Multidimensional single-cell analysis of the molecular characteristics and functional pathways of Regulatory T cells in the microenvironment of HR+ breast cancer

Abstract

Objective

The HR+/HER2- subtype represents the most prevalent form of breast cancer. T-cell heterogeneity and functional status within the breast cancer microenvironment significantly influence tumor progression and the efficacy of immunotherapy. Single-cell RNA sequencing(scRNA-seq) is a powerful tool that enables an in-depth analysis of diverse cell types and their molecular characteristics within tumor tissues.

Methods

This study analyzed scRNA-seq data from HR+/HER2- and ER+ breast cancer samples (GSE228499 and GSE176078) sourced from the GEO database. Initially, the Seurat package was employed for quality control and normalization of the data, followed by dimensionality reduction and clustering. Cell type identification was conducted using the SingleR and Garnett tools, with a focus on the extraction and annotation of T cells and their subsets. Subsequently, FindAllMarkers was utilized to screen for differentially expressed genes, in conjunction with Gene Set Enrichment Analysis (GSEA) pathway analysis using the GSEABase package. Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) Enrichment Analysis using the Sangerbox platform. CellChat was employed to analyze intercellular communication, and Monocle was used for pseudo-time analysis to visualize T cell differentiation trajectories.

Results

This study presents a multilayer analysis of scRNA-seq data derived from 29,540 HR+/HER2- cells and 41,103 ER+ cells, with a specific focus on T cells and Regulatory T (Treg) cells. By employing advanced single-cell sequencing techniques, we elucidate the distinct phenotypic and functional profiles of Treg cells, revealing their pivotal roles in tumor immune evasion and progression.

Conclusion

This approach systematically elucidates the heterogeneity and functional characteristics of T cells and Treg cells within the HR+ breast cancer microenvironment, thereby providing a solid data foundation for advancing our understanding of the tumor immune microenvironment.

Keywords

Breast cancer, scRNA-seq, Tumor immune microenvironment, Treg

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-026-04947-9.

Introduction

Breast cancer is one of the most prevalent cancers among women globally. Clinically, it is categorized into three primary subtypes based on hormone receptor (ER and PR) and HER2 (ERBB2) status: luminal ER-positive and PR-positive (further divided into luminal A and B), HER2-positive, and triple-negative breast cancer (TNBC) [1]. Hormone receptor-positive (HR+) and human epidermal growth factor receptor 2-negative (HER2-) breast cancer is characterized by the immunohistochemical expression of estrogen (ER) and/or progesterone (PR) receptors, representing the most common subtype. A study conducted in the United States found that this subtype accounts for approximately 70% of all phenotypes [2–3]. Most patients are diagnosed at an early stage and achieve favorable outcomes with a combination of treatments, including surgery, chemotherapy, and radiotherapy [4].

The tumor microenvironment of breast cancer is characterized by a diverse array of immune cells, with T cells playing a pivotal role in tumor progression. Among these, Treg cells represent a specialized subset that has garnered significant attention. Extensive research suggests that an increase in Treg cells correlates with enhanced tumor invasiveness, direct or indirect suppression of tumor immunity, and a reduction in overall cancer survival rates [5–7]. Therefore, deciphering the functional status of Tregs in HR+/HER2- breast cancer and their regulatory networks holds significant importance for developing novel immunotherapy strategies.

Single-cell RNA sequencing (scRNA-seq) is a technique utilized for characterizing the transcriptome of individual cells within a sample [8], and it has emerged as a crucial tool for deciphering the tumor immune microenvironment. Although previous studies have employed this technology to identify the presence and partial characteristics of Tregs in breast cancer, a systematic depiction of Treg subpopulation heterogeneity and continuous differentiation trajectories within the most prevalent subtype, HR+/HER2-, remains lacking. Therefore, this study constructed single-cell transcriptome data maps for HR+/HER2- and ER+ breast cancers, extracted T cells from the single-cell data for analysis, and examined Treg-associated expression clusters. We constructed differentiation trajectories and cellular communication networks, explored the potential impact of specifically expressed genes on HR+ breast cancer, and combined functional enrichment with cross-cancer comparisons to reveal the potential functional bias of this feature in breast cancer.

Materials and methods

Data source

Data sets were obtained from the GEO database. GSE228499 contained 9 HR+/HER2- samples, while GSE176078 included 26 primary tumors comprising 12 ER+, 5 HER2+, and 9 TNBC cases. 12 ER+ samples were selected for analysis.

scRNA-seq data preprocessing and integration

Using the CreateSeuratObject function from the Seurat package, a Seurat object was created for the data [9]. Cells expressing at least 300 genes and exhibiting expression in at least 3 cells were filtered. Hemoglobin and mitochondrial percentages were calculated, and cells with hemoglobin percentages below 3% and mitochondrial percentages below 20% were filtered out [10]. A for loop was used to apply the Seurat package’s NormalizeData function to standardize the raw UMI count data. The FindVariableFeatures function was employed to identify 3000 highly variable genes. The FindIntegrationAnchors function was utilized to integrate normalized data and locate anchors, followed by data consolidation using the IntegrateData function.

scRNA-seq dimension reduction and clustering

The data was scaled, and principal component analysis (PCA) was performed. Cluster analysis was conducted using the FindNeighbors and FindClusters functions with a resolution of 0.8. The data was visualized using the t-distributed stochastic neighbor embedding (tSNE) method and uniform manifold approximation and projection (UMAP).

Cell type identification

SingleR annotation was performed on the GSE228499 dataset, where clusters 6, 8, and 14 were identified as T cells. T cell clusters were extracted, and dimensionality reduction clustering was conducted using the same method described in Sect. 2.3, with a resolution set to 0.9. Relevant literature was referenced to obtain marker genes for T cell subsets [10–13]. By combining the results from Garnett and Azimuth annotations with the identified marker genes, the Treg cell cluster was successfully identified. Additionally, SingleR annotation was applied to the GSE176078 dataset, where clusters 1, 2, 8, 12, 18, and 24 were recognized as T cells [14]. These clusters were extracted for dimensionality reduction and clustering, maintaining a resolution of 0.9, in accordance with the methodology outlined in Sect. 2.3. By integrating the Garnett and Azimuth annotation results with the marker genes, we identified clusters in the SingleR annotation that exhibited low T cell expression. After filtering these clusters out, dimensionality reduction clustering was performed. T cells were subsequently clustered and annotated into distinct subtypes based on T cell type markers. The markers were sourced from relevant literature and retrieved from the CellMarker 2.0 database (http://bio-bigdata.hrbmu.edu.cn/CellMarker/) [15–27], ultimately identifying Treg cell clusters.

Differential expression analysis

The FindAllMarkers function was utilized to screen for marker genes in the GSE228499 dataset, resulting in the identification of differentially expressed genes (DEGs) with min.pct = 0.25, avg_log2FC > 1. The top five genes from each cluster were selected to generate expression heatmaps, while the one marker gene in each cluster was chosen for the creation of violin plots illustrating expression across various cell types. Furthermore, FindAllMarkers analysis was performed on the extracted T cells to obtain marker genes specific to T cell clusters. The top ten genes with avg_log2FC > 1 were selected to generate a heatmap, and the distribution of Treg cell marker genes was plotted.Using the FindMarkers function, we screened for differentially expressed genes between cluster 5 T cells and other clusters in GSE228499, applying criteria of min.pct = 0.01 and logfc.threshold = 0.01. We conducted differential analysis between Treg cell clusters and other T cell clusters utilizing RStudio’s GSEA function to identify up- and down-regulated differential pathways. Marker genes were identified from the GSE176078 dataset through differential analysis, producing a heatmap of the top ten T cell marker genes. The Treg cell cluster, as identified from T cell annotation results, underwent GSEA differential analysis against other cell clusters using the same methodology.

Enrichment analysis

We employed the FindMarkers function to identify differentially expressed genes between cluster 5 T cells and other clusters in GSE228499 (p_val < 0.05, avg_log2FC > 1). Subsequent filtering revealed differentially expressed genes with p_val_adj < 0.05, pct1-pct2 ≥ 0.1. These genes were also subjected to KEGG and GO enrichment analysis [28]. For the T cells in the GSE176078 dataset, we conducted FindMarkers analysis to identify differentially expressed genes between cluster 5 and other clusters, followed by KEGG and GO enrichment analysis using the same methodology as applied to GSE228499.

Intercellular communication analysis

We utilized the R package CellChat to analyze and visualize intercellular communication, facilitating the description of interactions between ligands, receptors, and their cofactors [29]. Communication analysis was performed between T cells and other cell types in GSE228499, as well as between Treg cells and other cell types in the GSE176078 dataset.

Pseudo-time series analysis

Monocle was employed to analyze the pseudo-time trajectories of T cells [30]. The function ‘reduceDimension’ was utilized with ‘max_components = 2’ and ‘reduction_method = “DDRTree"’ to analyze the dataset GSE228499. The ‘plot_cell_trajectory’ function was used for trajectory analysis, categorized by ‘color_by = “seurat_clusters"’, ‘State’, and ‘Pseudotime’. Additionally, ‘plot_pseudotime_heatmap’ was employed for cell sorting, branch differential gene analysis, and visualization. Trajectory analysis of T cells in GSE176078 was performed using ‘color_by = “seurat_clusters"’, ‘State’, and ‘Pseudotime’. Furthermore, Treg cell clusters were analyzed to visualize their distribution along trajectories. Jitter plots, violin plots, and pseudotime trajectory plots were generated for ten T cell marker genes.

KEGG and GO enrichment analysis of intersecting genes

To validate the consistency of the five differentially expressed genes in T cell subsets across the two datasets, we calculated the intersection of the differentially expressed genes identified in Sect.  2.6. A hypergeometric test was performed, a Venn diagram was plotted, and KEGG and GO enrichment analyses were conducted. The analysis of the intersecting genes using the CancerSEA database revealed the functional status of genes specifically expressed in the breast cancer Treg subset in comparison to other subsets in Lung Adenocarcinoma (LUAD) and Head and neck squamous cell carcinoma (HNSCC).

Results analysis

Data structures

This study included 9 HR+/HER2- samples from GSE228499 (Fig. 1A). Following preprocessing to eliminate low-quality cells, a total of 29,540 cells were retained for ‘NormalizeData’ and ‘FindVariableFeatures’. Anchor point integration was conducted using the ‘FindIntegrationAnchors’ and ‘IntegrateData’ functions. The integrated data underwent dimensionality reduction, retaining 20 dimensions for subsequent clustering, resulting in 17 clusters (Fig. 1B). Annotation of these 17 clusters using the SingleR package identified 6 distinct cell types: Epithelial_cells (c0, c1, c2, c3, c4, c5, c7, c9, c15), T_cells (c6, c8, c14), Tissue_stem_cells (c10, c13), Monocyte (c11), Endothelial_cells (c12), B_cell (c16) (Fig. 1C). Figure 1D displays key marker genes for each cell type, while Fig. 1E illustrates the presence of various cell types across different samples. Figure 1F displays violin plots showing the expression of a marker gene in different cell types within each cluster.

Fig. 1.

Integration and clustering of HR+/HER2- scRNA-Seq data. A t-SNE of the 9 HR+/HER2- samples. B t-SNE of the 17 cell clusters. C The 6 cell types were identified by SingleR. D Heat map showing top 5 marker genes in 17 clusters. E The cell types were present in the different samples. F Violin plot showing 1 marker gene expressions in 17 cluster across the 6 cell types

Clusters 6, 8, and 14 represent T cells. These clusters were extracted for analysis, followed by dimensionality reduction and clustering (Fig. 2A), and subsequently visualized to illustrate T cell distribution across samples (Fig. 2B). The results of the Garnett annotation are presented in Fig. 2C. Differentially expressed gene analysis comparing cluster 5 with other T cell clusters (Fig. 2D) identified FOXP3, IL2RA, TIGIT, and TNFRSF18 as specifically expressed in cluster 5. Consistent with the annotation results, these genes indicate Treg cell identity. The feature plot of marker genes (Fig. 2E) illustrates the distribution of Treg cell marker genes across clusters.

Fig. 2.

Integration and clustering of HR+/HER2- T cell subset data. A umap of the 13 T cell clusters. B The cell types were identified by Azimuth. C The cell types were identified by Garnett. D Heat map showing top 10 marker genes in 13 clusters. E umap plot highlighting the expression patterns of marker genes for cluster5

The 12 ER+ samples in the GSE176078 dataset (Fig. 3A) yielded a total of 41,103 cells after data processing. Following the same processing steps as in GSE228499, dimensionality reduction and clustering identified 26 clusters (Fig. 3B). SingleR annotation was performed (Fig. 3C), where clusters c1, c2, c8, c12, c18, and c24 were identified as T cells. After extraction and cleaning, 9,399 T cells were obtained and subjected to dimensionality reduction clustering, resulting in 14 clusters (Fig. 3D). By combining each cluster’s marker genes (Fig. 3F) with relevant literature, the cell subpopulations were annotated and identified (Fig. 3E): Cluster 0 specifically expressed CCR7 and SELL, representing naïve T cells, and was labeled as naïve (Tn, c0); Cluster 1, identified by CD8B expression as CD8 + T cells, specifically expressed NKG7, IFNG, CD69, and FOS, exhibiting characteristics of both effector T cells and tissue-resident T cells, and was labeled as effector/tissue-resident memory (CD8 + Teff/rm, c1); Cluster 2 exhibited activation through the expression of factors like FOSB, while mitochondrial-related gene expression indicated active mitochondrial function. These activated T cells also expressed HSPA1A and were labeled as stress response (T_STR, c2); Cluster 3 showed high enrichment of GZMA, GZMB, GZNK, and GZMH, along with elevated CXCR4 expression, combining features of effector memory CD8 + T cells and cytotoxic T cells, and was labeled as effector memory/cytotoxicity (Tem/cyto, c3); Cluster 4 was identified as CD8 + T cells based on CD8A expression, with specific expression of GZMB, NKG7, and GNLY indicating cytotoxic T cells, and was labeled as cytotoxic (Tcy, c4); Cluster 5 exhibited specific expression of FOXP3, TNFRSF4, TNFRSF18, CTLA4, TIGIT, and IL2RA, identifying it as regulatory T cells (Treg, c5); Cluster 6 possessed CCR7 and genes related to regulation and homing, representing central memory T cells (Tcm, c6). Cluster 7 exhibits high expression levels of ribosomal protein genes RPL8, RPL18, RPL18A, RPS15, and RPS5, indicating active cellular metabolism and elevated synthetic demands. The specific expression of the mitochondrial-related gene MT-ATP8 highlights vigorous cellular energy metabolism. Similarly, Cluster 8 shows high expression of ribosomal protein genes RPL13 and mitochondrial-related genes such as MT-ND4L, representing activated T cells labeled as activated (Tac, c7) and (Tac, c8). Cluster 9 is characterized by the absence of FOXP3 expression and the specific expression of CTLA4, TIGIT, and IL2RA, which classifies these cells as exhausted T cells, labeled as exhausted (Tex, c9). Based on the expression of CD8A and CD8B in cluster 10, these cells are identified as CD8 + T cells, which simultaneously express GZMK and cytotoxic markers (Tcy, c10). Cluster 11 displays extremely high CXCL13 expression along with elevated levels of TOX and CTLA4, indicating the presence of Follicular helper T cells, labeled as Follicular helper (Tfh, c11). Cluster 12 shows high expression of TYMS, MKI67, and TOP2A, signifying proliferating T cells, labeled as Proliferating (Tp, c12). Finally, Cluster 13 exhibits extremely high CXCL13 expression alongside specific expression of PDCD1, CTLA4, LAG3, and TOX, in addition to Tfh-associated molecules, indicating follicular helper/exhausted T cells (Tfh/ex, c13).

Fig. 3.

Integration and clustering of ER+ scRNA-Seq data. A t-SNE of the 12 ER+ samples. B t-SNE of the 26 cell clusters. C The 9 cell types were identified by SingleR. D t-SNE of the 14 cell clusters in T cells. E The cell types identified by marker genes. F Heat map showing top 10 marker genes in 14 clusters

Relevant enrichment analysis

The KEGG pathway enrichment results for the GSE228499 dataset indicate that 166 specific markers within the cluster 5 T cells are enriched in various pathways, including Cell Adhesion Molecules (CAMs), Th1 and Th2 cell differentiation, Rheumatoid Arthritis, Hematopoietic Cell Lineage, Th17 cell differentiation, NOD-like receptor signaling pathway, Epstein-Barr virus infection, Influenza A, Cytokine-cytokine receptor interaction, and Human T-cell leukemia virus 1 infection (Fig. 4A). The Gene Ontology (GO) functional enrichment analysis reveals that Biological Process (BP) enrichment occurs in pathways such as immune response, immune system process, T cell activation, cytokine-mediated signaling pathway, cellular response to cytokine stimulus, response to cytokine, cellular response to tumor necrosis factor, response to organic substances, T cell differentiation, and tumor necrosis factor-mediated signaling pathway. Cellular Component (CC) enrichment occurs in pathways such as the side of the membrane, MHC class II protein complex, COPII-coated ER to Golgi transport vesicle, MHC protein complex, coated vesicle, bounding membrane of organelles, ER to Golgi transport vesicle membrane, whole membrane, clathrin-coated endocytic vesicle, and cytoplasmic vesicle membrane. In contrast, Molecular Function (MF) enrichment occurs in pathways such as identical protein binding, MHC class II protein complex binding, tumor necrosis factor-activated receptor activity, peptidase regulator activity, MHC protein complex binding, death receptor activity, MHC class II receptor activity, enzyme binding, protein homodimerization activity, and tumor necrosis factor receptor binding (Fig. 4B). GSEA of cluster 5 Treg cells versus other clusters revealed that the TNFA_SIGNALING_VIA_NFKB, ESTROGEN_RESPONSE_EARLY, and HYPOXIA pathways exhibited p_adjust < 0.05 and were downregulated (Fig. 4C), while the IL2_STAT5_SIGNALING and IL6_JAK_STAT3_SIGNALING pathways exhibited p_adjust < 0.05 and high expression (Fig. 4D).

Fig. 4.

Rich analysis of T cells in HR+/HER2-. A KEGG enrichment analysis of T cells cluster 5 marker genes. B GO enrichment analysis of T cells cluster 5 marker genes. (BP, CC, MF). C GSEA of down-regulated pathways in cluster 5 versus other clusters. D GSEA of up-regulated pathways in cluster 5 versus other clusters

KEGG pathway enrichment analysis of 156 differentially expressed genes in the cluster 5 T cells of GSE176078 revealed that genes distinguishing Treg cell clusters from other cell types were significantly enriched in pathways such as Cytokine-Cytokine Receptor Interaction, Measles, Rheumatoid Arthritis, Viral Protein Interaction with Cytokine and Cytokine Receptor, Kaposi Sarcoma-Associated Herpesvirus Infection, Th17 Cell Differentiation, Type I Diabetes Mellitus, Autoimmune Thyroid Disease, Alzheimer’s Disease, Th1 and Th2 Cell Differentiation, Oxidative Phosphorylation, Huntington Disease, and the T Cell Receptor Signaling Pathway (Fig. 5A). GO functional enrichment analysis indicated that the differentially expressed genes were enriched in BP categories including the Cytokine-Mediated Signaling Pathway, Cellular Response to Cytokine Stimulus, Response to Cytokine, Cell Surface Receptor Signaling Pathway, Response to Organic Substance, Immune Response, Cellular Response to Chemical Stimulus, Extrinsic Apoptotic Signaling Pathway, Cell Death, and Apoptotic Signaling Pathway. In CC categories, the enrichment was observed in Extracellular Exosome, Extracellular Vesicle, Extracellular Organelle, Vesicle, Cytosol, Secondary Lysosome, Mitochondrion, Extracellular Space, Extracellular Region, and Mitochondrial Envelope. Furthermore, in MF categories, the enriched functions included Unfolded Protein Binding, Ubiquitin-Like Protein Ligase Binding, Protein Binding, Bridging, Signaling Adaptor Activity, SH3/SH2 Adaptor Activity, Ubiquitin Protein Ligase Binding, Chaperone Binding, ATPase Regulator Activity, Enzyme Binding, and Protein Binding Involved in Protein Folding (Fig. 5B). GSEA analysis comparing the Treg cluster 5 with other clusters revealed upregulation of the IL2_STAT5_SIGNALING and IL6_JAK_STAT3_SIGNALING pathways, with IL2_STAT5_SIGNALING exhibiting a p_adjust < 0.05 (Fig. 5C). Pathways such as ALLOGRAFT_REJECTION, COAGULATION, E2F_TARGETS, EPITHELIAL_MESENCHYMAL TRANSITION, and MYOGENESIS exhibited low expression levels with a p_adjust < 0.05 (Fig. 5D). The expression patterns of the IL2_STAT5_SIGNALING and IL6_JAK_STAT3_SIGNALING pathways were consistent with the GSEA analysis results for HR+/HER2- cases.

Fig. 5.

Rich analysis of T cells in ER+. A KEGG enrichment analysis of T cells cluster 5 marker genes. B GO enrichment analysis of T cells cluster 5 marker genes. (BP, CC, MF). C GSEA of up-regulated pathways in cluster 5 versus other clusters D GSEA of down-regulated pathways in cluster 5 versus other clusters

Intercellular communication analysis

Using CellChat to detect ligand-receptor pairs and molecular interactions between different cells (Fig. 6), we analyzed the HR+/HER2- dataset. The number of interactions in the cell-cell communication network is illustrated in Fig. 6A, while Fig. 6B presents the interaction weights/strength within the same network. Results indicate that ligand-receptor mediated cell interactions in the HR+/HER2- dataset manifest across 33 pathways, including MIF, COLLAGEN, and LAMININ, with a predominant focus on the MIF signaling pathway (Fig. 6C). The heatmap illustrating this pathway is shown in Fig. 6D, and the centrality analysis of cells within the MIF pathway is presented in Fig. 6E. Notably, the MIF signaling pathway encompasses MIF-CD74 + CXCR4 and MIF-CD74 + CD44 interactions, with the MIF-CD74 + CXCR4 ligand-receptor pair contributing most significantly to this pathway. Furthermore, gene expression of ACKR3 and CXCR2 is associated with this signaling pathway (Fig. 6F). Communication between T cells and other cell types across pathways is depicted in Fig. 6G.

Fig. 6.

Cell-cell communication analysis related to HR+/HER2- T cells. A The number of interactions in cell-cell communication network. B The interaction weights/strength in cell-cell communication network. C Cell-cell communication interaction in MIF signaling pathway D Heatmap showing MIF signaling pathway E Centrality analysis of T cells in MIF signaling pathway. F Violin plot showing the expression of related genes in the MIF signaling pathway G Bubble plot showing Multipathway signaling network of T cells to immune cell subsets

Ligand-receptor pairs and molecular interactions in ER+ samples are presented in Fig. 7. In the ER+ dataset, the number of interactions in the cell-cell communication network is shown in Fig. 7A, while Fig. 7B displays the interaction weights/strength. Interactions between T cells are reflected across 10 pathways, including MHC-I and MIF, with a predominant focus on the MHC-I pathway (Fig. 7C). The heatmap illustrating the MHC-I signaling pathway is presented in Fig. 7D, and centrality analysis of T cells within this pathway is shown in Fig. 7E, highlighting ligand-receptor pairs such as HLA-B-CD8A, HLA-C-CD8A, and HLA-A-CD8A. Among these, the HLA-B-CD8A pair contributes most significantly to MHC-I. A violin plot depicting the expression of related genes in the MHC-I signaling pathway is shown in Fig. 7F.

Fig. 7.

Cell-cell communication analysis related to ER+ T cells. A The number of interactions in cell-cell communication network. B The interaction weights/strength in cell-cell communication network. C Cell-cell communication interaction in MHC-I signaling pathway D Heatmap showing MHC-I signaling network E Centrality analysis of T cells in MHC-I signaling pathway. F Violin plot showing the expression of related genes in the MHC-I signaling pathway

Pseudotemporal analysis

Monocle2 conducted a pseudotemporal analysis on T cell subsets from GSE228499. Trajectory plots illustrating different clusters of T cells are presented in Fig. 8 A, while those showing various states are depicted in Fig. 8B. Additionally, trajectory plots representing different pseudotimes in T cells are displayed in Fig. 8C. The results reveal seven distinct states in T cell differentiation, as shown in Fig. 8D. States 4 and 5 manifest during the late developmental stages of pseudotime. The gene expression patterns of CD8A, GZMA, GZMB, GZMK, and NKG7 vary across differentiation stages, with elevated expression observed at the terminal differentiation state (Fig. 8E). The differential gene heatmap indicates similar expression trends for CD28, CD44, and RORA, all of which exhibit significant upregulation during late developmental stages. Conversely, LEF1 and TCF7 demonstrate a rapid decline in expression during these late stages, while GNLY and CCR6 are expressed at low levels in early pseudotime. PDCD1, LAG3, TOX, HAVCR2, MKI67, PCNA, GZMB, GZMA, and NKG7 all show a similar pattern of rapid upregulation in the late stage. CTLA4 exhibits higher expression in the early stage, followed by a decrease in the middle stage, and a subsequent increase in the late stage. CXCL13 and GZMK are notably highly expressed in the late stage (Fig. 8F). Furthermore, ER+ sample data underwent pseudotemporal analysis, with trajectory plots showing different clusters of T cells presented in Fig. 9A, different states in Fig. 9B, and different pseudotimes in Fig. 9C. Trajectory plots illustrating various cell types in T cells are shown in Fig. 9D, which are categorized into three developmental states (Fig. 9E). Treg cells are primarily located in state 2 (Fig. 9F), indicating the late developmental stage in the pseudotime analysis. The dynamic expression of ten T cell marker genes along pseudotime is illustrated in Fig. 9G, with CD8A, GNLY, GZMA, GZMB, LAG3, NKG7, and TIGIT exhibiting high expression levels in state 2, signifying the late developmental stage in the pseudotime analysis.

Fig. 8.

Trajectory analysis of T cells in HR+/HER2-. A Trajectory plots showing different clusters in T cells. B Trajectory plots showing different states in T cells. C Trajectory plots showing different pseudotimes in T cells. D Trajectory plots showing every state in T cells. E Dynamic expression of 6 T cells marker genes along pseudo time. F Heatmap showing the scaled expression of dynamic genes along pseudo time

Fig. 9.

Trajectory analysis of T cells in ER+. A Trajectory plots showing different clusters in T cells. B Trajectory plots showing different states in T cells. C Trajectory plots showing different pseudotimes in T cells. D Trajectory plots showing different celltypes in T cells. E Trajectory plots showing every state in T cells. F Trajectory plots showing Treg cells. G Dynamic expression of 10 T cell marker genes along pseudo time

Enrichment analysis of intersecting genes

In the hypergeometric test, the background gene set is defined as the intersection of genes identified as differentially expressed between T-cell cluster 5 and other clusters across two datasets using the FindMarkers function (N = 375). After removing genes present in only one dataset, the differentially expressed genes in cluster 5 numbered 106 in GSE228499 and 74 in GSE176078, with 60 overlapping genes. A hypergeometric test revealed that this overlap was significantly higher than expected by chance (p = 5.49e-27), indicating a high consistency in the differential expression profiles of T cell subset 5 across the two independent datasets. A Venn diagram was plotted to illustrate the intersection of the obtained differentially expressed genes (Fig. 10A), followed by KEGG (Fig. 10B) and GO enrichment analyses (Fig. 10C). KEGG pathways showed significant enrichment in Cytokine-cytokine receptor interaction, Rheumatoid arthritis, and Autoimmune thyroid disease pathways. Biological Process (BP) pathways were significantly enriched in the regulation of regulatory T cell differentiation, T cell activation, immune response, immune system processes, regulation of leukocyte activation, regulation of lymphocyte activation, regulation of T cell activation, and regulation of cell activation, as well as the tumor necrosis factor-mediated signaling pathway. Cellular Component (CC) showed significant enrichment in pathways related to the cell surface and the external side of the plasma membrane. Molecular Function (MF) exhibited significant enrichment in pathways such as tumor necrosis factor-activated receptor activity, death receptor activity, identical protein binding, SH3/SH2 adaptor activity, protein binding, bridging, and signaling adaptor activity. Gene set analysis revealed positive correlations in breast cancer with inflammation (+ 0.58), DNA repair (+ 0.35), differentiation (+ 0.31), and cell cycle (+ 0.30). Conversely, negative correlations were observed in LUAD with angiogenesis (− 0.35) and metastasis (− 0.32), as well as a negative correlation with stemness (− 0.47) in HNSCC. These results indicate that this gene set exhibits significantly distinct association patterns across different cancer types (Table 1).

Fig. 10.

Rich analysis of intersecting genes. A Venn diagram of intersecting genes. B GO enrichment analysis of intersecting genes (BP, CC, MF). C KEGG enrichment analysis of intersecting genes

Table 1.

Functional status and correlation coefficients of intersection difference genes across different cancers

Inflammation	BRCA	LUAD	HNSCC
	+ 0.58	–	–
Angiogenesis	–	-0.35	–
Metastasis	–	− 0.32	–
Stemness	–	–	− 0.47
DNA repair	+ 0.35	–	–
Differentiation	+ 0.31	–	–
Cell Cycle	+ 0.30	–	–

Discussion

HR+/HER2- represents a common subtype of breast cancer, accounting for approximately 70% of all cases. With comprehensive treatment approaches that include screening, surgery, chemotherapy, radiotherapy, endocrine therapy, and targeted therapy, this subtype generally has a favorable prognosis. Endocrine therapy is the standard treatment for ER + HR+/HER2- breast cancer. Targeted therapies, particularly cyclin-dependent kinase 4/6 (CDK4/6) inhibitors, combined with endocrine therapy, provide new, safe, and effective treatment options for this subtype [31–33]. However, treatment resistance and recurrence continue to pose significant clinical challenges. The influence of the immune system on cancer remains a topic of considerable interest. Treg cells play a crucial role in maintaining immune homeostasis and suppressing autoimmunity; they are also associated with increased aggressiveness, poor survival rates, and treatment resistance in breast tumors [34–36].

In the HR+/HER2- dataset, cluster 5 was identified as a subset of Treg. GSEA of this subset compared to other T cell subpopulations revealed a high expression of the IL2_STAT5_SIGNALING and IL6_JAK_STAT3_SIGNALING pathways. Similarly, GSEA analysis of the ER+ dataset demonstrated elevated expression of these pathways.

Research indicates that STAT5 activation is essential for the development of thymic Tregs and is critical for the expression and maintenance of the key transcription factor Foxp3 in Treg cells. The expression of Foxp3 determines the immunosuppressive function of Treg cells, and IL-2 receptor β-dependent STAT5 activation is a prerequisite for Foxp3-regulated T cell development [37–38]. IL-6 is produced by various cells within the tumor microenvironment. Studies indicate that the IL-6/STAT3 axis is a key factor in the mechanism of resistance to CDK4/6 inhibitors in HR+/HER2- metastatic breast cancer. STAT3 serves as a biomarker associated with lymph node metastasis and poor prognosis. The activation of this gene enables Treg cells to adapt to the inflammatory environment, while the ablation of Stat3 in myeloid cells suppresses Treg accumulation. The JAK/STAT pathway is a well-established signaling pathway involved in tumorigenesis and progression. This pathway regulates effective immune surveillance while simultaneously promoting immune evasion, which may lead to the development of an immune-excluding tumor microenvironment (TME) and the expansion of immunosuppressive cell populations. The signals associated with this pathway arise from various sources, including cytokines and secretory factors induced by interferon (IFN), transforming growth factor-β (TGF-β), and nuclear factor κB (NFκB) pathways, all of which are directly or indirectly regulated by JAK-STAT activation [39–46]. The consistent enrichment of highly expressed pathways in Treg cell subsets within HR+/HER2- and ER+ breast cancer tissues indicates that this subset may exist in a state of sustained activation and is closely linked to inflammatory signaling within the tumor microenvironment.

Nuclear factor-κB (NF-κB) shares similarities with STAT3. Upon activation, NF-κB influences cancer progression through processes such as promoting tumor growth and inflammation. Persistently activated STAT3 maintains constitutive NF-κB activity in tumors. STAT3 and NF-κB also jointly regulate certain inflammatory genes, and their sustained activation significantly impacts the tumor microenvironment [47–51]. Cytokines in the TNF family primarily induce rapid transcription of genes regulating inflammation, proliferation, and differentiation by activating the NF-κB pathway. Studies indicate that NF-κB and/or related proteins may participate in LPS-induced TNF-α gene transcription activation. Endogenous TNF-α detected at tumor sites is primarily secreted by activated macrophages and tumor cells [52–56]. Breast cancer is recognized as an immunologically cold tumor characterized by an immunosuppressive tumor microenvironment (TME). Immunosuppressive cells in cold tumors—including regulatory T cells (Tregs), myeloid-derived suppressor cells (MDSCs), and tumor-associated macrophages (TAMs)—significantly inhibit T cell activation and function, reinforcing the cold tumor phenotype [57–59]. The TNF-α/NF-κB pathway primarily drives pro-inflammatory responses. The low expression of TNFA_SIGNALING_VIA_NFKB in Tregs suggests that this subset primarily exerts immunosuppressive effects, while the pro-inflammatory role of the TNF-α/NF-κB pathway may be predominantly mediated by MDSCs or TAMs [60–62]. TNFA_SIGNALING_VIA_NFKB is downregulated in Treg cells from HR+/HER2- samples, potentially reflecting a functional division of labor between Tregs and MDSCs or TAMs, where Tregs primarily mediate immunosuppression while pro-inflammatory functions may be carried out by MDSCs and other cells. The low expression of estrogen_response_early indicates that the independence of the Treg cell population is not regulated by estrogen. Hypoxia weakens the cancer immune response [63], and low expression of hypoxia suggests that Treg cell subpopulations may have adapted to a microenvironment with relatively better oxygenation. ER+ samples exhibit low expression of pathways including allograft rejection, e2f target5, epithelial mesenchymal transition, myogenesis, and coagulation, indicating reduced cytotoxicity, proliferation, and migration/invasion in this cluster. High expression of the IL2_STAT5_SIGNALING pathway correlates with Treg cell survival and immunosuppressive function [64]. Enrichment results reflect potential survival and functional maintenance conditions for Tregs within the HR+ tumor microenvironment. A systematic comparison of GSEA results for Treg subpopulations across both datasets reveals both consistent and microenvironment-related differential enrichment patterns. Both datasets show significant enrichment of IL2-STAT5 and IL6-JAK-STAT3 signaling pathways in Treg subpopulations, indicating that these pathways are universally activated in HR+ breast cancer Tregs. However, in the GSE228499 dataset, Treg subsets also exhibit low expression of TNF-α/NF-κB, estrogen_response_early, and hypoxia pathways. Conversely, in the GSE176078 (ER+) dataset, Treg subsets show low expression of allograft rejection, e2f target5, epithelial mesenchymal transition, myogenesis, and coagulation pathways. These differences suggest that Tregs in distinct breast cancer subtypes may adaptively regulate their core inhibitory functions based on local microenvironmental signals. The observed Treg state in hormone receptor-positive breast cancer is substantiated by the consistency of core pathways, while the differential pathways offer insights into subtype-specific functions. The enrichment of IL2-STAT5 signaling confirms successful Treg identification, as IL2RA was among the markers used for annotation. The more informative findings are the pathways that distinguish Tregs from other T cell subsets.

Analysis of HR+/HER2- cell communication reveals enriched signaling between T cells and other cell types in pathways such as MIF and MHC-II. Furthermore, ER+ cell communication analysis shows that Treg cell clusters exhibit enriched signaling with other cell subpopulations via pathways including MHC-I, with co-expressed pathways encompassing MIF, CD99, CLEC, and CXCL. The tumor major histocompatibility complex class I and II (MHC-I and MHC-II) present tumor antigens to T cells; the loss of MHC-I correlates with poorer clinical prognosis [65–66]. MIF, secreted by multiple cell types (including cancer cells, T cells, and macrophages), promotes tumor microenvironment inflammation by inducing inflammatory cytokines such as IL-1β, IL-6, and TNFα. MIF is widely recognized to drive tumor growth through diverse mechanisms [67–68]. NF-κB significantly upregulates CXCL12 and CXCL1 in inflammatory breast cancer, while COX2 and CXCL1 are both elevated in metastatic breast cancer. Increased NF-κB gene expression drives tumor progression and metastasis [69–70]. This observation aligns with Treg cells expression patterns noted in GSEA analysis results.

The HR+/HER2- single-cell differential analysis revealed that Treg subsets specifically express genes such as FOXP3 and LAG3. Pseudo-time analysis indicated that the expression of PDCD1, LAG3, TOX, and HAVCR2 surged rapidly during the final differentiation stage (state 5). The ER+ pseudo-time analysis revealed that Treg cells were distributed at the late differentiation stage (state 2), with genes such as TIGIT and LAG3 showing dramatic upregulation at this stage. This suggests that Treg subsets at the late differentiation stage may exhibit a highly suppressive effector state. Previous studies analyzed 47,016 CD45 + cells from 8 treated patients with primary breast cancer (ER+, PR+, HER2+, TNBC). Genes closely associated with ‘terminal differentiation’ included co-stimulatory molecules (CD2, GITR, OX40, 4-1BB) and co-inhibitory receptors (CTLA-4, TIGIT), as well as Treg-specific genes (FOXP3, IL2RA, and ENTPD1) [13]. This aligns with the late-stage Treg differentiation features observed in our HR+ breast cancer cohort. While our study details this trajectory within HR+ subtypes, future multi-subtype analyses are needed to validate the subtype specificity of these features.

The intersection of differentially expressed genes identified in Sect.  2.6 from the two datasets underwent KEGG enrichment analysis. The results revealed significant enrichment in pathways such as cytokine-cytokine receptor interaction, with IL-2 and IL-6 emerging as key cytokines whose expression is regulated by this pathway. GO analysis further indicated that the molecular functions of these genes are concentrated in cytokine receptor binding and signal transduction (e.g., via SH2/SH3 domains). Cellular components are enriched in plasma membranes and outer membranes, while biological processes involve T cell activation, differentiation, and regulation of inflammatory responses. A comprehensive analysis suggests that the proteins encoded by these genes may participate in the functional execution of Tregs within the tumor microenvironment through membrane receptor-mediated signal sensing and transmission.

The GSE176078 primary study primarily focused on cellular heterogeneity within breast cancer, constructing a comprehensive transcriptional map of breast cancer cell architecture. This study performed a hypergeometric test on differentially expressed genes between regulatory T cell clusters and other clusters across two datasets, revealing an overlap significantly higher than random expectation (p = 5.49e^− 27). This indicates a high degree of consistency in the differential expression profiles of T cell subset 5 across the two independent datasets. Utilizing the CancerSEA database, functional state correlation analysis of 60 shared gene sets across BRCA, LUAD, and HNSCC revealed tumor-type-dependent functional associations: in breast cancer, it showed a strong positive correlation with inflammation, whereas in other cancers, it exhibited negative correlations with angiogenesis, metastasis, and stemness functions. This study not only replicated some observations from the original research but also constructed a working model of Treg functional states in HR+ breast cancer by integrating multidimensional data, including GSEA, cell communication, pseudo-time analysis, and functional enrichment.

Integrating the aforementioned multidimensional analyses, we propose the following working model: Within the HR+ breast cancer microenvironment, a specific subset of Tregs maintains its activated state by co-expressing genes associated with the IL2-STAT5 and IL6-JAK-STAT3 pathways. This subset extensively communicates with tumor cells and other immune cells through molecules such as MIF and MHC, thereby participating in the construction of a local immunosuppressive network. Its late-stage differentiation, characterized by high expression of immune checkpoint molecules, may be associated with enhanced inhibitory activity. These findings provide novel insights for overcoming drug tolerance and developing immunotherapies.

However, this study has limitations. The conclusions presented are primarily based on transcriptomic association analyses, aiming to propose testable hypotheses rather than establish causal mechanisms. For instance, the enrichment of the IL2-STAT5 and IL6-JAK-STAT3 pathways merely suggests their association with Treg activation, which can also be corroborated by relevant literature. However, Whether these pathways directly drive Treg inhibitory functions requires further validation through in vitro inhibition experiments, pathway-specific inhibitors, or knockout models. This study utilized publicly available data with a limited sample size; therefore, future research should incorporate larger clinical cohorts and functional experiments to validate these findings. Additionally, this study primarily focused on the transcriptional characteristics of Tregs and did not systematically analyze their association with patient clinical-pathological parameters (e.g., staging, grading, survival) or treatment responses (e.g., endocrine therapy, CDK4/6 inhibitor resistance). Future research should integrate large clinical cohorts and conduct survival analysis to evaluate the prognostic value of Treg-specific gene sets.

In summary, this study provides significant single-cell insights into the biological characteristics of Tregs in HR+ breast cancer. However, further validation through multicenter, multi-subtype, and multi-omics integrated studies, accompanied by functional experiments, is essential. These limitations also highlight future research directions, including: the incorporation of additional breast cancer subtypes to establish the universality of these characteristics; the integration of spatial transcriptomics to elucidate Treg microenvironmental localization; the validation of key molecular mechanisms through in vitro and in vivo functional experiments; and the exploration of the association between Treg characteristics and clinical outcomes, as well as their therapeutic translational value.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

P. L. , Y. Z. participated in the design of research ideas and the drafting of manuscripts, analysed the data. S. G. , X. C. made some well-intentioned suggestions. J. Z. , M. G. , H. L. helped draw relevant pictures.

Funding

This work has been supported by Beijing Natural Science Foundation (No. 7242230, China).

Data availability

This study only conduct analysis based on GEO public database, GSE228499 ([GSE228499[Accession] - GEO DataSets - NCBI] (https:/www.ncbi.nlm.nih.gov/gds/?term=GSE228499%5bAccession%5d)) and GSE176078( [GSE176078[Accession] - GEO DataSets - NCBI](https:/www.ncbi.nlm.nih.gov/gds/?term=GSE176078%5bAccession%5d)), T cell markers were obtained by screening the literature and searching the CellMarker 2.0 database ([CellMarker2.0](http:/bio-bigdata.hrbmu.edu.cn/CellMarker/CellMarkerSearch.jsp? index_species=Human&index_tissue=Breast&index_cellname=T%20cell&index_key=1)), with no experiments on humans and/or the use of human tissue samples involved.

Declarations

Ethics approval and consent to participate

This study utilized publicly available datasets from the Gene Expression Omnibus (GEO) database (accession numbers: GSE228499 and GSE176078).This study does not involve ethical approval.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.