Mechanosensors Collaborate to Regulate DNA Damage Repair in Hepatocellular Carcinoma

Matrix stiffness enhances DNA repair via dual pathways in hepatocellular carcinoma, suggesting that targeting mechanosensory cross-talk could be a potential strategy to overcome cytotoxic drug resistance.

Abstract

Increased matrix stiffness is a biomechanical feature of hepatocellular carcinoma (HCC) that functions as an important promoting factor driving tumor progression. Characterization of the mechanosensors and the underlying downstream mechanisms driving HCC could help identify potential treatment strategies. In this study, we developed a standard procedure for screening and identifying mechanosensors, revealing two HCC mechanosensors, SDC4 and PKD2. These two mechanosensory pathways collaboratively mediated matrix stiffness–promoted DNA damage repair (DDR). Specifically, SDC4 and PKD2 modulated matrix stiffness–increased BANF1 nuclear translocation by increasing protein phosphatase 2A assembly and activity. Subsequently, nuclear BANF1 boosted the formation of the MRE11–RAD50–NBS1 complex and phosphorylation of its downstream effectors to enhance DDR. These findings not only expand the theoretical understanding of matrix stiffness–regulated DDR but also offer a screening/identification approach for mechanosensors in cancer.

Significance:

Matrix stiffness enhances DNA repair via dual pathways in hepatocellular carcinoma, suggesting that targeting mechanosensory cross-talk could be a potential strategy to overcome cytotoxic drug resistance.

Graphical Abstract

Introduction

Transcatheter arterial chemoembolization (TACE) and hepatic arterial infusion chemotherapy are key interventional therapies for advanced hepatocellular carcinoma (HCC), and oxaliplatin is their most commonly used chemotherapeutic agent (1). Oxaliplatin exerts its antitumor effects by disrupting DNA replication and transcription and inducing double-strand breaks (DSB; ref. 2). However, oxaliplatin resistance is still an unsolved challenge in the treatment of advanced HCC (3). The DNA damage repair (DDR) ability of tumor cells is believed to be a pivotal factor in determining the efficacy and sensitivity of cytotoxic chemotherapy. Previous research has extensively focused on the roles of tumor cells themselves in chemotherapeutic resistance, such as gene mutations (4), metabolic reprogramming (5), oxidative stress (6), abnormal cell cycle and apoptosis (7), and stemness (8). In recent years, the contribution of the tumor microenvironment to DDR has been highlighted, including hypoxia (9), lactate accumulation (10), cancer-associated fibroblasts (11), and myeloid-derived suppressor cells (12). However, the effects of biomechanical signals within the microenvironment on DDR and their underlying mechanisms remain largely unknown, and the lack of novel mechanosensor screening/identification standardized pathways and ideal stiffness-related experimental platforms are two important limiting factors.

Increased matrix stiffness, a significant biomechanical feature of HCC, has now been regarded as a critical promoting factor driving the progression of HCC (13). Applying the established stiffness-related experimental platforms, we previously validated that increased matrix stiffness remarkably strengthened the malignant characteristics of HCC cells and facilitated their invasion and metastasis (14). Simultaneously, integrin β1/α5 (14) and Piezo1 (15) as the mechanosensors were also clarified to sense stiffness signals and mediate the transformation of biomechanical signals into biochemical signals. In addition to the above two mechanosensors, little is known about whether there are any other novel mechanosensors in HCC. The currently known mechanosensors are mainly classified into adhesion molecules/receptors (integrins) and ion channel proteins [Piezo1 (15) and TRPV4 (16)]; they serve as a direct “bridge” connecting extracellular mechanical signals and intracellular biochemical responses. Other reported molecules, including FAK (17), Src (18), YAP/TAZ (19), and Lamin A (20), are only downstream mechanical signal-responsive molecules; strictly speaking, they cannot be defined as mechanosensors. Bead pulling/magnetic tweezers have recently been used to determine whether adhesion molecule/receptor mechanosensors can directly sense and transmit mechanical signals in endothelial cells (21). Cell stretching devices combined with a fluorescent Ca²⁺ indicator also provide the possibility for verifying whether ion channel proteins sense extracellular mechanical signals (22). These new direct validation methods will undoubtedly help accelerate the discovery of novel mechanosensors.

Although there exists a potential effect of increased matrix stiffness on cytotoxic chemotherapy resistance (23), direct mechanism analysis from the key perspective of DDR is still insufficient. Deng and colleagues (24) recently reported that matrix stiffness mediates the recruitment of RNF8 to RNF168, BRCA1, and 53BP1 to regulate DDR via MAPK4/6/7 kinase-mediated ubiquitin phosphorylation. However, it did not address which upstream molecules respond to matrix stiffness stimulation and whether a collaborative regulation between different mechanosensory pathways participates in this process. Considering that adhesion molecules/receptors and ion channel proteins represent two different mechanosensory pathways, we speculate that there may be a cross-talk between the two mechanosensory pathways in matrix stiffness–enhanced DNA repair in HCC.

Here, we developed a standard procedure for screening and identifying new mechanosensors in HCC and proposed a new regulatory mechanism by which the two mechanosensory pathways collaboratively mediate matrix stiffness–promoted DDR. These findings refine the current understanding of matrix stiffness–regulated DDR and also bring a new idea for the discovery of novel mechanosensors in HCC.

Materials and Methods

Cell and cell culture

Human HCC cell (MHCC97H, RRID: CVCL_4972) was obtained from the Liver Cancer Institute of Zhongshan Hospital, Fudan University. Human HCC cell (Hep3B, RRID: CVCL_0326) and mouse HCC cell (Hepa1–6, cat. #RRID:CVCL_0327) were purchased from the Cell Bank of the Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences. MHCC97H and Hepa1-6 cells were cultured in DMEM (GenomBio; cat. #GNM12800) supplemented with 10% FBS (BioSun; cat. #BS-0002) and 1% penicillin–streptomycin (GenomBio; cat. #GNM15140). Hep3B cells were maintained in minimum essential medium (GenomBio, cat. #GNM41500) with the same 10% FBS and 1% penicillin–streptomycin supplementation. The used cells are guaranteed to be mycoplasma-free, and their identities were authenticated by short tandem repeat analysis.

Lentivirus and plasmid transfection

Lentiviral vectors (Homo-OE-SDC4, Homo-OE-PKD2, Homo-shSDC4,Homo-shPKD2, Mouse-shSDC4, Mouse-shPKD2) were designed, constructed, and packaged by Genechem Co. Ltd. Hep3B cells and MHCC97H cells were infected with Homo-series vectors at a multiplicity of infection (MOI) of 10 and 20, respectively. Hepa1-6 cells were infected with a vector containing Mouse-shSDC4 or Mouse-shPKD2 at an MOI of 40.

Plasmids encoding human BANF1 variants [wild type (WT), BANF1-dephosphorylated plasmid (MAAAQ), BANF1-hyperphosphorylated plasmid (MDDDQ), T3A/K54A/E83A/D86A, E83A/D86A] were prepared by Genechem Co. Ltd. Plasmids were transfected into MHCC97H and Hep3B cells using Lipofectamine 3000 (Invitrogen, cat. #3000015) following the manufacturer’s protocol. All sequences for overexpression/knockdown fragments are available in Supplementary Table S1.

Collagen I–coated polyacrylamide gel substrates with stiffness of 6 and 16 kPa

Referring to the method reported previously (25), polyacrylamide (PA) gel substrates with stiffness of 6 kPa (low, mimicking the tissue stiffness of normal liver) and 16 kPa (high, mimicking the tissue stiffness of cirrhotic liver) were prepared by mixing 30% acrylamide (Sigma-Aldrich, cat. #A8887), 2% diacrylamide (Sigma-Aldrich, cat. #146072), 10% ammonium persulfate (Sigma-Aldrich, cat. #A3678), Tetramethylethylenediamine (Beyotime Biotechnology, cat. #ST728), and ddH₂O (detailed formulas in Supplementary Table S2). Gel substrates were treated with UV irradiation for 30 minutes and then functionalized with 0.8 mg/mL dopamine hydrochloride [in 50 mmol/L HEPES (pH 6.8–8.2), Servicebio; cat. #G4210]. After being washed with 1× PBS three times, gel substrates were further coated with 1 mL of 0.1 mg/mL collagen I (COL1; Corning; cat. #CLS354236). The obtained COL1-coated PA gel substrates were used for subsequent cell culture.

A three-dimensional hydrogel culture system with stiffness of 6 and 16 kPa

A three-dimensional (3D) hydrogel culture system with distinct stiffness was fabricated to mimic the tissue stiffness of normal liver (6 kPa, low stiffness) and the tissue stiffness of cirrhotic liver (16 kPa, high stiffness). Hydrogels were prepared by mixing 4% modified sodium alginate (Aladdin, cat. #S100126) and 100 mmol/L CaCl₂ (Aladdin, cat. #C110766) at optimized volumes. Briefly, 1 g of sodium alginate was chemically modified by stirring with 274 mg of 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (Aladdin; cat. #E106172) and 484 mg of N-hydroxysuccinimide (Aladdin, cat. #H109330) in 100 mL of 0.1 mol/L MES buffer (Beyotime Biotechnology; cat. #ST370). The activated alginate was then covalently cross-linked with 100 mg of GGGGRGDSP peptide overnight at room temperature. The resulting modified alginate was extensively dialyzed using a 3,500 Da MW cutoff dialysis membrane, sterile filtered, and lyophilized. The modified alginate powder was dissolved in DMEM to prepare a 4% (w/v) stock solution. The 4% modified alginate solution was gently mixed with HCC cell suspensions and a 100 mmol/L CaCl₂ solution. The mixture was allowed to undergo complete gelation at 37°C for 1 hour. After gelation, excess CaCl₂ was removed, and 1 mL of complete culture medium was added. Cells were cultured in the 3D hydrogel system for 14 days prior to subsequent experimental analyses.

Western blot

Total proteins of cells were extracted using RIPA lysis buffer (Beyotime Biotechnology, cat. #P0013B). Nuclear proteins of cells were isolated with NE-PER Nuclear and Cytoplasmic Extraction Reagent (Thermo Fisher Scientific, cat. #78835), and membrane proteins of cells were prepared using the Minute Plasma Membrane Protein Isolation Kit (Invent, cat. #SM-005). Protein concentration was quantified using a BCA protein assay kit. Equal amounts of protein samples were separated by sodium dodecyl sulfate-PA gel electrophoresis (SDS-PAGE) and then transferred onto polyvinylidene difluoride membranes. After blocking with 5% skim milk in TBS containing 0.1% Tween 20, the membranes were incubated with diluted primary antibodies overnight at 4°C, followed by incubation with horseradish peroxidase (HRP)–conjugated secondary antibodies for 1 hour at room temperature. Protein bands were visualized using an enhanced chemiluminescence detection kit (Yeasen, cat. #36222ES76) and the ChemiDOC XRS imaging system (Bio-Rad). Band intensity was semiquantitatively analyzed using ImageJ software (version 1.8.0). For normalization, GAPDH or α-tubulin was used as the internal control for total proteins, Lamin B1 for nuclear proteins, and Na⁺/K⁺–ATPase for membrane proteins. Relative protein expression levels were calculated by normalizing to the corresponding internal control and compared with the control group based on the experimental design. Diluted primary antibodies were as follows: SDC4 (Abcam, cat. #ab74139, RRID:AB_1524464, 1:1,000), PKD2 (Proteintech, cat. #19126-1-AP, RRID:AB_10641050, 1:1,000), GAPDH (Proteintech, cat. #60004-1-Ig, RRID:AB_2107436, 1:5,000), α-tubulin (Proteintech, cat. #80762-1-RR, RRID:AB_2918911, 1:5,000), Lamin B1 (Proteintech, cat. #12987-1-AP, RRID:AB_2136290, 1:1,000), BANF1 (Proteintech, cat. #30525-1-AP, RRID:AB_3086349, 1:1,000), Flag (Proteintech, cat. #66008-4-Ig, RRID:AB_2918475, 1:1,000), phospho-FAK (Tyr397; Cell Signaling Technology, cat. #3283, RRID:AB_2173659, 1:1,000), FAK (Cell Signaling Technology, cat. #3285, RRID:AB_2269034, 1:1,000), phospho-YAP (Ser127; Cell Signaling Technology, cat. #4911, RRID:AB_2218913, 1:1,000), YAP (Cell Signaling Technology, cat. #14074, RRID:AB_2650491, 1:1,000), cleaved caspase 3 (Cell Signaling Technology, cat. #9661, RRID:AB_2341188, 1:500), caspase 3 (Cell Signaling Technology, cat. #9662, RRID:AB_331439, 1:1,000), Piezo1 (Proteintech, cat. #28511-1-AP, RRID:AB_2881161, 1:1,000), ITGB1 (Cell Signaling Technology, cat. #34971, RRID:AB_2799067, 1:1,000), γ-H2AX (Ser139; Cell Signaling Technology, cat. #9718, RRID:AB_2118009, 1:1,000), RAD51 (Cell Signaling Technology, cat. #8875, RRID:AB_2721109, 1:1,000), phospho-ATM (Ser1981; ABclonal, cat. #AP1030, RRID:AB_2863913, 1:1,000), ATM (ABclonal, cat. #A19650, RRID:AB_2862715, 1:1,000), phospho-CHK2 (Tyr68; ABclonal, cat. #AP0590, RRID:AB_2771000, 1:1,000), CHK2 (ABclonal, cat. #A19543, RRID:AB_2862658, 1:1,000), pan-Phospho (ABclonal, cat. #AP1475, RRID:AB_3698468, 1:1,000), MRE11 (Abcam, cat. #ab214, RRID:AB_302859, 1:1,000), RAD50 (ABclonal, cat. #A3078, RRID:AB_2764881, 1:1,000), NBS1 (ABclonal, cat. #A4197, RRID:AB_2863208, 1:1,000), protein phosphatase 2A scaffold subunit A (PP2A-A; ABclonal, cat. #A6175, RRID:AB_2863527, 1:1,000), protein phosphatase 2A catalytic subunit C (PP2A-C; ABclonal, cat. #A6702, RRID:AB_2767286, 1:1,000), PR130 (Abcam, cat. #ab126195, RRID:AB_11130990, 1:1,000), and Na⁺/K⁺–ATPase (Abcam, cat. #ab76020, RRID:AB_1310695, 1:1,000).

Immunohistochemistry

Tumor tissues were fixed in 4% paraformaldehyde, paraffin-embedded, and sectioned into 4-μm slices. After dewaxing, hydration, and antigen retrieval, sections were treated with 3% H₂O₂ for 20 minutes and then blocked in 3% bovine serum albumin (BSA) for 30 minutes. Diluted primary antibodies were incubated with sections at 4°C overnight, followed by HRP-labeled polymeric anti-rabbit/anti-mouse secondary antibodies (Gene Tech; cat. #GK6005) at room temperature for 30 minutes. Staining was developed with DAB (Gene Tech; cat. #GK6005) and counterstained with hematoxylin (Gene Tech; cat. #GT1005). The images were captured using an optical microscope (Olympus). For quantitative analysis, five random fields of view were selected per slide, and fluorescence intensity was quantified using ImageJ software (version 1.8.0).

Diluted primary antibodies were as follows: LOX (Proteintech, cat. #17958-1-AP, RRID:AB_2234541, 1:200), COL1A1 (Proteintech, cat. #67288-1-Ig, RRID:AB_2882554, 1:200), PKD2 (Proteintech, cat. #19126-1-AP, RRID:AB_10641050, 1:100), and SDC4 (Proteintech, cat. #11820-1-AP, RRID:AB_2877797, 1:100).

Immunofluorescence

Cell slides were fixed with 4% paraformaldehyde, permeabilized with 0.7% Triton X-100 for 15 minutes, and washed three times with PBS. Cells were then blocked in 3% BSA for 30 minutes at room temperature. After blocking, cells were incubated with diluted primary antibodies overnight at 4°C, followed by incubation with appropriate fluorescent secondary antibodies and 4′,6-diamidino-2-phenylindole (DAPI) for nuclear counterstaining. Primary antibodies are as follows (source, dilution): MRE11 (Abcam, cat. #ab214, RRID:AB_302859, 1:100), γ-H2AX (Abcam, cat. #ab22551, RRID:AB_447150, 1:100), and BANF1 (Abcam, cat. #ab129184, RRID: AB_11150422, 1:100). Confocal images were captured using a Leica confocal microscope (Leica Microsystems). Conventional fluorescence images were obtained using an Olympus fluorescence microscope (Olympus Corporation). For quantitative analysis, five random fields of view were selected per slide, and fluorescence intensity was quantified using ImageJ software (version 1.8.0).

Immunofluorescence (IF) staining for tumor tissue slides was performed using a protocol similar to immunohistochemistry, with the exception that fluorescently labeled secondary antibodies were used instead of enzyme-conjugated secondary antibodies, and sections were counterstained with DAPI for nuclear visualization.

qRT-PCR

Cell RNA was extracted using TRIzol reagent (Thermo Fisher Scientific; cat. #5596018). Reverse transcription synthesis of cDNA from up to 1 μg of total RNA was performed using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific; cat. #3438563). PCR amplification was conducted using Hieff qPCR SYBR Green Premix (Yeasen; cat. #1202ES60). qRT-PCR was performed on the QuantStudio 5 instrument (Thermo Fisher Scientific). All reactions were repeated three times. GAPDH was used as the reference gene to determine the relative mRNA expression level by the ΔΔCT method. The primer sequences are shown in Supplementary Table S3.

Young modulus measurement

A custom-designed indentation system, implemented with assistance from Dr. Yuan Feng’s laboratory at Shanghai Jiao Tong University, was utilized to characterize matrix stiffness, as detailed in previous studies (26). Briefly, indentation movement was driven by an electric motor (SMAC Corporation), whereas indenter displacement was monitored in real time via a laser displacement sensor (HG-C1050, Panasonic). Sensor signals were sampled at 1 kHz and acquired using a data acquisition board (PCI-1706U, Advantech Co., Ltd.). A customized graphical user interface (GUI) based on LabVIEW (National Instruments) was implemented to enable automated indentation control and real-time data collection.

A simple magnetic tweezers experimental system

Referring to the method described previously (21) and the manufacturer’s instructions, 4.5 μm Tosyl-activated paramagnetic beads (Thermo Fisher Scientific, cat. #14013) were coated with SCD4 antibody (Santa Cruz Biotechnology, cat. # sc-12766, RRID: AB_628314) or PLL (Beyotime Biotechnology; cat. #C0313). Before use, these magnetic beads needed to be incubated at room temperature in 0.2 mol/L Tris buffer (pH 7.4) overnight to remove any residual mesylate. Adherent HCC cells were incubated with antibody-coated magnetic beads for 30 minutes at 37°C, and then 1× PBS was used to remove the excess magnetic beads. Subsequently, HCC cells conjugated to magnetic beads were subjected to magnetic field force for 30 minutes. The cells bound to the magnetic beads were collected for subsequent experiments.

Intracellular Ca²⁺ measurements

Intracellular calcium ion concentration was detected using the Fluo-4 AM fluorescent probe (Beyotime Biotechnology). HCC cells were cultured on high- or low-stiffness substrates for 48 hours. The culture medium was discarded, and the cells were gently washed three times with 1× PBS. Cells were then incubated with 500 μL of 4 µmol/L Fluo-4 AM working solution (diluted in PBS) at 37°C for 30 minutes in the dark. After incubation, excess probe was removed by three washes with 1× PBS. Fluorescence images were captured using an Olympus fluorescence microscope (Olympus). For quantitative analysis, five random fields of view were acquired per group. The mean fluorescence intensity (MFI) was analyzed and quantified using ImageJ software (version 1.8.0).

Cell stretching assay

HCC cells with PKD2 overexpression were seeded on a COL1-coated polydimethylsiloxane (PDMS) surface. Before cell stretching, Calcium Fluorescent Probes Fluo-4 (Beyotime Biotechnology) were used for calcium ion staining according to the manufacturer’s instructions. The cells on the PDMS were stretched to 20% deformation for 1 minute using the Cell Sealing Culture System (CELLTANK), and five fields of view were randomly photographed using a fluorescence microscope (Olympus, Japan). The MFI of the cells was analyzed by ImageJ version 1.8.0.

Comet assay

HCC cells (1 × 10⁵) were treated with oxaliplatin for 48 hours (MHCC97H cells with 30 μmol/L, and Hep3B cells with 20 μmol/L) After treatment, the cells were collected and resuspended in 150 μL of 1× PBS. Three neutral comet assays were then performed according to the manufacturer’s instructions using the Trevigen Comet Assay Kit (Beyotime Biotechnology; cat. #C2041M). Their fluorescence images were obtained using an Olympus fluorescence microscope (Olympus). The fluorescence signals of at least 30 cells on each slide were measured, and the average tail moment (TM) was calculated using the Comet Assay Software Project (Beijing Bio-launching Technology Co., Ltd.) to determine DNA damage.

Co-immunoprecipitation and immunoprecipitation

Nuclear and total proteins from HCC cells were immunoprecipitated with indicated capture antibodies. The extracted nuclear or total proteins (400 μg) were incubated with capture antibodies [PP2RCA (ABclonal, cat. #A6702, RRID:AB_2767286, 1:100), Flag (Proteintech, cat. #66008-4-Ig, RRID:AB_2918475, 1:100), MRE11 (Abcam, cat. #ab214, RRID:AB_302859, 1:100)] or control IgG (Cell Signaling Technology, cat. #2729, RRID:AB_1031062, 1:100) at 4°C overnight under continuous rotation. Subsequently, 30 μL of protein A/G mixed magnetic beads (Millipore, cat. #618989) were added, and the mixtures were further incubated at room temperature for 30 minutes under rotation. Magnetic beads conjugated with antigen-antibody complexes were precipitated using a magnetic separator (Invitrogen). After washing with 1× PBS three times, the complexes were denatured in 1× SDS loading buffer containing dithiothreitol and boiled for 10 minutes at 95°C. Captured proteins in the supernatant were analyzed by Western blot.

PP2A enzyme activity

For each group, 5 × 10⁷ cells were collected and lysed. Protein concentrations were determined using the BCA Assay kit (Beyotime Biotechnology, cat. #P0012). PP2A enzyme activity was measured according to the manual of the Serine/Threonine Phosphatase Assay System kit (Promega, cat. #V2460).

Molecular docking analysis using Schrodinger software

Protein sequences and basic information were retrieved from the UniProt database. Two human proteins, MRE11 (UniProt ID: P49959) and BANF1 (UniProt ID: O75531), were selected for analysis, with their resolved crystal structures (PDB IDs: 3TLI for MRE11 and 7Z21 for BANF1) used for subsequent modeling. Structural energy minimization of both proteins was performed using the OPLS_4 force field. Protein–protein docking simulation was carried out using the Piper algorithm module embedded in the Schrodinger platform.

A total of 1,000 initial rotational conformations were generated, and these conformations were clustered according to the root-mean-square deviation of atomic coordinates. For each cluster, the conformation with the highest neighbor density was defined as the representative structure. All clustered conformations were further ranked by cluster size, and the dominant conformation from the largest cluster was chosen as the optimal binding mode for subsequent functional analysis.

Animal and breeding

Male C57BL/6 mice (RRID: MGI:2159769; 4–6 weeks old) were purchased from Weitong Lihua Lab Anim. Cent (Vital River Laboratory Animal Technology Co., Ltd). All mice were housed in a specific pathogen-free laboratory under a 12/12-hour dark/light cycle, at 22°C to 24°C, and 30% to 70% humidity, with ad libitum access to food and water. The diet was commercial feed (Rat/Mouse Growth & Reprod. Feed, cat. #P1100; ≥20.5% crude protein, ≥4% crude fat; Shanghai SLAC Laboratory Animal Co., Ltd.).

Liver cancer mouse models with high- and normal-liver-stiffness backgrounds

Mice in the high-liver-stiffness group were injected intraperitoneally with 10% carbon tetrachloride (in olive oil) twice a week for 8 weeks, and mice in the normal-liver-stiffness group received the same amount of olive oil twice a week for 8 weeks. On day 56, Hepa1-6 cells (1 $\mathrm{\times }$ 10⁶) were injected subcutaneously into the liver capsule. Three weeks after the implantation (on day 78), all mice were euthanized, and their livers were collected. The size of the tumor was measured, and the volume of the tumor (mm³) was calculated using the following formula: tumor volume = (long axis) × (short axis)² × 1/2. The weight of the tumor was measured using an electronic balance. Animal experiments were approved by the Animal Care Ethics Committee of Zhongshan Hospital, Fudan University (Shanghai, China).

The terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling assay

Apoptotic cells in paraffin-embedded tumor tissue sections were detected using a terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) detection kit (Servicebio, cat. #G1507) according to the manufacturer’s instructions. Stained sections were observed and imaged using an Olympus optical microscope (Olympus). For quantitative analysis, five random nonoverlapping fields of view were captured per tissue slide. The level of apoptosis in HCC tissues was determined by measuring the average optical density of TUNEL-positive nuclei using ImageJ software (version 1.8.0).

Patients with HCC and their tumor tissue specimens

Clinical data of 17 patients with HCC who underwent curative hepatectomy within 3 months after TACE at Zhongshan Hospital affiliated with Fudan University (April 2007–November 2009) were collected, and their tumor tissue sections were obtained from the Department of Pathology of the same hospital, with written informed consent obtained from the patients. The study was approved by the Ethics Committee of Zhongshan Hospital Affiliated with Fudan University, in compliance with the Declarations of Helsinki and Istanbul. Taking the median expression of LOX and COL1A1 in tumor tissues as the cutoff, the patients were stratified into two groups: COL1A1^Low/LOX^Low (n = 8, low-matrix-stiffness group) and COL1A1^High/LOX^High (n = 9, high-matrix-stiffness group).

Statistics and reproducibility

In vitro cell assays were performed with at least three biological replicates, with cells/conditions randomly assigned to experimental groups. In vivo animal experiments were conducted with at least six animals per group, and mice were randomized before treatment. Researchers were blinded to clinical information before analysis. Quantitative data are presented as mean ± standard deviation. Statistical analyses and graphs were generated using GraphPad Prism 9.0.0. An unpaired two-tailed Student t test was used to compare continuous variables between two independent groups. For comparisons across multiple groups, one-way ANOVA was used for data with normal distribution and equal variance. Two-way ANOVA was applied to evaluate the main effects of two independent variables and their interaction. Categorical variables were analyzed using Pearson χ² test. Survival curves were plotted using the Kaplan–Meier method, and differences between groups were assessed using the log-rank test. All statistical tests were two-sided, and a P value <0.05 was considered statistically significant.

Results

Screening and identification of novel mechanosensors in HCC

As cell membranes are the interface between cells and their surrounding microenvironment, individual specific proteins on them are likely to participate in sensing and delivering extracellular mechanical signals. Undoubtedly, it should be the most direct and efficient strategy to screen and identify mechanosensors in membrane proteins. Based on the structural characteristics of mechanosensors (27) and the phenotypic changes of HCC cells under mechanical signal stimulation (28), combined with single-molecule mechanical manipulation methods (29), we developed a standard procedure for screening and characterizing novel mechanosensors in membrane proteins of HCC cells (Fig. 1A). We first established low- and high-stiffness substrates (6 and 16 kPa) as per the protocol described previously (14) to culture HCC cells and observed that increased matrix stiffness substantially influenced the morphologic phenotype of HCC cells (Supplementary Fig. S1A). Our previous studies also showed that increased matrix stiffness prominently induced epithelial–mesenchymal transition in HCC cells (14) and enhanced their motility and invadopodia formation abilities (30). These morphologic changes indicate that specific membrane proteins are involved in matrix stiffness–mediated effects on malignant phenotypes through direct sensing and delivering extracellular mechanical signals. Subsequently, we isolated the membrane proteins of HCC cells grown on high- and low-stiffness substrates and evaluated their quality. The protein bands of extracted membrane proteins in SDS-PAGE were evenly distributed (Supplementary Fig. S1B). Simultaneously, Na⁺/K⁺–ATPase was highly expressed, but GAPDH was almost undetectable (Supplementary Fig. S1C). On the other hand, we analyzed the expression levels of Piezo1 and ITGB1, two known HCC mechanosensors (14, 15), to clarify the stiffness dependence of membrane protein expression. The expression levels of Piezo1 and ITGB1 in the high-stiffness group all exhibited an obvious increase (Supplementary Fig. S1D). These results suggest that the quality, purity, and stiffness dependence of the extracted membrane proteins are well qualified for subsequent proteomics screening analysis.

Figure 1.

Screening and identification of novel mechanosensors in HCC. A, Schematic for screening and identification of novel mechanosensors in HCC membrane proteins. B, Four-dimensional label-free proteomic analysis of stiffness-dependent membrane proteins. C, Heatmap of candidate differential membrane proteins. D and E, SDC4 and PKD2 expressions in membrane fractions and total protein lysates of HCC cells on low- vs. high-stiffness substrates (n = 3). P, unpaired t test. F, Schematic of 3D hydrogel culture system. G, Cell aggregates in low- and high-stiffness hydrogels on days 1 and 14. Scale bars, 50 μm. H, SDC4 and PKD2 expressions in low- vs. high-stiffness hydrogels (n = 3). P, unpaired t test. I, Magnetic tweezers system. J and K, p-FAK and p-YAP ratios in HCC cells under varying mechanical force (n = 3). P, two-way ANOVA. L, A workflow of cell stretching experiment. M, Intracellular Ca²⁺ levels in HCC cells before and after mechanical stretching (n = 5). Scale bars, 50 μm. P, two-way ANOVA. ns, P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001. A, created in BioRender. Li, J. (2026) https://BioRender.com/enyrrzr; F, created in BioRender. Li, J. (2026) https://BioRender.com/o1bmo42; I, created in BioRender. Li, J. (2026) https://BioRender.com/i8l52du; L, created in BioRender. Li, J. (2026) https://BioRender.com/z60z48e.

Applying a 4D label-free quantitative proteomics technology, we continued to screen stiffness-dependent differential membrane proteins between the high- and low-stiffness groups. Considering that ITGB1, as a mechanosensor, has been acknowledged in HCC (14), we defined the fold change of ITGB1 (>1.46 or <0.68) and statistical significance (P < 0.05) as the threshold to determine stiffness-dependent differential membrane proteins. Meanwhile, nonmembrane proteins among the identified differential proteins were further excluded using the CELLO website (http://cello.life.nctu.edu.tw/). We identified a total of 445 stiffness-dependent differential membrane proteins, including 426 upregulated membrane proteins and 19 downregulated membrane proteins in the high-stiffness group (Fig. 1B; Supplementary Fig. S1E). The currently known mechanosensors are mainly classified into ion channel proteins and receptor/adhesion molecule proteins (31). Calcium channel proteins are the most prominent type of ion channel proteins (32). Receptor/adhesion molecule proteins are mostly transmembrane proteins, and their extracellular domains are frequently linked to extracellular matrix (ECM) proteins (33). Therefore, among the upregulated membrane proteins, we mainly focused on the two types of proteins mentioned above. We used the Kyoto Encyclopedia of Genes and Genomes (KEGG) gene set (hsa04512: ECM–receptor interaction) as a screening gene set to determine candidate adhesion molecules/receptor proteins. On the other hand, we used UniProt (https://www.uniprot.org) in combination with a relevant literature review (34–39) to form a calcium ion channel gene set (Supplementary Table S4) for determining candidate calcium ion channel proteins. Nine candidate differential membrane proteins were ultimately obtained, including five calcium channel proteins and four receptors/adhesion molecules (Fig. 1C; Supplementary Fig. S1F; Supplementary Table S5). Among the four receptor/adhesion molecule proteins, syndecan-4 (SDC4), a leading differential protein, was selected as the target protein of interest. Consistently, among the 5 differential calcium channel proteins, polycystin-2 (PKD2) was identified as the leading differential protein to be the target protein of interest. The transmembrane region (https://services.healthtech.dtu.dk/services/TMHMM-2.0/) and characteristic fingerprint maps of mass spectrometry for SDC4 and PKD2 are presented in Supplementary Fig. S1G and S1H. Subsequently, we confirmed the high expressions of SDC4 and PKD2 in the membrane proteins and total proteins in the high-stiffness group (Fig. 1D and E), supporting that the expression changes of PKD2 and SDC4 are stiffness-dependent. To further corroborate this finding, we mixed RGD-modified sodium alginate with CaCl₂ to develop a 3D hydrogel culture system (Fig. 1F). There was no significant difference in the Young modulus between the surface and the internal parts of the hydrogels (Supplementary Fig. S1I). The Young modulus of the hydrogel with 25 mmol/L Ca²⁺ and that of the hydrogel with 45 mmol/L Ca²⁺ were close to 6 and 16 kPa (Supplementary Fig. S1I), which mirrored the normal liver tissue stiffness and cirrhotic liver tissue stiffness. Simultaneously, there was almost identical permeability and pore size uniformity between high- and low-stiffness hydrogels (Supplementary Fig. S1J and S1K). These data suggest that the established hydrogels can be used to explore matrix stiffness–mediated effects. We examined the tissue-like characteristics of cell aggregations in the hydrogel and discovered that compared with those in the low-stiffness hydrogel, the expressions of ICAM1, ITGB1, and paxillin were all increased in the high-stiffness hydrogel (Supplementary Fig. S1L and S1M). Furthermore, the expressions of Ki67, MMP2, and nuclear YAP were also elevated. These results support that the cell aggregations in the high-stiffness hydrogel exhibited tissue-like characteristics and malignant characteristics, indicating that they better simulate HCC tissues with normal- and high-liver-stiffness backgrounds. Additionally, the size of cell aggregations in the high-stiffness hydrogel was obviously larger than that in the low-stiffness hydrogel (Fig. 1G), suggesting a promotional effect of matrix stiffness on cell proliferation. High expression of Ki67 in cell aggregation in the high-stiffness hydrogel (Supplementary Fig. S1L and S1M) also supported this result. More importantly, the expression levels of SDC4 and PKD2 in the high-stiffness hydrogel were remarkably higher than those in the low-stiffness hydrogel (Fig. 1H), in accordance with the findings mentioned above. This evidence confirms that high-stiffness stimulation indeed increases the expression levels of PKD2 and SDC4 in HCC cells.

To clarify the bridging role of PKD2 and SDC4 in mechanical signal transmission, we further evaluated the effects of SDC4 overexpression/knockdown or PKD2 overexpression/knockdown (Supplementary Fig. S1N and S1O) on downstream classical stiffness-response molecules in HCC cells. The results demonstrated that the expression levels of SDC4 or PKD2 were positively correlated with the ratio of p-FAK but negatively correlated with the ratio of p-YAP (Supplementary Fig. S1P–S1S), indicating that PKD2 and SDC4 can deliver stiffness mechanical signals to activate downstream stiffness-response molecules in HCC cells. Afterward, a simple magnetic tweezers experimental system was constructed to directly corroborate the bridging role of SDC4 in mechanical signal transmission. The schematic diagrams of the experimental procedure and the device are shown in Fig. 1I and in Supplementary Fig. S1T. Using Maxwell simulation software, we clarified the relationship between the magnetic field intensity and distance in the positive direction of the Z-axis at the center point of the surface of the device (Supplementary Fig. S1U). According to the magnetic force calculation formula (21), we determined the magnetic force values of the high magnetic force group (approximately 12 pN) and the low magnetic force group (approximately 6 pN). Meanwhile, we prepared the coated magnetic beads with SCD4 antibody and co-cultured with HCC cells in a dish. The HCC cells conjugated to magnetic beads were then subjected to magnetic force for 30 minutes. We examined the expression changes in downstream classical stiffness-response molecules and found that compared with the nonspecific magnetic force group, the ratio of p-FAK was significantly increased, whereas the ratio of p-YAP was decreased in the SCD4-specific magnetic force group. Moreover, as the magnetic force increased, the ratio of p-FAK was prominently elevated, and the ratio of p-YAP was obviously reduced in the SCD4-specific magnetic force group (Fig. 1J and K). These data strongly suggest that SDC4 can directly respond to mechanical signals and alter downstream cell function.

Change in intracellular calcium ion content can indirectly reflect the activity of PKD2. We cultured HCC cells with PKD2 overexpression on a low-stiffness substrate and investigated the relationship between PKD2 expression and activity. We observed an increase in intracellular calcium ion content (Supplementary Fig. S1V), but the PKD2 inhibitor amiloride effectively attenuated this effect (Supplementary Fig. S1W), indicating a positive correlation between PKD2 expression and its functional activity. We used HCC cells on high- and low-stiffness substrates to test whether PKD2 can sense mechanical signal stimulation. The results showed that intracellular calcium ion content in the high-stiffness group was significantly higher than that in the low-stiffness group (Supplementary Fig. S1X). The above data indirectly reveal that PKD2 may serve as a mechanosensor to sense mechanical signal stimulation.

Using the cell stretching device, we further confirmed the role of PKD2 in sensing mechanical signals. The workflow of the cell stretching experiment and the curve of stretching distance and time are shown in Fig. 1L and Supplementary Fig. S1Y. We comparatively analyzed the intracellular calcium ion content in HCC cells before and after mechanical stretching and found that the intracellular Ca²⁺ content in HCC cells with PKD2 overexpression was significantly increased after cell stretching. Simultaneously, the PKD2 inhibitor amiloride effectively reversed this effect (Fig. 1M). Collectively, these findings, combined with the results in Supplementary Fig. S1V–S1X, confirm that PKD2 senses mechanical signals and therefore increases Ca²⁺ influx in HCC cells.

Increased matrix stiffness promotes DDR in HCC cells

In our preliminary work, we observed a phenomenon in which increased matrix stiffness attenuates the efficacy of oxaliplatin-induced apoptosis in HCC cells (40). To further confirm the influence of matrix stiffness on cytotoxic drug response, we applied oxaliplatin to treat HCC cells grown on high- and low-stiffness substrates for 48 hours and discovered that the number of apoptotic cells, as well as the ratio of cleaved caspase 3 and γ-H2AX in the high-stiffness group, were all significantly lower than those in the low-stiffness group (Fig. 2A and B). Consistently, analysis of a comet assay also revealed that tail moment (TM) of HCC cells in the high-stiffness group was substantially shorter compared with that in the low-stiffness group (Fig. 2C). To clarify that increased matrix stiffness causes oxaliplatin resistance through DDR rather than defective DNA damage perception, we performed a time course analysis of γH2AX and additional activated DDR markers (p-ATM and p-CHK2) in oxaliplatin-treated HCC cells grown on high- and low-stiffness substrates. Within the first 12 hours, there were no significant differences in p-ATM and p-CHK2 ratios and γ-H2AX expression between high- and low-stiffness groups. However, as time went on, γ-H2AX levels decreased more rapidly in the high-stiffness group, whereas p-ATM and p-CHK2 ratios exhibited an increasing trend (Supplementary Fig. S2A). Time course analysis of γ-H2AX expression in IF staining (Supplementary Fig. S2B) also showed the same trend. Together, matrix stiffness indeed increases oxaliplatin resistance in HCC cells by improving DDR.

Figure 2.

Increased matrix stiffness promotes DDR in HCC cells. A, Apoptosis in HCC cells treated with oxaliplatin for 48 hours on low- vs. high-stiffness substrates (n = 3). P, unpaired t test. B and C, Cleaved caspase 3 and γ-H2AX ratios (n = 3) and nuclear DNA TM (n = 30) in HCC cells treated with oxaliplatin for 48 hours on low- vs. high-stiffness substrates. Scale bars, 50 μm. P, unpaired t test. D, Schematic of high- and normal-liver-stiffness HCC mouse models for evaluating oxaliplatin response. E, Tumor appearance after oxaliplatin treatment in high- and normal-liver-stiffness HCC mouse models. F, Hematoxylin and eosin (H&E) staining showing HCC tumor cell morphology. Scale bars, 100 μm. G, Liver tumor volume and weight after oxaliplatin treatment in high- vs. normal-stiffness groups (n = 6). P, unpaired t test. H–J, TUNEL staining of apoptotic cells and expression levels of SDC4 and PKD2 in tumor tissues following oxaliplatin treatment, in normal- and high-liver-stiffness backgrounds (n = 6). Scale bars, 50 μm. P, unpaired t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. D, created in BioRender. Li, J. (2026) https://BioRender.com/72rh15c.

By the combined modeling approach (Fig. 2D), we developed liver cancer mouse models with high- and normal-liver-stiffness backgrounds to further evaluate oxaliplatin drug response in vivo. On the 52nd day of model establishment, six mice from both the high- and normal-liver-stiffness groups were randomly euthanized, and their liver tissues were collected for liver stiffness assessment. Surface of the liver in the high-liver-stiffness group was significantly rougher than that in the control group (Supplementary Fig. S2C). Analysis of Sirius red staining, reticular fiber staining, and Masson trichrome staining showed that the positive areas in the high-liver-stiffness group were all markedly greater than those in the control group (Supplementary Fig. S2D), indicating the formation of mouse models with high liver stiffness. On day 56, Hepa1-6 cells were orthotopically implanted into the livers of model mice with high or normal liver stiffness. On day 66, mice received an i.p. injection of 5 mg/kg oxaliplatin every 3 days. All mice were euthanized on day 78, and tumor tissues were harvested (Fig. 2E). Hematoxylin and eosin staining showed typical tumor cell morphology (Fig. 2F). The volume and weight of liver tumors in the high-liver-stiffness background group were distinctly greater than those in the control group (Fig. 2G). TUNEL staining analysis revealed that apoptosis levels in HCC cells in the high-liver-stiffness background group were remarkably lower than those in the control group (Fig. 2H). All these findings suggest that liver tumors in a high-liver-stiffness background exhibit an obvious resistance to oxaliplatin. Additionally, the expression levels of two mechanosensors, SDC4 and PKD2, were significantly elevated in the high-liver-stiffness background group (Fig. 2I and J), which corroborated the findings of cell assays in vitro.

SDC4 and PKD2 collaboratively participate in matrix stiffness–promoted DDR in HCC cells

HCC cells with SDC4 overexpression were cultured on low-stiffness substrates and treated with oxaliplatin for 48 hours. SDC4 overexpression significantly reduced the cleaved caspase 3/caspase 3 ratio, γ-H2AX expression, comet assay TM, and apoptotic cell number (Fig. 3A and B; Supplementary Fig. S3A, S3B), indicating that SDC4 overexpression enhanced oxaliplatin-induced DDR in low-stiffness cultures. Conversely, SDC4 knockdown yielded opposite effects in high-stiffness cultures (Fig. 3C and D; Supplementary Fig. S3C and S3D), impairing DDR. Parallel experiments in HCC cells with PKD2 overexpression or knockdown revealed analogous results. PKD2 overexpression in low-stiffness cultures recapitulated the DDR phenotypes of HCC (Fig. 3E and F; Supplementary Fig. S3E and S3F), whereas PKD2 knockdown reversed them in high-stiffness cultures (Fig. 3G and H; Supplementary Fig. S3G and S3H). Collectively, these data demonstrate that SDC4 and PKD2 can mediate matrix stiffness–promoted DDR together in HCC cells.

Figure 3.

SDC4 and PKD2 collaboratively participate in matrix stiffness–promoted DDR in HCC cells. A–D, Detection of cleaved caspase 3 and γ-H2AX ratios (n = 3) and nuclear DNA TM (n = 30) in oxaliplatin-treated HCC cells. Cells were cultured on low-stiffness substrates with SDC4 overexpression or high-stiffness substrates with SDC4 knockdown for 48 hours. Scale bars, 50 μm. P, unpaired t test and one-way ANOVA. E–H, Detection of cleaved caspase 3 and γ-H2AX ratios (n = 3) and nuclear DNA TM (n = 30) in oxaliplatin-treated HCC cells. Cells were cultured on low-stiffness substrates with PKD2 overexpression or high-stiffness substrates with PKD2 knockdown for 48 hours. Scale bars, 50 μm. P, unpaired t test & one-way ANOVA. I–L, Analysis of cleaved caspase 3 and γ-H2AX ratios (n = 3), and nuclear DNA TM (n = 30) in high-stiffness cultured HCC cells after 48 hours of oxaliplatin treatment, with combined intervention of SDC4 knockdown plus amiloride or PKD2 knockdown. Scale bars, 50 μm. P, two-way ANOVA. M and N, Effects of single SDC4 knockdown, single PKD2 knockdown, or combined SDC4/PKD2 knockdown on tumor growth and tumor tissue apoptosis in high-liver-stiffness background HCC mouse murine liver cancer models following 48 hours of oxaliplatin treatment (n = 6). Scale bars, 50 μm. P, two-way ANOVA. *, P < 0.05; **, P < 0.01; ***, P < 0.001. AOD, average optical density.

We further elucidated whether there was a collaborative role of SDC4 and PKD2 in mediating matrix stiffness–promoted DDR. Compared with SDC4 knockdown alone and amiloride alone, SDC4 knockdown combined with amiloride obviously elevated the cleaved caspase 3/caspase 3 ratio and γ-H2AX expression in HCC cells grown on a high-stiffness substrate (Fig. 3I) and prolonged the TM of HCC cells (Fig. 3J; Supplementary Fig. S3I). Similarly, compared with single downregulation, the combined downregulation of SDC4 and PKD2 also prominently elevated the cleaved caspase 3 ratio, γ-H2AX expression (Fig. 3K), comet TM (Fig. 3L; Supplementary Fig. S3J), and apoptotic cell proportion (Supplementary Fig. S3K). These results suggest that SDC4 and PKD2 can collaboratively mediate DDR in HCC cells.

These findings were validated using high-liver-stiffness mice orthotopically implanted with Hepa1-6 cells stably expressing shSDC4#3, shPKD2#2, or shSDC4 + shPKD2 (selected based on knockdown efficiency; Supplementary Fig. S3L). Ten days after implantation, mice received oxaliplatin (5 mg/kg, intraperitoneally, every 3 days) for 9 days, and then their tumor tissues were harvested (Supplementary Fig. S3M). SDC4 or PKD2 knockdown significantly reduced tumor weight and volume while increasing intratumoral apoptosis (Fig. 3M and N). Notably, the combined knockdown of SDC4/PKD2 elicited the most pronounced antitumor effects and highest apoptotic rates (Fig. 3M and N). These invivo data corroborate the findings invitro, also supporting that SDC4 and PKD2 collaboratively mediate matrix stiffness–promoted DDR in HCC.

SDC4 and PKD2 collaboratively mediate matrix stiffness–increased BANF1 nuclear translocation

Given a collaborative role between SDC4 and PKD2 in stiffness-driven DDR, nuclear-translocated proteins mediating this repair were further screened. Oxaliplatin-treated HCC cells on high- or low-stiffness substrates were subjected to nuclear protein extraction, with purity confirmed (Supplementary Fig. S4A). Nuclear γ-H2AX was suppressed, whereas RAD51 (a DDR marker) was elevated in the high-stiffness group, confirming enhanced repair capacity (Supplementary Fig. S4B). Nuclear proteomics analysis showed 729 differentially expressed proteins (DEPs, cutoffs of |fold change| > 2 and P < 0.05). Subcellular localization analysis (CELLO) was used to filter the proteins that simultaneously localized in both the cytoplasm and nucleus, yielding 145 candidate DEPs (94 upregulated and 51 downregulated in high stiffness; Fig. 4A). The top 20 DEPs were ranked by fold change (Fig. 4B; Supplementary Table S6).

Figure 4.

SDC4 and PKD2 collaboratively mediate matrix stiffness–increased BANF1 nuclear translocation. A, Analysis of differentially expressed nuclear proteins between the high- and low-stiffness groups by 4D label-free proteomics technology. B, Heatmap of the top 20 differentially expressed nuclear protein candidates. C and D, Analysis of BANF1 expression in nuclear proteins (n = 3) and total proteins (n = 3) of the oxaliplatin-treated HCC cells for 48 hours grown on high- and low-stiffness substrates. P, unpaired t test. E, The nuclear localization expression of BANF1 in tumor tissues from mouse liver cancer models with normal- and high-liver-stiffness backgrounds (n = 6). Scale bars, 50 μm. P, unpaired t test. F and G, Effect of SDC4 overexpression or knockdown on nuclear BANF1 levels in HCC cells treated with oxaliplatin for 48 hours (n = 3). P, unpaired t test and one-way ANOVA. H and I, Effect of PKD2 overexpression or knockdown on nuclear BANF1 levels in HCC cells treated with oxaliplatin for 48 hours (n = 3). P, unpaired t test & one-way ANOVA. J, Nuclear BANF1 levels in oxaliplatin-treated (48 hours) HCC cells on high-stiffness substrates under combined SDC4 knockdown and amiloride treatment (n = 3). P, two-way ANOVA. K, Nuclear BANF1 levels in oxaliplatin-treated (48 hours) HCC cells on high-stiffness substrates under combined SDC4 and PKD2 knockdown (n = 3). P, two-way ANOVA. L, Nuclear localization of BANF1 in tumor tissues from SDC4-, PKD2-, or SDC4/PKD2-downregulated HCC mouse models with high liver stiffness after oxaliplatin treatment (n = 6). Scale bars, 50 μm. P, two-way ANOVA. ns, P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

By intersection analysis between differential nuclear proteins and DDR proteins previously reported (41), BANF1 was determined to be the target nuclear protein that may participate in DDR. To validate the impact of matrix stiffness on the expression localization of BANF1, we examined the expression levels of BANF1 in total protein and nuclear protein from oxaliplatin-treated HCC cells grown on high- and low-stiffness substrates. The results displayed an obvious increase in the nucleoprotein level of BANF1 in the high-stiffness group (Fig. 4C). However, there was no significant difference in the total protein level of BANF1 between the high- and low-stiffness groups (Fig. 4D). Additionally, we analyzed the expression localization of BANF1 in tumor tissue sections from HCC mouse models with normal- and high-liver-stiffness backgrounds. BANF1 nuclear translocation was significantly increased in HCC cells in the high-liver-stiffness background group (Fig. 4E). Matrix stiffness enhanced BANF1 nuclear translocation without altering its total expression.

We next investigated whether SDC4 and PKD2 mediate this process. SDC4 overexpression increased nuclear BANF1 levels (Fig. 4F), whereas SDC4 knockdown decreased them (Fig. 4G), with no changes in total BANF1 (Supplementary Fig. S4C and S4D). Concordantly, PKD2 overexpression (Fig. 4H) and PKD2 knockdown also showed a similar effect on nuclear BANF1 levels, without affecting its total expression (Fig. 4I; Supplementary Fig. S4E and S4F). Cotreatment with SDC4 knockdown and a PKD2 inhibitor (amiloride) caused a more pronounced reduction in BANF1 nuclear translocation in high-stiffness cultures compared with single-agent treatment (Fig. 4J). Similarly, the combined knockdown of SDC4 and PKD2 exerted the strongest inhibitory effect (Fig. 4K). In tumor tissue sections from HCC mouse models with high-liver-stiffness backgrounds, SDC4 or PKD2 knockdown diminished intratumoral BANF1 nuclear translocation, with the combined knockdown group showing the most significant reduction (Fig. 4L), consistent with in vitro findings. Collectively, SDC4 and PKD2 collaboratively mediate matrix stiffness–induced BANF1 nuclear translocation in HCC.

SDC4 and PKD2 collaboratively regulate BANF1 dephosphorylation through PP2A enzyme assembly and activity and DDR in HCC cells

BANF1 phosphorylation determines its subcellular localization (42). To clarify the inhibitory effect of BANF1 phosphorylation on its nuclear translocation, we transfected the control plasmid, BANF1 WT plasmid (WT-BANF1), MAAAQ, and MDDDQ into HCC cells to analyze their BANF1 nuclear expression. Nuclear BANF1 was significantly higher in WT-BANF1 cells than in controls. Compared with WT-BANF1, nuclear BANF1 was markedly increased in the MAAAQ group but decreased in the MDDDQ group (Fig. 5A). These data demonstrate that BANF1 dephosphorylation promotes its nuclear translocation. We further explored how SDC4 and PKD2 caused BANF1 dephosphorylation. Immunoprecipitation (IP) assays revealed that the knockdown of SDC4 or PKD2 obviously increased BANF1 phosphorylation levels in HCC cells (Fig. 5B and C), indicating a regulatory role of SDC4 or PKD2 in BANF1 phosphorylation. A report shows that BANF1 can be dephosphorylated by PP2A (43), which is a heterotrimeric complex composed of a scaffold subunit (PP2A-A), a regulatory subunit (PP2A-B), and a catalytic subunit (PP2A-C; ref. 44). Therefore, PP2A-C plays a critical role in modulating BANF1 phosphorylation (45). Co-IP assays confirmed that there was an interaction between BANF1 and PP2A-C (Fig. 5D). Moreover, LB-100 intervention (PP2A inhibitor) significantly reduced the nuclear protein level of BANF1 (Supplementary Fig. S5A). Additionally, LB-100 intervention decreased the ratios of p-ATM and p-CHK2 and increased the level of γ-H2AX. These results suggest that PP2A affects DDR by BANF1 dephosphorylation (Supplementary Fig. S5B). Existing research demonstrates that integrin facilitates the assembly of the PP2A holoenzyme (46), and cytoskeletal remodeling is also one of the critical influencing factors for PP2A assembly (47). SDC4 is similar to integrin in terms of numerous structures and functions; its extracellular domain interacts with matrix proteins and influences cytoskeletal remodeling (48). Additionally, PP2A enzymatic activity was activated by Ca²⁺ binding to the EF-hand domain of the PR130 subunit in PP2A-B (49). Based on the above evidence, we speculated that SDC4 and PKD2 may collaboratively influence PP2A assembly and activity to dephosphorylate BANF1. PKD2 overexpression significantly enhanced PP2A activity in HCC cells, whereas amiloride markedly suppressed it (Fig. 5E). Combined with the data in Supplementary Fig. S1W, we concluded that PKD2-mediated Ca²⁺ influx increased PP2A activity. Additionally, SDC4 overexpression strengthened the interactions of PP2A-C with PP2A-A and PP2A-PR130 (Fig. 5F), whereas SDC4 knockdown reduced this effect (Supplementary Fig. S5C), suggesting that SDC4 promotes PP2A holoenzyme assembly. Together, SDC4 promotes PP2A assembly, whereas PKD2 enhances PP2A activity. They collaboratively enhance BANF1 dephosphorylation and increase its nuclear translocation.

Figure 5.

SDC4 and PKD2 collaboratively regulate BANF1 dephosphorylation through PP2A enzyme assembly and activity in HCC cells. A, A comparative analysis of the nucleoprotein level of BANF1 in HCC cells transfected with control plasmid, WT-BANF1, MAAAQ, or MDDDQ (n = 3). P, two-way ANOVA. B and C, IP assay for BANF1 phosphorylation in HCC cells with SDC4 or PKD2 knockdown. D, Co-IP assay detecting the interaction between BANF1 and PP2A-C in HCC cells. E, PP2A activity was measured in HCC cells with PKD2 overexpression, with or without amiloride intervention. P, one-way ANOVA. F, IP analysis of the interactions among PP2A-C, PP2A-A, and PP2A-PR130 in HCC cells with SDC4 overexpression. G, The expression of γ-H2AX in the oxaliplatin-treated HCC cells transfected with control plasmid, WT-BANF1, MAAAQ, and MDDDQ (n = 3). P, two-way ANOVA. H, The nuclear DNA TM of oxaliplatin-treated HCC cells for 48 hours transfected with control plasmid, WT-BANF1, MAAAQ, and MDDDQ (n = 30). Scale bars: 50 μm. P, two-way ANOVA. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

We further observed the impact of BANF1 nuclear translocation on DDR and noted that HCC cells in the WT-BANF1 plasmid group and in the MAAAQ plasmid group exhibited a low expression of γ-H2AX and a shortened TM (Fig. 5G and H). Conversely, HCC cells in the MDDDQ plasmid group presented high expression of γ-H2AX and a prolonged TM (Fig. 5G and H). Accordingly, BANF1 nuclear translocation prominently enhances DDR.

BANF1 enhances DDR in HCC cells through modulating the formation of the MRE11–RAD50–NBS1 complex

We continued to explore how BANF1 nuclear translocation regulated DDR. IF revealed obvious nuclear colocalization of BANF1 and γ-H2AX at 24 and 48 hours after oxaliplatin treatment (Fig. 6A; Supplementary Fig. S6A), indicating BANF1 enrichment at DNA DSB sites. Furthermore, we compared the changes in BANF1, p-ATM, p-CHK2, and γH2AX levels in nucleoprotein at 24 and 48 hours after oxaliplatin treatment, and the results showed that the expressions of nuclear BANF1 and repair markers (p-ATM, p-CHK2) gradually increased, but the ratio of γH2AX increased in the early stages of DNA damage and then decreased (Supplementary Fig. S6B), supporting the finding that BANF1 nuclear translocation enhances DDR. Subsequently, we applied IP coupled with tandem mass spectrometry (IP-MS/MS) to screen the interaction proteins of BANF1 in the nuclear proteins of HCC cells with Flag-BANF1. The quality of IP proteins was evaluated by Coomassie Brilliant Blue staining (Supplementary Fig. S6C). Using thresholds of peptide counts >2 and IP/IgG fold change >2 (or IgG = 0), 90 candidate interactors were identified. By an intersection analysis with the reported DDR gene set (50), the only common protein, MRE11, was determined to be the binding target protein with nuclear BANF1 (Fig. 6B). Next, we used Co-IP assays (Fig. 6C) and IF staining (Fig. 6D) to confirm the binding interaction between BANF1 and MRE11. The results all supported the binding interaction between them. During DDR, the MRE11–RAD50–NBS1 (MRN) complex, consisting of MRE11, RAD50, and NBS1, can be rapidly formed and localized at DSB sites. This complex activates ATM kinase to initiate the process of DNA repair (51). We inferred that BANF1 nuclear translocation may promote DDR by modulating MRN complex formation. The IP assay showed that BANF1 overexpression significantly enhanced MRE11 interaction with RAD50 and NBS1 (Fig. 6E), indicating that BANF1 nuclear translocation may facilitate MRN complex assembly.

Figure 6.

BANF1 nuclear translocation enhances DDR in HCC cells. A, A colocalization analysis between BANF1 and γ-H2AX in the nucleus of HCC cells after 48 hours of oxaliplatin treatment. Blue, DAPI; green, γ-H2AX; red, BANF1. Scale bars, 5 μm. B, An intersection analysis between the identified binding proteins with BANF1 and the reported DDR genes. C, Co-IP assays confirmed that there was a binding interaction between BANF1 and MRE11 in the nucleus of HCC cells after 48 hours of oxaliplatin treatment. D, A co-localization analysis between MRE11 and BANF1 in the oxaliplatin-treated HCC. Blue, DAPI; green, MRE11; red: BANF1. Scale bars, 10 μm. E, IP assay detection of MRN interactions in oxaliplatin-treated HCC cells with BANF1 overexpression. F, Prediction of the potential binding sites between MRE11 and BANF1 using Schrodinger software. Yellow, BANF1; cyan, MRE11. G, The effects of BANF1 site mutation on the interaction between BANF1 and MRE11 in the nuclear protein of oxaliplatin-treated HCC cells. H, The effects of BANF1 site mutation on the binding of MRE11 with RAD50 and NBS1 in the nuclear protein of oxaliplatin-treated HCC cells. I–K, Effects of BANF1 site mutation on the p-ATM and γ-H2AX ratios (n = 3), nuclear DNA TM (n = 30), and γ-H2AX–positive cells (n = 5) in oxaliplatin-treated HCC cells. Blue, DAPI; green, γ-H2AX. Scale bars, 50 μm. P, two-way ANOVA. **, P < 0.01; ***, P < 0.001.

Using Schrödinger software, we obtained the binding interfaces of MRE11–BANF1, including BANF1 Thr3/MRE11 Phe398, BANF1 Glu83/MRE11 Leu274, BANF1 Asp86/MRE11 His63, and BANF1 Lys54/MRE11 Glu64 (Fig. 6F; Supplementary Table S7). Molecular dynamics simulations revealed binding energies of −54.2 (T3A), −51.5 (E83A), −45.3 (D86A), and −56.2 kcal·mol⁻¹ (K54A)(Supplementary Table S8). BANF1 mutants (T3A/K54A/E83A/D86A, E83A/D86A) were generated to examine the role of nuclear BANF1 in MRN assembly. WT-BANF1 enhanced the nuclear BANF1–MRE11 interaction relative to the control vector (Fig. 6G), whereas T3A/K54A/E83A/D86A and E83A/D86A mutants significantly attenuated this binding. WT-BANF1 also promoted nuclear MRN complex formation, whereas the mutants impaired it (Fig. 6H). Therefore, nuclear BANF1 binding to MRE11 facilitates MRN assembly and DDR in HCC cells, with Thr3, Lys54, Glu83, and Asp86 serving as key interface residues.

To verify the regulatory role of nuclear BANF1 in DDR, we separately transfected control, WT-BANF1, T3A/K54A/E83A/D86A, or E83A/D86A plasmids into HCC cells to examine MRN downstream effectors (p-ATM, p-CHK2, γ-H2AX). WT-BANF1 increased p-ATM and p-CHK2 ratios and decreased γ-H2AX levels, whereas both mutants exerted the opposite effects (Fig. 6I; Supplementary Fig. S6D). Comet assays and IF staining confirmed that WT-BANF1 reduced comet TM and γ-H2AX–positive cell ratio, whereas the mutants increased both (Fig. 6J and K). Accordingly, nuclear BANF1 can effectively enhance DDR in HCC cells by modulating MRN complex formation.

Validation of SDC4 and PKD2 expressions, BANF1 nuclear localization in HCC tissues after oxaliplatin-based TACE therapy

Clinical data and tumor tissues of 17 patients with HCC, who underwent tumor radical resection within 3 months following TACE with oxaliplatin, were collected. As increased matrix stiffness primarily attributes to excessive deposition and cross-linking of ECM proteins (52), the expression of COL1A1 and LOX can be used to indicate tissue stiffness. Based on their median expression, patients were stratified into low (COL1A1^Low/LOX^Low, n = 8) and high (COL1A1^High/LOX^High, n = 9) matrix stiffness groups (Fig. 7A). High matrix stiffness was positively associated with reduced tumor shrinkage after TACE (Fig. 7B), consistent with our animal data. Patients in the high-stiffness group also exhibited significantly shorter overall survival and disease-free survival (Fig. 7C) and lower apoptotic indices (Fig. 7D), supporting that matrix stiffness increases oxaliplatin chemoresistance. Thus, impaired oxaliplatin response contributes to poor prognosis in patients with HCC with high matrix stiffness.

Figure 7.

Validation of SDC4 and PKD2 expressions and BANF1 nuclear localization in HCC tissues after TACE therapy with oxaliplatin. A, Continuous paraffin sections analysis of LOX and COL1A1 expressions in HCC tissues after TACE therapy with oxaliplatin (COL1A1^Low/LOX^Lown = 8; COL1A1^High/LOX^Highn = 9). Scale bars, 100 μm. B, The change in tumor size of patients with HCC before and after TACE therapy with oxaliplatin in the low- and high-matrix-stiffness groups. P, unpaired t test. C, Kaplan–Meier analysis of the overall survival (OS) curves and DFS curves of patients with HCC in the low- and high-matrix-stiffness groups. P, Log-rank test. D–F, TUNEL staining of apoptotic cells and expression levels of SDC4 and PKD2 in HCC tissues in high- and low-matrix-stiffness groups. Scale bars, 50 μm. P, unpaired t test. G, The relationship of SDC4 and PKD2 expressions between high- and low-matrix-stiffness groups. H, The nuclear localization expression of BANF1 in HCC tissues in the high- and low-matrix-stiffness groups. Scale bars, 50 μm. P, unpaired t test. *, P < 0.05; ****, P < 0.0001. AOD, average optical density; DFS, disease-free survival.

Additionally, SDC4 and PKD2 were highly expressed in the high-matrix-stiffness group (Fig. 7E and F), with elevated coexpression (Fig. 7G). Nuclear BANF1 was also significantly increased in the high-stiffness group (Fig. 7H), confirming that matrix stiffness promotes BANF1 nuclear translocation.

Discussion

The clinical challenge of cytotoxic drug resistance is still effectively unresolved. Therefore, it is of great significance to explore the mechanism of DDR from other new perspectives. Although increased matrix stiffness exhibits a potential promotion effect on DDR (53), its underlying mechanisms remain largely uncharacterized. Integrin β1/α5 and Piezo1, as HCC mechanosensors, have been validated to mediate matrix stiffness–induced EMT and angiogenesis (14, 15). However, apart from these two mechanosensors, little is known about whether there are other novel mechanosensors in HCC. Relying on the structural and functional characteristics of mechanosensors and their relevant research tips in HCC, combined with the new technologies of multiscale mechanical research, we developed a standard procedure for screening and identifying novel mechanosensors in membrane proteins of HCC cells. We utilized a quantitative proteomics technology to screen stiffness-dependent differential membrane proteins between the high- and low-stiffness groups, and then applied bioinformatics tools and literature prompts to determine receptor/adhesion molecule membrane proteins and calcium channel membrane proteins among them. Of the five calcium channel proteins obtained and four receptors/adhesion molecules, the leading differential membrane proteins SDC4 and PKD2 were selected as the target proteins of interest. Subsequently, their stiffness-dependent expressions were further corroborated in different stiffness substrates and a 3D hydrogel culture system. Importantly, multiscale indirect and direct mechanical experiments were performed to clearly characterize the function of PKD2 and SDC4 in sensing mechanical signals and mechanical signal transmission, suggesting that PKD2 and SDC4 can serve as two novel HCC mechanosensors to transmit mechanical signals.

As indicated by our data and that from other teams (14, 15, 30, 54), different cell types are likely to have different dominant mechanosensors, and the same mechanosensors are differentially expressed in different cell types. Therefore, the same mechanosensors are unlikely to play the same dominant role in all cell types, and the leading mechanosensors may vary among different types of tumor cells or normal cells. Other experiment contexts also address the mechanotransduction roles of SDC4 and PKD2. For example, SDC4 acts as a mechanosensor in pancreatic stellate cells (55), and PKD2 mediates mechanical stimulation–caused calcium ion influx in cilia (56). However, the biomechanical transduction roles of SDC4 and PKD2 in tumor cells, especially in HCC cells, have never been reported before. Moreover, there is currently no report on the collaborative regulation of SDC4 and PKD2 in matrix stiffness–promoted DDR. Accordingly, SDC4 and PKD2 can be considered two new HCC mechanosensors to explain matrix stiffness–mediated effects on DDR in HCC.

SDC4 and PKD2, two different types of functional proteins, regulate cellular biological behaviors/functions through different pathways. SDC4 activates FAK signaling to promote focal adhesion formation and cytoskeletal reorganization (57), whereas PKD2 causes calcium influx to influence the dynamic remodeling of actin (58). Notably, the SDC4-FAK pathway and PKD2-calcium signaling jointly modulate YAP nuclear translocation. Specifically, SDC4 promotes YAP nuclear translocation by inhibiting its phosphorylation. However, calcium signaling may indirectly influence YAP activity through the activation of LATS1/2 (59). The cross-talk between different pathways indicates that there may exist a collaborative regulatory mechanism in a high-stiffness microenvironment for HCC cells undergoing proliferation, invasion, and drug resistance. We speculate that SDC4- and PKD2-mediated different mechanosensory pathways have collaborative effects on DDR. Here, we presented sufficient evidence to validate that increased matrix stiffness prominently enhanced DDR and disclosed a novel cross-talk between two mechanosensory pathways (SDC4 and PKD2) in matrix stiffness–enhanced DNA repair in HCC. Our findings support that the cross-talk between different mechanosensory pathways may be a key regulatory mechanism for HCC cells in a high-stiffness microenvironment. A positive feedback loop between ITGB1 and Piezo1, described previously by our group (15), also favors this viewpoint.

As DDR occurs in the nucleus, we hypothesized that increased matrix stiffness may result in the nuclear translocation of specific proteins and then improve DDR. By a proteomics approach, CELLO analysis, and literature tips, we determined BANF1 as the target protein that might undergo nuclear translocation and participate in DDR in HCC cells. Subsequently, further experiments testified that SDC4 and PKD2 collaboratively modulated matrix stiffness–increased BANF1 nuclear translocation by increasing PP2A assembly and activity. There are currently five types of DDR in eukaryotic cells, including nucleotide excision repair, base excision repair, mismatch repair, homologous recombination repair (HRR), and nonhomologous end joining (NHEJ; ref. 60). Among them, HRR and NHEJ are the main repair methods for DNA DSBs. DNA DSB is the most prevalent form of oxaliplatin-induced DNA damage (60). HRR is considered a highly accurate method for repairing DSBs, and the MRN complexes serve as pivotal proteins in HRR. MRE11, a core component of the MRN complex, plays an essential role in the repair of DSBs (61). Once a DSB occurs, NBS1, MRE11, and RAD50 are recruited to DSB sites to form MRN complexes (62). MRE11 exhibits nuclease activity and processes the broken DNA ends, thereby initiating the HRR pathway (63). To clarify the contribution of nuclear BANF1 to DDR, we used IP-MS/MS and an intersection analysis to determine that MRE11 interacted with nuclear BANF1, indicating a potential linkage between nuclear BANF1 and the MRN complex. By site mutational experiments, nuclear BANF1 binding to MRE11 obviously enhanced MRN complex assembly and strengthened DDR by increasing ATM and CHK2 phosphorylation. A study shows that BANF1 overexpression enhances HRR but suppresses NHEJ, ultimately inhibiting DDR in osteosarcoma (41). Despite the promotional role of BANF1 being corroborated in HRR, its conclusion differs from our findings, which support that BANF1 nuclear entry, rather than its expression level, contributes to DDR. Overall, a new collaborative regulation between two different mechanosensory pathways has been established in matrix stiffness–promoted DDR. Additionally, another finding shows that SDC4 and PKD2 were upregulated in normal liver cells grown on a high-stiffness substrate, indicating that the same mechanisms may also occur in radiation-induced liver injury, which provides a new direction for subsequent research.

In addition to the regulatory pathway described above, there are other mechanistic pathways to explain matrix stiffness–enhanced drug resistance. One mechanistic pathway is the physical barrier, namely, stiffness increases cell aggregation, which forms a physical barrier to protect the inner cancer cells from exposure to the drugs (64). Another mechanistic pathway also belongs to the biomechanical signaling pathway; namely, CNN1^High cancer-associated fibroblast (CAF)–caused matrix stiffening contributes to 5-fluorouracil resistance in gastric cancer cells via activating YAP (11). However, the initiating factor of this study is a paracrine effect of CAFs; moreover, the activation of the YAP pathway depends on a mechanosensory pathway alone, which is different from the proposed regulatory pathway in this study. Therefore, there are multiple pathways to enhance drug resistance in a high matrix stiffness microenvironment, and from the two perspectives of the biomechanical regulatory pathway and physical barrier, it seems more comprehensive to explain matrix stiffness–enhanced drug resistance.

To date, most studies support that high matrix stiffness increases DDR in tumors, including cervical cancer (65), HCC (66, 67), and non–small cell lung cancer (64). Moreover, the promoting effect of high stiffness on cellular DSB repair may be shared among different cell types (24). However, the types of DDR mediated by matrix stiffness may be inconsistent in the above studies. Therefore, although high matrix stiffness mainly improved the efficiency of HRR in the study, other repair types for DSB (NHEJ) may still exist and be effective. Intriguingly, individual studies in 3D culture conditions have shown opposite findings. Organoids derived from breast cancer cells interacting with soft ECM increase resistance to chemotherapy and ionizing radiation (68). Additionally, nasopharyngeal carcinoma cells in low-stiffness hydrogel also exhibited striking radioresistance (69). These discrepancies for certain tumor cells in radioresistance may be associated with different cell culture methods. Additionally, cell–cell interaction and physical barriers may also have a certain impact on chemoradiotherapy resistance.

This study is subject to several limitations. The first is the survival analysis of patients with HCC in the small cohort size. Because most HCC patients receiving TACE treatment are in the middle and late stages, only a few achieved good conversion results after TACE treatment who met the surgical indications and standards. Therefore, it is very difficult to obtain tumor tissue samples in the clinic from these patients. Second, in addition to the proposed regulatory pathway, other DNA repair pathways still cannot be excluded, which merits further exploration in the future.

Conclusion

This study successfully developed a standard procedure for screening and identifying novel mechanosensors in membrane proteins and obtained two novel HCC mechanosensors (SDC4 and PKD2). Simultaneously, it also uncovered a novel cross-talk between two mechanosensory pathways in matrix stiffness–enhanced DDR in HCC. Specifically, SDC4 and PKD2 collaboratively modulated matrix stiffness–increased BANF1 nuclear translocation by enhancing PP2A assembly and activation, and nuclear BANF1 increased the formation of the MRN complex and its downstream effectors’ phosphorylation to improve DDR. To the best of our knowledge, this is the first report that two different mechanosensory pathways collaboratively promote DDR in HCC.

Supplementary Material

Supplementary Table 1

All sequences and overexpression/knockdown fragments.

Supplementary Table 2

6kPa、16kPa Col1-coated polyacrylamide gel stiffness substrate formulation table.

Supplementary Table 3

Primer sequence.

Supplementary Table 4

Calcium ion channel gene set.

Supplementary Table 5

A list of candidate differential membrane proteins.

Supplementary Table 6

The top 20 candidate differential nuclear proteins the between high- and low-stiffness groups.

Supplementary Table 7

The Schrödinger molecular docking software simulated the potential binding sites of BANF1 and MRE11.

Supplementary Table 8

The Schrödinger molecular docking software simulates the binding energy changes when the mutation is to alanine.

Supplementary Fig. 1

Screening and identification of novel mechanosensors in HCC.

Supplementary Fig. 2

Confirming the formation of mouse models with high liver stiffness.

Supplementary Fig. 3

SDC4 and PKD2 synergistically participate in matrix stiffness-promoted DNA damage repair in HCC cells.

Supplementary Fig. 4

SDC4 and PKD2 synergistically mediate matrix stiffness-increased BANF1 nuclear translocation.

Supplementary Fig. 5

SDC4 and PKD2 synergistically regulate BANF1 dephosphorylation through PP2A enzyme assembly and activity in HCC cells.

Supplementary Fig. 6

BANF1 nuclear translocation enhances DNA damage repair in HCC cells.

Data Availability

The data generated in this study are available within the article and its Supplementary Files. The proteomic data generated in this study are publicly available in iProX at PXD081109. The ECM–receptor interaction gene set used in this study is derived from KEGG hsa04512 (https://www.gsea-msigdb.org/gsea/msigdb/cards/KEGG_ECM_RECEPTOR_INTERACTION). All other raw data are available from the corresponding authors upon request.

Authors’ Disclosures

No disclosures were reported.

Authors’ Contributions

J. Li: Data curation, software, formal analysis, validation, investigation, visualization, writing–original draft. Z. Ren: Data curation, formal analysis, validation, investigation. J. Yao: Resources, validation. Y. Zhao: Validation, investigation. Y. Xu: Investigation. H. Yu: Investigation. C. Huang: Validation. M. Wang: Validation. F. Yao: Validation. M. Li: Resources. Y. Wang: Resources. D. Gao: Resources. Y. Zhao: Resources. B. Liu: Resources. Z. Wang: Resources. Y. Dong: Resources, funding acquisition, investigation. R. Chen: Conceptualization, resources, data curation, project administration, writing–review and editing. J. Cui: Conceptualization, resources, funding acquisition, writing–review and editing, project administration.