Targeting the ANXA2-CD63-exosomal PD-L1 axis in hepatocellular carcinoma: A novel mechanism of immune evasion and its implications for precision immunotherapy

Abstract

Background and Purpose

Hepatocellular carcinoma (HCC) remains one of the leading causes of cancer-related deaths globally due to late diagnosis and resistance to therapies, including immune checkpoint blockade (ICB). Tumor-derived exosomes carrying programmed death-ligand 1 (PD-L1) have emerged as key mediators of immune evasion by binding PD-1 on T cells, thereby inducing T cell exhaustion and systemic immunosuppression. This study investigated whether Annexin A2 (ANXA2) modulates exosomal PD-L1 expression in HCC cells to inhibit T-cell function and promote immune escape.

Results

Bioinformatic analyses (TIMER2.0, TCGA) showed ANXA2 upregulation in HCC, correlating with fewer CD8⁺ T cells and poor survival. ANXA2 knockout (KO) in Hepa1-6 cells reduced exosome secretion and exosomal PD-L1, confirmed by EM, nanoparticle tracking, and WB. Exosomes from controls suppressed CD8⁺ T cell activation (reducing CD69, IL-2, IFN-γ, TNF-α), while KO exosomes did not. ANXA2 KO tumors grew slower in immunocompetent C57BL/6 J mice but not in immunodeficient BALB/c Nude mice, highlighting immune dependence. ANXA2 depletion reduced CD63 protein stability without affecting mRNA. CD63 re-expression in KO cells restored exosomal PD-L1, indicating ANXA2 maintains CD63 stability for PD-L1 exosomal incorporation. Transcriptomic and pharmacologic evidence further supported a role for lysosome-associated turnover in CD63 loss upon ANXA2 depletion.

Conclusions

ANXA2 drives HCC immune evasion by upregulating exosomal PD-L1 via CD63 stabilization, offering a novel target to enhance ICB efficacy. This work highlights a tumor-specific exosome regulation mechanism, with potential implications for immunotherapy across exosome-dependent cancers.

Introduction

Hepatocellular carcinoma (HCC), a leading cause of cancer-related mortality globally, is characterized by late-stage diagnosis and therapeutic resistance, which contribute to its dismal prognosis [1]. While immune checkpoint blockade (ICB) targeting the PD-1/PD-L1 axis has provided clinical benefits for advanced HCC patients, objective response rates remain below 20%, with most patients developing resistance due to immune evasion mechanisms [2,3]^. The immunosuppressive tumor microenvironment (TME) is a central determinant of ICB resistance, and exosomes—30–150 nm lipid bilayer vesicles released via multivesicular body (MVB)-plasma membrane fusion—have emerged as critical mediators of intercellular communication, capable of remodeling the TME by transferring immunosuppressive cargos such as PD-L1 [[4], [5], [6]]. Exosomal PD-L1 binds PD-1 on T cells, driving T cell exhaustion and blunting antitumor immunit^y [7,8]^. In melanoma and lung cancer, tumor-derived exosomal PD-L1 not only suppresses local T cell activity but also impairs systemic immune surveillance by targeting distal lymph nodes [9,10]^. In HCC, elevated plasma exosomal PD-L1 correlates with anti-PD-1 resistance and poor survival, yet the molecular machinery governing its selective sorting and secretion remains poorly defined [11,12]. Although the ESCRT complex and tetraspanins have been implicated in exosomal PD-L1 packaging, their ubiquitous roles in diverse physiological processes limit their therapeutic targeting potential [13]_. Thus, identifying key regulators of exosomal PD-L1 sorting and dissecting their roles in HCC immune evasion represent urgent unmet needs.

Annexin A2 (ANXA2) is a calcium-dependent phospholipid-binding protein that plays multifaceted roles in tumor progression, metastasis, and therapy resistance [14,15]. In HCC, ANXA2 overexpression is strongly associated with aggressive phenotypes (vascular invasion, lymph node metastasis) and reduced patient survival [[16], [17], [18]]. Mechanistically, ANXA2 drives epithelial-mesenchymal transition (EMT) via PI3K/AKT/mTOR pathway activation and promotes angiogenesis by regulating VEGF secretion [15]. It is implicated in angiogenesis and the regulation of the tissue-type plasminogen activator (tPA) system, which catalyzes the conversion of plasminogen to plasmin, facilitating the degradation of the extracellular matrix (ECM) and thus promoting angiogenesis [19,20]. This process is crucial for enhancing the migratory and invasive capabilities of cancer cells. Furthermore, ANXA2 has been linked to the development of multidrug resistance, presenting challenges in cancer therapeutics. The role of ANXA2 in modulating the relationship between overexpression and proteolytic cleavage in various malignancies has been established. ANXA2 also participates in the degradation of ECM required for angiogenesis, tumor expansion, and invasion [21]^. Given its involvement in these processes, ANXA2 is considered a potential therapeutic target for the treatment of invasive breast cancer [22]. Additionally, ANXA2 can translocate to the nucleus, where its phosphorylated form can prevent DNA damage caused by radiotherapy toxicity. However, its role in shaping the HCC immune landscape remains unexplored.

Current studies on exosomal PD-L1 regulation focus on the ESCRT complex and endosomal trafficking pathways [4,23]. For instance, the ESCRT-I component TSG101 recruits PD-L1 into intraluminal vesicles (ILVs), while Rab5 microdomains stabilize PD-L1 on exosomal surfaces [24,25]. However, these mechanisms lack tissue or disease specificity and fail to explain why certain tumors exhibit markedly higher exosomal PD-L1 levels [26,27]. Based on this evidence, we hypothesize that ANXA2 orchestrates the immunosuppressive TME in HCC by governing exosomal PD-L1 sorting and secretion, thereby driving ICB resistance. This study investigates whether ANXA2 directly regulates PD-L1 enrichment in exosomes and elucidates the molecular mechanisms by which the ANXA2-PD-L1 axis suppresses CD8⁺T cell function. The innovation of this study lies in identifying ANXA2 as a disease-specific regulator of exosomal PD-L1, bypassing the limitations of ESCRT-centric mechanisms and offering a novel therapeutic avenue for HCC. By delineating the spatiotemporal dynamics of the ANXA2-PD-L1 axis, this work may yield prognostic biomarkers and personalized strategies to overcome ICB resistance in HCC patients.

Methods

Cell Culture

Human HCC cell lines (e.g.,JHH7) and a murine HCC cell line (e.g., Hepa1-6) were obtained from the National Cell Resource Center. All cell lines were authenticated by short tandem repeat (STR) profiling. Human cells were maintained in Roswell Park Memorial Institute (RPMI) 1640 medium (01–100-1ACS, BI, Israel) supplemented with 10% (v/v) fetal bovine serum (FBS) (01–052-1, BI, Israel) and cultured in a humidified incubator at 37 °C with 5% CO₂. Hepa1-6 cells were cultured in Dulbecco’s modified Eagle medium (DMEM) (01–051-1ACS, BI, Israel) supplemented with 10% FBS.

Establishment of ANXA2 KO Cell Lines

Single guide RNA (sgRNA) oligonucleotides targeting human or mouse ANXA2 were cloned into U6-sgRNA-EF1a-Cas9-FLAG-P2A-puro (for human cells) or U6-sgRNA-SFFV-Cas9-FLAG-P2A-mCherry (for mouse cells) and packaged as lentiviruses (Genechem, Shanghai, China). The lentiviruses were then transduced into human HCC cell lines (e.g., HepG2, Huh7) and murine HCC Hepa1-6 cells following a previously published protocol [22]. The culture medium was replaced 12 h post-infection, and transduction efficiency was assessed after 72 h. Single clones were isolated by dilution, and knockout (KO) clones were confirmed by Western blot analysis. The ANXA2 sgRNA sequences were as follows: for human cells, 5′-ATTATATCCAGGTAAGCCCG-3′; for mouse cells, 5′-CGGGCTTACCTGGATATAAT-3′ (sequences are provided as examples).

Lentiviral transduction for CD63 overexpression

To restore CD63 expression in ANXA2-knockout cells, the full-length cDNA sequence of murine Cd63 was chemically synthesized and cloned into the pLV-Puro lentiviral vector (Genechem, Shanghai, China). The recombinant plasmid (plv-CD63) or the empty vector control (plv-EV) was co-transfected with packaging plasmids into HEK293T cells to generate lentiviral particles. ANXA2-KO Hepa1-6 cells were infected with the viral supernatants in the presence of polybrene (8 μg/mL). Forty-eight hours after infection, stable cells were selected using puromycin (2 μg/mL) for 7 days. The overexpression efficiency of CD63 was confirmed by Western blot analysis prior to functional assays.

Isolation of exosomes by ultracentrifugation

HCC cells were cultured in a complete medium for 24 h, then switched to a 10% exosome-depleted FBS medium for an additional 36 h. The conditioned medium was collected and centrifuged at 5,000 × g for 30 min to remove cell debris, followed by a 10,000 × g centrifugation for 30 min to eliminate larger vesicles. Exosomes were pelleted by ultracentrifugation at 100,000 × g for 2 h, washed in phosphate-buffered saline (PBS), and ultracentrifuged again at 100,000 × g for 2 h. The final pellet was resuspended in PBS, filtered through a 0.22 μm filter, and stored at –80°C. Exosome concentrations were quantified using a bicinchoninic acid (BCA) assay kit (PC0020, Solarbio, China).

Isolation of exosomes by sucrose gradient centrifugation

Exosomes obtained via differential ultracentrifugation were further purified using sucrose gradient fractionation as previously described with modifications [28]. Briefly, exosome preparations were centrifuged at 100,000 × g for 2.5 h in gradients of 10–16%, 22–28%, 34–40%, 46–52%, 58–64%, and 70–82% sucrose solutions. Fractions were diluted 1:100 in PBS and centrifuged at 100,000 × g for another 2.5 h. The resulting pellets were resuspended in PBS for subsequent Western blot analysis using antibodies against exosomal markers (anti-CD63, ab59479, Abcam, USA; anti-ALIX, 2171S, CST, USA).

Western blot analysis

Total protein was extracted using RIPA lysis buffer supplemented with protease and phosphatase inhibitors. Lysates were incubated on ice and clarified by centrifugation at 12,000 × g for 15 min at 4°C. Protein concentration was determined using a BCA assay kit. Equal amounts of protein (typically 20 µg per lane) were separated by 10–12% SDS–PAGE and transferred onto PVDF membranes. Membranes were blocked with 5% non-fat milk in TBST for 1 h at room temperature and incubated with primary antibodies overnight at 4°C. The following primary antibodies were used: anti-ANXA2 (Cell Signaling Technology, CST; catalog #8235; 1:1000), anti-CD63 (Abcam; ab59479; 1:1000), anti-ALIX/PDCD6IP (CST; #2171; 1:1000), anti-PD-L1 (Proteintech; 28076-1-AP; 1:1000) and anti-β-actin (Proteintech; 66009-1-Ig; 1:5000). After washing with TBST, membranes were incubated with HRP-conjugated secondary antibodies for 1 h at room temperature: goat anti-rabbit IgG-HRP (CST; #7074; 1:5000) and goat anti-mouse IgG-HRP (CST; #7076; 1:5000). Signals were detected using enhanced chemiluminescence (ECL) reagents and imaged with a gel documentation system. Band intensities were quantified using ImageJ, and target proteins were normalized to β-actin.

RNA extraction and qRT–PCR

Total RNA was extracted from cells using TRIzol reagent (or an equivalent phenol–chloroform method) according to the manufacturer’s instructions. RNA concentration and purity were assessed by spectrophotometry (A260/280). cDNA was synthesized from 0.5 to 1 µg of total RNA using a reverse transcription kit. Quantitative real-time PCR (qRT–PCR) was performed using SYBR Green master mix on a real-time PCR system under standard cycling conditions (e.g., 95°C for initial denaturation followed by 40 cycles of 95°C and 60°C). Melt-curve analysis was used to confirm amplification specificity. Relative expression levels were calculated using the 2^−ΔΔCt method and normalized to GAPDH.

The primer sequences used for qRT–PCR were as follows:

• Human CD63: forward 5′-CAACCACACTGCTTCGATCCTG-3′,

• reverse 5′-GACTCGGTTCTTCGACATGGAAG-3′;

• Mouse Cd63: forward 5′-GGAATCCACTATCCATACCCAGG-3′,

• reverse 5′-CTCTTCACCAGACAGCAGGAGA-3′;

• Human GAPDH: forward 5′-GTCTCCTCTGACTTCAACAGCG-3′,

• reverse 5′-ACCACCCTGTTGCTGTAGCCAA-3′;

• Mouse Gapdh: forward 5′-CATCACTGCCACCCAGAAGACTG-3′,

• reverse 5′-ATGCCAGTGAGCTTCCCGTTCAG-3′.

Nanoparticle tracking analysis

Exosome size distribution and concentration were analyzed using Flow NanoAnalyzer (Flow Nano Biotechnology Co., Ltd., China) equipped with nanoscale flow cytometry (NanoFCM). Samples were filtered through 0.22 μm membranes and diluted 100-1000 fold in PBS to optimize detection parameters. Data were processed using NanoFCM analytical software.

Transmission electron microscopy (TEM)

Approximately 30 μL of exosome suspension was placed onto a carbon-coated copper grid and allowed to sit for 2–5 minutes. Excess liquid was blotted using filter paper, and the grid was dried at room temperature for 10 minutes. Subsequently, a droplet of uranyl acetate was applied for staining, left for 90 seconds, and then excess dye was removed. The grid was dried for an additional 3 hours prior to observation under an electron microscope.

Detection of exosomal PD-L1 by flow cytometry

For murine EVs (e.g., Hepa1-6–derived), exosomes from control and ANXA2 knockout (KO) cells were incubated with mouse CD63 capture beads (Mouse CD63 Capture Beads for Flow Detection, Immunostep; MO63CB-25) overnight at room temperature in the dark. Bead-bound exosomes were washed twice with PBS and then stained with PE-conjugated rat anti-mouse PD-L1 (CD274) antibody (BD Pharmingen; 568085, clone 10F.9G2; used at an optimized dilution) for 40 min at 4°C. A matched PE rat IgG2b, κ isotype control (BD Pharmingen; 555848, clone R35-38) was included to define background fluorescence. After additional washes, samples were acquired on a BD flow cytometer and analyzed by gating on the bead population.

For human HCC cell–derived exosomes, exosomes were captured using CD63 exosome capture beads (Abcam; ab239686) under the same conditions and stained with PE-conjugated anti-human PD-L1 antibody (BD; 557924; 1:10) for 40 min at 4°C, with PE-conjugated IgG isotype control (BD; 556650) as the negative control. Following final washes, samples were analyzed by flow cytometry with identical acquisition settings across groups.

Exosome PD-1/PD-L1 binding assay

High protein-binding 96-well plates were coated with human PD-1-Fc fusion protein (5 μg/mL, 10377-H02H, SinoBiological, China) and incubated at 37°C for 1 hour. Wells coated with bovine serum albumin (BSA, 5 μg/mL, A8020, Solarbio, China) were used as negative controls. After blocking with 5% BSA for 2 hours at room temperature, PKH26-labeled exosomes (labeled using MINI26-1KT, Sigma, Germany) were added and incubated at room temperature for 2 hours. Following thorough washing, the binding was visualized using fluorescence microscopy.

Primary mouse T-cell isolation and activation

Spleens and lymph nodes were harvested and mechanically dissociated through 70 µm strainers into cold RPMI-1640 supplemented with 10% heat-inactivated FBS, 1% Penicillin-Streptomycin (Gibco, 15140-122), 2 mM l-glutamine or GlutaMAX (Gibco, 35050-061), and 50 µM 2-mercaptoethanol (Gibco, 21985-023). Red blood cells were removed with ACK lysis buffer (Gibco, A10492-01). T cells were enriched by negative selection (standard mouse pan-T isolation kit) and counted for downstream assays.

Murine T cells were activated with Dynabeads™ Mouse T-Activator CD3/CD28 (Thermo Fisher Scientific, 11456D) at a 1:1 bead:cell ratio in complete RPMI + 10% FBS, supplemented with recombinant mouse IL-2 (PeproTech, 212-12, 50–100 IU/mL) for 24 h at 37°C, 5% CO₂; beads were then magnetically removed. Before staining, cells were blocked with anti-mouse CD16/32 (Fc block) (BioLegend, 101320) for 10 min at room temperature.

Flow cytometry

Live/dead discrimination used Zombie Aqua™ Fixable Viability Dye (BioLegend, 423101 or 423102). Surface markers (representative panel) included CD3ε-APC (clone 145-2C11; BioLegend, 100311/100312), CD8α-PE-Cy7 (clone 53-6.7; BioLegend, 100721/100722), and CD69-PE (clone H1.2F3; BioLegend, 104508; optional PE-Cy7 104512). Staining followed manufacturer instructions; compensation used single-stained controls.

Fluorescent labeling of exosomes and EV–T cell binding assay

Exosomes were fluorescently labeled with a green lipophilic membrane dye, DiO (Vybrant™ DiO Cell-Labeling Solution, Invitrogen, Thermo Fisher Scientific; Cat. No. V22886). DiO is supplied as a 1 mM stock solution in dimethylformamide (DMF). Exosomes were resuspended in Diluent C (or PBS) and incubated with DiO at a 1:200 dilution (5 μL DiO stock per 1 mL exosome suspension; final dye concentration, 5 μM) for 20 min at 37°C in the dark with gentle mixing. The labeling reaction was quenched by adding an equal volume of 1% BSA (or exosome-depleted FBS). Free dye was removed by ultracentrifugation at 100,000 × g for 70 min, followed by one additional PBS wash and a second ultracentrifugation step (100,000 × g for 70 min). The final pellet was resuspended in PBS and quantified by NanoFCM. For EV–T cell binding assays, activated murine T cells were incubated with particle-number-normalized DiO-labeled exosomes (1 × 10⁹ particles/mL) at 37°C for 2 h. Cells were then washed thoroughly with cold PBS containing 1% BSA and stained for surface markers and viability dye. EV binding/uptake was quantified by flow cytometry as the DiO mean fluorescence intensity (MFI) within live CD3⁺ (or CD8⁺) T cells.

Cytokine measurements

Supernatant IL-2, IFN-γ, and TNF-α were quantified using ELISA MAX™ Deluxe Sets (BioLegend; 431004/430804/430904) following kit manuals; absorbance was read at 450 nm with 570 nm reference, and concentrations were interpolated from 4-PL standard curves.

Immunofluorescence

Cells were washed with PBS and fixed in 4% neutral formaldehyde for 20 minutes at room temperature. After three washes with PBS, primary antibodies targeting PD-L1, ANXA2, and CD63 were applied overnight at 4°C. Cells were then washed and incubated with Alexa Fluor 488–conjugated anti-rabbit or Alexa Fluor 594–conjugated anti-mouse secondary antibodies (Thermo Fisher, USA) for 2 hours at room temperature. Nuclei were counterstained with DAPI (100 ng/mL), and coverslips were mounted using an anti-fade reagent. Images were acquired using a Leica TCS SP8 or Leica STELLARIS 8 STED confocal microscope and analyzed with Imaris software (Bitplane).

Immunohistochemistry Staining

Immunohistochemistry staining was performed on tumor tissues fixed in 10% formalin, embedded in paraffin, and sectioned at 5 μm. Sections underwent deparaffinization, rehydration, and antigen retrieval using EDTA or citrate buffer. After blocking with 5% goat serum, sections were incubated overnight at 4°C with primary antibodies. Subsequent incubation with peroxidase-conjugated secondary antibodies (Abcam, USA) was carried out for 2 hours at room temperature, followed by DAB staining (Abcam, USA). Slides were scanned using a Leica Aperio VERSA 8 (Leica Biosystems, Germany) and analyzed with Aperio image analysis software.

EV dose normalization in functional assays

For T-cell functional assays, exosomes were quantified by particle number using NanoFCM and added to cultures at a particle-number–normalized dose (equal particle numbers across groups) to compare per-particle potency rather than total EV yield. Unless otherwise indicated, exosomes were applied at 1 × 10⁹ particles/mL (NanoFCM-determined). Exosome protein amounts measured by BCA were reported as a reference but were not used for normalization in these in vitro assays. For in vivo experiments, exosome dosing was normalized by total exosomal protein mass (µg) per injection (see Animal Studies), and the corresponding particle numbers were measured by NanoFCM and recorded to facilitate reproducibility.

Animal studies

Six- to eight-week-old male mice were used in all experiments. C57BL/6 J and BALB/c nude mice were purchased from Cyagen Biosciences Inc. (Suzhou, China). Mice were housed in a specific pathogen-free (SPF) facility (12:12 h light–dark cycle) and acclimated for 1 week before experiments. Mice were randomized to groups by a computer-generated sequence at enrollment. To establish subcutaneous tumors, 2.5 × 10⁵ HCC cells (control or ANXA2 KO) in 100 μL sterile PBS were injected into each mouse (n = 6 per group). Tumor volume was measured every 3 days using digital calipers (volume = length × width² × 0.5). In in vivo exosome studies, when tumor volume reached approximately 100 mm³, mice were treated with 20 μg of exosomes via intratumoral injection twice weekly for 21 days. Tumors were then excised and weighed.

All survival procedures were performed under isoflurane inhalational anesthesia. Analgesia (buprenorphine and/or meloxicam) was provided as needed. Mice were monitored daily and euthanized at predefined humane endpoints (excess tumor burden, weight loss, or signs of distress). Euthanasia was by CO₂ inhalation followed by cervical dislocation to ensure death. All procedures complied with AVMA 2020 guidelines and were approved by the institutional IACUC.

Animal experiments were designed and reported in accordance with ARRIVE 2.0 guidelines. Details regarding study design, sample sizes, randomization, blinding, inclusion/exclusion criteria, outcome measures, and statistical methods are provided in the Methods. A completed ARRIVE checklist has been uploaded as Supplementary File 1.

Statistical analysis

Data were analyzed using GraphPad Prism 9 (GraphPad Software, USA). Data are presented as mean ± SD from at least three independent experiments, unless otherwise stated. For two-group comparisons, a two-tailed unpaired Student’s t-test was used. For comparisons among three or more groups, one-way ANOVA followed by Tukey’s multiple-comparisons test was applied. Tumor growth curves and cell proliferation curves measured over time were analyzed by two-way repeated-measures ANOVA (group × time) with appropriate multiple-comparisons correction; Normality and homogeneity of variance were assessed; when assumptions were not met, nonparametric tests were used (Mann–Whitney test for two groups, or Kruskal–Wallis test with Dunn’s multiple-comparisons test for multiple groups). P < 0.05 was considered statistically significant. Co-localization analyses were performed using ImageJ (NIH, USA).

Result

ANXA2 deficiency inhibits tumor progression through enhancing immune response in HCC

To investigate the role of ANXA2 in the progression of hepatocellular carcinoma (HCC), we performed a comprehensive analysis of ANXA2 expression across various cancer types using the TIMER2.0 platform. The results indicated that ANXA2 is upregulated in multiple cancer types, with a significant increase observed in HCC (Fig. 1A). Further analysis using The Cancer Genome Atlas (TCGA) database confirmed that ANXA2 levels were significantly elevated in tumor tissues compared to adjacent normal tissues (Fig. 1B and Fig. 1C). Patients with high ANXA2 expression levels had a notably reduced overall survival and disease free survival (Fig. 1D and E). Immunohistochemical (IHC) staining of a tissue microarray containing hepatocellular carcinoma (HCC) samples and matched adjacent normal tissues also revealed significantly elevated expression of ANXA2 in HCC tissues(Fig. 1F). To assess the in vivo relevance of ANXA2, we generated ANXA2-knockout (KO) Huh-7 and Hepa1-6 cells (Fig. 1G). ANXA2 loss did not measurably alter in vitro proliferation kinetics in either cell line (Fig. 1H). In immunocompetent C57BL/6 J mice, syngeneic implantation of Hepa1-6 ANXA2-KO cells resulted in significantly reduced tumor growth compared with control (CON) (Fig. 1I–K). In contrast, in an immunodeficient xenograft setting using Huh-7 cells in T cell–deficient BALB/c nude mice, tumor growth was not overtly different between WT and ANXA2-KO groups (Fig. 1L–N). While these models differ in both host immunity and tumor cell origin, the concordant lack of intrinsic proliferation changes in vitro together with the discrepant in vivo phenotypes is consistent with an immune-associated contribution to the tumor-suppressive effect observed upon ANXA2 loss in the syngeneic setting. Together, our findings suggest that ANXA2 may contribute to HCC progression through both tumor-intrinsic and immune-modulatory mechanisms.

Figure 1.

ANXA2 promotes HCC progression by suppressing T cell-mediated immune responses.

(A) ANXA2 expression levels in normal (N = 50) vs. tumor (T = 371) tissues from the TCGA-LIHC cohort, analyzed by one-way ANOVA. (B, C) Paired and unpaired analysis of ANXA2 expression in matched normal and tumor tissues from the TCGA database, further confirming its upregulation in HCC. Kaplan–Meier curves for the TCGA-LIHC cohort showing (D) overall survival (OS) and (E) disease-free survival (DFS) stratified by ANXA2 expression (high vs. low). Survival differences were assessed by the log-rank test. (F) Immunohistochemical staining of ANXA2 in a tissue microarray of HCC patient samples. Representative staining ANXA2 expression in primary HCC tissues and adjacent normal tissues (scale bar, 100 mm). (G) (H) Growth curves of Hepa1-6 and Huh7 cells in the presence or absence of ANXA2. (I, J, K) Tumor growth assessment in C57BL/6 J mice. At the experimental endpoint, xenograft tumor results were presented via photographs, growth curves, and weight charts. n = 5 mice per group. (L, M, N) Tumor growth assessment in immunodeficient BALB/c Nude mice. At the experimental endpoint, xenograft tumor results were presented via photographs, growth curves, and weight charts. n = 5 mice per group. Data represent the mean ± SD from three independent experiments. P values were determined by one-way ANOVA for group comparisons, P < 0.001 was considered statistically significant.

Knockdown of ANXA2 downregulates PD-L1 in HCC

To delineate the mechanism by which ANXA2 modulates antitumor immunity in hepatocellular carcinoma (HCC), we first interrogated the association between ANXA2 expression and immune‐cell features using the TIMER3.0 platform, which provides robust estimates of immune infiltration [36]. Across multiple immune subsets, ANXA2 showed a significant inverse correlation with infiltrating CD8⁺ T cells in HCC (Fig. 2A). Consistently, co-culture of HCC cells with anti-CD3/CD28 bead–activated T cells demonstrated that ANXA2 knockout increased T-cell–mediated killing of tumor cells (Fig. 2B). These findings implicate ANXA2 as a suppressor of T-cell responses that supports tumor growth.

Figure 2.

ANXA2 knockout downregulates PD-L1 and enhances T-cell-mediated cytotoxicity.

(A) Correlation analysis of ANXA2 with indicated immune cell infiltration in HCC by TIMER3.0. (B) Cell survival of hep 1-6 cells in the presence or absence of ANXA2, treated with anti-CD3/CD28-activated T cells. Scale bar = 40 μm. (C) Correlation analysis of ANXA2 with PD-L1 in HCC using TCGA database (n = 371). Expression of PD-L1 of contrl and ANXA2 knockout Hepa1-6 cell, mRNA (D), protein (E) and membrane PD-L1 (F). (G) Confocal images of PD-L1(red) and nucleus (blue) in the presence of ANXA2 or not. Scale bar =5 µm. (H)Expression and quantitative results of TSG101 in the cells presence or absence of ANXA2. The data are presented as the mean ± SD; n = 3; n.s, no significance, *P < 0.05; two-tailed unpaired Student’s t-test

Given that PD-L1 constrains T-cell activation via PD-1, we next examined its relationship with ANXA2. In TCGA-LIHC, ANXA2 expression showed a positive association with CD274 (PD-L1) mRNA. Given that TCGA profiles bulk tumor samples, we next examined PD-L1 regulation in controlled cell systems. Genetic ablation of ANXA2 in Hepa1-6 cells reduced PD-L1 abundance at both the mRNA and total protein levels. Because PD-L1 functions at the plasma membrane, we quantified surface PD-L1 and found no significant change upon ANXA2 loss (Fig. 2D-F). This reduction was observed not only in our primary model but was also recapitulated in human HCC cell lines (JHH7 and Huh-7), in which ANXA2 knockout decreased PD-L1 transcripts and protein, with corresponding changes in surface PD-L1 assessed by flow cytometry (Fig. S1A–F) We therefore investigated endosomal trafficking pathways governing PD-L1 synthesis, recycling, and degradation. Immunofluorescence analyses indicated that ANXA2 deletion enhanced intracellular recycling of PD-L1(Fig. 2G), and was accompanied by downregulation of the ESCRT-I component TSG101(Fig. 2H). Together, these results indicate that ANXA2 deletion lowers total PD-L1 while buffering surface PD-L1, consistent with altered endosomal trafficking marked by reduced ESCRT-dependent sorting into MVBs and exosomes and enhanced recycling, rather than generalized plasma-membrane loss. Consequently, extracellular (vesicular) PD-L1 is reduced, whereas membrane PD-L1 is preserved, providing a mechanistic link between ANXA2 and T-cell suppression in HCC.

ANXA2 knockout impairs exosome secretion and reduces functional exosomal PD-L1 in HCC

Previous reports have indicated that exosomal PD-L1 may also exert immunosuppressive functions. It is known that PD-L1 can be secreted via exosomes, a process that may be regulated by ANXA2. Based on this hypothesis, we isolated exosomes from liver cancer cells using differential ultracentrifugation and conducted scanning electron microscopy (SEM) and nano-flow cytometry analysis. The results demonstrated that exosomes derived from liver cancer cells were primarily sized between 40 and 120 nm (Fig. 3A and 3B). Western blot analysis revealed the presence of exosomal markers CD63 and ALIX, along with PD-L1, in exosomes from both control (CON) and ANXA2 knockout (ANXA2 KO) cells (Fig. 3C). However, PD-L1 levels were significantly reduced in exosomes from ANXA2 KO cells compared to those from control cells (Fig. 3D and Fig. 3E). Furthermore, the concentration of exosomes from ANXA2 KO cells was decreased, suggesting that the absence of ANXA2 could lead to a reduction in the secretion of exosomal PD-L1 (Fig. 3F). To ensure the functional integrity of exosomal PD-L1, we conducted PD-1/PD-L1 binding assays to evaluate the binding ability of exosomal PD-L1 to PD-1. Exosomes derived from HCC cells, labeled with PKH26, were found to bind to wells coated with recombinant PD-1, and this binding was significantly reduced in ANXA2 knocked-out HCC cells (Fig. 3G). Staining of exosomal markers on primary murine T cell surfaces likewise showed reduced binding of ANXA2 knockout exosomes compared with wild type exosomes (Fig. 3H). Importantly, to directly and quantitatively assess EV–T cell binding, we labeled EVs with a fluorescent dye and measured their binding to T cells by flow cytometry. Compared with EVs from control cells, EVs derived from ANXA2-knockout HCC cells showed a marked reduction in binding to T cells (∼70% decrease) when assayed at 4°C (Fig. 3I). Collectively, these findings indicate that ANXA2 loss in HCC cells reduces overall EV secretion and decreases the loading of functional PD-L1 into EVs, thereby weakening EV–T cell interactions and impairing the T cell–inhibitory capacity of tumor-derived exosomal PD-L1.

Figure 3.

ANXA2 regulates exosomal PD-L1 in hepatocellular carcinoma cells.

(A) Scanning electron microscopy (SEM) images show the morphology of exosomes derived from control (CON) and ANXA2 knockout (ANXA2 KO) HCC cells. (B) Nanoparticle tracking analysis (NTA) graphs depict the size distribution and concentration of exosomes from CON and ANXA2 KO cells, as measured by nano-flow cytometry. (C) Western blot analysis of exosomes reveals the expression of exosomal markers (CD63 and ALIX) and PD-L1 in ANXA2 KO exosomes compared to controls. (D) Western blot analysis of whole cell lysates confirms the protein levels of ALIX, CD63, ANXA2, and Actin in CON and ANXA2 KO cells. (E) Flow cytometry histograms compare PD-L1 expression on exosomes from CON and ANXA2 KO cells. (F) ELISA of PD-L1 levels in exosomes, n = 3. (G) PD-1/PD-L1 binding assay showing that PKH26-labeled exosomes derived from wild-type (WT) HCC cells and ANXA2 knockout (KO) cells adhere to recombinant PD-1–coated wells. (H) Representative image of exosomes labeled with PKH67 to monitor the level of T cell binding to exosomes. (I) Flow cytometry based EV-T cell binding assay measuring the mean fluorescence intensity (MFI) of EV-positive primary murine T cells after incubation at 4°C with fluorescently labeled EVs derived from control (CON) or ANXA2-knockout (ANXA2 KO) HCC cells. Data represent the mean ± SD from three independent experiments. P values were determined by one-way ANOVA; P < 0.001 was considered statistically significant.

ANXA2 deletion reverses the direct inhibition of T cells by HCC-derived exosomal PD-L1

To determine whether ANXA2 could modulate the immunosuppressive effects of exosomal PD-L1 through T cell receptor associated pathways, we established an ex vivo murine T cell stimulation model using primary splenocytes from C57BL/6 J mice, activated with CD3/CD28 stimulation and co-incubated with exosomes (EXO). CD69, a marker of CD8⁺ T cell activation, was significantly increased by CD3/CD28 stimulation, and its expression was markedly downregulated in CD8⁺ T cells treated with wild type EXO, whereas ANXA2 knockout EXO rescued the suppressed CD69 expression (Fig. 4A and B). Additionally, the cytokines secreted during T cell activation, IL-2, IFN-γ, and TNF-α, were suppressed by WT EXO and rescued by KO EXO (Fig. 4C to 4E). These findings consistently demonstrate that vesicles from liver cancer cells can directly inhibit TCR mediated murine T cell activation in a manner dependent on ANXA2 regulated exosomal PD-L1.

Figure 4.

ANXA2 deletion reverses the direct inhibition of T cells by HCC-derived exosomal PD-L1.

(A, B) Flow cytometric analysis of CD69 expression on CD8⁺ T cells stimulated with anti-CD3/CD28 beads and co-incubated with WT-EXO or KO-EXO and quantification of activated CD8⁺ T cells. (C–E) Measurement of key T cell–secreted cytokines (IFN-γ, TNF-α, and IL-2) by ELISA, n = 3. (F)Photographs depicting xenograft tumors at the experimental endpoint. n = 6 mice per group. (G, C) Tumor growth curves and final tumor weights in mice bearing ANXA2 KO tumors treated with exosomes derived from wild-type (WT-EXO) or ANXA2 KO (KO-EXO) cells for 21 days. (I) Representative hematoxylin and eosin (H&E) and immunohistochemical (IHC) staining for Ki-67, CD8, PD-L1 and PD-1 in tumor tissues, scale bar = 50 μm. (J) Quantification of Ki-67⁺ and CD8⁺ cells in tumor sections. (J) Quantification of PD-L1 and PD-1 cells in tumor sections. Data represent the mean ± SD from three independent experiments. P values were determined by one-way ANOVA with multiple-comparisons correction; Statistical significance is denoted as *P < 0.05, **P < 0.01, ***P < 0.001.

To validate our in vitro findings in vivo, we used a C57BL/6 J syngeneic Hepa1-6 CDX model. As observed previously, ANXA2 knockout (KO) significantly suppressed tumor growth in immunocompetent mice (Fig. 4F–H). To test whether tumor-derived EVs (exosomes) could modulate this phenotype, mice bearing ANXA2-KO tumors were administered EVs isolated from control (WT) Hepa1-6 cells (KO+EV-WT) or ANXA2-KO Hepa1-6 cells (KO+EV-KO) for 21 days. EVs from WT cells increased tumor volume and weight, whereas EVs from ANXA2-KO cells produced no obvious increase in tumor burden (Fig. 4G and 4H). Histological and immunohistochemical analyses showed no appreciable differences in Ki-67 signals among the groups, suggesting that the changes in tumor burden were not primarily attributable to altered proliferative activity (Fig. 4I and 4J). In contrast, CD8⁺ T-cell infiltration was higher in ANXA2-KO tumors than in WT tumors. Notably, WT-EV supplementation reduced CD8⁺ signals compared with unsupplemented KO tumors, whereas KO-EVs exhibited a weaker reduction (Fig. 4I and 4J). Consistently, PD-1–positive immune infiltrates were increased in KO tumors compared with WT tumors and were partially attenuated by WT-EV supplementation, while changes after KO-EV supplementation were less evident (Fig. 4I and 4K). In addition, PD-L1 staining was reduced in KO tumors compared with WT tumors, and WT-EV supplementation partially restored the PD-L1 signal (Fig. 4I and 4K). Collectively, these results suggest that WT HCC cell–derived EVs can partially remodel the immune-infiltration phenotype associated with ANXA2 loss in vivo.

ANXA2 regulates exosomal PD-L1 expression through modulating CD63 stabilization

To investigate how ANXA2 influences PD-L1 incorporation into exosomes, we first analyzed TCGA-LIHC and observed that ANXA2 expression positively correlated with multiple endosomal/exosome-related genes, including the tetraspanin CD63, a key component implicated in exosome biogenesis and cargo sorting (Fig. 5A). In whole-cell lysates, ANXA2 knockout markedly reduced CD63 protein abundance(Fig. 5B). CD63 mRNA levels were not significantly altered in murine Hepa1-6 cells (Fig. 5C). Similar results were observed in human HCC cell lines (Huh7 and JHH7), where CD63 transcript levels remained unchanged upon ANXA2 perturbation (Fig. S2). Consistent with reduced CD63 availability, immunofluorescence analyses showed diminished overlap between PD-L1 and CD63-positive compartments upon ANXA2 loss (Fig. 5D). To test whether CD63 contributes to the ANXA2-dependent regulation of exosomal PD-L1, we re-expressed CD63 in ANXA2-KO cells; restoration of CD63 was confirmed by Western blot (plv-CD63) (Fig. 5E). CD63 re-expression rescued the decrease in exosomal PD-L1 (Fig. 5F) and partially restored EV-mediated immunosuppressive activity in functional assays, as evidenced by increased tumor-cell survival in T-cell co-culture (Fig. 5G) and reduced T-cell activation cytokine production (IL-2, TNF-α, and IFN-γ) compared with the ANXA2-KO condition (Fig. 5H). Together, these data support a model in which ANXA2 maintains CD63 protein abundance, thereby promoting efficient PD-L1 sorting into exosomes and contributing to the immunoregulatory effects of tumor-derived EVs.

Figure 5.

ANXA2 regulates exosomal PD-L1 expression by modulating CD63 stabilization.

(A) Correlation matrix from the TCGA database showing the relationship between ANXA2 and multiple exosome-related genes. (B)Western blot analysis of ALIX, CD63, ANXA2, and Actin in CON and ANXA2 KO cells. (C) qPCR analysis of CD63 mRNA expression in CON and ANXA2 KO cells. (D) Immunofluorescence images of control (CON) and ANXA2 knockout (KO) cells stained for PD-L1 (green), ANXA2 (red), and CD63 (purple). (E)Western blot analysis of CD63 protein levels in control (CON), ANXA2-knockout (ANXA2 KO), and ANXA2-KO cells with CD63 overexpression (plv-CD63); (F) Quantification of PD-L1 levels in exosomes derived from CON cells, ANXA2 KO cells, and ANXA2 KO cells overexpressing CD63 was performed using ELISA. (G) Cell survival of control (CON) Hepa1-6 cells, ANXA2 KO cells, and ANXA2 KO cells overexpressing CD63, following co-culture with anti-CD3/CD28-activated T cells. (H) Measurement of key T cell–secreted cytokines (IFN-γ, TNF-α, and IL-2) by ELISA. Data represent the mean ± SD from three independent experiments. P values were determined by one-way ANOVA with multiple-comparisons correction or two-tailed unpaired Student’s t-test; Statistical significance is denoted as *P < 0.05, **P < 0.01, ***P < 0.001.

ANXA2 knockout regulates CD63 and PD-L1 through lysosomal degradation and ubiquitination pathways

To investigate how ANXA2 loss affects CD63 and exosomal PD-L1, we performed RNA-seq in Huh7 control and ANXA2-knockout (KO) cells. Principal component analysis showed clear separation between the two groups (Fig. 6A), and differential expression analysis identified robust transcriptional changes upon ANXA2 depletion, including upregulation of genes such as NEDD4, STX17, and RAB7 (Fig. 6B). GO enrichment highlighted terms related to autophagosome assembly and protein polyubiquitination (Fig. 6C). Consistently, GSEA supported enrichment of ubiquitin-mediated proteolysis and autophagy-related gene sets in ANXA2-KO cells, accompanied by coordinated changes in representative genes (Fig. 6D–G). qRT–PCR further validated increased expression of lysosome/autophagy-associated genes (e.g., LAMP1, ATG16L1, ULK1/2) and ubiquitination-related genes (e.g., NEDD4, STAMBP, SQSTM1) (Fig. 6H).

Figure 6.

ANXA2 Knockout modulates lysosomal degradation and ubiquitination pathways.

(A) Principal Components Analysis (PCA) plot illustrating the distinct clustering of control and ANXA2 knockout (KO) samples. (B) Volcano plot depicting differential gene expression between control and ANXA2 KO samples. Upregulated genes are marked in red, while downregulated genes are in blue. (C) Gene Ontology (GO) enrichment scatter plot showing enriched biological processes associated with ANXA2 KO. The size of the dots represents the number of genes, and the color gradient indicates the significance of enrichment. (D) Gene Set Enrichment Analysis (GSEA) plot for ubiquitin-mediated proteolysis. (F) Heatmap of ubiquitination-related genes comparing control and ANXA2 KO samples. (G) Heatmap of autophagy-related genes comparing control and ANXA2 KO samples. (H) Bar graph showing the relative mRNA expression levels of key autophagy-related genes in control and ANXA2 KO samples. (I) Western blot analysis of CD63 and p62/SQSTM1 in control (CON) and ANXA2-knockout (ANXA2 KO) cells, and in ANXA2 KO cells treated with bafilomycin A1 (BafA1) or TAK-243. (J) ELISA quantification of exosomal PD-L1 in EVs isolated from control (CON) cells, ANXA2-knockout (ANXA2 KO) cells, and ANXA2 KO cells treated with bafilomycin A1 (BafA1) or TAK-243. (K) Western blot analysis of LAMP1 and p62/SQSTM1 in CON and ANXA2 KO cells. P values were determined by one-way ANOVA with multiple-comparisons correction or two-tailed unpaired Student’s t-test; Statistical significance is denoted as *P < 0.05, **P < 0.01, ***P < 0.001.

We next performed pharmacological assays and quantified exosomal PD-L1. ANXA2-KO cells were treated with bafilomycin A1 or TAK-243, followed by exosome collection and PD-L1 measurement. Immunoblotting showed that bafilomycin A1 increased CD63 protein abundance in ANXA2-KO cells, whereas TAK-243 did not (Fig. 6I). Consistently, exosomal PD-L1 levels were restored in the bafilomycin A1–treated group but remained low after TAK-243 treatment (Fig. 6J). Finally, immunoblot analysis of lysosome/autophagy-related markers showed increased LAMP1 and p62/SQSTM1 accumulation in ANXA2-KO cells compared with control cells (Fig. 6K). Collectively, these findings suggest that ANXA2 loss promotes lysosome-dependent turnover of CD63, resulting in reduced CD63 protein abundance and diminished exosomal PD-L1 loading in hepatocellular carcinoma cells.

Discussion

In this study, we identify Annexin A2 (ANXA2) as a tumor-intrinsic regulator of exosomal PD-L1–mediated immune evasion in hepatocellular carcinoma (HCC). ANXA2 is upregulated in HCC and associates with unfavorable clinical outcome. Functionally, ANXA2 ablation restrains tumor progression in immunocompetent syngeneic models and is accompanied by enhanced CD8⁺ T-cell infiltration and effector activity, whereas this antitumor effect is largely diminished in immunodeficient hosts, highlighting an immune-dependent contribution. Mechanistically, our data support an ANXA2-linked axis that maintains CD63 protein abundance and thereby sustains exosomal PD-L1 output. ANXA2 loss reduces CD63 protein levels without a corresponding decrease in CD63 mRNA and is accompanied by activation of lysosome/autophagy-associated programs. Pharmacological blockade of lysosomal acidification with bafilomycin A1 restores CD63 levels and rescues exosomal PD-L1 output, whereas inhibition of ubiquitination with TAK-243 does not, suggesting that lysosome-dependent turnover is a major contributor to CD63 reduction after ANXA2 loss. Consistent with this mechanism, exosomes derived from ANXA2-expressing HCC cells significantly suppress T-cell activation in both in vitro and in vivo settings, reinforcing their role as a conduit of immunosuppression within the HCC tumor microenvironment (TME) (Fig. 7) .

Figure 7.

Schematic model of the ANXA2–CD63–exosomal PD-L1 axis regulating CD8⁺ T-cell function in HCC.

ANXA2 sustains CD63 protein abundance to support PD-L1 sorting and release on HCC-derived exosomes. Upon ANXA2 loss, CD63 undergoes lysosome-dependent degradation, leading to reduced exosomal PD-L1 output. Consequently, diminished exosomal PD-L1–PD-1 signaling relieves CD8⁺ T-cell suppression, restores effector activity (e.g., IL-2, IFN-γ, TNF-α), and contributes to antitumor immunity and tumor growth control.

The emergence of exosomal PD-L1 as a dynamic regulator of immune evasion has reshaped our understanding of tumor-immune crosstalk across malignancies such as melanoma and lung cancer [29,30]. Yet, the molecular determinants of its selective packaging and secretion, particularly in HCC, have remained enigmatic. Our identification of ANXA2 as a pivotal regulator of exosomal PD-L1 in HCC bridges this gap, illuminating a tumor-intrinsic axis that exploits exosomal cargo to subvert antitumor immunity. This finding gains heightened significance in the context of the liver’s tolerogenic immunological landscape, where baseline suppression of immune responses predisposes to chronic inflammation and oncogenesi^s [7,31]. By delineating the ANXA2-CD63-PD-L1 cascade, we provide a mechanistic basis for the elevated exosomal PD-L1 levels observed in HCC patients and their association with resistance to immune checkpoint blockade (ICB).

Exosome biogenesis is conventionally governed by the ESCRT machinery and tetraspanins such as CD63, which orchestrate cargo sorting into intraluminal vesicles [32,33]. Our data support an additional regulatory layer in which ANXA2 is linked to CD63 protein homeostasis and, in turn, to exosomal PD-L1 output. This regulatory layer introduces a disease-specific nuance to exosome assembly, distinguishing it from the ubiquitous ESCRT framework. Given ANXA2’s frequent overexpression in HCC and other aggressive cancers, this mechanism may underlie the variability in exosomal PD-L1 abundance across tumor types, offering a molecular explanation for HCC’s pronounced immunosuppressive phenotype. Intriguingly, the ANXA2-CD63 interaction hints at a broader role for annexins in tailoring exosomal content to the tumor’s evolutionary needs—an hypothesis ripe for exploration across cancer contexts.

In addition to regulating PD-L1 export via exosomes, ANXA2 may also influence PD-L1 at the transcriptional level. Notably, we observed that ANXA2 knockout decreased CD274 mRNA abundance, suggesting that ANXA2 loss can dampen PD-L1 gene expression in HCC cells. Because CD274 transcription is highly context dependent and is regulated by multiple signaling programs, including cytokine-driven IFNγ–JAK/STAT signaling as well as NF-κB, hypoxia, and oncogenic stress–associated pathways, the reduction in CD274 mRNA in ANXA2-deficient cells is likely indirect and may reflect broader changes in upstream signaling states. Importantly, our functional data support a distinct post-transcriptional layer in which ANXA2 maintains CD63 protein abundance and thereby sustains exosomal PD-L1 output, since blockade of lysosomal acidification restores CD63 and exosomal PD-L1 in ANXA2-KO cells. Together, these findings raise the possibility that ANXA2 contributes to immune evasion through dual regulation, coupling PD-L1 gene expression with exosomal PD-L1 export. Future studies evaluating CD274 promoter activity and pathway activation under basal and inflammatory cues (for example, IFNγ stimulation), along with ANXA2 rescue experiments, will help clarify the relative contribution of transcriptional versus exosomal mechanisms.

The therapeutic ramifications of these findings are profound. HCC remains a formidable clinical challenge, with ICB response rates languishing below 20%, largely due to adaptive immune evasion within the TME [34]. Our data suggests that targeting ANXA2 could disrupt exosomal PD-L1 secretion, thereby dismantling a critical immunosuppressive barrier and sensitizing tumors to PD-1/PD-L1 blockade. This approach holds particular promise given ANXA2’s pleiotropic contributions to HCC progression, including metastasis and chemoresistance [35,36], which could be co-opted for synergistic therapeutic gain. Moreover, the reduction in exosomal PD-L1 upon ANXA2 ablation not only enhances local T-cell cytotoxicity but may also mitigate systemic immune suppression—a phenomenon observed in other cancers where exosomal PD-L1 targets distal lymphoid tissues [37]. Such dual local-systemic effects position ANXA2 inhibition as a transformative strategy to amplify ICB efficacy.

Despite these advances, several questions remain. First, while our data support an ANXA2-dependent control of CD63 protein abundance and exosomal PD-L1 output, the molecular basis of this regulation requires further resolution, including whether ANXA2 influences CD63 stability through direct interaction or via broader changes in membrane trafficking and lysosome/autophagy pathways. In addition, although intratumoral (IT) administration of tumor-derived exosomes provided a tractable approach to validate local immune modulation in vivo, this bolus delivery does not fully recapitulate the continuous secretion, systemic pharmacokinetics, and biodistribution of endogenous exosomes. Future studies using systemic dosing paradigms and quantitative biodistribution analyses will be needed to define exposure dynamics and distal immune effects. Second, ANXA2 knockout is accompanied by transcriptional changes relevant to immune regulation, including reduced CD274 mRNA, raising the possibility that ANXA2 may contribute to PD-L1 control at both transcriptional and exosomal levels. Third, although we observe enhanced CD8⁺ T-cell infiltration and effector function in immunocompetent settings, the impact of ANXA2-regulated exosomes on other immune compartments, such as regulatory T cells, myeloid populations, and innate immune cells, remains to be defined. Addressing these points will benefit from complementary biochemical approaches to probe ANXA2–CD63 relationships, quantitative analyses of CD63 turnover, and systematic immune profiling across cellular compartments. In parallel, evaluating ANXA2-targeting strategies and their combination with immune checkpoint blockade in clinically relevant HCC models may help clarify therapeutic potential and define biomarkers of response.

In conclusion, our findings support ANXA2 as a tumor-intrinsic determinant of immune evasion in HCC by shaping PD-L1 availability through exosome-associated mechanisms. This work highlights that exosome output is not merely a constitutive process but can be tuned by oncogenic regulators to reinforce an immunosuppressive tumor microenvironment. Therapeutically, disrupting ANXA2-dependent pathways may provide a complementary strategy to enhance antitumor immunity and improve responses to immune checkpoint blockade. Further studies will be needed to define the precise molecular underpinnings and to evaluate translational potential in clinically relevant HCC models.

Ethics approval and consent to participate

All animal-related procedures were conducted in strict compliance with China's national regulations on the use of experimental animals and adhered to the guidelines established by the Experimental Animal Ethics Committee of Soochow University (No. AD2025074), which reviewed and approved the study.

Data availability

The data supporting the findings of this study are available within the article. Additional data related to this research may be requested from the corresponding authors.

Funding

This study was supported by the Suzhou Science and Education Strengthening Health Youth Project (KJXW2022014).

CRediT authorship contribution statement

Jian Zhang: Writing – review & editing, Writing – original draft, Visualization, Validation, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Ying Gu: Writing – original draft, Validation, Methodology, Investigation, Funding acquisition, Formal analysis, Conceptualization. Fangchao Zhao: Writing – original draft, Validation, Methodology, Investigation, Formal analysis, Data curation. Yong Chen: Writing – original draft, Validation, Methodology, Investigation, Formal analysis, Conceptualization. Yuqing Xia: Writing – original draft, Validation, Supervision, Resources, Methodology, Investigation, Conceptualization. Dekang Gao: Writing – original draft, Project administration, Methodology, Investigation, Formal analysis, Data curation. Qiang Yuan: Writing – original draft, Visualization, Validation, Supervision, Methodology, Investigation, Data curation, Conceptualization. Xuming Bai: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization.

Declaration of competing interest

The authors declare that they have no competing interests.