The GSE21422 dataset (GPL570 platform, expression profiling by array) was downloaded from the Gene Expression Omnibus (ncbi.nlm.nih.gov/geo/), and includes 5 healthy breast and 14 BRCA samples (ductal carcinoma in situ and invasive ductal breast carcinoma) (13). Gene expression data from TCGA-BRCA cohort were downloaded from the UCSC Xena Browser (https://xenabrowser.net/). Gene expression RNA-seq data in HTSeq-FPKM format were obtained from TCGA-BRCA dataset. TCGA-BRCA dataset encompasses all molecular subtypes of BRCA, including the triple-negative BRCA (TNBC), hormone receptor-positive and HER2-enriched subtypes. Using the GEO2R tool (https://www.ncbi.nlm.nih.gov/geo/geo2r/) and ‘limma’ R package (version 3.52.4) in R software (version 4.2.2), DEGs were analyzed on the basis of log₂[fold change (FC)]>1 for upregulation and <-1 for downregulation, with statistical significance (P<0.05). The GSE21422-DEGs and TCGA-BRCA-DEGs were combined and the overlapping genes were screened via a Venn diagram, which were then uploaded to the Database for Annotation, Visualization, and Integrated Discovery (https://david.ncifcrf.gov/tools.jsp) to conduct Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and Gene Ontology [GO; including Cell Component (CC), Biological Process (BP) and Molecular Function (MF)] analysis. P<0.05 was considered to indicate a statistically significant difference. To identify key genes among the DEGs, a protein-protein interaction (PPI) network was constructed using the STRING database (string-db.org/) with a minimum required interaction score of 0.4. The resulting network was imported into Cytoscape software (version 3.9.1), and the CytoHubba plugin was applied to analyze network topology. The BottleNeck algorithm within CytoHubba was used to rank genes based on their network importance.

LASSO analysis was performed via the ‘glmnet’ package in R software. This model was applied to the overlapping genes identified in the aforementioned analysis. The coefficients of the signature genes were determined at λ_min, which was obtained by examining the association between the partial likelihood deviation and log(λ). The samples from TCGA-BRCA dataset were divided into high-risk and low-risk groups based on the median value (−0.0085) of the calculated risk scores derived from the expression levels of the overlapping genes. A risk score model was then designed to compare the survival status of patients with BRCA and the differences in the expression of signature genes between the two groups. Kaplan-Meier (KM) survival curves for progression-free survival (PFS) were generated using the ‘survival’ package. The hazard ratio (HR), confidence interval (CI) and log rank P-value were calculated to compute the differences in survival outcomes between the high-risk and low-risk groups. To assess the prognostic value of the risk model, receiver operating characteristic (ROC) curves were established using the ‘timeROC’ package, specifically for the 1-, 3- and 5-year time points. The predictive accuracy of the model was quantified by calculating the area under the curve (AUC) values.

GSCA is a computational method used to assess the variation in gene set activity across different samples in gene expression data. In the present study, the GSCA database (https://guolab.wchscu.cn/GSCA/#/) was used to perform gene mutation analysis on signature genes, including single nucleotide variants (SNVs), copy number variations (CNVs), variant types, variant classifications, variant classification summaries and variants per sample.

Expression, ROC and KM survival analyses were also conducted to evaluate the individual clinical relevance of each signature gene. Unlike the risk model-based analyses described above, which assessed the combined prognostic value of the entire gene signature using a composite risk score, these analyses focused on the diagnostic and prognostic value of each gene independently. Specifically, the signature gene expression levels in the TCGA-BRCA tumor and normal groups were observed. ROC analysis was subsequently used to evaluate the diagnostic value of the signature genes in BRCA. ROC analysis was performed using the ‘pROC’ package. The larger the AUC value, the greater the diagnostic value. KM survival curves were constructed to visualize the differences in survival outcomes. KM survival analysis was conducted using the ‘survminer’ package in R. The log-rank test was used to estimate the significant difference between the groups, and HRs were calculated to estimate the risk associated with the gene expression levels. Finally, the overall survival (OS) probability of genes with diagnostic value was evaluated.

Univariate and multivariate Cox regression analyses were performed on signature genes that showed significant association with OS together with clinical factors, including age, pT stage and pTNM stage, to identify the 5 candidate genes linked to BRCA survival. These analyses were conducted via the ‘forestplot’ package, and prognostic genes with a significant association (P<0.05) were identified. A prognostic nomogram incorporating the significant prognostic variables was subsequently designed via the ‘rms’ package. The nomogram aimed to predict the 1-, 3- and 5-year survival probabilities of patients with BRCA with identified prognostic factors. Additionally, a calibration curve was used to estimate the performance and accuracy of the survival nomogram model.

The Human Protein Atlas (HPA; proteinatlas.org/) is a comprehensive and publicly available resource that provides information about the expression and localization of proteins in various human tissues and cells (14). In the present study, the protein expression of Rac/Cdc42 guanine nucleotide exchange factor 6 (ARHGEF6) in BRCA was assessed. Data from the University of Alabama at Birmingham Cancer Data Analysis Portal (https://ualcan.path.uab.edu/index.html) was used to conduct clinical feature analysis on the basis of individual cancer stage, age, nodal metastasis status and TP53 mutation status. The Genomics of Drug Sensitivity in Cancer database (cancerrxgene.org/) was used to analyze the relationship between ARHGEF6 expression levels and the IC₅₀ of paclitaxel in BRCA. We collected genes involved in apoptosis and performed single-sample gene set enrichment analysis (ssGSEA) using the GSVA package (version 3.21) in R with the method set to ‘ssgsea’. Finally, Spearman correlation analysis was used to assess the relationship between gene expression and pathway scores.

The BRCA cell lines, Hs578t (cat. no. QS-H157; Keycell Biotechnology Co., Ltd.) and MDA-MB-231 (cat. no. BFN608008564; The Cell Bank of Type Culture Collection of the Chinese Academy of Sciences), were subjected to short tandem repeat analysis and mycoplasma testing. These cell lines were stored in liquid nitrogen at −80°C and subsequently cultured in Roswell Park Memorial Institute 1640 medium (cat. no. E600028; Sangon Biotech Co., Ltd.) supplemented with 10% fetal bovine serum (FBS; cat. no. C0235; Beyotime Institute of Biotechnology) and penicillin-streptomycin (cat. no. 60162ES76; Shanghai Yeasen Biotechnology Co., Ltd.). For subsequent studies, the cells were cultured at 37°C in a humidified environment with 5% CO₂.

The ARHGEF6 overexpression plasmid [pRP(Exp)-CMV>hARHGEF6; NM_004840.3] and a negative control [pRP(Exp)-CMV>ORF_Stuffer) were constructed by VectorBuilder, Inc. Lipofectamine 2000 (cat. no. 11668019; Invitrogen; Thermo Fisher Scientific, Inc.) was used for cell transfection according to the manufacturer's instructions. Briefly, 2 µg of plasmid was transfected into cells cultured at 37°C in a humidified incubator with 5% CO₂. The transfection mixture was incubated with cells for 6 h at 37°C, after which the medium was replaced with fresh complete medium. Subsequent experiments were performed 48 h post-transfection.

After cell transfection, total RNA was isolated from cells with TRIzol reagent (cat. no. R0016; Beyotime Institute of Biotechnology), followed by RT of the RNA into complementary DNA with a PrimeScript RT kit (cat. no. 04379012001; Roche Diagnostics) according to the manufacturer's instructions. qPCR was conducted with a SYBR Green PCR Mix Kit (cat. no. 4344463; Thermo Fisher Scientific, Inc.). The reaction conditions were as follows: Initial denaturation at 95°C for 5 min, followed by 40 cycles of denaturation at 95°C for 5 sec, annealing at 60°C for 30 sec and extension at 74°C for 30 sec. GAPDH served as the internal control. The 2^−ΔΔCq method was used to determine the relative expression levels of the target genes (15). The primers used were as follows: ARHGEF6 forward, 5′-TCCAGGAATGGTTGGAGCAG-3′ and reverse, 5′-GGCTGTCCGGTAGAGCTAAA-3′; BCL2 forward, 5′-AAAAATACAACATCACAGAGGAAGT-3′ and reverse, 5′-GTTTCCCCCTTGGCATGAGA-3′; caspase3 forward, 5′-TGAGGCGGTTGTAGAAGAGTTTC-3′ and reverse, 5′-TTATTAACGAAAACCAGAGCGCC-3′; p53 forward, 5′-AAGTCTAGAGCCACCGTCCA-3′ and reverse, 5′-GACGCTAGGATCTGACTGCG-3′; GAPDH forward, 5′-AATGGGCAGCCGTTAGGAAA-3′ and reverse, 5′-GCGCCCAATACGACCAAATC-3′.

Cell lysis was performed using RIPA lysis buffer (cat. no. P0013; Beyotime Institute of Biotechnology) and total protein was extracted. Then, 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (cat. no. P0690; Beyotime Institute of Biotechnology) was used to separate the proteins (25 µg/lane), before the proteins were transferred to PVDF membranes (cat. no. FFP24; Beyotime Institute of Biotechnology). The membranes were then blocked for 2 h at room temperature with 5% non-fat milk. Primary antibodies were added to the membranes and incubated overnight at 4°C. Antibodies were as follows: anti-ARHGEF6 (1:1,000, ab184569, Abcam), anti-BCL2 (1:1,000, ab182858, Abcam, UK), anti-p53 (1:1,000, ab32509, Abcam, UK), anti-caspase-3 (1:1,000, ab32351, Abcam, UK), anti-c-caspase-3 (1:1,000, ab32042, Abcam, UK) and anti-GAPDH (1:3,000, ab9485, Abcam, UK). Then the membranes were incubated with goat Anti-Rabbit IgG H&L (HRP) (1:5,000, ab6721, Abcam, UK) for 4 h at room temperature. GAPDH served as the internal control, and protein expression was visualized with an enhanced chemiluminescence kit (cat. no. KGC4603; KeyGene).

Cell proliferation was measured using Cell Counting Kit-8 (CCK-8; cat. no. C0038; Beyotime Institute of Biotechnology) assay. After transfection for 48 h, the cells were seeded into 96-well plates at a density of 2×10³ cells per well. Then, 10 µl of CCK-8 solution was added to each well at various time intervals (0, 24, 48, 72 and 96 h) to assess the viability and proliferation of the cells, followed by incubation for 1 h at 37°C. A microplate reader (Bio-Rad Laboratories, Inc.) was then used to measure the absorbance of the cells at 450 nm.

For the invasion experiment, the transfected cells (5×10⁴) were plated in the upper chamber of a Transwell insert covered with Matrigel (cat. no. 356234; BD Biosciences), which had been incubated at 37°C for 1 h before cell seeding. The cells were cultured in serum-free media in the upper chamber. The lower chamber was filled with media containing 10% FBS. The cells were incubated at 37°C for 24 h. The invading cells were fixed with 4% paraformaldehyde (cat. no. P0099; Beyotime Institute of Biotechnology) and labeled with 4′,6-diamidino-2-phenylindole (cat. no. C1006; Beyotime Institute of Biotechnology). Randomly chosen fields of view were used to study the cells under a microscope. Same procedures were followed for the cell migration experiment, except that Matrigel was not applied to the Transwell insert.

Transfected cells were harvested using trypsin (cat. no. 40101ES25; Shanghai Yeasen Biotechnology Co., Ltd.). The cell suspension was then centrifuged at 300 × g for 5 min at room temperature to obtain a pellet. The pellet was subsequently resuspended in phosphate-buffered saline (cat. no. C0221A; Beyotime Institute of Biotechnology). The resuspended cells were stained with an Annexin V-fluorescein isothiocyanate (FITC)/propidium iodide (PI) (cat. no. C1383L; Beyotime Institute of Biotechnology) staining kit according to the manufacturer's instructions. Following incubation, the stained cells were analyzed by a flow cytometer (BD Biosciences). Data were analyzed with FlowJo software (version 10.8.1; BD Biosciences).

All cellular experiments were performed in triplicate, and data are expressed as mean ± standard deviation. Statistical analyses were conducted using GraphPad Prism 8.0 software (Dotmatics). Comparisons between two groups were performed using unpaired independent samples t-test. P<0.05 was considered to indicate a statistically significant difference.

The overlapping DEGs were identified by integrating data from the GSE21422 dataset and TCGA-BRCA cohort. Using the GSE21422 dataset, 1,081 downregulated and 973 upregulated DEGs were identified (Fig. 1A). From TCGA-BRCA samples, 1,498 downregulated and 903 upregulated DEGs were obtained (Fig. 1B). The intersection of these datasets yielded overlapping DEGs (Fig. 1C). To explore the biological significance of these overlapping DEGs, GO enrichment analysis was performed. In the BP category, the overlapping DEGs were significantly enriched in processes such as the ‘Regulation of angiogenesis’, ‘External encapsulating structure organization’ and ‘Regulation of cell population proliferation’ (Fig. 1D). For CC, the genes were predominantly associated with the ‘Collagen-containing extracellular matrix’, ‘Cell-cell junction’ and ‘Spindle’ (Fig. 1D). In terms of MF, significant enrichment was observed for ‘Metal ion binding’, ‘Androsterone dehydrogenase activity’, ‘ErbB-3 class receptor binding’, ‘CXCR3 chemokine receptor binding’ and ‘NAD-retinol dehydrogenase activity’ (Fig. 1D). KEGG pathway analysis further revealed that the overlapping DEGs were involved in several critical pathways, including those related to the Regulation of lipolysis in adipocytes, PPAR signaling pathway, ECM-receptor interaction, AMPK signaling pathway and Pyruvate metabolism (Fig. 1E). To identify potential hub genes, a protein-protein interaction network was constructed utilizing the BottleNeck algorithm, which pinpointed 30 key interactive genes (Fig. 1F).

Subsequent LASSO Cox regression analysis of the 30 previously identified interactive genes led to the selection of 8 prognostic signature genes on the basis of the λ_min value (Fig. 2A and B). The risk model analysis (Fig. 2C) demonstrated that patients in the high-risk group had a higher mortality rate and a lower survival rate than did those in the low-risk group. Notably, the expression levels of the 8 signature genes, fibronectin (FN1), IL4I1, PIK3R1, integrin α1 (ITGA1), inhibitor of DNA binding 1 (ID1), aldehyde dehydrogenase 1 family member A1 (ALDH1A1), ARHGEF6 and enhancer of zeste 2 polycomb repressive complex 2 subunit (EZH2), were significantly different between the two groups. KM survival curve analysis for PFS revealed that patients in the high-risk group had significantly poorer prognoses (log-rank P=0.000363; 95% CI, 1.312-2.543). The calculated HR between the groups was 1.826 (>1), underscoring the utility of the risk model as an effective prognostic indicator (Fig. 2D). Furthermore, ROC curve analysis (Fig. 2E) indicated that the risk model exhibited strong predictive performance (AUC >0.6) for survival outcomes at 1, 3 and 5 years, with the highest accuracy observed at the 1-year time point (AUC=0.699). These results underscore the prognostic value of the identified signature genes and the robustness of the developed risk model in predicting survival outcomes for the studied population.

Next, a comprehensive analysis of gene mutations in the signature genes in BRCA was performed, which revealed significant genomic alterations across multiple samples, providing valuable insights into the landscape of somatic mutations in cancer. CNV analysis revealed diverse CNV percentages among the 8 genes in patients with BRCA (Fig. 3A). Notably, ID1 exhibited the highest percentage of heterozygous amplification, whereas PIK3R1 showed the highest percentage of heterozygous deletion. By contrast, nearly all the genes displayed a low percentage of homozygous amplification. The SNV heatmap further revealed that PIK3R1 had the highest mutation frequency in BRCA samples, followed by FN1 and ITGA1 (Fig. 3B). Among the mutation types, missense mutations were the most prevalent, with single nucleotide polymorphisms identified as the primary variant type. Additionally, C>T transitions were the most common substitution pattern in the SNV class, with PIK3R1 emerging as the most frequently mutated gene (Fig. 3C and D).

The expression levels of the 8 signature genes in the normal and BRCA tumor groups were further analyzed. The results revealed that EZH2, IL4I1 and FN1 were upregulated in the tumor group, whereas the remaining genes were downregulated (Fig. 4A). ROC curve analysis demonstrated that ALDH1A1, ARHGEF6, EZH2, ITGA1 and PIK3R1 exhibited high diagnostic value for BRCA (Fig. 4B-I). KM survival analysis further revealed the prognostic significance of these genes. Specifically, high expression levels of ALDH1A1, ARHGEF6, ITGA1 and PIK3R1 were associated with improved OS, whereas high EZH2 expression was associated with a poorer OS (Fig. 4J-N). These findings highlight ALDH1A1, ARHGEF6, EZH2, ITGA1 and PIK3R1 as potential diagnostic biomarkers for patients with BRCA, with prognostic implications.

Univariate and multivariate Cox regression analyses were performed to estimate the prognostic significance of the candidate prognostic genes (Fig. 5A and B). In univariate analysis, ARHGEF6 was significantly associated with better overall survival, while advanced pT-stage and pTNM-stage predicted worse outcomes (Fig. 5A). In multivariate analysis, ARHGEF6 remained an independent protective factor, whereas ITGA1 and pTNM-stage were independent risk factors (Fig. 5B). To facilitate accurate prognosis prediction for patients with BRCA, a nomogram was subsequently created on the basis of the independent prognostic factors identified via multivariate Cox regression analysis. This nomogram integrated multiple clinical characteristics, including ALDH1A1, ARHGEF6, ITGA1 and the pTNM stage (Fig. 5C), which were evaluated via a calibration curve. These curves were generated to assess the agreement between the predicted survival probabilities from the nomogram and the actual observed survival rates at 1, 3 and 5 years. The results in Fig. 5D show that the predicted results closely corresponded to the actual results, which indicated that this nomogram had high predictive accuracy. On the basis of these findings, ARHGEF6 was targeted as the key gene in the present study.

Analysis using the HPA database revealed that the ARHGEF6 protein levels were significantly lower in BRCA tumor tissues than in normal tissues (Fig. 6A). Higher ARHGEF6 expression was associated with lower IC₅₀ values in the paclitaxel sensitivity analysis, suggesting its potential role in guiding personalized therapeutic strategies (Fig. 6B). The downregulation of ARHGEF6 was observed across various clinical subgroups, including different cancer stages (Fig. 6C), age groups (Fig. 6D), nodal metastasis statuses (Fig. 6E) and TP53 mutation statuses (Fig. 6F). However, the differences in ARHGEF6 expression among these clinical subgroups were relatively minor. ARHGEF6 expression across molecular subtypes of BRCA was further compared and it was found that ARHGEF6 was downregulated in the Luminal, HER2-positive and TNBC subtypes compared with normal tissues (Fig. 6G). Compared with the Luminal subtype, the ARHGEF6 levels were significantly lower in both the HER2-positive and TNBC groups (Fig 6G). Although ARHGEF6 expression was slightly higher in TNBC than in the HER2-positive group, the difference was not statistically significant (Fig 6G). Although both HER2-positive and TNBC subtypes showed reduced ARHGEF6 expression compared with the normal breast tissue and Luminal subtypes, TNBC was selected for functional validation due to its poor prognosis and lack of targeted therapies.

For the cellular experiments, ARHGEF6 was first overexpressed in Hs578t and MDA-MB-231 cells, as confirmed by PCR and western blotting assays (Fig. 7A and B). Subsequent functional experiments demonstrated that ARHGEF6 overexpression significantly suppressed the proliferation (Fig. 7C and D), invasion and migration (Fig. 7E and F) of BRCA cell lines. Furthermore, ARHGEF6 was found to be positively correlated with apoptosis (Fig. 8A), a finding further validated by flow cytometry (Fig. 8B), PCR and western blotting analyses. Mechanistically, ARHGEF6 overexpression downregulated BCL2 while upregulating caspase-3 and p53 at both the mRNA and protein levels, along with increased cleaved caspase-3 protein expression, thereby promoting apoptosis in BRCA cells (Fig. 8C-F). Collectively, these results highlight ARHGEF6 as a tumor suppressor gene in BRCA that regulates cell growth, invasion, migration and apoptosis.