aging, chromosomal region, secretory granule lumen, cytoplasmic vesicle lumen, receptor ligand activity, signaling receptor activator activity, cytokine activity, IL-17 signaling pathway, Cell cycle, Calcium signaling pathway, AMPK signaling pathway (Fig. 2C). The complete results of GO and KEGG analysis are shown in Table S3.

Fig. 2.

Identification and functional enrichment analysis of endoplasmic reticulum stress and apoptosis-related genes between the TCGA-BRCA cohort and normal breast samples. A Volcano plot of differentially expressed genes in BRCA based on data from TCGA-BRCA cohort. B Venn diagram of the intersection between endoplasmic reticulum stress, apoptosis-related genes and DEGs identified by the DESeq2 algorithm. CTerms of GO enrichment analysis and KEGG pathways related to the 137 endoplasmic reticulum stress and apoptosis-related genes.

3.2. Construction and assessment of the risk scoring system

Initially, the univariate Cox regression analysis was performed to investigate the association between the expression levels of 137 DE-ERAGs and the OS times of patients in the TCGA-BRCA cohort. A total of 25 genes related to OS were identified using the Cox proportional hazards model with a significance threshold of P < 0.05 (Table 1). Then, the gene sets were refined using LASSO regression analysis. (Fig. 3A and B). Sixteen genes were selected as the most significant predictive genes, and subsequently, a risk-scoring system was constructed utilizing the formula. The patients were categorized into high- and low-risk groups based on the calculated median risk score. Furthermore, the analysis of the survival time distribution suggested that higher risk scores were correlated with worse prognosis (Fig. 3C). The areas under the time ROC curves (AUCs) were 0.747, 0.701, and 0.695 for the 1-, 3- and 5-year OS times, respectively (Fig. 3D). The KM analysis showed that the prognosis of the high-risk group was significantly worse than that of the low-risk group (P < 0.001, Fig. 3E). The time ROC curves and KM analysis of the 16 individual genes is shown in Fig. S1 and Fig. S2.

Table 1.

25 ER stress and apoptosis related genes identified by univariate COX regression analysis.

Gene	Description	HR (95% CI)	P value
CASP14	Caspase 14	1.486 (1.070–2.063)	0.018
FOXM1	Forkhead Box M1	1.399 (1.008–1.941)	0.045
PLK1	Polo Like Kinase 1	1.433 (1.035–1.983)	0.03
IVL	Involucrin	1.787 (1.283–2.490)	<0.001
IGF2BP1	Insulin Like Growth Factor 2 MRNA Binding Protein 1	1.411 (1.018–1.956)	0.038
TFF1	Trefoil Factor 1	0.690 (0.496–0.959)	0.027
CDKN2A	Cyclin Dependent Kinase Inhibitor 2A	0.704 (0.506–0.980)	0.038
SLC7A5	Solute Carrier Family 7 Member 5	1.598 (1.149–2.224)	0.005
RAD51	RAD51 Recombinase	1.468 (1.055–2.042)	0.023
CCNB1	Cyclin B1	1.396 (1.007–1.935)	0.045
IL24	Interleukin 24	0.676 (0.487–0.938)	0.019
ABCB1	ATP Binding Cassette Subfamily B Member 1	0.688 (0.496–0.955)	0.026
TF	Transferrin	0.621 (0.448–0.860)	0.004
ESR2	Estrogen Receptor 2	0.669 (0.480–0.932)	0.018
NR3C2	Nuclear Receptor Subfamily 3 Group C Member 2	0.639 (0.458–0.892)	0.008
NSG1	Neuronal Vesicle Trafficking Associated 1	0.663 (0.477–0.920)	0.014
ANXA1	Annexin A1	0.713 (0.513–0.991)	0.044
GFAP	Glial Fibrillary Acidic Protein	0.684 (0.494–0.949)	0.023
TP63	Tumor Protein P63	0.696 (0.503–0.964)	0.029
NTRK2	Neurotrophic Receptor Tyrosine Kinase 2	0.634 (0.457–0.878)	0.006
TACR1	Tachykinin Receptor 1	0.706 (0.510–0.977)	0.036
S100B	S100 Calcium Binding Protein B	0.660 (0.475–0.918)	0.013
PPARG	Peroxisome Proliferator Activated Receptor Gamma	0.716 (0.517–0.992)	0.045
CXCL2	C-X-C Motif Chemokine Ligand 2	0.649 (0.468–0.901)	0.01
CSF3	Colony Stimulating Factor 3	0.682 (0.492–0.946)	0.022
Annotation: HR Hazard Ratio, 95%CI 95% confidence interval

Fig. 3.

Construction and verification of the risk scoring system. A The LASSO regression model of the 25 endoplasmic reticulum stress and apoptosis-related genes performed by Lasso-ten-fold cross-validation. B The coefficient distribution in the LASSO regression model. C The risk score, survival time distributions and gene expression heat map of immune infiltration-related genes in the TCGA-BRCA cohort. D The ROC curves of the risk scoring model predicting OS of 1-, 3- and 5-year in the TCGA-BRCA cohort. E Kaplan–Meier survival analysis of the OS between the risk groups in the TCGA-BRCA cohort.

3.3. Construction and evaluation of the prognostic model

Firstly, we found significant differences in risk scores between different clinical groups (Fig. 4 A–G): age (≤60 vs. > 60, P < 0.001), pathologic T stage (T1 & T2 vs. T3 & T4, P < 0.001), pathologic N stage (N0 & N1 vs. N2 & N3, P < 0.05), pathologic M stage (M0 vs. M1, P < 0.05), PR stage (Negative vs. Positive, P < 0.001), ER status (Negative vs. Positive, P < 0.01), and HER2 status (Negative vs. Positive, P < 0.001). Then, we performed univariate and multivariate Cox regression analyses of predictors, including age, pathologic T stage, pathologic N stage, pathologic M stage, PR stage, ER status, HER2 status, and risk group (Table 2). The results showed that age, pathologic T stage, pathologic N stage, pathologic M stage, ER status, and risk group were independent risk factors for OS in BRCA patients (Fig. 5A). The nomogram model included the above independent predictors (Fig. 5B). The C-index of the nomogram model was 0.751 (0.726–0.776). The nomogram AUCs for the 1-, 3- and 5-year OS rates were 0.807, 0.765, and 0.750, respectively (Fig. 5C). The KM analysis suggested that the prognosis of the nomogram high-risk group was significantly worse (Fig. 5D). Moreover, the calibration plots of the nomogram model demonstrated a high level of agreement for the 1-, 3- and 5-year OS rates compared with the ideal model (Fig. 5E).

Fig. 4.

The different risk scores in the different clinical groups. A Age (≤60 vs > 60), B Pathologic T stage (T1&T2 vs T3&T4), C Pathologic N stage (N0&N1 vs N2&N3), D Pathologic M stage (M0 vs M1), E PR stage (Negative vs Positive), F ER status (Negative vs Positive). G HER2 status (Negative vs Positive). *P < 0.05, **P < 0.01, ***P < 0.001. Wilcoxon signed rank test was applied.

Table 2.

Univariate and multivariate Cox regression analysis in the TCGA-BRCA cohort.

Characteristics	Total(N)	Univariate analysis		Multivariate analysis
		Hazard ratio (95% CI)	P value	Hazard ratio (95% CI)	P value
Age	1051		< 0.001
≤60	586	Reference		Reference
>60	465	2.036 (1.468–2.822)	< 0.001	2.691 (1.643–4.406)	< 0.001
Pathologic T stage	1048		0.010
T1&T2	882	Reference		Reference
T3&T4	166	1.673 (1.152–2.429)	0.007	3.137 (1.697–5.799)	< 0.001
Pathologic N stage	1032		< 0.001
N0&N1	843	Reference		Reference
N2&N3	189	2.248 (1.525–3.313)	< 0.001	1.836 (1.014–3.323)	0.045
Pathologic M stage	890		< 0.001
M0	870	Reference		Reference
M1	20	4.315 (2.501–7.444)	< 0.001	4.191 (1.737–10.110)	0.001
PR status	1000		0.125
Negative	334	Reference
Positive	666	0.762 (0.541–1.074)	0.120
ER status	1003		0.068
Negative	237	Reference		Reference
Positive	766	0.704 (0.487–1.017)	0.062	0.487 (0.289–0.821)	0.007
HER2 status	695		0.069
Negative	542	Reference		Reference
Positive	153	1.611 (0.981–2.644)	0.059	0.915 (0.522–1.604)	0.756
Riskgroup	1055		< 0.001
Low	527	Reference		Reference
High	528	2.603 (1.840–3.683)	< 0.001	2.708 (1.565–4.688)	< 0.001

Annotation: HR Hazard Ratio, 95%CI 95% confidence interval.

Fig. 5.

Construction and evaluation of the prognostic model. A The forest plot of risk factors using multivariate Cox regression analysis. B The independent risk factors that affect the OS of BRCA patients screened by multiple Cox regression were incorporated into the nomogram model. C The ROC curves for predicting the nomogram of 1-, 3- and 5-year OS in the TCGA-BRCA cohort. D The nomogram calibration curves for predicting the 1-, 3- and 5-year OS of the TCGA-BRCA cohort.

3.4. Correlation and verification of the DE-ERAGs

The Spearman correlation among 16 DE-ERAGs was shown in the correlation heatmap (Fig. 6A). The PPI network of 16 DE-ERAGs was exhibited in Fig. 6B. The gene expression of the DE-ERAGs in the data set GSE42568 of the GEO database was shown in Fig. 6C. Compared with normal samples, FOXM1, PLK1, IVL, TFF1, and SLC7A5 were upregulated and TF, NR3C2 and GFAP were downregulated in BC samples of GSE42568, while others have no significance. This result could be related to the limited sample size. Based on the results of PCR experiments, FOXM1, PLK1, TFF1, ESR2, NR3C2 were upregulated in MDA-MB-231, while IVL, IGF2BP1, SLC7A5, IL24, ABCB1, TF, NSG1, GFAP, TP63, S100B were downregulated (Fig. 6D). CDKN2A primers cannot be designed for PCR amplification due to the high G/C content of the gene sequence. Moreover, the immunohistochemical data of DE-ERAGs between breast cancer and normal breast tissue were shown by The Human Protein Atlas database (https://www.proteinatlas.org/) (Fig. 7A–O). There is no protein expression of NSG1 in breast tissue and breast cancer according to the human protein atlas data.

Fig. 6.

The correlation and verification of the 16 endoplasmic reticulum stress and apoptosis-related genes. A Spearman correlation among 16 endoplasmic reticulum stress and apoptosis-related genes. B The PPI network constructed by using STRING database. C The differential expression of 16 genes between the normal and BRCA group in GSE42568. D RT-qPCR validation of ERAGs expression between the breast cancer cell line (MDA-MB-231) and the mammary epithelial cell line (MCF-10A). ***: P < 0.001, **: P < 0.01, *: P < 0.05, ns: not significant. Student's t-test was applied.

Fig. 7.

Immunohistochemical analysis of BRCA and normal breast tissue determined by HPA database. A FOXM1; B PLK1; C IVL; D IGF2BP1; E TFF1; F CDKN2A; G SLC7A5; H IL24; I ABCB1; J TF; K ESR2; L NR3C2; M GFAP; N TP63; O S100B.

3.5. Mutation analysis in BRCA

The somatic mutation profiles of 969 BRCA patients were acquired from TCGA. We used the “maftools” package to analyze and visualize the results. The mutation information was exhibited in a waterfall plot. The different mutation types were represented by different colors with annotations at the bottom (Fig. 8A). In Fig. 8B, the coincident and exclusive correlations among mutant genes were shown. Green denoted co-occurrence, while red denoted the mutually exclusive relationships. These mutations were further categorized, with missense mutation accounting for the majority of those categories (Fig. 8C), the frequency of single nucleotide polymorphism (SNP) was higher than that of insertion (INS) (Fig. 8D), and the single nucleotide variation (SNV) in BRCA with the highest frequency was C > T (Fig. 8E). We calculated the number of changed bases in each sample and colored different types of mutation (Fig. 8F and G). Finally, we displayed the top 10 BRCA mutations with sorted percentages, including TF (13%), ABCB1 (13%), NR3C2 (11%), FOXM1 (8%), ESR2 (5%), GFAP (6%), IVL (5%), IGF2BP1 (5%), PLK1 (3%) and IL24 (3%) (Fig. 8H).

Fig. 8.

Summary of the mutation information with statistical calculations. A Landscape of mutation profiles in TCGA-BRCA samples. Mutation information of each gene in each sample was shown in the waterfall plot, in which various colors with annotations at the bottom represented the different mutation types. The barplot above the legend exhibited the mutation burden. B The coincident and exclusive associations across mutated genes. C, D, E Classification of mutation types according to different categories, in which missense mutation accounts for the most fraction, SNP showed more frequency than insertion, and C > T was the most common of SNV; F, G tumor mutation burden in specific samples; H the top 10 mutated genes in BRCA. SNP, single nucleotide polymorphism; SNV, single nucleotide variants. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3.6. Immune cell infiltration level analysis

We analyzed the levels of immune cell infiltration in the TCGA-BRCA cohort by CIBERSORTx. According to the results, the high-risk group had lower levels of partial immune cell infiltration, such as B cells naïve (P < 0.001), Plasma cells (P < 0.001), T cells CD4 memory resting (P < 0.001), dendritic cells resting (P < 0.001), and mast cells resting (P < 0.01). The other part of immune cell infiltration levels in the high-risk group were higher, such as B cells memory (P < 0.01), T cells regulatory (Tregs) (P < 0.05), Macrophages M0 (P < 0.001), and Macrophages M2 (P < 0.001) (Fig. 9).

Fig. 9.

Analysis of immune cell infiltration in TCGA-BRCA cohort. The box plot showed the levels of immune cell infiltration between the high-risk group and low-risk group in BRCA patients. *: P < 0.05, **: P < 0.01, ***:P < 0.001, ns: not significant. Wilcoxon signed rank test was applied.

4. Discussion

According to current research, traditional prognostic systems have limitations in reflecting the biological heterogeneity of breast cancer [27]. Novel prognostic and predictive models have been developed due to bioinformatics advancements. Continuous ER stress is a newly identified cancer hallmark [5]. ER stress with concomitant ferroptosis, ER stress leading to autophagy, and the combinatorial effects of oxidative and ER stress could change tumor cell homeostasis to induce apoptosis [6]. Further research is required to explore the prognostic value and molecular mechanism of ERAGs in BRCA.

In our study, we obtained gene expression data and clinical information from the TCGA database. Following the DEG screening process and the intersection analysis with ERAGs, 137 DE-ERAGs were identified and selected. Subsequently, 16 genes were identified through univariate Cox and LASSO regression analyses. We utilized the 16 DE-ERAGs to calculate risk scores. The findings demonstrated that the risk score model could properly predict the OS of BRCA patients. Using univariate and multivariate Cox regression analysis, we incorporated risk group, age, pathologic T stage, pathologic N stage, pathologic M stage, and ER status into the nomogram model. The ROC, KM analysis, and calibration plots revealed that the OS nomogram had a reliable predictive value. Furthermore, the functional enrichment analysis, correlation, and mutation information of ERAGs provided evidence for the underlying interaction mechanism study.

Multiple studies have demonstrated that ERAGs are involved in the pathological process of breast cancer. FOXM1 serves as a transcriptional target of ERα, facilitating its mitogenic activities and exhibiting involvement in the modulation of endocrine sensitivity and resistance in breast cancer. The risk of relapse in ER + breast cancer is higher than in ER-breast cancer, indicated by the expression of the FOXM1 gene in the initial tumor [28]. PLK1, as an enzyme, plays a crucial role in regulating various aspects of the mitotic phase. The use of PLK1 siRNAs in orthotopic breast cancer models has been observed to substantially impair the development of human breast cancer and induce apoptosis in breast cancer cell lines [29]. The overexpression of Plk1 in patients has exhibited a positive correlation with improved survival in specific subtypes of breast cancer [30]. Studies have indicated that approximately 25% of breast tumors express IVL, suggesting a potential association between differentiation processes and breast cancer [31].

According to a study, the expression of TFF1 was significantly higher in 34.6% of patients with metastatic breast cancer compared to 0% of nonmetastatic (P = 0.002) [32]. The study of Wang et al. [33] found that a significant proportion of patients (43.3%) diagnosed with bone metastases demonstrated a notable upregulation of TFF1 expression. ABCB1 plays a pivotal role in the chemoresistance in BC [34]. Iron has been identified as a precipitating factor in the proliferation of breast cancer cells, which might lead to metastasis. Endocytosis of transferrin receptors with bound iron-carrying transferrin (TF) is the principal way of iron delivery into cancer cells [35]. A study found that higher NR3C2 expression was related with longer OS, progression free interval (PFI), and disease specific survival (DSS) in BC patients [36]. The role of NSG1 has been explored in lung adenocarcinoma [37] and esophageal squamous cell carcinoma [38], this study initially discovered its function and mechanism in breast cancer. Darlix A. et al. [39] found that GFAP is a predictor of brain metastases, and elevated serum GFAP levels were independently associated with poor outcomes. The results of the above research support the findings of this study, emphasizing the role of ERAGs in breast cancer.

Based on the KEGG analysis, ERAGs were significantly enriched in IL-17 signaling pathway, p53 signaling pathway, calcium signaling pathway, PI3K-Akt signaling pathway, AMPK signaling pathway, and so on. One study suggested that Twist1, a transcription factor (TF) related to epithelial-mesenchymal transition (EMT), plays a role on IL-17 signaling in HER2+ BC [40]. Camilla T. et al. [41] found that missense mutant p53 oncoproteins increase the synthesis of de novo serine/glycine synthesis and the intake of essential amino acids, which accelerates the development of breast cancer. In breast cancer, flubendazole increases the inhibitory effect of paclitaxel through the HIF1α/PI3K/AKT signaling pathways [42]. The reliability of our findings has been confirmed by these investigations. Nevertheless, further research is still needed to identify the specific mechanism in BRCA.

The findings of systematic reviews have demonstrated a positive correlation between higher levels of tumor-infiltrating B cells or CD8⁺ T-cell infiltration and extended overall survival [43,44]. The decrease in CD4 (+) lymphocytes was associated with reduced OS among patients with metastatic breast cancer [45]. Giorello M B et al. [46] found that patients with high dendritic cell numbers had a lower risk of metastatic occurrence. According to a study, tumor-associated macrophages promote the development and invasion of cancer cells inside the tumor microenvironment in breast cancer, particularly when evasion of immunoreactions due to T and B tumor-infiltrating lymphocytes occurs in breast cancer [47]. The studies mentioned above have effectively validated the dependability of our findings.

However, it is necessary to recognize the limits of our research. First, we obtained the datasets to build and verify the ERAGs-related prognostic model from TCGA and GEO database. Further data collection and rigorous verification through a substantial sample of prospective research is still required. Second, further mechanistic investigations were needed to be carry out in laboratory settings.

5. Conclusion

In this study, we developed and verified a risk scoring system that utilizes DE-ERAGs for prognostic evaluation and risk categorization of BC patients. A nomogram model was developed to predict the 1-, 3- and 5-year OS, and it demonstrated favorable predictive accuracy. The 16 ERAGs in the risk score system might become potential targets for investigating the biological mechanisms of BC. The possibly significant genes were subsequently screened for further studies by correlation analysis, expression validation, and mutation analysis. In addition, GOKEGG and tumor immune infiltration analysis suggested that ERAGs are involved in the biological processes of BC. All these results might provide novel ideas for BC research. Our study systematically evaluated the prognostic values and potential molecular mechanisms related to ERAGs in BRCA, offering novel insights and methods for investigating ER stress and apoptosis in BRCA.

Data availability statement

The data associated with our study has been deposited into the publicly available repository. The datasets are available in the TCGA (https://portal.gdc.cancer.gov/) and GEO (https://www.ncbi.nlm.nih.gov/geo/). Further inquiries can be directed to the corresponding authors.

Ethics declarations

Review and approval by an ethics committee was not needed for this study because the data resources were entirely collected from internet databases.

Funding

This work was supported by the Natural Science Foundation of Shanxi Province [201901D111373] and the Science and Technology Innovation Program of Higher Education Institutions in Shanxi Province [2021L196].

CRediT authorship contribution statement

Hao Fan: Writing – review & editing, Writing – original draft, Validation, Investigation, Formal analysis, Data curation, Conceptualization. Mingjie Dong: Writing – review & editing, Validation, Data curation. Chaomin Ren: Writing – review & editing, Validation, Investigation. Pengfei Shao: Writing – review & editing, Visualization. Yu Gao: Writing – review & editing, Investigation. Yushan Wang: Writing – review & editing, Funding acquisition, Data curation. Yi Feng: Writing – review & editing, Supervision, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article.

Table S1

The list of endoplasmic reticulum stress related genes from the GeneCards database.

Table S2

The list of apoptosis related genes from the GeneCards database.

Table S3

The complete results of GO and KEGG analysis.

Table S4

The primer sequences of DE-ERAGs.

Fig. S1.

Time ROC analysis of the 16 DE-ERAGs. A PLK1, B IVL, C IGF2BP1, D SLC7A5, E CDKN2A, F GFAP, G FOXM1, H TF, I ESR2, J IL24, K NSG1, L TFF1, M ABCB1, N NR3C2, O TP63, P S100B.

Fig. S2.

KM survival curves of the 16 DE-ERAGs. A ABCB1, B FOXM1, C NR3C2, D SLC7A5, E CDKN2A, F IGF2BP1, G NSG1, H TF, I ESR2, J IL24, K PLK1, L TFF1, M FOXM1, N IVL, O S100B, P TP63.