Targeting the MAGED2-TRIM28-FLNC axis overcomes chemoresistance in TNBC via EMT suppression Authors: Qifan Luo, Le Chen, Sai Zhao, Lei Wei, Yihao Tian, Jingwei Zhang Published: 2025 ABSTRACT Triple negative breast cancer (TNBC) is an aggressive subtype of breast cancer with a poor response to chemotherapy. Fewer than 40% of patients achieve a complete pathological response to neoadjuvant chemotherapy, demonstrating the chemoresistance characteristic of TNBC. Here, we analyze proteomic and transcriptomic datasets of TNBC patients treated with neoadjuvant chemotherapy and identify MAGED2 as a critical regulator of chemoresistance. Elevated MAGED2 expression in TNBC cells is associated with poor treatment response. Both in vitro and in vivo experiments demonstrate that MAGED2 promotes epirubicin resistance in TNBC. Mechanistically, MAGED2 interacts with TRIM28 to form an E3 ubiquitin ligase complex. MAGED2 enhances the ubiquitin ligase activity of TRIM28. MAGED2 L278/L279 residues are essential for its interaction with TRIM28. The interaction promotes MAGED2-induced epirubicin resistance. Further biochemical assays identify FLNC as a new target protein of the MAGED2-TRIM28 complex, which functions as a downstream effector of MAGED2-induced chemoresistance. The MAGED2-TRIM28 complex interacts with FLNC, resulting in its ubiquitination and subsequent proteasomal degradation. Loss of FLNC activates EMT, a well-established driver of chemoresistance. Notably, targeting TRIM28 via AAV-mediated gene therapy reverses MAGED2-induced epirubicin resistance in vivo, offering a promising therapeutic approach to overcome chemoresistance in TNBC.Graphical abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s12964-025-02529-w. Introduction Triple negative breast cancer (TNBC) is an aggressive subtype of breast cancer, characterized by the absence of estrogen receptor, progesterone receptor, and Her-2 expression. Due to the lack of targeted therapies, chemotherapy remains the primary treatment option [1]. However, fewer than 40% of TNBC patients achieve a complete pathological response to neoadjuvant chemotherapy [2], reflecting a high degree of intrinsic resistance and leading to poor survival outcomes. This limited response rate underscores the critical need to dissect the molecular basis of chemoresistance and develop alternative therapeutic strategies. The melanoma-associated antigen (MAGE) family has emerged as oncogenic drivers in various malignancies [3]. MAGE proteins contribute to tumorigenesis by forming E3 ubiquitin ligase complexes with RING domain proteins [4]. For example, MAGEA3/A6-TRIM28 complexes promote tumorigenesis through the ubiquitination and degradation of AMPK [5]. Similarly, MAGEA3/C2-TRIM28 complexes facilitate metabolic reprogramming by targeting FBP1 in hepatocellular carcinoma [6]. Despite these advances in understanding MAGE-mediated oncogenesis, the role of MAGE family members in chemotherapy resistance, particularly in TNBC, remains poorly characterized. Tripartite motif-containing 28 (TRIM28) is a pivotal regulator of epithelial-mesenchymal transition (EMT). Previous studies have demonstrated that TRIM28 regulates EMT-responsive gene expression at the transcriptional level [7, 8]. However, TRIM28 contains a TRIM structure in which the RING domain endows it with the ability to function as an E3 ubiquitin ligase [9]. While ubiquitination is an essential component of post-translational modification, it remains unclear whether TRIM28 regulates EMT at the post-translational level. Furthermore, MAGE proteins reportedly bind to E3 RING ubiquitin ligases, but with specificity [4]. The mechanism by which individual MAGE proteins regulate the E3 ubiquitin ligase activity of TRIM28 is not fully understood. In this study, we integrate proteomic and transcriptomic datasets from TNBC patients treated with neoadjuvant chemotherapy and identify MAGED2 as a key molecule associated with poor treatment response. Functional experiments reveal that MAGED2 confers resistance to epirubicin, a first-line TNBC chemotherapeutic agent, both in vitro and in vivo. Mechanistically, MAGED2 interacts with TRIM28 to form an E3 ubiquitin ligase complex that ubiquitinates filamin C (FLNC), targeting it for proteasomal degradation. Loss of FLNC triggers epithelial-mesenchymal transition (EMT), a well-established driver of chemoresistance. Notably, adeno-associated virus (AAV)-mediated targeting of TRIM28 effectively reverses MAGED2-induced epirubicin resistance in vivo, suggesting a promising therapeutic strategy against chemoresistance in TNBC. Method Cell lines and cell culture MDA-MB-231 (RRID: CVCL_0320), MDA-MB-468 (RRID: CVCL_0419), and MCF-10 A (RRID: CVCL_0598) cells were obtained from the American Type Culture Collection (ATCC). The embryonic kidney cell line 293 T (RRID: CVCL_0063) was obtained from the National Collection of Authenticated Cell Culture. All cell lines were cultured in DMEM (Procell, China) supplemented with 10% fetal bovine serum (Lonsera, China). Tissue samples of TNBC patients TNBC tumor samples and adjacent non-tumor tissue samples were collected from TNBC patients who underwent surgery at Zhongnan Hospital of Wuhan University. All samples were collected post-surgery. Ethical approval was obtained from the Medical Ethics Committee of Zhongnan Hospital of Wuhan University (approval number 2023002 K). Analysis of transcriptomic data from the HPA website Transcriptomic data from the GTEx dataset, HPA dataset, and consensus dataset were accessed from the Human Protein Atlas (HPA) website (https://www.proteinatlas.org/). Online analysis tools on the HPA website were used to compare mRNA expression levels. The log-rank test and Cox regression were applied to evaluate the prognostic impact of gene expression. P < 0.05 was considered statistically significant. Bioinformatic analysis of the proteomic dataset The proteomic dataset was downloaded from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) website (https://pdc.cancer.gov/pdc/browse). The unshared value was used for subsequent analysis. After normalization via log2 transformation, differentially expressed proteins between responders and non-responders were compared. Differential expression analysis was performed via the LIMMA method implemented in the R package ‘limma’ [10]. According to the referenced methodology, thresholds were set at log2 (fold change) >0.8 and P < 0.05 [11]. To ensure the abundance of proteins in each group, the top 20 proteins were selected based on the rank of their average expression levels. The protein-protein interaction network of the differentially expressed proteins was constructed via Cytoscape (https://cytoscape.org/). Bioinformatic analysis of GEO datasets Transcriptomic data from GSE25055 and GSE25065 were downloaded from the Gene Expression Omnibus (GEO) website (https://www.ncbi.nlm.nih.gov/geo/). The R package ‘Affy’ was used for preprocessing raw microarray data. Cox regression and log-rank tests were performed for survival analysis. MAGED2 mRNA expression levels were compared among the different groups. Univariate and multivariate logistic regression analyses were conducted to evaluate the association between MAGED2 expression and response to neoadjuvant chemotherapy. P < 0.05 was considered statistically significant. Bioinformatic analysis of cell line sensitivity to epirubicin Epirubicin (EPI) AUC data for all breast cancer cell lines were obtained from the Cancer Dependency Map (DepMap) portal (https://depmap.org/portal/). Triple-negative breast cancer cell lines were manually classified based on negative expression of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor (Her-2). Based on the AUC values of EPI in TNBC cell lines, the sensitivity of these cell lines to EPI was ranked. Spearman correlation analysis was conducted to assess the correlation between MAGED2 expression levels and EPI sensitivity in breast cancer cell lines. RNA-seq and analysis RNA-seq was performed on the control group and MAGED2 knockdown groups. Total RNA was extracted and sequenced by Wuhan Igenebook Biotechnology Co., Ltd. STAR was used for alignment and mapping. RSEM was used for quantification of gene expression levels. Gene set enrichment analysis (GSEA) was used to analyze enriched biological pathways. The RNA-seq data generated in this study was submitted to the China National Center for Bioinformation with project number PRJCA045992. Protein extraction and Western blotting Cells were washed with PBS twice before harvesting. SDS buffer, containing 62.5% Tris–HCl (pH 6.8), 2% SDS, 10% glycerol, 0.1% bromophenol blue, and 50 mM DTT, was used for cell lysis. The cell lysate was then subjected to SDS-PAGE and transferred to 0.22 μm PVDF membranes (Millipore Corporation). The PVDF membranes were incubated with different primary antibodies, the details of which were provided in Supplementary Table 1. The secondary antibody, conjugated with HRP, was visualized using the Tanon system. The intensity of the Western blotting signals was normalized to the intensity of internal control proteins, including GAPDH, β-actin, and vinculin. Details of the primary antibodies used for the internal control proteins were also provided in Supplementary Table 1. Immunoprecipitation (IP) Cells were washed twice with PBS prior to harvesting. IP lysis buffer (catalog #P0013, Beyotime, Inc.) was used for cell lysis. Following cell lysis, the supernatant was collected and incubated with antibody at 4 °C for 2 h. Magnetic beads (catalog #L1004, Biolinkedin) were then added to the mixture and incubated at 4 °C for an additional 2 h. The magnetic beads linked to the immune complexes were subsequently washed with PBST buffer and then boiled in 2x loading buffer (10 mM Tris–HCl pH 6.8, 1% SDS, 25% glycerol, 0.1 mM β-mercaptoethanol, 0.003% bromophenol blue). The boiled samples were then subjected to Western blotting as described above. Immunofluorescence of cell lines MDA-MB-231 and MDA-MB-468 cells were seeded in 6-well plates. Paraformaldehyde was used to fix the cells at 37 °C for 15 min. Cell slides were washed with PBS for 5 min. Triton X-100 (1%) was used to permeabilize the cells for 1 h. Cell slides were then washed with PBS. Goat serum was used to block the slides at room temperature for 30 min. Cell slides were incubated with primary antibodies (anti-MAGED2 and anti-TRIM28) at 4 °C overnight, followed by incubation with anti-rabbit and anti-mouse Alexa Fluor-conjugated secondary antibodies. DAPI was then added to each cell slide, which was subsequently incubated at room temperature for 10 min and washed with PBS in the dark. Immunofluorescence images of cell slides were captured using Olympus CellSens software. Immunohistochemistry staining Paraffin sections were placed in a 60 °C oven for 120 min. Xylene and ethanol were used for dewaxing and hydration. Sections were washed with distilled water and incubated with 3% H₂O₂ for 30 min at room temperature. Sections were then washed with distilled water. Antigen retrieval was performed using microwave heating. Sections were washed with PBS. Primary antibody was applied to the sections. Sections were incubated with the corresponding primary antibody overnight at 4 °C. After washing with PBS, secondary antibody was added. Sections were incubated at 37 °C for 60 min. PBS was used to wash the sections. DAB was used for development, followed by thorough rinsing with tap water. Hematoxylin was used to counterstain the sections. Sections were dehydrated and cleared for 10 min each. Neutral resin was used for mounting, and the sections were allowed to dry naturally. A bright-field microscope was used to capture images of the IHC staining. ImageJ software (v1.53) was used to analyze the positive area and mean optical density (OD) value of the IHC images. In vitro epirubicin sensitivity assay Transfected cells were seeded into 96-well plates and cultured for 24 h before EPI treatment. After 24 h, the culture medium was replaced with medium containing EPI. Cells were further cultured for 48 h. The medium was then discarded and replaced with fresh medium containing CCK-8 reagent at a ratio of 1:10. After incubation for 30 min, the absorbance of each well was measured at 450 nm. Colony formation assay Transfected cells were seeded into 6-well plates at a density of 1000 cells per well. The plates were cultured in medium containing EPI for 14 days. Afterwards, the cells were washed twice with PBS, fixed with methanol, and stained with 0.1% crystal violet solution for 30 min. Migration and invasion assays Transwell chambers (Labgic Technology Co., China) were used for both migration and invasion assays. For invasion assays specifically, chambers were pre-coated with Matrigel (Yeasen BioTechnologies Co., China) 24 h prior to cell seeding. Cells were starved in serum-free DMEM medium before seeding. Cells were then seeded at a density of 1 × 10⁵ cells per well. The lower chamber contained DMEM medium supplemented with 10% FBS as a chemoattractant. After 36 h of incubation, non-migrated/non-invaded cells were removed from the upper chamber. Migrated/invaded cells were fixed with 4% paraformaldehyde, washed with PBS, and stained with either Giemsa solution (Beyotime, Inc.) or 0.1% crystal violet (Biosharp). Following PBS washes, cells were imaged using an inverted microscope (Olympus) and quantified. GraphPad Prism 8 was used to perform t-test analysis between different groups, with error bars representing mean ± standard deviation. A p-value < 0.05 was considered statistically significant. Plasmid construction and lentivirus generation Short hairpin RNA (shRNA) sequences targeting MAGED2, TRIM28, and FLNC were synthesized by Tsingke Biotechnology. The sequences of the cloning primers and shRNAs were shown in Supplementary Table 2. The sequences of sgRNA targeting FLNC and MAGED2 were shown in Supplementary Table 2. The shRNA sequences were inserted into the pLKO.1 vector using EcoRI and AgeI restriction enzymes and T4 ligase (Thermo Fisher Scientific). The sgRNA sequences were inserted into the lentiCRISPR v2 vector. Ectopic expression plasmids for MAGED2 and TRIM28 were constructed by cloning the coding sequences (CDSs) of the MAGED2 and TRIM28 genes into the pLVX vector containing a Flag tag. The CDSs of MAGED2 and TRIM28 were obtained from the NCBI website (https://www.ncbi.nlm.nih.gov/gene). Cell transfection and lentivirus generation were performed using Lipofectamine 2000 (Thermo Fisher Scientific) according to the manufacturer’s instructions. Recombinant AAV-shRNA virus construction Adeno-associated virus serotype 2 (AAV2) capsids were selected for in vivo experiments because of their breast tropism. TRIM28 shRNA sequences were designed and packaged into AAV2 viruses (Supplementary Table 2). The constructs were transfected into HEK-293T cells to generate AAV-shTRIM28 viruses. The homologous AAV2 empty virus (AAV-shNC) was used as a blank control. RNA extraction RNA was extracted from cells using the New Cell Tissue Total RNA Kit (New Cell & Molecular Biotech, China) according to the manufacturer’s instructions. The purity of RNA was assessed using a NanoDrop spectrophotometer (Thermo Fisher Scientific). Animal models Female BALB/c-nude mice (5 weeks old) were purchased from GemPharmatech. All animal experiments were conducted in accordance with the guidelines of the Institutional Animal Care and Use Committee. This study was approved by the Laboratory Animal Management and Use Committee of Wuhan University (approval number WQ20210094). A total of 106 cells were subcutaneously injected into the nude mice. The tumor volume was measured twice a week. The volume was calculated using the formula: volume (mm3) = width2×length/2. EPI treatment (7 mg/kg) was initiated when the subcutaneous tumor became palpable. EPI was purchased from MCE company. When the mice were sacrificed, a portion of the xenograft tumor was fixed for staining, and another portion was processed for Western blotting. Statistical methods All statistical graphs were all generated using GraphPad Prism 8, with error bars representing mean ± standard deviation. In vitro experiments were performed in three replicates. For in vivo experiments, each group contained 5–7 mice. Comparisons between two independent groups were performed using a t-test. For comparisons involving three or more groups, ANOVA was used. A p- value < 0.05 was considered statistically significant. Data availability Proteomic data was downloaded from the CPTAC website. Transcriptomic data was downloaded from the GEO website, the HPA website, and the GEPIA website. Results Proteomic and transcriptomic analysis identifies MAGED2 as a key regulator of chemoresistance To identify molecules associated with chemoresistance in TNBC, we analyzed proteomic and transcriptomic datasets of TNBC patients treated with neoadjuvant chemotherapy [12] (Fig. 1A). Differential expression analysis of proteins was performed in patients with distinct responses to neoadjuvant chemotherapy (Fig. 1B). The top 20 proteins associated with chemoresistance were identified (Table 1). Subsequent analysis of protein-protein interactions demonstrated that MAGED2 was a key regulator of chemoresistance (Fig. 1C). The analysis of transcriptomic datasets (GSE25055, GSE25065) validated the association of MAGED2 expression with chemoresistance in TNBC. MAGED2 expression was significantly upregulated in TNBC patients resistant to neoadjuvant chemotherapy (Fig. 1D and E). High MAGED2 expression was significantly correlated with residual tumors after neoadjuvant chemotherapy (Fig. 1F, Supplementary Tables 3–4). Compared to stage II, MAGED2 expression was significantly upregulated in stage III (Fig. 1G). Moreover, compared to MCF-10 A, MAGED2 expression was significantly elevated in MDA-MB-231, MDA-MB-468, and MDA-MB-453 cell lines, which are negative for estrogen receptor, progesterone receptor, and Her-2 expression (Fig. 1H). Immunohistochemistry (IHC) staining of TNBC tumor and adjacent non-tumor tissues demonstrated that MAGED2 was significantly upregulated in TNBC compared to the adjacent normal tissue (Fig. 1I). Together, these results suggested that MAGED2 might promote chemoresistance in TNBC. Fig. 1Proteomic and transcriptomic analysis identifies MAGED2 as a key regulator of chemoresistance. A Schematic diagrams show the procedures for identifying potential molecules associated with chemoresistance in TNBC. B Volcano plot illustrating the differentially expressed proteins in TNBC patients with distinct responses to neoadjuvant chemotherapy. C The protein interaction map of MAGED2 is generated by Cytoscape website. D MAGED2 expression levels in patients sensitive and resistant to neoadjuvant chemotherapy based on the GSE25055 dataset. E MAGED2 expression levels in patients sensitive and resistant to neoadjuvant chemotherapy based on the GSE25065 dataset. F The forest plot shows the association between MAGED2 expression and treatment outcomes of breast cancer patients receiving neoadjuvant chemotherapy. G MAGED2 expression levels in patients with different AJCC stages based on the GSE25055 dataset. H MAGED2 expression levels in normal breast epithelial cells (MCF-10A) and TNBC cells (MDA-MB-231, MDA-MB-468, and MDA-MB-453). I Representative immunohistochemistry (IHC) images and quantification for MAGED2 expression showing that MAGED2 is significantly upregulated in TNBC tissues. J The EPI sensitivity map illustrates the EPI sensitivity of TNBC cell lines and breast cancer cell lines. K The correlation analysis demonstrates the correlation between MAGED2 expression and EPI sensitivity of breast cancer cell lines Proteomic and transcriptomic analysis identifies MAGED2 as a key regulator of chemoresistance. A Schematic diagrams show the procedures for identifying potential molecules associated with chemoresistance in TNBC. B Volcano plot illustrating the differentially expressed proteins in TNBC patients with distinct responses to neoadjuvant chemotherapy. C The protein interaction map of MAGED2 is generated by Cytoscape website. D MAGED2 expression levels in patients sensitive and resistant to neoadjuvant chemotherapy based on the GSE25055 dataset. E MAGED2 expression levels in patients sensitive and resistant to neoadjuvant chemotherapy based on the GSE25065 dataset. F The forest plot shows the association between MAGED2 expression and treatment outcomes of breast cancer patients receiving neoadjuvant chemotherapy. G MAGED2 expression levels in patients with different AJCC stages based on the GSE25055 dataset. H MAGED2 expression levels in normal breast epithelial cells (MCF-10A) and TNBC cells (MDA-MB-231, MDA-MB-468, and MDA-MB-453). I Representative immunohistochemistry (IHC) images and quantification for MAGED2 expression showing that MAGED2 is significantly upregulated in TNBC tissues. J The EPI sensitivity map illustrates the EPI sensitivity of TNBC cell lines and breast cancer cell lines. K The correlation analysis demonstrates the correlation between MAGED2 expression and EPI sensitivity of breast cancer cell lines Table 1Differentially expressed proteins of TNBC patients resistant to neoadjuvant chemotherapySpot NumberEntry nameProtein nameLog2FCP Value1COL9A1Collagen alpha-1 chain3.4195080.0091511862HAPLN1Hyaluronan and proteoglycan line protein 12.2109370.0239729683KRT6AKeratin, type II cytoskeletal 6 A2.2043580.0419288624AKR1B10Aldo-ketoreductase family 1 member B101.989290.0318355955SERPINE2Glia-derived nexin1.900710.0421273416LGALS7Galectin-71.8528780.0309388757ALG6Dolichyl pyrophosphate Man9GlcNAc2 alpha-1,3-glucosyltransferase1.3022080.0491862268KRT17Keratin, type I cytoskeletal 171.1817470.0494065929AIF1LAllograft inflammatory factor 1-like1.1603980.04351447710CTGFCCN family member 21.1381010.04939390611HES1Transcription factor HES-11.1188090.04895723412ARHGAP39Rho GTPase-activating protein 391.0426910.04609496913MYEF2Myelin expression factor 21.0039610.0432813114ISYNA1Inositol-3-phosphate synthase10.9873620.04010651115PFN2Profilin-20.98550.04145841316PVRPoliovirus receptor homolog0.9454610.04911590917PCSK1NProSAAS0.8801250.04749322818S100A16Protein S100-A160.8353150.04329931319MAGED2Melanoma-associated antigen D20.8124610.04106138520ISCA1Iron-sulfur cluster assembly 1 homolog0.8090280.039424892 Table 1Differentially expressed proteins of TNBC patients resistant to neoadjuvant chemotherapySpot NumberEntry nameProtein nameLog2FCP Value1COL9A1Collagen alpha-1 chain3.4195080.0091511862HAPLN1Hyaluronan and proteoglycan line protein 12.2109370.0239729683KRT6AKeratin, type II cytoskeletal 6 A2.2043580.0419288624AKR1B10Aldo-ketoreductase family 1 member B101.989290.0318355955SERPINE2Glia-derived nexin1.900710.0421273416LGALS7Galectin-71.8528780.0309388757ALG6Dolichyl pyrophosphate Man9GlcNAc2 alpha-1,3-glucosyltransferase1.3022080.0491862268KRT17Keratin, type I cytoskeletal 171.1817470.0494065929AIF1LAllograft inflammatory factor 1-like1.1603980.04351447710CTGFCCN family member 21.1381010.04939390611HES1Transcription factor HES-11.1188090.04895723412ARHGAP39Rho GTPase-activating protein 391.0426910.04609496913MYEF2Myelin expression factor 21.0039610.0432813114ISYNA1Inositol-3-phosphate synthase10.9873620.04010651115PFN2Profilin-20.98550.04145841316PVRPoliovirus receptor homolog0.9454610.04911590917PCSK1NProSAAS0.8801250.04749322818S100A16Protein S100-A160.8353150.04329931319MAGED2Melanoma-associated antigen D20.8124610.04106138520ISCA1Iron-sulfur cluster assembly 1 homolog0.8090280.039424892 Differentially expressed proteins of TNBC patients resistant to neoadjuvant chemotherapy Anthracyclines are administered in the first four cycles of neoadjuvant chemotherapy [13] and demonstrate cross-resistance with other chemotherapeutic drugs [14]. If patients develop resistance to anthracyclines, they may be resistant to subsequent chemotherapeutic drugs, leading to reduced objective response rates in neoadjuvant chemotherapy. Therefore, we primarily focused on the resistance to anthracyclines. Epirubicin (EPI) is a commonly used anthracycline agent in clinical practice. Subsequently, we compared EPI sensitivity among different breast cancer cell lines. Our results revealed that different breast cancer cell lines exhibited heterogeneous EPI sensitivity, with MDA-MB-231 and MDA-MB-468 cells being in the middle (Fig. 1J). We further explored the correlation between MAGED2 RNA expression and EPI sensitivity in breast cancer cell lines. MAGED2 RNA expression was positively correlated with EPI resistance in breast cancer cell lines, although this correlation did not reach statistical significance (Fig. 1K). This might be attributed to the limited number of cell lines analyzed in DepMap dataset (n = 27). These findings suggested that MAGED2 might promote EPI resistance, but further validation was required. MAGED2 promotes EPI resistance in vitro and in vivo through upregulation of EMT To investigate the role of MAGED2 in EPI resistance, we designed shRNA sequences targeting MAGED2 (Supplementary Table 2) and constructed stable MDA-MB-231 and MDA-MB-468 cell lines with MAGED2 knockdown (Fig. 2A). The knockdown of MAGED2 significantly impaired the viability of MDA-MB-231 and MDA-MB-468 cells at different EPI concentrations (Fig. 2B, Supplementary Fig. 1A–1B). Conversely, MAGED2 overexpression increased cell viability at different EPI concentrations (Fig. 2C and D, Supplementary Fig. 1C-1D). Compared to the empty vector group, ectopic expression of MAGED2 significantly increased the number and mean size of clones in MDA-MB-231 and MDA-MB-468 cells at each EPI concentration (Fig. 2E and F), indicating enhanced EPI resistance. Fig. 2MAGED2 promotes EPI resistance in vitro and in vivo through upregulation of EMT. A Immunoblotting analysis of whole-well lysates (WCLs) derived from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLKO.1, MAGED2 shRNA#1, and MAGED2 shRNA#2. B The EPI sensitivity of MDA-MB-231 cells and MDA-MB-468 cells with MAGED2 knockdown is determined by CCK-8 assays. C-D The EPI sensitivity of MDA-MB-231 cells and MDA-MB-468 cells stably expressing the empty vector and MAGED2-Flag is determined by CCK-8 assays. E-F Ectopic expression of MAGED2 significantly enhances the resistance of MDA-MB-231 cells and MDA-MB-468 cells to EPI treatment, as assessed by colony formation assays. G Gross images of subcutaneous tumors in the control group, MAGED2-overexpressing group, and MAGED2 knockout group. H Graphs showing statistical analysis of subcutaneous tumor volume in the control group, MAGED2-overexpressing group, and MAGED2 knockout group. I Graphs showing statistical analysis of subcutaneous tumor weight in the control group, MAGED2-overexpressing group, and MAGED2 knockout group. J Graphs showing statistical analysis of subcutaneous tumor volume in the control group and MAGED2-overexpressing group treated with EPI. K Gross images of subcutaneous tumors in the control group and MAGED2-overexpressing group treated with EPI. L Graphs showing statistical analysis of subcutaneous tumor weight in the control group and MAGED2-overexpressing group treated with EPI. M Representative images from hematoxylin and eosin (H&E) staining of subcutaneous tumors in the control group and MAGED2 stable expression group treated with EPI. N Enrichment analysis of gene expression is performed via hallmark gene sets from the MSigDB database. The top 20 enriched pathways are selected. O GSEA shows that knockdown of MAGED2 downregulates EMT. Q The migration ability of MDA-MB-231 and MDA-MB-468 stably expressing pLVX-Flag and MAGED2-Flag is assessed using the transwell assay. R The invasion ability of MDA-MB-231 and MDA-MB-468 stably expressing pLVX-Flag and MAGED2-Flag is assessed using the transwell assay coated with a layer of Matrigel MAGED2 promotes EPI resistance in vitro and in vivo through upregulation of EMT. A Immunoblotting analysis of whole-well lysates (WCLs) derived from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLKO.1, MAGED2 shRNA#1, and MAGED2 shRNA#2. B The EPI sensitivity of MDA-MB-231 cells and MDA-MB-468 cells with MAGED2 knockdown is determined by CCK-8 assays. C-D The EPI sensitivity of MDA-MB-231 cells and MDA-MB-468 cells stably expressing the empty vector and MAGED2-Flag is determined by CCK-8 assays. E-F Ectopic expression of MAGED2 significantly enhances the resistance of MDA-MB-231 cells and MDA-MB-468 cells to EPI treatment, as assessed by colony formation assays. G Gross images of subcutaneous tumors in the control group, MAGED2-overexpressing group, and MAGED2 knockout group. H Graphs showing statistical analysis of subcutaneous tumor volume in the control group, MAGED2-overexpressing group, and MAGED2 knockout group. I Graphs showing statistical analysis of subcutaneous tumor weight in the control group, MAGED2-overexpressing group, and MAGED2 knockout group. J Graphs showing statistical analysis of subcutaneous tumor volume in the control group and MAGED2-overexpressing group treated with EPI. K Gross images of subcutaneous tumors in the control group and MAGED2-overexpressing group treated with EPI. L Graphs showing statistical analysis of subcutaneous tumor weight in the control group and MAGED2-overexpressing group treated with EPI. M Representative images from hematoxylin and eosin (H&E) staining of subcutaneous tumors in the control group and MAGED2 stable expression group treated with EPI. N Enrichment analysis of gene expression is performed via hallmark gene sets from the MSigDB database. The top 20 enriched pathways are selected. O GSEA shows that knockdown of MAGED2 downregulates EMT. Q The migration ability of MDA-MB-231 and MDA-MB-468 stably expressing pLVX-Flag and MAGED2-Flag is assessed using the transwell assay. R The invasion ability of MDA-MB-231 and MDA-MB-468 stably expressing pLVX-Flag and MAGED2-Flag is assessed using the transwell assay coated with a layer of Matrigel To explore the role of MAGED2 in vivo, we constructed MDA-MB-231 cells stably expressing MAGED2 (MAGED2-Flag) and an empty vector (pLVX-Flag) as the negative control (Supplementary Fig. 1E). In addition, we constructed a stable cell line with MAGED2 knockout (Supplementary Fig. 1 F). We injected the control cells, MAGED2-overexpressing cells, and MAGED2 knockout cells subcutaneously into nude mice. Compared to the control group, MAGED2 overexpression significantly promoted tumor growth, increased tumor volume, and elevated tumor weight, while MAGED2 knockout significantly suppressed tumor growth, reduced tumor volume, and decreased tumor weight in vivo (Fig. 2G and I, Supplementary Fig. 1G-1 H). To assess MAGED2’s role under EPI treatment, we injected the control group cells and MAGED2-overexpressing MDA-MB-468 cells subcutaneously. When the subcutaneous tumors became palpable, the nude mice were treated with EPI once a week. Compared to the control group, ectopic expression of MAGED2 significantly promoted tumor growth and increased tumor volume and tumor weight during EPI treatment in vivo (Fig. 2J and L, Supplementary Fig. 1I). More tumor cell infiltration was observed in the MAGED2-overexpressing group (Fig. 2M). Together, these results demonstrated that MAGED2 promoted EPI resistance in vivo. Transcriptome sequencing was used to investigate the mechanisms underlying the effects of MAGED2 on TNBC. Differential gene expression analysis was conducted between cells with MAGED2 knockdown and those expressing the empty vector (pLKO.1). Furthermore, gene set enrichment analysis (GSEA) was performed to generate the top 20 pathways affected by the knockdown of MAGED2 (Fig. 2N). The knockdown of MAGED2 significantly downregulated epithelial-mesenchymal transition (EMT) (Fig. 2O). A previously reported MAGED2-associated process, the PI3K/Akt/mTOR pathway, was also included in the top 20 pathways we identified [15], confirming the validity of our pathway enrichment analysis. Since EMT is widely recognized for mediating drug resistance in various cancer types, the potential regulation of EMT by MAGED2 is of significance. To further validate the role of MAGED2 in EMT, we conducted migration and invasion assays comparing the control group and the MAGED2-overexpressing group in MDA-MB-231 and MDA-MB-468 cells. The results demonstrated that ectopic expression of MAGED2 significantly enhanced migration and invasion abilities of MDA-MB-231 and MDA-MB-468 cells (Fig. 2Q and R). Identifying TRIM28 as a MAGED2-interacting protein that promotes EMT To elucidate how MAGED2 regulats EMT, we identified MAGED2-interacting proteins (Fig. 3A). We performed immunoprecipitation (IP) using an anti-Flag antibody in lysates of MDA-MB-231 cells stably overexpressing MAGED2-Flag, with an empty vector (pLVX-Flag) as a negative control. The immuno-complex was analyzed by mass spectrometry (MS) to identify potential MAGED2-interacting proteins in cells (Fig. 3B, Supplementary Table 5). Since MAGED2 has a conserved MAGE homology domain, which was previously reported to bind E3 RING domain proteins [16], we integrated MS analysis results with the E3 ubiquitin ligase database [17] to identify overlapping E3 RING domain proteins (Fig. 3A). Notably, six E3 RING domain proteins were identified in the immuno-complex from cell lysates overexpressing MAGED2, suggesting that these RING domain proteins might be potential binding partners for MAGED2. We ranked these six E3 RING domain proteins based on the intensity ratio by comparing the MAGED2-overexpressing group to the control group, and found that TRIM28 had the highest intensity ratio (Fig. 3C). Thus, we hypothesized that TRIM28 was the E3 RING domain protein that interacted with MAGED2. Fig. 3Identification of TRIM28 as a MAGED2-interacting protein that promotes EMT. A Immunoprecipitation (IP) coupled with mass spectrometry analysis is used to identify potential MAGED2-interacting proteins. B Immunoblotting analysis of WCLs and anti-Flag IPs derived from MDA-MB-231 cells stably expressing MAGED2-Flag or pLVX-Flag as the negative control. C The intensity ratio of the E3 RING domain proteins between the MAGED2-overexpressing group and the control group. D Immunoblotting analysis of WCLs and anti-MAGED2 IPs derived from MDA-MB-231 cells. E Immunoblotting analysis of WCLs and anti-TRIM28 IPs derived from MDA-MB-231 cells. F Immunoblotting analysis of WCLs and anti-MAGED2 IPs derived from MDA-MB-468 cells. G Immunoblotting analysis of WCLs and anti-TRIM28 IPs derived from MDA-MB-468 cells. H Representative immunofluorescence images and colocalization analysis of MAGED2 (green), TRIM28 (red), and DAPI nuclear staining (blue) in MDA-MB-231 cells and MDA-MB-468 cells. I-J Immunoblotting analysis of WCLs and anti-Flag denatured IPs derived from the lysates of MDA-MB-231 cells and MDA-MB-468 cells stably expressing TRIM28 using the indicated antibodies. The cells are transfected with shMAGED2 or pLKO.1 as a negative control before IP performed via an anti-Flag antibody. K-L Immunoblotting analysis of MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLVX-Flag and TRIM28-Flag using the indicated antibodies Identification of TRIM28 as a MAGED2-interacting protein that promotes EMT. A Immunoprecipitation (IP) coupled with mass spectrometry analysis is used to identify potential MAGED2-interacting proteins. B Immunoblotting analysis of WCLs and anti-Flag IPs derived from MDA-MB-231 cells stably expressing MAGED2-Flag or pLVX-Flag as the negative control. C The intensity ratio of the E3 RING domain proteins between the MAGED2-overexpressing group and the control group. D Immunoblotting analysis of WCLs and anti-MAGED2 IPs derived from MDA-MB-231 cells. E Immunoblotting analysis of WCLs and anti-TRIM28 IPs derived from MDA-MB-231 cells. F Immunoblotting analysis of WCLs and anti-MAGED2 IPs derived from MDA-MB-468 cells. G Immunoblotting analysis of WCLs and anti-TRIM28 IPs derived from MDA-MB-468 cells. H Representative immunofluorescence images and colocalization analysis of MAGED2 (green), TRIM28 (red), and DAPI nuclear staining (blue) in MDA-MB-231 cells and MDA-MB-468 cells. I-J Immunoblotting analysis of WCLs and anti-Flag denatured IPs derived from the lysates of MDA-MB-231 cells and MDA-MB-468 cells stably expressing TRIM28 using the indicated antibodies. The cells are transfected with shMAGED2 or pLKO.1 as a negative control before IP performed via an anti-Flag antibody. K-L Immunoblotting analysis of MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLVX-Flag and TRIM28-Flag using the indicated antibodies To test our hypothesis, we performed endogenous co-immunoprecipitation (co-IP) using MAGED2 and TRIM28 antibodies in lysates of MDA-MB-231 and MDA-MB-468 cells. Our co-IP assays demonstrated that MAGED2 had a strong binding affinity for TRIM28. Conversely, TRIM28 was confirmed to interact with MAGED2. The interaction between MAGED2 and TRIM28 was validated in MDA-MB-231 and MDA-MB-468 cells (Fig. 3D and G). We then investigated whether MAGED2 interacted with other E3 ubiquitin ligases, such as TRIM25 or TRIM29, and found that MAGED2 did not interact with either TRIM25 or TRIM29 (Supplementary Fig. 1J–1 K). We further investigated whether MAGED2 colocalized with TRIM28 and performed immunofluorescence in MDA-MB-231 and MDA-MB-468 cells. Our results revealed that MAGED2 colocalized with TRIM28 (Fig. 3H). Since TRIM28 belongs to the TRIM family and contains the E3 RING domain [18], the interaction and the colocalization with MAGED2 suggested that MAGED2 interacted with TRIM28 to form an E3 ubiquitin ligase complex. We next investigated whether MAGED2 regulates the ubiquitin ligase activity of TRIM28. The results of endogenous ubiquitin-IP revealed that knockdown of MAGED2 suppressed the ubiquitin modification of TRIM28-interacting proteins in MDA-MB-231 and MDA-MB-468 cells (Fig. 3I and J). These results suggested that MAGED2 interacted with TRIM28 to form an E3 ligase complex and regulated the ubiquitin ligase activity of TRIM28. Additionally, ectopic expression of TRIM28 increased the expression of EMT markers in MDA-MB-231 and MDA-MB-468 cells (Fig. 3K and L, Supplementary Fig. 1L and 2 A-2B), suggesting that TRIM28 contributed to the upregulation of EMT. Interaction with TRIM28 promotes MAGED2-induced EPI resistance To explore the role of TRIM28 in MAGED2-induced EPI resistance, we designed and constructed shRNA sequences targeting the coding sequence of TRIM28 (Supplementary Table 2). Stable MDA-MB-231 and MDA-MB-468 cells with TRIM28 knockdown were then generated. TRIM28 knockdown inhibited EPI resistance in MDA-MB-231 and MDA-MB-468 cells (Fig. 4A and B, Supplementary Fig. 2C-2D). Conversely, ectopic TRIM28 expression increased cell viability at different EPI concentrations (Fig. 4C and D, Supplementary Fig. 2E–2 F). Moreover, ectopic expression of TRIM28 rescued the EPI resistance inhibited by knockdown of MAGED2 (Fig. 4E and F, Supplementary Fig. 2G-2 H), indicating TRIM28 contributes to MAGED2-induced EPI resistance. Fig. 4Interaction with TRIM28 promotes MAGED2-induced EPI resistance. A Immunoblotting analysis of WCLs derived from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLKO.1, TRIM28 shRNA#1, or TRIM28 shRNA#2. B The EPI sensitivity of MDA-MB-231 cells and MDA-MB-468 cells with TRIM28 knockdown is determined by CCK-8 assays. C-D The EPI sensitivity of MDA-MB-231 cells and MDA-MB-468 cells stably expressing the empty vector and TRIM28-Flag is determined by CCK-8 assays. E-F The EPI sensitivity of MDA-MB-231 cells and MDA-MB-468 cells transfected with the indicated constructs is determined by CCK-8 assays. G A schematic illustration of the MAGED2 protein sequence showing its different domains and deletion truncations. H Immunoblotting analysis of WCLs and anti-Flag IPs from MDA-MB-231 cells transfected with the indicated constructs. I Immunoblotting analysis of WCLs and anti-Flag IPs from MDA-MB-231 cells transfected with the indicated constructs. J The EPI sensitivity of MDA-MB-231 cells transfected with the indicated constructs is determined by CCK-8 assays Interaction with TRIM28 promotes MAGED2-induced EPI resistance. A Immunoblotting analysis of WCLs derived from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLKO.1, TRIM28 shRNA#1, or TRIM28 shRNA#2. B The EPI sensitivity of MDA-MB-231 cells and MDA-MB-468 cells with TRIM28 knockdown is determined by CCK-8 assays. C-D The EPI sensitivity of MDA-MB-231 cells and MDA-MB-468 cells stably expressing the empty vector and TRIM28-Flag is determined by CCK-8 assays. E-F The EPI sensitivity of MDA-MB-231 cells and MDA-MB-468 cells transfected with the indicated constructs is determined by CCK-8 assays. G A schematic illustration of the MAGED2 protein sequence showing its different domains and deletion truncations. H Immunoblotting analysis of WCLs and anti-Flag IPs from MDA-MB-231 cells transfected with the indicated constructs. I Immunoblotting analysis of WCLs and anti-Flag IPs from MDA-MB-231 cells transfected with the indicated constructs. J The EPI sensitivity of MDA-MB-231 cells transfected with the indicated constructs is determined by CCK-8 assays Since MAGED2 interacted with TRIM28, we hypothesized that the interaction between MAGED2 and TRIM28 contributed to EPI resistance. The functional domain of the MAGE family is the MAGE homology domain (MHD), which is composed of winged-helix A (WH-A) and WH-B motifs [4]. To determine which motif on MAGED2 was important for its interaction with TRIM28, we generated MAGED2 truncations lacking WH-A or WH-B (Fig. 4G). Our results demonstrated that the deletion of WH-A resulted in a lower binding affinity with TRIM28 compared to MAGED2-WT (Fig. 4H), suggesting that MAGED2 interacted with TRIM28 through the WH-A motif. To determine the residues of MAGED2 important for its interaction with TRIM28, we mutated all dileucine residues in the WH-A motif of MAGED2 and constructed MAGED2 L278A/L279A, L290A/L291A, and L348A/L349A plasmids. Previous studies have demonstrated that the dileucine motif is conserved across MAGE proteins. For example, MAGEG1 L96/L97 forms part of the hydrophobic core of the MAGEG1 WH-A motif and contributes to the NSE1 interaction [4]. Our results revealed that the MAGED2 L278A/L279A mutation reduced the binding affinity with TRIM28 compared to MAGED2-WT (Fig. 4I). Meanwhile, the L278/L279 mutation did not affect the stability of MAGED2 (Supplementary Fig. 2I-2 J). We further explored the role of the interaction between MAGED2 and TRIM28 in EPI resistance. Our results revealed that the knockdown of TRIM28 inhibited EPI resistance in cells with ectopic expression of MAGED2-WT, but not in MAGED2 L278A/L279A mutant (Fig. 4J, Supplementary Fig. 2K-2 L). Together, these results suggested that the interaction with TRIM28 contributed to MAGED2-induced EPI resistance. Identifying FLNC as a new target protein of the MAGED2-TRIM28 E3 ligase complex To identify the proteins targeted by the MAGED2-TRIM28 E3 ligase complex, we performed immunoprecipitation using an anti-Flag antibody in lysates of MDA-MB-231 cells stably expressing TRIM28-Flag or an empty vector as a negative control. The immunocomplex was analyzed by MS to identify potential TRIM28-interacting proteins in cells (Fig. 5A and B, Supplementary Fig. 3 A, Supplementary Table 6). Fig. 5Identification of FLNC as the target protein of the MAGED2-TRIM28 complex. A IP coupled with mass spectrometry analysis is used to identify potentially target proteins of the MAGED2-TRIM28 complex. B Immunoblotting analysis of WCLs and anti-Flag IPs derived from MDA-MB-231 cells stably expressing TRIM28-Flag or pLVX-Flag as the negative control. C-D Immunoblotting analysis of WCLs derived from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLKO.1, MAGED2 shRNA#1, or MAGED2 shRNA#2. E-F Immunoblotting analysis of WCLs from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLKO.1 and shMAGED2. The cells are treated with 200 µg/mL cycloheximide (CHX) for the indicated durations. The FLNC band intensity is quantified by ImageJ, normalized to that of vinculin, and compared to that at the t = 0 time point. G-H Immunoblotting analysis of WCLs derived from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLKO.1 and shTRIM28. The cells are treated with 200 µg/mL CHX for the indicated durations. The FLNC band intensity is quantified by ImageJ, normalized to that of vinculin, and compared to that at the t = 0 time point. I-J Immunoblotting analysis of WCLs derived from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLVX-Flag and MAGED2-Flag before and after MG132 treatment. The cells were treated with 10 µM MG132 for 10 h before being harvested. K-L Immunoblotting analysis of WCLs derived from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLVX-Flag and TRIM28-Flag before and after MG132 treatment. The cells are treated with 10 µM MG132 for 10 h before being harvested Identification of FLNC as the target protein of the MAGED2-TRIM28 complex. A IP coupled with mass spectrometry analysis is used to identify potentially target proteins of the MAGED2-TRIM28 complex. B Immunoblotting analysis of WCLs and anti-Flag IPs derived from MDA-MB-231 cells stably expressing TRIM28-Flag or pLVX-Flag as the negative control. C-D Immunoblotting analysis of WCLs derived from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLKO.1, MAGED2 shRNA#1, or MAGED2 shRNA#2. E-F Immunoblotting analysis of WCLs from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLKO.1 and shMAGED2. The cells are treated with 200 µg/mL cycloheximide (CHX) for the indicated durations. The FLNC band intensity is quantified by ImageJ, normalized to that of vinculin, and compared to that at the t = 0 time point. G-H Immunoblotting analysis of WCLs derived from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLKO.1 and shTRIM28. The cells are treated with 200 µg/mL CHX for the indicated durations. The FLNC band intensity is quantified by ImageJ, normalized to that of vinculin, and compared to that at the t = 0 time point. I-J Immunoblotting analysis of WCLs derived from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLVX-Flag and MAGED2-Flag before and after MG132 treatment. The cells were treated with 10 µM MG132 for 10 h before being harvested. K-L Immunoblotting analysis of WCLs derived from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLVX-Flag and TRIM28-Flag before and after MG132 treatment. The cells are treated with 10 µM MG132 for 10 h before being harvested By integrating MS results of MAGED2-interacting proteins and TRIM28-interacting proteins, we selected overlapping proteins as potential targets of the MAGED2-TRIM28 E3 complex. Seven overlapping proteins were identified (Supplementary Fig. 3B). To confirm which protein was targeted by the MAGED2-TRIM28 E3 complex for degradation, we examined protein abundance of these candidates in MDA-MB-231 and MDA-MB-468 cells with knockdown of MAGED2 or TRIM28. Notably, MAGED2 or TRIM28 knockdown dramatically increased FLNC protein levels but not ATN1/2 protein levels, which are encoded by SLC25A5 or SLC24A4 (Fig. 5C and D, Supplementary Fig. 3C–3 F). Thus, we hypothesized that FLNC was the targeted protein of the MAGED2-TRIM28 E3 complex, which had not been previously reported as a substrate for TRIM28. To validate whether FLNC was targeted by the MAGED2-TRIM28 complex, we utilized cycloheximide (CHX) to inhibit protein translation and analyzed the effect of the MAGED2-TRIM28 complex on the FLNC protein half-life. Knockdown of MAGED2 or TRIM28 significantly extended FLNC half-life in both cell lines (Fig. 5E and H, Supplementary Fig. 3G-3 J). These results demonstrated that the MAGED2-TRIM28 complex destabilized FLNC protein. MG132 was further used to inhibit proteasome function. Our results showed that overexpression of MAGED2 or TRIM28 significantly downregulated FLNC protein levels without MG132 treatment, whereas this effect was suppressed by MG132 treatment (Fig. 5I and L), indicating that FLNC degradation induced by the MAGED2-TRIM28 E3 complex was mediated through the proteasome. The MAGED2-TRIM28 complex interacts with FLNC to ubiquitinate FLNC Since the MAGED2-TRIM28 complex induced FLNC degradation, we hypothesized that it interacts with FLNC to promote ubiquitination and degradation. Consistent with this, our co-IP assay demonstrated that MAGED2 and TRIM28 exhibited strong binding affinities with FLNC in MDA-MB-231 cells (Fig. 6A and B). This interaction was confirmed at endogenous levels in MDA-MB-468 cells (Fig. 6C and D), with FLNC exhibiting high affinity for the MAGED2-TRIM28 complex in both cell lines (Fig. 6E, Supplementary Fig. 3K–3 L). Additionally, in vivo ubiquitination assays revealed that ectopic expression of MAGED2 or TRIM28 promoted FLNC ubiquitination (Fig. 6F and G). These results suggested that the MAGED2-TRIM28 complex interacted with FLNC to ubiquitinate it. Fig. 6The MAGED2-TRIM28 complex interacts with FLNC to ubiquitinate FLNC. A Immunoblotting analysis of WCLs and anti-MAGED2 IPs derived from MDA-MB-231 cells. B Immunoblotting analysis of WCLs and anti-Flag IPs derived from MDA-MB-231 cells stably expressing pLVX-Flag or TRIM28-Flag. C Immunoblotting analysis of WCLs and anti-MAGED2 IPs derived from MDA-MB-468 cells. D Immunoblotting analysis of WCLs and anti-TRIM28 IPs derived from MDA-MB-468 cells. E Immunoblotting analysis of WCLs and anti-FLNC IPs derived from MDA-MB-231 cells. F-G Immunoblotting analysis of WCLs and anti-FLNC IPs derived from MDA-MB-231 cells and MDA-MB-468 cells transfected with the indicated constructs. The cells are treated with 10 µM MG132 for 8 h before being harvested. H-I Immunoblotting analysis of WCLs derived from MDA-MB-231 cells and MDA-MB-468 cells transfected with the indicated constructs before and after MG132 treatment. J-K Immunoblotting analysis of WCLs derived from MDA-MB-231 cells transfected with the indicated constructs. The cells are treated with 200 µg/mL CHX for the indicated durations. The FLNC band intensity is quantified by ImageJ, normalized to that of vinculin, and compared to that at the t = 0 time point. L-M Immunoblotting analysis of WCLs derived from MDA-MB-468 cells transfected with the indicated constructs. The cells are treated with 200 µg/mL CHX for the indicated durations. The FLNC band intensity is quantified by ImageJ, normalized to that of vinculin, and compared to that at the t = 0 time point. N-O Immunoblotting analysis of WCLs derived from MDA-MB-231 and MDA-MB-468 cells transfected with the indicated constructs. The cells are treated with 200 µg/mL CHX for the indicated durations. The FLNC band intensity is quantified by ImageJ, normalized to that of vinculin, and compared to that at the t = 0 time point The MAGED2-TRIM28 complex interacts with FLNC to ubiquitinate FLNC. A Immunoblotting analysis of WCLs and anti-MAGED2 IPs derived from MDA-MB-231 cells. B Immunoblotting analysis of WCLs and anti-Flag IPs derived from MDA-MB-231 cells stably expressing pLVX-Flag or TRIM28-Flag. C Immunoblotting analysis of WCLs and anti-MAGED2 IPs derived from MDA-MB-468 cells. D Immunoblotting analysis of WCLs and anti-TRIM28 IPs derived from MDA-MB-468 cells. E Immunoblotting analysis of WCLs and anti-FLNC IPs derived from MDA-MB-231 cells. F-G Immunoblotting analysis of WCLs and anti-FLNC IPs derived from MDA-MB-231 cells and MDA-MB-468 cells transfected with the indicated constructs. The cells are treated with 10 µM MG132 for 8 h before being harvested. H-I Immunoblotting analysis of WCLs derived from MDA-MB-231 cells and MDA-MB-468 cells transfected with the indicated constructs before and after MG132 treatment. J-K Immunoblotting analysis of WCLs derived from MDA-MB-231 cells transfected with the indicated constructs. The cells are treated with 200 µg/mL CHX for the indicated durations. The FLNC band intensity is quantified by ImageJ, normalized to that of vinculin, and compared to that at the t = 0 time point. L-M Immunoblotting analysis of WCLs derived from MDA-MB-468 cells transfected with the indicated constructs. The cells are treated with 200 µg/mL CHX for the indicated durations. The FLNC band intensity is quantified by ImageJ, normalized to that of vinculin, and compared to that at the t = 0 time point. N-O Immunoblotting analysis of WCLs derived from MDA-MB-231 and MDA-MB-468 cells transfected with the indicated constructs. The cells are treated with 200 µg/mL CHX for the indicated durations. The FLNC band intensity is quantified by ImageJ, normalized to that of vinculin, and compared to that at the t = 0 time point Moreover, co-overexpression of MAGED2 and TRIM28 induced the highest FLNC ubiquitination (Fig. 6F and G), suggesting that MAGED2 enhances TRIM28-mediated FLNC ubiquitination. Consistent with this finding, MAGED2 overexpression also accelerated TRIM28-mediated FLNC degradation via the proteasome (Fig. 6H and I). Crucially, dual knockdown of MAGED2 and TRIM28 extended FLNC half-life more than TRIM28 knockdown alone (Fig. 6J and L, Supplementary Fig. 3M and 3O), confirming MAGED2 potentiates TRIM28 function. Further, TRIM28 overexpression shortened FLNC half-life versus controls (Supplementary Fig. 3Q-3R). While MAGED2 knockout extended FLNC half-life, TRIM28 overexpression reversed this effect (Fig. 6K and M, Supplementary Fig. 3 N and 3P). This suggested that FLNC degradation mediated by MAGED2 was dependent on TRIM28. Because two cysteine residues in the RING domain of TRIM28 (C65 and C68) are critical for its catalytic activity [19], we constructed catalytically inactive TRIM28 (TRIM28 C65A/C68A) to explore its effect on FLNC stability. Ectopic expression of catalytically inactive TRIM28 (TRIM28 C65A/C68A) failed to rescue the effect of MAGED2 knockdown on the half-life of FLNC (Fig. 6N and O, Supplementary Fig. 3S-3T), providing additional evidence for TRIM28-mediated FLNC ubiquitination and degradation. FLNC functions as a downstream effector of MAGED2-induced EPI resistance via EMT FLNC belongs to the filamin family, which plays a key role in maintaining cell shape and stabilizing the cytoskeleton. Filamin C is part of the mechanosensory complex, where changes in the actin cytoskeleton result in downstream effects on intermediate filaments [20, 21]. However, the role of FLNC in cancer has rarely been studied. We hypothesized that FLNC might function as a tumor suppressor to inhibit EPI resistance. Analysis of the TCGA BRCA dataset revealed that FLNC expression was significantly downregulated in breast cancer samples (Fig. 7A). High FLNC expression was associated with longer overall survival in breast cancer patients (Fig. 7B). Compared to MCF-10 A cells, FLNC expression was downregulated in MDA-MB-231 and MDA-MB-468 cells (Fig. 7C). Analysis of pan-cancer cell lines revealed that FLNC mRNA expression was upregulated in non-cancerous cell lines (Supplementary Fig. 4 A). Together, these results suggested that FLNC might function as a tumor suppressor. Fig. 7Loss of FLNC contributes to MAGED2-induced EPI resistance through activating EMT. A FLNC expression levels in normal tissues and breast cancer tissues. B The prognostic effect of FLNC expression on overall survival in the GSE25055 dataset. C FLNC expression levels in normal breast epithelial cells (MCF-10A) and TNBC cells (MDA-MB-231 and MDA-MB-468). D-E The EPI sensitivity of MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLKO.1, FLNC shRNA#1, or FLNC shRNA#2 is determined by CCK-8 assays. F-G The EPI sensitivity of MDA-MB-231 cells transfected with the indicated constructs is determined by CCK-8 assays. H-I The EPI sensitivity of MDA-MB-468 cells transfected with the indicated constructs is determined by CCK-8 assays. J-K The EPI sensitivity of MDA-MB-231 cells and MDA-MB-468 cells transfected with the indicated constructs is determined by CCK-8 assays. L-M Immunoblotting analysis of WCLs derived from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLKO.1, FLNC shRNA#1, or FLNC shRNA#2 Loss of FLNC contributes to MAGED2-induced EPI resistance through activating EMT. A FLNC expression levels in normal tissues and breast cancer tissues. B The prognostic effect of FLNC expression on overall survival in the GSE25055 dataset. C FLNC expression levels in normal breast epithelial cells (MCF-10A) and TNBC cells (MDA-MB-231 and MDA-MB-468). D-E The EPI sensitivity of MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLKO.1, FLNC shRNA#1, or FLNC shRNA#2 is determined by CCK-8 assays. F-G The EPI sensitivity of MDA-MB-231 cells transfected with the indicated constructs is determined by CCK-8 assays. H-I The EPI sensitivity of MDA-MB-468 cells transfected with the indicated constructs is determined by CCK-8 assays. J-K The EPI sensitivity of MDA-MB-231 cells and MDA-MB-468 cells transfected with the indicated constructs is determined by CCK-8 assays. L-M Immunoblotting analysis of WCLs derived from MDA-MB-231 cells and MDA-MB-468 cells stably expressing pLKO.1, FLNC shRNA#1, or FLNC shRNA#2 To investigate the role of FLNC in MAGED2-induced EPI resistance, we designed shRNAs targeting the coding sequence of FLNC (Supplementary Table 2, Supplementary Fig. 4B). Knockdown of FLNC promoted EPI resistance in MDA-MB-231 and MDA-MB-468 cells (Fig. 7D and E, Supplementary Fig. 4C–4 F). Furthermore, FLNC knockdown rescued the EPI resistance suppressed by MAGED2 knockdown (Fig. 7F and I). Since the MAGED2-TRIM28 complex induced FLNC degradation, we investigated whether the loss of FLNC rescued the EPI resistance suppressed by dual knockdown of MAGED2 and TRIM28. FLNC knockdown enhanced EPI resistance inhibited by MAGED2 and TRIM28 knockdown in both cell lines (Fig. 7J and K, Supplementary Fig. 4G–4 H). Together, these results suggested that FLNC served as a downstream effector of the MAGED2-induced EPI resistance. Because FLNC helps stabilize cell shape, we hypothesized that the loss of FLNC might activate EMT, which was downregulated by MAGED2 knockdown. Consistent with this hypothesis, FLNC knockdown dramatically increased the expression of mesenchymal markers (Fig. 7L and M). Moreover, we established stable FLNC knockout cell lines and analyzed the expression of EMT markers. Compared to the control group, the levels of N-cad and Vimentin increased, while the levels of E-cad decreased, suggesting that FLNC knockout could induce EMT in MDA-MB-231 and MDA-MB-468 cells (Supplementary Fig. 4I-4 J). Targeting TRIM28 via AAV-mediated gene therapy reverses MAGED2-induced EPI resistance in vivo Since MAGED2-mediated FLNC degradation depended on TRIM28, we speculated that targeting TRIM28 might reverse MAGED2-induced EPI resistance in TNBC. We designed and constructed AAV shRNA sequences for targeting the coding sequence of TRIM28 (Supplementary Table 2). We packaged AAV viruses using TRIM28 AAV shRNA sequences (AAV-shTRIM28) and an empty vector (AAV-shNC) as a negative control. Next, we utilized a subcutaneous mouse tumor model to evaluate the efficacy of AAV-shTRIM28 gene therapy in vivo. The same number of MDA-MB-231 cells expressing the empty vector (pLVX) and overexpressing MAGED2 (MAGED2-Flag) were injected subcutaneously into nude mice. When the subcutaneous tumors became palpable, the AAV-shTRIM28 and AAV-shNC viruses were injected intratumorally. All the nude mice were treated with EPI once a week (Fig. 8A). Compared to the control group with AAV-shNC, AAV-shTRIM28 significantly inhibited tumor growth and reduced tumor volume during EPI treatment. Furthermore, treating the MAGED2-overexpressing group with AAV-shTRIM28 significantly reduced tumor volume and tumor weight, which had been increased by MAGED2 overexpression (Fig. 8B and C, Supplementary Fig. 5 A). Fig. 8Targeting TRIM28 via AAV-mediated gene therapy reverses MAGED2-induced EPI resistance in vivo. A A schematic treatment plan for targeting TRIM28 via AAV-mediated gene therapy in vivo. B The growth of subcutaneous MDA-MB-231 tumors stably expressing MAGED2 or pLVX as a negative control is analyzed under the indicated treatment conditions. C Statistical graphs showing the weight of subcutaneous MDA-MB-231 tumors stably expressing MAGED2 or pLVX treated with the indicated treatment. D Statistical graphs showing the body weights of nude mice bearing MDA-MB-231 tumors stably expressing MAGED2 or pLVX treated with the indicated treatments. E Representative images from IHC staining of subcutaneous MDA-MB-231 tumors stably expressing MAGED2 or pLVX treated with the indicated treatment. F-H Quantification for MAGED2 (F), TRIM28 (G), and Ki-67 (H) positive area were counted from mice tumor sections. n = 6 mice per group. Data are presented as mean ± standard deviation. I Working model for MAGED2-induced EPI resistance in TNBC Targeting TRIM28 via AAV-mediated gene therapy reverses MAGED2-induced EPI resistance in vivo. A A schematic treatment plan for targeting TRIM28 via AAV-mediated gene therapy in vivo. B The growth of subcutaneous MDA-MB-231 tumors stably expressing MAGED2 or pLVX as a negative control is analyzed under the indicated treatment conditions. C Statistical graphs showing the weight of subcutaneous MDA-MB-231 tumors stably expressing MAGED2 or pLVX treated with the indicated treatment. D Statistical graphs showing the body weights of nude mice bearing MDA-MB-231 tumors stably expressing MAGED2 or pLVX treated with the indicated treatments. E Representative images from IHC staining of subcutaneous MDA-MB-231 tumors stably expressing MAGED2 or pLVX treated with the indicated treatment. F-H Quantification for MAGED2 (F), TRIM28 (G), and Ki-67 (H) positive area were counted from mice tumor sections. n = 6 mice per group. Data are presented as mean ± standard deviation. I Working model for MAGED2-induced EPI resistance in TNBC The TRIM28 expression of normal tissues from the control and AAV-shTRIM28 groups was analyzed. No significant difference was detected in TRIM28 expression between normal tissues from the control and AAV-shTRIM28 treatment groups (Supplementary Fig. 5B-5 F). This finding suggested that injecting AAV-shTRIM28 into subcutaneous tumors did not affect TRIM28 function in normal tissues, potentially preventing the side effects of this therapy. Moreover, mouse weight did not change significantly during treatment, suggesting minimal toxicity of AAV-mediated gene therapy (Fig. 8D). To assess tumor progression, we conducted immunostaining for MAGED2, TRIM28, and Ki-67 across four experimental groups: the control group treated with AAV-shNC, the control group treated with AAV-shTRIM28, the MAGED2-overexpressing group treated with AAV-shNC, and the MAGED2-overexpressing group treated with AAV-shTRIM28. Compared to the control group with AAV-shNC, AAV-shTRIM28 significantly reduced TRIM28 expression (Fig. 8E and G). Additionally, administration of AAV-shTRIM28 to the MAGED2-overexpressing group led to a significant decrease in Ki-67 expression, which was promoted by MAGED2 overexpression (Fig. 8F and H). These findings demonstrated that targeting TRIM28 via AAV-mediated gene therapy could serve as a potential therapeutic strategy to counteract EPI resistance promoted by MAGED2. Discussion Neoadjuvant chemotherapy is a standard treatment for TNBC [22]. However, fewer than 40% of TNBC patients achieve a complete response to neoadjuvant chemotherapy. The limited objective response rate necessitates further investigation into the mechanisms underlying chemoresistance and the development of novel therapeutic strategies. In this study, we demonstrate that MAGED2 promotes chemoresistance by forming an E3 ubiquitin ligase complex with TRIM28, which targets FLNC for proteasomal degradation. Loss of FLNC activates EMT, a well-established driving force of chemoresistance (Fig. 8I). Our findings provide new insights into the development of innovative strategies to overcome chemoresistance in TNBC. MAGED2 belongs to the MAGE family. MAGE proteins have been identified as oncogenic drivers in various cancers [23]. The oncogenic functions of the MAGE-I subclass, including MAGE-A, MAGE-B, and MAGE-C, have been extensively studied [24–27]. However, research on the MAGE-II subclass, including MAGE-D, MAGE-E, MAGE-F, MAGE-G, and necdin, remains limited. In this study, by analyzing proteomic and transcriptomic datasets of TNBC, we found that MAGED2 expression was upregulated in TNBC patients resistant to neoadjuvant chemotherapy. Functional assays validated the role of MAGED2 in promoting EPI resistance both in vitro and in vivo. Pathway enrichment analysis demonstrated that the knockdown of MAGED2 downregulated EMT. To further investigate how MAGED2 regulated EMT, we identified MAGED2-interacting proteins and confirmed that MAGED2 formed an E3 ligase complex with TRIM28. Subsequent mass spectrometry analysis identified FLNC as a new target protein of the MAGED2-TRIM28 complex. By binding to TRIM28, MAGED2 enhanced the ubiquitination and degradation of FLNC via the proteasome. Loss of FLNC activated the EMT process, contributing to MAGED2-induced EPI resistance. Overall, our study expands the role of the MAGE-II subclass in chemoresistance, reveals its underlying regulatory mechanism, and identifies a novel target protein of the E3 ubiquitin ligase complex. Tumor cells adapt to cancer therapies by undergoing transitions in cell states, fostering drug resistance [28]. The epithelial to mesenchymal transition serves as a proxy for phenotypic plasticity [29]. The association of EMT with therapy resistance has been observed in various cancer types [30]. Our study demonstrates that MAGED2 upregulates EMT by binding TRIM28 to form an E3 ligase complex. Previous studies have investigated the role of TRIM28 in regulating the EMT process, primarily at the transcriptional level. In non-small cell lung cancer, TRIM28 influences the acetylation and methylation of histones at the E-cadherin and N-cadherin promoters [7]. TRIM28 regulates the expression of a broad spectrum of EMT-responsive genes by occupying the FTS-1 site in chromatin [31]. However, the role of TRIM28 in regulating the EMT process at the post-translational level remains poorly understood. Here, our study reveals that the MAGED2-TRIM28 complex promotes EMT by ubiquitinating the FLNC protein, leading to its degradation via the proteasome. FLNC, a member of the filamin family, interacts with intermediate filaments, such as cytokeratin and vimentin. The filamin family and intermediate filaments contribute to the formation of the cellular skeleton. FLNC is a component of mechanosensory complexes that transmit signals from the actin cytoskeleton to intermediate filaments, where they induce downstream effects [20, 21]. In our study, loss of FLNC triggers EMT, suggesting that cellular skeleton instability promotes the transition of epithelial cells to a mesenchymal phenotype. However, the potential mechanism deserves further investigation. Moreover, TRIM28 plays distinct roles within the ubiquitin-proteasome system. In gastric cancer, TRIM28 stabilizes the PD-L1 protein by inhibiting PD-L1 ubiquitination and promoting PD-L1 SUMOylation, thereby blocking its proteasome-dependent degradation [32]. In prostate cancer, TRIM28 interacts with TRIM24 to prevent its ubiquitination and degradation by SPOP [33]. Moreover, the SETDB1-TRIM28 complex destabilizes PD-L1 protein and downregulates PD-L1 expression in ovarian cancer [34]. Our study reveals that MAGED2 forms an E3 ligase complex with TRIM28 and enhances the ubiquitin ligase activity of TRIM28. The MAGED2-TRIM28 complex facilitates FLNC ubiquitination and degradation via the proteasome. On the basis of these findings, we speculate that the distinct roles of TRIM28 in the ubiquitin-proteasome system are context-specific. However, the mechanism by which TRIM28 switches from promoting substrate degradation to stabilizing substrates remains unclear and warrants further investigation. This study has some potential limitations. The association between MAGED2 expression and chemotherapy resistance was primarily identified and validated in public datasets. Although we compared MAGED2 expression in TNBC tumor tissues and adjacent normal tissues using in-house patient sections, additional proteomic data from the in-house cohort would further verify the significance of MAGED2 in promoting chemoresistance in TNBC patients. In this study, we used EPI as the representative drug and revealed the underlying mechanism. However, the combination of different drugs in chemotherapy is common, which renders chemotherapy resistance a complex issue. The mechanism by which MAGED2 promotes EPI resistance is just one aspect of this complexity. The MAGED2-TRIM28-FLNC axis revealed in this study will be further tested for resistance to other drugs. In summary, our study identifies a novel MAGED2-TRIM28-FLNC axis that promotes chemoresistance by upregulating EMT. Findings from AAV-shTRIM28 gene therapy, which effectively reverses MAGED2-induced EPI resistance in vivo, suggest a novel therapeutic strategy to overcome chemotherapy resistance in TNBC. Supplementary Information Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2.