RGS Family Remodeling in Breast Cancer: A Meta‐Analysis and Multi‐Omics Profiling of Prognostic Biomarkers

ABSTRACT

Background

Breast cancer (BC) is a heterogeneous malignancy with diverse molecular subtypes and variable clinical outcomes. Despite diagnostic and therapeutic advances, recurrence and metastasis contribute to poor prognosis in subsets of patients. Regulators of G protein signaling (RGS) proteins, negative modulators of G protein‐coupled receptor (GPCR) pathways, influence tumor progression, but their expression profiles, genomic alterations, immune associations, and prognostic roles in BC remain incompletely understood. This study systematically investigated the RGS family to identify potential biomarkers and therapeutic targets.

Methods

Transcriptomic and clinical data from TCGA and seven independent GEO datasets were evaluated. A random‐effects meta‐analysis established cross‐cohort expression consensus. Diagnostic value was assessed via ROC curve analysis. A prognostic signature was constructed using LASSO and multivariate Cox regression. Genomic alterations, DNA methylation, immune mapping, and pharmacogenomic profiling (DepMap/Broad Institute) were comprehensively analyzed.

Results

Meta‐analysis identified eight robustly dysregulated RGS genes across BC cohorts. A combined 8‐gene panel demonstrated diagnostic accuracy (AUC = 0.98). Furthermore, a LASSO‐derived 6‐gene signature successfully stratified patients into high‐ and low‐risk prognostic groups (p < 0.0001). Immune infiltration profiling, validated by scRNA‐seq, confirmed that RGS18 expression is robustly correlated with immune cells and originates predominantly from the tumor microenvironment rather than malignant cells.

Conclusions

This study identifies the RGS gene family as a multidimensional framework for BC stratification. Through meta‐analysis, we confirmed that RGS3 and RGS4 act as independent oncogenic risk factors, while RGS1 and RGS18 serve as key immunoregulatory biomarkers. We established a high‐accuracy 8‐gene diagnostic panel (AUC = 0.98) and a 6‐gene prognostic signature (RGS1, 2, 3, 10, 16, 19) that independently predicts patient survival. Our findings reveal that these dysregulations are driven by genomic amplifications and CpG methylation. These results position the RGS family as robust clinical biomarkers and actionable targets for precision oncology.

Integrated multi‐omics analyses revealed significant dysregulation of RGS family genes in breast cancer and identified robust diagnostic and prognostic biomarkers. RGS expression was associated with genomic and epigenetic alterations, immune microenvironment features, and drug response patterns.

Abbreviations

BC

breast cancer

EMT

epithelial‐mesenchymal transition

MAPK

mitogen‐activated protein kinase

CCLE

Cancer Cell Line Encyclopedia

ER

estrogen receptor

GEO

Gene Expression Omnibus

GO

Gene Ontology

GPCR

G protein‐coupled receptor

HER2

human epidermal growth factor receptor 2

HR

hazard ratio

KEGG

Kyoto Encyclopedia of Genes and Genomes

NF‐κB

nuclear factor kappa

OS

overall survival

PI3K/AKT

phosphoinositide 3‐kinase/protein kinase

PR

progesterone receptor

RFS

relapse‐free survival

RGS

regulator of G protein signaling

TCGA

The Cancer Genome Atlas

TIMER

tumor immune estimation resource

TNBC

triple‐negative breast cancer

1. Background

Breast cancer (BC) is among the most frequently diagnosed malignancies and remains a leading cause of cancer‐related death in women worldwide [1, 2]. This disease encompasses diverse biological subtypes, each defined by unique molecular profiles and clinical characteristics. Based on gene expression patterns, BC can be categorized into molecular subtypes, including luminal A, luminal B, triple‐negative breast cancer (TNBC), basal‐like, HER2‐positive, and normal‐like [3, 4, 5]. Despite advancements in early diagnosis and systemic treatment, which have significantly improved the clinical outcome of BC, the prognosis for some patients remains poor due to metastasis and recurrence. Therefore, it is necessary to identify new sensitive and specific biomarkers for the prognosis of BC patients [6].

G protein‐coupled receptors (GPCRs) play pivotal roles in mediating cellular responses to external signals [7]. As a major class of membrane proteins, GPCRs are critical to a wide range of signaling pathways throughout the body [8]. Regulators of G protein signaling (RGS) function as GTPase‐activating proteins (GAPs) for GPCRs, accelerating the hydrolysis of GTP to GDP on Gα subunits. This renders them negative regulators of GPCR activity, thereby modulating various downstream signaling pathways [9, 10, 11]. Growing evidence indicates that aberrant expression of certain RGS proteins is closely linked to multiple cancers. These proteins influence tumor biology by controlling cell growth, differentiation, proliferation, and migration [12, 13].

Previous studies have examined the expression patterns and prognostic implications of RGS family members across several malignancies, including breast, gastric, colorectal, pancreatic, and bladder cancers, as well as gliomas [14, 15, 16, 17, 18, 19]. While RGS proteins exhibit diverse biological functions in BC, multiple members have been implicated in promoting or suppressing tumor progression, metastasis, migration, and cell proliferation [20, 21, 22, 23, 24]. On this basis, we carried out a comprehensive bioinformatic analysis to investigate the expression profiles, biological functions, and prognostic value of these genes in BC.

2. Methods

2.1. UALCAN

To explore the expression patterns of the RGS gene family in BC, UALCAN (https://ualcan.path.uab.edu/) was used to analyze mRNA expression data from The Cancer Genome Atlas (TCGA) and MET500 databases [25]. The “TCGA gene analysis” module was used to compare RGS gene expression between BC tumor tissues and adjacent non‐tumor tissues, providing insights into differential expression profiles. Statistical significance was determined using Student's t‐test, defined as an adjusted p < 0.05.

2.2. bc‐GenExMiner

The bc‐GenExMiner v5.0 platform (http://bcgenex.centregauducheau.fr/BCGEM/) was utilized to investigate associations between RGS gene expression and clinical characteristics in BC patients [26]. The expression levels of RGS genes were analyzed in relation to age, nodal status, progesterone receptor (PR) status, estrogen receptor (ER) status, epidermal growth factor receptor 2 (HER2) status, Scarff‐Bloom‐Richardson (SBR) grade, and intrinsic molecular subtypes defined by the Prediction Analysis of Microarray 50 (PAM50), enabling a comprehensive evaluation of clinical relevance. Welch's t‐test was applied for two‐group comparisons, and the Dunnett–Tukey–Kramer test was used for multigroup comparisons, with significance set at p < 0.05.

2.3. Gepia

Gepia (http://gepia.cancer‐pku.cn/) was employed to assess RGS gene expression across different pathological stages of BC, leveraging TCGA RNA profiling data [27]. This multifunctional platform facilitated the exploration of expression patterns to understand stage‐specific variations in RGS gene activity. Statistical significance across disease stages was evaluated using analysis of variance (ANOVA) (p < 0.05).

2.4. Kaplan–Meier Plotter

To evaluate the prognostic significance of RGS gene expression in BC, the Kaplan–Meier Plotter (https://kmplot.com/analysis/) was used to analyze overall survival (OS) and relapse‐free survival (RFS) [28]. Patient samples were categorized into high and low expression groups based on median gene expression, allowing assessment of survival outcomes associated with RGS genes.

Survival outcomes were assessed using the log‐rank test. To control the false discovery rate (FDR) across multiple RGS genes, the Benjamini–Hochberg correction was applied, with adjusted p $\text{values}$ < 0.05 considered significant and reported in Table S6. Hazard ratios (HR) and 95% confidence intervals (CI) were also extracted.

2.5. cBioPortal

cBioPortal (https://www.cbioportal.org/) is an online tool used to investigate and analyze cancer genomic data [29, 30]. We utilized elements from the invasive breast carcinoma dataset (TCGA, Firehose legacy), which contains information on 1108 samples, to examine changes in the genomic profiles of the core and subtype‐specific differentially expressed genes (DEGs). We generated a list of the 50 most altered genes co‐occurring with RGS members. Furthermore, the platform was used to evaluate the association between genomic copy number alterations (CNAs) from GISTIC and mRNA expression Z‐scores. Additionally, we investigated the correlation between CpG DNA methylation levels (HM450) and mRNA expression (RNA‐Seq V2 RSEM). Spearman's rank correlation coefficient (r) was used for methylation‐expression associations, and statistical significance across CNA categories (Deep Deletion, Diploid, Amplification, etc.) was defined as adjusted p < 0.05.

2.6. STRING

The STRING database (https://string‐db.org/) was utilized to construct a protein–protein interaction (PPI) network for RGS family members and the top 50 most frequently altered genes in BC [31, 32]. The resulting network was visualized using Cytoscape (version 3.10.0) to elucidate potential functional interactions. Interactions in STRING were filtered for a confidence score > 0.4, and visualizations were performed using Cytoscape (version 3.10.0).

2.7. GeneMANIA

GeneMANIA (http://www.genemania.org/) [33] is an online tool that provides information on genetic and protein interactions, co‐localization, co‐expression, pathways, and similarity between protein domains of submitted genes. We utilized GeneMANIA to investigate a gene–gene interaction network for RGS family members.

2.8. Enrichr

Enrichr (https://maayanlab.cloud/Enrichr/) is a web‐based analysis tool for gene functional annotation [34, 35]. We conducted Gene Ontology (GO) enrichment, which involves the study of biological processes, molecular functions, and cellular components. Additionally, we used the Kyoto Encyclopedia of Genes and Genomes (KEGG) to conduct pathway enrichment analysis, miRNA prediction using the miRTarBase, and transcription factor analysis using the Chip Enrichment Analysis (ChEA) database of the RGS family via the Enrichr database. We utilized the ggplot2 R package to produce figures for KEGG and GO analyses. The ranking of the results was performed according to the Rich factor, which is calculated as the ratio of the number of differentially expressed genes (DEGs) to the genes that had annotations in the relevant pathway. Thus, increasing the Rich factor, the degree of enrichment also increases.

2.9. MethSurv

MethSurv (https://biit.cs.ut.ee/methsurv/) was used to analyze CpG methylation patterns and their prognostic significance in BC patients, with a focus on RGS genes [36]. This platform enables the evaluation of methylation‐based survival outcomes. The log‐rank test was utilized to evaluate methylation‐based survival outcomes, setting significance at p < 0.05.

2.10. Immune Infiltration Analysis and Single‐Cell Expression Mapping

The association between RGS gene expression and tumor‐infiltrating immune cells in BC was initially assessed using the TIMER database (https://cistrome.shinyapps.io/timer/) [37]. To rigorously account for the confounding effects of tumor cellularity, we subsequently performed purity‐adjusted partial Spearman's correlation analyses using the TIMER 3.0 platform (https://compbio.cn/timer3/) [38]. This analysis was evaluated across multiple immune deconvolution algorithms, including CIBERSORT, QUANTISEQ, xCell, MCP‐counter, EPIC, and TIDE, ensuring robust findings across BC molecular subtypes. Survival outcomes related to immune infiltration were evaluated using Kaplan–Meier and Cox proportional hazards models. Finally, to map the cell‐of‐origin for specific RGS transcripts and distinguish malignant from nonmalignant expression sources, single‐cell RNA sequencing (scRNA‐seq) datasets from the TISCH database (http://tisch.comp‐genomics.org/) were analyzed to visualize gene expression at single‐cell resolution [39].

2.11. Validation Using GEO Datasets

To validate the differential expression of RGS family genes observed in TCGA, we analyzed seven independent BC datasets from the Gene Expression Omnibus (GEO), including GSE65194, GSE45827, GSE42568, GSE29431, GSE65216, GSE54002, and GSE10810 [40, 41, 42, 43, 44, 45, 46, 47]. These datasets were selected based on high‐quality gene expression profiles and sufficient sample sizes (total samples ranging from 58 to 417), ensuring robust differential expression analysis. Preprocessed gene expression data were obtained from GEO (https://www.ncbi.nlm.nih.gov/geo/) [48]. Preprocessed data were quantile‐normalized and log₂ transformed to ensure inter‐dataset consistency.

Initially, differential expression analysis for each dataset was independently performed, comparing BC tumor samples to normal controls using the limma package in R [49]. Genes with a Benjamini–Hochberg adjusted p value < 0.05 and an absolute log₂ fold change (|log₂ FC|) > 0.5 were considered significant. Subsequently, to synthesize these findings into a robust consensus, a formal meta‐analysis was conducted. A random‐effects meta‐analysis model was applied using the meta package (version 4.2.1) in R [50] to calculate the pooled log₂ FC, standard errors (SE), and 95% confidence intervals (CIs) for each RGS family gene. Heterogeneity among datasets was evaluated using the I ² statistic, with statistical significance defined as an adjusted p value < 0.05.

2.12. Cell Line Expression Analysis

To complement the tissue‐based expression analysis, we examined the mRNA expression of RGS family members in BC cell lines using data from the Cancer Dependency Map (DepMap) portal [51]. Gene expression data (log₂ FC ($\mathrm{TPM}$ + 1) format) were retrieved from the dataset (OmicsExpressionTPMLogp1HumanProteinCodingGenes.csv). Cell line metadata, including patient age, race, and pathology (primary tumor vs. metastasis), were sourced from the corresponding DepMap models annotation file (Model.csv). The dataset was filtered to include 73 BC cell lines.

To standardize expression levels across the cohort, the expression values for the targeted RGS genes were normalized to Z‐scores (calculated as (x − μ)/σ for each gene). Overexpressed genes in specific models were typically characterized by a Z‐score > 2. Hierarchical clustering on both genes and cell lines was performed using Euclidean distance. The expression patterns, alongside clinical annotations, were visualized as a comprehensively annotated heatmap using the ComplexHeatmap package [52]. All data processing and visualization were conducted in R utilizing the dplyr, readr, ComplexHeatmap, and circlize packages.

2.13. Drug Sensitivity and Pharmacogenomic Analysis

To investigate the relationship between RGS gene expression and pharmacological response, we integrated transcriptomic data with large‐scale drug screening datasets. Drug response profiles for 6790 compounds, alongside compound and cell line metadata, were acquired from the Broad Institute Drug Repurposing Hub [51, 53]. Cell lines were cross‐referenced using DepMap identifiers to match the drug screening matrix with our RGS mRNA expression dataset (Z‐scores). This alignment yielded a common set of 66 BC cell lines.

For each gene‐compound pair, we calculated the Pearson correlation coefficient (r) to evaluate the association between baseline gene expression and drug response. Pairwise complete observations were utilized to account for missing values in the screening data. Statistical significance was determined by calculating p values based on a two‐tailed t‐distribution, adjusting for the exact number of valid samples per interaction. All data wrangling, statistical analyses, and visualizations—including the matrix bubble plots—were executed in the R statistical computing environment, primarily utilizing the ggplot2 package [54]. Statistical significance was determined using two‐tailed t‐distributions, adjusting for the number of valid samples (p < 0.05). Visualizations were executed in R using ggplot2.

2.14. Statistical Analysis

All local data processing and statistical computing that were not web‐based were performed in the R statistical environment (version 4.2.1). Cox proportional hazards regression analysis was performed to evaluate the association between RGS expression, clinical‐pathological features (Age, Stage, ER, and PR status), and overall survival. Subsequently, a multivariate Cox regression model was constructed to identify independent prognostic factors. Across all tests, a two‐sided p value < 0.05 was considered statistically significant unless otherwise specified (e.g., FDR adjustments).

3. Results

3.1. Expression Patterns of RGSs in Patients With BC

To explore differences in the mRNA expression of RGS family members between normal breast tissue and BC tumors, we analyzed data from the UALCAN database. Our findings revealed significantly elevated mRNA levels of RGS1, RGS4, RGS7, RGS8, RGS10, RGS11, RGS14, RGS16, and RGS19 in BC patients compared to normal samples. In contrast, the mRNA expression of RGS2, RGS3, RGS5, RGS6, RGS12, RGS17, RGS18, RGS20, and RGS22 was notably reduced in tumor tissues. Moreover, the transcript levels of RGS9 and RGS13 showed no significant differences from normal samples, whereas RGS21 displayed only a minor, albeit statistically significant, variation (Figure 1a).

FIGURE 1.

(a) The mRNA expression of various members of the RGS family in BC tissues and normal breast tissues (UALCAN). The X‐axis indicates the sample types (normal and primary tumor), while the Y‐axis shows transcripts per million of each RNA molecule. Significant results for RGS members are indicated in the figure with *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (b) Violin plots illustrating the association of RGS1, RGS2, RGS9, and RGS18 expression with major pathological stages (Stage I to Stage IV) in BC, generated using the GEPIA database. Significant differences across stages are indicated by p < 0.05.

3.2. Association of RGS mRNA Expression With Clinicopathological Features, Molecular Subtypes, and Tumor Stage

We further examined the associations between RGS mRNA expression and key clinicopathological features in BC via the bc‐GenExMiner v5.0 database. As detailed in Table 1, we observed upregulated expression of RGS2, RGS5, RGS8, and RGS11, alongside downregulated expression of RGS4, in patients older than 51 years compared with those 51 years or younger. Additionally, BC patients with positive lymph nodes; receptor status analysis revealed distinct expression patterns. In ER‐negative and PR‐negative BC, mRNA levels of RGS2, RGS6, RGS10, RGS19, and RGS20 were significantly elevated, whereas multiple other RGS members (including R GS4, RGS5, RGS7, RGS8, RGS11, RGS12, RGS13, RGS14, RGS17, and RGS22) were downregulated compared to receptor‐positive tumors. For HER2‐positive BC, RGS4 expression was increased, while RGS1, RGS2, RGS3, RGS6, RGS7, RGS9, RGS12, RGS13, RGS14, RGS18, and RGS22 were decreased. Notably, Triple‐Negative Breast Cancer (TNBC) and basal‐like subtypes were characterized by the high expression of RGS2, RGS6, RGS9, RGS10, RGS19, and RGS20, coupled with the low expression of RGS3, RGS4, RGS5, RGS7, RGS8, RGS11, RGS12, RGS13, RGS17, and RGS22.

TABLE 1.

Association between the mRNA levels of RGSs and the clinicopathological characteristics of BC patients (bc‐GenExMiner v5.0).

Gene	Age		Nodal status		ER		PR		HER2		TNBC/not TNBC		Basal‐like BC/not basal‐like
	≤ 51	> 51	(−)	(+)	(−)	(+)	(−)	(+)	(−)	(+)
RGS1
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA							↓			↓
p value	0.0566		0.2327		0.2161		0.0042 *		0.0286 *		0.5251		0.6921
RGS2
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA		↑			↑		↑			↓	↑		↑
p value	< 0.0001 *		0.1976		< 0.0001 *		< 0.0001 *		0.0072 *		< 0.0001 *		< 0.0001 *
RGS3
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA										↓	↓		↓
p value	0.2719		0.1385		0.1182		0.6468		0.0002 *		0.0036 *		< 0.0001 *
RGS4
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA	↓			↑	↓		↓			↑	↓		↓
p value	< 0.0001 *		< 0.0001 *		< 0.0001 *		< 0.0001 *		0.0166 *		< 0.0001 *		< 0.0001 *
RGS5
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA		↑		↑	↓		↓				↓		↓
p value	0.0286 *		< 0.0001 *		< 0.0001 *		< 0.0001 *		0.1681		< 0.0001 *		< 0.0001 *
RGS6
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA					↑		↑			↓	↑		↑
p value	0.0580		0.3708		< 0.0001 *		< 0.0001 *		< 0.0001 *		< 0.0001 *		< 0.0001 *
RGS7
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA				↑	↓		↓			↓	↓		↓
p value	0.2697		0.0005 *		< 0.0001 *		< 0.0001 *		< 0.0001 *		< 0.0001 *		< 0.0001 *
RGS8
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA		↑			↓		↓				↓		↓
p value	0.0002 *		0.3641		< 0.0001 *		0.0005 *		0.1910		0.0033 *		< 0.0001*
RGS9
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA					↑					↓	↑		↑
p value	0.4290		0.4954		0.0305 *		0.8868		0.0002 *		0.0014 *		< 0.0001 *
RGS10
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA					↑		↑				↑		↑
p value	0.7591		0.4008		< 0.0001 *		< 0.0001 *		0.1275		< 0.0001 *		< 0.0001 *
RGS11
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA		↑			↓		↓				↓		↓
p value	0.0303 *		0.0663		< 0.0001 *		< 0.0001 *		0.2645		< 0.0001 *		< 0.0001 *
RGS12
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA					↓		↓			↓	↓		↓
p value	0.1281		0.3593		< 0.0001 *		< 0.0001 *		0.0158 *		< 0.0001 *		< 0.0001 *
RGS13
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA					↓		↓			↓	↓		↓
p value	0.4874		0.5885		< 0.0001 *		< 0.0001 *		0.0323 *		0.0258 *		< 0.0001 *
RGS14
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA					↓		↓			↓
p value	0.3371		0.7629		0.0019 *		< 0.0001 *		0.0049 *		0.0978		0.0743
RGS16
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA
p value	0.2074		0.2697		0.8351		0.4025		0.0524		0.7615		0.5785
RGS17
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA					↓		↓				↓		↓
p value	0.4515		0.2313		0.0082 *		0.0012 *		0.0905		0.0217 *		0.0037 *
RGS18
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA										↓
p value	0.2996		0.4776		0.8677		0.0507		0.0089 *		0.8081		0.9336
RGS19
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA	↑				↑		↑				↑		↑
p value	0.0323 *		0.4263		< 0.0001 *		0.0030 *		0.6591		< 0.0001 *		< 0.0001 *
RGS20
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA					↑		↑				↑		↑
p value	0.8389		0.0713		< 0.0001 *		< 0.0001 *		0.0754		< 0.0001 *		< 0.0001 *
RGS21
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA
p value	0.9685		0.3533		0.6922		0.7063		0.0835		0.7400		0.5977
RGS22
Number	267	476	332	358	187	530	243	470	396	109	87	578	136	605
mRNA					↓		↓			↓	↓		↓
p value	0.6676		0.0977		< 0.0001 *		< 0.0001 *		0.0027 *		< 0.0001 *		< 0.0001 *

Note: ↑ means upregulated; ↓ means downregulated. Bold values indicate statistical significance (p < 0.05).

Abbreviations: ER, estrogen receptor; HER‐2, human epidermal growth factor 2; PR, progesterone receptor; TNBC, triple‐negative breast cancer.

^* p < 0.05.

When stratifying patients by SBR grade, higher grades were associated with decreased levels of RGS5, RGS7, RGS13, and RGS17, but increased levels of RGS19 (p < 0.05; Figure S1). Analysis across PAM50 molecular subtypes demonstrated elevated expression of RGS4 and RGS16 in the HER2‐enriched subtype, RGS10 and RGS20 in Basal‐like, RGS11 and RGS14 in Luminal A, and RGS13 in normal breast‐like compared to other intrinsic subtypes (Figure 2). Finally, analyzing RNA‐seq data via GEPIA revealed that the expression levels of RGS1, RGS2, RGS9, and RGS18 varied significantly across BC pathological stages (Stage I–IV), while other family members remained relatively stable (Figure 1b).

FIGURE 2.

The expression distribution of RGSs based on BC's PAM50 subtype (bc‐GenExMiner v5.0). Significant differences are indicated in the figure with *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

3.3. Prognostic Value of RGS Family Members in BC

We analyzed the prognostic value of the RGS gene family in BC patients using the Kaplan–Meier Plotter database. Kaplan–Meier curves and log‐rank tests (Figure 3) showed that high mRNA levels of RGS5 (HR = 0.76, p = 0.005), RGS12 (HR = 0.77, p = 0.0075), RGS13 (HR = 0.62, p = 6.9e‐07), RGS18 (HR = 0.64, p = 0.0011), and RGS22 (HR = 0.74, p = 0.027) were associated with better OS. In contrast, elevated mRNA levels of RGS3 (HR = 1.27, p = 0.013) and RGS16 (HR = 1.24, p = 0.024) were linked to worse OS.

FIGURE 3.

Prognostic value of RGSs mRNA expression (Kaplan–Meier Plotter). The relationship between the mRNA expression of the RGS family and OS (a) and RFS (b) in BC patients is shown, with significant differences indicated in the figure. Survival curves for patient groups (higher and lower than median expression levels) are represented by red and black lines, respectively. Confidence intervals are displayed in brackets, and hazard ratios (HR) are provided. MGC26458 refers to RGS9L.

For RFS, higher mRNA levels of RGS3 (HR = 0.84, p = 0.00085), RGS5 (HR = 0.64, p < 1e‐16), RGS6 (HR = 0.81, p = 3.2e‐05), RGS7 (HR = 0.75, p = 4.5e‐08), RGS8 (HR = 0.79, p = 0.0024), RGS9 (HR = 0.78, p = 2.4e‐06), RGS10 (HR = 0.82, p = 0.00014), RGS11 (HR = 0.75, p = 2.5e‐08), RGS12 (HR = 0.75, p = 4.6e‐14), RGS13 (HR = 0.71, p = 3e‐11), RGS14 (HR = 0.76, p = 9.6e‐08), RGS17 (HR = 0.81, p = 4.2e‐05), RGS18 (HR = 0.77, p = 0.00082), and RGS22 (HR = 0.64, p = 1.1e‐08) were significantly associated with better RFS in BC patients. However, increased RGS4 expression (HR = 1.23, p = 4.5e‐05) was tied to shorter RFS.

To identify the prognostic significance of RGS family members and clinical features, univariate and multivariate Cox regression analyses were conducted. The univariate analysis revealed that Age, advanced tumor stages (Stage III and IV), and the expression of RGS3 (p = 0.005), RGS5 (p = 0.038), RGS9BP (p = 0.020), and RGS20 (p = 0.001) were significantly associated with overall survival. However, hormone receptor status (ER and PR) did not show a significant association with OS. Furthermore, the multivariate Cox regression analysis demonstrated that, along with Age and Stage III/IV, the expression levels of RGS3 (HR = 1.214, 95% CI: 1.058–1.392, p = 0.006), RGS4 (HR = 1.167, 95% CI: 1.029–1.325, p = 0.017), and RGS17 (HR = 1.150, 95% CI: 1.039–1.273, p = 0.007) served as independent prognostic factors for poor survival in BC patients (Table 2).

TABLE 2.

Univariate and multivariate Cox proportional hazards regression analysis of RGS family genes and clinical characteristics for OS in BC patients.

Variables	Univariate Cox			Multivariate Cox
	HR	95% CI	p	HR	95% CI	p
Age	1.033	1.020–1.045	< 0.001	1.037	1.023–1.051	< 0.001
Stage II	1.554	0.899–2.686	0.115	1.593	0.910–2.791	0.103
Stage III	2.991	1.687–5.306	< 0.001	3.670	2.048–6.574	< 0.001
Stage IV	13.643	6.725–27.675	< 0.001	12.942	6.255–26.781	< 0.001
ER‐positive	0.744	0.516–1.072	0.112	0.679	0.387–1.193	0.179
PR‐positive	0.761	0.543–1.066	0.112	0.739	0.439–1.241	0.253
RGS1	0.842	0.701–1.011	0.066	0.988	0.823–1.186	0.899
RGS2	0.856	0.681–1.075	0.181	0.924	0.755–1.131	0.444
RGS3	1.199	1.055–1.364	0.005	1.214	1.058–1.392	0.006
RGS4	1.105	0.962–1.269	0.160	1.167	1.029–1.325	0.017
RGS5	1.110	1.006–1.226	0.038	1.081	0.984–1.187	0.104
RGS6	1.050	0.913–1.207	0.494	1.089	0.945–1.256	0.238
RGS7	1.038	0.893–1.207	0.626	1.028	0.884–1.195	0.721
RGS8	0.982	0.717–1.344	0.908	0.928	0.547–1.573	0.781
RGS9	0.762	0.578–1.003	0.053	0.949	0.748–1.202	0.662
RGS9BP	1.146	1.022–1.284	0.020	1.142	0.988–1.321	0.073
RGS10	0.942	0.801–1.108	0.471	0.954	0.793–1.147	0.614
RGS11	0.857	0.704–1.045	0.127	0.840	0.658–1.071	0.160
RGS12	0.931	0.783–1.106	0.415	0.915	0.756–1.108	0.362
RGS13	0.836	0.652–1.072	0.158	0.919	0.715–1.180	0.507
RGS14	0.851	0.707–1.023	0.085	0.920	0.758–1.115	0.396
RGS16	1.069	0.928–1.231	0.356	1.123	0.961–1.312	0.146
RGS17	1.103	0.997–1.221	0.058	1.150	1.039–1.273	0.007
RGS18	0.935	0.790–1.106	0.432	1.035	0.867–1.236	0.702
RGS19	0.924	0.768–1.112	0.402	0.952	0.779–1.163	0.630
RGS20	1.171	1.064–1.289	0.001	1.080	0.979–1.193	0.125
RGS21	0.977	0.796–1.199	0.823	0.938	0.737–1.193	0.601
RGS22	0.834	0.639–1.087	0.179	0.867	0.622–1.209	0.400

Note: Bold values indicate statistical significance (p < 0.05).

Abbreviations: CI, confidence interval; ER, estrogen receptor; HR, hazard ratio; PR, progesterone receptor.

3.4. RGS Family Genomic Alterations and Gene Dosage Effects

Genomic alterations of the RGS family members in BC were assessed using the cBioPortal database. To accurately reflect structural genomic changes and avoid inflating alteration prevalence with transcriptional events, we restricted our query strictly to somatic mutations and copy number alterations (CNAs), excluding mRNA expression outliers. Based on these criteria, we identified genomic alterations of RGS genes in 45% of the 498 evaluated BC patients. Among the family members, RGS22 exhibited the highest rate of structural alterations (17%), followed by RGS7 (14%), RGS4 (12%), and RGS5 (11%). Notably, the predominant type of genomic alteration across the RGS family was copy number amplification, rather than somatic mutations (Figure 4a).

FIGURE 4.

Genomic alteration of RGSs (a) in BC patients (cBioPortal). PPI network of RGSs (b) and the 50 most frequently altered neighboring genes in BC using STRING. Gene–gene interaction network of RGS family members (c) using GeneMANIA. (d) Functional enrichment analysis of RGS genes in BC. (a) GO molecular function, (b) GO biological process, (c) GO cellular component, and (d) KEGG pathway enrichment analysis. The Y‐axis represents the enriched terms, while the X‐axis denotes the rich factor. Dot size corresponds to the number of genes involved, and the color gradient indicates the statistical significance (p value), with lower p values represented by warmer colors.

To address whether these copy number changes functionally impact gene expression levels, we evaluated the association between GISTIC‐level CNAs and mRNA expression. As illustrated in Figure S2, a clear positive correlation was observed across the majority of RGS genes. Specifically, the progression from genomic deletions (deep or shallow) to amplifications was consistently accompanied by a stepwise increase in median mRNA expression levels. This finding establishes a significant gene dosage effect, indicating that DNA copy number variations directly contribute to the dysregulation of RGS transcripts in BC. To explore the functional and physical interactions of RGS family members, protein–protein interaction (PPI) and gene–gene interaction networks were constructed. The PPI network incorporated RGS members alongside their 50 most frequently altered neighboring genes (Figure 4b). The top 10 most frequently altered genes associated with RGS family members are detailed in Table S1, and the gene–gene interaction network highlighted the complex regulatory relationships among the family members (Figure 4c). Furthermore, the Enrichr database was used to identify the pathways and functions of RGS family as well as their frequently altered neighbor genes (Figure 4d). The most significantly enriched molecular functions (MF) were mainly associated with GTPase activity. Enrichment analysis showed that the most significant biological processes (BP) were related to the regulation of G protein‐coupled receptor signaling pathways, and the most commonly enriched cellular components (CC) were associated with neuron projection. Finally, KEGG pathway analysis revealed that the NF‐kappa B signaling pathway prominently interacts with RGS family members and their neighbor genes.

3.5. Predicting miRNAs and Transcription Factors (TFs) Linked With RGS Members

Tables S2 and S3 summarize the TFs and miRNAs predicted to regulate RGS members, derived from the ChEA and miRTarBase databases, respectively. To evaluate their clinical relevance, Kaplan–Meier plotter analyses were performed in BC. Higher expression of the TF RUNX1 was significantly associated with improved OS, whereas elevated levels of EZH2 and CBX2 correlated with shorter OS (all p < 0.05; Figure S3).

Similarly, Kaplan–Meier curves indicated that increased expression of hsa‐miR‐4667, hsa‐miR‐4700, and hsa‐miR‐6763 was significantly linked to reduced OS in BC patients (all p < 0.05; Figure S4).

3.6. The Prognostic Value of Single CpG Methylation of RGSs in BC Patients

DNA methylation data of RGS family genes were analyzed, and the results are shown as heatmaps (Figure 5). Several CpG sites, including cg02586212 (RGS1), cg24547840 (RGS2), cg13938035 (RGS3), cg07835293 (RGS4), cg10604646 (RGS5), cg10050157 (RGS6), cg17480466 (RGS7), cg17124700 (RGS8), cg05511685 (RGS9), cg09872088 (RGS10), cg02899402 (RGS11), cg05275153 (RGS12), cg07178206 (RGS13), cg11302624 (RGS14), cg05302036 (RGS16), cg05102015 (RGS17), cg05207870 (RGS19), cg16319136 (RGS20), and cg06569139 (RGS22), exhibited the highest methylation levels.

FIGURE 5.

Heatmap visualization of CpG methylation levels in RGS genes among BC patients (MethSurv). Each row corresponds to a distinct CpG site, while each column represents an individual patient. Methylation values range from 0 (completely unmethylated) to 1 (fully methylated) and are displayed using a gradient from blue (indicating low methylation) to red (representing high methylation). The heatmap also includes additional annotations highlighting factors such as patient ethnicity, race, age, event status, and genomic context, including UCSC_CpG_island association and UCSC_refGene_Group classification.

In addition, survival analysis revealed significant associations between single CpG methylation and prognosis in BC patients. Specifically, 1 CpG site of RGS1, 5 sites of RGS2, 6 sites of RGS3, 2 sites of RGS4, 1 site of RGS5, 10 sites of RGS6, 14 sites of RGS7, 2 sites of RGS8, 6 sites of RGS9, 8 sites of RGS10, 8 sites of RGS11, 33 sites of RGS12, 1 site of RGS13, 16 sites of RGS14, 6 sites of RGS16, 8 sites of RGS17, 1 site of RGS19, 11 sites of RGS20, 1 site of RGS21, and 9 sites of RGS22 were significantly correlated with patient outcomes (Table S4; Figure S5).

To address whether DNA methylation functionally impacts gene expression, we evaluated the correlation between CpG methylation levels and mRNA expression for the RGS family genes. Consistent with the epigenetic silencing mechanism, we observed significant negative correlations (Spearman r ≤ −0.2, p < 0.05) between methylation and expression for several key genes. Notably, the strongest negative correlations were found for RGS10 (Spearman r = −0.53, p = 1.02 × 10⁻⁵⁷) and RGS2 (Spearman r = −0.51, p = 3.76 × 10⁻⁵⁴). Moderate but highly significant negative correlations were also identified for RGS19 (r = −0.42), RGS14 (r = −0.39), RGS1 (r = −0.38), RGS12 (r = −0.36), RGS20 (r = −0.34), RGS16 (r = −0.32), RGS3 (r = −0.31), RGS6 (r = −0.27), and RGS22 (r = −0.21) (all p < 0.001) (Figure S6).

3.7. Correlation Analysis of RGS Family Genes and Infiltrating Immune Cells

We explored the association between RGS expression and immune cell infiltration in BC. Using the TIMER database, significant correlations were observed between RGS1, RGS8, RGS13, RGS16, RGS18, and RGS19 and the infiltration of six immune cell types, including neutrophils, macrophages, B cells, CD4⁺ T cells, CD8⁺ T cells, and dendritic cells (all p < 0.05; Figure S6). Among these, RGS18 showed the strongest associations, displaying positive correlations with neutrophils (Cor = 0.668, p = 6.98 × 10⁻¹²⁴), macrophages (Cor = 0.447, p = 2.03 × 10⁻⁴⁹), dendritic cells (Cor = 0.653, p = 9.21 × 10⁻¹¹⁷), B cells (Cor = 0.380, p = 5.37 × 10⁻³⁵), CD8⁺ T cells (Cor = 0.634, p = 9.73 × 10⁻¹¹⁴), and CD4⁺ T cells (Cor = 0.582, p = 1.92 × 10⁻⁸⁸) (Figure 6).

FIGURE 6.

The correlation between immune cell infiltration and RGSs in BC (TIMER).

To ensure the robustness of these findings across different deconvolution methodologies, we further analyzed purity‐adjusted partial correlations using the multi‐algorithm TIMER 3.0 platform (Figure S7). Robust and consistent correlations were confirmed for several RGS genes, notably RGS1, RGS16, and RGS18, across various immune cell populations and multiple deconvolution algorithms (including CIBERSORT, xCell, and MCP‐counter). RGS18 exhibited the most striking immune associations across BRCA and its molecular subtypes, consistently demonstrating strong positive correlations with the infiltration of myeloid lineage cells (macrophages, monocytes, and dendritic cells) as well as T cells and B cells across almost all algorithms evaluated (Partial Cor > 0, p < 0.05).

Given that RGS18 is known to be highly expressed in specific hematopoietic lineages, we sought to determine whether these correlations reflect active tumor biology or merely the presence of infiltrating nonmalignant cells. To validate the cellular origin of RGS18, we analyzed scRNA‐seq datasets of BC from the TISCH database (Figure S8). Single‐cell resolution confirmed that RGS18 expression is nearly absent in malignant breast epithelial cells. Instead, its expression is predominantly localized within the tumor microenvironment, specifically enriched in infiltrating immune cells such as M1/M2 macrophages, monocytes, conventional dendritic cells (cDC2), and to a lesser extent, in certain stromal populations. This scRNA‐seq validation confirms that the strong immune correlations observed in bulk tumor data are driven by the robust expression of RGS18 within the infiltrating immune compartment rather than the cancer cells themselves.

Furthermore, the TIMER‐based Cox proportional hazards model, which accounts for tumor‐infiltrating immune cells, revealed that the expression of RGS2 (p = 0.006), RGS3 (p = 0.001), RGS18 (p = 0.027), and RGS20 (p = 0.003) was significantly associated with clinical outcomes in BC patients (Table S5).

3.8. Validation of RGS Expression via Random‐Effects Meta‐Analysis

To robustly validate the differential expression of RGS family genes identified in the TCGA cohort (Figure 1a) and to mitigate potential biases arising from platform heterogeneity, we evaluated seven independent GEO datasets. As a representative case, the expression profiles from the GSE65194 dataset are visualized in Figure S9. Rather than relying solely on these isolated cohort trends, a formal random‐effects meta‐analysis was conducted. Based on our stringent inclusion criteria (|logFC| > 0.5, p < 0.05, and evaluation in ≥ 5 studies), the pooled meta‐analysis revealed eight robustly dysregulated RGS genes in BC (Table 3). Specifically, RGS1, RGS4, RGS10, RGS16, and RGS19 were significantly upregulated, whereas RGS2, RGS3, and RGS5 exhibited significant downregulation across the datasets. The cross‐dataset consistency and heterogeneity (I ²) for key representative genes are illustrated in the generated Forest plots (Figure 7a).

TABLE 3.

Meta‐analysis of RGS family gene expression in BC compared to normal tissues.

Gene symbol	Pooled logFC	95% CI lower	95% CI upper	p	I 2 heterogeneity	Studies
RGS4	2.05	1.50	2.61	6.00E‐13	71.5%	7
RGS1	1.95	0.50	3.41	0.0086	96.4%	7
RGS19	0.93	0.49	1.38	4.14E‐05	93.8%	7
RGS16	0.91	0.44	1.38	0.00017	78.5%	7
RGS10	0.90	0.54	1.27	1.12E‐06	76.8%	7
RGS2	−1.96	−3.00	−0.92	0.00021	92.0%	7
RGS5	−1.05	−1.46	−0.64	5.29E‐07	55.4%	7
RGS3	−0.64	−1.17	−0.11	0.0185	94.5%	7

Note: Genes are ranked by absolute pooled logFC. p values indicate the significance of the pooled effect size across the included studies.

Abbreviations: CI, confidence interval; I ², I‐squared statistic for heterogeneity; logFC, Log₂ fold change.

FIGURE 7.

Validation of RGS expression via meta‐analysis, diagnostic evaluation, and construction of a prognostic signature. (a) Forest plots displaying the random‐effects meta‐analysis of eight dysregulated RGS genes (RGS1, RGS2, RGS3, RGS4, RGS5, RGS10, RGS16, and RGS19) across seven independent GEO datasets. The plots illustrate the mean difference (MD), 95% confidence interval (CI), and heterogeneity (I ²) for each gene. (b) ROC curves demonstrating the diagnostic value of individual RGS genes and the combined 8‐gene panel in distinguishing BC tissues from normal tissues. The 8‐gene panel achieved the highest area under the curve (AUC = 0.989). (c) Forest plot of the multivariate Cox proportional hazards regression analysis for the 6‐gene prognostic signature (RGS1, RGS2, RGS3, RGS10, RGS16, and RGS19). Hazard ratios (HR) with 95% CIs and p values are shown. (d) Kaplan–Meier survival curves for overall survival (OS) of TCGA‐BRCA patients, stratified into high‐risk and low‐risk groups based on the median risk score derived from the LASSO‐based 6‐gene signature (p < 0.0001).

3.9. Diagnostic Value of the RGS Gene Panel in BC

To evaluate the diagnostic accuracy of the RGS family genes in distinguishing BC tissues from normal tissues, the receiver operating characteristic (ROC) curve analysis was performed. The individual diagnostic values varied, with AUCs ranging from 0.607 to 0.890. Notably, RGS2 (AUC = 0.890), RGS19 (AUC = 0.867), and RGS5 (AUC = 0.819) exhibited strong individual diagnostic potentials.

Furthermore, we established an 8‐gene combined diagnostic panel. ROC analysis revealed that this comprehensive panel achieved an exceptionally high diagnostic accuracy, with an area under the curve (AUC) of 0.989, demonstrating a substantial improvement over the diagnostic power of any single RGS gene (Figure 7b). These findings indicate that the 8‐gene RGS signature holds significant promise as a highly sensitive and specific diagnostic biomarker for BC.

3.10. Construction and Validation of the RGS‐Based Prognostic Signature

To construct a robust prognostic model, we applied the Least Absolute Shrinkage and Selection Operator (LASSO) Cox regression analysis to the initially evaluated eight RGS genes. Based on the optimal penalty parameter (λ = 0.0042) determined by 10‐fold cross‐validation, a 6‐gene prognostic signature was established, comprising RGS1, RGS2, RGS3, RGS10, RGS16, and RGS19 (Figure 7c).

The risk score for each patient was calculated using the following formula:

$$\begin{array}{c}\text{Risk Score}=\left(-0.174\times \mathrm{RGS}1\text{expression}\right)+\left(-0.145\times \mathrm{RGS}2\text{expression}\right)+\left(0.215\times \mathrm{RGS}3\text{expression}\right)\hfill \\ +\left(-0.017\times \mathrm{RGS}10\text{expression}\right)+\left(0.148\times \mathrm{RGS}16\text{expression}\right)+\left(-0.062\times \mathrm{RGS}19\text{expression}\right).\hfill \end{array}$$

Based on the median risk score, patients in the TCGA‐BRCA cohort were stratified into high‐risk and low‐risk groups. Kaplan–Meier survival analysis demonstrated that patients in the high‐risk group exhibited significantly worse OS compared to those in the low‐risk group (p < 0.0001) (Figure 7d).

3.11. Expression Profiling of RGS Family Members in BC Cell Lines

Z‐score analysis of 73 BC cell lines (Figure 8a) demonstrated heterogeneous expression patterns among RGS genes. Several genes exhibited elevated levels in specific cell line models. Validating our prior analyses, the key immune‐correlated gene RGS18 presented the most remarkable outlier overexpression across the matrix, specifically in the UACC893 cell line (Z = 5.09). RGS4 and RGS16 exhibited strong overexpression in CAL120 (TNBC) and MCF7 (Luminal A) models (Z = 4.34 and Z = 2.92, respectively), which is consistent with their clinical upregulation in patient cohorts. RGS19, which showed significant diagnostic value (AUC = 0.867) in TCGA‐BRCA, exhibited a highly variable expression profile across different cell lines, suggesting context‐specific regulatory mechanisms.

FIGURE 8.

RGS gene expression profiles and their correlation with drug sensitivity in BC cell lines. (a) Heatmap of Z‐score‐normalized mRNA expression for 21 RGS genes across 73 BC cell lines from the DepMap database. Both genes (rows) and cell lines (columns) are hierarchically clustered. The color scale represents the Z‐score, with red indicating high expression and blue indicating low expression. Clinical annotations for Race, Age, and Pathology are shown above the heatmap. (b) Drug sensitivity correlation matrix visualizing the top 50 most significant associations between RGS gene expression and drug response (AUC). Each bubble corresponds to a specific gene–drug pair. The color scale represents the Pearson correlation coefficient, where blue indicates a negative correlation (higher gene expression correlates with increased drug sensitivity) and red indicates a positive correlation (higher gene expression correlates with drug resistance). The size of each bubble is proportional to the statistical significance, scaled by the −log₁₀(p value).

Conversely, consistent with their overarching downregulation driven by epigenetic silencing in patient datasets, genes such as RGS2 and RGS5 generally demonstrated lower mean expression trends. Nevertheless, specific aggressive models revealed unique activation profiles; for example, TNBC models such as HCC1187 (RGS6, Z = 4.24) and HS578T (RGS5, Z = 3.83) showed notable gene‐specific overexpression. Overall, this profiling confirms that while key hub genes like RGS2 remain largely repressed, others like RGS18, RGS4, and RGS16 can be selectively hyperactivated in distinct BC lineage models.

3.12. Pharmacogenomic Profiling Links RGS Expression to Specific Drug Responses

We next evaluated whether baseline RGS expression levels could predict therapeutic responses in BC models. By integrating our transcriptomic data with high‐throughput pharmacological screens across 66 matched cell lines, we mapped a landscape of significant gene–drug interactions (Figure 8b). The analysis revealed robust pharmacological dependencies, highlighting both potential therapeutic vulnerabilities and resistance phenotypes. Notably, the most profound sensitization effect was observed between high RGS12 expression and PF‐03463275 (r = −0.72), alongside a strong correlation between RGS10 upregulation and sensitivity to Monensin (r = −0.76). Furthermore, elevated RGS4 levels were significantly associated with increased sensitivity to CUDC‐427, an inhibitor of the X‐linked inhibitor of apoptosis protein (XIAP). Conversely, our profiling captured transcripts indicative of resistance mechanisms; for instance, high RGS3 expression exhibited a significant positive correlation with the RAF inhibitor AZ‐628 (r = 0.76), and RGS19 upregulation was strongly linked to Delanzomib resistance (r = 0.76). Collectively, these pharmacogenomic associations suggest that specific RGS expression signatures could serve as predictive biomarkers for targeted interventions in BC.

4. Discussion

In the present study, we systematically evaluated the dysregulation and clinical significance of the RGS gene family in BC. Recognizing the limitations of single‐cohort studies, we initially conducted a family‐wide meta‐analysis integrating data from seven independent datasets. Our findings revealed consistent dysregulation of key RGS genes, indicating upregulation of RGS1, RGS4, RGS10, RGS16, and RGS19, and downregulation of RGS2, RGS3, and RGS5 across the TCGA and multiple GEO datasets. These patterns suggest subtype‐specific roles; for instance, tissue‐level upregulation (e.g., RGS4, RGS16 in UALCAN, Figure 1a) was mirrored by cell line data showing elevated RGS16 in luminal models (e.g., MCF7), RGS4 and RGS14 in TNBC (e.g., CAL120, DU4475, respectively), and RGS1 and RGS13 in HER2+ lines (e.g., HCC1954). These expression trends closely align with their individual diagnostic potentials (e.g., RGS2 AUC = 0.89, RGS4 AUC = 0.77, RGS5 AUC = 0.81, RGS10 AUC = 0.79, and RGS19 AUC = 0.86 in TCGA‐BRCA, Figure 7b). Building upon these individual evaluations, we developed an 8‐gene diagnostic panel that exhibited excellent diagnostic performance (AUC = 0.989) in distinguishing BC from normal tissues, demonstrating a notable improvement over the predictive capability of any single gene alone.

Translating these expression profiles into prognostic outcomes, we utilized LASSO regression analysis to derive a 6‐gene prognostic signature (comprising RGS1, RGS2, RGS3, RGS10, RGS16, and RGS19) capable of stratifying patient survival. Among the identified RGS members, our analyses highlighted RGS3 and RGS4 as potential independent adverse prognostic factors. RGS3 showed reduced expression in more aggressive BC subtypes (including HER2‐positive, TNBC, and basal‐like tumors), yet higher RGS3 expression was independently associated with poorer OS (Multivariate HR = 1.214, 95% CI: 1.058–1.392, p = 0.006). Likewise, clinicopathological evaluation demonstrated that RGS4 was significantly upregulated in ER‐positive, PR‐positive, and HER2‐positive tumors, as well as in older patients (Age > 51); however, multivariate Cox regression analysis established that its elevated expression also predicted worse survival outcomes (Multivariate HR = 1.167, 95% CI: 1.029–1.325, p = 0.017). These findings suggest that, despite their subtype‐dependent expression patterns, both genes may exert clinically relevant effects in BC progression. Our observations align with emerging evidence proposing that specific RGS proteins might adopt oncogenic‐like roles in certain solid tumors, potentially through the dysregulation of GPCR and GTPase signaling pathways [55, 56, 57, 58]. For instance, a recent study demonstrated that RGS3 acts as a tumor promoter in ovarian cancer by facilitating TGF‐β signaling and promoting epithelial‐mesenchymal transition (EMT) [56].

In parallel, recent evidence indicates that RGS4 drives gastric cancer progression by enhancing cell proliferation and invasion via the FAK/PI3K/AKT signaling pathway and EMT, which correlates with poorer patient survival [58].

Notably, epigenetic evaluations revealed that RGS3‐associated methylation sites included cg13938035, with six CpG sites significantly correlated with patient outcomes, while RGS4 harbored cg07835293 alongside two prognostic CpG sites. Together, these observations imply that specific RGS members may function not only as prognostic markers but also as potential predictive biomarkers for personalized therapy in BC. Our pharmacogenomic analysis suggested distinct therapeutic associations for these alterations. Elevated RGS3 expression correlated with a potential resistance to AZ‐628, implying that RGS3 might be involved in adaptive survival pathways or bypass mechanisms against specific targeted therapies. Conversely, higher RGS4 expression appeared to predict increased sensitivity to CUDC‐427 (an IAP antagonist). While requiring experimental validation, these dual prognostic and predictive implications suggest that specific RGS members may serve not only as indicators of disease progression but could also help inform future personalized therapeutic strategies in BC.

To better understand the underlying mechanisms driving the widespread downregulation of RGS family members in BC, we evaluated their epigenetic profiles. Our findings indicate that DNA methylation serves as a crucial regulatory mechanism. For instance, RGS12—despite not being part of our primary prognostic signature—exhibited the highest number of methylated CpG sites among the family (e.g., cg05275153), which likely contributes to its transcriptional silencing. Interestingly, this downregulation was paradoxically associated with improved OS in aggressive subtypes such as TNBC and HER2‐positive BC. While RGS12 acts as a tumor suppressor in other cancers [59, 60, 61]. Its silencing in BC might attenuate specific survival‐promoting pathways in a highly context‐dependent manner. A similar epigenetic phenomenon was observed for RGS5, where hypermethylation (e.g., cg10604646) and subsequent downregulation may disrupt Wnt/β‐catenin and MAPK/ERK signaling. This disruption could potentially limit angiogenesis and metastasis in certain contexts [62, 63, 64]. Together, these observations highlight the critical role of epigenetic modifications in shaping the complex, subtype‐specific functions of RGS proteins in BC.

Elevated RGS5 expression in TNBC cell lines, such as HS578T (Z = 3.83), and reported upregulation in TNBC vasculature, and its reduction correlates with higher SBR grades. Consistent with a tumor‐suppressive role, elevated RGS5 expression strongly predicts favorable clinical outcomes, significantly improving both OS and RFS. However, its regulation appears highly nuanced; elevated expression was observed in specific demographics (patients > 51 years). Furthermore, previous reports linking RGS5 to tumor vasculature under stress conditions [65, 66] suggest that its role might be highly localized. This dichotomy implies that RGS5 may exert divergent effects depending on whether its expression originates from the bulk tumor mass or the heterogeneity of specific microenvironmental niches influencing subtype‐specific prognostic and diagnostic trends.

Beyond individual tumor cell dynamics, RGS proteins act as crucial modulators of the tumor microenvironment (TME) and antitumor immunity. Among them, RGS18 emerged as a highly robust immunoregulatory biomarker. Bulk transcriptomic data via TIMER initially indicated strong correlations between RGS18 expression and the infiltration of neutrophils (r = 0.668, p = 6.98 × 10⁻¹²⁴) and dendritic cells (r = 0.653, p = 9.21 × 10⁻¹¹⁷). Crucially, we validated these findings at single‐cell resolution using scRNA‐seq datasets, which unequivocally demonstrated that RGS18 is practically absent in malignant epithelial cells, originating predominantly from TME‐resident immune populations—specifically dendritic cells, macrophages, monocytes, and mast cells. This immune‐centric profile, supported by favorable OS and RFS, corresponds with RGS18's regulation of the MAPK pathway [67], which enhances antitumor immunity. Furthermore, a recent study identified RGS18 as a plasma biomarker in gastric cancer [68]. Together, these findings suggest RGS18 contributes to BC prognosis by modulating immune responses and inhibiting tumor‐promoting pathways, aligning with its known regulatory roles in platelet and osteoclast activity [67, 69].

Based on our results, RGS1 is a key member exhibiting a confirmed upregulation in BC patients (validated by meta‐analysis), with expression levels varying significantly across different pathological stages. One of the most prominent features of RGS1 is its strong association with the tumor immune microenvironment (TME). Its expression shows a consistent positive correlation with the infiltration of multiple immune cells, including macrophages, neutrophils, and T cells (such as CD4⁺ and CD8⁺), which was corroborated by various algorithms (TIMER 3.0). This strong immunological link aligns with recent findings in other malignancies; for instance, Zhang et al. [70] identified RGS1 as a crucial gene affecting immune status and a potential target for immunotherapy in cervical cancer, while Shengnan et al. [71] highlighted its oncogenic role in facilitating tumor development in ovarian cancer. Furthermore, epigenetic analyses revealed that RGS1 expression is significantly regulated by DNA methylation (negative correlation r = −0.38), and at least one of its prognosis‐related CpG sites is directly associated with patient clinical outcomes.

Our analysis identified RGS2 as a significantly downregulated gene in BC, a trend robustly validated across multiple independent datasets via meta‐analysis. Despite this overarching suppression in tumor tissues, clinicopathological stratification revealed a complex expression pattern, whereby RGS2 mRNA levels were paradoxically elevated in highly aggressive subtypes—including ER/PR‐negative, basal‐like, and TNBC tumors—while remaining decreased in HER2‐positive BC. The overarching silencing of RGS2 appears to be primarily driven by epigenetic mechanisms, as evidenced by a striking negative correlation between its expression and DNA methylation (r = −0.51, p = 3.76 × 10⁻⁵⁴), alongside the prognostic significance of five specific CpG sites.

Clinically, RGS2 demonstrated substantial individual diagnostic value (AUC = 0.89) and was selected as a core component of our 6‐gene LASSO prognostic signature. These findings regarding the downregulation of RGS2 and its clinical implications align closely with recent evidence in other malignancies. Notably, Ihlow et al. [72] demonstrated that RGS2 is barely detectable at the mRNA level in high‐grade serous ovarian cancer, where its suppression is intrinsically linked to chemoresistance and an unfavorable prognosis.

RGS19 is of particular importance due to its high diagnostic potential and predictive role in drug response. This gene showed substantial upregulation in receptor‐negative tumors, TNBC, the basal‐like subtype, and high‐grade (SBR) tumors. Our investigations proved that RGS19 alone possesses high diagnostic accuracy (AUC = 0.867) and constitutes a core pillar of both the 8‐gene diagnostic panel (AUC = 0.989) and our LASSO prognostic model. The regulation of RGS19 is notably influenced by epigenetic modifications, evidenced by its significant negative correlation with DNA methylation (r = −0.42). In the realm of pharmacogenomics, elevated RGS19 expression is highly correlated with resistance to Delanzomib (r = 0.76), rendering it a critical candidate for studying therapeutic evasion mechanisms in aggressive BC subtypes. Interestingly, the biological function of RGS19 in tumor progression appears highly context‐dependent and multifaceted. While a previous study by Li et al. [73] suggested a potential antimetastatic mechanism where RGS19 upregulates Nm23‐H1/2 metastasis suppressors, recent compelling evidence by Ke et al. [74] strongly supports an oncogenic role. They demonstrated that RGS19 is significantly upregulated in hepatocellular carcinoma, correlates with poor prognosis, and actively promotes tumor proliferation and metastasis by activating the MYH9/β‐catenin/c‐Myc positive feedback loop [74]. This recent finding of its pro‐tumorigenic capacity perfectly aligns with our observations regarding the elevation of RGS19 in highly aggressive, treatment‐resistant BC subtypes, further highlighting its complex role in cancer biology.

Similarly, RGS16 displays a dual, context‐dependent role in BC. Previous functional studies have demonstrated that RGS16 can act as a tumor suppressor by negatively regulating the PI3K/Akt signaling pathway and, to a lesser extent, the MAPK pathway in BC cells [22, 75], thereby inhibiting proliferation and sensitizing cells to tyrosine kinase inhibitors. Conversely, our current bioinformatic findings underscore a potentially adverse role for RGS16 in BC progression. We observed a significant upregulation of this gene in clinical tumor tissues, specifically within the HER2‐enriched subtype and in cell lines such as MCF7. This elevated expression is significantly associated with shorter OS (HR = 1.24, p = 0.024). This pro‐tumorigenic behavior mirrors its reported role in other malignancies; for instance, Huang et al. demonstrated that RGS16 acts as a crucial oncogene, promoting malignant progression, epithelial‐mesenchymal transition, and immunological and inflammatory responses in gliomas [19]. Hoshi et al. [76] stated that RGS16 acts as a tumorigenesis suppressor in the early stages, but low expression levels can lead to increased aggressiveness during the progression of cancer. In certain breast tumors, the loss of RGS16 enhances PI3K signaling that is triggered by growth factors, ultimately leading to the promotion of cell proliferation [22]. In line with these immunological associations, our study found that RGS16 expression correlates with immune cell infiltration within the BC tumor microenvironment (TME). Furthermore, our data imply that epigenetic alterations are instrumental in the aberrant activation of this gene in breast tumors, as evidenced by a significant negative correlation with DNA methylation (r = −0.32) and the identification of six distinct prognosis‐related CpG sites. Together, these data suggest that RGS16 plays a multifaceted role in regulating critical signaling pathways, ultimately influencing BC progression and clinical outcomes.

Consistent with these findings, enrichment analyses via Enrichr and STRING/GeneMANIA highlighted significant involvement of GTPase activity, GPCR signaling, NF‐κB signaling, and general “pathways in cancer.” Additionally, predicted miRNAs (e.g., hsa‐miR‐4667, hsa‐miR‐4700, hsa‐miR‐6763) and transcription factors such as RUNX1 (associated with better OS) and EZH2/CBX2 (associated with worse OS) further emphasize the potential regulatory network through which RGS family members modulate BC biology.

EZH2, a subunit of the Polycomb repressor complex, is frequently mutated in various cancers and highly expressed in many others, including BC, due to its function as a histone methyltransferase [77, 78]. In line with our findings, Wang et al. [79] found that increased expression of EZH2 in BC patients was directly correlated with a poor OS. Li et al. also discovered that elevated expression levels of meR342‐EZH2 are linked with metastasis and poor clinical outcomes in patients with BC [78] and may contribute to the development of BC through the activation of NF‐κB and NOTCH1 targets [80, 81]. RUNX1 activity in human BC remains controversial, and little is known about how it contributes to tumor progression. Emerging data indicate that RUNX1 suppresses tumor growth in ER+ BC but promotes tumor aggressiveness in TNBC [82, 83, 84, 85]. Zheng et al. [86] discovered that increased levels of CBX2 expression were notably linked with lower OS in BC patients.

Our analysis positions the RGS gene family as a promising source of prognostic biomarkers, immune regulators, and therapeutic targets in BC. The consistent upregulation of RGS1, RGS4, RGS10, RGS16, and RGS19, alongside the downregulation of RGS2, RGS3, and RGS5 across multiple cohorts, highlights their potential.

The field of cancer biomarker discovery has expanded rapidly, and many recent studies have reported single‐gene or pathway‐specific prognostic and diagnostic markers in cancer. Building on these important findings, our study takes a broader view by examining the entire RGS gene family. A family‐level approach is valuable because RGS members are functionally linked through GPCR signaling, can compensate for each other, and exhibit subtype‐specific patterns of dysregulation that contribute to shaping tumor heterogeneity. By integrating expression, methylation, genomic alterations, immune infiltration, and drug sensitivity data across TCGA, GEO, and CCLE, we were able to identify coordinated trends that may not be apparent from single‐gene analyses. This perspective complements existing biomarker research and suggests that the RGS family could represent a collective source of prognostic and therapeutic insights in BC.

4.1. Limitations

This study leverages comprehensive multi‐omics data from public databases (e.g., TCGA, GEO, DepMap) to provide robust insights into the roles of the RGS family in BC. However, several limitations must be acknowledged. First, despite rigorous preprocessing (e.g., quantile normalization) and random‐effects meta‐analysis, inherent dataset heterogeneity and batch effects cannot be entirely eliminated. Second, while we utilized scRNA‐seq to validate the microenvironmental origin of specific genes like RGS18, our overarching findings remain fundamentally bioinformatic and correlative. The diagnostic and prognostic signatures require rigorous validation in independent, prospective clinical cohorts before translation into practice. Finally, the pharmacogenomic dependencies identified via cell line models necessitate further in vitro and in vivo functional assays (e.g., gene knockout/overexpression models) to confirm causal mechanistic links between RGS expression and drug sensitivity or resistance.

5. Conclusions

In conclusion, this comprehensive multi‐omics study highlights the multifaceted diagnostic, prognostic, and therapeutic significance of the RGS gene family in BC. Through stringent meta‐analysis, we confirmed the robust dysregulation of key members across multiple independent cohorts. Notably, we established a highly accurate 8‐gene diagnostic panel (AUC = 0.989) and a robust 6‐gene prognostic signature (RGS1, RGS2, RGS3, RGS10, RGS16, and RGS19) capable of independently stratifying patient survival. Furthermore, our findings demonstrate that RGS expression in BC is intricately regulated by genomic amplifications and CpG methylation. By integrating bulk and single‐cell transcriptomics, we validated that the strong immune correlations of RGS18 are driven by its specific expression within the infiltrating microenvironment rather than in malignant cells. Finally, linking baseline RGS expression to distinct pharmacological responses reveals novel therapeutic vulnerabilities. Together, these findings position RGS family members as critical biomarkers for risk stratification and as promising therapeutic targets for future precision medicine approaches in BC.

Author Contributions

Zohreh Mirzaei: conceptualization, methodology, writing – original draft, formal analysis. Madiheh Mazaheri Moghaddam: writing – review and editing, data curation, validation. Tahereh Barati: investigation, methodology, writing – original draft. Amir Ebrahimi: formal analysis, resources. Mahmoud Shekari Khaniani: writing – review and editing, project administration.

Funding

The authors have nothing to report.

Ethics Statement

Ethical approval was not required for this study since all the data were obtained from publicly available databases. No individual‐level or identifiable human data were used.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Association between the SBR criterion and the RGS family (bc‐GenExMiner v5.0). Significant associations are indicated in the figure with *p < 0.05.

Figure S2: Association between genomic copy number alterations (CNAs) and mRNA expression of RGS family genes. Box plots display the mRNA expression levels (RNA‐Seq V2 RSEM) of RGS genes stratified by GISTIC‐level putative copy number alteration status (Deep Deletion, Shallow Deletion, Diploid, Gain, and Amplification) in BC patients.

Figure S3: The prognostic value of the TFs regulating RGSs (Kaplan–Meier plotter). The association of mRNA expression of EZH2, CBX2, and RUNX1 with OS in BC patients. A value of p < 0.05 was defined as significant. The confidence intervals are represented in brackets. Red: high expression level; black: low expression level. HR, hazard ratio.

Figure S4: The prognostic value of the miRNAs regulating RGSs (Kaplan–Meier plotter). The association of hsa‐miR4667, hsa‐miR4700, and hsa‐miR6763 expression with OS in BC patients. A value of p < 0.05 was defined as significant. The confidence intervals are represented in brackets. Red: high expression level; black: low expression level. HR, hazard ratio.

Figure S5: Kaplan–Meier curves are used to visualize the prognostic value of single CpG methylation of RGSs in BC patients.

Figure S6: Correlation between DNA methylation and mRNA expression of RGS family genes in BC. Scatter plots illustrate the significant negative correlations between CpG methylation levels (HM450) and mRNA expression (RNA‐Seq V2 RSEM) for RGS family genes. Spearman's rank correlation coefficients (r) and p values are indicated in each panel.

Figure S7: Correlation between RGS family expression and immune cell infiltration in BC. The heatmap illustrates the purity‐adjusted partial Spearman's rank correlation (ρ) between the expression level of RGS genes and the abundance of various tumor‐infiltrating immune cells across multiple deconvolution algorithms (such as TIMER, EPIC, CIBERSORT, xCell, etc.) obtained from the TIMER 3.0 database. Red indicates a positive correlation, while blue indicates a negative correlation. The color intensity reflects the strength of the correlation. Statistical significance is denoted by asterisks (*p < 0.05, **p < 0.01, ***p < 0.001).

Figure S8: Single‐cell transcriptomic profiling of RGS family members in the BC microenvironment. Heatmaps display the average expression levels (log(TPM/10 + 1)) of selected RGS genes across distinct cell populations derived from multiple scRNA‐seq datasets in the TISCH database. The analysis highlights the cell‐of‐origin, distinguishing between expression in malignant epithelial cells versus infiltrating immune and stromal populations.

Figure S9: Validation of RGS gene expression in GSE65194. (a) Expression distribution after normalization. (b) PCA showing separation between cancer and normal samples. (c) Volcano plot highlighting differentially expressed RGS genes. Comparable patterns were consistently observed across other GEO datasets.

Table S1: The top 10 frequently altered genes with RGS family members in BC.

Table S2: The most significant TFs related to RGSs.

Table S3: The most significant miRNAs related to RGSs.

Table S4: The significant prognostic value of single CpG methylation of RGS in BC patients (MethSurv).

Table S5: The Cox proportional hazard model of RGSs and six tumor‐infiltrating immune cells in BC (TIMER).

Table S6: Adjusted p values for overall survival (OS) and relapse‐free survival (RFS) of RGS genes using Benjamini–Hochberg correction.

Data Availability Statement

All data used in this study are publicly available from the corresponding databases. In addition, the codes generated and/or analyzed during the current study are available in the GitHub repository: https://github.com/mirzaeizohreh8‐sudo/RGS_BreastCancer_Analysis.