In‐Outside Hypoxia Remodeling Strategy Triggers Ferroptosis to Boost TACE Therapy in Hepatocellular Carcinoma

ABSTRACT

The therapeutic efficacy of transarterial chemoembolization (TACE) is strongly limited by the hypoxic and immunosuppressive tumor microenvironment (TME). Herein, monodisperse perfluorooctyl bromide (PFOB)‐loaded fucoidan/gelatin microspheres (P@FGMS) with excellent oxygen‐carrying capability were prepared for enhanced TACE treatment of hepatocellular carcinoma (HCC). It was found that the introduction of fucoidan endowed P@FGMS with excellent electrostatic affinity for positively charged doxorubicin (DOX) and metformin (MET), enabling their efficient co‐loading and sustained release. The resulting drug‐loaded microspheres (DMP@FGMS) largely enhanced chemotherapy sensitivity through an in‐outside hypoxia remodeling strategy: the supplement of exogenous O₂ via PFOB and inhibition of intracellular oxygen consumption by MET. Notably, DMP@FGMS could not only modulate hypoxic TME, but also generate intracellular ROS and disrupt the redox homeostasis, thereby triggering cancer ferroptosis. Moreover, DMP@FGMS effectively remodeled the immunosuppressive TME by alleviating tumor hypoxia and inducing immunogenic cell death (ICD), thereby enhancing CD8⁺ T cell infiltration while reducing immunosuppressive cell populations. Consequently, DMP@FGMS could effectively suppress tumor growth in an orthotopic HCC rat model, with this effect further potentiated by combination with immune checkpoint blockade therapy. Therefore, this study demonstrates that DMP@FGMS is a potent embolic agent capable of remodeling the hypoxic/immunosuppressive TME, representing a promising strategy for TACE treatment of HCC.

Hypoxia and immunosuppression limit TACE efficacy in HCC. Oxygen‐carrying fucoidan‐gelatin microspheres (DMP@FGMS) co‐load doxorubicin and metformin to achieve inside‐out hypoxia relief: exogenous O₂ supply and inhibition of oxygen consumption. This induces ferroptosis via ROS, alleviates immunosuppression through immunogenic cell death, and enhances CD8⁺ T cell infiltration. Combined with immune checkpoint blockade, DMP@FGMS suppresses tumor growth, providing a potent embolic strategy.

1. Introduction

Hepatocellular carcinoma (HCC), accounting for 75%–85% of primary liver cancer cases, is characterized by high incidence and mortality worldwide [1]. Due to its asymptomatic nature in the early stage, more than 70% of patients are diagnosed at intermediate or advanced stages, thereby missing the optimal window for surgical resection [1, 2]. Transarterial chemoembolization (TACE), which blocks the arterial blood supply to tumors while enabling the sustained local release of chemotherapeutic agents, has become the standard therapeutic approach for patients with unresectable HCC [3, 4, 5]. Beyond its mechanical and cytotoxic effects, accumulating evidence indicates that TACE can also induce immunogenic cell death (ICD) in tumor cell [4, 6]. This process is characterized by the release of tumor‐associated antigens, damage‐associated molecular patterns (DAMPs), and pro‐inflammatory cytokines into the tumor microenvironment (TME), thereby activating dendritic cells (DCs) and initiating tumor‐specific T‐cell immune responses [7]. However, the clinical efficacy of TACE is often suboptimal, primarily due to the highly complex tumor microenvironment (TME) [8].

Hypoxia, a hallmark characteristic of the TME in solid tumors including HCC, serves as a critical reason for the suboptimal clinical efficacy of TACE [9]. Under hypoxia conditions, malignant behaviors such as tumor cells proliferation, migration, invasion, and epithelial‐mesenchymal transition (EMT) are markedly enhanced, thereby accelerating disease progression [10]. The central regulatory factor in this process is hypoxia‐inducible factor‐1α (HIF‐1α), which is activated under low‐oxygen conditions and functions as a master transcriptional regulator orchestrating the expression of a broad array of hypoxia‐responsive genes [10, 11]. For instance, HIF‐1α upregulates vascular endothelial growth factor A (VEGF‐A) to promote tumor angiogenesis, while also enhancing the expression of efflux transporters such as P‐glycoprotein (P‐gp), thereby contributing to chemoresistance [12, 13, 14]. In addition to directly promoting malignant tumor progression, the hypoxic microenvironment also markedly compromises the efficacy of immunotherapy [15]. Specifically, it suppresses the infiltration of CD8⁺ T cells, increases the proportion of immunosuppressive cell populations, and drives the polarization of tumor‐associated macrophages (TAMs) toward an anti‐inflammatory phenotype, collectively contributing to the establishment of an immunosuppressive microenvironment [16]. Notably, although TACE kills tumor cells, it simultaneously aggravates intratumoral hypoxia, which fosters chemoresistance and immune suppression and ultimately constrains overall therapeutic efficacy [17, 18]. Accordingly, relieving tumor hypoxia has emerged as a key strategy to enhance the benefit of TACE.

The direct delivery of exogenous oxygen to the tumor site represents an effective strategy for relieving tumor hypoxia. Perfluorochemicals (PFCs) have been extensively investigated as potential artificial blood substitutes due to their excellent biocompatibility and high oxygen‐binding capacity [19, 20]. These compounds can reversibly absorb, and release oxygen based on the partial pressure of oxygen, allowing for the automatic regulation of oxygen levels within tumors [21]. However, PFCs‐based oxygen delivery alone is inadequate for prolonged modulation of the tumor hypoxic microenvironment. As an FDA‐approved drug widely used for type II diabetes, metformin (MET) has been shown to alleviate tumor hypoxia by inhibiting mitochondrial complex I and consequently reducing oxygen consumption, which significantly improves outcomes in chemotherapy, immunotherapy, photodynamic therapy, and related treatments [22, 23]. Therefore, we propose a cooperative “inside‐outside” hypoxia remodeling strategy by combining PFCs with MET to establish a synergistic “oxygen‐supplementation and oxygen‐sparing” paradigm: PFCs act as external oxygen donors to increase local oxygen supply, while MET reduces cellular oxygen demand from within, together achieving fundamental modulation of the hypoxic tumor microenvironment and markedly potentiating the therapeutic efficacy of TACE.

In this study, monodisperse perfluorooctyl bromide (PFOB)‐loaded fucoidan/gelatin microspheres (P@FGMS) with an excellent oxygen‐carrying capability were first fabricated using a simplified microfluidic system. Fucoidan is a natural sulfated polysaccharide that not only facilitates the adsorption of positively charged DOX and MET via electrostatic interactions but also provides sustained drug release properties to the microspheres. Then, the DOX and MET‐loaded P@FGMS (DMP@FGMS) was employed as a novel embolic agent for enhanced TACE therapy (Scheme 1). As an in‐outside hypoxia remodeling strategy, DMP@FGMS significantly enhances chemotherapy sensitivity. Notably, DMP@FGMS triggers a burst accumulation of ROS, which subsequently disrupts redox homeostasis and ultimately activates the ferroptosis pathway. Concurrently, by alleviating hypoxia and inducing ICD, DMP@FGMS enhances CD8⁺ T cells infiltration and reduces immunosuppressive cell populations, thereby effectively reversing the immunosuppressive state. Animal experiments confirmed that DMP@FGMS significantly inhibited tumor growth in an orthotopic HCC rat model, and its therapeutic efficacy was further enhanced when combined with immune checkpoint blockade therapy. Overall, this study demonstrates that DMP@FGMS serves as a promising embolic agent, offering a novel and effective strategy for safe interventional treatment of HCC.

SCHEME 1.

Schematic illustration showing the application of DMP@FGMS + αPD‐L1 for enhanced TACE‐immunotherapy of HCC.

2. Results and Discussion

2.1. Hypoxia Promotes HCC Progression and Chemoresistance

Chemotherapy resistance poses a major challenge in the clinical treatment of HCC. To systematically investigate the relationship between transcriptomic alterations and chemotherapy resistance in HCC, we performed KEGG pathway enrichment analysis using data from The Cancer Genome Atlas (TCGA) (Figure 1A). The results revealed significant enrichment of multiple metabolic pathways in HCC tissues, including “Drug metabolism‐cytochrome P450,” “Drug metabolism‐other enzymes,” and “ABC transporters” (Figure 1B). These findings indicate substantial remodeling of the drug metabolism system in HCC cells, particularly in pathways involved in drug metabolism, suggesting their potential critical role in mediating chemotherapy resistance.

FIGURE 1.

Hypoxia induces chemotherapy resistance in tumors. (A) Research Workflow. (B) KEGG pathway enrichment analysis between tumor tissues (n = 369) and normal tissues (n = 160) of HCC patients from the TCGA TARGET GTEx database. (C,D) GSEA (gene sets “KEGG_APOPTOSIS” and KEGG_ABC_TRANSPORTERS). (E) Expression levels of ABCB1 in tumor tissues and normal tissues. (F) Kaplan‐Meier survival curves for overall survival (OS) in HCC patients with low versus high expression of ABCB1. (G) Expression levels of HIF1A in normal tissues and tumor tissues. (H) Validation of the protein expression levels of HIF1A in HCC tissues compared with normal tissues by HPA database. (I) Expression of HIF1A in various stages of liver cancer. (J) Kaplan‐Meier survival curves for overall survival (OS) in HCC patients with low versus high expression of HIF1A. (K) The Spearman correlation of gene expression between HIF1A and ABCB1. (L) The abundance of CD8⁺ T cells in normal tissues and tumor tissues in HCC patients. (M) Correlation analysis of HIF1A gene expression with CD8⁺ T cell infiltration levels in normal tissues and tumor tissues in HCC patients. (N) Immunofluorescence images of DOX treated Hepa1‐6 cells under hypoxic and normoxic conditions (Pimonidazole, VEGF, P‐gp, HIF‐1α). (O) Cytotoxicity of DOX on Hepa1‐6 cells under hypoxic and normoxic conditions. (P) Flow cytometry plots and Quantification of proliferation CD8⁺ T cells under hypoxic and normoxic conditions. (Q) Flow cytometry plots and Quantification of TIM‐3⁺ population in CD8⁺ T cells under hypoxic and normoxic conditions. Data are presented as mean ± SD (n = 3). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Further comparison of expression profiles between HCC and normal liver tissues using gene set enrichment analysis (GSEA) demonstrated that drug metabolism pathways were markedly upregulated in HCC, while apoptosis pathways were significantly suppressed (Figure 1C,D). This suggests that HCC cells may enhance their capacity for drug detoxification and elimination, thereby attenuating chemotherapy‐induced apoptotic cell death. Additionally, analysis of the multidrug resistance gene ABCB1 revealed significantly elevated expression in HCC compared to normal liver tissues (Figure 1E). Kaplan‐Meier survival analysis further indicated that high ABCB1 expression was associated with significantly shorter overall survival in patients (Figure 1F), providing clinical evidence supporting the crucial role of ABCB1 in multidrug resistance in HCC. These findings are consistent with the activation of the “ABC transporters” pathway identified in the KEGG analysis, collectively suggesting that HCC cells may establish a multidrug‐resistant phenotype through the coordinated upregulation of both drug‐metabolizing enzymes and efflux transporters.

The tumor hypoxic microenvironment serves as a critical background for regulating drug resistance [24]. We observed that HIF1A expression is elevated in HCC (Figure 1G,H) and increases with advancing tumor stage (Figure 1I). Kaplan‐Meier survival analysis further demonstrated that high HIF1A expression was significantly associated with poorer patient prognosis (Figure 1J). Notably, a significant positive correlation was found between HIF1A and ABCB1 expression (correlation coefficient = 0.164, p value = 0.00158), suggesting that HIF1A may participate in regulating ABCB1‐mediated drug resistance (Figure 1K). To further explore the underlying mechanism, we analyzed the tumor immune microenvironment and found that CD8⁺ T cell infiltration was significantly reduced in HCC (Figure 1L) and negatively correlated with HIF1A expression (Figure 1M), indicating that HIF1A may impair antitumor immune responses by suppressing CD8⁺ T cell infiltration, thereby affecting treatment efficacy.

To functionally validate the regulatory role of hypoxia on the HIF1A‐ABCB1 axis, we used pimonidazole to label hypoxic regions in hepatocellular carcinoma cell lines and compared protein expression under normoxic and hypoxic conditions. Immunofluorescence results demonstrated that hypoxia significantly up‐regulated HIF1A expression, accompanied by synchronous increases in ABCB1 and VEGF levels (Figure 1N). Cytotoxicity assays further revealed that the cytotoxic effects of doxorubicin (DOX) on hepatocellular carcinoma cells were significantly attenuated under hypoxic conditions, suggesting that the hypoxia‐induced HIF1A‐ABCB1 pathway may enhance cellular drug resistance (Figure 1O, S1).

Additionally, to investigate the impact of hypoxia on T cell function, we assessed CD8⁺ T cell proliferation using CFSE labeling and found that hypoxia markedly inhibited their proliferative capacity (Figure 1P). Flow cytometry analysis further showed a significant increase in the proportion of exhausted CD8⁺ T cells (TIM‐3⁺) under hypoxic conditions, which may be an important factor contributing to T cell dysfunction (Figure 1Q). In summary, this study systematically reveals, through both clinical data and in vitro experiments, the potential mechanism by which hypoxia promotes multidrug resistance and an immunosuppressive microenvironment in HCC via activation of the HIF1A‐ABCB1 axis, providing new experimental evidence for understanding chemotherapy resistance in HCC.

2.2. Preparation and Characterization of PFOB‐Loaded Fucoidan/Gelatin Microspheres (P@FGMS)

P@FGMS was prepared according to the procedure shown in Figure 2A. To achieve uniform dispersion of the hydrophobic liquid PFOB, nanoemulsions with an average diameter of 327.63 ± 19.14 nm were first prepared via probe‐assisted ultrasonication (Figure 2B, S2). Subsequently, monodisperse P@FGMS microspheres were fabricated using a microfluidic device. The prepared P@FGMS showed a regular spherical structure with an average diameter of 62.69 ± 2.25 µm and excellent size uniformity, as indicated by a coefficient of variation (CV) of 3.58% (Figure 2B,C). In contrast, the blank gelatin microspheres (GMS) showed an average particle size of 48.47 ± 3.71 µm with a CV of 7.65% (Figure 2B,C). The difference in particle size is likely caused by changes in the viscosity or interfacial tension at the interface of the dispersed and continuous phases upon incorporation of fucoidan and PFOB nanoemulsions. Next, the elemental composition of P@FGMS was further characterized by energy‐dispersive X‐ray spectroscopy (EDS), which demonstrated uniform spatial distribution of sulfur (S), oxygen (O), nitrogen (N), bromine (Br), and fluorine (F) within the microsphere (Figure S3). These results confirm the successful incorporation of PFOB (indicated by F and Br signals) and fucoidan (indicated by the S signal) into the gelatin microspheres. Subsequently, the swelling capacity of P@FGMS was evaluated by immersing the microspheres in water. It was observed that the particle sizes of P@FGMS, P@GMS, and GMS increased significantly after the swelling capacity test (Figure S4). The swelling capacity of P@FGMS, P@GMS, and GMS was calculated to be 121.67 ± 8.50%, 174.67 ± 7.57%, and 174 ± 13.45%, respectively (Figure S5). The decreased swelling capacity of P@FGMS may be attributed to electrostatic interactions or crosslinking between fucoidan's anionic functional groups and the amino groups of gelatins, which limits matrix hydration.

FIGURE 2.

Characteristics of DMP@FGMS. (A) Schematic illustration of the preparation DMP@FGMS with a simplified microfluidic device. (B) TEM images of PFOB‐NE (left). SEM images of GMS and P@FGMS (right). (C) Size distribution of GMS and P@FGMS (n = 100). (D) The efficiency of MET and DOX loading of GMS or P@FGMS. (E) The release behavior of MET and (F) DOX of DM@GMS or DMP@FGMS. (G) Oxygen concentration at each time points after the introduction of P@FGMS, GMS, or PFOB‐NE into pure oxygen. (H) The change in oxygen concentration over time after the addition of P@FGMS, G@MS, or PFOB‐NE to deoxygenated water. (I) Schematic diagram of decellularized liver model. (J) Macroscopic appearance of a fresh liver, before and after decellularization. (K) Decellularized liver model: Microsphere distribution within the vascular network under stereomicroscopy or fluorescence microscopy. (L) Rabbit kidney embolism model: DSA images of the rabbit kidney before (left) and after (right) embolization. (M) The computed tomography angiography of rabbit at 2‐week post embolization. Data are presented as mean ± SD (n = 3).

Fucoidan is a highly negatively charged polysaccharide that may enable the microspheres to efficiently adsorb positively charged DOX and MET via electrostatic interactions [25]. To verify this hypothesis, the drug loading capacities of P@FGMS and P@GMS were compared. As shown in Figure 2D, the encapsulation efficiencies of P@GMS for DOX and MET were 40.46 ± 1.84% and 21.92 ± 0.55%, respectively. In contrast, the drug loading capability of P@FGMS was remarkably higher, with encapsulation efficiencies of 95.17 ± 1.48% for DOX and 59.43 ± 2.84% for MET, respectively. Afterward, in vitro drug release behavior of P@FGMS and P@GMS was displayed in Figure 2E,F, both DOX and MET could be quickly released from P@GMS, as more than 80% total drug content released within the first 3 days. In contrast, P@FGMS demonstrated a sustained release behavior, with approximately 60% of DOX and 75% of MET released within 14 days. Collectively, these results demonstrate that fucoidan serves a dual role by significantly enhancing the microspheres' ability to encapsulate DOX and MET while simultaneously promoting a sustained drug release profile.

PFOB has attracted considerable interest as a blood oxygen carrier substitute due to its exceptional oxygen‐carrying capacity and excellent biocompatibility [26]. The oxygen loading and release kinetics of the microspheres were assessed by monitoring the dissolved oxygen concentration at various time intervals after exposure to pure oxygen, using a calibrated dissolved oxygen meter. As expected, GMS alone showed negligible ability to retain or carry oxygen (Figure 2G). Conversely, P@FGMS displayed a pronounced ability to rapidly load oxygen, achieving a maximum dissolved oxygen concentration of around 25 mg/L within 6min. This result highlights the indispensable role of PFOB in imparting strong oxygen‐capturing ability to the microspheres. The dissolved oxygen concentration exceeding the equilibrium solubility of water may be explained by the high oxygen affinity of PFOB. Oxygen enriched in the PFOB phase can be gradually released into the surrounding aqueous medium, thereby creating a transient oxygen‐supersaturated microenvironment, which is consistent with previous reports on perfluorocarbon‐based oxygen delivery systems [27, 28, 29]. Distinct from the rapid release behavior observed in PFOB nanoemulsions, P@FGMS served as a reservoir for controlled and prolonged oxygen release (Figure 2H). Taken together, the high oxygen‐loading efficiency and sustained release characteristics support the use of P@FGMS as a promising oxygen carrier for hypoxia‐targeted cancer treatment.

The embolization performance of DMP@FGMS was first evaluated using a rat decellularized liver model [30]. As displayed in Figure 2I,J, the decellularized liver model with well‐preserved vascular architecture was successfully generated via a rapid perfusion protocol as previously reported [30]. It was found that DMP@FGMS was able to access peripheral vascular structures following injection into the decellularized rat liver model (Figure 2K). This finding strongly suggests that DMP@FGMS possesses an excellent capacity to effectively embolize tumor‐supplying arteries, highlighting its potential as a promising embolic agent for HCC. Subsequently, in vivo embolization effect of DMP@FGMS was investigated using a rabbit kidney model. The feasibility of delivering DMP@FGMS via a microcatheter to the left renal artery was confirmed under a real‐time DSA navigation. Prior to embolization, clear vascular imaging of the rabbit kidney was observed (Figure 2L). Post‐embolization imaging revealed complete vascular disappearance (Figure 2L), confirming the high embolization efficacy of DMP@FGMS. Furthermore, CTA performed 2 weeks post‐treatment showed that the left kidney was no longer visible (Figure 2M), confirming the long‐term embolic efficacy of DMP@FGMS. To further evaluate the embolization efficiency of DMP@FGMS in vivo, the kidneys were harvested 2 weeks after embolization. The normal right kidney appeared elastic and bright red, while the embolized left kidney turned white due to ischemic necrosis (Figure S6).

Overall, the incorporation of fucoidan into DMP@FGMS was not only designed to enhance drug‐loading capacity and achieve sustained drug release, but also to regulate the internal network structure through electrostatic interactions with gelatin. Although the introduction of fucoidan reduced the swelling behavior of the microspheres, DMP@FGMS still exhibited acceptable injectability as well as excellent embolization performance both in vitro and in vivo. These results collectively demonstrate that the incorporation of fucoidan did not compromise the embolic capability of the microspheres, while simultaneously providing additional structural and drug‐delivery advantages for HCC therapy.

2.3. DMP@FGMS Reverses Hypoxia‐Induced Chemotherapy Resistance

To assess the hypoxia‐alleviating capability, the hypoxic Hepa1‐6 cells were treated with various types of DOX‐unloaded microspheres, including P@FGMS, M@FGMS and MP@FGMS. After the treatment, the cells were stained with pimonidazole, HIF‐1α, and VEGF for immunofluorescence analysis. As shown in Figure 3A, a significantly intense pimonidazole fluorescence signal was observed in Hepa1‐6 cells cultured under hypoxia in the PBS group. The fluorescence intensity of pimonidazole was significantly reduced after the cells were treated with P@FGMS and M@FGMS. Notably, the hypoxic cells treated with MP@FGMS displayed the weakest pimonidazole signal when compared to the other groups (Figure 3A). Consistently, MP@FGMS exhibited the strongest ability to reduce the expression levels of HIF‐1α and VEGF within the hypoxic cells, as confirmed by both immunofluorescence and western blot analyses (Figure 3B–D, S7). The ABCB1 gene encodes P‐glycoprotein (P‐gp), an efflux pump that drives tumor drug resistance. Western blot results further showed that the MP@FGMS‐treated cells had the lowest expression level of P‐gp protein (Figure 3C,D). Next, in vivo toxicity of different types of DOX‐loaded microspheres against hypoxic Hepa1‐6 cells was evaluated using the MTT assay. As shown in Figure 3E, the cytotoxicity of DOX‐loaded microspheres followed a concentration‐dependent manner, regardless of the type of microsphere. According to MTT results, the half maximal inhibitory concentration (IC₅₀) values of D@FGMS, DP@FGMS, DM@FGMS and DMP@FGMS were calculated to be 255.50, 114.80, 140.50 and 53.02 µg/mL, respectively (Figure 3E). In addition, the apoptosis and live‐dead cell assays further confirmed the best in vitro anticancer efficacy of DMP@FGMS compared to the other DOX‐loaded microspheres (Figure 3F,G, S8). Taken together, these findings the encapsulation of PFOB and MET endows the microspheres with a superior ability to effectively alleviate hypoxia and overcome hypoxia‐induced drug resistance in vitro (Figure 3H).

FIGURE 3.

DMP@FGMS reverses hypoxia‐induced chemotherapy resistance. (A,B) Immunofluorescence staining of Hepa1‐6 cells with different treatments using pimobendazole (A) or HIF‐1α (B). (C,D) Expression of P‐gp, HIF‐1α, and VEGF proteins in Hepa1‐6 cells under different treatments. Protein expression levels were independently normalized to GAPDH and subsequently calculated relative to the corresponding control group for each target protein. (E) MTT assay to assess the toxicity of different microspheres to Hepa1‐6 cells under hypoxic conditions. (F,G) Annexin V‐FITC/7‐AAD assay to assess the apoptosis of Hepa1‐6 cells treated with different microspheres under hypoxic conditions. (H) Schematic illustration of DMP@FGMS enhancing chemotherapy sensitivity. (I) Volcano plot of differentially expressed genes (DEGs) in Hepa1‐6 cells treated with DMP@FGMS under hypoxic conditions. (J) Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of DEGs between the two groups. (K) Heatmap of hypoxia‐related DEGs expression between the two groups. (L‐M) GSEA (gene sets “ABC transporter pathway” and “Ferroptosis pathway”). Data are presented as mean ± SD (n = 3). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

To clarify the mechanisms driving the increased chemotherapy sensitivity of DMP@FGMS in HCC cells, RNA‐seq analysis was conducted between the control group and the DMP@FGMS group. It was found that DMP@FGMS treatment resulted in a total of 4712 differentially expressed genes (DEGs), including 2167 up‐regulated genes and 2545 down‐regulated genes (Figure 3I, S9 left). The heatmap displayed the DEGs between the control and DMP@FGMS group (Figure S9 right). Based on the observed differences, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were performed to elucidate the biological roles of differentially expressed mRNAs and their associated pathways. KEGG analysis revealed the anti‐HCC effect of DMP@FGMS was associated with various signaling pathway, including Ferroptosis, Chemical carcinogenesis‐reactive oxygen, HIF‐1 signaling pathway, Response to hypoxia and ABC transporters (Figure 3J). GO classification showed that DEGs were involved in various biological processes, such as rRNA processing, apoptotic processes, and autophagy (Figure S10). In terms of ‘Molecular Function’, DEGs were chiefly related to ‘binding’, including protein binding, nucleotide binding, and ATP binding (Figure S10). In the ‘Cellular Component’ category, DEGs were predominantly associated with the cytoplasm, nucleus, and cytosol (Figure S10). In addition, we found that 18 hypoxia‐related genes in cells were significantly downregulated following treatment with DMP@FGMS (Figure 3K). GSEA revealed a significant downregulation in the gene expression of the ABC transporter pathway following DMP@FGMS treatment compared to the control group (Figure 3L). The suppression of the ABC transporter pathway, a major mediator of multidrug resistance where members like ABCB1 significantly reduce intracellular chemotherapeutic drug concentrations through active efflux, indicates that DMP@FGMS effectively counteracts hypoxia‐induced chemoresistance, thereby establishing a key condition for enhancing chemotherapeutic efficacy. Notably, transcriptome sequencing uncovered a pivotal finding: the ferroptosis pathway was markedly activated in DMP@FGMS‐treated tumor cells (Figure 3M). This observation suggests that, beyond engaging traditional apoptotic pathways, DMP@FGMS exerts its antitumor effects by inducing a form of regulated cell death characterized by the accumulation of lipid peroxides. This discovery significantly expands our understanding of the therapeutic mechanism, indicating that the strategy enhances treatment efficacy by cooperatively triggering distinct cell death programs.

2.4. DMP@FGMS Activating Ferroptosis Through Oxidative Stress

To validate the finding of ferroptosis pathway activation from RNA sequencing, we systematically evaluated the impact of DMP@FGMS on cellular oxidative stress levels under hypoxic conditions. Using DCFH‐DA (DCF) as a fluorescent probe for intracellular reactive oxygen species (ROS) and C11‐BODIPY (C11) for lipid peroxidation, we assessed the oxidative stress status of Hepa1‐6 cells by flow cytometry and confocal microscopy. The results showed no significant difference in fluorescence intensity between Group I and Group II, indicating that both normoxic and hypoxic cells maintained basal levels of ROS and lipid peroxidation (Figure 4A,B, S11 and S12). Compared with Group II, D@FGMS treatment significantly increased the fluorescence intensities of DCF and C11, demonstrating a marked elevation in intracellular oxidative stress (Figure 4A,B, S11 and S12). This enhanced oxidative stress effect was further amplified in the DP@FGMS groups, suggesting that combining DOX with exogenous oxygen could collectively exacerbate cellular oxidative stress (Figure 4A,B, S11 and S12).

FIGURE 4.

DMP@FGMS Activating ferroptosis through oxidative stress. (A,B) Intracellular ROS and lipid peroxidation levels were measured by flow cytometry. (C) Representative confocal images of ROS generated from mitochondria stained with MitoSox. (D) Representative confocal image of mitochondrial membrane potential stained with JC‐1. (E) GSH/GSSG ratios of Hepa1‐6 cells treated with different microspheres. (F) MDA level of Hepa1‐6 cells treated with different microspheres. (G) Expression of GPX4, SLC7A11, and GAPDH proteins in Hepa1‐6 cells under different treatments. (H). Cell viability of DMP@FGMS‐treated cells with or without Fer‐1. (I). Schematic illustration of DMP@FGMS‐induced ferroptosis. Data are presented as mean ± SD (n = 3). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

As the primary source of intracellular ROS, mitochondria play a critical role in maintaining cellular redox homeostasis, and their structural and functional integrity is linked to this balance [31]. To clarify the mechanism by which DMP@FGMS induces oxidative stress in hypoxic Hepa1‐6 cells, we further focused on mitochondrial function and the associated oxidative stress status. MitoSOX Red staining revealed that DMP@FGMS treatment significantly increased mitochondrial ROS levels compared to the DP@FGMS group (Figure 4C, S13). This effect can be attributed to MET induced inhibition of mitochondrial complex I, which subsequently leads to mitochondrial dysfunction and elevated ROS production. JC‐1 staining further demonstrated a notable decrease in red fluorescence and a corresponding increase in green fluorescence in the DMP@FGMS group, indicating severe mitochondrial membrane depolarization and functional impairment (Figure 4D).

Furthermore, the GSH/GSSG ratio was the lowest in the DMP@FGMS‐treated group, indicating a severe depletion of the cellular antioxidant defense capacity (Figure 4E). The resulting collapse of the antioxidant system led to uncontrolled lipid peroxidation, as evidenced by a significant increase in malondialdehyde (MDA) levels (Figure 4F). These results collectively demonstrate that DMP@FGMS causes severe disruption of cellular redox homeostasis. Western blot analysis confirmed the activation of ferroptosis, with significant downregulation of the key proteins GPX4 and SLC7A11, consistent with the transcriptional sequencing data at the protein level (Figure 4G, S14). To determine whether ferroptosis contributed to DMP@FGMS‐induced cell death, we applied the specific ferroptosis inhibitor Ferrostatin‐1 (Fer‐1). The protective effect of Fer‐1, which partially reversed the cytotoxicity of DMP@FGMS and enhanced cell survival, provides additional evidence underscoring the contribution of ferroptosis (Figure 4H). Based on these findings, we propose the following mechanism: (1) The combination of DOX and exogenous oxygen exacerbates cellular oxidative stress; (2) MET‐mediated complex I inhibition induces mitochondrial dysfunction, resulting in increased ROS production; (3) The accumulation of ROS and the decline in cellular antioxidant capacity induced by DMP@FGMS collectively lead to the downregulation of GPX4 and SLC7A11 protein expression, ultimately triggering ferroptosis (Figure 4I).

2.5. ICD Induction and DC Maturation of DMP@FGMS In Vitro

ICD is characterized by the surface presentation of calreticulin (CRT) and the release of HMGB1 from the nucleus [32, 33]. The potential of DMP@FGMS to induce ICD in vitro was first evaluated using the immunofluorescence staining method. Compared with the control group, treatment with D@FGMS resulted in enhanced CRT fluorescence and reduced nuclear HMGB1 fluorescence in Hepa1‐6 cells (Figure 5A–D), suggesting that D@FGMS could effectively trigger ICD. Notably, the efficacy of D@FGMS in inducing ICD could be substantially improved by incorporating PFOB or MET into the microspheres, as both DP@FGMS and DM@FGMS showed a greater ability to enhance CRT fluorescence and reduce nuclear HMGB1 fluorescence. Among all the treatment groups, the cells treated with DMP@FGMS showed the highest level of CRT fluorescence and the lowest level of nuclear HMGB1 fluorescence (Figure 5A–D). The strongest ability of DMP@FGMS to trigger ICD was further confirmed by measuring the release levels of HMGB1 and ATP in the culture medium, which showed the highest levels compared to other treatments (Figure 5E,F).

FIGURE 5.

ICD Induction and DC Maturation. (A,B) Under hypoxic conditions, CRT (A) exposure and HMGB1 (B) analysis were performed by immunofluorescence after different treatments. (C, D) Statistical analysis of CRT (C) and HMGB1 (D) immunofluorescence. (E,F) After different treatments under hypoxia, the concentration of ATP (E) or HMGB1 (F) in the cell culture medium was measured. (G) Schematic diagram of the Hepa1‐6 and BMDC co‐culture system. (H) Representative FCM plots of CD11c⁺CD80⁺CD86⁺ DCs in the co‐culture system. (I) Quantitative analysis of CD11c⁺CD80⁺CD86⁺ DCs in the co‐culture system. (J–L) ELISA assay to measure the secretion levels of IL‐12p70 (J), IL‐6 (K), and IL‐10 (L) cytokines by DCs in the supernatant of the co‐culture system. Data are presented as mean ± SD (n = 3). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

The release of DAMPs from dying cells enhances the tumor immune response through the robust promotion of dendritic cell (DCs) maturation, which plays a crucial role as an antigen‐presenting cell in the immune system [34]. Consequently, DMP@FGMS may facilitate the maturation of DCs to strengthen potent antitumor immune response. To confirm this hypothesis in vitro, a two‐dimensional coculture system was established by adding bone marrow‐derived dendritic cells (BMDCs) to the dying tumor cells (Figure 5G). Flow cytometry analysis revealed that DMP@FGMS‐induced dying cells remarkably promoted the maturation of DCs, achieving a percentage of approximately 40% (Figure 5H,I). This ratio was significantly higher than that of D@FGMS (25.40 ± 0.79%), DP@FGMS (30.63 ± 1.27%), and DM@FGMS (33.03 ± 1.27%) (Figure 5H,I). Furthermore, the levels of IL‐12p70, IL‐6, and IL‐10 secreted by DCs were assessed using ELISA kits. As shown in Figure 5J–L, DMP@FGMS treatment caused the most significant increase in IL‐12p70 and IL‐6 release, along with a reduction in IL‐10 release among all groups. These results further confirmed that DMP@FGMS has a superior capacity in inducing ICD in HCC cells.

2.6. Evaluation of DMP@FGMS: Biodistribution, Hypoxia Alleviation, Induction of ICD and Ferroptosis In Vivo

To evaluate the in vivo release profile of DMP@FGMS, we first employed the IVIS Spectrum imaging system for detection. Cy5.5 was used to replace DOX and MET adsorbed onto PFGMS, thereby minimizing potential interference from tissue autofluorescence. The prepared Cy5.5@PFGMS was administered intratumorally to Hepa1‐6 tumor‐bearing mice, whereas mice treated with free Cy5.5 served as the control group. Fluorescence signals of Cy5.5 in each group were acquired on days 0, 1, 3, 5, and 7 using the IVIS Spectrum system. As shown in Figure 6A and S15, the Cy5.5 fluorescence intensity in the tumor area of the control group decreased rapidly from day 1. In contrast, tumors treated with Cy5.5@PFGMS maintained sustained fluorescence signals over time, indicating that Cy5.5@PFGMS enables the prolonged release of DOX and MET in vivo.

FIGURE 6.

Evaluation of DMP@FGMS: Biodistribution, Hypoxia Alleviation, and Induction of ICD and Ferroptosis In Vivo. (A) Evaluating the in vivo release behavior of Cy5.5@PFGMS using the IVIS Spectrum imaging system (n = 3 mice per group). (B) Immunofluorescence staining images of HIF1A in tumor slices after different treatments. (C,D) Immunofluorescence staining images of CRT (C) and ROS (D) in tumor slices after different treatments. (E) Representative imaging of GPX4 analysis was presented.

The tumor‐bearing mice were intratumorally injected with PBS, D@FGMS, DP@FGMS, DM@FGMS and DMP@FGMS, respectively. The mice were sacrificed for 5 days after different treatment, and their tumors were collected for slicing and subsequent staining. The results revealed that treatment with D@FGMS had a negligible effect on intratumoral hypoxia, with HIF‐1α and pimonidazole fluorescence levels remaining comparable to those in the control group (Figure 6B, S16 and S17). In comparison, DMP@FGMS treatment effectively alleviated tumor hypoxia in vivo, and its efficacy surpassed that of both DP@FGMS and DM@FGMS (Figure 6B, S16 and S17). These collective findings confirm that the concurrent application of MET (to suppress oxygen consumption) and PFOB (to provide oxygen supply) creates a synergistic effect, thereby demonstrating a more potent resolution of tumor hypoxia compared to individual monotherapies (Figure 6B, S16 and S17). The fluorescence intensity of CRT in the DMP@FGMS‐treated group was notably higher compared to the other treatment groups (Figure 6C, S18). Conversely, a significantly lower HMGB1 fluorescence signal was observed in the tumor tissues of the DMP@FGMS‐treated mice (Figure S19). These results suggest that the incorporation of MET and PFOB enables the DOX‐loaded microspheres to synergistically trigger ICD in vivo. Tumor sections from the DMP@FGMS‐treated group exhibited the most intense ROS fluorescence and the least GPX4 expression, consistent with the most potent induction of ferroptosis (Figure 6D,E, Figure S20). In summary, DMP@FGMS functions as a multifaceted therapeutic strategy that precisely remodels the tumor microenvironment by simultaneously overcoming hypoxia, activating antitumor immunity, and executing ferroptotic cell death.

2.7. In Vivo Antitumor Efficiency of DMP@FGMS

Here, we first employed a subcutaneous tumor model to evaluate the therapeutic potential of DMP@FGMS in combating HCC. The Hepa1‐6 tumor‐bearing mice were randomly divided into 5 groups and subjected to the following treatments: (I) Control, (II) D@FGMS, (III) DP@FGMS, (IV) DM@FGMS and (V) DMP@FGMS (Figure 7A). After various treatments, the changes of tumor volumes were recorded by caliper measurement starting from day 0. Compared to the control group, the treatment of D@FGMS resulted in only partial inhibition of tumor growth (Figure 7B, S21). In contrast, treatment with DP@FGMS and DM@FGMS exhibited a moderate capacity to suppress tumor growth. Notably, the anticancer efficacy of DMP@FGMS was significantly stronger than that of DP@FGMS or DM@FGMS, as it exhibited the slowest tumor growth rate among all the treatment groups (Figure 7B, S21). Besides, we observed that mice in the DMP@FGMS group exhibited a significantly longer survival time compared to the other treatment groups (Figure 7C). The therapeutic efficacy was further assessed through H&E and Ki67 staining of tumor sections after various treatments. In line with the tumor growth and survival time data, the treatment of DMP@FGMS led to the most extensive tumor damage compared to the other treatment group (Figure 7D), providing additional evidence of its strong tumor‐killing efficacy. Collectively, our findings demonstrate that DMP@FGMS represents a promising combinatorial regimen capable of overcoming key therapeutic barriers in HCC, including chemoresistance and immunosuppression, leading to significantly improved survival outcomes.

FIGURE 7.

In vivo antitumor efficiency of DMP@FGMS on subcutaneous tumors. (A) Schematic illustration of animal experiments. (B) Average tumor growth curves on Hepa1‐6‐tumor‐bearing mice after different treatments. (C) Survival rates of mice during the treatment period. (D) Representative imaging of H&E and Ki67 analysis was presented. (E) Representative flow cytometric plots and intratumoral frequencies of CD80⁺CD86⁺DCs. (F) Representative flow cytometric plots and intratumoral frequencies of CD3⁺CD8⁺ T cells. (G) Quantitative analysis of CD4⁺ Foxp3⁺ Tregs in tumors. (H) Quantitative analysis of CD3⁻NK1.1⁺ NK cells in tumors (n = 3). (I,J) Quantitative analysis of CD11b⁺CD86⁺ M1 macrophages (I) and CD11b⁺CD206⁺ M2 macrophages (J), ratio of M1/M2 macrophage in tumors. Data in B and C are presented as mean ± SD (n = 6 mice per group). Data in E–J are presented as mean ± SD (n = 3 mice per group). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

2.8. In Vivo Anticancer Immune Response of DMP@FGMS

The release of DAMPs such as HMGB1 and CRT from dying cancer cells contributed to the tumor immune response by significantly promoting DCs maturation, which plays a crucial role as an antigen‐presenting cell in the immune system [34]. We then harvested the tumor‐adjacent lymph nodes and investigated the ability of different microspheres to promote DCs maturation by using flow cytometry. As shown in Figure 7E, the percentage of mature DCs in lymph nodes collected from Group I, II, III, IV and V was 18.20 ± 0.79%, 25.80 ± 2.61%, 35.00 ± 1.55%, 34.80 ± 3.10%, and 44.10 ± 2.92%, respectively. This result can be attributed to the fact that the dual hypoxia alleviation provided by MET and PFOB, which endows DMP@FGMS with a synergistic ability to induce ICD, thereby showing the excellent capability in promoting DCs maturation. Additionally, the intratumoral infiltration of various immune cells was analyzed using flow cytometry. It was observed that DMP@FGMS treatment led to the highest percentages of tumor‐infiltrating CD4⁺ and CD8⁺ T cells compared to the other treatment groups (Figure 7F, S22). Notably, the proportion of CD8⁺ T cells increased approximately sevenfold in Group V (40.23 ± 2.46%) compared to the control group (5.78 ± 1.20%), indicating the excellent anticancer immune response of DMP@FGMS (Figure 7F). The DMP@FGMS treatment group exhibited the lowest level of CD8⁺ T cell exhaustion among all groups (Figure S23). We also found that DMP@FGMS was more efficient at decreasing the intratumoral frequencies of immunosuppressive regulatory T cells (Tregs) than the other types of microspheres (Figure 7G, S24). Besides, DMP@FGMS exhibited the highest effectiveness in boosting the percentages of natural killer (NK) cells (Figure 7H, S25), which is an important indicator of innate anticancer immunity. In the tumor microenvironment, M1‐type tumor‐associated macrophages (TAMs) exert anticancer effects by promoting inflammation and killing tumor cells, whereas M2 phenotype has an immunosuppressive role, inhibiting immune responses and promoting tumor growth [35]. As expected, DMP@FGMS demonstrated the best ability to increase the frequency of M1‐type macrophages, while reducing the infiltration of M2‐type macrophages into the tumor (Figure 7I,J, S26). Furthermore, it was revealed that DMP@FGMS also exhibited the superior ability to induce the secretion of cytotoxic cytokines, including the tumor necrosis factor‐α (TNF‐α) and interferon‐γ (IFN‐γ) (Figure S27). Taken together, it can be concluded that the dual hypoxia‐alleviating capability of DMP@FGMS enables it to not only induce cancer apoptosis but also effectively reverses the immunosuppressive TME, thereby promoting conditions advantageous for eliciting robust antitumor immunity.

2.9. DMP@FGMS Synergizes With αPD‐L1 Immunotherapy to Combat Distant HCC

HCC metastasis, a primary cause of therapeutic failure and patient mortality, remains refractory to conventional therapeutic modalities, including surgical resection, chemotherapy, radiotherapy, and chemoembolization [36, 37]. Immune checkpoint blockade (ICB), utilizing therapeutic antibodies targeting PD‐1/PD‐L1 or CTLA‐4, has proven to be a promising clinical intervention for metastatic cancers such as HCC [38]. Inspired by the strong antitumor immune response induced by DMP@FGMS, we proposed that the combination of DMP@FGMS with anti‐PD‐L1 antibody (αPD‐L1) could potentially enhance the therapeutic efficacy against HCC. For this purpose, a bilateral Hepa1‐6 tumor model was established by injecting the cancer cells into the right‐side flank of mice at day ‐7 (designated as the primary tumor) and into the left‐side flank at day ‐4 (designated as the distant tumor). The mice with bilateral tumors were then divided into four groups and treated as follows: (I) Control, (II) αPD‐L1, (III) DMP@FGMS, and (IV) DMP@FGMS + αPD‐L1 (Figure 8A). After various treatments, the changes in bilateral tumor volumes were monitored every 2 days. As depicted in Figure 8B and S28, the administration of αPD‐L1 immunotherapy demonstrated limited efficacy in inhibiting the growth of bilateral tumors. On the other hand, DMP@FGMS treatment effectively suppressed the growth of the primary tumor, while its efficacy against the distant tumor was relatively modest. Notably, the combination treatment of DMP@FGMS and αPD‐L1 led to a markedly greater suppression of both primary and distant tumors compared to the monotherapy (Figure 8B and S28). In line with the tumor volume data, the survival time of mice receiving the combination treatment was significantly extended compared to those treated with DMP@FGMS or αPD‐L1 monotherapy (Figure 8C). These positive findings suggest that this combinational therapeutic strategy holds considerable potential for inhibiting HCC metastasis, particularly in targeting small metastatic foci that remain undetectable prior to treatment.

FIGURE 8.

DMP@FGMS synergizes with αPD‐L1 to combat distant tumor. (A) Scheme of bilateral tumor modeling and treatment strategy. The tumor volumes at both primary and distant sites were measured every 2 days. (B) Growth curves of primary and distant HCC tumors following various treatments. (C) Survival rates of mice during the treatment period. (D) Quantitative analysis of CD86⁺CD80⁺ DCs in tumors. (E, F) Quantitative analysis of CD3 T cells (E) and CD3⁺CD8⁺ T cells (F) in tumors. (G) Quantitative analysis of CD3⁻NK1.1⁺ NK cells in tumors. (H) Ratio of M1/M2 macrophages in tumors. (I) Quantitative analysis of CD4⁺ Foxp3⁺ Tregs in tumors. (J) 'Schematic representation of the mechanism by which DMP@FGMS combined with aPD‐L1 induces antitumor immune responses. Data in B and C are presented as mean ± SD (n = 6 mice per group). Data in D‐I are presented as mean ± SD (n = 3 mice per group). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

To clarify the mechanism underlying the enhanced antitumor effect mediated by DMP@FGMS in combination with αPD‐L1, the systemic immune responses were assessed in the tumor‐bearing model at day 7 following the respective treatments. Compared with DMP@FGMS and αPD‐L1 monotherapy, the combination treatment was more effectively to promote the DCs maturation neighboring the distant tumor (Figure 8D, S29). In addition, we found that the combination treatment of DMP@FGMS and αPD‐L1 markedly elevated the proportion of total T cells within the distant tumor (Figure 8E, S30). This finding indicates that the combination therapy effectively stimulates robust systemic immune responses, consequently facilitating substantial recruitment of immune cells into the distant tumor. Similarly, a significant increase in the percentages of CD8⁺ T cells in distant tumors was observed following the combinational therapy (Figure 8F, S31). Furthermore, the distant tumors receiving the combination treatment exhibited a notably higher percentage of tumor‐infiltrating NK cells and M1‐type macrophages than those treated with DMP@FGMS or αPD‐L1 monotherapy (Figure 8G,H, S32 and S33). Conversely, the combination treatment resulted in a significantly reduced level of tumor‐infiltrating M2‐type macrophages and Tregs within the distant tumors (Figure 8H,I, S34 and S35). These positive results collectively demonstrate that DMP@FGMS could effectively modulate the TME and induce cancer cell apoptosis, thereby activating robust antitumor immunity, while PD‐L1 blockade further enhances the systemic antitumor immune response (Figure 8J).

2.10. In Vivo TACE‐Immunotherapy of DMP@FGMS With αPD‐L1

TACE is frequently employed in clinical practice for intermediate and advanced HCC due to its minimally interventional nature [39]. TACE works by embolizing the blood vessels that supply HCC, causing a rapid vessel closure and resulting in ischemic necrosis of the tumor [40]. In practice, TACE is frequently combined with ICB therapy to enhance therapeutic efficacy and extend the survival of patients with unresectable HCC [41]. The therapeutic efficacy of DMP@FGMS‐based TACE combined with αPD‐L1 immunotherapy was therefore evaluated using an orthotopic N1S1 tumor model. The tumor‐bearing rat models were created by injecting N1S1 cells into the left lobe of the liver, followed by the treatments described below: (I) Control; (II) αPD‐L1; (III) D@FGMS, (IV) DMP@FGMS, and (V) DMP@FGMS + αPD‐L1 (Figure 9A). The tumor progression was monitored by MRI 1 week after tumor implantation and at 3, 7, and 14 days after treatment. It was discovered that the mice αPD‐L1 monotherapy still displayed a rapid progression of tumor growth. The tumor volume quickly increased to approximately 1500 mm³ within 2 weeks, which was only slightly smaller than that of the control group (∼1700 mm³) (Figure 9B–D). D@FGMS‐based TACE treatment was effective in slowing the tumor growth, but its ability to suppress tumor growth remained lower compared to DMP@FGMS. Notably, the combination of DMP@FGMS‐based TACE and αPD‐L1 immunotherapy yielded the most effective results for treating orthotropic HCC, leading to the total elimination of tumors in two rats (Figure 9B–D). Moreover, the histopathological analyses with H&E and Ki‐67 staining revealed that the tumor sections from group V exhibited the highest level of cell apoptosis and the lowest level of cell proliferation among the treatment groups (Figures S36 and S37), further validating the anticancer efficacy could be largely enhanced by the combination of DMP@FGMS‐based TACE and αPD‐L1 immunotherapy.

FIGURE 9.

DMP@FGMS amplifies TACE‐immunotherapy in an orthotopic N1S1 model. (A) Treatment procedure of DMP@FGMS combined with αPD‐L1 for hepatic arterial embolization in a rat. (B) T2‐weight MRI images of N1S1 tumor‐bearing rats were obtained 1 week after tumor implantation and at 3, 7, and 14 days following the indicated TACE treatments. (C) The average tumor growth curves before and after embolization. (D) Tumor suppression rate at 14 days post‐embolization. (E) Flow cytometric quantitative analysis of CD3⁺CD8⁺ T cells in N1S1 tumors. (F) Flow cytometric quantitative analysis of CD4⁺ Foxp3⁺ Tregs in N1S1 tumors. (G, H) Secretion levels of IFN‐γ (G) and TNF‐α (H) in tumors of rats after various treatments as indicated. (I, J) Immunofluorescence staining of CD8⁺ and Tregs cell in tumor sections from N1S1 tumor‐bearing rats subjected to the indicated treatments. CD8 and Foxp3 Immunofluorescence staining were obtained from different rats within the same treatment group. Data in B‐D are presented as mean ± SD (n = 5 rats per group). Data in E‐H are presented as mean ± SD (n = 3 rats per group). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

The remarkable therapeutic response observed with the combination treatment could be attributed to the ability of DMP@FGMS to effectively remodel the tumor TME through dual hypoxia alleviation and trigger ICD, thereby enhancing the activation of antitumor immunity. To confirm this explanation, the rats were euthanized on the seventh day after various treatments, and the tumor tissues were collected for immunofluorescence staining analyses. It was found that the TACE‐induced hypoxia states could be remarkably alleviated by employing DMP@FGMS, as evidenced by the pimonidazole and HIF‐1α staining results (Figure S38). DMP@FGMS‐based TACE treatment demonstrated the superior efficacy in enhancing CRT exposure and HMGB1 release within tumor tissues, highlighting its exceptional ability to induce ICD (Figure S39). To further investigate the tumor immune microenvironment, tumor tissues were homogenized into single‐cell suspensions and analyzed using flow cytometry. As shown in Figure 9E and S40, the combination treatment of DMP@FGMS and αPD‐L1 exhibited a significantly higher ratio of CD4⁺ and CD8⁺ T cells than the other group. In contrast, such combination therapy resulted in the lowest ratio of Tregs within the tumor tissue (Figure 9F, S41). The combination treatment of DMP@FGMS and αPD‐L1 induces the highest secretion of cytotoxic cytokines (TNF‐α and IFN‐γ) (Figure 9G,H). Similarly, immune‐histochemistry analysis also indicate that DMP@FGMS‐based TACE in combination with αPD‐L1 showed the greatest efficacy in increasing tumor‐infiltrating CD8⁺ T cells and Tregs to the other groups (Figure 9I,J, S42 and S43). To effectively trigger ICD and stimulate a strong immune response within TME, thereby remarkably enhancing the therapeutic outcomes of TACE‐immunotherapy.

To evaluate the safety and biocompatibility of the proposed strategy, another batch of rats with orthotopic N1S1 tumors were euthanized after various treatments, and major organs and blood samples were collected. No significant damage or inflammation was observed in the major organs (liver, heart, spleen, lungs, and kidneys) of the rats (Figure S44). Additionally, there were no notable alterations in key serum biochemistry parameters, such as white blood cells (WBC), red blood cells (RBC), hemoglobin (HGB), platelets (PLT), alanine transaminase (ALT), and aspartate transaminase (AST), further demonstrating the excellent biocompatibility of this combined therapy (Figure S45). These results indicate that the DMP@FGMS‐based TACE in combination αPD‐L1 immunotherapy effectively inhibits the growth of N1S1 tumors without causing noticeable adverse effects.

3. Conclusion

In summary, this study successfully developed and characterized a multifunctional microsphere platform, DMP@FGMS, which integrates chemoembolization, hypoxia remodeling, and immunotherapy to enhance the therapeutic efficacy against HCC. The microspheres achieved high encapsulation efficiencies for DOX and MET through electrostatic interactions and exhibited controlled, sustained drug release. This in‐outside hypoxia remodeling strategy effectively reversed hypoxia‐induced chemoresistance by inhibiting intracellular oxygen consumption via MET and enhancing oxygen delivery through PFOB, thereby downregulating HIF‐1α, VEGF, and P‐gp both in vitro and in vivo. Remarkably, we discovered and confirmed that DMP@FGMS could not only modulate hypoxic TME, but also remarkably generate intracellular ROS and disrupt the redox homeostasis, thereby triggering cancer ferroptosis. Concurrently, DMP@FGMS effectively induced ICD and promoted DC maturation, which in turn elicited potent antitumor immune responses characterized by increased infiltration of CD8⁺ T cells, NK cells, and M1‐type macrophages, as well as elevated levels of TNF‐α and IFN‐γ, accompanied by decreased populations of Tregs and M2‐type macrophages. Furthermore, the immunostimulatory effect of DMP@FGMS exhibited strong synergy with immune checkpoint blockade therapy, leading to significant tumor suppression and prolonged survival in both subcutaneous and orthotopic HCC models. Overall, DMP@FGMS‐based TACE provides a transformative and translationally promising approach to alleviate tumor hypoxia and immune evasion, opening a new avenue for effective treatment of advanced HCC.

4. Materials and Methods

4.1. Materials

Gelatin was purchased from Sigma‐Aldrich (Missouri, USA). Perfluorooctyl bromide (PFOB), metformin (MET), doxorubicin hydrochloride (DOX), Span 80, and liquid paraffin were purchased from Aladdin Chemicals Co., Ltd (Shanghai, China). Penicillin‐streptomycin, trypsin, PBS, and FBS were procured from Thermo Fisher Scientific (Massachusetts, USA). Annexin V‐FITC apoptosis detection kit, MTT assay kit, BCA protein quantification kit, ATP assay kit, DAPI, Calcein AM/PI cell viability/cytotoxicity assay kit, GSH and GSSG assay kit, MDA assay kit, ROS probes, mitoSOX probes, and C11‐BODIPY probes was purchased from Beyotime Biotechnology Co., Ltd. (Shanghai, China). Anti‐calreticulin (Cat #: 27298‐1‐AP), anti‐VEGFA (Cat #: 19003‐1‐AP), anti‐GAPDH (Cat #: 60004‐1‐Ig), anti‐HIF‐1α (Cat #: 66730‐1‐Ig), anti‐P glycoprotein (Cat #: 22336‐1‐AP), anti‐GPX4 (67763‐1‐Ig), anti‐SLC7A11 (26864‐1‐AP) were purchased from Proteintech Co., Ltd (Wuhan, China). Protease inhibitor was purchased from Phygene Life Sciences Co., Ltd (Fuzhou, China). Hypoxyprobe‐1 plus kit was purchased from Beijing Biolead Biology SCI & TECH Co. Ltd (Beijing, China). The following antibodies are available from BioLegend, Inc. (San Diego, CA): Purified anti‐mouse CD3 antibody (Cat# 100202), Purified anti‐mouse CD28 antibody (Cat# 102102), anti‐mouse‐Tim‐3 (Cat#: 134010), anti‐mouse‐CD3 (Cat#: 100236), anti‐mouse‐CD4 (Cat#: 100405), anti‐mouse‐CD8a (Cat#: 100708), anti‐mouse‐CD11c (Cat#: 117306), anti‐mouse‐CD80 (Cat#: 104714), anti‐mouse‐CD86 (Cat#: 159204), anti‐mouse‐CD206 (Cat#: 141726), anti‐mouse‐CD11b (Cat #: 101228), anti‐mouse ‐F4/80 (Cat #: 123108), anti‐rat‐CD3 (Cat#: 201403), anti‐rat‐CD8 (Cat#: 200609), anti‐mouse‐CD4 (Cat#: 201515), and anti‐mouse‐Foxp3 (Cat #: 101228). The HMGB1, IL‐10, IL‐6, IL‐12p, TNF‐α and IFN‐γ ELISA kits were obtained from Enzyme‐free Industrial Co., Ltd (Jiangsu, China). An anti‐PD‐L1 checkpoint blocker was obtained from InVivoPlus (Cat # BP0101, B7‐H1, Bio cell, USA).

4.2. Cell Lines and Animals

The murine Hepa1‐6 and rat N1S1 liver cancer cells were purchased from American Type Culture Collection (ATCC; Manassas, VA, USA). The Hepa1‐6 cells were cultured in DMEM medium supplemented with 10% fetal bovine serum (FBS, Gibco BRL, USA) and 1% of penicillin/streptomycin (37°C, 5% CO₂). The N1S1 cells were cultured in Iscove’s Modified Dubecco’s Medium (IMDM) supplemented with 10% of FBS and 1% of penicillin/streptomycin (37°C, 5% CO₂). The cells requiring hypoxic conditions (1% O₂) were cultured in a hypoxia chamber with 5% CO₂ at 37°C.

C57BL/6 female mice (6–8 weeks), Sprague‐Dawley (SD) rats (200–250 g) and New Zealand white rabbits (2–2.5 kg) were purchased from Shanghai SLAC Laboratory Animal Co., Ltd. (Shanghai, China). The animal procedures were carried out in compliance with Lishui University's Guidelines for Care and Use of Laboratory Animals and were approved by the Animal Ethics Committee at Lishui University. To construct a unilateral subcutaneous HCC tumor model, Hepa1‐6 cells were injected into the right flank of each C57BL/6 mouse. To construct a bilateral subcutaneous HCC tumor model, Hepa1‐6 cells were injected into the right flank of each C57BL/6 mouse on the ‐seventh day and the left flank of each C57BL/6 mouse on the ‐fourth day. To construct the orthotopic HCC tumor model, N1S1 cells were injected into the right lower lobe of the liver in SD rats using a 25‐gauge syringe needle under anesthesia.

4.3. Bioinformatics Analysis

Clinical information and RNA‐Seq data of 371 hepatocellular carcinoma (HCC) tumor tissues and 50 paired adjacent non‐tumor liver tissues were obtained from The Cancer Genome Atlas (TCGA) dataset (https://portal.gdc.cancer.gov/). To expand the sample size of normal liver controls and improve the statistical power of differential expression analysis, additional 228 normal liver tissues were downloaded from the Genotype‐Tissue Expression (GTEx) database (https://gtexportal.org/). Survival analysis was conducted using the Kaplan‐Meier method, and survival rates were compared utilizing the log‐rank test. All statistical analyses and R package implementations were performed using R version 4.0.3. Statistical significance was determined using a threshold of p ≤ 0.05.

4.4. Preparation for P@FGMS

Monodisperse PFOB‐loaded fucoidan/gelatin microspheres containing (denoted as P@FGMS) were prepared by using a simplified microfluidic device. The microfluidic device was homemade according to a procedure described in our previous study [42]. First, PFOB nanoemulsions were prepared by homogenizing 10% of PFOB with 90% of aqueous solution containing 1% of BSA under a probe ultrasound (SCIENTZ‐IID, Scientz, Zhejiang, China). Then, an aqueous phase containing gelatin (100 mg/mL), fucoidan (10 mg/mL) and PFOB (50 µL/mL) were prepared at 50 °C as the dispersed phase. The liquid paraffin containing 5 wt% of Span 80 was prepared as the continuous phase. Next, the dispersed and continuous phases were pumped into the microfluidic chip with a flow rate of 5 and 300 µL/min, respectively. The monodispersed droplets were then collected into a beaker containing the continuous phase and kept in an ice‐cold water bath under gentle stirring. After solidification, the microspheres underwent crosslinking with 25% glutaraldehyde for 6h. Afterward, the microspheres were repeatedly washed with isopropyl alcohol (50%, 75% and 100%) to eliminate the remaining oil and excess free cross‐linker. The collected P@FGMS was then naturally dried and stored at 4°C. Besides, the blank gelatin microspheres (GMS) and PFOB‐loaded gelatin microspheres (P@GMS) were prepared by the same procedure apart from introducing different corresponding components.

4.5. Sample Characterization

The average size and size distribution of PFOB nanoemulsions were measured by a dynamic light scatterometer (DLS, Zetasizer Lab ZSU3100, Malvern, UK). Transmission electron microscopes (TEM, HT7800, Hitachi, Japan) were used for morphological observations of PFOB nanoemulsions. The morphology of microspheres was analyzed with a scanning electron microscope (SEM, Phenom Pharos, Phenom Scientific, Netherlands). The chemical compositions of P@FGMS were analyzed using SEM along with energy dispersive X‐ray spectrometry (EDS). The average size of the microspheres was determined using Image‐J software, and the size distribution was analyzed with Origin software. The coefficient of variation (CV) of microspheres were calculated by Equation 1:

$$CV=\sigma /d_{av}\times 100\%$$ (1)

where σ represents the standard deviation of the particle diameter, and d _av is the average particle diameter of the microspheres (n = 100). The swelling capacity of microspheres were calculated using Equation 2:

$$Swellingcapacity\left(\%\right)=\left(W_1-W_0\right)/W_0\times 100\%$$ (2)

where W₀ and W₁ represent the mass of the microspheres before and after immersion in the aqueous solution for 10 min.

4.6. In Vitro Drug Loading and Release Behavior

Drug loading capacity test: In this part, P@FGMS, P@GMS or GMS (25 mg) were separately placed in a 2 mL centrifuge tube. Then, 400 µL of aqueous solution containing DOX (2.5 mg/mL) and MET (2.5 mg/mL) were added to the centrifuge tube. After co‐incubation for 1h, the microspheres were subjected to centrifugation and washing three times, during which the supernatants were collectively collected. Subsequently, the concentration of DOX in the supernatants were measured using a UV–vis spectrophotometer (UV1900i, Shimadzu, Japan) at wavelengths of 480 nm. The concentration of MET in the supernatant was determined using HPLC (Waters e2695 Alliance, Waters, USA) with UV detection at a wavelength of 233 nm. Therefore, the percentage of drug loading can be quantified based on the concentration of DOX and MET detected in the supernatants.

Drug release behavior: In this part, 25 mg of the drug‐loaded microspheres (DMP@GMS or DMP@FGMS) prepared using the method described above were placed into a 15 mL centrifuge tube, followed by the addition of 10 mL of PBS as the release medium. Then, the centrifuge tube was subjected to a constant temperature oscillating water bath at 37°C (100 rpm). At a specific time, 1 mL of the release medium was removed and substituted with the same amount of fresh release medium. Finally, the concentrations of DOX and MET in the release medium at each time point were measured, and these values were used to calculate the cumulative drug release rates from the microspheres.

4.7. In Vitro Oxygen Loading and Release Behavior

In vitro oxygen loading and release behavior was evaluated by a dissolved oxygen meter. In brief, P@FGMS, GMS or PFOB nanoemulsions was added into 10 mL of water. Then, the pure oxygen was introduced into the mixture at a flow rate of 5 L/min for 10 min. Then, the oxygen concentration at each time point was recorded by the dissolved oxygen meter. After stopping the oxygen flow, the dissolved oxygen meter was used to continuously record the oxygen concentration, in each sample, which represents the oxygen release behaviors.

4.8. Evaluation of the Embolization Effect of DMP@FGMS

Decellularized liver model: The normal liver was harvested from SD rats and frozen overnight at −80°C. Then, the frozen liver was completely thawed at room temperature. To obtain a decellularized liver matrix, the portal vein and hepatic vein of rat livers were sequentially perfused with distilled water for 30 min, followed by a 4% Triton X solution for 3h, and then with a 0.5% sodium dodecyl sulfate (SDS) solution for 12 h at a flow rate of 7 mL/min, until the liver tissue became translucent. Afterward, the translucent liver was thoroughly rinsed with saline to eliminate the remaining SDS. The DMP@FGMS was then slowly injected into the hepatic vein of the decellularized liver via syringe, and the images were taken using fluorescence microscopy.

Rabbit kidney embolism model: The embolization efficacy of DMP@FGMS was assessed in vivo using a rabbit renal embolization model. Rabbits were anesthetized via intravenous administration of pentobarbital sodium through the auricular vein, after which the right femoral artery was surgically exposed. A 4‐F vascular sheath (Cook, Bloomington, USA) was introduced directly into the femoral artery. Under digital subtraction angiography (DSA, Allura Xper, Philips Healthcare, Netherlands) guidance, a 2.7‐F microcatheter (Terumo, Tokyo, Japan) was selectively navigated into the targeted renal artery. Prior to embolization, renal angiography was performed by injecting iohexol contrast agent through the microcatheter to confirm arterial anatomy and blood flow. Subsequently, a homogeneous suspension of DMP@FGMS and iohexol was slowly injected into the renal artery to achieve effective embolization. The embolization endpoint was defined as the appearance of contrast agent reflux during injection, indicating that sufficient embolization had been achieved and microsphere injection was terminated. After that, the microcatheter was removed and the wound was sutured. To prevent potential infection and inflammation, the rabbits were administered ampicillin at a dosage of 50 000–200 000 units/kg for two cycles. Angiography was performed on the 14th day post‐embolization to confirm the absence of blood flow in the kidney.

4.9. In Vitro Immunofluorescence Staining

Hepa1‐6 cells were seeded in 12‐well plates at a density of 1 × 10⁵ cells per well and cultured under normoxic or hypoxic conditions. After overnight incubation, cells were further incubated with oxygen‐loaded microsphere (0.25 mg/mL) for an additional 12h. Subsequently, cells were gently washed and subjected to immunofluorescent staining using pimonidazole, HIF‐1α, and VEGF antibodies according to the manufacturer's instructions. Fluorescence images were acquired using a confocal laser scanning microscope (CLSM; STELLARIS 5, Leica, Germany).

4.10. Western Blot Analysis

Hepa1‐6 cells treated with various microspheres were washed with PBS and subsequently harvested. Total protein was extracted from the harvested cells using RIPA lysis buffer (Beyotime Biotechnology, Shanghai, China). Cell lysates were clarified by centrifugation at 12 000 rpm for 10 min at 4°C, and the resulting supernatants were collected. Protein concentrations were quantified using the Bradford protein assay kit (Bio‐Rad Laboratories, Shanghai, China). Equivalent amounts of protein from each sample were separated via SDS‐PAGE and subsequently transferred onto polyvinylidene fluoride (PVDF) membranes. Membranes were blocked with 5% non‐fat dry milk in Tris‐buffered saline containing 0.1% Tween 20 at room temperature. Following washing steps, the membranes were incubated overnight at 4°C with primary antibodies targeting HIF‐1α, VEGF, P‐gp, GPX4, SLC7A11 and GAPDH. After incubation with the appropriate secondary antibodies, protein bands were visualized using enhanced chemiluminescence reagents (Pierce Biotechnology, USA), and their intensities were quantified using ImageJ software. The expression level of each target protein was first normalized to the corresponding GAPDH loading control. Relative protein expression was then independently calculated for each target protein using its corresponding control group as the reference.

4.11. Cytotoxicity In Vitro

Cytotoxicity was assessed using an MTT assay and a Calcein/PI cell viability/cytotoxicity assay kit. Hepa1‐6 cells were cultured overnight in a 12‐well plate under normoxic or hypoxic conditions at 37°C. The cells were further incubated with oxygen‐loaded microspheres (1 mg/mL) for 24 h. Next, the cells were stained with Calcein AM/PI for 30 min at 37°C, and fluorescence images of live/dead cells were captured with a fluorescence microscope (DMi8, Leica, Germany).

In the MTT assay, each well containing differently treated cells received 100 µL of a 5% MTT solution and was incubated for 4 h. The medium was then removed, and 1 mL of DMSO was added to dissolve the purple formazan crystals. The absorbance at 570 nm for each well was measured using a microplate reader (BioTek, Winooski, VT, USA).

The apoptosis of Hepa1‐6 cells was assessed using the Annexin V‐7‐AAD staining method. Hepa1‐6 cells were cultured at a density of 1 × 10⁵ cells per well in a 12‐well cell plate and cultured overnight under hypoxic or normoxic conditions. The oxygen‐loaded microspheres (1 mg/mL) were added for a 24h co‐incubation, and flow cytometry (BD FACSCanto, BD Biosciences, USA) was employed to analyze the apoptotic cell percentage in each group.

4.12. In Vitro ICD Induction

Hepa1‐6 cells were placed in 12‐well plates at a density of 1 × 10⁵ cells per well and incubated under hypoxic condition overnight. The cells were then exposed to different types of oxygen‐loaded microspheres (0.25 mg/mL) under hypoxic conditions for another 12 h. Afterward, the cells were washed with PBS, fixed with 4% paraformaldehyde for 15 min, and stained with CRT and HMGB1 antibodies for fluorescence imaging by CLSM. Moreover, the release levels of ATP and HMGB1 in the culture medium were analyzed by ELISA kits according to the manufacturer's protocols.

4.13. In Vitro DCs Maturation

Bone marrow cells were extracted from the femurs and tibias of C57BL/6 mice. To obtain bone marrow‐derived dendritic cells (BMDCs), the bone marrow cells were stimulated with 20 ng/mL of recombinant mouse GM‐CSF for 6 days. Meanwhile, the hypoxic Hepa1‐6 cells were treated with different oxygen‐loaded microspheres (0.5 mg/mL) for 24 h. Afterward, the dying Hepa1‐6 cells were collected and co‐incubated with BMDCs at a ratio of 2:1 for 48 h. Then, cells were collected and analyzed by flow cytometry to measure the percentages of CD11c⁺CD80⁺CD86⁺ in DCs. Moreover, the cytokine levels of IL‐12p70, IL‐6, and IL‐10 secreted by DCs in the supernatant were measured using ELISA.

4.14. RNA Transcriptome Sequencing

Hepa1‐6 cells were placed in 12‐well plates at a density of 1 × 10⁵ cells per well and incubated under hypoxic conditions overnight. For the experimental group, the cells were treated with DMP@FGMS (0.25 mg) under hypoxic conditions for 12 h. The hypoxic cells without any treatment were used as the control group. After co‐incubation for another 24 h, the cells were collected for RNA sequencing (RNA‐seq) analysis. RNA‐seq was performed by the MGISEQ‐T7 platform at Beijing Genome Institute (BGI‐tech, Shenzhen, China). Differential gene expression analysis was conducted using DESeq2. Significantly differentially expressed genes were further subjected to Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and Gene Ontology (GO) enrichment analyses to elucidate their potential biological functions and pathways. According to the results of differential gene detection, the R package pheatmap was used to perform hierarchical clustering analysis on the union set differential genes. The results were visualized using the Dr. Tom system (BGI‐tech, Shenzhen, China).

4.15. In Vitro Ferroptosis Evaluation

To investigate the effect of DMP@FGMS on oxidative stress in hepatocellular carcinoma cells, we assessed its impact on intracellular ROS levels, lipid peroxidation (using C11‐BODIPY 581/591), mitochondrial membrane potential, and mitochondrial oxidative status in Hepa1‐6 cells. Briefly, Hepa1‐6 cells were seeded in 12‐well plates at a density of 1 × 10⁵ cells per well and incubated overnight under hypoxic conditions. For the treatment group, cells were exposed to DMP@FGMS (0.25 mg) under hypoxia for 12 h, while untreated hypoxic cells served as the control. After a total of 24 h of incubation, fluorescence intensities of DCF (for ROS), C11‐BODIPY (for lipid peroxidation), JC‐1 (for mitochondrial membrane potential), and MitoSOX Red (for mitochondrial superoxide) were measured by flow cytometry and confocal microscopy.

In a separate experiment, the GSH/GSSG ratio was evaluated using a commercial GSH/GSSG assay kit according to the manufacturer's protocol. Malondialdehyde (MDA) levels were determined using an MDA assay kit. Cell viability was assessed after co‐treatment with 1 µM Ferrostatin‐1 (Fer‐1) and DMP@FGMS.

4.16. Biodistribution Evaluation In Vivo

Tumor‐bearing mice with tumor volumes reaching 100 mm³ were selected. Cy5.5 was used as a substitute for DOX and MET loaded onto the PFGMS to minimize potential interference from tissue autofluorescence. The Cy5.5‐labeled PFGMS (Cy5.5@PFGMS) and free Cy5.5 were resuspended in 100 µL of PBS and administered intratumorally. Their biodistribution was then visualized at designated time points using the IVIS Spectrum imaging system (Lumina III, PerkinElmer, USA).

4.17. In Vivo Anticancer Efficacy of DMP@FGMS

In vivo anticancer efficacy of DMP@FGMS was first evaluated using a unilateral subcutaneous HCC tumor model. When tumor volumes reached 80–100 mm³, mice were randomly assigned to five groups and received the following treatments: (I) Control, (II) D@FGMS, (III) DP@FGMS, (IV) DM@FGMS, and (V) DMP@FGMS (n = 6 per group). Different types of oxygen‐loaded microsphere suspensions (MSs, 25 mg) were injected into the tumor sites at day 0. Tumor volumes were measured every 2 days throughout the treatment period and calculated using Equation 3:

$$V=\left(L\times W^2\right)/2$$ (3)

where L and W represent the tumor length and width, respectively. Additionally, on day 14, ROS, H&E, and Ki‐67 staining of tumor tissue were performed among the different groups.

4.18. In Vivo Immune Response Evaluation

C57BL/6 mice bearing subcutaneous Hepa 1–6 tumors were subjected to different treatments as described in Section 2.15. Tumor tissues were excised from each group on day 5, followed by the preparation of single‐cell suspensions using a standard protocol. The collected single‐cell suspensions were stained with the following fluorescent antibodies for flow cytometry analysis. Tumoral infiltration of CD8⁺ T cells (CD3⁺CD4⁻CD8⁺), CD4⁺ T cells (CD3⁺CD4⁺CD8⁻), exhausted CD8⁺ T cells (TIM‐3⁺CD8⁺), regulatory T cells (Tregs; CD3⁺CD4⁺Foxp3⁺), NK cells (CD49b⁺CD3⁻CD45⁺), M1 macrophages (F4/80⁺CD11b⁺CD86⁺), and M2 macrophages (F4/80⁺CD11b⁺CD206⁺) in tumors were analyzed. Additionally, the concentrations of the TNF‐α and IFN‐γ in the tumor tissue were assessed using appropriate ELISA kits. In addition, the tumor tissues stained with HIF1‐α, CRT, HMGB1, ROS and pimonidazole fluorescent antibodies were assessed through immunofluorescence staining.

4.19. In Vivo Combination Therapy Against Bilateral Hepa 1–6 Tumor

The bilateral subcutaneous tumor‐bearing mouse model was utilized to investigate the in vivo combination therapy of DMP@FGMS and αPD‐L1. The tumor‐bearing mice were randomly divided into four groups and received the following treatments: (I) Control, (II) αPD‐L1, (III) DMP@FGMS, and (IV) DMP@FGMS + αPD‐L1 (n = 6 per group). For groups III and IV, oxygen‐loaded microsphere suspensions (MS, 25 mg) were directly injected into the primary tumor site on day 0. In group IV, αPD‐L1 (10 µg per rat) was intravenously administered on days 1, 3, and 6. The tumor volumes at both primary and distant sites were measured every 2 days. In an independent experiment, distant tumors were collected on day 7 for flow cytometry to quantify CD8⁺ T cells, CD4⁺ T cells, regulatory T cells, NK cells, and M1/M2 macrophages.

4.20. In Vivo Interventional Embolization Therapy

In vivo interventional embolization therapy was assessed using an orthotopic N1S1 tumor‐bearing rat model. Tumor formation was confirmed 1‐week post‐implantation using a 3.0 T MRI scanner (Achieva 3.0TX, Philips, Netherlands). The rats were then randomly assigned to five groups: (I) Control, (II) αPD‐L1, (III) D@FGMS, (IV) DMP@FGMS, and (V) DMP@FGMS + αPD‐L1 (n = 5 per group). For groups III, IV, and V, oxygen‐loaded microsphere suspensions (25 mg MS per rat) were injected into the hepatic artery of the orthotopic liver tumors using a catheter immediately after surgical exposure of the abdominal cavity on day 0. In groups II and V, αPD‐L1 (100 µg per rat) was administered intravenously on day 1, 3, and 6. MRI scans were performed 1 week after tumor implantation and at 3, 7, and 14 days aftertreatment to calculate tumor volumes. In a separate experiment, tumor tissues were harvested 7 days after treatment and subjected to H&E and Ki‐67 staining to evaluate the therapeutic efficacy of the combined DMP@FGMS and αPD‐L1 treatment against orthotopic HCC. In addition, the infiltration of CD8⁺ T cells and Tregs in tumor tissues was analyzed by flow cytometry. Immunofluorescence staining was performed using antibodies against CD3, FoxP3, HIF‐1α, CRT, HMGB1, and pimonidazole.

In another experiment, tumors from rats treated with iodized oil TACE or without TACE were collected on the fourth day. Immunohistochemistry was employed to evaluate the expression of hypoxia and drug resistance genes (HIF‐1α, VEGF, P‐gp) following embolization. Additionally, tumor sections were stained with pimonidazole to investigate tumor hypoxia.

4.21. In Vivo Biosafety Evaluation

Moreover, on the 14th day, H&E staining was employed to evaluate the safety of various microspheres, focusing on the liver, heart, lungs, spleen, and kidneys. In addition, we assessed the concentrations of white blood cells (WBC), red blood cells (RBC), platelets (PLT), hemoglobin (HGB), alanine aminotransferase (ALT), and aspartate aminotransferase (AST) in the blood of rats from each group.

4.22. Statistical Analysis

All data are presented as mean ± standard error of the mean and were analyzed using GraphPad Prism 8.0 software. Statistical significance was evaluated using an unpaired two‐tailed Student's t‐test, one‐way ANOVA or two‐way ANOVA, as appropriate. A p‐value < 0.05 was considered statistically significant. Statistical significance was indicated as follows: *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001.

Conflicts of Interest

All other authors declare they have no competing interests.

Supporting information

Supporting File: smll74482‐sup‐0001‐SuppMat.docx.

Data Availability Statement

All data are available in the main text or the supplementary materials.