ISG15-driven immune modulation and tumor progression in breast cancer metastasis: insights from single-cell and spatial transcriptomics.

Background
Breast cancer (BRCA) is a malignant tumor [1–3] that significantly affects women’s health. Although early detection and treatment methods for BRCA have improved, lymph node metastasis remains one of the major causes of poor prognosis for patients [4–6]. Recent studies have highlighted the important role of cancer stem cells (CSCs) in the development and metastasis of BRCA [7–9]. However, the molecular mechanisms of CSCs in BRCA lymph node metastasis and their interaction with the immune microenvironment are still unclear [10, 11].
The tumor microenvironment (TME) exerts a profound influence on tumor development and metastasis [12, 13]. Among immune components, tumor-associated macrophages (TAMs) and T cells are particularly critical in BRCA [14, 15]. TAMs typically express markers such as CD206 and ARG1 and possess strong immunosuppressive activity. They remodel the immune microenvironment by secreting IL-10, suppressing antigen presentation, and promoting Treg infiltration, thus accelerating immune escape [16, 17]. However, the crosstalk between BRCA stem cells and immune cell populations remains unclear [7, 18, 19]. Elucidating these interactions could reveal novel mechanisms of immune escape and inform therapeutic strategies.

In the immune microenvironment of cancer, macrophages and T cells are key regulatory factors, and their polarization status and functions determine the balance between immune surveillance and tumor progression [20]. Interferon-stimulated gene 15 (ISG15), an immune-related molecule, has recently attracted attention in various cancers. ISG15 not only has antiviral effects during viral infections but has also been shown to play an important role in tumor immune evasion and immune microenvironment regulation [21]. Recent studies have shown that ISG15 is involved in regulating CSC self-renewal and immune modulation across multiple malignancies, such as breast, lung, and glioma cancers. It regulates the stability of stemness-related proteins via ISGylation modification and facilitates the formation of an immunosuppressive microenvironment [22–24]. However, the specific role of ISG15 in breast CSCs and its mechanisms in regulating the immune microenvironment have not been fully studied.
ISG15 has been reported to influence macrophage polarization and T cell activation within tumors [25]. Furthermore, ISG15 expression levels in BRCA are closely related to tumor progression and metastasis, indicating its potential importance in disease advancement. However, the specific function of ISG15 in breast CSCs and the mechanisms by which it may promote tumor metastasis through immune cell regulation remain unknown. Therefore, in-depth research on ISG15’s role in CSCs and its microenvironment is essential for understanding BRCA metastasis mechanisms and for developing new targeted therapeutic strategies.
In this study, we employed a mouse model of BRCA to examine the role of BRCA stem cells in lymph node metastasis. scRNA-seq was applied to primary and lymph node metastatic tumor tissues, and bioinformatics analyses were conducted to delineate cellular composition differences [26–28]. Key findings were further validated through both in vitro and in vivo experiments.
Our findings revealed an elevated proportion of CSCs in BRCA lymph node metastasis tumor tissue. These CSCs appeared to facilitate M2 polarization of TAMs via cellular interactions and modulate T cell activation by influencing specific signaling pathways. These observations underscore the pivotal role of CSCs in promoting lymph node metastasis.
This study aims to systematically investigate ISG15’s role in breast CSCs through in vivo and in vitro models, uncovering its mechanisms in regulating macrophage polarization and T cell activity, and exploring its potential function in BRCA lymph node metastasis. By gaining insights into the interactions between breast CSCs and immune cells, we hope to identify new therapeutic targets and strategies for BRCA treatment. Our findings will contribute to understanding the mechanisms behind BRCA progression and metastasis and provide theoretical and clinical evidence for developing personalized treatment plans and immunotherapies. Ultimately, this research has the potential to improve the prognosis of BRCA patients and achieve significant breakthroughs in BRCA treatment.

Methods

Experimental animals
Wild-type BALB/c mice (6 weeks old) and BALB/c nude mice (Cat#: 211, 401, Beijing Vitonlihua Experimental Animal Technology Co., Ltd.) were maintained under specific-pathogen-free (SPF) conditions at 22–25 °C and 60–65% humidity. After a 1-week acclimatization period, experimental procedures were initiated with continuous health monitoring. All animal experiments were approved by the Animal Ethics Committee of China Medical University (CMUXN2023046) [29].

Analysis of single-cell tumors
A total of 2 × 105 murine BRCA cells 4T1 (Cat#: MZ-0007, Ningbo Mingzhou Biotechnology Co., Ltd) were implanted into the fourth mammary fat pad of wild-type female BALB/c mice. On day 14, when tumors reached ~ 250 mm3, primary tumor tissues (BRCA_PT dataset) were collected. After tumor removal, the surgical incisions were sutured, allowing the mice to develop spontaneous lymph node metastasis. On day 28, mice were euthanized, and metastatic tumor tissues from inguinal lymph nodes were harvested to generate the lymph node metastasis dataset (BRCA_LNMT).
Tumor tissues were rinsed with ice-cold PBS and enzymatically dissociated in 1 mg/mL collagenase (Cat#: C2674; Sigma-Aldrich, USA) at 37 °C for 10 min, followed by a 5-min digestion with pancreatin/EDTA (Cat#: 25,200,072; Gibco, USA) at 37 °C to generate single-cell suspensions. Cells were isolated using the C1 Single-Cell Auto Prep System (Fluidigm, USA), lysed on-chip to release mRNA, and reverse-transcribed into cDNA. The cDNA was fragmented, pre-amplified within the microfluidic chip, and subsequently processed for library construction before sequencing on an Illumina HiSeq 4000 platform (paired-end, 2 × 75 bp, ~ 20,000 reads per cell) [30].
scRNA-seq data were processed in Seurat (v4.3.0). Quality control thresholds were nFeature_RNA > 500, nCount_RNA > 3000, nCount_RNA < 20,000, and percent.mt < 10. Data normalization was performed with the LogNormalize method, followed by principal component analysis (PCA) on the top 2000 highly variable genes. Significant principal components were selected using JackStrawPlot and ElbowPlot for UMAP clustering. Cluster-specific marker genes were identified with FindAllMarkers and annotated using the CellMarker database and literature. Gene expression patterns were visualized with FeaturePlot and VlnPlot. Intercellular signaling pathway activities were analyzed using the “CellChat” R package (v1.6.1) [31, 32]. CytoTRACE2 stemness scoring analysis was performed based on the raw UMI count matrix from the RNA assay. Inference was conducted using the CytoTRACE2 R package (CytoTRACE2 v1.0.0, https://github.com/digitalcytometry/cytotrace2) [33].

Spatial transcriptome data analysis
The spatial transcriptome sequencing dataset GSE198353 of mouse BRCA tissue was downloaded from the GEO database (http://www.ncbi.nlm.nih.gov/geo/). The Seurat package’s Load10X_spatial function was employed to create a Seurat object, combining the raw gene expression matrix, spatial coordinates, and tissue H&E images. Data were normalized and reduced to the top 20 principal components via PCA. Marker genes were identified with FindAllMarkers, and spatially variable genes were detected using FindSpatiallyVariableFeatures with default parameters. To integrate scRNA-seq data with 10× Visium spatial transcriptomics, we applied Seurat’s anchor-based integration pipeline, enabling the transfer of cell-type annotations from scRNA-seq to the spatial transcriptome. The predicted cell types were visualized and annotated using the SPOTlight package (v0.1.7) [34].
The spatial transcriptomic dataset GSE198353 used in this study includes only mouse BRCA primary tumor samples and was used to analyze the spatial distribution of cell types within the primary tumor (PT). Due to the current lack of high-quality spatial transcriptomic data for BRCA lymph node metastasis in public databases, the spatial analysis results reflect only the local immune microenvironment of the PT and cannot be directly extrapolated to lymph node metastasis tissues.

BRCA-related the cancer genome atlas (TCGA) data acquisition
The RNA-Seq data of BRCA in TCGA, comprising 1104 tumor and 113 normal tissue samples, were acquired from the UCSC Xena database (https://xena.ucsc.edu/). Perl language was utilized to group the samples, and then the GENCODE Gene Set-09.2019 version annotation file was utilized to convert the ensemble ID of the samples. Ensemble IDs absent from GENCODE, including those of lncRNAs and mRNAs, were removed. As the dataset is publicly accessible, neither ethical approval nor informed consent was necessary [35].

Differential gene expression analysis
Differential expression between sequencing data and TCGA samples was analyzed using the limma package (v3.54.1) in R (v4.2.1). P values were adjusted via the false discovery rate (FDR) method, with FDR < 0.05 as the significance threshold. Heatmaps and volcano plots were generated with pheatmap (v1.0.12) and ggplot2 (v3.4.2), respectively. Group comparisons were conducted using the Wilcoxon test [36].

Enrichment analysis of pathways
Gene Set Enrichment Analysis (GSEA) was performed using the MSigDB gene set c5.go.v2022.1.Hs.symbols.gmt as reference. Functional enrichment of differentially expressed genes (DEGs) was assessed via ConsensusPathDB (v4.6.0). Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were carried out with the R package ClusterProfiler, and results were visualized using ggplot2. The top 10 significant GO/KEGG terms, ranked by adjusted P values, were obtained through enrichGO and enrichKEGG functions [37–39].

Protein–protein interaction (PPI) network analysis
PPI networks for DEGs upregulated in BRCA_LNMT-derived CSCs were generated using the STRING database (https://string-db.org/, version 11.5). The top 50 upregulated DEGs were analyzed with a minimum interaction score of 0.7 to identify major interaction hubs and calculate interaction counts for each gene [40].

Analysis of immune cell correlations
The CIBERSORT algorithm was employed to estimate the relative proportions of 22 immune cell types in BRCA-related TCGA samples. Correlations between gene expression and immune cell abundance were calculated, retaining only results with P < 0.05 [41].

Construction of lentivirus vectors
Lentiviral interference (pSIH1-H1-copGFP, Cat#: SI501A-1, System Biosciences, USA) and overexpression (pCDH-CMV-MCS-EF1α-copGFP, Cat#: CD511B-1, System Biosciences, USA) vectors were used to generate ISG15 knockdown and overexpression constructs. Lentiviral particles were packaged in HEK-293 T cells (Cat#: iCell-h237, Saibaoqiang Biotechnology, Shanghai, China) with a commercial packaging kit (A35684CN, Invitrogen, USA). After 48 h, the supernatant containing lentivirus (1 × 108 TU/mL) was collected. shRNA sequences were as follows: sh-NC, AGGCTACAATGATCAGACTAAT; sh-ISG15-1, CTGAGCATCCTGGTGAGGAAT; sh-ISG15-2, CATGTCGGTGTCAGAGCTGAA [42].

BRCA stem cell sorting and grouping
Mouse BRCA cell lines 4T1 and 4T07 were cultured in DMEM basal medium containing 10% fetal bovine serum (FBS, Cat#: S9020, Solarbio, Beijing, China) in a humidified incubator at 37 °C with 5% CO2. When cells reached 90% confluence, they were collected using 0.25% trypsin–EDTA (Cat#: 25,200,072, Gibco, USA) and neutralized with a complete medium. Cells were centrifuged at 300 × g for 5 min, the supernatant was discarded, and the pellet was resuspended in PBS containing 2% FBS at a final concentration of 1 × 107 cells/mL. CD133-positive cells were then isolated using immunomagnetic beads (Cat#: 130–105–226, Miltenyi Biotec, Germany). 4T1 breast cancer cells were sorted using antibodies against CD133 and CD44 to isolate the breast CSC population. Following fixation and permeabilization, OCT4 antibodies (Cat#: 75,463, CST, USA) were used for immunofluorescence staining or flow cytometry (FCM) to assess the expression of stemness-related transcription factors within the sorted cell population.
The cell suspension was first incubated with CD133-specific antibody (Cat#: ab271092, Abcam, Cambridge, UK), CD44-specific antibody (Cat#: 113,824, BioLegend, USA), and OCT4-specific antibody (Cat#: 75,463, CST, USA) at 4 °C for 30 min. Unbound antibodies were removed by washing with PBS. The cells were then incubated with anti-rabbit IgG-conjugated magnetic beads (Miltenyi Biotec, Germany) at 4 °C for 15 min, and CD133, CD44, and OCT4-positive cells were isolated by magnetic separation. These cells were cultured in serum-free DMEM/F12 medium under suspension conditions.
Finally, stem cell markers (CD133, CD44, OCT4, Nestin) were validated by immunofluorescence, and the populations were designated as BRCA stem cells (4T1-S, 4T07-S). For in vitro assays, freshly sorted cells were used to preserve stemness. Cells were assigned to oe-NC, oe-ISG15, sh-NC, or sh-ISG15 groups, infected with 1 mL lentivirus, and analyzed for infection efficiency by western blot after 48 h [43].

Western blot
Total protein was extracted from cells using RIPA lysis buffer containing PMSF (P0013C, Biyun Tian, Shanghai, China). The samples were incubated on ice for 30 min, followed by centrifugation at 4 °C and 8000 g for 10 min to collect the supernatant. The total protein concentration was determined using the BCA assay kit (Cat#: 23,227, ThermoFisher, USA). Fifty micrograms of protein was dissolved in 2 × SDS loading buffer (Cat#: P0750, Beyotime, China), boiled at 100 °C for 5 min, separated by SDS-PAGE gel electrophoresis (Cat#: P1200, Solarbio, China), and then transferred to a PVDF membrane (Cat#: YA1701, Solarbio, China). The membrane was blocked with 5% skim milk at room temperature (RT) for 1 h, then incubated overnight at 4 °C with primary antibodies against ISG15 (1:1000, ab308219), NESTIN (1:1000, ab221660), OCT4 (1:10,000, ab200834), SOX2 (1:1500, ab92494), STAT3 (1:1500, ab68153), p-JAK2 (1:1500, ab32101), JAK2 (1:5000, ab108596), PD-L1 (1:1000, ab213480), Arg (1:1000, ab203490), CD206 (1:1200, ab64693), CD163 (1:1000, ab182422), CD86 (1:1200, ab220188), and GAPDH (1:2500, ab9485) as the internal reference (all from Abcam, Cambridge, UK), as well as p-STAT3 (1:1000, NB100-82213, Novus Biologicals). Membranes were washed three times with TBST (10 min each), then incubated for 1 h at RT with HRP-conjugated goat anti-rabbit IgG H&L secondary antibody (1:2000, ab97051, Abcam, Cambridge, UK). After additional TBST washes, equal volumes of solutions A and B from the enhanced chemiluminescence kit (Cat#: abs920, ABclonal Technology, Shanghai, China) were applied to the membrane. Signals were captured using a Bio-Rad imaging system (Bio-Rad, USA), and band intensities were quantified with Quantity One v4.6.2 by normalizing to GAPDH. Each experiment was performed in triplicate, and mean values were reported [44].

Immunofluorescence staining
Tissue sections (20 µm thick) were fixed with 4% paraformaldehyde (Sigma-Aldrich, USA) for 10 min and then permeabilized with 0.3% Triton X-100 (Sigma-Aldrich, USA) for 10 min. Subsequently, the sections were blocked with 5% bovine serum albumin (BSA, Sigma-Aldrich, USA) for 1 h. The sections were then incubated overnight at 4 °C with the following primary antibodies: Alexa Fluor® 488 anti-CD133 (1:200, ab252126, Abcam, USA), Alexa Fluor® 555 anti-CD44 (1:200, ab313309, Abcam, USA), Alexa Fluor® 647 anti-EPCAM (1:200, ab313669, Abcam, USA), Alexa Fluor® 647 anti-KRT8 (1:200, ab192468, Abcam, USA), and Alexa Fluor® 647 anti-KRT18 (1:200, ab206269, Abcam, USA). Images were captured using a confocal fluorescence microscope (Leica SP8, Germany) and analyzed with ImageJ software (v1.52, National Institutes of Health, USA) to calculate the proportion of M2-type microglia.

Suspension of stem cell experiment
The 4T1-S and 4T07-S cell lines were seeded into low-adhesion 96-well plates at 2 × 103 cells/mL in serum-free DMEM/F12 (Cat#: 11,320,033, ThermoFisher, USA) enriched with 20 ng/mL EGF (Cat#: 315–09-500UG, Peprotech, USA) and 20 ng/mL FGF-β (Cat#: 450–33-50UG, Peprotech, USA). The medium was replaced every 3 days. After 10 days, cell spheres were imaged and counted using a CKX4L inverted microscope (OLYMPUS, Japan) [45].

Colony formation experiment
The cells digested with trypsin (Cat#: T8150, Beijing Solabio Technology Co., Ltd., Beijing, China) were collected in a sterile 15-mL centrifuge tube. A total of 500 cells were selected and plated on a six-well plate, followed by the addition of 2 mL of 10% FBS culture medium and thorough mixing. The culture medium was replaced every 3 days. The 4T1-S and 4T07-S cells were cultured for 10 days. After aspirating the culture medium, the six-well plates were rinsed twice with PBS. Then, each well was treated with a total of 1 mL of 4% paraformaldehyde (Cat#: P1110, Beijing Solabio Technology Co., Ltd.) for 30 min. After removing the 4% paraformaldehyde, the wells were stained with a 0.1% crystal violet solution (Cat#: C8470, Beijing Solabio Technology Co., Ltd.) for 20 min. After rinsing with purified water, the cells in the six-well plate were photographed [45].

Transwell experiment
The Matrigel (Cat#: 356,234, Haoyang Biological Technology Co., Ltd., Shanghai, China) was thawed at 4 °C overnight, diluted 1:1 with serum-free medium, and 50 μL was added to the upper chamber of an 8-μm Transwell insert (G4740, Solarbio, China) for solidification at 37 °C for 2–3 h.
Cells were digested, counted, adjusted to 2 × 105 cells/mL in serum-free medium, and treated with 10 μM mitomycin C [46]. Then, 200 μL of the suspension was seeded into the upper chamber, while 800 μL of complete medium with 20% FBS was added to the lower chamber. After 24 h at 37 °C, non-invaded cells were removed, and invaded cells were fixed with 4% formaldehyde for 10 min, stained with 0.1% crystal violet for 30 min, rinsed, and counted in at least four random microscopic fields (TE2000, Nikon, China).
Migration assays followed the same protocol without Matrigel coating. The incubation time remained 24 h.

Cell proliferation assay using the CCK-8 method
Cell proliferation experiments were conducted using the CCK-8 assay kit (Cat#: No: CA1210, Beijing Solarbio Science & Technology Co., Ltd., Beijing, China). Cells in the logarithmic growth phase were seeded in 96-well plates at 1 × 104 cells/well and cultured for 24 h. After grouping-based transfection, 10 μL of CCK-8 reagent was added at 0, 24, 48, and 72 h, followed by a 3-h incubation at 37 °C. Absorbance at 450 nm was measured with a microplate reader, and proliferation curves were plotted based on the absorbance values [47].

Co-culture system
4T1-S and 4T07-S cells were transfected and incubated for 24 h, after which the conditioned medium (CM) was collected for co-culture with mouse macrophages (Raw264.7, Cat#: MZ-2039, Mingzhou Biotechnology, China).
Peripheral blood mononuclear cells (PBMCs) were obtained from the whole blood of the mouse BRCA orthotopic tumor model via Ficoll density gradient centrifugation. Specifically, the whole blood was first diluted in PBS at a 1:1 ratio and carefully layered over Ficoll-Paque solution (density 1.077 g/mL, GE Healthcare; Cat#: F4375, Sigma, USA) without mixing. This mixture was centrifuged at 400 × g for 30–40 min at RT. Four layers were obtained: plasma, PBMCs, Ficoll, and erythrocytes. The PBMC layer was carefully collected, transferred to a clean tube, and washed twice with PBS to remove residual Ficoll.
Subsequently, T cells were purified from the isolated PBMCs using magnetic beads conjugated with the CD3 antigen (Cat#: 11151D, ThermoFisher, USA). The transfected 4T1-S or 4T07-S cells were then co-cultured with the purified T cells, with the optional addition of a 10-μM JAK/STAT pathway inhibitor, WP1066 (Cat#: HY-15312, MCE, China), for 24 h, as needed for further experiments [48].
In the 4T1-S or 4T07-S and T cell co-culture system, PD-1/PD-L1 signaling was blocked by adding 10 μg/mL anti-PD-L1 (ab213480, Abcam, UK) and anti-PD-1 (ab214421, Abcam, UK) antibodies. Anti-IgG (ab172730, Abcam, UK) served as an isotype control [49]. The FCM gating strategy is shown in Additional file 2: Fig. S1.

Detection of macrophage markers using FCM
FCM was applied to detect surface markers of macrophages. Initially, the suspension of macrophages was resuspended in a staining buffer (Cat#: No: abs9475, Aibixinxin Biotechnology Co., Ltd., Shanghai, China). Subsequently, the cells were stained and incubated with F4/80_PE antibody (NB600-404PE, Novus Biologicals), CD206_FITC antibody (FAB25351F, Novus Biologicals), and CD86_FITC (NB100-65255, NovusBio). CD206+ cells were defined as M2 macrophages, while CD86+ cells were identified as M1 macrophages. Stained samples were analyzed utilizing a flow cytometer, and data were processed with FlowJo software [48].

ELISA experiment
IL-10 levels in culture supernatants were quantified using an ELISA kit (ab255729, Abcam, Cambridge, UK). After equilibrating reagents to RT for 20 min, standards and samples (100 μL/well) were loaded onto the plate. Plates were incubated at 37 °C for 90 min, washed, and sequentially incubated with 100 μL biotinylated antibody (60 min, 37 °C), enzyme-conjugated reagent (30 min, 37 °C, protected from light), and substrate solution (15 min, 37 °C, protected from light), with washes between steps. Optical density was measured at 450 nm using a BioTek Synergy 2 microplate reader, and IL-10 concentrations were determined from a standard curve [50].

T cell proliferation and activity assays
Collected T cells following co-culture or treatment with the JAK/STAT pathway inhibitor WP1066 for 24 h were labeled with anti-CD8_FITC (NBP1-49045F, NovusBio) to identify CD8+ subsets. Cells were then fixed, permeabilized, and stained with anti-Ki-67_PE (NB500-170PE, NovusBio) and anti-IFN-γ_PerCP (FAB485C, NovusBio) antibodies. Proliferation and functional activity were analyzed using a BD FACS Canto™ II flow cytometer (BD Immunocytometry Systems).

Dual-luciferase assay
PD-L1 promoter activity was assessed utilizing a dual-luciferase reporter system. Vectors containing either the wild-type PD-L1 promoter sequence (PD-L1-WT: TCCCAGGAAG) or a mutant promoter with altered transcription factor binding sites (PD-L1-MUT: AGGGTCCTTC) were transfected into cells with Lipofectamine 2000 (11,668,019, ThermoFisher, USA). Renilla luciferase was used as an internal control. After 48 h, cells were lysed, and firefly and Renilla luciferase activities were measured using a luciferase assay kit (K801-200, Biovision, USA) on a dual-luciferase reporter analysis system (Promega, Madison, WI, USA). Promoter activity was expressed as the firefly/Renilla RLU ratio [51].

ChIP experiment
Chromatin immunoprecipitation (ChIP) was performed using a commercial kit (Cat#: KT101-02, SAIK Biological Technology Co., Ltd., Guangzhou, China) to assess p-STAT3 binding to the PD-L1 promoter. Cells (70–80% confluent) were cross-linked with 1% formaldehyde for 10 min at RT, then sonicated (10 s on/10 s off, 15 cycles) to obtain appropriately sized DNA fragments. After centrifugation (13,000 rpm, 4 °C), supernatants were divided into two aliquots and incubated overnight at 4 °C with either rabbit anti-IgG (1:100, ab172730, Abcam, UK) as a negative control or rabbit anti-p-STAT3 (ab30647, Abcam, UK). Immune complexes were captured with Protein Agarose/Sepharose, washed, and reverse cross-linked at 65 °C overnight. DNA was purified by phenol–chloroform extraction and analyzed for PD-L1 promoter enrichment via PCR using primers: forward primer: 5′-AGTACCTTGCTTCGGCAGAG-3′, reverse primer: 5′-GAGCGCTTAAGGGAGCCAAA-3′ [52].

Murine subcutaneous BRCA orthotopic model
4T1-S cells from the sh-NC and sh-ISG15 groups were prepared at 2 × 105 cells/mL, and 0.2 mL of the suspension was injected subcutaneously into the left axillary region of BALB/c nude mice (n = 6 per group) using a 1-mL syringe. Mice were housed under SPF conditions, and tumor growth was recorded on days 4, 8, 12, 16, 20, and 24. On day 25, mice were euthanized by cervical dislocation, and tumors were excised and weighed [53].

Mouse mammary carcinoma in situ tumor model
4T1-S cells (sh-NC or sh-ISG15) were prepared at 2 × 105/mL, and 0.2 mL of the suspension was injected into the left fourth mammary fat pad of BALB/c mice. After 4 weeks, mice were euthanized via cervical dislocation, and inguinal lymph nodes were harvested to assess metastasis. Tumors were dissociated into single-cell suspensions for flow cytometric analysis of M2 macrophages and T cells [29, 54, 55].

Isolation of mouse tumor stem cells
Single-cell suspensions were prepared from mouse orthotopic tumor tissues, followed by isolation of CD133-positive cells using immunomagnetic beads (Cat#: 130–105–226, Miltenyi Biotec, Germany). First, the cell suspension was incubated with CD133-specific antibody (Cat#: ab271092, Abcam, Cambridge, UK), CD44-specific antibody (Cat#: 113,824, BioLegend, USA), and OCT4-specific antibody (Cat#: 75,463, CST, USA) at 4 °C for 30 min. The cells were then washed with PBS to remove unbound antibodies. Next, the cells were incubated with anti-rabbit IgG-conjugated magnetic beads (Miltenyi Biotec, Germany) at 4 °C for 15 min, and CD133, CD44, and OCT4-positive cells were isolated through magnetic separation. The isolated positive cells were resuspended in serum-free DMEM/F12 medium for further culture. All downstream experiments were performed within 6 h after cell sorting to ensure that cells had not yet entered the proliferation cycle, thereby minimizing the confounding effects of cell division on experimental outcomes.

Immunohistochemistry
BRCA orthotopic tumor tissues and lymph nodes were fixed in 10% formalin, deparaffinized twice in xylene (10 min each), and rehydrated through graded ethanol (100%, 95%, 75%, 50%; 10 min each). Antigen retrieval was performed in 0.01 mol/L citrate buffer by microwave heating for 20 min. After cooling, sections were blocked with 5% goat serum at RT for 5 min, followed by incubation with primary antibodies overnight at 4 °C: anti-CD133 (1:200, ab222782, Abcam, USA), anti-CD31 (1:200, ab28364, Abcam, USA), anti-CD206 (1:200, ab64693, Abcam, USA), anti-CD8 (NBP2-29,475, Novus Bio), anti-CD44 (1:200, ab243894, Abcam, USA), and anti-Ki-67 (1:200, ab15580, Abcam, USA).
The following day, sections were incubated at 37 °C for 1 h, followed by the addition of biotinylated goat anti-rabbit secondary antibody (1:500, ab150077, Abcam, USA) and incubation at 37 °C for 30 min, and developed with freshly prepared DAB solution (Cat#: DA1015, Solarbio, Beijing, China) for 1–2 min. Nuclei were counterstained with hematoxylin for 1 min, and sections were dehydrated, cleared, and mounted with neutral gum. Five random high-power fields were examined under a light microscope [56].

H&E staining
Mouse lymph nodes with primary BRCA lesions were fixed in 10% neutral formalin at RT for 24 h, followed by dehydration through a graded ethanol series (75%, 85%, 95%, and 100%, each for 1 h) and clearing in xylene (twice, 15 min each). The tissues were embedded in paraffin and sectioned at a thickness of 4 μm. Sections were dewaxed with xylene (twice, 10 min each) and rehydrated through a reverse ethanol gradient (100%, 95%, 85%, 75%, 2 min each). Hematoxylin staining (Cat#: H9627, Sigma-Aldrich, USA) was performed for 5 min at RT, followed by bluing under running water. Eosin counterstaining (Cat#: E4009, Sigma-Aldrich, USA) was applied for 2 min at RT. After light rinsing, sections were dehydrated again with graded ethanol and cleared in xylene (twice, 5 min each), then mounted with neutral resin (Cat#: 03989, Sigma-Aldrich, USA). Tissue morphology and lymph node metastases were evaluated and imaged under a light microscope (BX53, Olympus, Japan) by two independent pathologists [57].

Transcriptome sequencing of BRCA in situ tumor tissue
BRCA in situ tumor tissue was collected for RNA high-throughput second-generation sequencing. Initially, total RNA was extracted utilizing Trizol reagent (15,596,026, Invitrogen, Carlsbad, CA, USA). The concentration and purity of RNA samples were then determined using the Nanodrop 2000 spectrophotometer (1011U, Nanodrop, USA). Samples with RNA integrity number (RIN) ≥ 7.0 and 28S:18S ratio ≥ 1.5 were retained.
Sequencing libraries were prepared by CapitalBio Technology (Beijing, China) using 5 μg of RNA per sample. rRNA was removed with the Ribo-Zero™ Magnetic Kit (MRZE706, Epicentre Technologies, USA), and libraries were constructed with the NEBNext Ultra RNA Library Prep Kit (#E7775, NEB, USA). RNA was fragmented (~ 300 bp) in NEBNext First Strand Synthesis Buffer (5 ×), and first-strand cDNA was synthesized with reverse transcriptase and random primers, followed by second-strand synthesis in buffer containing dUTP (10 ×). cDNA fragments were end-repaired, poly(A)-tailed, and ligated to Illumina adapters. Strand-specific libraries were generated by USER Enzyme (#M5508, NEB, USA) digestion, PCR-amplified, and purified. Library quality was assessed using an Agilent 2100 Bioanalyzer and quantified with the KAPA Library Quantification Kit (KK4844, KAPA Biosystems). Sequencing was performed on an Illumina NextSeq CN500 (paired-end).
Raw reads were quality-checked with FastQC v0.11.8 and processed with Cutadapt v1.18 to remove adapters and poly(A) sequences. Reads containing > 5% N bases were discarded with Perl scripts. The FASTX Toolkit v0.0.13 was used to retain reads with ≥ 70% bases having quality scores > 20, and BBMap was applied for paired-end repair. Clean reads were aligned to the reference genome using HISAT2 v0.7.12 [58].

Statistical analysis
All statistical analysis of the data was conducted utilizing GraphPad Prism 9 software and R programming language. Descriptive statistics were initially performed to determine the central tendency and dispersion of the data. The normality of the dataset was examined using the Shapiro–Wilk and Kolmogorov–Smirnov tests. For normally distributed data, unpaired Student’s t-test was used for comparing between two groups, while non-normally distributed data were analyzed utilizing the Mann–Whitney U test or Wilcoxon signed-rank test. For comparing multiple groups, one-way ANOVA was utilized, followed by tests for homogeneity of variances and Tukey’s post hoc test. Multiple hypothesis testing, such as in gene expression analysis, was controlled by Bonferroni correction or the FDR method. Survival analysis for time-to-event data was performed using the Kaplan–Meier method and Cox proportional hazards model. The significance level for all tests was set at P < 0.05, and effect size and confidence intervals were reported alongside significant test results to provide practical significance of the findings.

Results

BRCA metastasis: single-cell sequencing reveals microenvironment heterogeneity and potential roles of CSCs
BRCA is one of the most prevalent malignancies in women worldwide, and metastasis is a significant prognostic factor associated with poor outcomes. Recent advances in scRNA-seq have enabled high-resolution analysis of the TME, providing insights into immune heterogeneity and mechanisms underlying metastasis [59]. To investigate these mechanisms and identify potential therapeutic targets, we performed scRNA-seq on primary BRCA tumors (BRCA_PT) and matched lymph node metastatic tumors (BRCA_LNMT) (Fig. 1A).
The cohort consisted of two primary BRCA samples (BRCA_PT; N1/2) and two lymph node metastatic BRCA samples (BRCA_LNMT; N3/4). scRNA-seq data were processed in Seurat, with quality control and normalization applied. Post-filtering, the correlation between nCount and percent.mt was − 0.12, and between nCount and nFeature was 0.9 (Additional file 2: Fig. S2A), indicating good cell quality. PCA on the top 2000 highly variable genes (RunPCA; Additional file 2: Fig. S2B) showed no detectable batch effects across the four samples (Additional file 2: Fig. S2C). The JackStrawPlot function was then utilized for visualizing the top 40 PCs (Additional file 2: Fig. S2D), and the leading genes for the first six PCs are shown in Additional file 2: Fig. S2E and F.
Uniform manifold approximation and projection (UMAP) identified 18 cell clusters (Fig. 1B), which were annotated based on canonical marker genes into six cell types: monocytes, B cells, T cells, NK cells, dendritic cells (DCs), and CSCs (Fig. 1C and D, Additional file 2: Fig. S2G). Comparative analysis of cell type proportions revealed substantial compositional differences between BRCA_PT and BRCA_LNMT (Fig. 1E and F). Although the same six cell categories were present in both groups, their relative abundances varied substantially. For example, cancer stem cells and monocytes were more abundant in BRCA_LNMT, whereas T cells were less represented compared with BRCA_PT. We further applied CytoTRACE2 to assess stemness in tumor-derived cells from BRCA_PT and BRCA_LNMT. The results demonstrated that BRCA_LNMT exhibited a significantly higher CytoTRACE2 stemness score than BRCA_PT, indicating a greater proportion of stem-like cells within lymph node metastases. According to the CytoTRACE2_Potency classification, we observed that cells with higher stemness potential—namely oligopotent and multipotent populations—were predominantly localized within regions corresponding to the CSC cluster (Additional file 2: Fig. S2H). Importantly, the spatial distribution of high-stemness regions inferred from CytoTRACE2 was highly consistent with the CSC cluster defined by canonical stem cell markers, further validating the accuracy of our CSC annotation from a computational transcriptomics perspective.
Moreover, multiplex immunofluorescence staining of classical epithelial markers EPCAM, KRT8, or KRT18 co-stained with CD133 or CD44 showed a lack of stromal or epithelial components. These results are consistent with our scRNA-seq findings, which did not detect stromal or epithelial cells in the samples (Fig. 1G and Additional file 2: Fig. S2I). Notably, BRCA_LNMT exhibited a higher abundance of CSCs and monocytes, while the proportion of T cells was lower. CSCs are widely recognized as a major challenge in cancer therapy due to their abilities to initiate tumors, promote metastasis, resist chemotherapy, and self-renew [60]. Based on our scRNA-seq findings and previous studies, we propose that CSCs may contribute to BRCA lymph node metastasis, potentially through crosstalk with immune cells in the TME.

ISG15 promotes self-renewal and metastatic capability of BRCA stem cells
To further investigate the potential molecular mechanisms by which CSCs regulate lymph node metastasis in BRCA at the genetic level, we analyzed DEGs in CSCs from BRCA_PT and BRCA_LNMT tissue samples. We identified 40 DEGs (Additional file 2: Fig. S3A; Additional file 1: Table S1), including 31 upregulated and 9 downregulated genes. Protein interaction analysis of the 31 upregulated genes in BRCA_LNMT CSCs revealed that the top genes were B2m, Irf7, Isg15, and Ifit3 (Fig. S3B). Bubble plot analysis confirmed that ISG15 expression was markedly elevated in CSCs_BRCA_LNMT. Although expressed in a smaller subset of CSCs (smaller dots), these cells exhibited high expression levels (darker color), indicating strong ISG15 upregulation within this subpopulation (Fig. 2A).
To validate ISG15 expression in CSCs, CD133+ subpopulations from 4T1 and 4T07 cells were isolated via FCM. The specific gating strategy is shown in Fig. 2B. Immunofluorescence staining of the sorted cells confirmed the expression of CSC markers CD133, CD44, Nestin, and OCT4 (Fig. 2C and Additional file 2: Fig. S3C), indicating successful sorting of 4T1-S and 4T07-S cells. Notably, CD44+ CSCs predominated in primary tumors, whereas CD133+ CSCs were more prevalent in LNMT tissues, suggesting site-specific CSC subtype heterogeneity.
Gene expression analysis revealed that ISG15 was most highly expressed in CSCs (Additional file 2: Fig. S3D). Western blot analysis of CD133+ and CD133− CSCs (sorted via immunomagnetic separation) revealed a significantly elevated expression of ISG15 in CD133+ cells, indicating that ISG15 primarily functions in this CSC subpopulation. In contrast, bulk tumor cells (unfractionated) exhibited intermediate ISG15 levels, falling between those of CD133+ and CD133− cells (Fig. 2D). Additionally, compared to unbound control cells, the expression of CSC markers CD133, Nestin, CD44, and OCT4 was markedly upregulated in 4T1-S cells (Fig. 2E). Previous studies have linked ISG15 to poor BRCA prognosis [61]. However, research on the specific role and mechanism of ISG15 in CSCs remains limited and warrants further study.
Next, lentivirus-based manipulation of ISG15 expression in 4T1-S and 4T07-S cells was performed. Western blot analysis showed (Fig. 2F and G and Additional file 2: Fig. S3E and F) that ISG15 expression was significantly increased in 4T1-S and 4T07-S upon overexpression and significantly decreased upon silencing, with sh-ISG15-1 exhibiting the strongest knockdown effect, which was selected for subsequent experiments.
Subsequently, western blot, sphere formation, and colony formation assays were used to assess the self-renewal capacity of 4T1-S and 4T07-S cells. Results showed (Fig. 2H–J and Additional file 2: Fig. S3G and H) that overexpression of ISG15 significantly increased the expression of stemness markers NESTIN, OCT4, and SOX2, along with a marked increase in both sphere and colony number, indicating enhanced self-renewal capacity of 4T1-S and 4T07-S cells. In contrast, ISG15 silencing significantly inhibited stem cell phenotypes, such as sphere and colony formation.
CCK-8 and Transwell assays were used to examine proliferation, migration, and invasion capacities. ISG15 overexpression significantly promoted proliferation, migration, and invasion in both 4T1-S and 4T07-S cells, while ISG15 silencing substantially inhibited these properties (Fig. 2K and L and Additional file 2: Fig. S3I–K).
Collectively, these results indicate that high ISG15 expression promotes self-renewal and migratory/invasive capacity of 4T1-S and 4T07-S cells, implicating ISG15 as a key contributor to BRCA lymph node metastasis.

The role and mechanisms of CSC–immune cell communication in BRCA metastasis
The crosstalk between CSCs and immune cells within the TME is critical for primary tumor growth, metastatic progression, and immune evasion [62]. In this study, we utilized the “CellChat” package to analyze cellular communication between CSCs and non-CSCs with different cell types. The analysis uncovered a complex network of intercellular interactions in BRCA_LNMT tissue, particularly highlighting strong signaling between CSCs and monocytes (Additional file 2: Fig. S4A and B, Fig. 3A). Additionally, compared to the BRCA_PT group, the communication intensity between CSCs and monocytes was significantly enhanced in the BRCA_LNMT group (Fig. 3B).
Furthermore, we performed clustering analysis on spatial transcriptomic data of mouse BRCA tissue using the Seurat package (Fig. 3C, Additional file 2: Fig. S4C). Integration with scRNA-seq data enabled the annotation of ST-seq cell populations and the inference of spatial enrichment patterns (Fig. 3D). The distribution of CSCs and monocytes is illustrated in Additional file 2: Fig. S4D. Utilizing the “SPOTlight” package in R, we obtained information on the spatial interactions between cells and generated circular plots depicting the strength of the intercellular interactions (Fig. 3E), revealing the strongest interaction between CSCs and monocytes.
Monocytes, circulating white blood cells of the innate immune system, infiltrate tissues and differentiate into macrophages or DCs [63]. Subgroup analysis of monocytes identified 14 distinct clusters annotated as macrophages and DCs (Fig. 3F and G, Additional file 2: Fig. S5A), with macrophages predominating. Based on this finding, we hypothesize that the cellular communication between CSCs and monocytes primarily occurs between CSCs and macrophages.
Macrophages infiltrating tumor tissues are referred to as TAMs. These TAMs consist of two subtypes, M1 and M2, with the M2 subtype playing a significant role in promoting tumor growth and metastasis [64]. By annotating macrophage subtypes using M1 and M2 macrophage marker genes (Additional file 2: Fig. S5B), we observed that the majority of cells belonged to the M2 subtype, and TAMs in the BRCA_LNMT tissue predominantly exhibited the M2 subtype (Fig. 3H–J). Integrating these findings with our cell–cell communication analysis, we propose that CSC–TAM interactions may promote M2 macrophage polarization, thus contributing to tumor growth and metastasis. Additionally, we performed immunohistochemical and flow cytometric analyses on BRCA_LNMT and BRCA_PT tissues and found that the proportion of CSCs, particularly CD44+ cells, was higher compared to primary tumor tissues (Fig. 3K), validating the results of single-cell transcriptome sequencing. Furthermore, the proportion of M2 macrophages increased (Fig. 3L), while the proportion of T cells decreased significantly (Fig. 3M and N).
It is important to note that the spatial transcriptomic dataset GSE198353 used in this study was derived from primary BRCA tumors. Therefore, the observed spatial co-localization of CSCs and monocytes reflects only potential interaction patterns in PT tissues. For LNM tissues, immune interaction characteristics were verified through a combination of scRNA-seq, FCM, and immunohistochemistry, enhancing the comprehensiveness of the analysis.

Molecular mechanisms of ISG15 promoting macrophage M2 polarization in 4T1-S and the role of IL10
To explore how CSCs influence macrophage polarization via intercellular communication and to investigate the underlying molecular mechanisms, Raw264.7 macrophages were stimulated in vitro using CM derived from 4T1-S cells (4T1-S_CM). Western blot analysis (Fig. 4A) demonstrated that overexpression of ISG15 in 4T1-S cells markedly increased the expression of M2 macrophage markers (Arg1, CD163, CD206) and decreased M1 marker CD86. Conversely, ISG15 knockdown decreased M2 marker expression and elevated CD86 levels. FCM (Fig. 4B and C) confirmed that ISG15 overexpression promoted M2 macrophage enrichment and suppressed M1 macrophage populations, while ISG15 silencing produced the opposite effect. These findings indicate that ISG15 in 4T1-S cells facilitates macrophage polarization toward the M2 phenotype.
Previous studies have reported that ISG15 can enhance the secretion of IL-10 [65], and IL-10 can induce TAMs to polarize into M2 macrophages [64]. Consistent with these reports, ELISA quantification of IL-10 in 4T1-S_CM demonstrated that ISG15 overexpression significantly elevated IL-10 levels, whereas ISG15 knockdown significantly reduced IL-10 secretion (Fig. 4D). These results suggest that ISG15 in 4T1-S can promote the secretion of IL10.
Based on the aforementioned findings, we hypothesize that ISG15 in 4T1-S may promote macrophage M2 polarization by enhancing IL-10 secretion (Fig. 4E). Given that CSC-conditioned media likely contain multiple cytokines, including TGF-β, CSF1, and IL-4, follow-up cytokine array profiling could help identify key drivers of M2 polarization and determine whether ISG15 indirectly regulates IL-10/TGF-β expression to mediate its effects.

High expression of ISG15 in BRCA tumor tissue correlates with M2 macrophage polarization and T cell activation
Based on these results, we further examined the involvement of ISG15 in CSCs and its influence on the immune landscape of BRCA using transcriptomic data from the TCGA-BRCA cohort. Gene differential expression analysis (Additional file 2: Fig. S6A) revealed that BRCA tumor tissue exhibited 614 significantly upregulated and 1075 significantly downregulated genes compared to normal tissue, with ISG15 among the most markedly upregulated genes (Additional file 2: Fig. S6B). Based on the median ISG15 expression level, BRCA tumor samples were stratified into high and low ISG15 expression groups. Comparison between these groups (Additional file 2: Fig. S6C) identified 52 significantly upregulated and 24 significantly downregulated genes in the high-ISG15 group. Further functional enrichment analysis of these DEGs (Additional file 2: Fig. S6D) indicated their significant enrichment in immune regulation-related functions, such as “Immune System,” “Cytokine Signaling in the Immune system,” and “cell-specific immune response.” Moreover, the results of GSEA enrichment analysis (Additional file 2: Fig. S6E) showed a strong association between elevated ISG15 expression and pathways involved in immune response regulation.
Immune infiltration analysis (Additional file 2: Fig. S6F–H) revealed that BRCA tumors with high ISG15 expression contained a significantly higher proportion of M2 macrophages and a reduced proportion of M1 macrophages, consistent with our experimental finding that ISG15 promotes M2 polarization in CSCs. Additionally, radar plot analysis (Additional file 2: Fig. S6G) showed that high ISG15 expression samples exhibited elevated proportions of various immune regulatory subtypes, including activated NK cells, CD8+ T cells, and DCs. Notably, the increase in immunosuppressive cells such as M2 macrophages, Tfh cells, and regulatory T cells (Tregs) supports our hypothesis that ISG15 promotes immune tolerance and M2 polarization, thereby contributing to lymph node metastasis in BRCA. Moreover, a pronounced decrease in total T cell abundance was observed in the high-ISG15 group, suggesting that ISG15 may influence immune escape in the tumor microenvironment by modulating immune cells.

Silencing of ISG15 inhibits BRCA growth and lymph node metastasis: the role of immune regulation
We further performed in vivo experiments in 6–8-week-old female BALB/c mice, establishing subcutaneous xenograft and orthotopic mammary fat pad tumor models by injecting 4T1-S cells with different ISG15 interventions into the dorsal flank or the left fourth mammary fat pad, respectively. Tumor growth monitoring in the orthotopic model (Fig. 5A–C) showed that ISG15 silencing markedly inhibited tumor progression, as indicated by significant reductions in tumor volume and weight. Immunohistochemical analysis of lymph node metastasis in orthotopic BRCA models (Fig. 5D) demonstrated that silencing ISG15 significantly reduced the number of lymph node. This finding suggests that ISG15 knockdown inhibits both BRCA growth and lymph node metastasis, providing evidence for its potential role in BRCA progression.
Flow cytometric analysis of immune cell subsets in orthotopic tumors demonstrated (Fig. 5E and F) that ISG15 silencing reduced the number of M2 macrophages and the expression of their marker CD206, while increasing the number of M1 macrophages and the expression of their marker CD86. Additionally, the number of T cells increased significantly, with higher expression of CD8 (Fig. 5G) and IFN-γ (Fig. 5H). Immunohistochemical results (Fig. 5I) corroborated these findings, showing decreased CD206+ cells and increased CD8+ cells in ISG15-silenced tumors. Collectively, these results suggest that ISG15 knockdown suppresses BRCA progression by inhibiting M2 macrophage polarization and promoting T cell activation, thereby reducing immune evasion and limiting both primary tumor growth and lymph node metastasis (Fig. 5J).

ISG15 regulation of the JAK-STAT signaling pathway influences BRCA metastasis and immune escape
To further elucidate the mechanisms by which ISG15 silencing affects BRCA growth and metastasis, we utilized transcriptome sequencing to analyze the transcriptional changes in situ tumor tissue from mouse mammary cancer. The results revealed (Fig. 6A) that silencing ISG15 led to significant upregulation of expression in 481 genes, while the expression of 330 genes was significantly downregulated. A heatmap displays the top 50 downregulated genes (Fig. 6B). GO enrichment analysis (Fig. 6C) revealed that these DEGs were enriched in immune regulation–related processes, including “alpha–beta T cell activation negative” and “negative regulation of the immune system.” KEGG pathway analysis (Fig. 6D) identified prominent enrichment in the JAK–STAT signaling axis and the PD-L1 expression/PD-1 checkpoint pathway in cancer. As PD-L1 binding to PD-1 on T cells drives exhaustion and promotes tumor immune evasion [66], and considering the availability of clinically validated PD-1/PD-L1 inhibitors, we focused on this pathway for further validation. Notably, ISG15 silencing led to a marked downregulation of PD-L1 expression (Fig. 6E), suggesting a link between ISG15 and PD-L1–mediated T cell suppression. Additionally, studies have shown that ISG15 can activate the JAK/STAT signaling, and in silico motif analysis identified a STAT3 binding site (TCCCAGGAAG) within the PD-L1 regulatory region (Fig. 6F).
In conclusion, it is hypothesized that ISG15 may upregulate PD-L1 expression by modulating the JAK-STAT signaling pathway, thereby affecting T-cell activation and mediating immune escape.

ISG15 promotes immune escape by suppressing T cell activation and upregulating PD-L1 expression
To verify our hypothesis, we performed co-culture experiments with 4T1-S cells and T cells in vitro. FCM analysis (Fig. 7A and B) showed that ISG15 overexpression markedly suppressed T cell proliferation and activity, whereas ISG15 silencing significantly restored both parameters. These findings indicate that ISG15 in 4T1-S cells can inhibit T-cell activation.
Western blot analysis (Fig. 7C) revealed that ISG15 overexpression increased phosphorylation of JAK2 and STAT3 and elevated PD-L1 protein levels, while ISG15 silencing had the opposite effect. Dual-luciferase assays (Fig. 7D) confirmed that ISG15 overexpression enhanced PD-L1 promoter activity, whereas knockdown reduced it. ChIP assay results (Fig. 7E) showed that in both in vitro cultured 4T1-S cells (left panel) and CSCs isolated from the mouse tumor model (right panel), ISG15 overexpression significantly increased p-STAT3 enrichment on the PD-L1 promoter, while ISG15 silencing significantly reduced it. These findings indicate that ISG15 in 4T1-S cells can promote PD-L1 transcription through activation of the JAK–STAT signaling axis.
To block the effects of PD-L1 or PD-1, anti-PD-L1 or anti-PD-1 treatment was introduced into the co-culture medium of 4T1-S cells and T cells. Experimental results (Fig. 7F and G) showed that T cell proliferation and activity were markedly reduced in the oe-ISG15 + anti-IgG group compared with the oe-NC + anti-IgG control. Importantly, both anti-PD-L1 and anti-PD-1 treatments significantly restored T cell function in the ISG15-overexpressing setting, indicating that blockade of the PD-L1/PD-1 axis can reverse ISG15-mediated T cell suppression. To test the upstream signaling pathway, we treated the co-culture with WP1066, a JAK2/STAT3 inhibitor. WP1066 treatment significantly enhanced T cell proliferation and activity in the ISG15-overexpressing system (Fig. 7H and I) and markedly reduced PD-L1 expression levels (Fig. 7J).
Overall, these results suggest that ISG15 promotes immune escape by activating the JAK–STAT3 pathway to upregulate PD-L1, thereby suppressing T cell activation (Fig. 7K).

Discussion
BRCA is a prevalent malignancy, with lymph node metastasis serving as a critical prognostic factor that influences therapeutic decisions and patient survival [4, 67, 68]. Clarifying its underlying mechanisms is essential for improving clinical outcomes. This study explored the function of CSCs in BRCA lymph node metastasis and investigated their interactions with the immune microenvironment. Using a mouse BRCA model, we collected primary tumor tissue (BRCA_PT) and lymph node metastatic tissue (BRCA_LNMT). ScRNA-seq was applied to characterize the cellular landscape and transcriptomic features of the TME. In conjunction with in vitro cell experiments, we examined how specific cytokines and immune cell subsets influence CSC characteristics and metastatic capacity.
Our findings reveal that specific immune cell types, particularly TAMs, play a key role in maintaining and proliferating CSCs. In the early stages of lymph node metastasis, CSCs showed enhanced proliferation and self-renewal, closely linked to immune microenvironmental alterations. These findings reveal intricate CSC–immune cell crosstalk in BRCA metastasis and provide a basis for developing targeted therapeutic strategies.
Previous research has shown that CSCs are essential in tumor progression and metastasis [69–71]. Our study confirmed an elevated proportion of CSCs in lymph node metastasis samples, highlighting their importance in this process. Analysis of BRCA_LNMT samples revealed significantly increased expression of CSC surface markers such as CD44, indicating a potential role for these cells in driving lymph node metastasis. Prior studies have demonstrated that CSCs possess self-renewal, migratory, and invasive abilities; our in vitro experiments further confirmed these functions [72–75]. We found that under specific conditions, CSCs exhibited strong proliferative and migratory abilities, especially when interacting with immune cells.
The immune microenvironment is a critical regulator of metastasis [76–78]. Notably, the interactions between glioma stem cells (GSCs) and non-GSCs are complex, with bidirectional transformation occurring as the inflammatory stroma, characterized by elevated NF-kB signaling, leads non-GSCs to acquire a GSC-like phenotype. This plasticity, potentially induced by stressors like immunotherapy, allows non-GSCs to mimic GSCs, aiding immune evasion and fostering resistance to immunotherapy [79, 80]. These findings suggest that dynamic CSC–immune cell crosstalk is a central mechanism in tumor progression.
Our findings highlight the critical role of CSC–immune cell interactions in BRCA lymph node metastasis, particularly CSC-mediated polarization of TAMs toward the M2 phenotype, fostering an immunosuppressive niche that supports tumor growth and metastasis [18, 69, 81]. Additionally, chemotherapy and radiotherapy significantly impact immune cell populations and CSC proportions within the TME. For instance, chemotherapy promotes M2 macrophage polarization, supporting an immunosuppressive environment that accelerates tumor recurrence and metastasis [82, 83]. Radiotherapy, although activating antitumor M1 macrophages, may also promote chronic inflammation, potentially advancing tumor progression [84]. Both therapies can lead to T cell exhaustion, potentially through the PD-1/PD-L1 pathway [85, 86].
In our model, lymph node metastases displayed reduced T cell and elevated monocyte/macrophage proportions, differing from some scRNA-seq studies reporting higher T and B cell levels [87]. This discrepancy may result from the high heterogeneity of the TME, with immune cell distribution varying significantly between cancer patients, models, and metastatic sites [88]. The highly invasive 4T1 mouse BRCA model used in this study tends to exhibit an immunosuppressive environment with increased monocyte and macrophage roles and relatively reduced T cell proportions [89]. Additionally, while lymph nodes are generally T and B cell-rich immune organs, tumor cells may modify the lymph node microenvironment to favor the accumulation of specific immune cells, such as macrophages, thereby enhancing immunosuppressive functions and facilitating tumor metastasis [90]. Differences in single-cell sequencing classification standards and analytical methods may also contribute to varying results [91].
ISG15 has been extensively studied in immune regulation and tumor progression [92–94]. Single-cell annotation revealed a decreased proportion of T cells and an increased proportion of monocytes/macrophages in LNMT tissues, indicative of an immunosuppressive trend. Furthermore, In BRCA, ISG15 overexpression correlated with M2 macrophage polarization and T cell suppression, extending its known functions to immune microenvironment modulation, metastasis promotion, and immune evasion. In cell-based experiments, high ISG15 expression promoted CSC self-renewal and migration/invasion abilities, aligning with existing literature on ISG15’s role in cancer [22, 25]. While ISG15’s role in cancer immune responses has been investigated, its specific function in CSCs remained relatively unexplored until now. Our study fills this gap, showing that ISG15 enhances BRCA metastasis by regulating CSC self-renewal and invasion.
The interaction between CSCs and immune cells is critical during BRCA metastasis. Our results support previous findings that ISG15 promotes M2 macrophage polarization, suppressing effective antitumor immunity and advancing BRCA metastasis [20]. Communication analysis further validated the interaction between CSCs and M2 macrophages, highlighting ISG15’s role in this process and supporting its dual function in BRCA metastasis and immune regulation.
Given its multifaceted functions, ISG15 emerges as a promising therapeutic target. Inhibiting ISG15 or disrupting its interaction with macrophages could restore antitumor immunity and limit metastasis. Previous studies have shown that ISG15 regulates immune responses through interactions with signaling molecules such as STAT1/STAT2, IRF9, and USP18 [95, 96]. Additionally, as a secreted interferon-inducible factor, ISG15 may influence T cell activation and macrophage function by modulating the cytokine network [97, 98]. Although our study did not directly identify molecular targets of ISG15, future work utilizing co-immunoprecipitation and proteomics approaches will help elucidate its downstream signaling network. This study provides a solid foundation for further development of ISG15 as a therapeutic target. Future clinical studies should explore ISG15’s specific role in BRCA metastasis and evaluate its potential application in immunotherapy.
This study has several limitations. Although multiple approaches were employed to analyze CSCs and the immune microenvironment, further validation experiments are required to confirm the reliability of the findings. The sample size and models used were limited; future studies will incorporate patient-derived xenograft (PDX) models to more accurately recapitulate the human breast cancer microenvironment. Moreover, both the spatial transcriptomic data and TCGA analysis in this study were based on primary tumor samples and did not include lymph node metastasis (LNM)-specific datasets. As a result, direct comparisons of ISG15 expression levels and M1/M2 polarization states in lymph nodes were not possible and warrant further investigation using LNM-targeted transcriptomic or spatial omics analyses. Nevertheless, our findings regarding the primary tumor microenvironment provide important theoretical insights into the mechanisms underlying lymph node metastasis.
Future research should also delve deeper into the interactions between CSCs and the immune microenvironment, which may reveal novel preventive and therapeutic strategies for metastatic breast cancer. Given that the spatial transcriptomic dataset employed in this study did not include LNM tissues, it cannot be used to characterize spatial features within the metastatic microenvironment. Integrating LNM-specific spatial omics data in future studies will help elucidate the spatiotemporal dynamics of cell–cell interactions during metastatic progression. Furthermore, considering the dynamic lineage features of tumor-associated macrophages (TAMs), future analyses should incorporate spatial and functional phenotyping to better resolve their heterogeneity and roles in the metastatic niche.

Conclusions
Based on this research, we propose that ISG15 in breast CSCs may regulate the immune microenvironment by promoting M2 macrophage polarization and inhibiting T cell activation, thereby accelerating lymph node metastasis. Specifically, ISG15 appears to promote IL-10 secretion, enhancing M2 macrophage polarization, and activate the JAK-STAT signaling pathway, upregulating PD-L1 to inhibit T cell activation.

Supplementary Information