NK cell vesicles-based nanoplatform loaded with sorafenib synergizing ferroptosis and antitumor immunity against hepatocellular carcinoma Authors: Chulan Gong, Shenjie Zhang, Yuting Fan, Dan Liang, Zailing Yang, Jing Zhang, Xiong Liu, Ting Guo, Xinlan Li, Tao Shen, Changya Li, Xing Zhao, Shi Zuo Published: 2026 ABSTRACT AbstractBackgroundFerroptosis is a type of cell death that relies on iron and is marked by an excess of intracellular iron and the buildup of lipid peroxides. This process presents a new and promising approach for cancer treatment. Improving ferroptosis can effectively mitigate the limitations of conventional treatment methods.MethodsWe developed a synergistic nanotherapeutic platform. Natural killer cell derived nanovesicles (NK-NVs) were fabricated using an extrusion method and subsequently loaded with sorafenib (SRB) via electroporation, resulting in the NK-NVs-SRB complex for targeted delivery.ResultsThe therapeutic strategy not only escaped recognition by drug efflux pumps, thus lowering drug expulsion, but also effectively triggered ferroptosis in cancer cells. This process induced significant release of damage associated molecular patterns (DAMPs), ultimately leading to immunogenic cell death (ICD). In vivo experiments demonstrated that NK-NVs-SRB intervention achieved a tumor growth inhibition rate of 44.54%, effectively drove the ferroptosis process, and activated anti-tumor immunity by recruiting CD8 + T cell infiltration.ConclusionsThis study proposes a novel therapeutic strategy for hepatocellular carcinoma (HCC) that acts by synergistically inducing ferroptosis and activating anti-tumor immunity, thereby paving a new avenue for HCC treatment.Graphical abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s12951-026-04550-8.Keywords: NK cell vesicles, Sorafenib, Ferroptosis, Antitumor immunity, Hepatocellular carcinoma Introduction Globally, liver cancer is the sixth most diagnosed cancer and the third leading cause of cancer-related deaths, representing a significant challenge for healthcare systems worldwide [1, 2]. Viral infections, high alcohol intake, and metabolic conditions are major contributors to hepatocellular carcinoma (HCC) [3]. As HCC typically advances without distinct early symptoms, most individuals are diagnosed when the disease is at an intermediate or advanced stage. For those with early stage HCC, surgical intervention remains the most effective treatment, with a five-year survival rate exceeding 70% [4]. However, only about 40% of patients are diagnosed early, leaving the majority ineligible for surgery [5]. Sorafenib (SRB), the initial FDA-approved treatment for advanced HCC, acts as a tyrosine kinase inhibitor to curb tumor growth and proliferation. It also exhibits antiangiogenic effects and has been shown to extend overall survival in HCC patients [6]. Nevertheless, only about one-third of patients respond to sorafenib treatment, and drug resistance frequently develops, often leading to relapse within six months [7]. Overcoming sorafenib resistance and enhancing HCC sensitivity to the drug remain critical challenges that require urgent solutions. Ferroptosis refers to a regulated cell death mechanism that is triggered by lipid peroxidation dependent on iron [8]. Studies have shown that ferroptosis inducers can inhibit tumor growth, remodel the tumor microenvironment, and overcome resistance to conventional therapies in various cancers by enhancing the response to immunotherapy [9]. The core regulatory mechanisms of ferroptosis involve glutathione (GSH) depletion, inhibition of glutathione peroxidase 4 (GPX4) activity, and accumulation of lipid reactive oxygen species (ROS). Sorafenib, a multi-kinase inhibitor, plays an important role in the treatment of advanced hepatocellular carcinoma. Several studies show that sorafenib therapy markedly elevates intracellular ROS production, malondialdehyde (MDA) levels, and iron content, while decreasing GSH levels [10–12]. These changes are hallmark features of ferroptosis. Together, this evidence suggests that sorafenib effectively induces ferroptosis in tumor cells by modulating iron metabolism and oxidative stress pathways, providing important insight into its antitumor mechanism. Research also indicates that sorafenib resistance is closely associated with suppression of ferroptosis [13, 14]. In sorafenib resistant tumor cells, the ferroptosis pathway is often significantly dysregulated, leading to reduced sensitivity to the drug. Therefore, interventional strategies targeting ferroptosis regulation offer a new direction for overcoming sorafenib resistance. PI3K-AKT-mTOR signaling pathway is one of the most critical survival pathways in cells, extensively involved in regulating cell proliferation, metabolism, survival, and apoptosis [15]. It is also one of the most frequently dysregulated signaling cascades in cancer patients, playing a key role in tumor initiation, progression, and treatment response [16]. Multiple studies have indicated that abnormal activation of the PI3K-AKT-mTOR pathway is associated with drug resistance [17–19]. Recent findings have shown that the activation of the PI3K-AKT-mTOR pathway can greatly inhibit ferroptosis. For instance, in a lipopolysaccharide-induced renal podocyte model, overexpression of CIAPIN1 activated the PI3K-AKT pathway and upregulated key antiferroptosis proteins such as SLC7A11 and GPX4, thereby mitigating ferroptosis related injury [20]. Additional research indicates that the PI3K-AKT-mTOR pathway might suppress ferroptosis via various mechanisms, and blocking this pathway can enhance the induction of ferroptosis [21, 22]. Moreover, there is evidence linking PI3K-AKT-mTOR pathway activation to sorafenib resistance [23, 24], while inhibiting this pathway has been shown to improve sorafenib sensitivity [25, 26]. These findings provide a rationale for developing combination strategies that target the PI3K-AKT-mTOR pathway to overcome ferroptosis associated treatment resistance. NK cells, which are cytotoxic lymphocytes, have the capability to directly target and eliminate virus-infected and tumor cells [27]. NK cell based therapy combined with chemotherapy has demonstrated efficacy against various cancers [28, 29]. NK cell-derived vesicles (NK-NVs), preserve the cytotoxic abilities of NK cells and function as drug carriers that can specifically target tumor cells. This positions them as a promising “cell-free” alternative for antitumor treatment. The inherent tumor targeting capability of NK-NVs is largely due to surface receptors such as NKG2D, which recognize and specifically bind to ligands on tumor cells, enabling targeted delivery. Owing to their biocompatibility, low immunogenicity, and tumor targeting properties, NK-NVs represent a highly promising drug delivery platform [30]. Our previous study reported that doxorubicin encapsulated in NK-NVs significantly enhanced antitumor efficacy [31]. In this study, we developed an immune cell-based nanoplatform using NK-NVs to efficiently deliver SRB to HCC cells. By entering tumor cells via endocytosis, NK-NVs help evade recognition by drug efflux pumps, thereby reducing SRB expulsion. Moreover, the nanoplatform induced ferroptosis in sorafenib resistant HCC cells by inhibiting the PI3K-AKT-mTOR signaling pathway, while potentially triggering immunogenic cell death to activate antitumor immunity. Together, these mechanisms work synergistically to overcome drug resistance. In a tumor bearing animal model, treatment with NK-NVs-SRB resulted in a promising tumor growth inhibition rate. This approach aims to achieve targeted delivery of SRB and effectively promote both ferroptosis and antitumor immune responses, demonstrating significant potential for HCC therapy. Methods Cell culture The human hepatocellular carcinoma cell line SNU398 was obtained from Sanxun Biological Technology (Guangzhou, China). Huh7 cells were sourced from Wuhan Servicebio Technology. MHCC97H, HLF, and HepG2 cells, along with the ovarian cancer cell line SKOV3, were acquired from the China Center for Type Culture Collection. The human pancreatic cancer cell line BXPC-3 was purchased from Haixing Biotechnology Co., Ltd. Human NK92MI cells were procured from Procell Life Science & Technology Co., Ltd. SKOV3 cells were maintained in McCoy’s 5a medium (Gibco, Gaithersburg, USA). Panc-1 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco). Human NK92MI cells were grown in MEMα medium (Gibco). All media were supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (Gibco). Cells were incubated at 37 °C in a humidified atmosphere containing 5% CO₂ and 95% air. Generation of sorafenib resistant cells To determine the half-maximal inhibitory concentration (IC50) of sorafenib in different HCC cell lines, several HCC cells were seeded in 96 well plates and treated with varying concentrations of sorafenib. Cell viability was measured after 48 h. Huh7 cells in the logarithmic growth phase were cultured in 6 well plates and exposed to sorafenib at a concentration slightly below their IC50. After stable growth was achieved, the sorafenib concentration was gradually increased. Following six months of selection, a sorafenib resistant cell line, designated Huh7/SR, was established. These cells were maintained under continuous sorafenib treatment to preserve the resistant phenotype. Isolation and characterization of NK-NVs Following the expansion of NK92MI cells in vitro, the cells were washed twice with PBS and counted. The cells were resuspended in 1 mL of saline at a concentration of 2 × 10⁶ cells/mL. After low temperature ultrasonication, the sample was subjected to low-speed centrifugation to remove cellular debris, and the supernatant was collected. The collected supernatant was sequentially extruded through membranes of different pore sizes in a gradient manner. The resulting suspension was then centrifuged at 10,000 × g for 30 min, and the supernatant was filtered through a 0.22 μm filter. The protein concentration of NK-NVs was determined using a BCA protein quantification kit. The prepared NK-NVs were stored at −80 °C for subsequent experiments. The hydrodynamic diameter and size distribution profile of NK-NVs were measured by dynamic light scattering (DLS) using a Malvern Zetasizer Nano ZS instrument. The expression of specific surface markers on NK-NVs was confirmed by Western blot analysis using the following primary antibodies: anti-CD56 (CST, 99746 S), anti-CD63 (Affinity, AF5117), anti-CD81 (CST, 56039 S), anti-Alix (MCE, HY-P80011), and anti-Calnexin (Biodragon, BD-PT0613). Cellular uptake of NK-NVs in vitro and targeting specificity in vivo Huh7/SR cells (1 × 10⁵ cells/well in confocal dishes) were cultured overnight, after which the original medium was removed. DIO-stained NK-NVs (MCE, Cat# HY-D0969) were added to Huh7/SR cells and incubated for different time periods. Cells were stained with Actin-Tracker Red-Rhodamine (Beyotime, C2207S) at 37 °C for 30 min, followed by DAPI staining to visualize nuclei. Cellular uptake was observed using a confocal microscope (Nikon, Japan). For in vivo targeting evaluation, DIR-labeled NK-NVs (MCE, HY-D1048) were administered to C57BL/6J mice via intraperitoneal injection after removing free dye through ultrafiltration. Fluorescence intensity in different organs was measured at various time points using an IVIS imaging system (PerkinElmer, USA) to assess organ targeting specificity. Preparation and characterization of NK-NVs-SRB NK-NVs, SRB, and electroporation buffer were mixed at a ratio of 1:1:2. Electroporation was performed using a Celetrix electroporator under the conditions of 450 V for 120 ms. Immediately after electroporation, the mixture was placed on ice for 5–10 min to facilitate membrane recovery. Following drug loading, the sample was centrifuged at 100,000 × g for 70 min to separate the pellet and supernatant. The pellet was resuspended in deionized water for subsequent measurements of particle size and zeta potential. The supernatant was analyzed by HPLC to determine the amount of unencapsulated SRB. The encapsulation efficiency was calculated using the following formula: Encapsulation Efficiency (%) = (Total Drug Amount - Free Drug Amount)/Total Drug Amount ×100%. Cell viability assay To evaluate the cytotoxicity of NK-NVs-SRB, Huh7/SR and SNU398 cells were seeded in 96 well plates at a density of 1 × 10⁴ cells per well and cultured overnight until full attachment was achieved. The cells were then treated with equivalent amounts of NK-NVs, free SRB, and NK-NVs-SRB, followed by incubation for 48 h (37 °C, 5% CO₂). Subsequently, 10 µL of Cell Counting Kit-8 (CCK8; Meilun Biotechnology, China) reagent was added to each well, and the plates were incubated at 37 °C for 2 h. The OD at 450 nm was measured for each well (Synergy H1, BioTek Instruments, Inc.). Each experimental group was set up in triplicate wells, and all experiments were independently repeated three times. The cell survival rate was calculated using the following formula: Cell Survival Rate (%) = (OD of experimental group - OD of blank group)/(OD of control group - OD of blank group) × 100%. Measurement of MDA Levels Malondialdehyde (MDA), a final product of lipid peroxidation, is positively correlated with ferroptosis. The MDA assay kit (Nanjing Jiancheng Bioengineering Institute, China) was used to measure MDA levels in cells from each experimental group. Briefly, cells were scraped and collected. After adding the corresponding reagents according to the manufacturer’s instructions, the samples were sonicated to lyse the cells. Protein concentration was determined using a BCA protein assay kit (Solarbio, Cat# PC0020). Following the preparation of the MDA working solution as described in the kit protocol, sample aliquots or blank controls were added for analysis. The reaction mixture was heated in a boiling water bath for 45 min, cooled under running water, and then centrifuged to obtain the supernatant. Finally, the absorbance at 530 nm was measured using a microplate reader. Measurement of GSH GSH serves as the most critical endogenous antioxidant defense mechanism against ferroptosis. Its cellular level directly influences cellular susceptibility to ferroptotic cell death. The GSH content in cells from each experimental group was determined using a Reduced Glutathione Assay Kit (Nanjing Jiancheng Bioengineering Institute, Cat# A006-2-1). After cell collection and counting, reagents were added according to the manufacturer’s instructions, followed by cell lysis. A standard solution was prepared accordingly. After adding the required reagents as specified, the absorbance was measured at 405 nm using a microplate reader. Detection of mitochondrial membrane potential Huh7/SR and SNU398 cells were seeded in confocal dishes at a density of 1 × 10⁵ cells per well and cultured overnight. The cells were then treated with DMEM maintenance medium, NK-NVs, SRB, or NK-NVs-SRB for 48 h. After treatment, JC-1 staining solution (MCE, HY-K0601) was added, and the cells were incubated at 37 °C for 20 min. Mitochondrial membrane potential changes were subsequently observed using a confocal microscope (Nikon, Japan). Reactive oxygen species (ROS) detection Huh7/SR and SNU398 cells were seeded in 6well plates at a density of 3 × 10⁵ cells per well and cultured overnight. The cells were then treated with DMEM maintenance medium, NK-NVs, SRB, or NK-NVs-SRB for 24 h. After treatment, cells were collected and washed twice with ice-cold PBS, followed by resuspension in 1× binding buffer. The cells were incubated with a ROS detection reagent for 30 min in the dark. Subsequently, 1× binding buffer was added and gently mixed. ROS levels were analyzed using a flow cytometer (BD LSR Fortessa 4, USA), and the data were processed with FlowJo 10 software. For confocal imaging, cells were stained directly without harvesting and immediately observed under the microscope. ICD effect Huh7/SR cells (1 × 10⁴) were seeded on coverslips and cultured overnight. The following day, cells were treated with different drugs for 48 h. For immunofluorescence, cells were stained with primary antibodies against HMGB1 (ABclonal, A19529) and CRT (ABclonal, A20986) at 4 °C overnight, followed by incubation with fluorescently-labeled secondary antibodies at room temperature for 1 h. After three washes with PBS, cells were co-stained with Actin-Tracker and DAPI, and finally imaged under a CLSM. For HMGB1 release analysis, cell culture supernatant was collected and measured using an HMGB1 ELISA kit (Xiamen, China) according to the manufacturer’s instructions. For in vivo analysis, tumor tissues were harvested, embedded in paraffin, and sectioned for immunohistochemical (IHC) staining as described above. Western blot analysis Cells from each group were lysed on ice using RIPA lysis buffer (Solarbio) for 10 min, followed by centrifugation to extract total protein. Protein concentrations were determined using a BCA assay kit (Solarbio). Proteins were separated by 10% polyacrylamide gel electrophoresis and subsequently transferred to activated PVDF membranes. The membranes were blocked with RapidBlock™ Buffer for 1 h, then incubated with primary antibody solutions at 4 °C overnight. After three washes with TBST, the membranes were incubated with secondary antibodies diluted 1:10,000 at room temperature for 1 h. Following additional TBST washes, protein signals were detected using the BeyoECL Plus kit. Protein expression levels were quantified and analyzed with ImageJ software. TEM imaging Samples were fixed with 3% glutaraldehyde, followed by post-fixation in 1% osmium tetroxide. After dehydration, the specimens were infiltrated and embedded in resin. Ultrathin Sects. (60–90 nm) were prepared using an ultramicrotome and mounted on copper grids. The sections were then stained with uranyl acetate for 10–15 min and lead citrate for 1–2 min before being visualized under the transmission electron microscope. Animal model An orthotopic liver tumor model was established in this study. Mice were anesthetized using isoflurane inhalation anesthesia. Tumor implantation was performed by injection of 1 × 10⁶ luc-Hepa1-6 cells. The harvested cells were maintained on ice until the moment of injection. A 1.5 cm incision was made in the abdominal wall, followed by a 1 cm peritoneal incision to expose the liver. Cells were injected into the left hepatic lobe of each mouse, with visible color change observed in the liver during the injection procedure. Following injection, both the peritoneal incision and skin were surgically closed. Treatments were initiated according to experimental groups three days after tumor cell implantation. Three weeks post-injection, all mice were euthanized. At the experimental endpoint, livers and tumors were harvested, along with heart, kidney, spleen, and lung tissues for subsequent analysis. In vivo therapeutic efficacy and safety evaluation In the antitumor therapeutic study, Luc-Hepa1-6 tumor-bearing mice were randomly assigned to four treatment groups: PBS, NK-NVs, SRB, and NK-NVs-SRB. All treatments were administered at a dose of 10 mg/kg. Tumor fluorescence and mouse body weight were monitored every other day for 14 days, after which the tumors were harvested. The tumor growth inhibition (TGI) rate was calculated using the formula: TGI = 100 × [1 − (mt/mc)] where mt represents the average tumor weight in the treatment group and mc denotes the average tumor weight in the control group. Blood samples were collected from the mice, and serum levels of ALT and AST were measured using commercial kits (Nanjing Jiancheng Bioengineering Institute) to evaluate safety. Intratumoral Fe²⁺ levels were determined using an iron assay kit (Solarbio). Major organs and tumor tissues were fixed in 4% paraformaldehyde for subsequent H&E staining and immunohistochemical analysis. Statistical analysis Statistical analysis was performed using GraphPad Prism 9 software, employing one-way ANOVA and two-way ANOVA where appropriate. Data are presented as mean ± standard deviation. A p-value of less than 0.05 was considered statistically significant. Results and discussion Synthesis and characterization of NK-NVs-SRB Many researchers have used nanocarriers for drug delivery to enhance tumor cell response to SRB, increasing cellular sensitivity with materials like polymer, lipid, silica, and metal nanoparticles [32, 33]. However, synthetic materials may produce adverse effects [34], including potential toxicity to cardiovascular and neurological systems [35, 36]. Natural killer cells play a key role in the immune system with their innate ability to fight tumors without the need for antigen presentation [37]. Our previous research demonstrated that NK cell derived exosomes could enhance chemotherapeutic efficacy [38]. Nevertheless, isolating exosomes is still technically difficult and yields are limited. We therefore developed an extrusion method to generate NK-NVs (Fig. 1A), achieving substantially higher particle numbers and protein concentrations compared to natural exosomes [31]. Preparation and characterizations of NK-NVs-SRB. (A) Schematic illustration of NK-NVs preparation (created in BioRender). (B) Particle size of NK-NVs. (C) TEM images of NK-NVs. (Scale bar = 200 nm). (D) Western blot analysis of NK cell marker proteins on NK-NVs. (E) Confocal microscopy analysis of Huh7/SR after co-incubation with DiO-labelled NK-NVs at different time periods (6, 12 and 24 h)(Scale bar = 10 μm). (F) Schematic illustration of the electroporation-mediated loading of SRB into NK-NVs (created in BioRender). (G) Particle size of NK-NVs-SRB. (H) Zeta potential. (I) The colloidal stability and (J) PDI variations of NK-NVs-SRB at different temperatures (−20℃ and −80℃) for 5 days (n = 5). (K) Serum stability of NK-NVs-SRB evaluated by hydrodynamic diameter changes. (L) In vitro release profiles of NK-NVs-SRB under different pH conditions Initially, we examined the biochemical and physical characteristics of NK-NVs. Dynamic light scattering analysis indicated an average particle size of 134.53 ± 3.41 nm (Fig. 1B). The morphology of the vesicles, as shown by transmission electron microscopy, was typically that of a saucer (Fig. 1C). Protein expression analysis revealed that NVs had positive markers like Alix, CD63, and CD81, while calnexin, an endoplasmic reticulum marker, was not detected. CD56 expression was expressed in both NK cells as well as NK-NVs (Fig. 1D). These results confirm successful isolation of NK-NVs from NK92MI cells. To evaluate the cellular uptake efficiency, Huh7/SR cells were cultured with NK-NVs and analyzed at various times using confocal laser scanning microscopy. Fluorescent signals from DiO labeled NK-NVs were detectable as early as 6 h after incubation (Fig. 1E). The increase in signal intensity over time, culminating in maximum uptake at 24 h, indicates that NK-NVs are internalized by cells depending on time. We then refined the loading conditions by systematically comparing different ratios of the drug to the carrier and employing various loading strategies. The optimal formulation was achieved at an NK-NVs to SRB mass ratio of 1:1 using electroporation, which provided improved particle size distribution and stability (Figure S1). This method was selected for subsequent preparation (Fig. 1F). The resulting NK-NVs-SRB exhibited a uniform size of approximately 167.08 ± 2.05 nm with a polydispersity index (PDI) of 0.2348 (Fig. 1G). Successful drug loading was further confirmed by changes in zeta potential: blank NK-NVs showed a surface charge of −11.00 ± 0.45 mV, which is consistent with the negatively charged phospholipid membrane of natural vesicles. After SRB loading, the zeta potential significantly shifted to −7.51 ± 0.64 mV (Fig. 1H). This notable change not only confirms successful encapsulation but may also influence the biological behavior of the nanoparticles. A decrease in negative charge might lessen the electrostatic repulsion between vesicles and negatively charged cancer cell membranes, facilitating closer interaction and boosting cellular uptake through endocytosis. This mechanism may contribute to the superior efficacy of NK-NVs-SRB in overcoming drug resistance compared to free drug. With the optimized electroporation protocol, a drug loading efficiency of 34.8% was obtained, as quantified using HPLC. The successfully constructed NK-NVs-SRB complex provided a reliable tool for subsequent investigations. To evaluate stability under different storage conditions, we monitored the size and PDI of NK-NVs-SRB over time at −20 °C and −80 °C. Assessment of long-term stability indicated that NK-NVs-SRB retained high stability at −80 °C, with negligible variations in particle size or PDI over a period of 120 h (Fig. 1I-J). On the other hand, we assessed the serum stability of NK-NVs-SRB by monitoring its hydrodynamic diameter in fresh mouse serum, aiming to mimic its in vivo half-life, and the results showed that NK-NVs-SRB maintained stable size for 48 h (Fig. 1K). Moreover, NK-NVs-SRB exhibited accelerated release under acidic pH (5.5), achieving a cumulative release of more than 90% at 8 h, while a much slower release profile was observed at pH 7.4 (Fig. 1L). These results collectively demonstrate that the prepared NK-NVs-SRB meets standard criteria for nanomedicine applications. Preparation of NK-NVs-SRB and its antitumor efficacy in vitro Considering the inadequate response and frequent resistance development to sorafenib observed in clinical practice, we exposed several hepatocellular carcinoma cell lines to the drug. As shown in Fig. 2A and S2A–B, The cytotoxicity of sorafenib was concentration dependent, and its effects varied significantly among different cell lines. Huh7 and HepG2 cells were relatively sensitive, with IC50 values of 2.050 µM and 2.263 µM, respectively, whereas SNU398 cells were less responsive, showing an IC50 of 18.50 µM. Subsequently, we established a sorafenib resistant Huh7 subline (Huh7/SR) by chronically exposing the cells to gradually increasing concentrations of the drug over six months (Fig. 2B). The successful development of Huh7/SR cells was confirmed by a resistance index (RI) greater than 5 (Fig. 2C). P-glycoprotein (Pgp), a transmembrane transporter belonging to the ATP-binding cassette (ABC) family [39], plays a key role in drug resistance by actively extruding drugs from cells in an ATP-dependent manner [40]. Western blot analysis verified a notable rise in Pgp expression in the developed Huh7/SR cells (Figure S3A-B). These findings demonstrate the successful development of a Huh7 cell line that is resistant to sorafenib. Construction of Huh7/SR and in vitro antibody tumor effects of NK-NVs-SRB. (A) Sensitivity of different hepatocellular carcinoma cell lines to SRB. (B) Schematic illustration of the establishment of the sorafenib-resistant Huh7/SR cell line (created in BioRender). (C) Sensitivity to sorafenib in Huh7 and Huh7/SR. (D) The anti-tumor effect of NK-NVs-SRB on Huh7/SR and SNU398 cells. (E) Cytotoxic effect of NK-NVs-SRB on other hepatocellular carcinoma cell lines. (F) In vitro safety of NK-NVs-SRB. (G) Effect of NK-NVs-SRB on the colony formation ability of Huh7/SR and (H) SNU398 cells. (I) In vitro cell migration assay. (J) In vitro cell invasion assay (scale bar = 100 μm) (∗∗∗/### P < 0.001, ∗∗/##P < 0.01, ∗/#P < 0.05, ns, not significant, #means vs. control group) Extracellular vesicles share lipid and protein membrane components with their parent cells, enabling targeted delivery of bioactive cargo to recipient cells [41]. Consequently, vesicles produced by different cell types possess distinct biological functions, permitting them to serve as both inherent therapies and engineered delivery mechanisms for precise molecular transport. In Supplement S4A, we performed proteomic analysis on isolated NK-NVs and conducted KEGG enrichment analysis to investigate their biological functions. The results indicated that NK-NVs retain functional similarities to their parent NK cells. Notably, we found that NK-NVs had cytotoxic effects on different tumor cell lines, and this activity was positively correlated with the concentration of vesicles (Fig. S4B). These findings further confirm that NK-NVs possess inherent antitumor efficacy. Evidence indicates that nanoformulations can amplify the tumor-suppressing impact of drugs and minimize the dosage needed [42, 43]. To increase the sensitivity of resistant cells to sorafenib, SRB was loaded in NK-NVs and their therapeutic potential was evaluated. Comparative analysis revealed that NK-NVs loaded via electroporation achieved the strongest tumor cell killing, while increasing the mass of NK-NVs did not enhance the cytotoxicity of NK-NVs-SRB (Figure S5A-B). We subsequently examined the cytotoxic effects of NK-NVs-SRB on hepatocellular carcinoma cells with CCK-8 assays (Fig. 2D-E, Figure S5C-E). Notably, at equivalent concentrations, both free SRB and blank NK-NVs showed minimal toxicity against resistant cells, whereas the NK-NVs-SRB formulation induced strong dose dependent cell killing. Meanwhile, NK-NVs-SRB caused less than 20% damage to normal cells, demonstrating favorable safety (Fig. 2F, Figure S6). To assess the long-term inhibitory effects, colony formation assays were performed to evaluate proliferative capacity following treatment. Consistent with CCK-8 results, NK-NVs-SRB treatment nearly completely abolished colony formation in resistant cells, while free SRB and blank NK-NVs showed negligible effects (Fig. 2G-H). This discovery clearly shows that NK-NVs-SRB not only causes immediate cell death but also successfully eradicates proliferative cancer cell groups, which frequently lead to cancer relapse and unsuccessful treatments. Transwell assays revealed that NK-NVs-SRB treatment significantly suppressed the invasive and migratory capacities of resistant cells, whereas free SRB and blank NK-NVs had minimal impact (Fig. 2I-J). Multiple studies indicate that cells primarily internalize extracellular vesicles through classical endocytic mechanisms [44]. Inhibition of caveolae-mediated endocytosis with MβCD led to a decrease in the cytotoxic effects of NK-NVs-SRB on HCC cells (Figure S5F), which is in agreement with our prior findings [31]. Summarizing, our results show that the NK-NVs-SRB we constructed is a stable nanomedicine system for efficient SRB delivery. The marked difference in efficacy underscores the importance of our NK-NVs delivery platform for improving drug sensitivity. Moreover, the lack of cytotoxicity observed with blank NK-NVs at the tested dosage indicates that the therapeutic effect is not solely mediated by the delivered SRB. This observation provides a foundation for further investigation into the potential synergistic role of NK-NVs themselves, as revealed in subsequent mechanistic studies. The core mechanisms of nanotechnology-driven drug delivery systems involve either bypassing or suppressing the Pgp efflux pump [45]. This version places the mechanisms at the forefront and uses “bypass” and “suppress” to clearly distinguish between the two distinct actions [46, 47]. Consistent with these reports, the stark contrast between the ineffectiveness of free SRB and the potent efficacy of NK-NVs-SRB strongly suggests that our NK-NVs delivery system bypasses recognition by the key resistance protein Pgp (Supplementary Fig. 6). This mechanism enables the encapsulated SRB to evade being expelled from the cells. These findings position NK-NVs-SRB not merely as a cytotoxic agent, but as an integrated therapeutic strategy addressing a critical clinical challenge in advanced liver cancer. Inducing of ferroptosis in HCC cells by NK-NVs-SRB To clarify how NK-NVs-SRB causes the death of resistant cells, we initially conducted RNA sequencing analysis. RNA sequencing revealed that NK-NVs-SRB treatment could affect the ferroptosis pathway (Fig. 3A). This critical evidence led us to hypothesize that ferroptosis, a process initiated by iron accumulation and lipid peroxidation [48], serves as a key mediator of the observed cell death. To validate this, we applied different cell death inhibitors, including the ferroptosis inhibitor Ferrostatin-1 (Fer-1), autophagy inhibitor 3-Methyladenine (3-MA), apoptosis inhibitor Z-VAD-FMK, and pyroptosis inhibitor 2-Bromohexadecanoic acid (2-BP). Notably, only Fer-1 effectively reversed the cytotoxic effect of NK-NVs-SRB on hepatocellular carcinoma cells (Fig. 3B-G). Characteristics of ferroptosis of Huh7/SR cells by NK-NVs-SRB. (A) Using transcriptomic data, we performed pathway analysis of differentially expressed genes in the Kyoto Encyclopedia of Genes and Genomes (KEGG) between the control group and the NK-NVs-SRB group. (B-E) Inhibitory effects of different inhibitors on NK-NVs-SRB. (F) Assessment of cell death using PI staining (scale bar = 50 μm). (G) PI staining fluorescence statistical plot. (H) Effect of NK-NVs-SRB on Fe2+ levels and (I) MDA and (J) GSH in Huh7/SR cells with or without Fer-1 after the cells were treated. (K-L) Effects of NK-NVs-SRB on intracellular ROS levels in Huh7/SR cells, detected by confocal microscopy (scale bar = 100 μm) and (M-N) flow cytometry, in the presence or absence of Fer-1. (O) TEM images of mitochondria with characteristic morphologies due to ferroptosis, and arrowheads point to an abnormal morphology with an increase in the mitochondrial membrane density in the cells treated with NK-NVs-SRB. (scale bar = 2 μm for a low-power microscope (left); 500 nm for a high-power microscope (right)) (P) Effect of NK-NVs-SRB on mitochondrial membrane potential (ΔΨm) with or without Fer-1 (scale bar = 10 μm). (Q-R) Western blotting confirmed the expression levels of ferroptosis-associated proteins (ACSL4, SLC7A11, GPX4, FTH1) in cells treated with NK-NVs, SRB, NK-NVs-SRB (∗∗∗/### P < 0.001, ∗∗/##P < 0.01, ∗/#P < 0.05, ns, not significant. #means vs. control group) Ferroptosis depends on two key factors: high levels of iron and unchecked lipid peroxidation [49, 50]. Our experimental data confirmed both prerequisites. On one hand, NK-NVs-SRB treatment significantly increased intracellular ferrous iron (Fe²⁺) levels (Fig. 3H). On the other hand, the concentration of malondialdehyde (MDA), a terminal product of lipid peroxidation, was markedly elevated (Fig. 3I). Glutathione (GSH), one of the most crucial intracellular antioxidants, plays a central role in scavenging reactive oxygen species (ROS) and protecting cells from oxidative damage. In the process of ferroptosis, it mainly operates by regulating the activity of glutathione peroxidase 4 (GPX4). As shown in Fig. 3J-N, NK-NVs-SRB treatment notably reduced GSH activity and led to a substantial accumulation of ROS in Huh7/SR. Additionally, ferroptosis leads to unique modifications in mitochondrial structure [51]. Transmission electron microscopy revealed that while untreated control cells exhibited normal, intact mitochondria, NK-NVs-SRB treated cells showed characteristic features of ferroptosis, including mitochondrial shrinkage, increased membrane density, and disrupted cristae (Fig. 3O). These ultrastructural characteristics offer visual morphological evidence that distinguishes ferroptosis from other forms of cell death, such as apoptosis or necrosis. JC-1 staining further revealed a specific loss of mitochondrial membrane potential (ΔΨm) (Fig. 3P). Consistent with our earlier findings, these results collectively confirm the induction of ferroptosis. A strong defense mechanism against ferroptosis is present in healthy cells, mainly facilitated by the essential proteins SLC7A11 and GPX4 [52–54]. Significantly, Western blot analysis revealed that NK-NVs-SRB treatment greatly reduced the expression of SLC7A11, GPX4, and FTH1 (Fig. 3Q-R), further indicating that ferroptosis is the main mechanism of cell death in tumor cells. Subsequent experiments with the ferroptosis inhibitor Fer-1 confirmed this mechanism. Fer-1 treatment effectively suppressed membrane lipid peroxidation, reduced intracellular Fe²⁺ levels, and restored GSH concentration. Concurrently, we observed partial recovery of ROS accumulation and mitochondrial membrane potential. These findings indicate that our delivery system not only enhances drug accumulation at tumor sites but also synergistically amplifies the antitumor efficacy by promoting ferroptosis. This multi-step process involves a cascade of events, initiating with iron overload, progressing to inactivation of the SLC7A11-GPX4 system and ensuing lipid peroxidation, and culminating in characteristic mitochondrial damage. NK-NVs-SRB suppresses the PI3K-AKT-mTOR axis to sensitize resistant HCC cells to ferroptosis pathway Having confirmed that NK-NVs-SRB induces cell death through ferroptosis, we next sought to elucidate its underlying mechanisms. Analysis of the NK-NVs proteomics data revealed their regulatory role in the PI3K-AKT-mTOR signaling pathway within the cell death subpopulation (Fig. 4A). Furthermore, we performed GSEA enrichment analysis on the transcriptomic data following NK-NVs-SRB treatment and found that NK-NVs-SRB also modulated the PI3K-AKT signaling pathway (Fig. 4B). The PI3K-AKT-mTOR signaling pathway is a traditional sequence that plays a role in cell survival and growth [55]. However, its frequent dysregulation during cancer development has made it an important therapeutic target [56]. NK-NVs-SRB promotes ferroptosis by inhibiting the PI3K-AKT-mTOR signaling pathway. (A) KEGG pathway analysis shows significant enrichment of the PI3K-AKT signaling pathway in response to NK-NVs. (B) GSEA reveals a correlation between ferroptosis-related gene sets and the PI3K-AKT signaling pathway. (C-D) Comparison of PI3K-AKT-mTOR signaling axis protein expression levels and (E) mRNA expression levels in Huh/SR cells to their parental cells. (F) NK-NVs-SRB decreases both total and phosphorylated forms of PI3K, AKT and mTOR. (G) 740Y-P reverses the cytotoxic effect of NK-NVs-SRB. (H-O) 740Y-P reverses the hallmarks of ferroptosis induced by NK-NVs-SRB (including Fe²⁺, MDA, GSH, ROS, and JC-1) as well as in the protein levels of SLC7A11 and GPX4 (JC-1 scale bar = 10 μm; ROS scale bar = 100 μm). (P-R) The effect of AKT1 knockdown on ferroptosis. (S) Fer-1 does not restore the suppression of p-PI3K and p-AKT expression induced by NK-NVs-SRB (∗∗∗/### P < 0.001, ∗∗/##P < 0.01, ∗/#P < 0.05, ns, not significant.#means vs. control group) According to current studies, the PI3K-AKT-mTOR pathway is abnormally activated in HCC cells that are resistant to sorafenib [57]. We first characterized the baseline status of this pathway in our resistant model. Both mRNA and Western blot analyses confirmed hyperactivation of the PI3K-AKT-mTOR axis in resistant cells, showing significantly elevated phosphorylation levels of PI3K, AKT, and mTOR compared to parental cells (Fig. 4C-E). Remarkably, the activation state was substantially reversed by NK-NVs-SRB treatment, which markedly reduced the phosphorylation levels of PI3K, AKT, and mTOR (Fig. 4F). This powerful inhibitory action provides a mechanistic understanding of how NK-NVs-SRB restores drug sensitivity in resistant cells by directly targeting and disrupting their key survival pathway. To determine whether pathway inhibition directly causes subsequent ferroptosis, we employed the PI3K activator 740Y-P. Exposure to the PI3K activator 740Y-P prior to treatment partially reduced NK-NVs-SRB-induced cell death (Fig. 4G). Co-treatment with the activator significantly reversed the accumulation of Fe²⁺, MDA, and ROS induced by NK-NVs-SRB, while also mitigating the loss of mitochondrial membrane potential (Fig. 4H-N). This rescue experiment demonstrates that inhibition of the PI3K-AKT-mTOR axis is a core driver, not just a correlate, of the ensuing ferroptosis. The therapy interferes with this vital survival mechanism, triggering a series of events that remove the inhibition on ferroptosis, allowing lipid peroxidation to continue unchecked until cell death occurs. To establish a direct connection between PI3K-AKT pathway activation and ferroptosis execution at the protein level, we examined key markers following treatment with a PI3K activator. Consistent with the restored ferroptosis phenotype (Fig. 4O-P), 740Y-P effectively restored the phosphorylation of PI3K and AKT that had been suppressed by NK-NVs-SRB. Crucially, this recovery of upstream survival signaling coincided with increased protein expression of the key ferroptosis inhibitors SLC7A11 and GPX4. In summary, when the PI3K pathway suppressed by NK-NVs-SRB was pharmacologically reactivated, all hallmark events of ferroptosis were attenuated or prevented. Collectively, these findings demonstrate that NK-NVs-SRB blocks the PI3K-AKT-mTOR signaling pathway, causing a decrease in the antioxidant defense system (SLC7A11/GPX4). This loss of defensive capacity creates permissive conditions for lipid peroxidation to accumulate to lethal levels, ultimately triggering ferroptosis. The discovery that PI3K activation simultaneously counteracts pathway inhibition, reinstates antioxidant defenses, and prevents cell death. To directly validate the regulation of ferroptosis by the PI3K‑AKT pathway, we knocked down AKT1 in Huh7/SR cells (Figure S8). CCK‑8 assays showed that AKT1 knockdown significantly reduced cell viability, and this effect was largely reversed by co‑treatment with a ferroptosis inhibitor (Fer-1). Consistently, AKT1 knockdown markedly increased the levels of ferrous iron (Fe2+) and malondialdehyde (MDA), both of which were restored to near‑baseline levels upon ferroptosis inhibitor addition. These results indicate that AKT1 suppression alone is sufficient to recapitulate the ferroptotic phenotype induced by NK-NVs-SRB. Importantly, combined treatment with AKT1 knockdown and an AKT activator did not further alter Fe2+ or MDA levels compared with AKT1 knockdown alone, suggesting that the activator requires functional AKT1 to exert its effect (Fig. 4P-R). Furthermore, western blot analysis revealed that NK-NVs-SRB treatment decreased the expression of p‑PI3K and p‑AKT in Huh7/SR cells. Co‑incubation with Fer-1 failed to restore the phosphorylation levels of PI3K or AKT (Fig. 4S), indicating that NK-NVs-SRB’s inhibition of the PI3K-AKT signaling pathway is not an outcome of ferroptosis but rather likely an upstream event. This study conclusively demonstrates that the signal cascade reaction serves as the pivotal link between NK-NVs-SRB therapy and the occurrence of ferroptosis. More interestly, although the combination of the PI3K inhibitor LY294002 and free SRB significantly increased intracellular ferrous iron and MDA levels, its effect was substantially weaker than that of NK-NVs-SRB (Figure S9). These data demonstrate that the nanoformulation achieves a significantly stronger ferroptotic response than simple pharmacological combination. This crucial finding reveals the multidimensional superiority of the NK-NVs-SRB nanocomplex and suggests a multiple mechanistic basis for its potent activity. DAPMs induced by ferroptosis Ferroptosis triggers a major adaptive immune response system [58], which plays a critical role in immune surveillance and targeted attack against tumor cells to suppress cancer progression. Specifically, lipid oxidation during ferroptosis leads to tumor cell membrane rupture and the release of damage associated molecular patterns (DAMPs) [59]. The release of high mobility group box 1 (HMGB1) from the nucleus and the exposure of calreticulin (CRT) on the cell surface are important DAMPs that can lead to the maturation of dendritic cells (DCs) and stimulate the growth of immune cells. This adaptive immune response ultimately inhibits tumor growth [60]. Initially, we examined the release of DAMPs, such as HMGB1 and CRT, to assess the activation of adaptive immunity. In normal situations, HMGB1 is primarily situated in the nucleus, where it acts as a DNA chaperone, but it can also relocate to the cytoplasm as a typical DAMP. CRT, a crucial calcium-binding protein, is usually found in the endoplasmic reticulum. By being overexpressed and then displayed on the cell surface, this protein can boost cellular immunogenicity. Our in vitro experiments demonstrated that NK-NVs-SRB treatment led to a reduction of HMGB1 within the nucleus and its subsequent translocation to the cytoplasm, as observed by confocal laser scanning microscopy (CLSM) (Fig. 5A). This finding was further confirmed by detecting increased HMGB1 levels in the cell culture supernatant (Fig. 5B). Calreticulin, which exhibited a homogeneous distribution in the cytoplasm, was prominently exposed on the cell surface in the NK-NVs-SRB treatment group, with the most significant signal intensity observed among all groups (Fig. 5C). And ATP serves as an important molecule, functioning not only in cellular energy but also in inflammatory signaling. Released from damaged cells, ATP can activate receptors on immune cells [61]. We measured extracellular ATP release from Huh7/SR cells following NK-NVs-SRB treatment using a detection kit. Compared with other groups, the NK-NVs-SRB group showed markedly higher ATP release (Fig. 5D).These results indicate that ferroptosis triggers immunogenic cell death (ICD) by inducing tumor cells to release DAMPs. These patterns transmit an “eat me” signal to the immune surveillance system, thereby initiating the antigen presentation process. DAMPs induced by NK-NVs-SRB. (A) Representative CLSM images for HMGB1 releasing from the cell nuclear in Huh7/SR cells after coculturing with NK-NVs, SRB and NK-NVs-SRB for 24 h (scale bar = 20 μm). (B) Measurement of HMGB1 levels in cell culture supernatant by ELISA after drug treatment. (C) Representative CLSM images for CRT exposure on the cell surface in Huh7/SR cells after coculturing with NK-NVs, SRB and NK-NVs-SRB for 24 h (scale bar = 20 μm). (D) After 48 h of treatment with the drug, the ATP content in the cell changed (n = 3) (∗∗∗/### P < 0.001, ∗∗/##P < 0.01, ∗/#P < 0.05, ns, not significant. #means vs. control group) In vivo anticancer effects of NK-NVs-SRB To assess the antitumor effectiveness of NK-NVs-SRB in living organisms, mice with tumors were divided into groups and given various formulations through intraperitoneal injection. The 15-day study, spanning from tumor inoculation to termination, involved implanting Hepa1-6 cells into the livers of 6-week-old C57 mice. Treatments commenced two days post inoculation (Fig. 6A). Before conducting the therapeutic study, we monitored the in vivo distribution of NK-NVs tagged with the near-infrared fluorescent dye DiR through live imaging. As shown in Fig. 6B, a clear fluorescent signal was observed in the liver at 12 h post injection. The signal intensity peaked in the hepatic region at 24 h, indicating maximum accumulation of NK-NVs at this time point. Although the overall signal diminished by 48 h, detectable fluorescence persisted in the liver, suggesting partial retention of NK-NVs within the tumor tissue. According to these biodistribution results, NK-NVs accumulate in the liver and stay for up to 48 h, indicating their potential as a continuous drug release system. Throughout the treatment period, tumor volume was monitored every other day. Control group tumors exhibited continuous growth, whereas NK-NVs-SRB administration significantly suppressed tumor progression after the first injection. This inhibitory effect persisted until the end of the study, with the NK-NVs-SRB group showing superior therapeutic outcomes compared to other groups (Fig. 6C). At the endpoint, tumor tissues were excised and weighed to calculate the tumor growth inhibition rate. Despite the absence of significant body weight differences across all groups, marked variations in tumor suppression were observed. The average tumor weight was 1.32 ± 0.23 g in the NK-NVs-SRB group versus 2.38 ± 0.27 g in the control group (Fig. 6D-F). The TGI values ranged from 15.97% to 44.54% across treatment groups, with NK-NVs-SRB demonstrating the most potent suppression. These results confirm the synergistic antitumor effect of NK-NVs-SRB. In vivo anticancer efficacy study. (A) Experimental flow chart (created in BioRender). (B) The distribution of NK-NVs in organs was observed at 6, 12, 24, and 48 h following intraperitoneal administration of NK-NVs. (C) Tumor fluorescence intensity in mice from different groups was measured after treatment. (D) Body weight of different treatment groups during the study (n = 5). (E) Mean tumor weight of each treatment group at the terminal time point (n = 5). (F) The anatomical image of the mouse tumor after the treatment. (G) H&E and immunohistochemistry staining images of tumors (H&E scale bar = 312.5 μm; Ki67 scale bar = 50 μm) Further in vivo evaluation of antitumor efficacy was performed through H&E staining and immunohistochemical analysis of tumor tissues. The marker Ki67, which is widely used to indicate cell proliferation [62], demonstrated reduced positive staining in all treatment groups (NK-NVs, SRB, and NK-NVs-SRB), with the NK-NVs-SRB group having the most notable decline in cell proliferation (Fig. 6G). We next investigated whether the potent tumor suppression observed in mice treated with NK-NVs-SRB was driven by the same ferroptosis mechanism identified in our in vitro experiments. We examined the expression of key ferroptosis‑related proteins and PI3K‑AKT‑mTOR signaling components in mouse HCC tissues using mIHC and western blotting. mIHC analysis revealed that, compared with the other groups, NK-NVs-SRB group exhibited the lowest expression levels of SLC7A11 and GPX4 (Fig. 7A-B). Consistently, the levels of ferrous iron (Fe²⁺) in tumor tissues, were highest in the NK-NVs-SRB group (Fig. 7C). These results indicate that NK-NVs-SRB effectively suppresses the cystine‑glutamate antiporter system and GPX4‑mediated antioxidant defense in vivo, thereby promoting iron accumulation and lipid peroxidation. And parallel mIHC staining further showed that the protein levels of p‑PI3K, p‑AKT, and p‑mTOR were also markedly decreased in the NK-NVs-SRB group compared with others (Fig. 7D-G). Western blot analysis of total tumor lysates corroborated these findings, confirming that NK-NVs-SRB significantly downregulated p‑PI3K, p‑AKT, SLC7A11, and GPX4 (Fig. 7H-I). Collectively, these in vivo data are highly consistent with our cellular observations that NK-NVs-SRB suppresses the PI3K‑AKT‑mTOR axis, which in turn leads to reduced expression of SLC7A11 and GPX4, culminating in ferroptosis. Ferroptosis and ICD effects induced by NK-NVs-SRB in vivo. (A-B) Multiplex IF staining for detecting ferroptosis-related protein levels in tumor tissues across different groups (scale bar = 50 μm). (C) Ferrous iron (Fe²⁺) levels in tumor tissues (n = 3). (D-G) Multiplex immunofluorescence staining was performed to observe the expression of p-PI3K, p-AKT, and p-mTOR in tumor tissues of different groups (scale bar = 50 μm). (H-I) Expression of PI3K-AKT pathway and ferroptosis-related proteins in tumor tissues of each group detected by Western blot. (J) The relocalization of HMGB1 and CRT in tumor tissues from different treatment groups (NK-NVs, SRB, NK-NVs-SRB) was assessed by immunohistochemistry (scale bar = 50 μm). Assessment of DC maturation (L-M) and CD8 + T cell activation (N-O) within tumor tissues by flow cytometry. (P-Q) Differences in CD8 + T cell infiltration among various treatment groups were observed by multiplex immunofluorescence (scale bar = 50 μm) Recent research suggests that nanoparticles that encourage the release of DAMPs can work effectively with checkpoint inhibitors by reversing T cell exhaustion and boosting cytotoxic responses [63]. Our in vitro data show that NK-NVs-SRB induced ferroptosis is accompanied by DAMPs release. We next investigated whether this process initiates an effective antitumor immune response in vivo. Immunohistochemical analysis of mouse tumor tissues revealed more pronounced relocalization of HMGB1 and CRT from the nucleus to the extracellular space in the NK-NVs-SRB treated group (Fig. 7J). This shift in the distribution of these key DAMPs clearly indicates ICD. These findings suggest that NK-NVs-SRB effectively sends “danger” and “eat me” signals to the immune system while killing tumor cells in vivo. As crucial immune sentinels, DCs were analyzed by flow cytometry. Tumors from the NK-NVs-SRB group showed significantly elevated levels of the costimulatory molecules CD86 and CD80 on DCs (Fig. 7L-M). These molecules are critical markers of DC functional maturation. The results demonstrate that ICD induced by NK-NVs-SRB successfully converts immature DCs into functionally mature DCs, enabling efficient T cell activation. As expected, this DC maturation promoted the recruitment of adaptive immune effector cells. Both flow cytometry and multiplex immunofluorescence staining consistently showed a substantial increase in the number and density of CD8 + T cells within tumors from the NK-NVs-SRB group compared to other treatment groups (Fig. 7N-Q). These findings explain the potent and sustained antitumor efficacy of NK-NVs-SRB. Beyond direct cytotoxicity, NK-NVs-SRB enhances antitumor immune responses through DAMPs release, effectively suppressing tumor progression and significantly improving therapeutic outcomes. The safety profile of NK-NVs-SRB was systematically evaluated both in vitro and in vivo. Body weight was monitored throughout the experimental period to assess systemic toxicity, with no statistically significant differences observed among all groups. Upon study completion, serum was collected for biochemical analysis. Liver and kidney function parameters in peripheral blood remained within normal limits across treatment groups (Fig. 8A-C), indicating negligible hepatorenal toxicity. Furthermore, H&E staining of major organs revealed no apparent morphological abnormalities in any treatment group (Fig. 8D). These collective findings demonstrate the favorable biosafety profile of NK-NVs-SRB and support its potential for clinical translation. In vivo biocompatibility assessment. Blood biochemistry indicators for the function of the liver (A, B) and kidney (C) (n = 3). (D) Representative pictures of H&E stainstaining of mice organs (scale bar = 312.5 μm) Conclusions By loading sorafenib into naturally tumor targeting NK cell derived vesicles, we have successfully constructed a synergistic therapeutic nanoplatform that integrates dual functions of direct tumor killing and immune activation. This nanocomplex achieves precise intracellular delivery via endocytosis and triggers robust ferroptosis by inhibiting the PI3K-AKT-mTOR pathway, subsequently activating CD8 + T cell mediated long-term immune surveillance. Our work transcends the conventional boundaries between targeted therapy and immunotherapy, offering a novel integrated approach to enhance sorafenib sensitivity in hepatocellular carcinoma. Electronic Supplementary Material Supplementary Material 1 Acknowledgements Not applicable. Author contributions Conceptualization: CLG, SZ, XZ. Methodology: CLG, SJZ, DL, ZLY, YTF. Software: XLL, TS. Formal analysis: CLG, DL, ZLY, YTF, TG. Investigation: CLG, TYF, CYL. Project administration: XZ, SZ. Writing original draft preparation: CLG. Writing-review and editing: XL, JZ. Supervision: XZ, SZ. All authors read and approved the final manuscript. Funding The National Natural Science Foundation of China (Grant No: 82260535); Guizhou Provincial Science and Technology Projects (Grant No: ZK[2024]039), (Qiankehe Basic MS [2026]587); Guizhou Province High-Quality Development Medical Research Joint Fund (Grant No: 2024GZYXKYJJXM0020); Tumor Immunotherapy Technology Engineering Research Center (University Engineering Center 2024 [001]); Guizhou Provincial Science and Technology Support Program (Qiankehe Support (2025) General 126); 2024 Guizhou Provincial Graduate Scientific Research Foundation (2024YJSKYJJ275); Science and Technology Foundation Project of Guizhou Health Committee, Grant Number: 2024GZWJKJXM0771, 2026GZWJKJXM1278. Data availability The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval This animal experiment has been approved by the Ethics Committee of Guizhou Medical University, with approval number NO. 2502825. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Xing Zhao, Email: 351497622@qq.com. Shi Zuo, Email: drzuoshi@gmc.edu.cn. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.