Introduction

Triple negative breast cancer (TNBC) defined by deficiency of estrogen/progesterone receptors and human epidermal growth factor receptor 2 represents 15%~20% of all breast cancers1[,2. TNBC tumors exhibit notable heterogeneity, characterized by polyclonal cell populations, and coexist in diverse phenotypic characteristics, both phenomena associated with tumor malignancy and therapeutic resistance3,4. Due to the absence of targetable proteins, TNBC treatment has relied on cytotoxic chemotherapy for decades5. Currently, TNBC responsible for 15–20% of breast cancer cases and the 5-year overall survival (OS) rate for metastatic TNBC is approximately 11%, with a median OS ranging from 11 to 13 months6,7. In the operable settings, 30 ~ 40% of TNBC females relapse within 5 years8, and relapse is more common in ~ 50% of patients who cannot achieve pathological complete response (pCR) following neoadjuvant systemic therapy9. In the metastatic settings, the median overall survival (OS) is still less than 2 years, though systemic therapy has advanced10,11. Immunotherapy has revolutionized therapy for several solid tumor types12. In breast cancers, TNBC benefits the most from immunotherapy owing to its higher immunogenicity level versus other subtypes12,13. Our in-depth recognition of TNBC biology and growing awareness of the potential for personalized therapy has resulted in a paradigm shift in the early and late settings.

Exosomes are lipid-bilayer membranous extracellular vesicles, with diameter of 30 ~ 150 nm. The endosomal sorting complex required for transport (ESCRT) -dependent and ESCRT-independent processes can be utilized for categorizing the biogenesis mechanisms of exosomes14. Exosomes are secreted into the extracellular microenvironment by nearly all cell populations, with subsequent internalization by recipient cells15. Exosome-containing bioactive components exert a crucial function in intercellular interplay, affecting the fate of recipient cells via paracrine or autocrine machinery16. The outer membrane of exosomes owns the traits of cytoplasmic phospholipid bilayer, which potently maintains the activity and stability in a variety of body fluids. Exosomes are essential mediators in TNBC, serving as biomarkers for diagnosis and potential therapeutic agents17. For example, exosomal annexin A6 leads to gemcitabine resistance via attenuating ubiquitination and degradation of EGFR in TNBC18. The tumor microenvironment (TME) comprises cancer cells and neighbor components. Exosomes are rich within the TME, functioning as an indispensable information commutation tool between cancer cells and the TME via delivering specific exosomal contents. For instance, exosomes from TNBC cells facilitate pro-inflammatory macrophages correlated to more favorable clinical outcomes19. Activated T cell-secreted exosomal PD-1 decreases PD-L1-mediated immune evasion in TNBC20. Given the importance of exosomes in TNBC biology, this work proposed a novel exosome-based gene signature for individualized prognostication and assisting clinical management as well as discovered FAM129B as a promising therapeutic target against TNBC.

Materials and methods

Data sources

Based upon previous research, 2700 exosome genes were acquired (Supplementary Table 1). Transcriptomics data of 116 TNBC tumors and 113 normal tissues from The Cancer Genome Atlas (TCGA)-TNBC were included in this work. All the TNBC specimens owned the complete prognostic data. From the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/), two TNBC datasets: GSE58812 (n = 107)21, and GSE135565 (n = 84)22 were gained, which were merged into the meta-cohort (n = 190) following correction of batch effects via sva tool23. The meta-cohort was utilized for external verification. The prognostic features of TNBC patients are summarized in Supplementary Table 2. Homologous recombination deficiency (HRD), Copy number variations (CNVs), somatic mutations and tumor mutational burden (TMB) of TCGA-TNBC were also obtained. From the GSE148673 dataset, the work curated single-cell RNA sequencing (scRNA-seq) profiles of 5 TNBC tumors24.

Feature selection of candidate exosome genes

Firstly, differential analysis of exosome genes was executed in TCGA-TNBC tumors in comparison to healthy controls utilizing limma tool25. Differential exosome genes were chosen with adjusted p < 0.05 and |log2 fold-change|>0.58526. For subsequent selection of prognostic exosome genes, univariate Cox regression approach was adopted across TCGA-TNBC. p < 0.05 was utilized as the threshold. Through running glmnet package, least absolute shrinkage and selection operator (LASSO) analysis was executed to select characteristic exosome genes27. The optimal λ was chosen for avoiding the data overfitting. Kaplan-Meier (K-M) curves of OS were drawn for TCGA-TNBC patients stratified by the median transcript value of each characteristic exosome gene.

Construction of an exosome-based prognostic signature

An exosome-based gene signature was defined with the formula of risk score = sum (transcript value of characteristic exosome gene × matched coefficient). The risk score was computed in each TCGA-TNBC patient, with subsequent stratification into low- or high-risk group with the median risk score. The distinction between groups was proven via principal component analysis (PCA). Receiver operating characteristic curves (ROCs) were plotted for estimation of the predictive effectiveness in prognostication. K-M curves of OS were also conducted. Above analyses were externally verified in the meta-cohort. Uni- and multivariate analyses were executed for verifying whether the exosome-based signature functioned as an independent prognostic indicator across TCGA-TNBC.

Functional enrichment analysis

By running clusterProfiler package28, GO and KEGG analyses on differential exosome genes were carried out29,30,31. Gene set enrichment analysis (GSEA) was utilized for characterizing KEGG pathways of low- and high-risk groups under the “c2.cp.kegg.v7.5.1.entrez.gmt” as a reference set32. The activity of well-established pathways and each step within cancer-immunity cycle33 was estimated by running single-sample GSEA (ssGSEA) from GSVA package34.

Drug sensitivity prediction

Drug susceptibility was estimated by use of pRRophetic algorithm35. The prediction of chemotherapeutic response was based upon the Genomics of Drug Sensitivity in Cancer. The IC50 value was computed with ridge regression approach. In accordance with drug sensitivity information of human cancer cell lines were curated from the CTRP and PRISM projects, area under the curve (AUC) value of small-molecule compounds was inferred36.

Analysis of genetic mutations

CNVs were investigated by use of GISTIC2.0 for determining arm-level alterations with q-value < 0.2537. Somatic mutations were estimated via maftools tool by default38.

ScRNA analysis

Seurat software 5.1.0 was utilized for reading10× Genomics data39. Utilizing PercentageFeatureSet() function, proportion of mitochondrial genes in each cell was estimated. Cells with > 20% mitochondrial proportion were discarded. ScRNA-seq data normalization was subsequently implemented. The first 1,500 genes with high variability were chosen for downstream analysis. After scaling data, PCA was conducted. For avoiding technical noise, JackStrawPlot program was run to conduct a resampling test. The optimal principal components were selected. Cell clusters were determined utilizing FindClusters function, which were mapped into t-SNE. Through FindMarkers function, differential genes in each cluster were analyzed. Subsequently, cell types were defined utilizing prior knowledge40.

Estimation of TME components

TIMER41, CIBERSORT42, CIBERSORT-ABS42, QUANTISEQ43, MCP-counter44, xCell45, and EPIC46 computational approaches were utilized for inferring the abundance of components within TCGA-TNBC.

Immune checkpoint blockade (ICB) response estimation

The expression of immune checkpoints was computed. Tumor Immune Dysfunction and Exclusion (TIDE) was executed for inferring the response to ICB based upon immune escape machinery47. Response to PD-1 or CTLA4 antibody48,49 was estimated by use of Submap method50.

Cell culture and transfection

MDA-MB-231 and MDA-MB-468 cells (ATCC, USA) were cultured in DMEM medium (Gibco, USA) containing 10% fetal bovine serum (Gibco) and 1% penicillin/streptomycin (Gibco) at 37 °C with 5% CO2. Small interfering RNAs (siRNAs) of FAM129B (si-FAM129B) and negative control (si-NC) were synthesized by GenePharma (Shanghai, China). Based upon the manufacturer’s specification, TNBC cells were transfected with 50 nM siRNA pre-mixed with Lipofectamine 3000 (Invitrogen, USA). After 48-h transfection, the knockdown efficiency of FAM129B was assessed via western blot.

Western blot

Cells were lysed via RIPA lysis buffer (BL504A; Biosharp, Guangzhou, China). Total protein content was measured through BCA kit (BL521A; Biosharp). Protein was separated via 6% or 10% SDS-PAGE electrophoresis as well as transferred onto PVDF membrane. Following blockade, incubation with primary antibody against FAM129B (1:1000; 22553-1-AP; Proteintech, Wuhan, China) or GAPDH (1:5000; 60004-1-Ig; Proteintech) at 4 ℃ overnight. After washing with PBST, the membrane was placed into horseradish peroxidase-labeled secondary antibody (1:10000; A21020 or A21010; abbkine, Wuhan, China) diluted with 5% milk/PBST and incubated in a room temperature shaker for 1 h. The protein was developed via ECL luminescent solution (BMU101-CN; abbkine) as well as investigated utilizing automatic chemiluminescence image system (5200Multi; Tanon, USA).

5-ethynyl-2’-deoxyuridine (EdU) assay

TNBC cells were added to a 24-well plate (8*105 cells/well). Following 48 h, EdU assay was carried out by use of BeyoClick™ EdU-594 cell proliferation assay kit (C0078S; Beyotime, Beijing, China) in accordance with the manufacturer’s specifications. Cell proliferation was investigated under a BX53 fluorescence microscope (Olympus, Japan), with subsequent estimation of the percentage of EdU-positive cells.

Wound healing assay

Cells were inoculated onto a 6-well plate (5*106 cells/well) and maintained overnight. After the cells uniformly covered the culture plate, scratches were made utilizing a 10-µL sterile pipette tip with subsequent rinse by PBS. Photographs were recorded at 0 h and 24 h under an IX71 microscope (Olympus).

Statistical analysis

Data were analyzed utilizing appropriate R packages (version 3.6.1) or GraphPad Prism software (version 9.0.1). Continuous variables between groups were compared through Student’s t-test or Wilcoxon test. Pearson or Spearman test was executed for correlation analysis. K-M survival curves were drawn, with subsequent log-rank test. P-value < 0.05 was statistically significant.

Results

Selection of characteristic exosome genes in TNBC

According to the previously study, we firstly collected 2700 known exosome-related genes and further analyzed their expression level in TNBC and normal tissue. Among these exosome genes, 474 were significantly dysregulated in TNBC tumors compared to normal tissues (Fig. 1A, B). Their prognostic significance in TNBC was subsequently observed through univariate cox regression analysis. Consequently, 27 exosome genes were identified that significantly connected to patient prognosis (Table 1). The GO enrichment result of the differential exosome genes were divided into three parts, including biological process (BP), cellular component (CC) and molecular function (MF). In the BP part, genes mainly involved in establishment of localization, response to chemical, transport, vesicle-mediated transport, secretion by cell, exocytosis, regulated exocytosis (Fig. 1C). In the CC part, extracellular region, vesicle, extracellular region part, extracellular space, extracellular vesicle, extracellular exosome were significant enriched (Fig. 1D). In the MF component, anion binding, cadherin binding, amide binding, cell adhesion molecule binding, peptide binding were mainly enriched (Fig. 1E). In addition, KEGG pathway enrichment result indicated that genes mainly involved in phagosome, tight junction, regulation of actin cytoskeleton, carbon metabolism, glycolysis/gluconeogenesis (Fig. 1F). Based upon feature selection via LASSO algorithm, 7 characteristic exosome genes were finally chosen, comprising ALCAM, FAM129B, GNB2, KRT6A, PGK1, SERPINE1, and THY1 (Fig. 1G, H). Notably, all the feature genes presented the notable overexpression in TNBC tumors versus controls (Fig. 1I). In addition, they were connected to poor OS outcomes (Fig. 1J-P), further uncovering their involvement in TNBC.

Fig. 1
figure 1

Selection of characteristic exosome genes in TCGA-TNBC. (A, B) Differential exosome genes in TCGA-TNBC tumors versus normal tissues. Gene ontology (GO) enrichment result of differential exosome genes that categorized into biological process (BP) (C), cellular component (CC) (D) and molecular function (MF) (E). (F) KEGG pathway enrichment result of differential exosome genes. (G, H) LASSO for feature selection of exosome genes in TNBC. (I) Differential expression of characteristic exosome genes in TNBC tumors versus control tissues. (J-P) OS outcomes of patients with lowly and highly expressed characteristic exosome genes: ALCAM, FAM129B, GNB2, KRT6A, PGK1, SERPINE1, and THY1.

Full size image
Table 1. Prognostic exosome genes in TCGA-TNBC.
Full size table

An exosome-based gene signature robustly estimates TNBC clinical outcomes

Based upon the coefficients of the feature genes and their expression values, an exosome-based gene signature was proposed, with the calculation of the risk score of 0.427199592 * PGK1 + 0.143753595 * GNB2 + 0.969065967 * FAM129B + 0.573649346 * THY1 + 0.620743219 * ALCAM + 0.076646291 * KRT6A + 0.053824999 * SERPINE1 across TCGA-TNBC (Fig. 2A). The TCGA-TNBC patients were then categorized into low- and high-risk groups based on the median risk score. PCA demonstrated a clear distinction between low- and high-risk tumors (Fig. 2B). The AUC values under one-, three- and five-year OS all exceeded 0.85, uncovering that the signature enabled to accurately estimate clinical outcomes (Fig. 2C). The reproducibility was validated in the meta-cohort (GSE58812 and GSE135565) (Fig. 2D-F). Moreover, patients in high-risk group were correlated worse OS outcome compared to those in low-risk group in TCGA-TNBC cohort (p < 0.05) (Fig. 2G), which were externally proven in the meta-cohort (Fig. 2H). To further validated the independence of the signature in the TCGA-TNBC cohort, we conducted univariate and multivariate cox regression analysis on the pathological factors and signature score, and the results suggested the signature can be served as an independent risk factor of TNBC prognosis (Fig. 2I, J).

Fig. 2
figure 2

Definition of an exosome-based gene signature for robust estimation of TNBC clinical outcomes. (A) Calculation of the risk score across TCGA-TNBC based upon the feature genes. (B) PCA for uncovering the differentiation between low- and high-risk tumors. (C) ROCs of verification of the predictive effectiveness of the risk score in prognostication. (D-F) External verification of the risk score distribution, the distinction between low- and high-risk tumors, and ROCs in the meta-cohort (GSE58812 and GSE135565). (G) Differential OS outcomes between low- and high-risk TCGA-TNBC patients. (H) Validation of OS difference between groups in the meta-cohort. (I, J) Uni- or multivariate analysis of the connections of the risk score and clinical parameters with TCGA-TNBC survival.

Full size image

High-risk TNBC tumors are more sensitive to docetaxel and several small-molecule agents

To explore the involved pathway in the signature, we conducted gene set enrichment analysis (GSEA) between high and low risk group. As showed in Fig. 3A, the remarkable enrichment in key tumorigenic pathways (Notch signaling pathway, ECM receptor interaction, pathway in cancer) was discovered in high-risk tumors. In addition, the risk score presented the negative connections to DNA damage repair machinery, with the positive connections to stromal activation (Fig. 3B). Further drug sensitivity analysis uncovered that patients in the high risk group owned the remarkably lower IC50 of docetaxel and lower HRD score in comparison to those with low risk (Fig. 3C, D), indicated that patients in high-risk group have the higher possibility to benefit from docetaxel-based chemotherapeutics. Moreover, we also explored the candidate drugs to the high-risk patients by integrating CTRP and and PRISM drug database, and a few small-molecule agents were discovered for personalized therapy of high-risk tumors, e.g., MK-0752, BRD-K33199242, IC-87,114, fumonisin B1, ilomastat, GW-788,388, afobazole, and batimastat (Fig. 3E, F).

Fig. 3
figure 3

Analysis of heterogeneity in well-established pathways and responses to docetaxel and small-molecule agents between low- and high-risk TNBC tumors. (A) The enrichment of tumorigenic pathways in high- versus low-risk tumors through GSEA. The pathways are marked by unique colors. (B) Connections of the risk score to well-established mechanisms. (C, D) Analyses of heterogeneity in IC50 of docetaxel and HRD score between groups. (E, E) Connections of the risk score with AUC of small-molecule agents from the CTRP or PRISM and the differential AUC between tumors. *P < 0.05; **p < 0.01; ***p < 0.001.

Full size image

Low- and high-risk TNBC tumors present distinct genetic mutations

By comparing CNV mutation in the TNBC tumors, we observed that heterogeneous CNVs were highly occurred in high risk group (Fig. 4A-D). Moreover, low-risk tumors appeared to have more frequent CNVs. The frequency of mutated genes also displayed the extensive heterogeneity between high- and low-risk tumors (Fig. 4E, F). The top twenty mutated genes, such as TP53, TTN etc. owned the higher frequency in low- versus high-risk tumors. Like previous findings51, high TMB group owned the longer OS time versus another group (Fig. 4G). In accordance with the risk score and TMB, TNBC patients were stratified into four subgroups. Low-risk patients with high TMB owned the best OS outcomes, with those with high risk and low TMB possessing the poorest OS, indicating that the signature could independently predict OS outcome (Fig. 4H). Above findings uncovered the heterogeneity in genetic mutation characteristics between low- and high-risk TNBC.

Fig. 4
figure 4

Heterogeneous genetic mutations across low- and high-risk TNBC tumors. (A, B) Copy number amplifications and losses in high-risk tumors. The threshold was set as q-value of 0.25. (C, D) Copy number amplifications and losses in low-risk tumors. (E, F) Waterfall plots illustrating the first twenty mutated genes in high- or low-risk tumors. (G) Differential OS outcomes between low and high TMB tumors. (H) Differential OS outcomes among four groups stratified by the risk score and TMB.

Full size image

Characteristic exosome genes are specifically expressed in the TNBC microenvironment

To further assess the expression of exosome genes at the single cell level, 5 TNBC tumors from GSE148673 were included in this work. Following quality control, data normalization as well as dimensionality reduction (Supplementary Fig. 1A-G), 21 cell clusters were determined (Fig. 5A). Based upon differential genes in each cell cluster and prior cell type markers (Fig. 5B), 6 cell types within TNBC tumors were defined, composed of endothelial, epithelial, stromal, mast, eosinophil, and neutrophil cells (Fig. 5C). These cell types were heterogeneous across diverse TNBC tumors (Fig. 5D). The characteristic exosome genes (ALCAM, FAM129B, GNB2, KRT6A, PGK1, SERPINE1, and THY1) were extensively expressed in the diverse cell populations within the TNBC microenvironment (Fig. 5E-G), indicating their possible implications in modulating the TME.

Fig. 5
figure 5

Single-cell analysis of characteristic exosome genes in the TNBC microenvironment. (A) Single-cell atlas of diverse cell clusters marked by unique colors. (B) Heatmap illustrating the top eight differential genes in each cell cluster versus other clusters. (C) Definition of six cell types across TNBC tumors. (D) Specific distribution of diverse cell types in each TNBC tumor. (E-G) Specific expression of characteristic genes: ALCAM, FAM129B, GNB2, KRT6A, PGK1, SERPINE1, and THY1 across distinct cell types.

Full size image

High-risk TNBC patients better respond to ICB

High- and low-risk TNBC tumors exhibited heterogeneous TME components based upon multiple computational methods (Fig. 6A). Innate and adaptive immune cells appeared to present more immune infiltration in high-risk tumors. Cancer-immunity cycle is composed of multiple processes needed for immune-based control of cancer growth52. The risk score owned positive connections to almost all steps (Fig. 6B). Interrupting one or more processes facilitate immunosurveillance escape. Hence, the risk score was proven to associate with antitumor immunity. Immune checkpoints: CD276, NRP1, TNFRSF4 and TNFSF4 presented remarkably higher transcript values in high- versus low-risk tumors (Fig. 6C), indicating that high-risk tumors potentially benefited from CD276, NRP1, TNFRSF4 or TNFSF4 blockade. Additionally, high-risk patients owned the significant similarity to those respond to CTLA4 blockade (Fig. 6D). In accordance with higher dysfunction/exclusion/TIDE scores, high-risk TNBC tumors appeared to occur notable immune evasion (Fig. 6E-G). Altogether, high-risk patients might better respond to ICB.

Fig. 6
figure 6

Heterogeneous anticancer immunity between low- and high-risk TNBC tumors. (A) Estimation of the abundance of components within the low- or high-risk TNBC microenvironment utilizing common computational approaches. (B) Connections of the risk score with the activity of steps within cancer-immunity cycle. (C) Differential expression of immune checkpoints in low- and high-risk TNBC. (D) Submap for inferring the similarity of low- or high-risk samples with response to PD-1/CTLA4. (E-G) Differential dysfunction/exclusion/TIDE scores between low- and high-risk TNBC. *P < 0.05; **p < 0.01; ***p < 0.001.

Full size image

FAM129B knockdown attenuates proliferation and motility in TNBC cells

Among the characteristic exosome genes (ALCAM, FAM129B, GNB2, KRT6A, PGK1, SERPINE1, and THY1), little is known about the function of FAM129B, which belongs to a family of Niban proteins, in TNBC. Therefore, this study investigated the role of FAM129B in TNBC cells. Transient transfection of FAM129B siRNAs into two TNBC cell lines, MDA-MB-231 and MDA-MB-468, effectively knocked down FAM129B expression, as confirmed by Western blot (Fig. 7A-D). Subsequent functional analyses revealed that FAM129B knockdown significantly impaired cellular behaviors. Specifically, EdU assays demonstrated that FAM129B knockdown attenuated the proliferative capacity of both MDA-MB-231 and MDA-MB-468 cells (Fig. 7E-G). Furthermore, wound healing assays showed that FAM129B knockdown prominently restrained the cell motility of these TNBC cells (Fig. 8A-D), indicating that targeting FAM129B attenuates aggressive behaviors, including proliferation and migration, in TNBC cells.

Fig. 7
figure 7

Knockdown of FAM129B impairs proliferative capacity of TNBC cells. (A, B) Representative western blot images of FAM129B and quantification results in MDA-MB-231 cells transfected with siRNAs of FAM129B or its controls. GAPDH served as a reference control. (C, D) Representative western blot images of FAM129B and quantification results in MDA-MB-468 cells with si-FAM129B or si-NC transfection. (E) Representative photographs of EdU assay in si-FAM129B- or si-NC-transfected MDA-MB-231 and MDA-MB-468 cells. Scale bar, 20 μm. (F, G) Evaluation of EdU-positive MDA-MB-231 and MDA-MB-468 cells in the context of si-FAM129B or si-NC transfection. **p < 0.01; ***p < 0.001.

Full size image
Fig. 8
figure 8

Knockdown of FAM129B attenuates aggressive behaviors of TNBC cells. (A) Representative 0-h and 24-h wound healing photographs of MDA-MB-231 cells with si-FAM129B or si-NC transfection. Scale bar, 200 μm. (B) Quantification of cell motility of si-FAM129B- or si-NC-transfected MDA-MB-231 cells. (C) Representative 0-h and 24-h wound healing photographs of MDA-MB-468 cells in the context of si-FAM129B or si-NC transfection. Scale bar, 200 μm. (D) Quantification of cell motility of MDA-MB-468 cells with si-FAM129B or si-NC transfection. **p < 0.01; ***p < 0.001.

Full size image

Discussion

TNBC represents an aggressive subtype with the features of widespread intra-tumoral heterogeneity53,54. Despite there being some prognostic signatures identified in TNBC, such as a methylation signature and ferroptosis, etc., the tumor heterogeneity and deficiency of biomarkers remain challenging55,56,57. In this work, seven characteristic exosome genes were discovered to be implicated in TNBC, comprising ALCAM, FAM129B, GNB2, KRT6A, PGK1, SERPINE1, and THY1, which were utilized for defining the exosome-based gene signature. We found that the signature enabled to accurately estimate patient prognostic outcomes, with functioning as an independent predictor. In addition, it could potentially infer patients’ therapeutic response.

Surgery is a primary treatment for triple-negative breast cancer (TNBC), often involving lumpectomy (removing the tumor and surrounding tissue) or mastectomy (removing the entire breast). The impact of anesthetics on breast cancer biology represents a critical intersection of oncology and anesthesiology, with emerging evidence suggesting that the choice of anesthetic during surgical resection may influence long-term cancer outcomes. Lin et al. elucidate the impact of anesthetics during surgery for breast cancer58,59,60,61,62. The neoadjuvant chemotherapy remains the standard of care for early TNBC63. BRCA1/2-mutated and some sporadic TNBC own DNA repair defects and present the sensitivity to DNA-damaging therapies. HRD score that is an unweighted sum of loss of heterozygosity, telomeric allelic imbalance, and large-scale state transition scores enables to identify TNBC tumors more likely to respond to platinum-containing neoadjuvant chemotherapy64. The pCR rate of TNBC patients receiving docetaxel in combination with carboplatin is notably higher compared with those receiving epirubicin/cyclophosphamide after docetaxel65. Patients treated with docetaxel combined with carboplatin who achieve pCR present higher HRD scores versus those with non-pCR66. HRD is thus a promising predictor for clinical benefits from docetaxel plus carboplatin. Based upon the lower IC50 of docetaxel and lower HRD score, high-risk TNBC tumors were more likely to be clinically sensitive to docetaxel-based chemotherapy67.

Several small-molecule agents (MK-0752, BRD-K33199242, IC-87114, fumonisin B1, ilomastat, GW-788388, afobazole, and batimastat) were discovered to be suitable for high-risk individuals. Preclinical studies have demonstrated the anti-breast cancer properties of these agents. For instance, MK-0752 in synergy with Tocilizumab effectively attenuates breast cancer stem cells as well as hinders cancer growth68. IC-87,114 p110δ-selective inhibitor suppresses breast carcinoma progression through targeting tumor cells and macrophages69. Ilomastat perturbes the SNAI1-based activation of breast cancer stem cell phenotypes70. Batimastat exerts anti-tumor and anti-metastasis property in TNBC71.

The breast cancer microenvironment (TME) is a dynamic ecosystem comprising tumor cells, immune cells, stromal components, and extracellular matrix. It orchestrates immunosuppressive networks that facilitate tumor progression, metastasis, and therapy resistance72. For example, Regulatory T cells (Tregs) are enriched in metastatic breast cancer. Single-cell RNA sequencing reveals Tregs overexpress FOXP3 and *CTLA-4*, directly suppressing effector T cells. Despite the notable advances in cancer immunotherapy, clinical therapy of TNBC remains tough to make a breakthrough73. The undesirable treatment effectiveness may be attributed to the deficiency of tumor immunogenicity as well as immunosuppressive TME74. In addition, based upon the heterogeneity of TNBC, suitable biomarkers can be a potently powerful tool for selection of patients for therapy75. Herein, characteristic exosome genes: ALCAM, FAM129B, GNB2, KRT6A, PGK1, SERPINE1, and THY1 were extensively distributed in the TME, further uncovering their significance in remodeling the TME. Among the characteristic exosome genes, Activated leukocyte cell adhesion molecule (ALCAM), also known as CD166, is a transmembrane immunoglobulin weighing between 100 and 105 KDa. It plays a crucial role in the activation of T-cells, hematopoiesis, neutrophils trans-endothelial migration, angiogenesis, inflammation, and tumor propagation and invasiveness by forming homophilic and heterophilic interactions76. FAM129B was highly expressed in TNBC tissue compared to normal tissue. Moreover, high expression level of FAM129B was correlated with a poor survival outcome, indicated that over-expressed FAM129B might promote tumor evasion. Zhou et al. identified that FAM129B can be served as an prognostic predictor of non-small cell lung cancer (NSCLC) patients and promote tumor invasion and proliferation of NSCLC cells through the activation of FAK signaling pathway77. In our study, it was experimentally validated that suppression of FAM129B effectively attenuated proliferative and aggressive behaviors of TNBC cells. Shi et al. discovered that miR-142-3p can enhance PTX resistance by targeting GNB2, revealing that knocking down GNB2 expression can activate the AKT-mTOR pathway in breast cancer78. KRT6A was found a prognostic biomarker and related to the progression of invasive areas of colorectal cancer in a recent study79. Chen et al. identified that phosphoglycerate kinase 1 (PGK1) play a crucial role in TNBC by modulating metabolism and HIF-1 signaling pathway80. Zhang et al. demonstrated that miR-30d-5p inhibits cell proliferation, invasion, and metastasis by directly targeting SERPINE1 and enhancing fatty acid β-oxidation in breast cancer81. THY1 was correlated with metastasis and poor patient survival in the basal-like subtype, and considered as a promising new therapeutic target for the treatment of breast cancer82. The evidence uncovered the potential of FAM129B as a possible treatment target against TNBC. High-risk individuals were inferred to be responsive to ICB (CD276, NRP1, TNFRSF4, TNFSF4, and CTLA4). Thus, the exosome-based model might potentially estimate ICB response of TNBC individuals. Nonetheless, the effectiveness of the exosome-based model in estimation of clinical outcomes and treatment response requires to be further verified in larger cohorts.

Several limitations of our present study need to be elucidated. Firstly, despite demonstrating a high performance of the exosome-related prognostic signature for TNBC, the sample size of the cohort is relatively small and requires larger multi-center cohorts. Secondly, the cohort from the TCGA database may introduce biases due to technical and biological limitations83,84. Thirdly, these findings still needs to be validated in in vivo and in vitro experiments, such as patient-derived xenograft models experiment85.

Conclusion

Taken together, the exosome-based gene signature (ALCAM, FAM129B, GNB2, KRT6A, PGK1, SERPINE1, THY1) serves as a practical prognostic tool for TNBC patients. In addition, the signature enabled to individually estimate the response to docetaxel, small-molecule agents (MK-0752, BRD-K33199242, IC-87114, fumonisin B1, ilomastat, GW-788388, afobazole, and batimastat) and ICB (CD276, NRP1, TNFRSF4, TNFSF4, and CTLA4), which strengthened the new implication of exosomes in TNBC. More importantly, suppression of FAM129B was experimentally evidenced to hinder proliferation and aggressiveness of TNBC cells. Thus, FAM129B acted as a possible treatment target against TNBC.