Single-cell RNA sequencing reveals age-related heterogeneity in the tumor microenvironment of breast cancer patients

Abstract

Breast cancer (BC) is the most common malignancy among women, with its progression and prognosis significantly influenced by the tumor microenvironment (TME). Age-related differences in TME composition lead to distinct tumor behaviors: young patients (≤ 40 years) exhibit aggressive tumors, while elderly patients (> 70 years) experience immunosenescence and reduced therapy responses. We performed single-cell RNA sequencing (scRNA-seq) analysis on tumors from 10 breast cancer patients (5 ≤ 40 years, 5 ≥ 70 years), encompassing 33,664 high-quality cells. After cell annotation and batch correction, malignant epithelial cells were identified using inferCNV. We applied pseudotime trajectory analysis, pathway enrichment, and cell–cell communication profiling to investigate age-specific TME dynamics. Survival relevance was assessed using a GEO cohort (GSE20685) of young breast cancer patients, and immunohistochemical staining was performed on clinical tumor and fibroadenoma tissues to validate protein-level expression of key ISGs. In young patients, malignant epithelial cells showed gradual upregulation of interferon-stimulated genes (ISGs) such as IFI44, IFI44L, IFIT1, and IFIT3 along the pseudotime trajectory, suggesting their involvement in early tumorigenesis. High expression of these ISGs was significantly associated with poor overall survival in a young BC cohort (GSE20685). Immunohistochemical validation further confirmed elevated IFIT3 protein levels in young tumor tissues. In contrast, elderly patients had a TME enriched in macrophages and fibroblasts, with activation of immunosuppressive pathways (e.g., SPP1, COMPLEMENT). Our integrative analysis identifies ISGs as key transcriptional drivers of tumorigenesis in young breast cancer, with potential prognostic and therapeutic value. Despite limited sample size, the combination of single-cell transcriptomics, clinical survival data, and protein-level validation provides robust evidence of age-specific TME remodeling. These findings support the development of age-tailored immunotherapy strategies targeting interferon signaling in young patients and immune checkpoint pathways (e.g., LAG3, CTLA4) in elderly individuals.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-24720-2.

Introduction

Breast cancer (BC) is the most common malignancy among women worldwide and remains one of the leading causes of cancer-related mortality in females^1,2. According to 2020 statistics, there were 2.3 million new cases of breast cancer, accounting for 11.7% of all newly diagnosed cancers globally, ranking fifth among causes of cancer-related deaths (6.9%)³. Based on the expression levels of estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2), and other molecular markers, breast cancer can be classified into five molecular subtypes: Luminal A, Luminal B, HER2-positive (hormone receptor-positive), HER2-positive (hormone receptor-negative), and triple-negative breast cancer (TNBC)^4–6.

The progression and prognosis of breast cancer are closely associated with the tumor immune microenvironment (TME), a complex cellular environment surrounding the tumor that includes immune and inflammatory cells, fibroblasts, neuroendocrine cells, extracellular matrix, and other stromal components. By interacting with tumor cells, the TME plays a critical role in the malignant biological behavior of tumors⁷. Targeting the tumor microenvironment (TME) has become a key strategy for improving the prognosis of breast cancer patients, particularly those with triple-negative breast cancer (TNBC). Current immunotherapy approaches for breast cancer mainly include immune checkpoint blockade (ICB) therapies, such as PD-1/PD-L1 inhibitors and CTLA-4 inhibitors, as well as immunotherapy combined with chemotherapy. While these strategies have yielded encouraging results in certain patient populations, the increasing proportion of elderly breast cancer patients due to population aging has not been adequately addressed. Existing treatment regimens and clinical trials are primarily designed for patients under 70 years old, leaving the therapeutic needs and unique response characteristics of elderly patients insufficiently explored. This gap highlights the urgent need for tailored treatment strategies that specifically address the distinct characteristics of elderly breast cancer patients.

There are significant differences in the clinical-pathological features and TME between young and elderly breast cancer patients. In elderly patients, the TME exhibits weakened immune function, characterized by a reduced proportion of CD4 + stem-like T cells (scT), naïve B cells, and NK cells, alongside an increased proportion of memory B cells and myeloid-derived suppressor cells (MDSCs). Additionally, cancer-associated fibroblasts (CAFs) in young patients are more involved in extracellular matrix (ECM) remodeling, whereas in elderly patients, CAFs exhibit a more pro-inflammatory phenotype and heightened metabolic activity⁸. In contrast, young breast cancer patients tend to have more aggressive tumor proliferation, higher invasiveness, and an increased recurrence rate, often accompanied by enhanced immunosuppressive signals, such as elevated PD-L1 and CTLA-4 expression, as well as an enrichment of regulatory T cells (Tregs)^9,10. These age-related TME characteristics have profound implications for both tumor growth dynamics and immunotherapy responses. Therefore, exploring personalized and precise therapeutic strategies that account for the heterogeneity between young and elderly breast cancer patients is of critical importance¹¹.

Compared to traditional high-throughput sequencing, single-cell RNA sequencing (scRNA-seq) technology allows for the resolution of tumor and TME heterogeneity at the single-cell level, providing unique advantages in uncovering tumor immune evasion mechanisms. This study aims to utilize scRNA-seq to systematically analyze differences in the tumor microenvironment and immune responses between young and elderly breast cancer patients. By identifying age-related immune targets and shared potential therapeutic targets, we aim to provide a theoretical basis and technical support for the precision and individualized treatment of breast cancer.

Method

Sc-RNA seq data processing

This study utilized single-cell RNA sequencing (scRNA-seq) data from 5 young (≤ 40 years old) and 5 elderly (> 70 years old) breast cancer patients, downloaded from the GEO database (GSE161529, GSE176078). Data processing was performed using the Seurat R package (version 5.1.0) and R software (https://www.r-project.org/, version 4.4.1).

Initially, single-cell sequencing data were batch-loaded, and Seurat objects were created using the “CreateSeuratObject” function with parameters set to “min.cells = 3” and “min.features = 300” to filter low-quality cells. The mitochondrial gene (MT) and red blood cell gene (HB) proportions were calculated for each sample, and cells were further filtered based on the following criteria:

nFeature_RNA: Number of expressed genes per cell between 300 and 7000;

nCount_RNA: UMI count greater than 1000, excluding the top 3% of cells with the highest expression;

mt_percent: Mitochondrial gene proportion less than 10%;

HB_percent: Red blood cell gene proportion less than 3%.

After filtering, data were normalized, and highly variable genes were identified. PCA was used for dimensionality reduction, and batch effects were corrected using the Harmony algorithm. Clustering analysis was conducted with a resolution parameter of “resolution = 0.2” and PCA dimensions of “dim = 1:10”, resulting in 13 cell clusters.Group information was added to the metadata (“group”) to facilitate subsequent analyses.

InferCNV analysis for malignant cell identification

To analyze the heterogeneity between tumor and normal cells in breast cancer tissues, InferCNV (https://github.com/broadinstitute/inferCNV/wiki) was used to infer copy number variation (CNV) from scRNA-seq data. Genome-stable B/plasma cells were selected as the reference group, while epithelial cells were designated as the observation group to evaluate genomic instability and potential tumorigenic characteristics.

During the analysis, a genome annotation file (hg38_gencode_v27.txt, https://data.broadinstitute.org/Trinity/CTAT/cnv/) was downloaded. Default hidden Markov model (HMM) settings were applied with the “denoise” parameter enabled, and the threshold was set to 0.1.

Monocle analysis

Quality-controlled and normalized scRNA-seq data were imported into the Monocle3 framework. Cell subpopulations were extracted, ensuring that metadata included cell type annotations. Dimensionality reduction was performed using the UMAP algorithm, and preliminary clustering was conducted based on gene expression patterns. To maintain biological consistency, cell type annotations adhered to standards established in prior analyses.

The “learn_graph” function in Monocle3 was employed to construct a cell trajectory map, with normal epithelial cells designated as the starting point to simulate the progression from normal to tumor states. Genes significantly altered along the pseudotime trajectory were identified, and pseudotime trajectory plots were generated to illustrate dynamic changes in cell states.

GEO data processing and survival analysis

The breast cancer dataset GSE20685 was obtained from the Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/). This dataset was generated using the Affymetrix Human Genome U133 Plus 2.0 Array platform (GPL570) and includes gene expression profiles and clinical follow-up information for 327 patients. Raw CEL files were processed using the “affy” package in R, applying the Robust Multi-array Average (RMA) algorithm for background correction, normalization, and log2 transformation. Probe annotations were matched to gene symbols using the GPL570 annotation file, and where multiple probes mapped to a single gene, their average expression value was used.

Based on the available clinical data, patients younger than 40 years were selected for further analysis, yielding a total of 71 cases. For each of the four genes—IFIT1, IFIT3, IFI44, and IFI44L—patients were stratified into high and low expression groups using the median expression level as the cutoff. Kaplan–Meier survival curves were generated using the “survival” and “survminer” R packages, and group differences in overall survival (OS) were assessed using the log-rank test. A p-value less than 0.05 was considered statistically significant.

Immunohistochemistry and quantification of AOD values

Formalin-fixed, paraffin-embedded breast cancer tissue sections were subjected to standard immunohistochemical (IHC) staining. Briefly, sections were deparaffinized in xylene, rehydrated through graded ethanol, and underwent heat-induced antigen retrieval in citrate buffer (pH 6.0). Endogenous peroxidase activity was blocked using 3% hydrogen peroxide, followed by incubation with 5% bovine serum albumin to reduce nonspecific binding. Sections were then incubated with primary antibodies overnight at 4 °C, followed by secondary antibody incubation and visualization using a DAB chromogenic substrate. Counterstaining was performed with hematoxylin. After dehydration and mounting, slides were imaged using a brightfield microscope.

To quantify IHC staining intensity, average optical density (AOD) was calculated using ImageJ software (NIH, USA). Images were processed using the Colour Deconvolution plugin with the “H DAB” vector to isolate the DAB signal. Following thresholding and binary mask generation, regions of interest (ROIs) were analyzed. For each image, the integrated density (IntDen) and area of DAB-positive regions were measured, and AOD was calculated as:

AOD = Integrated Density / Area.

This approach provides a semi-quantitative assessment of protein expression levels and enables consistent comparisons across experimental groups.

To ensure antibody specificity, all primary antibodies were validated by the manufacturer for IHC on human formalin-fixed paraffin-embedded tissues, and staining patterns were cross-referenced with published Human Protein Atlas data. Negative control sections omitting primary antibodies were included in each staining batch to monitor for non-specific background.

Five breast cancer (BC) and five age-matched fibroadenoma (FA) samples were selected to represent tumor versus benign tissues in young patients. All samples were histopathologically reviewed to confirm diagnosis and tissue integrity before inclusion. Semi-quantitative analysis using average optical density (AOD) was chosen for its reproducibility and robustness in small cohort studies, minimizing inter-observer bias.

Pathway enrichment analysis

Differentially expressed genes (DEGs) were identified between cell subpopulations of young and elderly groups using the “FindMarkers” function in Seurat, with cell type as the identifier (“Idents”). Parameters were set to “logfc.threshold = 0.2” and “min.pct = 0.1”. DEGs were filtered based on the following criteria:

pct.1 > 0.1

p_val_adj < 0.05.

abs(avg_log2FC) > 0.5

Genes with an absolute average log2 fold change greater than 2 were further analyzed. Genes with an average log2FC > 1 were selected to create a new gene set, and volcano plots were generated.

Gene Ontology (GO), KEGG, and Gene Set Enrichment Analysis (GSEA) were performed on the selected gene set. For KEGG analysis, pathways were identified using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database¹². For GSEA, each gene was assigned an ENTREZ ID, and genes lacking ENTREZ IDs were excluded. Genes were ranked in descending order by “avg_log2FC”, and enrichment results were sorted by the absolute value of the normalized enrichment score (NES) to highlight key pathways.

Cell–cell communication analysis

Cell–cell interaction analysis was performed using the CellChat package (version 2.1.20, https://github.com/jinworks/CellChat). Following the official guidelines, data were imported from Seurat objects into CellChat. A CellChat object was created, and communication probabilities for ligand-receptor interactions were calculated for each signaling pathway. To ensure reliability, the ‘filterCommunication’ function was applied to exclude communication networks with fewer than 10 cells. Finally, cell–cell communication probabilities at the signaling pathway level were inferred and visualized.

Result

Single-cell RNA sequencing analysis

We performed single-cell RNA sequencing (scRNA-seq) on tumor samples from 10 breast cancer (BC) patients, including 5 elderly patients (old, > 70 years) and 5 young patients (young, < 40 years) (Fig. 1A).Clinical characteristics of the patients are summarized in Table S1. All patients were treatment-naïve prior to surgery. After stringent quality control and data filtering, a total of 33,664 cells were retained for transcriptomic analysis, with 10,467 cells from the elderly group and 23,197 cells from the young group. Differential gene analysis was conducted using the “FindAllMarkers” function in the Seurat package (parameters: “min.pct = 0.5”, “logfc.threshold = 0.25”), identifying marker genes with significance thresholds of P < 0.05. Heatmaps were generated to visualize the expression of these marker genes (Fig. 1E). Based on classic cell markers, 11 major cell types were identified, including B/plasma cells (IGHG4, IGHG1), basal cells (MYLK, KRT17), endothelial cells (VWF, EMCN), epithelial cells (KRT19, KRT8), fibroblasts (DCN, POSTN), macrophages (C1QA, MSR1), perivascular cells (RGS5, PDGFRB), proliferative cells (MKI67), and T cells (CD3D, CD3E). UMAP dimensionality reduction was performed to display the distribution of different cell types (Fig. 1B). Additionally, Fig. 1C and D illustrate the expression levels of marker genes across cell types.

Fig. 1.

Cellular composition and marker gene expression in breast cancer single-cell data. A Workflow diagram of the study. B UMAP plot showing 11 major cell types identified in the dataset. C Marker gene expression heatmap for identified cell types. D Dot plot of top marker genes for each cell type. E Heatmap of top differentially expressed genes (DEGs) across clusters. F Comparison of cell type proportions between young and elderly groups. G Group-specific UMAP distributions, highlighting differences in cell composition

Comparing the cell composition between young and elderly patients (Fig. 1F), we observed a higher proportion of macrophages and fibroblasts in the elderly group, suggesting a more immunosuppressive tumor microenvironment. Figure 1G further depicts the UMAP results for both groups, showing a higher prevalence of T cells and B/plasma cells in the young group, while fibroblasts and macrophages were more abundant in the elderly group. These findings highlight significant differences in the immune microenvironment between young and elderly BC patients.

Identification of malignant cells via InferCNV analysis

To evaluate the genomic characteristics of tumor cells, we conducted copy number variation (CNV) analysis on the scRNA-seq data using the InferCNV tool. B/plasma cells, which exhibited stable CNV signals across the genome, were selected as the reference group, indicating high genomic integrity (Fig. 2A). Epithelial cells were defined as the observation group to assess potential tumorigenic features. Results revealed that epithelial subclusters 0, 3, 5, 6, and 9 displayed significant CNV features, characterized by increased (red) or decreased (blue) copy numbers in multiple chromosomal regions, suggesting these subclusters were likely tumor cells. In contrast, other epithelial cells exhibited no significant CNV changes, indicating they were normal epithelial cells.

Fig. 2.

Identification and trajectory analysis of malignant epithelial cells and survival validation of ISGs. A InferCNV heatmap identifying malignant epithelial subclusters based on CNV profiles. B–F Pseudotime analysis of epithelial cells from the young group showing gradual transcriptional changes; expression of IFI44, IFI44L, IFIT1, and IFIT3 increased along the trajectory. G–J Kaplan–Meier survival curves for IFI44, IFI44L, IFIT1, and IFIT3 in young breast cancer patients (GSE20685); high expression correlated with poorer prognosis

Pseudotime analysis of normal and tumor epithelial cells

To investigate the dynamic transcriptional changes during tumor progression, we selected Epithelial cell_0 (Ep_0) as the representative tumor subcluster and Epithelial cell_4 (Ep_4) as the presumptive normal epithelial population to construct a pseudotime trajectory. Ep_0 was chosen based on a combination of factors, including high cell count, robust CNV alterations indicative of malignancy, and coherent trajectory alignment with Ep_4. Among all inferred tumor clusters (Ep_0, 3, 5, 6, and 9), Ep_0 exhibited the clearest transcriptional progression and strongest ISG expression gradient along pseudotime.

Figure 2B illustrates the spatial distribution of normal and tumor cells along the pseudotime trajectory, with normal epithelial cells concentrated at the starting point and tumor cells distributed progressively along later stages of the trajectory. Figure 2C highlights normal epithelial cells as the starting point, while Fig. 2D and E show cell type and pseudotime coloring, respectively, revealing the dynamic expression of key genes, including IFI44, IFI44L, IFIT1, and IFIT3, which increased gradually along the pseudotime. These genes may play pivotal roles in tumor progression.

Figure 2F confirms the dynamic expression patterns of these genes through gene trajectories, showing peak expression in later stages of the pseudotime trajectory. Normal epithelial cells displayed stable transcriptional features in the early pseudotime stages, while tumor cells exhibited progressive distribution into mid-to-late stages. Further analysis revealed that interferon-stimulated genes (ISGs) associated with tumor cells, such as IFI44, IFI44L, IFIT1, and IFIT3, showed significantly increased expression along the pseudotime.

Dynamic gene expression patterns were identified: IFI44 and IFI44L displayed rapid increases in early pseudotime, maintaining high expression levels in subsequent stages, suggesting their involvement in the initial phases of tumorigenesis. IFIT1 exhibited significant upregulation in mid-pseudotime, potentially enhancing immune signaling pathways. IFIT3 peaked in later pseudotime stages, indicating its association with tumor invasion and metastasis. In contrast, normal epithelial cells exhibited stable gene expression without notable dynamic changes. These results underscore the transcriptional differences between tumor and normal epithelial cells, reflecting their distinct biological states.

Survival analysis of interferon-stimulated genes in the GEO cohort

To further validate the clinical relevance of IFI44, IFI44L, IFIT1, and IFIT3 in breast cancer patients, we performed survival analysis using the GSE20685 dataset from the GEO database. A total of 71 breast cancer patients aged under 40 were selected, and patients were divided into high- and low-expression groups based on the median expression level of each gene. Kaplan–Meier survival curves were generated (Fig. 2G–J). The results showed that high expression of IFI44 was significantly associated with shorter overall survival (P = 0.0423), and elevated expression of IFI44L, IFIT1, and IFIT3 also correlated with poorer prognosis (P = 0.0139, 0.0123, and 0.00259, respectively). These findings indicate that these genes serve as reliable prognostic markers in clinical breast cancer samples, aligning with their dynamic expression patterns revealed by single-cell trajectory analysis.

Taken together with the pseudotime results, the clinical data further support a potential pro-tumorigenic role for these four interferon-stimulated genes (ISGs) in young breast cancer patients. Previous studies have also demonstrated that high expression of IFIT1 and IFIT3 is significantly associated with advanced T stage, lymph node metastasis, and poor outcomes in patients with oral squamous cell carcinoma (OSCC)¹³. Similarly, IFI44 and IFI44L have been shown to positively correlate with ACSL4 levels in head and neck squamous cell carcinoma (HNSCC), predicting worse survival¹⁴. We hypothesize that these ISGs may promote tumor progression in young breast cancer by activating interferon signaling and cooperatively modulating EGFR pathways and the remodeling of the immune microenvironment, highlighting their potential as targets for both prognostic evaluation and therapeutic intervention.

Immunohistochemical validation of key ISGs expression

To further validate the expression of IFI44, IFI44L, IFIT1, and IFIT3 at the protein level, we performed immunohistochemical (IHC) staining on clinical breast tissue samples. We collected paraffin-embedded surgical specimens from five breast cancer (BC) patients under the age of 40, including two Luminal-type, two HER2-positive, and one triple-negative breast cancer (TNBC) case. As a control group, five age-matched cases of fibroadenoma (FA) were included. Representative staining images are shown in Fig. 3.

Fig. 3.

Immunohistochemical (IHC) validation of interferon-stimulated gene expression in young breast cancer tissues. Representative IHC images and AOD quantification for A IFI44, B IFI44L, C IFIT1, and D IFIT3 in breast cancer tissues (n = 5) and age-matched fibroadenomas (n = 5). IFI44, IFI44L, and IFIT3 showed significantly elevated expression in tumor tissues; IFIT1 showed no significant difference

Quantitative analysis of average optical density (AOD) values revealed that IFI44 expression was significantly higher in BC tissues compared to FA controls (Fig. 3A, p = 0.0462). Similarly, IFI44L showed markedly elevated expression in BC tissues (Fig. 3B, p = 0.0489). IFIT3 expression was also significantly upregulated in the BC group (Fig. 3D, p = 0.0024), consistent with its peak expression in the later stages of the pseudotime trajectory. In contrast, IFIT1 did not exhibit a statistically significant difference in expression between BC and FA samples (Fig. 3C, p = 0.4384).

These IHC findings support the results from the single-cell pseudotime analysis and survival data, further underscoring the potential involvement of IFI44, IFI44L, and IFIT3 in early tumorigenic processes in young breast cancer patients. The lack of significant upregulation of IFIT1 in tissue samples may reflect context-dependent expression regulation or post-transcriptional modifications not captured at the mRNA level. Taken together, these results strengthen the evidence for a pro-tumorigenic role of interferon-stimulated genes in young breast cancer and highlight their potential as diagnostic and therapeutic targets.

Differential analysis of tumor cells

Tumor cell subpopulations

Data from tumor cell subclusters (Epithelial cell_0, 3, 5, 6, 9) were extracted for differential expression analysis between young and elderly patients. The volcano plot (Fig. 4A) highlights differentially expressed genes (DEGs), with several genes significantly upregulated in tumor cells from the young group. Notably, genes such as KRT5, KRT14, and CDKN2A exhibited strong upregulation and statistical significance in elderly tumor cells, consistent with basal-like or senescence-associated phenotypes. Although these genes are located at the upper edge of the volcano plot due to extremely low adjusted p-values, they are explicitly referenced here for interpretability.

Fig. 4.

Transcriptomic features of tumor cells in young vs. elderly breast cancer patients. A Volcano plot of differentially expressed genes (DEGs) in tumor cells.Dot size reflects expression proportion; color indicates log2 fold change. Genes with extreme significance (e.g., KRT5, KRT14, CDKN2A) are discussed in the main text. B GO enrichment of upregulated genes in young group. C KEGG pathway enrichment showing oxidative phosphorylation and proteasome activity. D GSEA enrichment plots highlighting cell cycle and proliferation-related pathways in young tumor cells. Pathway analysis in (C) was performed using the KEGG database.

Source: Kanehisa Laboratories, www.kegg.jp/kegg/kegg1.html

GO functional enrichment analysis (Fig. 4B) revealed that tumor cells from the young group were significantly enriched in functions related to cell cycle regulation, DNA replication, and mitochondrial biogenesis, all of which are closely associated with high proliferative activity. KEGG pathway analysis (Fig. 4C) showed that pathways such as oxidative phosphorylation and proteasome activity were prominently enriched in young tumor cells, indicating the importance of metabolic activity and protein degradation in their biological characteristics.

Additionally, GSEA analysis (Fig. 4D) demonstrated higher activity in cell cycle-related pathways in young tumor cells, consistent with their rapid proliferation. Collectively, tumor cells from the young group exhibited enhanced proliferative and metabolic characteristics, potentially contributing to their higher aggressiveness.

Differential analysis of B/plasma cells

Differential analysis of B/plasma cells revealed several genes significantly upregulated in the young group, such as MZB1 and XBP1, as shown in the volcano plot (Fig. 5A). These genes are closely associated with antibody secretion, endoplasmic reticulum stress, and immune response in B cells. In contrast, fewer genes were significantly upregulated in the elderly group, suggesting reduced immune activity.

Fig. 5.

Transcriptomic differences in B/plasma cells and T cells between young and elderly patients. A–D DEG volcano plot, GO, KEGG, and GSEA analyses for B/plasma cells. Young group showed enrichment in ribosome biogenesis and immune signaling. E–H DEG and pathway enrichment analyses for T cells. Young group T cells had elevated antiviral and cytotoxic markers; elderly group showed enrichment of immune checkpoint genes and inflammation-related pathways. Pathway analysis in B–C includes KEGG pathway enrichment.

Source: Kanehisa Laboratories, www.kegg.jp/kegg/kegg1.html

GO enrichment analysis (Fig. 5B) indicated that B/plasma cells in the young group were significantly enriched in functions related to ribosome biogenesis, translation, and protein synthesis, while those in the elderly group were primarily enriched in protein complex assembly and rRNA processing. KEGG pathway analysis (Fig. 5C) further revealed that young group B/plasma cells were significantly enriched in ribosome biogenesis and COVID-19-related pathways, suggesting greater adaptability under pathophysiological conditions.

Moreover, GSEA analysis (Fig. 5D) confirmed higher activity in cytokine signaling and protein degradation pathways (e.g., proteasome) in the young group’s B/plasma cells. These findings suggest that B/plasma cells in the young group may respond more effectively to challenges within the tumor microenvironment through enhanced immune signaling and metabolic adaptation.

Differential analysis of T cells

Differential expression analysis of T cells (Fig. 5E) showed that antiviral immunity-related genes, such as IFITM1 and GZMB, were significantly upregulated in the young group, reflecting stronger cytotoxicity and antiviral capabilities. Conversely, the elderly group exhibited upregulation of negative regulatory genes, such as LAG3 and CTLA4, indicative of a more immunosuppressive phenotype.

GO analysis (Fig. 5F) revealed significant enrichment of functions related to viral response, immune signaling activation, and defense mechanisms in the young group’s T cells, while the elderly group’s T cells were enriched in functions associated with genome replication and viral lifecycle regulation. KEGG pathway analysis (Fig. 5G) further demonstrated enrichment of antiviral pathways (e.g., influenza and hepatitis B) and oxidative phosphorylation in the young group, whereas the elderly group showed enrichment in inflammation and oxidative stress-related pathways.

Additionally, GSEA analysis (Fig. 5H) highlighted significant enrichment of the MAPK signaling pathway in young group T cells, suggesting a stronger stress response in the tumor microenvironment. In contrast, elderly group T cells were enriched in LPS-mediated inflammatory signaling pathways, potentially linked to chronic inflammation. These results indicate that T cells in young patients exhibit higher immune activity, while those in elderly patients may reflect immunosenescence.

Differences in intercellular interactions within the tumor microenvironment of young and old breast cancer patients

Overall intercellular communication patterns

Comparing intercellular communication within the tumor immune microenvironment between the young and elderly groups (Fig. 6A), the total number of communications was similar (young: 1398; old: 1407). However, the communication strength was significantly higher in the elderly group (old: 42,243; young: 32,118), suggesting that intercellular signaling is more active in the tumor microenvironment of elderly patients, potentially associated with chronic inflammation.

Fig. 6.

Cell–cell communication patterns reveal age-specific TME signaling. A Total number and strength of cell–cell interactions. B Communication network topology showing key communicating cell types. C–D Heatmaps of interaction strengths between cell types. E Enriched signaling pathways in young and elderly groups: young group enriched in CXCL, PTN, and CD40; elderly group enriched in SPP1, COMPLEMENT, and PERIOSTIN

Network topology of intercellular communication

The intercellular communication network (Fig. 6B) illustrated the distribution of communication quantity (left) and strength (right) in both groups. In the elderly group, B/plasma cells, epithelial cells, and macrophages displayed more complex network connections. Regarding communication strength, interactions between B/plasma cells and epithelial cells were the most prominent in the elderly group, suggesting these cell types play central roles in the tumor microenvironment of elderly breast cancer patients.

Cell-type-specific communication strengths

The heatmap of intercellular communication strengths (Fig. 6C) revealed distinct interaction patterns between cell types. In the young group, communication between epithelial and basal cells was stronger, indicating active signaling between tumor and epithelial cells in young patients. In contrast, macrophage-epithelial cell interactions were significantly enhanced in the elderly group, suggesting macrophages may contribute to tumor progression through immunomodulatory effects.

Quantitative and qualitative differences in intercellular interactions

A matrix heatmap (Fig. 6D) displayed differences in intercellular communication in terms of both quantity and strength. In the young group, communication between proliferative cells and basal cells was more frequent, reflecting the active proliferation signaling in their tumor microenvironment. In the elderly group, communication strength was significantly elevated between epithelial cells and macrophages, as well as between epithelial cells and B/plasma cells, highlighting their roles in tumor progression.

Signal pathway analysis reveals conserved and specific pathways in the two groups (Fig. 6E)

Conserved signaling pathways

Both the young and elderly groups exhibited active COLLAGEN and LAMININ signaling pathways, which are closely related to extracellular matrix (ECM) regulation. This suggests that ECM remodeling is a common feature in the tumor microenvironment of breast cancer patients regardless of age.

Young group-specific signaling pathways

In the young group, significantly enriched pathways included CXCL, PTN, CHEMERIN, SELP, and CD40. CXCL and PTN pathways were associated with cell migration and growth, indicating greater tumor invasiveness. CHEMERIN, SELP, and CD40 signaling were linked to leukocyte recruitment, adhesion, and antigen presentation, suggesting a more active immune microenvironment in young patients. Additionally, the ANGPTL pathway, which promotes angiogenesis and tumor dissemination, was more active in the young group.

Elderly group-specific signaling pathways

Signaling pathways significantly enriched in the elderly group included SPP1, COMPLEMENT, PERIOSTIN, and GAL. These pathways are involved in matrix remodeling, immune evasion, and chronic inflammation regulation. SPP1 and COMPLEMENT pathways may enhance the immunosuppressive functions of tumor-associated macrophages (TAMs), promoting tumor progression. High activity in the GAL signaling pathway suggests its role in supporting tumor survival and growth through metabolic regulation.

Discussion

This study utilized single-cell transcriptomics to uncover age-associated heterogeneity in the tumor microenvironment (TME) of breast cancer, with a particular focus on the dynamic expression and functional implications of interferon-stimulated genes (ISGs), including IFI44, IFI44L, IFIT1, and IFIT3. These genes exhibited progressive upregulation along the pseudotime trajectory of malignant epithelial cells in young patients, suggesting a coordinated role in early tumorigenesis and immune microenvironment remodeling.

Among the malignant epithelial clusters identified by inferCNV, Ep_0 was selected as the tumor population for pseudotime modeling due to several features: (i) it displayed prominent copy number variation (CNV) patterns, indicative of chromosomal instability commonly associated with malignancy; and (ii) it exhibited a clearer transcriptional trajectory when aligned with the Ep_4 cluster (normal epithelial cells), allowing more consistent inference of gene activation dynamics. Preliminary attempts using other tumor subclusters (e.g., Ep_3, Ep_5) yielded noisier pseudotime distributions and weaker gradients of ISG expression, thus reinforcing the biological representativeness of Ep_0 in modeling early tumor evolution.

IFI44 and IFI44L were elevated in the early pseudotime phase, implicating them in initiating interferon-related immune responses. In contrast, IFIT3 peaked at later stages, potentially promoting tumor cell invasion. The clinical relevance of these ISGs was confirmed through survival analysis in a young breast cancer cohort (GSE20685), where higher expression correlated with poorer overall survival. Furthermore, immunohistochemical staining validated the protein-level upregulation of IFI44, IFI44L, and IFIT3 in young breast cancer tissues versus age-matched fibroadenomas, reinforcing their potential as diagnostic or therapeutic targets. Quantitative assessment using average optical density (AOD) provided a robust and reproducible measure of staining intensity across samples. In fibroadenoma (FA) tissues, ISG expression was evaluated primarily within epithelial ducts; adjacent stromal regions were also inspected and consistently exhibited weak or undetectable staining, supporting the tumor-specific expression patterns of IFI44, IFI44L, and IFIT3. Notably, although IFIT1 showed transcriptional elevation in pseudotime analysis and bulk data, its protein expression did not significantly differ between groups. This discrepancy may be attributed to post-transcriptional regulation, protein stability differences, or context-dependent translational control, highlighting the complexity of ISG regulation in tumor biology.

In parallel with ISG expression, tumor cells in young patients displayed enhanced proliferative and metabolic states, with enrichment in cell cycle, oxidative phosphorylation, and DNA replication pathways. This aggressive phenotype was accompanied by active immune components. B/plasma cells in young tumors upregulated MZB1 and XBP1 and were enriched in ribosome biogenesis, indicating vigorous antibody production. T cells exhibited increased expression of cytotoxic and antiviral markers (e.g., IFITM1, GZMB), while elderly counterparts showed elevated levels of immune checkpoints (LAG3, CTLA4), suggesting immunosenescence.To ensure the accuracy of our cell-type annotation, we also compared our marker-based classification with publicly available human single-cell references such as Tabula Sapiens. The identified populations (e.g., T cells, B/plasma cells, fibroblasts, macrophages) showed high consistency in canonical marker expression profiles, supporting the validity of our annotations.

Cell–cell communication analysis revealed that although the number of interactions was comparable between age groups, the elderly group exhibited stronger signaling intensity, particularly among macrophages, epithelial cells, and B/plasma cells. These findings suggest a chronically inflamed, immunosuppressive TME in older patients. The young group showed enrichment in immune activation pathways such as CXCL, PTN, and CD40, whereas the elderly group exhibited pathways like SPP1, COMPLEMENT, and PERIOSTIN, promoting matrix remodeling and immune evasion.

While these findings offer valuable insight, several limitations should be acknowledged. First, the small sample size (5 patients per group) may restrict generalizability; however, we ensured rigorous selection of treatment-naïve, clinically annotated cases and supplemented the manuscript with detailed metadata (Table S1). Second, the absence of spatial transcriptomic data limits direct inference of physical cell–cell proximity. Although CellChat analysis was performed with stringent filtering, future integration with spatial data will enhance our understanding of TME architecture. Lastly, while we performed immunohistochemical validation, further functional assays such as flow cytometry or knockdown studies are needed to directly establish the mechanistic roles of ISGs.

In summary, this study reveals that IFI44, IFI44L, IFIT1, and IFIT3 are not only dynamically regulated during tumor progression in young breast cancer patients but also linked to poor prognosis and immunological reshaping. These findings, supported by transcriptional, clinical, and histological evidence, suggest that ISGs may serve as age-specific biomarkers and therapeutic targets. Moreover, our results highlight fundamental differences in immune architecture and intercellular signaling across age groups, underscoring the need for age-tailored immunotherapy strategies in breast cancer.

Author contributions

ZL: Writing—original draft, Methodology, Data curation. JF: Writing—review and editing. MA: Data curation. SK: Writing—review and editing. ZZ: Writing—review and editing. JY: Writing—review and editing. BL: Conceptualization, Funding acquisition, Supervision, Writing—review and editing. FS: Methodology, Investigation, Conceptualization, Writing—original draft.

Funding

This research was supported by the Hebei Provincial Key Research Projects (No. 22372409D), the Hebei Province Outstanding Youth Fund (No. BJ2025056), the Natural Science Foundation of Hebei Province (No. H2022206418), the Health Commission of Hebei Province (No. ZF2025023, No. 20240643, No. 20231063), and the Department of Education of Hebei Province (No. CXZZBS2020111).

Data availability

The datasets analyzed in this study were downloaded from the GEO database (GSE161529, GSE176078, and GSE20685) and are publicly available at [https://www.ncbi.nlm.nih.gov/geo/].

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval and consent to participate

This study includes both publicly available datasets and newly collected clinical tissue samples. The single-cell RNA sequencing and bulk transcriptome data were obtained from the GEO database and are covered by the ethical approvals of the original studies. For the immunohistochemical analysis, clinical breast tissue specimens were collected from patients at The First Hospital of Hebei Medical University with written informed consent. The study protocol was approved by the Ethics Committee of The First Hospital of Hebei Medical University (Approval No. S00680). All methods were performed in accordance with the relevant guidelines and regulations.

Consent for publication

Not applicable, as this study used publicly available data.