Tumor-derived type IX collagen regulates dormancy in triple-negative breast cancer cells Authors: Wei Liu, Zhe Song Published: 2026 ABSTRACT Metastasis and recurrence of triple-negative breast cancer (TNBC), are mainly attributed to the presence of dormant cancer cells (DCCs), yet the molecular mechanisms governing the maintenance and reactivation of DCC dormancy remain poorly elucidated. This study identified collagen IX alpha 1 chain (Col9a1) as a gene specifically expressed in TNBC dormant cells and confirmed that it is a key molecule sustaining DCC dormancy. Mechanistic studies revealed that COL9A1 sustains TNBC cell dormancy by suppressing the FAK-AKT-p27 signaling axis, thereby inducing G0/G1 cell cycle arrest. Col9a1 knockdown reactivates DCCs, and the reactivated cells recruit natural killer (NK) cells by upregulating retinoic acid early inducible 1γ (RAE-1γ) and secreting proinflammatory cytokines (IL-15/IL-18), thus abrogating the immune evasion capacity of DCCs and enhancing the chemosensitivity of reactivated TNBC cells to docetaxel. This study uncovers a novel mechanism by which COL9A1 regulates TNBC cell dormancy, indicating that COL9A1 can serve as a potential biomarker for assessing the dormant phenotype of TNBC and predicting tumor recurrence, as well as a novel therapeutic target for TNBC. This study also provides preclinical evidence for developing COL9A1-targeted combination therapies for TNBC.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-49687-6. Introduction In 2022, female breast cancer ranked as the second leading cause of cancer incidence globally, with an estimated 2.3 million new cases, and it was the fourth leading cause of cancer-related deaths worldwide1. Triple-negative breast cancer (TNBC) is a subtype of breast cancer defined by the absence of estrogen receptors, progesterone receptors, and HER2/ERBB2 expression. This subtype is known for its aggressive nature and poor prognosis2. TNBC is the most aggressive subtype of breast cancer when compared to hormone receptor-positive and HER2-amplified tumors. It is characterized by higher and earlier rates of recurrence and mortality in the operable setting (stages I-III)3 and a shorter overall survival rate in the inoperable setting (stage IV)4, 5. In 1954, Geoffrey Hadfield proposed that relapse occurs when dormant cancer cells(DCCs) enter a state of “temporary mitotic arrest”6. Quiescent cancer cells (QCCs), characterized by reversible G0/G1 cell cycle arrest, are the core cellular component of DCCs. Cell dormancy denotes a broader in vivo biological state grounded in cellular quiescence, which is dynamically regulated by the crosstalk between disseminated tumor cells (DTCs) and their surrounding microenvironment7. Under favorable conditions, disseminated dormant tumor cells “reawaken” and resume proliferation, subsequently colonizing distant sites. Chronic inflammation induces the formation of neutrophil extracellular traps (NETs), which can reactivate DCCs8. In contrast, type III collagen maintains tumor dormancy by inhibiting DCC reawakening in the primary tumor microenvironment9. Changes in the tumor microenvironment, such as fibrosis10, tissue remodeling11, obesity12, inflammation13, or exposure to cigarette smoke8, may trigger the awakening of DCCs. However, the specific factors that influence the balance between signals that inhibit diffuse tumor cell growth and those that promote awakening and overt metastasis remain unclear. Dormant tumor cells possess the capability to evade immune destruction. In immune-competent mouse models, dormant growth-arrested acute myeloid leukemia cells evade immune surveillance by upregulating immune checkpoint proteins (PD-L1 and CTLA-4)14. Additionally, DCCs establish a reservoir of immunotherapy resistance by creating a local hypoxic and immunosuppressive environment that impedes T cell function15. Recent studies have also indicated that dormant DTCs evade clearance by natural killer (NK) cells, and that only upon awakening does activation render them recognizable to the immune system16. Notably, distant metastatic TNBC cells often enter a dormant state, which is a major cause of chemotherapy resistance and clinical recurrence17. Moreover, research has confirmed that the proportion of QCCs increases over time following treatment with docetaxel, suggesting that chemotherapeutic agents may promote the expansion of the QCC population, thereby inducing drug resistance18. This observation aligns with clinical realities: ovarian cancer cells isolated from avascular collagen nodules are often in a quiescent state post-chemotherapy19, and these surviving QCCs, which subsequently lose their quiescent state and resume proliferation, have been confirmed as important mechanisms leading to chemotherapy resistance and cancer recurrence20, 21. Furthermore, tumors with a high proportion of chemotherapy-resistant QCCs exhibit significantly earlier recurrence compared to those with low QCC abundance22. There is substantial evidence that certain extracellular matrix (ECM) proteins can promote cellular cancer dormancy. Research has demonstrated that TGFβ-mediated secretion of fibronectin enables cells to maintain a dormant state23. Furthermore, dormant MCF-7 cells exhibit increased expression of collagen type VII (ColVII)24, and the downregulation of COL7A1 is associated with reduced 5-year survival rates in breast cancer25. The expression of COL17A1 can prevent dormant tumor cells from re-entering the cell cycle by inhibiting the activation of the FAK-YAP axis26. Conversely, type I collagen-enriched fibrosis in breast cancer metastases may serve as a critical factor influencing cytoskeletal reorganization in dormant tumor cells, facilitating their transition from dormancy to metastatic growth27. Additionally, increased expression of COL3A1 remodels the ECM by enhancing its curliness, thereby driving the cells into a dormant state sustained by DDR1 binding9. However, the specific roles and underlying mechanisms of ECM proteins in regulating TNBC dormancy remain largely unknown. In this study, we identified collagen IX alpha 1 chain (Col9a1) as a key ECM gene specifically expressed in TNBC DCCs, and further investigated its role and molecular mechanisms in regulating TNBC dormancy, including cell cycle quiescence, immune evasion and chemosensitivity. This study confirms that Col9a1 is highly expressed in dormant breast cancer cells. In a TNBC model, overexpression of Col9a1 significantly inhibits cell proliferation and migration; conversely, knockdown of Col9a1 promotes the proliferation and migration of 4T07 cells, likely related to the activation of the FAK–AKT–p27 signaling pathway following the loss of COL9A1. In vivo experiments further demonstrate that reducing Col9a1 expression can induce reactivation of dormant tumor cells, evidenced by increased tumor volume and enhanced infiltration of peripheral immune cells, particularly a significant accumulation of NK cells. Mechanistically, downregulation of Col9a1 makes reactivated tumor cells more susceptible to recruiting NK cells and enhances their sensitivity to NK cell-mediated cytotoxicity. This process may be associated with the release of cytokines such as IL-15 by reactivated cancer cells, as well as the upregulation of NK cell activation receptor ligand retinoic acid early inducible 1γ (RAE-1γ) and downregulation of the inhibitory ligand major histocompatibility complex class I (MHCI) molecules. However, this immune response is insufficient to completely suppress tumor growth; combined treatment with docetaxel significantly enhances the antitumor effect and effectively curtails tumor progression. Methods and materials Cell cultures The 4T1, 293T, MDA-MB-231, MDA-MB-468, BT-549, and Hs578T cell lines were purchased from Procell Life Science and Technology. 4T07 cells were provided by Dr. Hu Guohong (Chinese Academy of Sciences). 4T1 and 4T07 (ER-/PR-/HER2-) are murine mammary carcinoma cell lines derived from the same BALB/c spontaneous breast tumor, serving as a well-validated classic model pair for investigating TNBC dormancy: 4T1 cells exhibit high metastatic potential (non-dormant), while 4T07 cells have low metastasis and a stable dormant phenotype in syngeneic mice, consistent with human TNBC molecular features28, 29, 30. All cell lines were authenticated by short tandem repeat (STR) analysis (ANSI/ATCC ASN-0002-2021) and passaged for fewer than 6 passages. 4T1, 293T, and MDA-MB-231 cells were cultured in DMEM + 10% fetal bovine serum (FBS); 4T07 cells in DMEM + 10% FBS + 1% non-essential amino acids (NEAA); and other lines in specialized media. All cells were maintained at 37 °C with 5% CO₂ in media containing penicillin and streptomycin. RNA sequencing (RNA-seq) Total RNAs from 4T07 and 4T1 cells were extracted with TRIzol (Invitrogen). Libraries were constructed and sequenced following Illumina’s standard protocol. Low-quality reads were filtered, and differentially expressed genes (DEGs) were identified using DEGseq and PoissonDis (fold change > 2, p < 0.05). GO and KEGG enrichment analyses were performed via R software (phyper function), with terms considered significantly enriched at FDR ≤ 0.05. Lentiviral production and transduction For Col9a1 knockdown, four shCol9a1 constructs were cloned into pLVX-shRNA2-Puro. Lentivirus was produced by transient co-transfection of 293T cells and concentrated. 4T07 cells were transduced with the virus plus polybrene, followed by selection with 10 µg/mL puromycin (Solarbio) 48 h later. For Col9a1 overexpression, the Col9a1 fragment was cloned into a lentiviral vector; virus production and transduction were performed as above, with 4T1 cells as recipients. Transient transfection was conducted using Lipofectamine 3000 (Invitrogen). Sequences of siRNAs, shRNAs, and plasmids are provided in Supplementary Tables S2–5. Animal studies All animal procedures were conducted in accordance with relevant guidelines, and all animal procedures in this study were approved by the Institutional Animal Care and Use Committee of Suzhou University and strictly adhered to relevant animal welfare and ethical guidelines. The methods are reported following the ARRIVE guidelines. Female BALB/C mice (6–8 weeks old, weighing 18–20 g) were purchased from Hangzhou Qizhen Laboratory Animal Co., Ltd. For all invasive procedures (cell inoculation and euthanasia), mice were anesthetized with 3% isoflurane in 100% oxygen (1 L/min) using a precision vaporizer. Two murine breast cancer cell lines (4T07 and 4T1) were used. For orthotopic transplantation: Col9a1-overexpressing 4T1 cells (5 × 10⁶ cells/mL, 0.1 ml) or Col9a1-knockdown 4T07 cells (2 × 10⁷ cells/mL, 0.1 ml) were injected into the fourth mammary gland. For docetaxel (DTX) treatment: 2 × 10⁶ 4T07 cells were subcutaneously inoculated, followed by weekly intraperitoneal injection of DTX (15 mg/kg)31 or solvent (5% DMSO, 30% PEG300, 5% Tween80 in ddH₂O) one week later. Tumor volume was measured every 3 days (formula: (L×W²)/2). Lung metastasis was assessed via tail vein injection of equal cell numbers. For NK cell depletion: mice received intraperitoneal anti-Asialo-GM1 antibody (100 µg, Invitrogen) on days 0, 3, 7, 11, and 15 (isotype antibody as control), with 2 × 10⁶ 4T07 cells injected into the mammary gland on Day 1. For survival analysis: 2 × 10⁶ 4T07 cells were tail vein-injected one week prior to weekly intraperitoneal DTX (10 mg/kg). RT-qPCR Total RNA was extracted with the RNA-Quick Purification Kit (ESScience). cDNA was synthesized using NovoScript Plus All-in-one 1st Strand cDNA Synthesis SuperMix (Novoprotein), and real-time qPCR was performed with Novostart SYBR qPCR SuperMix Plus (Novoprotein). Primers are listed in Supplementary Table S6. Western blot analysis and coIP Proteins were extracted from cell lysates and culture supernatants, separated by electrophoresis under denaturing/reducing conditions, transferred to nitrocellulose membranes, and detected by specific antibody binding and chemiluminescence. Conditioned media were centrifuged to remove cell debris and concentrated by ultrafiltration. For supernatant-derived proteins, loading was normalized to equal cell number and verified by Coomassie brilliant blue total protein staining; for cell lysates, GAPDH was used as the loading control (antibody details are shown in Supplementary Table S7).For co-immunoprecipitation assays, cell lysates were incubated with protein A/G agarose beads (Santa Cruz, sc-2003) in ice-cold non-denaturing lysis buffer according to the manufacturer’s protocol. Cell proliferation For proliferation assays, cells were seeded at 5000 cells/well in a standard cell culture 96-well plate. Proliferation was measured using the CCK-8 proliferation assay kit (Apex Bio) according to manufacturer instructions. Cell cycle and cell apoptosis Cell cycle distribution was assessed using the Cell Cycle Analysis Kit (Beyotime). Cells were fixed in 70% ethanol at 4 °C, stained with propidium iodide/RNase A in the dark, and analyzed via flow cytometry (FlowJo software). Apoptosis was detected using Annexin V APC or Annexin V-FITC/7-AAD kits (Biogems) according to the manufacturer’s instructions and analyzed by flow cytometry. Migration and wound healing assay For Transwell migration assays: 1.5 × 10⁵ cells in 200 µl serum-free medium were seeded into the upper chamber (Corning). Complete medium was added to the lower chamber, and cells were incubated at 37 °C for 48 h. Migrated cells were fixed with 4% paraformaldehyde, stained with crystal violet (Invitrogen), imaged, and counted via ImageJ. For wound healing assays: cells were seeded into Ibidi inserts in 35 mm dishes, grown to confluence (37 °C, 5% CO₂), and inserts were removed. Serum-free medium was added, and wound closure was imaged and analyzed by ImageJ after 24 h. All experiments were performed in triplicate. Supplementary validation assays (wound healing with mitomycin C, Transwell with 24 h incubation) were conducted to exclude proliferation interference (Fig. S2C-D). Flow cytometry Tumor tissues were ground, sieved, and lysed to obtain single-cell suspensions. Dead cells were excluded with Fixable Viability Dye eFluor™ 506 (eBioscience), and non-specific binding was blocked with FcR blocking reagent (Miltenyi). Surface staining was performed with anti-mouse CD45, CD3, CD4, CD8a, CD49b, RAE-1γ, MHCI, and EP-CAM antibodies (4 °C, 20 min). For intracellular staining, cells were treated with stimulation and transport inhibitor cocktails (eBioscience), fixed, permeabilized, and incubated with anti-mouse IFN-γ, Perforin, and Granzyme B antibodies (room temperature, 15 min, dark). Antibody details are provided in Supplementary Table S7. For immune cell subset quantification, gating was performed sequentially: main cell populations were gated from FSC/SSC plots to exclude debris, singlet cells selected by FSC-A/FSC-H gating, and viable CD45⁺ immune cells identified by excluding dye-positive dead cells. T cells were defined as CD45⁺CD3⁺, and NK cells as CD45⁺CD3⁻CD49b⁺ subsets in viable CD45⁺ cells. For tumor cell marker analysis (RAE-1γ, MHCI), viable EP-CAM⁺ cells were gated as tumor cells. The complete gating strategy is provided in the Supplementary Information. For intracellular staining of NK cell effector molecules (GZMB, Perforin, IFN-γ), the staining and flow cytometric analysis were specifically performed on the pre-gated CD45⁺CD3⁻CD49b⁺ NK cell subset. Immunofluorescence staining 4T07 cells grown on glass coverslips were fixed with 4% paraformaldehyde (15 min, RT), permeabilized with PBSTx (0.3% Triton X-100), and incubated overnight at 4 °C with anti-COL9A1 and anti-p27 primary antibodies. After washing with TBSTx, cells were stained with Cy3- and FITC-conjugated secondary antibodies (Beyotime) for 1 h (RT, dark). Nuclei were counterstained with DAPI, and images were acquired using a Leica SP5 confocal microscope. Multiplex immunohistochemistry mIHC was performed using a tyramide signal amplification (TSA) protocol. Slides were baked at 65 °C, dewaxed, rehydrated, and subjected to antigen retrieval (pH 8.0 EDTA or pH 6.0 citrate buffer, microwave). Non-specific binding was blocked with 3% H₂O₂ and goat serum. Primary antibodies were incubated overnight at 4 °C, followed by HRP-conjugated secondary antibodies and TSA fluorescence labeling. Staining cycles were repeated for sequential markers. Nuclei were counterstained with DAPI, and images were acquired using a multiplex fluorescence imaging system. Quantitative analysis normalized target protein signal intensity to per cell to avoid cell density interference. NK cell cytotoxicity assay Splenic NK cells from BALB/C mice were isolated using the NK Cell Isolation Kit (Miltenyi). Isolated NK cells were resuspended in RPMI 1640 medium supplemented with 10% FBS, 1% penicillin, and 50 µM β-mercaptoethanol. Effector (NK) and target (4T07) cells were co-cultured at specified ratios in triplicate round-bottom 96-well plates for 4 h at 37 °C. Cytotoxicity was assessed using the CCK-8 kit (Apex Bio). The NK cell-mediated cytotoxicity rate was calculated as follows: Cytotoxicity (%) = [1 − (OD(NK+Tumor) − OD(NK))/OD(Tumor)] × 100%, where OD₍NK+Tumor₎ = absorbance of NK and 4T07 co-culture group, OD₍NK₎ = absorbance of NK cell-only group, and OD₍Tumor₎ = absorbance of 4T07 cell-only group. All absorbance data were calculated as the mean value of three technical replicates, and subsequent statistical analysis of biological replicates was performed using GraphPad Prism 8. In vitro transwell NK migration assay 4T07 cells were washed with PBS and cultured in serum-free medium for 16 h to generate conditioned medium (CM). NK cells were washed and resuspended in serum-free medium. A total of 2.5 × 10⁵ NK cells in 100 µl medium were seeded into the upper chamber, and 600 µl of CM was added to the lower chamber. After 4 h incubation at 37 °C, migrated NK cells in the lower chamber were collected, counted via flow cytometry using Precision Count Beads™ (BioLegend). ELISA analysis Concentrations of IL-15 and IL-18 in culture medium were quantified using specific ELISA kits (Elabscience) according to the manufacturer’s instructions, with concentrations calculated from standard curves. Statistical analysis All experiments were performed at least three times unless indicated otherwise. Two-sided paired Student’s t-tests were used for two-group comparisons. Survival curves were generated by the Kaplan–Meier method and compared via the log-rank test. Analyses were conducted using Prism 8 (GraphPad). Data are presented as mean ± SD, with exact P values reported for all comparisons. Results Dormant TNBC cells specifically express COL9A1 To investigate genes associated with TNBC dormancy, we conducted RNA sequencing (RNA-seq) transcriptomic profiling of two murine TNBC cell lines with different dormant capacities, specifically 4T07 and 4T1. This cell line pair is a classic validated model for TNBC dormancy research, with 4T07 and 4T1 representing the dormant and proliferative/metastatic phenotypes, respectively28, 29, 30. To ensure transparency and reproducibility of RNA-seq, supporting data are provided in the Supplementary Materials. Supplementary Table S1A shows KEGG pathway enrichment results corresponding to Fig. 1A, and Supplementary Table S1B contains raw RNA-seq and DEG data underlying Fig. 1D. Transcriptomic profiling of dormant 4T07 and metastatic 4T1 cells revealed enrichment of pathways related to focal adhesion, PI3K-Akt signaling, and ECM-receptor interaction (Fig. 1A, B). To visualize the core differentially expressed genes in these pathways, we selected DEGs overlapping with the three pathways, filtered out those with log₂ (fold change) < 3, and ranked the remaining genes by log₂ (fold change) in descending order for heatmap analysis (Fig. 1C). The complete list of genes included in the heatmap is available in Supplementary Table S1B. Among differentially expressed genes, Col9a1 emerged as a top candidate, with significantly higher expression in 4T07 cells compared to 4T1 cells, as validated by qPCR (Fig. 1E). Western blot analysis of cell culture supernatants confirmed elevated COL9A1 protein levels in 4T07 cells (Fig. 1F), with Coomassie brilliant blue total protein loading control (Fig. S1). Furthermore, analysis using the Gene Expression Profiling Interactive Analysis 2(GEPIA2) database demonstrated that high COL9A1 expression correlates with improved overall survival in breast cancer patients, as determined by the log-rank test (HR = 0.63, p = 0.0073; Fig. 1G). Collectively, these data identify COL9A1 as a key regulator associated with the dormant state of TNBC cells. Fig. 1DEGs and pathway enrichment Profiling via RNA-seq in 4T1 versus 4T07 Cells. A KEGG analysis comparing gene-expression data from dormancy(4T07) and metastasis(4T1) cells (n = 3)32, 33, 34. B GSEA of the transcriptional signature in 4T07 cells compared with 4T1 cells (n = 3). C Heatmap of core differentially expressed genes (DEGs) between 4T07 and 4T1 cells. Genes were selected by: (1) overlapping with Focal adhesion, PI3K-Akt signaling pathway, and ECM-receptor interaction pathways; (2) log₂ (fold change) ≥ 3; (3) ranked in descending order of log₂ (fold change). Detailed information of these genes is provided in Supplementary Table S1B. D Volcano plots showing DEGs in 4T07 cells and 4T1 cells (n = 3). E Quantitative PCR with reverse transcription (RT–qPCR) analysis of DEGs (n = 3). F COL9A1 protein expression in 4T1 and 4T07 cell culture supernatants. Western blotting of debris-removed, concentrated conditioned media; equal loading confirmed by Coomassie staining (Fig. S1). G Overall survival of breast cancer patients stratified by median Col9a1 expression (high vs. low). Survival analysis was performed using the log-rank test (hazard ratio (HR) = 0.63, 95% confidence interval shown by dashed lines, p = 0.0073).Data are the mean ± SD. p Values were calculated by two tailed unpaired Student’s t test. All experiments were performed with 3 biological replicates. DEGs and pathway enrichment Profiling via RNA-seq in 4T1 versus 4T07 Cells. A KEGG analysis comparing gene-expression data from dormancy(4T07) and metastasis(4T1) cells (n = 3)32, 33, 34. B GSEA of the transcriptional signature in 4T07 cells compared with 4T1 cells (n = 3). C Heatmap of core differentially expressed genes (DEGs) between 4T07 and 4T1 cells. Genes were selected by: (1) overlapping with Focal adhesion, PI3K-Akt signaling pathway, and ECM-receptor interaction pathways; (2) log₂ (fold change) ≥ 3; (3) ranked in descending order of log₂ (fold change). Detailed information of these genes is provided in Supplementary Table S1B. D Volcano plots showing DEGs in 4T07 cells and 4T1 cells (n = 3). E Quantitative PCR with reverse transcription (RT–qPCR) analysis of DEGs (n = 3). F COL9A1 protein expression in 4T1 and 4T07 cell culture supernatants. Western blotting of debris-removed, concentrated conditioned media; equal loading confirmed by Coomassie staining (Fig. S1). G Overall survival of breast cancer patients stratified by median Col9a1 expression (high vs. low). Survival analysis was performed using the log-rank test (hazard ratio (HR) = 0.63, 95% confidence interval shown by dashed lines, p = 0.0073).Data are the mean ± SD. p Values were calculated by two tailed unpaired Student’s t test. All experiments were performed with 3 biological replicates. The proliferation and migration abilities of TNBC cells were diminished following the overexpression of Col9a1 The overexpression of Col9a1 significantly reduced the proliferation and migration abilities of TNBC cells. Our findings indicate that the expression levels of Col9a1 are lower in two human TNBC cell lines, MDA-MB-231 and Hs578T (Fig. S2A). After successful transfection of the OECol9a1 plasmid into 4T1 cells and the aforementioned TNBC cell lines (Fig. 2A), with Coomassie brilliant blue loading control (Fig. S2B). CCK8 assay results demonstrated a significant decrease in the proliferation ability of these cells (Fig. 2B). Additionally, wound healing and Transwell assays revealed that the migration abilities of TNBC cells overexpressing Col9a1 were markedly diminished (Fig. 2C and D). Supplementary proliferation control assays confirmed that the observed migratory differences were independent of cell proliferation (Fig. S2C–D). Furthermore, TNBC cells overexpressing Col9a1 exhibited G0/G1 phase cell cycle arrest (Fig. 2E), accompanied by a significant increase in their apoptosis rate (Fig. 2F). Following orthotopic injection of Col9a1-overexpressing 4T1 cells into BALB/C mice, tumor volume in the Col9a1 overexpression group was significantly smaller than that in the control group from Day 12 onwards (Fig. 2G). At the experimental endpoint, euthanasia and dissection confirmed that tumor volume in the overexpression group was notably reduced (Fig. 2H–I). Moreover, tail vein injection experiments showed a significant decrease in the number of lung nodules in the Col9a1 overexpression group (Fig. 2J). Taken together, these findings confirm that overexpression of Col9a1 significantly inhibits cell proliferation and migration, and induces G0/G1 phase cell cycle arrest, thereby contributing to a dormant phenotype in TNBC cells. Fig. 2Overexpression of Col9a1 reduces the proliferation and migration abilities of proliferative TNBC cells. A Western blot (WB) analysis confirms the successful establishment of the Col9a1 overexpression cell model; equal loading confirmed by Coomassie staining (Fig. S2B). B The CCK8 assay demonstrates the proliferation ability of TNBC cells with Col9a1 overexpression. (4T1 cells, n = 9; MDA-MB-231 and Hs578T cells, n = 6). C Wound healing assay assesses the wound healing capability of three types of TNBC cells that overexpress Col9a1(n = 3). (Proliferation control: Fig. S2C) D The Transwell migration assay indicates that Col9a1 overexpression inhibits migration (n = 3). (Proliferation control: Fig. S2D) E Flow cytometry results reveal that Col9a1 overexpression induces cell cycle arrest in the G0/G1 phase in TNBC cells (n = 3). F Flow cytometry is employed to evaluate apoptosis in TNBC cells (n = 3). G–I Tumor volume changes in mice following orthotopic injection of 4T1 cells overexpressing Col9a1 are shown (n = 7). J A significant reduction in the number of lung nodules is observed after Col9a1 overexpression (n = 6). Data are the mean ± SD. p Values were calculated by two tailed unpaired Student’s t test. All experiments were performed with 3 biological replicates. Overexpression of Col9a1 reduces the proliferation and migration abilities of proliferative TNBC cells. A Western blot (WB) analysis confirms the successful establishment of the Col9a1 overexpression cell model; equal loading confirmed by Coomassie staining (Fig. S2B). B The CCK8 assay demonstrates the proliferation ability of TNBC cells with Col9a1 overexpression. (4T1 cells, n = 9; MDA-MB-231 and Hs578T cells, n = 6). C Wound healing assay assesses the wound healing capability of three types of TNBC cells that overexpress Col9a1(n = 3). (Proliferation control: Fig. S2C) D The Transwell migration assay indicates that Col9a1 overexpression inhibits migration (n = 3). (Proliferation control: Fig. S2D) E Flow cytometry results reveal that Col9a1 overexpression induces cell cycle arrest in the G0/G1 phase in TNBC cells (n = 3). F Flow cytometry is employed to evaluate apoptosis in TNBC cells (n = 3). G–I Tumor volume changes in mice following orthotopic injection of 4T1 cells overexpressing Col9a1 are shown (n = 7). J A significant reduction in the number of lung nodules is observed after Col9a1 overexpression (n = 6). Data are the mean ± SD. p Values were calculated by two tailed unpaired Student’s t test. All experiments were performed with 3 biological replicates. DCCs awaken following Col9a1 knockdown Following the knockdown of Col9a1, DCCs were activated. We first established an in vitro model to assess the effects of Col9a1 knockdown. Screening revealed that both shCol9a1-1 and shCol9a1-3 resulted in a significant reduction of Col9a1 expression at the RNA and protein levels, with Coomassie brilliant blue loading control (Fig. S3A-B and Fig. 3A). CCK8 assay results indicated that Col9a1 knockdown significantly enhanced cell proliferation capacity (Fig. 3B), while wound healing and Transwell assays demonstrated markedly elevated migration abilities in the knockdown group (Fig. 3C, D). Supplementary proliferation control assays confirmed that the observed migratory differences were independent of cell proliferation (Fig. S3C-D). Flow cytometry analysis showed a reduced proportion of cells in quiescent G0/G1 phase arrest, accompanied by increased S and G2/M phase populations in the knockdown group (Fig. 3E). Additionally, Col9a1 knockdown led to a significant decrease in the apoptosis rate of 4T07 cells (Fig. 3F). In vivo experiments further validated these effects, with the knockdown group exhibiting significantly increased tumor volume (Fig. 3G, H). Corresponding tail vein injection assays revealed a notable increase in the number of metastatic lung nodules in the Col9a1 knockdown group (Fig. 3I). Altogether, these results indicate that Col9a1 knockdown enhances cell viability, increases the proportion of dividing cells, and promotes migration, confirming the reactivation of dormant TNBC cells. Fig. 3Decreased expression of Col9a1 enhances the proliferation and migration capabilities of dormant TNBC cells. A The expression levels of Col9a1 in 4T07 cells transfected with shCol9a1-1 and shCol9a1-3 were assessed using WB; equal loading confirmed by Coomassie staining (Fig. S3B). B The impact of Col9a1 knockdown on the viability of 4T07 cells was evaluated through the CCK8 assay (n = 9). C, D Following the reduction of COL9A1 protein expression, both wound healing and Transwell migration capabilities were found to be enhanced (n = 3). (Proliferation control: Fig. S3C-D). E The effect of COL9A1 protein knockdown on the distribution of cell cycle stages in dormant tumor cells was investigated (n = 3). F The apoptosis rate of 4T07 cells after Col9a1 knockdown was measured using flow cytometry (n = 3). G, H Tumor volume changes in mice were monitored following the orthotopic injection of Col9a1 knockdown 4T07 cells (n = 5). I The number of metastatic lung nodules was quantified after injection of GFP-tagged shCol9a1 lentivirus-transfected Col9a1-knockdown 4T07 cells into mice (n = 6); green fluorescence denotes stable shCol9a1 transfection. Data are the mean ± SD. p Values were calculated by two tailed unpaired Student’s t test. All experiments were performed with 3 biological replicates. Decreased expression of Col9a1 enhances the proliferation and migration capabilities of dormant TNBC cells. A The expression levels of Col9a1 in 4T07 cells transfected with shCol9a1-1 and shCol9a1-3 were assessed using WB; equal loading confirmed by Coomassie staining (Fig. S3B). B The impact of Col9a1 knockdown on the viability of 4T07 cells was evaluated through the CCK8 assay (n = 9). C, D Following the reduction of COL9A1 protein expression, both wound healing and Transwell migration capabilities were found to be enhanced (n = 3). (Proliferation control: Fig. S3C-D). E The effect of COL9A1 protein knockdown on the distribution of cell cycle stages in dormant tumor cells was investigated (n = 3). F The apoptosis rate of 4T07 cells after Col9a1 knockdown was measured using flow cytometry (n = 3). G, H Tumor volume changes in mice were monitored following the orthotopic injection of Col9a1 knockdown 4T07 cells (n = 5). I The number of metastatic lung nodules was quantified after injection of GFP-tagged shCol9a1 lentivirus-transfected Col9a1-knockdown 4T07 cells into mice (n = 6); green fluorescence denotes stable shCol9a1 transfection. Data are the mean ± SD. p Values were calculated by two tailed unpaired Student’s t test. All experiments were performed with 3 biological replicates. Tumors formed by reactivated dormant cells exhibit enhanced immune infiltration Given the immune tolerance properties of dormant tumor cells, we investigated immune cell infiltration in tumors derived from reactivated dormant cells in murine models. Flow cytometry analysis showed a marked increase in the proportion of CD45+ immune cells in tumor tissues from the Col9a1 knockdown group, indicative of enhanced immune infiltration. Notably, the number of NK cells was significantly elevated in these tumors, whereas no obvious differences were detected in the T cell population (Fig. 4A). In contrast, tumor tissues formed by Col9a1-overexpressing 4T1 cells displayed impaired immune infiltration, with a prominent reduction in NK cell numbers (Fig. 4B). Multiplex fluorescent immunohistochemistry analysis of tumor tissues yielded results consistent with those of flow cytometry (Fig. 4C, D). Furthermore, the expression levels of GZMB, IFN-γ, and Perforin in NK cells adjacent to tumors formed by Col9a1-knockdown 4T07 cells were significantly upregulated (Fig. 4E), while the levels of GZMB, PRF1, and IFN-γ in NK cells surrounding Col9a1-overexpressing 4T1 cell-derived tumors were decreased (Fig. 4F, Fig. S4C). Collectively, these data suggest that Col9a1 downregulation reactivates dormant TNBC cells, leading to reduced immune tolerance and increased immune infiltration in the tumor microenvironment, with NK cells potentially exerting a critical role in this process. Initial observations suggested that NK cell infiltration around reactivated cancer cells did not suppress tumor growth, which seemingly contradicts the aforementioned findings; however, this interpretation is potentially misleading. Subsequent functional depletion experiments demonstrated that selective NK cell depletion in Col9a1-knockdown mice resulted in a significant increase in tumor volume (Fig. S5), which conclusively confirmed the tumor-suppressive cytotoxic function of NK cells in this model. We therefore speculate that the anti-tumor inhibitory effect of NK cells may be counteracted by the rapid proliferative capacity of reactivated tumor cells. Fig. 4Enhanced immune infiltration of awakened dormant tumor cells. T cells were defined as CD45⁺CD3⁺ subset and NK cells as CD45⁺CD3⁻CD49b⁺ subset in viable immune cells for all flow cytometry analyses. A Flow cytometry results showing the immune infiltration of tumor tissues from mice treated with 4T07 cells expressing low levels of Col9a1(n = 5). B Flow cytometry results showing the immune infiltration of tumor tissues from mice treated with 4T1 cells overexpressing Col9a1(n = 5). C, D Immunofluorescence staining and quantitative analysis results of specified molecules in tumor tissues. Scale bar, 50 μm (n = 5). E, F Flow cytometry results showing the expression levels of NK cell effector molecules GZMB, PRF1, and IFN-γ in the pre-gated CD45⁺CD3⁻CD49b⁺ NK cell subset in mouse tumor tissues (n = 5). Data are the mean ± SD. p Values were calculated by two tailed unpaired Student’s t test. Enhanced immune infiltration of awakened dormant tumor cells. T cells were defined as CD45⁺CD3⁺ subset and NK cells as CD45⁺CD3⁻CD49b⁺ subset in viable immune cells for all flow cytometry analyses. A Flow cytometry results showing the immune infiltration of tumor tissues from mice treated with 4T07 cells expressing low levels of Col9a1(n = 5). B Flow cytometry results showing the immune infiltration of tumor tissues from mice treated with 4T1 cells overexpressing Col9a1(n = 5). C, D Immunofluorescence staining and quantitative analysis results of specified molecules in tumor tissues. Scale bar, 50 μm (n = 5). E, F Flow cytometry results showing the expression levels of NK cell effector molecules GZMB, PRF1, and IFN-γ in the pre-gated CD45⁺CD3⁻CD49b⁺ NK cell subset in mouse tumor tissues (n = 5). Data are the mean ± SD. p Values were calculated by two tailed unpaired Student’s t test. Awakened cancer cells upregulate NK cell activation ligands, triggering immune responses from adjacent NK cells Given the increased NK cell infiltration around awakened cancer cells, we first assessed NK cell migration using conditioned media (CM) from cultured cancer cells in vitro. Results showed that CM from awakened cancer cells recruited significantly more NK cells (Fig. 5A). Furthermore, cancer cell awakening was accompanied by elevated transcriptional levels of Il2, Il15, Il18, and Il21 (Fig. 5B), and ELISA assays confirmed the increased secretion of IL-15 and IL-18 by these cells (Fig. 5C). Co-culture experiments subsequently demonstrated that awakened cancer cells were significantly more sensitive to NK cell-mediated cytotoxicity (Fig. 5D). We next evaluated MHC I expression, which was markedly higher in dormancy-associated cell populations (Fig. S6A–B). Flow cytometry analysis, performed in both in vitro and in vivo settings, revealed a higher proportion of RAE-1γ⁺ cells in the EP-CAM⁺ tumor cell subset of proliferative (awakened) versus DCCs /tissues (Fig. 5E–H). To elucidate the functional relevance of RAE-1γ, we blocked the NKG2D receptor on NK cells prior to co-culture with reactivated cancer cells, which resulted in attenuated NK cell cytotoxicity (Fig. 5I). Consistent with this, siRNA-mediated knockdown of Raet1c in reactivated cancer cells yielded similar results, indicating that RAE-1γ contributes to the enhanced sensitivity of reactivated cancer cells to NK cell killing (Fig. 5J–K).Since RAE-1 expression is regulated by E2F transcription factors35, and phosphorylation of RB at T373 is a rate-limiting step for E2F activation36, we investigated the interaction between RB and E2F2 (a key pro-proliferative E2F family member) in our model. As shown in Fig. 5L, pRB-T373 levels were elevated in reactivated cancer cells. Co-immunoprecipitation (CoIP) assays confirmed that E2F2 binds to RB in both parental and Col9a1-overexpressing 4T1 cells; notably, this interaction was significantly enhanced in Col9a1-overexpressing (dormant) 4T1 cells (Fig. 5M). Additionally, knockdown of E2f2 resulted in a significant reduction in Raet1c mRNA expression, indicating that E2f2 positively regulates Raet1c transcription (Fig. 5N). Taken together, these findings demonstrate that in awakened cancer cells, E2F2 dissociates from RB, gains transcriptional activity, and thereby upregulates Raet1c expression. Fig. 5The awakened cancer cells recruit and activate NK cells to exert cytotoxic effects. A A schematic diagram of the Transwell migration assay, showing the number of NK cells that migrated into the culture medium of 4T07 cells (n = 3). B Quantitative detection of mRNA levels of IL-2, IL-15, IL-18, and IL-21 by qPCR (n = 3). C The content of IL-15(n = 3) and IL-18 (n = 5) in the culture medium of 4T07 cells with NC or Col9a1 knockdown. D Schematic diagram of NK cell-mediated cytotoxicity assay, and comparative analysis of cytotoxic killing efficiency after co-culturing NK cells with negative control (NC) or Col9a1-knockdown 4T07 cells for 4 h under different effector-to-target (E: T) ratios (n = 3). E Flow cytometric quantification of RAE-1γ in NC and Col9a1-overexpressing 4T1 cells (n = 3). F Flow cytometric quantification of RAE-1γ in NC and Col9a1 knockdown 4T07 cells (n = 3). G Flow cytometric quantification of RAE-1γ in the EP-CAM⁺ tumor cell subset from tumor tissues of NC and Col9a1-overexpressing 4T1 cell-implanted mice (n = 5). H Flow cytometric quantification of RAE-1γ in the EP-CAM⁺ tumor cell subset from tumor tissues of NC and Col9a1-knockdown 4T07 cell-implanted mice (n = 5). I The percentage of NK-driven cytotoxicity was measured in control and Col9a1-knockdown 4T07 cells co-cultured with NK cells after the addition of isotype control antibodies or anti-NKG2D antibodies to block the interaction between the NKG2D receptor on NK cells and the NKG2D ligands expressed on tumor cells. Cytotoxicity was quantified by CCK8 assay (n = 5). J, K After knockdown of Raet1c, the NK-driven cytotoxicity percentage of control and Col9a1-knockdown 4T07 cells was assessed (J, n = 3; K, n = 5). L In 4T07 cells, the knockdown of Col9a1 increased the phosphorylation of RB at T373. M In 4T1 cells, co-immunoprecipitation of RB and E2F2 showed increased binding of RB and E2F2 after overexpression of Col9a1. N After knocking down E2f2 in Col9a1-knockdown 4T07 cells and Col9a1-overexpressing 4T1 cells, the expression levels of E2f2 and Raet1c were quantitatively detected using qPCR technology (n = 3). Data are the mean ± SD. p Values were calculated by two tailed unpaired Student’s t test. The awakened cancer cells recruit and activate NK cells to exert cytotoxic effects. A A schematic diagram of the Transwell migration assay, showing the number of NK cells that migrated into the culture medium of 4T07 cells (n = 3). B Quantitative detection of mRNA levels of IL-2, IL-15, IL-18, and IL-21 by qPCR (n = 3). C The content of IL-15(n = 3) and IL-18 (n = 5) in the culture medium of 4T07 cells with NC or Col9a1 knockdown. D Schematic diagram of NK cell-mediated cytotoxicity assay, and comparative analysis of cytotoxic killing efficiency after co-culturing NK cells with negative control (NC) or Col9a1-knockdown 4T07 cells for 4 h under different effector-to-target (E: T) ratios (n = 3). E Flow cytometric quantification of RAE-1γ in NC and Col9a1-overexpressing 4T1 cells (n = 3). F Flow cytometric quantification of RAE-1γ in NC and Col9a1 knockdown 4T07 cells (n = 3). G Flow cytometric quantification of RAE-1γ in the EP-CAM⁺ tumor cell subset from tumor tissues of NC and Col9a1-overexpressing 4T1 cell-implanted mice (n = 5). H Flow cytometric quantification of RAE-1γ in the EP-CAM⁺ tumor cell subset from tumor tissues of NC and Col9a1-knockdown 4T07 cell-implanted mice (n = 5). I The percentage of NK-driven cytotoxicity was measured in control and Col9a1-knockdown 4T07 cells co-cultured with NK cells after the addition of isotype control antibodies or anti-NKG2D antibodies to block the interaction between the NKG2D receptor on NK cells and the NKG2D ligands expressed on tumor cells. Cytotoxicity was quantified by CCK8 assay (n = 5). J, K After knockdown of Raet1c, the NK-driven cytotoxicity percentage of control and Col9a1-knockdown 4T07 cells was assessed (J, n = 3; K, n = 5). L In 4T07 cells, the knockdown of Col9a1 increased the phosphorylation of RB at T373. M In 4T1 cells, co-immunoprecipitation of RB and E2F2 showed increased binding of RB and E2F2 after overexpression of Col9a1. N After knocking down E2f2 in Col9a1-knockdown 4T07 cells and Col9a1-overexpressing 4T1 cells, the expression levels of E2f2 and Raet1c were quantitatively detected using qPCR technology (n = 3). Data are the mean ± SD. p Values were calculated by two tailed unpaired Student’s t test. The COL9A1-AKT-p27 axis orchestrates the dormancy of breast cancer cells We first analyzed the expression of matrix metalloproteinase (MMP) family members in 4T07 and 4T1 cells using RNA-seq data from the Gene Expression Omnibus (GEO) database (accession number: GSE203296). This analysis revealed that Mmp9, Mmp10, Mmp13, and Mmp24 were expressed at significantly lower levels in dormant 4T07 cells compared to metastatic 4T1 cells (Fig. 6A). Given that MMP9 is a key effector of tumor progression, responsible for degrading basement membrane components such as collagen and laminin, we validated its expression and found that MMP9 protein levels were indeed higher in 4T1 cells (Fig. 6B). Co-immunoprecipitation (CoIP) assays further confirmed a physical interaction between COL9A1 and MMP9 in both cell lines (Fig. 6C). Functional validation demonstrated that Mmp9 overexpression in 4T07 cells reduced COL9A1 protein levels (Fig. 6D), whereas Mmp9 knockdown in 4T1 cells led to a concomitant increase in COL9A1 expression (Fig. 6E, Fig. S7A), with Coomassie brilliant blue loading control for both blots (Fig. S7B), confirming that MMP9 enzymatically degrades COL9A1.Building on our earlier transcriptomic findings, which identified focal adhesion and PI3K-Akt signaling as pathways enriched in dormant cells (Fig. 1A–B), we assessed the expression of key signaling proteins in these cascades. In dormant breast cancer cells, the phosphorylation ratios of FAK (pFAK/FAK) and AKT (pAKT/AKT) were decreased, while the dormancy marker p27 was elevated37 (Fig. 6F). Notably, Col9a1 knockdown reversed this profile, leading to the upregulation of pFAK/FAK and pAKT/AKT, and a decrease in p27 expression (Fig. 6G). Conversely, Col9a1 overexpression in 4T1 cells suppressed pFAK/FAK and pAKT/AKT levels and elevated p27 expression (Fig. 6H). Immunofluorescence staining showed that Col9a1 knockdown resulted in reduced p27 protein expression in both in vitro (Fig. 6I) and in vivo (Fig. 6J) models, with knockdown efficiency confirmed in in vitro (Fig. S8A) and in vivo (Fig. S8B) samples. Altogether, these results indicate that COL9A1, whose stability is negatively regulated by MMP9, sustains the dormant state of breast cancer cells by suppressing the focal adhesion and PI3K-Akt signaling pathways and upregulating the cyclin-dependent kinase inhibitor p27. Fig. 6The COL9A1-AKT-p27 axis regulates the tumor cell dormancy program. A RNA-seq heatmap and volcano plot display the differential Mmp genes in 4T07 and 4T1 cells. B Expression of MMP9 protein in 4T07 and 4T1 cells (n = 3). C coIP assay detects the binding between COL9A1 and MMP9. D Expression of COL9A1 protein in 4T07 cells after overexpression of Mmp9; equal loading confirmed by Coomassie staining (Fig. S6B). E Expression of COL9A1 protein in 4T1 cells after knockdown of Mmp9, equal loading confirmed by Coomassie staining (Fig. S6B). F Expression of proteins related to the FAK-AKT-p27 axis in 4T07 and 4T1 cells. G Expression of proteins related to the FAK-AKT-p27 axis in NC and Col9a1-knockdown 4T07 cells. H Expression of FAK-AKT-p27 axis-related proteins in NC and Col9a1-overexpressing 4T1 cells. I Immunofluorescence staining of p27 protein in NC and Col9a1-knockdown 4T07 cells (n = 3). J Immunofluorescence staining of p27 protein in tumor tissue sections formed in situ from NC and Col9a1-knockdown 4T07 cells (n = 5). Data are the mean ± SD. p Values were calculated by two tailed unpaired Student’s t test. The COL9A1-AKT-p27 axis regulates the tumor cell dormancy program. A RNA-seq heatmap and volcano plot display the differential Mmp genes in 4T07 and 4T1 cells. B Expression of MMP9 protein in 4T07 and 4T1 cells (n = 3). C coIP assay detects the binding between COL9A1 and MMP9. D Expression of COL9A1 protein in 4T07 cells after overexpression of Mmp9; equal loading confirmed by Coomassie staining (Fig. S6B). E Expression of COL9A1 protein in 4T1 cells after knockdown of Mmp9, equal loading confirmed by Coomassie staining (Fig. S6B). F Expression of proteins related to the FAK-AKT-p27 axis in 4T07 and 4T1 cells. G Expression of proteins related to the FAK-AKT-p27 axis in NC and Col9a1-knockdown 4T07 cells. H Expression of FAK-AKT-p27 axis-related proteins in NC and Col9a1-overexpressing 4T1 cells. I Immunofluorescence staining of p27 protein in NC and Col9a1-knockdown 4T07 cells (n = 3). J Immunofluorescence staining of p27 protein in tumor tissue sections formed in situ from NC and Col9a1-knockdown 4T07 cells (n = 5). Data are the mean ± SD. p Values were calculated by two tailed unpaired Student’s t test. Awakened cancer cells exhibit increased sensitivity to docetaxel chemotherapy Docetaxel, a first-line chemotherapeutic agent for breast cancer, exerts its anti-tumor effects by stabilizing microtubule assembly and inhibiting microtubule depolymerization, thereby arresting cancer cells in the mitotic M phase. To determine whether the awakening of DCCs alters their chemosensitivity, we treated both dormant and awakened TNBC cells with a gradient of docetaxel concentrations in vitro. Notably, awakened cancer cells displayed marked sensitivity to low docetaxel concentrations (Fig. 7A), and this finding was further validated by colony formation assays (Fig. 7B–C). We then extended these observations to an in vivo orthotopic tumor model, where docetaxel treatment was initiated one week after inoculation of either dormant or awakened cancer cells (Fig. 7D). Tumor volume in the docetaxel-treated awakened cell group began to decrease from Day 12 onwards (Fig. 7E). At the experimental endpoint (21 days), docetaxel treatment resulted in a significant reduction in tumor volume in mice bearing awakened tumors compared to their untreated counterparts; in contrast, the treatment had no discernible effect on tumor volume in the dormant tumor group (Fig. 7F, G). Correspondingly, docetaxel treatment significantly prolonged the overall survival of mice bearing awakened tumors (Fig. 7H). These results demonstrate that the awakening of dormant TNBC cells reverses their chemoresistant phenotype, enhances their susceptibility to docetaxel, and ultimately translates to improved survival outcomes. In summary, our study delineates a comprehensive regulatory network governing TNBC cell dormancy and immune evasion. Upstream, COL9A1 expression is negatively regulated by MMP9-mediated degradation. In TNBC cells, COL9A1 suppresses the FAK-AKT signaling pathway, which in turn promotes the accumulation of the cell cycle inhibitor p27, inducing G0/G1 cell cycle arrest and sustaining the dormant state. Concurrently, reduced AKT phosphorylation facilitates the physical interaction between RB and the transcription factor E2F2. This interaction sequesters E2F2, abrogating its transcriptional activation of Raet1c and leading to decreased cell surface expression of RAE-1γ. The downregulation of this NKG2D ligand attenuates NK cell-mediated cytotoxicity, enabling dormant tumor cells to evade immune surveillance (Fig. 7I). Fig. 7The awakened cancer cells are more sensitive to chemotherapy drugs. A Cell viability of NC group and Col9a1 knockdown group treated with Docetaxel at concentrations of 0, 0.1, 0.5, and 1 µg/mL (n = 6). B, C NC group and Col9a1 knockdown group 4T07 cells were seeded at 2 × 103 in six-well plates and cultured for 7 days to form colonies (n = 3). There was no significant change in Colony formation activity in the NC group after treatment, while the Colony formation activity in the Col9a1 knockdown group significantly decreased after treatment. D Representation of the experimental design from (E to G). E Tumor volume (mm³) growth curves of mice treated with NC + solvent, NC + DTX, shCol9a1 + solvent, and shCol9a1 + DTX (n = 5). F, G Comparison of tumor volume in mice after euthanasia on Day 21. H Evaluate the survival time of mice injected with 2 × 106 4T07 cells via the tail vein and treated with intraperitoneal injection of DTX (n = 8). I Schematic summary of COL9A1 regulation of dormancy mechanisms in TNBC cells (by Figdraw (https://www.figdraw.com)). Data are the mean ± SD. p Values were calculated by two tailed unpaired Student’s t test. The awakened cancer cells are more sensitive to chemotherapy drugs. A Cell viability of NC group and Col9a1 knockdown group treated with Docetaxel at concentrations of 0, 0.1, 0.5, and 1 µg/mL (n = 6). B, C NC group and Col9a1 knockdown group 4T07 cells were seeded at 2 × 103 in six-well plates and cultured for 7 days to form colonies (n = 3). There was no significant change in Colony formation activity in the NC group after treatment, while the Colony formation activity in the Col9a1 knockdown group significantly decreased after treatment. D Representation of the experimental design from (E to G). E Tumor volume (mm³) growth curves of mice treated with NC + solvent, NC + DTX, shCol9a1 + solvent, and shCol9a1 + DTX (n = 5). F, G Comparison of tumor volume in mice after euthanasia on Day 21. H Evaluate the survival time of mice injected with 2 × 106 4T07 cells via the tail vein and treated with intraperitoneal injection of DTX (n = 8). I Schematic summary of COL9A1 regulation of dormancy mechanisms in TNBC cells (by Figdraw (https://www.figdraw.com)). Data are the mean ± SD. p Values were calculated by two tailed unpaired Student’s t test. Discussion Col9a1 is a core gene for maintaining TNBC cell dormancy. Col9a1 overexpression inhibits breast cancer cell proliferation and migration by inducing G0/G1 quiescent cell cycle arrest. In contrast, Col9a1 knockdown in 4T07 cells reverses this quiescent phenotype to promote proliferation and migration. In vivo, reduced Col9a1 reactivates dormant tumor cells, recruits NK-dominant immune infiltration, and disrupts the core quiescent features that maintain TNBC cell dormancy. Additionally, reactivated dormant cells show significantly enhanced sensitivity to first-line TNBC chemotherapeutics. These findings deepen our understanding of the molecular mechanisms driving TNBC dormant cell reactivation and identify Col9a1 as a potential therapeutic target for TNBC. Notably, its dual role in dormancy regulation and immune modulation distinguishes COL9A1 from other ECM-derived regulators. Our RNA-seq results confirm that TNBC dormancy is closely associated with the ECM-receptor interaction pathway, consistent with Dong et al.38. ECM components (type I/III/VII collagens, fibronectin) are well-characterized regulators of tumor dormancy via modulating cell cycle or cell-matrix interactions9, 24, 26. As a FACIT collagen family member, COL9A1 differs fundamentally from fibrillar collagens: it does not form fibers, but modulates ECM stiffness by binding to fibers and proteoglycans, endowing it with a unique role in dormancy regulation. Mechanistically, COL9A1 inhibits the FAK-AKT pathway to upregulate the dormancy marker p27 and maintain cell cycle arrest, consistent with the core dormancy mechanisms of type III collagen (DDR1-dependent) and type VII collagen (FAK-YAP axis-inhibitory)9, 26. The key innovation of this study is the identification of COL9A1’s dual regulatory function: it regulates the cell cycle via the FAK-AKT-p27 axis and mediates immune evasion via RAE-1γ/MHC I modulation, a feature unreported in other dormancy-related collagens9, 24. Notably, COL9A1 is functionally distinct from other matrix proteins: type I collagen promotes dormancy-to-metastasis transition, while COL9A1 is highly expressed in dormant cells to suppress metastasis; type III collagen acts via DDR1, while COL9A1 relies on the FAK-AKT-p27 axis; type VII collagen lacks immune regulatory functions, while COL9A1 has both immune regulatory activity and prognostic value; fibronectin secretion is TGFβ-mediated, while COL9A1 is tumor-derived and regulates cell-autonomous and non-cell-autonomous mechanisms9, 23, 24, 26. In summary, COL9A1 is a novel ECM-derived dormancy regulator that links cell cycle control and immune evasion, complementing the collagen family in the tumor dormancy regulatory network and providing a new theoretical basis for TNBC recurrence research and targeted therapy. Accumulating evidence has shown that the MMP family, especially MMP9, regulates tumor microenvironment homeostasis by degrading ECM collagen components. For example, MMP9 degrades type I collagen to participate in ECM remodeling and tumor progression39, 40, consistent with the MMP9-COL9A1 interaction and enzymatic degradation observed in our study. COL9A1 expression follows classic MMP-mediated ECM remodeling pathway, consistent with other ECM collagens. Additionally, collagens mediate tumor immune evasion by regulating immune cell infiltration: DDR1, for instance, binds to collagen to form an immune barrier that inhibits T cell infiltration41. In contrast, COL9A1 impairs NK cell cytotoxicity by downregulating RAE-1γ and upregulating MHC I molecules, identifying a novel non-barrier-dependent mechanism of collagen-mediated immune evasion. This aligns with Bushnell et al.42, who reported that dormant cells evade immunity by regulating NK cell-activating ligands, clarifying COL9A1’s unique role in the ECM-immune evasion regulatory network. GEPIA2 database analysis showed that breast cancer patients with high Col9a1 expression had significantly higher overall survival rates than those with low expression. Our study identified Col9a1 as the most significantly differentially expressed gene between dormant 4T07 and proliferative 4T1 cells. Col9a1 overexpression in 4T1 cells significantly reduced tumor volume and enhanced cellular dormancy maintenance in vivo. Conversely, Col9a1 knockdown in 4T07 cells enhanced cell proliferation and migration, reduced apoptosis, decreased the proportion of G0/G1 phase cells and increased dividing phase cells, confirming dormant cell reactivation. DCCs evade T cell attack by forming a hypoxic niche containing immunosuppressive fibroblasts, dysfunctional dendritic cells, and highly exhausted T cells15. Bushnell et al.42 reported that NK cells are critical for maintaining the tumor cell dormant phenotype, with dormant cancer stem cells (CSCs) resistant to NK cell cytotoxicity and proliferative CSCs more susceptible—consistent with our results. Furthermore, our study clarifies that the resistance of TNBC dormant CSCs to NK cell cytotoxicity is driven by the COL9A1-mediated quiescent cell cycle state, which directly links the core cytological features of dormancy to its immune evasion properties. Additionally, we found that quiescent cells have elevated MHC I expression and reduced RAE-1γ expression. Quiescent cells do not evade immune detection via antigen presentation; instead, they downregulate NK cell-activating ligands and induce a quiescent stem cell-like phenotype to resist NK cell cytotoxicity. Reactivated dormant TNBC cells have increased susceptibility to NK cell-mediated cytotoxicity. In vivo, tumor-infiltrating NK cells in Col9a1-knockdown tumors had elevated IFN-γ, perforin, and granzyme expression, and Col9a1-knockdown cells secreted higher levels of NK cell-recruiting cytokines. Col9a1 downregulation triggers dormant cell reactivation, and reactivated tumor cells secrete cytokines (e.g., IL-15, IL-18) to recruit NK cells to the tumor microenvironment, thereby promoting NK cell cytotoxicity and tumor cell elimination. Col9a1-knockdown 4T07 injection in mice promotes tumor growth and lung metastasis (increased volume, more metastatic nodules). Flow cytometry shows enhanced immune infiltration in Col9a1-knockdown tumors, with NK cells as the predominant population. NK cell depletion confirms increased tumor volume in mice, indicating NK cells’ anti-tumor role in tumor microenvironment. We speculate NK cells’ anti-tumor effect is offset by reactivated cells’ rapid proliferation. DCCs exhibit resistance to standard treatments43, 44, 45, 46. These cells have entered a quiescent G0–G1 cell cycle arrest phase (the core characteristic of dormancy), rendering them less responsive to cytotoxic chemotherapy agents that primarily target actively proliferating cells17. Our findings indicate that in situ tumors formed by dormant cells do not experience significant changes in volume following the administration of chemotherapy drugs, which confirms that the chemoresistance of TNBC dormant cells is closely associated with their quiescent cell cycle state regulated by COL9A1. Notably, dormancy in TNBC is regulated not only by cell cycle quiescence (the basis of chemoresistance) but also closely associated with immune regulation; recent studies have verified that the CXCL10/CXCR3 axis is a key mediator of immune dormancy in TNBC models, further highlighting the complex relationship between cellular quiescence and immune surveillance during tumor latency47. Col9a1 knockdown enhances tumor immune infiltration and initiates NK killing program. When combined with the first-line chemotherapeutic agent, docetaxel, there was a significant reduction in tumor volume, and the survival rate of tumor-bearing mice was notably improved. Notably, reactivated cells’ enhanced docetaxel sensitivity depends on proliferation and NK-mediated cytotoxicity. NK cell depletion experiment confirmed NK cells’ critical tumor-suppressive role in vivo, as their depletion increased tumor volume in Col9a1-knockdown models—laying the foundation for NK cell-chemotherapy synergy. Reactivated cells enhance NK cell recruitment and activation via elevated cytokine secretion and activation ligand upregulation. Mechanistically, docetaxel targets proliferating reactivated cells, while NK cells clear residuals via specific signaling axes, complementing each other to improve tumor clearance. Assessing docetaxel responses in NK cell depletion models and correlating efficacy with NK cell infiltration will be our subsequent research focus, further validating the clinical potential of combination strategies. A key limitation of this study is that docetaxel treatment only reduced tumor burden in the Col9a1 KD group to WT levels, with no additional therapeutic benefit beyond baseline. The 15 mg/kg in vivo dose was selected based on widely accepted preclinical references, and while higher docetaxel concentrations were not tested in vitro, our results confirm Col9a1 knockdown enhances reactivated dormant cells’ sensitivity to low-dose docetaxel; we will explore tolerable dose escalation in future research to improve tumor control. Our research focuses on dormancy-based therapeutic strategies: maintaining proliferative tumors in a dormant state and reactivating dormant cells to enhance their sensitivity to antiproliferative drugs. While our in vivo data confirm COL9A1’s critical role in tumor dormancy, metastasis, and survival, two main limitations exist. First, no overt toxicity was observed in mouse models, but long-term safety assessments (off-target effects, organ toxicity) were not conducted. Second, findings are based on preclinical TNBC models; clinical validation in patient cohorts is needed to confirm COL9A1 expression’s clinical relevance and correlation with TNBC recurrence/survival. Future studies will address these limitations to advance the clinical translation of COL9A1-targeted strategies. Finally, COL9A1 knockdown-mediated dormant TNBC cell reactivation significantly enhances NK cell infiltration and cytotoxicity in the tumor microenvironment, providing specific clinical insights and molecular targets for optimizing TNBC immunotherapy. COL9A1 can serve as a dual target for TNBC dormancy regulation and immunotherapy. Targeted inhibition of COL9A1 can abrogate the immune evasion of dormant cells from NK cells and endogenously activate the tumor surveillance function of NK cells30, 42. On this basis, a triple-combination therapeutic strategy of “dormancy awakening + NK cell activation + chemotherapy” can be constructed: targeted intervention of Col9a1 to awaken dormant cells, combined with NK cell agonists such as IL-15/IL-18 to amplify their cytotoxic effects30, 48, and further combined with first-line chemotherapeutic agents like docetaxel to specifically kill awakened cells entering the proliferative phase49. This achieves dual clearance of dormant and proliferating cells, significantly improving the treatment response rate. In addition, the expression level of COL9A1 can be used as a predictive biomarker for the efficacy of NK cell-mediated immunotherapy in TNBC. Low COL9A1 expression patients are more likely to benefit from NK activation-based immunotherapy, providing a molecular basis for individualized clinical plans and precise TNBC recurrence intervention. In this study, we identified type IX collagen as a marker of dormant breast cancer cells. By sustaining dormancy and inducing dormancy awakening, we observed an increase in immune infiltration around the tumor following dormancy awakening, with NK cells playing a pivotal role in this process. After awakening, tumor cells secrete cytokines that recruit NK cells, facilitating immune infiltration and exerting cytotoxic effects. Furthermore, awakened tumor cells exhibit heightened sensitivity to first-line therapeutic drugs. As we advance our understanding of tumor dormancy and immunotherapy, the integration of dormancy induction with therapies specifically targeting dormant cells is poised to ultimately prevent tumor recurrence and metastasis. Electronic Supplementary Material Below is the link to the electronic supplementary material. Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2. Supplementary Material 3. Supplementary Material 3. Supplementary Material 4. Supplementary Material 4.