Synergistic immunotherapeutic effects of irreversible electroporation and CAR-NK cell therapy against hepatocellular carcinoma

TL;DR

Combining IRE and CAR-NK cell therapy improves treatment of liver cancer by enhancing immune response and reducing tumor resistance.

Contribution

A novel combination of IRE and LNP-engineered CAR-NK cells to overcome tumor resistance in hepatocellular carcinoma.

Findings

IRE promotes immunogenic cell death and enhances NK cell infiltration by releasing chemokines like CX3CL1.

IRE-treated HCC cells show increased susceptibility to NK-mediated cytotoxicity due to higher reactive oxygen species.

The combination of IRE and CAR-NK cells leads to synergistic antitumor activity in HCC models without systemic toxicity.

Abstract

Chimeric antigen receptor natural killer (CAR-NK) cell therapy has emerged as a promising immunotherapeutic modality with potent cytotoxicity and a favorable safety profile. However, its therapeutic efficacy is often limited by poor infiltration into tumors and the profoundly immunosuppressive tumor microenvironment (TME). In hepatocellular carcinoma (HCC), one of the leading causes of cancer-related mortality, this suppressive TME severely compromises the function of CAR-NK cells. To overcome this limitation, we developed a combinatorial strategy integrating irreversible electroporation (IRE), a clinically approved nonthermal ablation modality capable of reshaping the TME, with glypican-3 (GPC3)-targeted CAR-NK cells generated via 1,2-dioleoyl-3-trimethylammonium-propane (DOTAP)-functionalized lipid nanoparticle (LNP)-mediated gene delivery. IRE promoted immunogenic cell death and reprogrammed the TME through the release of damage-associated molecular patterns and chemokines, notably CX3CL1, thereby enhancing NK cell infiltration. Moreover, IRE-treated HCC cells exhibited heightened susceptibility to NK-mediated cytotoxicity through elevated intracellular reactive oxygen species, establishing a mechanism of immune sensitization. When combined with LNP-engineered GPC3-specific CAR-NK cells, this approach elicited synergistic antitumor activity, as demonstrated by superior tumor lysis in vitro and robust tumor regression in various HCC models without systemic toxicity. By coupling TME remodeling of IRE with the precision and safety of LNP-engineered CAR-NK cells, we propose a durable, clinically actionable treatment paradigm to overcome resistance in solid tumors, such as HCC.

Figures

Irreversible electroporation modulates the tumor immune microenvironment. a Experimental scheme. Orthotopic liver tumors were established in C57BL/6 mice using Hepa1c1c7 cells. After 10 days, IRE was directly administered to the tumor region in the liver (*n* = 3 for the control group, *n* = 4 for each experimental group). b H&E staining of liver tissue sections 72 h post-IRE in the orthotopic Hepa1c1c7-bearing HCC tumor model. Scale bar = 200 μm. c Chemokine expression levels and d heatmap in Hepa1c1c7 tumors at the indicated time points following IRE treatment. Reference sequence, R.S. Flow

PEF exposure to liver cancer cells induces the recruitment of NK cells via CX3CL1 chemotaxis. a Representative images from live cell analysis of Hepa1c1c7 and Huh-7 cells after 24 h of PEF exposure. The red and yellow areas indicate the ablation zone. b ATP release from Hepa1c1c7 and Huh-7 cells after 2 h of PEF exposure. c HMGB1 released from Hepa1c1c7 and Huh-7 cells after 24 h of PEF exposure d (Left) Representative flow cytometric histogram and (right) quantified bar graph of calreticulin-presented Hepa1c1c7 and Huh-7 cells after 24 h of PEF exposure. e Migration analysis of CFSE-labeled N

The PEF-mediated HMGB1-ROS axis sensitizes residual liver cancer cells to NK cells. a Analysis of NK92MI-mediated cytotoxicity against PEF-treated Huh-7 cells at each effector cell/target cell (E/T) ratio for 4 h. b (Left) Representative flow cytometry plots and (right) graph of mNK-mediated cytotoxicity against Hepa1c1c7 control and PEF-treated cells for 4 h at each E/T ratio. c Fluorescence images of PEF-treated Huh-7 cells with NK92MI cells (E/T = 2.5). Cancer cells were labeled with CFSE. Scale bar = 50 μm. d Cleaved caspase-3 expression in PEF-treated Huh-7 cells cocultured with NK92MI ce

Synergetic effects of PEF and anti-GPC3 CAR-NK cells on liver cancer cells in vitro. a Design of *anti-human GPC3 CAR* (*anti-hGPC3*) and *anti-murine Gpc3 CAR* (*anti-mGpc3*) DLNPs. b Particle size of *Fluc*, *anti-hGPC3-, or anti-mGpc3*-DLNPs measured by DLS. c Electron microscopy images of *anti-hGPC3* or *anti-mGpc3*-DLNPs. Scale bar = 100 nm. d Zeta potential of *Fluc*-, *anti-hGPC3-*, or *anti-hGPC3-*DLNPs by DLS. e Encapsulation efficiency (%) of *Fluc-, anti-hGPC3-*, or *anti-mGpc3-*mRNA within DLNPs using a RiboGreen assay. f Immunoblot analysis of CD3ζ expression in *anti-hGPC3*-DLNP

Full text

Introduction

Liver cancer is one of the leading causes of cancer-related deaths.^1^ Hepatocellular carcinoma (HCC), the predominant type of primary liver cancer, presents a serious global health concern, with rising mortality rates. Current therapeutic options, such as radiofrequency ablation, transarterial chemoembolization, and surgical resection, can be effective for selected patients but often fail to prevent recurrence due to tumor heterogeneity and underlying liver disease.^2,3^ Consequently, there remains a critical urgent need for an innovative therapeutic approach that improves long-term clinical outcomes in HCC.

In the landscape of cancer immunotherapy, chimeric antigen receptor (CAR)-based strategies have reshaped therapeutic paradigms by enabling immune cells to selectively eliminate malignant cells.^4,5^ CAR-modified natural killer (NK) cells have emerged as an alternative cellular immunotherapy^6–8^, offering advantages such as reduced risk of severe graft-versus-host disease and cytokine release syndrome compared to CAR-T-cell therapies.^9–11^ In HCC, NK cells play a central role in the innate immune surveillance, and their recruitment to tumor sites can greatly enhance cancer reduction efficacy.^12,13^ However, the immunosuppressive tumor microenvironment (TME) in HCC remains a major barrier, limiting NK cell infiltration and impairing cytotoxic function.^14,15^ Therefore, overcoming resistance is essential to improve the therapeutic efficacy of CAR-NK cell therapy in HCC. To address these challenges, novel therapeutic strategies that modulate the immune landscape from a desert to an inflamed state by enhancing immune cell infiltration into tumors are essential.

Irreversible electroporation (IRE), which has been clinically approved in the US by the Food and Drug Administration (FDA), has emerged as a promising nonthermal ablation modality that uses high voltage pulsed electric fields (PEFs) to reshape the immune landscape of solid tumors, such as HCC and pancreatic ductal adenocarcinoma (PDAC).^16,17^ IRE is considered suitable for anatomically challenging locations because it preserves the extracellular matrix and vasculature while inducing irreversible permeabilization of tumor cell membranes, ultimately triggering immunogenic cell death (ICD) through loss of cellular homeostasis.^18^ As a consequence, this process releases damage-associated molecular patterns (DAMPs), along with proinflammatory cytokines and chemokines, resulting in an in situ vaccination effect by modulating the tumor immune microenvironment (TIME).^19^ Furthermore, recent studies on various tumor models, such as rat sarcoma^20^, murine renal carcinoma^21^, and canine glioma^22^, have demonstrated that IRE enhances antitumor efficacy in immunocompetent animals, suggesting the potential participation of the host’s immunity. Nonetheless, the durability of antitumor control achieved by IRE alone remains limited, underscoring the need for rational combination strategies to maximize its long-term clinical benefit.

In this study, we integrated IRE as a TME-remodeling modality with glypican-3 (GPC3)-targeted CAR-NK cells generated using a virus-free DOTAP-functionalized lipid nanoparticle (DLNP) platform. Beyond its direct ablative effects, IRE reshapes the TIME toward an immune-permissive state, thereby promoting NK cell recruitment and increasing the susceptibility of cancer cells to NK cell-mediated cytotoxicity. Within the remodeled niche by IRE, engineered CAR-NK cells for GPC3 exhibit robust antitumor activity in the absence of systemic toxicity, and the combination therapy of IRE with CAR-NK cells further enhances adaptive immune responses. These findings highlight a synergistic therapeutic effect in which local TME modulation via IRE augments the efficacy of CAR-NK cell therapy in HCC.

Results

Irreversible electroporation modulates the tumor immune microenvironment of hepatocellular carcinoma

We previously confirmed that IRE induces ICD of tumors in CT-26 tumor-bearing mice, releasing immunomodulatory proteins and molecules such as DAMPs, cytokines, and chemokines, and regulating immunity.^23^ To expand the effects of IRE on HCC in terms of moderating immunity, we established an orthotopic liver tumor mouse model by inoculating Hepa1c1c7 murine cancer cells into C57BL/6 mice, followed by treatment with IRE directly into the tumors (Fig. 1a). After IRE treatment, histological observations revealed infiltrated immune cells in the IRE-treated area (Fig. 1b). Given the observed infiltration of immune cells, we next examined the chemokine profile of IRE-treated HCC tumors to assess whether chemotactic signals contribute to immune cell recruitment. IRE-treated HCC tumors showed elevated expression levels of various chemokines, including CX3CL1, CCL6, CCL8, CCL9/10, CCL12, CCL22, CXCL1, CXCL2, CXCL6, CXCL9, and chemerin (Fig. 1c, d). To evaluate the IRE-mediated alteration of tumor-infiltrating innate immune cell populations, we analyzed the populations of NK cells, neutrophils, and DCs in HCC tumors (Supplementary Fig. 1). As shown in Fig. 1e, f, h, i, NK cells were significantly increased as early as 3 h post-IRE, while neutrophils began to accumulate at 6 h, as reported.^24,25^ DCs were significantly elevated at 72 h (Fig. 1g, j). We focused on NK cells, which showed the highest degree of immune cell infiltration after IRE. Fluorescence imaging analysis revealed an increased distribution of NKp46^+^ cells in HCC tumor tissues after IRE treatment (Fig. 1k). These results demonstrated that IRE treatment of HCC modulates the TIME by promoting the release of various chemokines and recruiting innate immune cells, particularly NK cells.Fig. 1. Irreversible electroporation modulates the tumor immune microenvironment. a Experimental scheme. Orthotopic liver tumors were established in C57BL/6 mice using Hepa1c1c7 cells. After 10 days, IRE was directly administered to the tumor region in the liver (n = 3 for the control group, n = 4 for each experimental group). b H&E staining of liver tissue sections 72 h post-IRE in the orthotopic Hepa1c1c7-bearing HCC tumor model. Scale bar = 200 μm. c Chemokine expression levels and d heatmap in Hepa1c1c7 tumors at the indicated time points following IRE treatment. Reference sequence, R.S. Flow cytometry analysis of Hepa1c1c7-bearing HCC tumor-infiltrated immune cells at the indicated times after IRE, including e, h NK cells (NK1.1^+^), f, i neutrophils (Ly6G^+^), and g, j DCs (CD11c^+^MHCⅡ^+^). Data are presented as the mean ± S.D. (Control, n = 3; Experimental groups, n = 4). k Representative immunofluorescence images and quantified data of NKp46^+^ cells in Hepa1c1c7-bearing HCC tumors 72 h after IRE treatment. Scale bar = 50 μm. Data represent the mean ± S.D. (n = 3 biologically independent samples per group). Statistical significance was assessed using a two-tailed Student’s t test (h, k). Panel a was created with BioRender.com

Pulsed electric field promotes NK cell recruitment to liver cancer cells via CX3CL1-mediated chemotaxis

Because infused allogeneic NK cells preferentially accumulate in the liver^26,27^ and IRE markedly enhances NK cell infiltration into HCC lesions as early as 3 h post-IRE (Fig. 1h), we hypothesized that combining IRE with allogeneic NK cell therapy could further enhance therapeutic efficacy against HCC. To explore the immunological basis of this therapeutic potential, we next examined whether PEF, an in vitro-applicable pulsed electrical stimulation, could trigger ICD in liver cancer cells in vitro. For this purpose, human (Huh-7) and murine (Hepa1c1c7) liver cancer cells were exposed to a PEF. This procedure generated a distinct ablation zone in both cell lines (Fig. 2a). Since DAMPs primarily mediate ICD, we quantified extracellular ATP and HMGB1, two representative DAMPs, in culture supernatants. PEF exposure increased ATP release and HMGB1 (Figs. 2b and 2c). In addition, the cell-surface presentation of calreticulin, another ICD marker, was elevated following PEF treatment (Fig. 2d).Fig. 2PEF exposure to liver cancer cells induces the recruitment of NK cells via CX3CL1 chemotaxis. a Representative images from live cell analysis of Hepa1c1c7 and Huh-7 cells after 24 h of PEF exposure. The red and yellow areas indicate the ablation zone. b ATP release from Hepa1c1c7 and Huh-7 cells after 2 h of PEF exposure. c HMGB1 released from Hepa1c1c7 and Huh-7 cells after 24 h of PEF exposure d (Left) Representative flow cytometric histogram and (right) quantified bar graph of calreticulin-presented Hepa1c1c7 and Huh-7 cells after 24 h of PEF exposure. e Migration analysis of CFSE-labeled NK92MI and f mouse primary NK cells (mNK) toward conditioned medium (CM) from PEF-exposed liver cancer cell lines (PEF-CM). Scale bar = 100 μm and 50 μm, respectively. g Concentration of CX3CL1 in PEF-CM from Hepa1c1c7 and Huh-7 cells. h Concentration of CX3CL1 at the indicated time points after IRE treatment in orthotopic Hepa1c1c7 tumors in C57BL/6 mice. i Migration analysis of CFSE-labeled NK92MI cells toward PEF-CM, including CX3CL1-neutralizing antibody. Scale bar = 100 μm. j Caspase-3/7 activation in PEF-exposed Huh-7 cells cocultured with migrated NK92MI cells. Scale bar = 50 μm. k Scheme of PEF-mediated CX3CL1 release and NK cell recruitment. Data represent the mean ± S.D. (n = 3; biological replicates). A two-tailed Student’s t test was used for the P value (b–j). Panels e, f, j, and k were created with BioRender.com

Next, to determine whether PEF-induced soluble factors mediate NK cell recruitment, we performed transwell migration assays using conditioned medium (CM) from control liver cancer cell lines (Con-CM) or PEF-treated liver cancer cell lines (PEF-CM). NK92MI cells, a human NK cell line, exhibited significantly higher migration toward PEF-CM than Con-CM from Huh-7 cells (Fig. 2e). Similarly, mouse primary NK cells (mNK) showed enhanced migration toward Hepa1c1c7 PEF-CM (Fig. 2f). These results suggest that soluble factors released from liver cancer cells following PEF exposure promote NK cell recruitment. Based on the increased expression levels of various chemokines in IRE-treated HCC tumors (Fig. 1c, d), we investigated the chemokines related to the NK cell migration in PEF-CM from Huh-7 cells (Supplementary Fig. 2). Given that CX3CL1 plays a crucial role in the NK cell-mediated chemotaxis^28,29^, we measured the amount of released CX3CL1 in PEF-CM. CX3CL1 secretion was significantly elevated in the PEF-CM group compared with the Con-CM group (Fig. 2g). Consistently, in vivo evaluation revealed that the expression level of CX3CL1 in IRE-treated HCC tumors was significantly increased, with notable elevations observed at 3, 6, and 24 h post-IRE (Fig. 2h). Importantly, blockade of CX3CL1 with a neutralizing antibody significantly reduced NK92MI migration toward PEF-CM (Fig. 2i). To evaluate whether migrated NK cells exhibit an effective cancer-killing effect, PEF-exposed Huh-7 cells were cocultured with NK92MI cells using a transwell system. Figure 2j shows that Caspase-3/7 was highly stimulated in PEF-exposed Huh-7 cells by migrated NK92MI cells compared to control cells. Notably, blocking CX3CL1 in PEF-exposed Huh-7 cells diminished Caspase-3/7 activation, although activation remained higher than in control cells. Overall, these results indicate that PEF induces ICD in liver cancer cells, accompanied by the release of DAMPs and increased secretion of CX3CL1. This PEF-induced elevation of CX3CL1 markedly promotes NK cell chemotaxis, providing a mechanistic basis for the observed synergistic potential of IRE and allogeneic NK cell therapy (Fig. 2k).

Pulsed electric field-induced immunogenic signals and oxidative stress sensitize residual liver cancer cells to NK cells

To assess whether PEF enhances the susceptibility of residual viable liver cancer cells to NK cell-mediated cytolysis, we cocultured NK92MI cells with Huh-7 cells exposed to PEF. PEF-treated Huh-7 cells exhibited significantly higher lysis by NK92MI cells than untreated controls across various effector-to-target (E/T) ratios (Fig. 3a). A similar increase in sensitivity was observed when Hepa1c1c7 cells were cocultured with mNKs (Fig. 3b). Coculture imaging further confirmed the rapid clearance of CFSE-labeled PEF-treated Huh-7 cells by NK92MI cells (Fig. 3c). These results are supported by immunoblot analysis of cleaved caspase-3, which indicates increased apoptosis in PEF-treated Huh-7 cells cocultured with NK92MI cells (Fig. 3d). We next sought to elucidate the mechanism by which PEF treatment sensitizes liver cancer cells to NK cells. Since NKG2D ligands such as stress-inducible protein ULBPs are well-known mediators of NK recognition^30^, we examined their expression after PEF treatment; however, no significant changes were observed (Supplementary Fig. 3).Fig. 3. The PEF-mediated HMGB1-ROS axis sensitizes residual liver cancer cells to NK cells. a Analysis of NK92MI-mediated cytotoxicity against PEF-treated Huh-7 cells at each effector cell/target cell (E/T) ratio for 4 h. b (Left) Representative flow cytometry plots and (right) graph of mNK-mediated cytotoxicity against Hepa1c1c7 control and PEF-treated cells for 4 h at each E/T ratio. c Fluorescence images of PEF-treated Huh-7 cells with NK92MI cells (E/T = 2.5). Cancer cells were labeled with CFSE. Scale bar = 50 μm. d Cleaved caspase-3 expression in PEF-treated Huh-7 cells cocultured with NK92MI cells for 4 h (E/T = 2.5). β-Actin was a loading control. Quantification was performed with ImageJ. e (Left) Representative images and (right) quantified bar graph of DCF-DA staining for ROS levels of PEF-treated Huh-7 cells with or without N-acetyl-L-cysteine (NAC). Scale bar = 200 μm. Fluorescence intensities were measured by ImageJ. f Analysis of NK92MI-mediated cell lysis of PEF-treated Huh-7 cells with or without NAC for 4 h (E/T = 10). g Cleaved caspase-3 expression in PEF-treated Huh-7 cells with NK92MI according to NAC treatment (E/T = 2.5). h Representative images and quantified graph of DCF-DA staining in Huh-7 and Hepa1c1c7 cells treated with PEF-CM from Huh-7 and Hepa1c1c7 cells, respectively. H_2_O_2_ was a positive control. i Representative images and quantified data of ROS levels in HMGB1KD Huh-7 cells after PEF exposure. j (Left) Representative flow cytometry plots and (right) quantification of NK92MI-mediated cytolysis against control or HMGB1KD Huh-7 cells after PEF exposure (E/T = 10). k Schematic illustration of the proposed mechanism. Data represent mean ± S.D. (n = 3). A two-tailed Student’s t test was evaluated using statistical significance (a, b, and d–j). Panel k was created with BioRender.com

Previously, some studies reported that PEF treatment triggers the production of intracellular ROS in cancer cells, and NK cell cytotoxicity toward the target cells is ROS dependent.^31–35^ Thus, we hypothesized that PEF-mediated ROS production in cancer cells was linked with susceptibility to NK cells. As expected, PEF treatment induced an ~2-fold elevation in ROS levels in liver cancer cells (Supplementary Fig. 4). Importantly, the increased susceptibility of PEF-exposed Huh-7 cells to NK92MI was abolished when ROS were scavenged with N-acetyl-L-cysteine (NAC), restoring resistance to levels comparable to those of the untreated control (Fig. 3e, f). Consistently, Caspase-3 activation induced by NK92MI in PEF-exposed cells was attenuated in the presence of NAC (Fig. 3g).

To further investigate ICD-mediated ROS accumulation, we examined the effect of CM from PEF-exposed liver cancer cells. Treatment with PEF-CM increased intracellular ROS production in both Huh-7 and Hepa1c1c7 cells (Fig. 3h), indicating that soluble factors released by PEF contribute to paracrine ROS elevation. Given that HMGB1 plays a role in ROS production^36^, blockade of HMGB1 with a neutralizing antibody significantly reduced ROS generation compared to a control antibody in PEF-CM (Supplementary Fig. 5). Supporting this result, when PEFs were exposed to HMGB1 knockdown (HMGB1KD), ROS were not induced in Huh-7 cells compared to Huh-7 control cells (Supplementary Fig. 6 and Fig. 3i), implicating DAMPs such as HMGB1 as key mediators of ROS generation. Consistently, PEF-mediated sensitization to NK cells was diminished by suppressing HMGB1 in Huh-7 cells (Fig. 3j). Collectively, these findings demonstrate that PEF treatment sensitizes residual liver cancer cells to NK cell-mediated killing through a DAMP-ROS axis. By promoting intracellular ROS accumulation, PEF treatment renders liver cancer cells more susceptible to NK cells (Fig. 3k).

Genetic engineering of NK cells using cationic lipid nanoparticles

Building on prior evidence implicating IRE in modulating antitumor immune responses, we next explored whether enhancing NK cell functionality through genetic engineering could further improve therapeutic outcomes. Previous clinical studies combining IRE with allogeneic NK cell therapy have reported modest survival benefits compared with IRE alone in patients with HCC.^37,38^ Thus, we hypothesized that the combination of IRE and CAR-NK cells could enhance antitumor efficacy in HCC. However, the broader application of CAR-NK cells remains limited by inefficient gene delivery to NK cells.^39^ To address this challenge, we employed DLNPs, which we previously developed for successful genetic engineering of NK cells.^40^ To confirm the capability of DLNPs to enable mRNA delivery to NK cells, we synthesized DLNPs encapsulating EGFP mRNA as previously established (Supplementary Fig. 7). The particle sizes of conventional LNPs (as control LNPs) and DLNPs were approximately 162 and 155 nm, respectively (Supplementary Fig. 8a). Owing to the cationic lipid DOTAP, the zeta potential of the DLNPs shifted positively (Supplementary Fig. 8b). Additionally, DLNPs showed remarkable mRNA encapsulation efficiency, exceeding 90% (Supplementary Fig. 8c). To investigate the efficacy of DLNP-mediated mRNA delivery to NK cells, NK92MI cells were treated with EGFP mRNA encapsulating DLNPs. While NK92MI cells exhibited low EGFP expression in the conventional LNP group, DLNPs noticeably increased EGFP expression in NK92MI cells (Supplementary Fig. 8d). Flow cytometry analysis confirmed that the EGFP mRNA transfection efficiency was above 50% in DLNP-treated NK92MI cells (Supplementary Fig. 8e).

Because GPC3 is frequently overexpressed in HCC and serves as a well-established therapeutic target for CAR-based cell therapy^2,41^, we focused on generating anti-GPC3 CAR-NK cells using DLNPs. Before establishing anti-GPC3 CAR-NK cells, we determined whether PEF treatment altered GPC3 expression in liver cancer cells. In Huh-7 cells, the expression of GPC3 was increased by PEF, whereas in Hepa1c1c7 cells, GPC3 expression was not significantly altered by PEF (Supplementary Fig. 9). These results indicate that GPC3 is a suitable target for CAR-NK cells combined with IRE.

Synergistic effects of pulsed electric field with anti-glypican-3 CAR-NK cells on liver cancer cells in vitro

To evaluate the therapeutic potential of combining PEF with anti-GPC3 CAR-NK cells to liver cancer cells, we generated CAR-NK cells using DLNPs encapsulating either anti-human GPC3 CAR mRNA (Anti-hGPC3-DLNPs) or anti-murine Gpc3 CAR mRNA (Anti-mGpc3-DLNPs) (Fig. 4a, Supplementary Fig. 10, and Supplementary Fig. 11). Fluc-DLNPs served as vehicle controls. DLS revealed particle sizes of ~148 nm for Fluc-DLNPs, 123 nm for anti-hGPC3-DLNPs, and 126 nm for anti-mGpc3-DLNPs (Fig. 4b). Cryo-EM confirmed the spherical morphology of both anti-hGPC3- and anti-mGpc3-DLNPs (Fig. 4c). The zeta potential of all DLNP formulations was approximately 30 mV, with encapsulation efficiencies exceeding 90% (Fig. 4d, e). Immunoblotting validated CAR expression in NK92MI and mNK cells following transfection of anti-hGPC3-DLNPs and anti-mGpc3-DLNPs, respectively (Fig. 4f).Fig. 4. Synergetic effects of PEF and anti-GPC3 CAR-NK cells on liver cancer cells in vitro. a Design of anti-human GPC3 CAR (anti-hGPC3) and anti-murine Gpc3 CAR (anti-mGpc3) DLNPs. b Particle size of Fluc, anti-hGPC3-, or anti-mGpc3-DLNPs measured by DLS. c Electron microscopy images of anti-hGPC3 or anti-mGpc3-DLNPs. Scale bar = 100 nm. d Zeta potential of Fluc-, anti-hGPC3-, or *anti-hGPC3-*DLNPs by DLS. e Encapsulation efficiency (%) of Fluc-, anti-hGPC3-, or *anti-mGpc3-*mRNA within DLNPs using a RiboGreen assay. f Immunoblot analysis of CD3ζ expression in anti-hGPC3-DLNP-treated NK92MI (CAR-NK92MI) and *anti-mGpc3-*DLNP-treated mNK (CAR-mNK) cells. g Schematic illustration of the production of CAR-NK cells cocultured with PEF-treated liver cancer cells. h Analysis of CAR-NK92MI-mediated cytotoxicity against PEF-treated Huh-7 and control cells for 4 h (E/T = 10). Fluc-DLNP-treated NK92MI (Con-NK92MI) cells were used as control NK cells. i Analysis of CAR-mNK-mediated cytotoxicity against PEF-treated Hepa1c1c7 and control cells for 4 h (E/T = 10). Fluc-DLNP-treated mNK cells were used as control NK cells (Con-mNK). j Immunoblot of cleaved caspase-3 in PEF-treated Huh-7 cells cocultured with Con- or CAR-NK92MI cells for 4 h (E/T = 2.5). Band intensities were measured by ImageJ, with values represented beneath the blot. k (Right) Fluorescence images and (left) quantified data of PEF-treated Huh-7 cells cocultured with Con- or CAR-NK92MI cells for 4 h (E/T = 2.5). Huh-7 cells and NK92MI cells were stained with CFSE and Far-red, respectively. Scale bar = 50 μm. l ROS levels of cocultured Huh-7 cells with Con- or CAR-NK92MI cells after PEF exposure (E/T = 2.5). Data represent mean ± S.D. (n = 3). A two-tailed Student’s t test was used (h, k, and l). Panels a and g were created with BioRender.com

We next examined whether combining PEF with anti-GPC3 CAR-NK cells potentiates their cytolytic activity against liver cancer cell lines in vitro. As illustrated in Fig. 4g, Huh-7 and Hepa1c1c7 cells exposed to PEF were cocultured with either anti-GPC3 CAR-NK or Fluc-NK control (Con-NK) cells. As shown in Fig. 4h, PEF-exposed Huh-7 cells exhibited increased susceptibility to anti-hGPC3 CAR-NK92MI (CAR-NK92MI) compared with Con-NK92MI. Similarly, the combination of PEF and anti-mGPC3 CAR mNK (CAR-mNK) yielded the highest lysis of Hepa1c1c7 cells compared with Con-mNK (Fig. 4i). Immunoblotting further confirmed robust Caspase-3 activation in PEF-exposed Huh-7 cells cocultured with CAR-NK92MI cells (Fig. 4j). Imaging assays revealed that CAR-NK92MI cells exhibited markedly enhanced recognition and binding to PEF-exposed Huh-7 cells (Fig. 4k). To investigate whether PEF treatment modulates ROS generation in target cells upon NK cell engagement, DCF-DA staining was performed after coculturing Huh-7 cells with either Con-NK92MI or CAR-NK92MI cells. While untreated Huh-7 cells showed minimal ROS induction, PEF-pretreated Huh-7 cells exhibited markedly elevated ROS levels upon NK cell coculture. Notably, CAR-NK92MI further amplified ROS production in PEF-exposed Huh-7 cells compared with untreated cells, indicating that PEF sensitizes Huh-7 cells to NK cell-mediated oxidative stress and that CAR engineering potentiates this effect (Fig. 4l). Taken together, these findings demonstrate that PEF treatment sensitizes liver cancer cells to CAR-NK cell-mediated killing and that the combination of PEF treatment with CAR-NK cell therapy produces synergistic effects on liver cancer cell lysis and apoptosis.

Enhanced antitumor effects of irreversible electroporation with anti-glypican-3 CAR-NK cell therapy against hepatocellular carcinoma

To evaluate the clinical applicability of our approach, we established HCC patient-derived organoids (HCC-PDOs). Following PEF treatment, HCC-PDOs maintained stable GPC3 expression levels, indicating that PEF treatment does not significantly alter the target antigen profile of HCC cells (Supplementary Fig. 12). When PEF-exposed HCC-PDOs were cocultured with CAR-NK92MI cells, we observed markedly increased Caspase-3/7 activation compared with either PEF-treated PDOs cocultured with Con-NK92MI cells or untreated HCC-PDOs cocultured with CAR-NK92MI cells (Fig. 5a, b). To validate these findings in vivo, we next evaluated the antitumor efficacy of IRE with CAR-NK92MI therapy using an Huh-7 xenograft model in NSG mice (Fig. 5c). The body weights did not change appreciably after the combination therapy of IRE with CAR-NK92MI (Supplementary Fig. 13a). While IRE alone modestly inhibited tumor progression, combining IRE with Con-NK92MI produced a marked reduction in tumor growth, and the CAR-NK92MI combination achieved the most pronounced antitumor effect (Figs. 5d and 5e). In the nude mouse Huh-7 xenograft model (Fig. 5f), IRE alone resulted in no tumor-free mice (0/5). The combination of IRE with Con-NK92MI partially suppressed tumor growth, achieving tumor-free outcomes in 2 out of 5 mice. Notably, IRE with CAR-NK92MI elicited the strongest antitumor activity, resulting in tumor-free status in 3 out of 5 mice during the indicated time and markedly extending survival without body weight changes (Fig. 5g, h, and Supplementary Fig. 13b). Furthermore, we also confirmed the antitumor efficacy of IRE with CAR-mNK therapy in vivo using an orthotopic HCC model in nude mice. CAR-mNKs were intravenously administered immediately after IRE treatment (Fig. 5i). Consistent with the findings of Huh-7 xenograft models, the combination therapy with IRE and CAR-mNK exhibited superior therapeutic efficacy compared to either treatment alone (Fig. 5j). Histological analysis and TUNEL assays further revealed increased necrotic and apoptotic regions within HCC tumors in the combination therapy group (Fig. 5k, l). Additionally, the frequency of tumor-infiltrating NK cells was significantly greater in the combination therapy group than in the groups treated with IRE or CAR-mNK alone (Fig. 5m). This combination therapy not only increased intratumoral IFN-γ and TNF-α levels compared with either IRE alone or the CAR-mNK group (Fig. 5n) but also upregulated a broad spectrum of chemokines, such as CCL11, CCL12, CCL21, CCL22, CXCL9, CXCL10, CXCL11, CXCL13, CXCL16, and Chemerin (Supplementary Fig. 14). Biochemical analyses of blood samples, which are critical for monitoring HCC progression, showed that combination therapy significantly restored aspartate aminotransferase (AST), alanine aminotransferase (ALT), total bilirubin (TBIL), and lactate dehydrogenase (LDH) levels to near-normal values (Fig. 5o). Histological analysis of the major organs confirmed that the combination therapy of IRE and CAR-mNK did not cause any adverse effects (Supplementary Fig. 15). Taken together, the combination therapy of IRE and anti-GPC3 CAR-NK cells showed synergistic immunotherapeutic antitumor effects in vivo, significantly enhancing tumor suppression and restoring liver function markers without causing adverse effects.Fig. 5. Enhanced tumor suppression with IRE with anti-GPC3 CAR-NK combination therapy. a Representative image of Caspase-3/7 activation and b quantified data of PEF-exposed HCC-PDOs cocultured with CAR- or CAR-NK92MI cells for 4 h. Con- or CAR-NK92MI cells were stained with Far-red. Scale bar = 50 μm. The fluorescence of cleaved caspase-3/7 was measured by ImageJ. c Experimental timeline. 5 × 10^6^ Huh-7 cells were subcutaneously inoculated into an NSG mouse. After 10 days, IRE was administered directly to the tumor, followed by intravenous injection of Con- or CAR-NK92MI cells on days 10 and 12 (1 × 10^7^ cells/mouse; n = 3). d Average tumor growth curves for mice bearing Huh-7 tumors treated with IRE alone, IRE with Con-NK92MI, or IRE with CAR-NK92MI. e Representative tumor images and weights from each group at 28 days. Scale bar = 1 cm. f Experimental scheme of combination therapy of IRE with CAR-NK92MI to Huh-7 xenograft in nude mice (n = 5 per group). g Individual tumor volume and h survival curves for each group. i Experimental timeline. Hepa1c1c7 cells were orthotopically injected into the livers of nude mice (4 × 10^5^ cells/mouse). After 10 days, IRE was administered directly to the tumor region of the liver, followed by intravenous injection of CAR-mNK (2 × 10^6^/mouse; n = 3). j Representative HCC tumor images and the tumor area from each group. Scale bar = 1 cm. Blue dotted lines indicate HCC tumor regions. k H&E and TUNEL analysis in sections of HCC tumors. Scale bar = 100 μm. l Percentage of TUNEL^+^ area in sections of HCC tumors from each group (n = 3 biologically independent samples per group). Data represent the mean ± S.D. m (Left) Immunofluorescence images and (right) quantified data of NKp46^+^ cells in sections of HCC tumors from each group. n Expression levels of local IFN-γ and TNF-α in HCC tumors from each group. o Serum biochemical analysis of each mouse group. Dotted lines represent average marker values for normal mice (n = 3). Data are presented as the mean ± S.D. Statistical analyses were calculated using two-tailed Student’s t test (b, e, j, l, m, and o), two-way ANOVA (d), and log-rank test (h). Panels c, f, and i were created with BioRender.com

Irreversible electroporation combined with anti-glypican-3 CAR-NK cells enhances dendritic and cytotoxic T-cell immunity

Given that IRE combined with CAR-NK therapy substantially increased intratumoral levels of IFN-γ and TNF-α (Fig. 5n), key cytokines involved in shaping adaptive immunity, we next investigated whether these effects extend beyond the innate response to influence the adaptive immune compartment. We established an orthotopic HCC model in immunocompetent C57BL/6 mice using Hepa1c1c7 cells. Mice were treated with IRE followed by intravenous infusion of either Con-mNK or CAR-mNK, and tumor-infiltrating and systemic immune responses were analyzed 30 days post-treatment (Fig. 6a). Notably, the IRE with CAR-mNK group exhibited a pronounced increase in DC frequency within HCC tumors compared with the IRE alone or IRE with Con-mNK groups (Fig. 6b, c, and Supplementary Fig. 16a). Moreover, the expression levels of the costimulatory molecules CD86 and CD80 were markedly elevated in this group, indicating enhanced DC activation within the TME and spleen (Fig. 6d, e, and Supplementary Fig. 17). Consistent with increased DC activation, IRE with the CAR-mNK group also demonstrated enhanced T-cell immunity. A marked increase in tumor-infiltrating CD8^+^ and CD4^+^ T cell populations was observed in mice receiving CAR-mNKs after IRE compared with the IRE alone groups (Fig. 6f, g, and Supplementary Fig. 16b). Consequently, an increased proportion of splenic CD8^+^ T cells was detected in the IRE with CAR-mNK combination therapy group compared with the IRE alone group (Supplementary Fig. 18). To further assess systemic adaptive responses, antigen-specific CD8^+^ T cells were examined in the splenic samples. The CD8^+^ T cells from the spleen of the IRE with CAR-mNK treatment group showed a higher frequency of IFN-γ expression following stimulation with Hepa1c1c7 antigens than those from the control, IRE alone, and IRE with Con-mNK-treated groups (Fig. 6h, i). In summary, these findings demonstrate that IRE combined with anti-GPC3 CAR-NK therapy not only shows enhanced antitumor effects but also drives robust DC maturation and primes antigen-specific CD8^+^ T-cell responses, thereby promoting a coordinated adaptive immune program capable of supporting durable antitumor immunity.Fig. 6. Enhanced DC maturation and activation of tumor-reactive CD8^+^ T-cells following combined IRE and anti-GPC3 CAR-NK therapy. a Experimental timeline. C57BL/6 mice received intraperitoneal injections of TAA, followed by orthotopic inoculation of Hepa1c1c7 cells into the liver. Ten days later, tumors underwent IRE, after which Con-mNKs or CAR-mNKs were administered intravenously (n = 3). b Representative flow cytometry plots and c quantified bar graph show CD11c^+^ MHCII^+^ DCs (% of CD45^+^ cells) in HCC tumor tissue from each treatment group. d Representative histograms and e quantified MFI data of CD80 and CD86 on DCs (CD11c^+^ MHCII^+^) in HCC tumors. f Flow cytometry plots and g quantified bar graphs display frequencies of CD8^+^ T cells and CD4^+^ T cells (% of CD11b^-^CD3^+^) in HCC tumor tissue for each treatment group. h Representative flow cytometry plots showing IFN-γ expression (% of CD8^+^) of antigen-specific CD8^+^ T cells from the spleens of each group stimulated with Hepa1c1c7 cell lysate. i Frequencies of IFN-γ^+^ antigen-specific CD8^+^ T cells and the corresponding MFI. Data are presented as the mean ± S.D. Statistical significance was assessed using a two-tailed Student’s t test (c, e, g, and i). Panel a was created with BioRender.com

Discussion

HCC presents a profoundly immunosuppressive TME that restricts immune cell infiltration and weakens antitumor immunity. Among innate immune cells, NK cells are critical mediators of early tumor control; however, their recruitment and activation are markedly impaired within the HCC niche.^12,14^ Therefore, strategies that enhance NK cell infiltration and cytotoxicity are essential for improving therapeutic outcomes in HCC.

PEF is a term that describes the application of electrical pulses that modulate membrane permeability, among which IRE represents a high-intensity regimen capable of inducing permanent membrane disruption.^18^ IRE employs ultrashort, high-voltage electrical pulses to ablate tumors while minimizing collateral damage, and concurrently promotes ICD and remodeling of the TIME via the release of various chemokines and cytokines. Owing to these immunomodulatory effects, IRE serves as an ideal partner for combination with immunotherapy, enhancing the antitumor immune response and potentially improving therapeutic outcomes.^42–45^ Here, our findings reveal that NK cell therapy represents the most suitable adoptive cell strategy for HCC during the early window following IRE treatment. By analyzing the extracellular and intracellular factors in liver cancer cells after PEF treatment, we identified the imperative need for an immune reaction in PEF-treated liver cancer cells through CX3CL1-mediated NK cell chemotaxis, as well as the increase in sensitizing liver cancer cells to NK cells after PEF treatment by the DAMPs-ROS axis (Figs. 2 and 3).

Beyond its local ablative role, IRE has also been recognized as an approach that enhances antitumor immunity, particularly by potentiating responses to immune checkpoint inhibition through the activation and recruitment of diverse immune cells.^19,42,43^ Analysis of IRE-treated tumors revealed a broad upregulation of chemokines, including CX3CL1, CCL6, CCL8, CCL9/10, CCL12, CCL22, CXCL1, CXCL2, CXCL6, CXCL9, and chemerin, which are associated with immune cell recruitment and activation (Fig. 1c). Notably, the kinetics of chemokine induction following IRE were not uniform; this temporal heterogeneity suggests that the immune-modulatory landscape created by IRE is dynamically evolving rather than static. Consistent with this pattern, early time points following IRE were characterized by a marked accumulation of tumor infiltration of NK cells together with elevated levels of chemokines involved in NK cell recruitment, including CX3CL1 and CXCL9. Given that CX3CL1 is a major chemokine responsible for directing NK cells to lesion sites in multiple tissues, including HCC, lung, and uterus^46–48^, these findings indicate that the timing of adoptive NK cell infusion is crucial after IRE. In particular, the crucial role of CX3CL1, as confirmed in Fig. 2, suggests that IRE generates a transient but favorable window for NK cell-based immunotherapy. Therefore, early administration of NK cells is likely to maximize therapeutic benefit, highlighting the importance of defining the temporal chemokine profile following IRE to optimize infusion scheduling. Although IRE treatment upregulates CX3CL1 expression in HCC (Fig. 2h), the mechanisms underlying the IRE-mediated increases in chemokine and cytokine levels require further study. IRE forms permanent pores in the cell membrane that can forcibly release intracellular chemokines and cytokines.^49,50^ However, based on the fact that some, but not all, secreted factors are increased, it is speculated that there is a specific mechanism for the increase in each factor. One plausible explanation is the paracrine action of DAMPs released by IRE, which may activate TLR signaling in neighboring viable cells. This process, in turn, could amplify NF-κB and MAPK pathways, resulting in increased production of inflammatory cytokines and chemokines.^51,52^

Beyond enhancing chemokine secretion, IRE also sensitized residual tumor cells to NK-mediated cytotoxicity. A clinical trial utilizing IRE and allogeneic NK cell therapy for treating patients with HCC already exists, but the molecular mechanism remains incompletely defined.^37,38^ Importantly, our results suggested that PEF-induced intracellular ROS accumulation in residual liver cancer cells increases their susceptibility to NK cell killing, aligning with previous reports that ROS are essential for NK cell-mediated tumor clearance. (Fig. 3). This suggests that IRE not only primes the TME for NK cell recruitment but also modulates tumor-intrinsic vulnerabilities. Given that NK cell-mediated cancer cell killing requires the accumulation of ROS in cancer cells^34,35^, these results suggest that not only ablation therapy but also chemotherapeutic agents known to increase ROS, such as sorafenib or regorafenib, may synergize similarly, offering additional avenues for combination strategies.^53,54^

In parallel, generating CAR-NK cells using DLNPs represents a major advancement in cellular immunotherapy.^40^ DLNPs achieved high transfection efficiency in NK cells and enabled the rapid generation of anti-GPC3 CAR-NK cells, supporting potential as scalable, off-the-shelf therapeutics (Fig. 4f and Supplementary Fig. 8). However, the transient expression of mRNA-based engineering remains a key limitation, as the expression generally diminishes within a short time frame. To achieve more durable antitumor activity, further strategies may incorporate platforms that enable prolonged or stable CAR expression, such as circular RNA engineering, Sleeping Beauty-mediated transposition, or CRISPR-based genomic knock-in approaches.^55–57^ Integrating these technologies with the nonviral DLNP system could further enhance the therapeutic persistence and clinical applicability of CAR-NK cell therapy.

Beyond the ex vivo-engineered CAR-NK strategy demonstrated here, integrating IRE with emerging in situ CAR-NK generation approaches represents a promising but still technically challenging direction.^58,59^ IRE induces ICD, enhances NK cell infiltration, and remodels the TME in ways that could theoretically amplify the efficiency of NK-targeted genetic delivery. Thus, combining IRE with future nanoplatforms capable of in situ CAR induction may enable a more minimally invasive and tumor-responsive therapy. However, in situ engineering of NK cells requires highly sophisticated nanotechnology systems with organ- and cell-type specificity, and NK cells remain inherently resistant to gene transfer.^59^ Importantly, combining IRE with cell therapy inevitably increases the procedural burden for patients, as IRE involves anesthesia and hardware-dependent tumor ablation, while CAR-NK infusion requires additional clinical logistics. Therefore, while our current ex vivo allogeneic CAR-NK approach offers practical translational advantages and a manageable clinical workflow, ongoing advances in NK-targeted nanotechnology could help reduce procedural complexity and pave the way toward a more procedure-consolidated in situ CAR-NK with IRE strategy.

Our findings further revealed that IRE combined with CAR-NK therapy induced more pronounced tumor regression in vivo than either treatment alone. This enhanced efficacy is likely due to the synergistic interaction between IRE-mediated immunomodulation and CAR-NK-mediated targeted cytotoxicity. Additionally, increased IFN-γ and TNF-α levels in combination-treated tumors suggest activation of adaptive immunity, supported by an increase in tumor infiltration of DCs, CD4⁺, and CD8⁺ T cells (Fig. 6). These results demonstrate that IRE with CAR-NK therapy effectively bridges innate and adaptive immune responses, generating a more comprehensive and durable antitumor effect.

From a translational perspective, several considerations warrant further investigation. Because IRE is already applied clinically for unresectable HCC, integrating CAR-NK infusion into existing procedures is highly feasible. Nonetheless, determining the optimal timing between IRE and NK cell administration is critical, given the transient chemokine alteration observed in our study. Additionally, mRNA-engineered CAR-NK cells may require repeated dosing or enhanced persistence strategies for sustained clinical benefit. Safety remains an important concern: although both IRE and NK cell therapies exhibit favorable profiles independently, combined therapy necessitates clinical evaluation to assess potential cytokine release, off-tumor cytotoxicity, and liver-related adverse effects. Ultimately, systematic optimization of dosage, scheduling, patient selection (e.g., GPC3 expression), and IRE parameters will be essential for successful translation into clinical practice.

Collectively, our results demonstrate that combining IRE with CAR-NK cell therapy elicits synergistic antitumor effects by enhancing NK cell recruitment, tumor cell sensitivity, and adaptive immune activation. Because IRE, LNP-based nanotechnologies, and NK cell-based therapies are clinically accessible, our study provides a mechanistic justification and a strong preclinical foundation for translating the combination therapy of IRE with CAR-NK cells into future HCC treatment strategies.

Methods

Cell culture

The liver cancer cell lines Huh-7 and Hepa1c1c7 and the NK cell line NK92MI were purchased from the Korean Cell Line Bank (Republic of Korea) and American Type Culture Collection (USA). Huh-7 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (FBS; HyClone) and 1% penicillin‒streptomycin (P/S; Gibco, Grand Island, NY, USA). Hepa1c1c7 cells were cultured in minimum essential medium alpha (MEM-α; Gibco) supplemented with 10% FBS and 1% P/S. NK92MI cells were cultured in MEM-α supplemented with 2 mM L-glutamine (Gibco), 1% P/S, 0.1 mM 2-mercaptoethanol (Gibco), 0.02 mM folic acid (Sigma‒Aldrich, St. Louis, MO, USA), 0.2 mM inositol (Sigma‒Aldrich), and 12.5% FBS. All cells were incubated at 37 °C in a 5% CO_2_ atmosphere.

Following the manufacturer’s protocol, mouse primary NK cells (mNK) were extracted from mouse spleens using a commercially available mouse NK cell isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany). The isolated mNK cells were then cultured in RPMI medium (HyClone) containing 10% FBS, 1% antibiotic/antimycotic solution (Gibco), 2 mM L-glutamine, 1% nonessential amino acids (Gibco), and 50 µM 2-mercaptoethanol. Mouse interleukin-2 (IL-2; PeproTech, Cranbury, NJ, USA) was also added to the culture.

In vitro PEF exposure and DAMP expression analysis

To assess the effects of PEF on liver cancer cells, 2 × 10^5^ Huh-7 and Hepa1c1c7 cells were seeded into 12-well plates and treated using an ECM 830 electroporator (BTX, Holliston, MA, USA). After 12 h, the cells were washed with phosphate-buffered saline (PBS) and treated with a bipolar PEF setting in serum-free medium (voltage: 500 V, pulse duration: 100 µs, electrode spacing: 5 mm). After 2 h, the medium was collected to analyze the amount of ATP released. Analysis of the released ATP was performed using an ATP bioluminescence kit (Sigma‒Aldrich). HMGB1 release was assessed by collecting CM after 24 h of PEF treatment and performing an enzyme-linked immunosorbent assay (ELISA) according to the manufacturer’s protocol (ABclonal Biotechnology, Woburn, MA, USA). To confirm the ablation zone caused by PEF, PEF-treated cancer cells were washed with PBS and stained with calcein-AM (Invitrogen, Waltham, MA, USA). After fixation with 4% paraformaldehyde (PFA), the area of living cells was observed using the area scan mode of a microplate reader (Synergy H1; BioTek, Winooski, VT, USA).

NK cell migration assay

The migration activity of NK cells was analyzed using a 24-well insert 8.0 μm Transwell chamber (SPL, Pocheon, Republic of Korea). Complete medium and 75% Con-CM or PEF-CM were added to the bottom chamber. A total of 1 × 10^5^ NK cells suspended in serum-free medium and stained with 1 μM CellTrace CFSE (Invitrogen) were loaded into the upper transwell insert. After 16 h, the CFSE-labeled NK cells in the bottom chamber were imaged using a fluorescence microscope (TM-30iF; TAESHIN BIO, Namyangju, Republic of Korea) and quantified using a Luna Cell Counter (Logos, Anyang, Republic of Korea). Anti-CX3CL1 antibody (5 μg/mL, R&D Systems, Minneapolis, MN, USA) was added to PEF-CM to neutralize CX3CL1 released by PEF treatment.

NK cell-mediated cytotoxicity assay

To analyze NK-mediated cytotoxicity by flow cytometry, NK cells were first stained with CellTrace CFSE or Far-red dye (Invitrogen) and then cocultured with target cancer cells for 4 h at effector-to-target (E/T) ratios of 10:1, 5:1, 2.5:1, and 1.25:1. Total cells were subsequently stained with 7-aminoactinomycin D (7AAD; Thermo Fisher Scientific, Waltham, MA, USA), and cell lysis was determined using a CytoFLEX flow cytometer (Beckman Coulter, Brea, CA, USA) and analyzed using FlowJo Software (BD Bioscience, Franklin Lakes, NJ, USA).

Coculture imaging of cancer cells with NK cells

To evaluate the targeting capability of NK cells in vitro, Huh-7 cells were stained with CellTrace CFSE, and 1 × 10³ Huh-7 cells per well were seeded into a 96-well confocal dish (SPL). After 24 h, NK cells were stained with CellTrace Far-Red and added to wells containing CFSE-labeled cancer cells at an E/T ratio of 2.5. Following a 4 h incubation, 4% PFA was added to each well for fixation. Images of NK cells bound to cancer cells were captured using a confocal laser scanning microscope (Carl Zeiss, Oberkochen, Germany) and analyzed using ZEN software (Carl Zeiss). The number of NK cells bound to cancer cells was counted at 10 cancer cells per well across a total of three wells. To detect the activation of apoptosis in cancer cells by NK cells, far-red-stained NK cells were cocultured with cancer cells. Cocultured samples were stained with CellEvent^TM^ Caspase-3/7 reagent (Invitrogen) according to the manufacturer’s protocol. Fluorescence images of stained samples were acquired using a confocal laser scanning microscope.

Immunoblotting

Cells were lysed with PRO-PREP lysis buffer (iNtRON Biotechnology, Seongnam, Republic of Korea) containing phosphatase inhibitors (Roche Applied Science, Penzberg, Germany) and then heated for 10 min at 95 °C. Cell lysates were resolved on sodium dodecyl sulfate‒polyacrylamide gels, and the separated proteins were transferred to polyvinylidene difluoride membranes (Bio-Rad, Hercules, CA, USA). The membranes were blotted with primary antibodies against Caspase-3, Cleaved Caspase-3, and GAPDH (Cell Signaling); CD3ζ (Abcam, Cambridge, UK); and β-actin and HMGB1 (ABclonal) at a 1:1000 dilution. After washing, the blots were incubated with a horseradish peroxidase (HRP)-conjugated secondary antibody at RT for 1 h (1:1000; Abcam). Immunoreactivity was detected using an enhanced chemiluminescence solution (Thermo Fisher Scientific).

Chemokine and cytokine analysis

To screen for alterations in chemokine expression induced by IRE treatment in HCC tumors, tumors were mechanically disrupted and lysed with PRO-PREP lysis buffer. The lysates were centrifuged at 13,000 rpm for 30 min, and the resulting supernatants were pooled from at least three independent biological samples. The pooled samples were then processed for chemokine array according to the manufacturer’s protocol (R&D Systems). To determine the levels of human CX3CL1 and mouse CX3CL1 in PEF-CM from Huh-7 and Hepa1c1c7 cells, ELISA was performed using reagents from ABclonal Biotechnology. Huh-7 and Hepa1c1c7 cells (2 × 10^5^) were seeded in 12-well plates. After 12 h, the cells were washed with PBS and treated with bipolar PEF pulses in serum-free medium. After 24 h, the CM was harvested and processed for ELISA according to the manufacturer’s protocol. To evaluate the local concentrations of IFN-γ and TNF-α within HCC tumors, ELISA was performed using reagents from R&D Systems. Tumors were mechanically disrupted and lysed with PRO-PREP lysis buffer, and the tissue lysates were centrifuged at 13,000 rpm for 30 min. The supernatants were then processed for ELISA according to the manufacturer’s protocol.

Imaging of ROS

To detect ROS levels by PEF, PEF-treated liver cancer cells were incubated with 20 μM 2′,7′-dichlorofluorescin diacetate solution (DCF-DA; Sigma‒Aldrich) for 30 min at 37 °C and washed twice with PBS. The fluorescence intensity was imaged under a fluorescence microscope TM-30iF and measured using ImageJ software.

Synthesis of DLNPs

Lipid components, including D-Lin-MC3-DMA (MC3; MedChemExpress, Monmouth Junction, NJ, USA), 1,2-dioleoyl-3-trimetylammonium-propane (DOTAP; Avanti Polar Lipids, Alabaster, AL, USA), 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC; Avanti), cholesterol (Sigma‒Aldrich), and 1,2-dimyristoyl-rac-glycero-3-methoxypolyethlene glycerol-2000 (PEG-DMG; Avanti), were individually solubilized in ethanol to generate stock solutions at concentrations of 25 mM for DOTAP, MC3, DSPC, and cholesterol and 12.5 mM for PEG-DMG. The mRNA encoding enhanced green fluorescent protein (EGFP) was purchased from TriLink Biotechnology (San Diego, CA, USA). To generate anti-hGPC3 CAR mRNA and anti-mGpc3 CAR mRNA, we used the mMESSAGE mMACHINE^TM^ T7 Ultra Kit (Thermo Fisher Scientific). The anti-hGPC3 CAR and anti-mGpc3 CAR constructs were designed as shown in Supplementary Figs. 10 and 11. The lipid solutions were equilibrated at 37 °C for 10 min before mixing. The lipids were mixed at a molar ratio of 30:35:7:26.95:1.05 (DOTAP: MC3: DSPC: Cholesterol: PEG-DMG). This lipid formulation was then combined with mRNA in 25 mM sodium acetate buffer (pH 5.2), maintaining a volume ratio of 1:3 and a weight ratio of 1:20 (mRNA to lipids). The synthesized DLNPs were dialyzed in PBS for 1 h to eliminate residual ethanol and equilibrate the pH to neutral before application.

Characterization of DLNPs

DLNPs were characterized using dynamic light scattering (DLS; Malvern Nano ZS, Malvern Instruments Ltd., Worcestershire, UK) to ascertain their diameter and surface zeta potential. Morphological assessments were conducted using cryo-electron microscopy (Cryo-EM; Glacios, Thermo Fisher Scientific). The encapsulation efficiency of mRNA within the DLNPs was quantified using a Quanti-it^TM^ RiboGreen RNA assay kit (Thermo Fisher Scientific) according to the manufacturer’s protocol. This procedure involved lysing the DLNPs with 0.5% Triton X-100, staining the released mRNA with Quant-it RiboGreen reagent, and quantifying the fluorescence using a Synergy H1 microplate reader. The encapsulation efficiency (%) was calculated using the following formula:

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Encapsulation}\,{efficeincy}\,\left( \% \right)=\left(1-\frac{free\,mRNA\,concentration}{total\,mRNA\,concentration}\right)\times 100$$\end{document}

Generation and expansion of HCC patient-derived organoids (HCC-PDOs)

Patient-derived HCC cells were isolated using a digestion method as described in Broutier et al.^60,61^ (SKKU2024-04-003). Briefly, the tumor tissue (1 ~ 3 mm tissue) was minced and washed using basal media (Advanced DMEM/F-12 (Gibco) with 1% P/S, 1% GlutaMAX (Gibco), and 1% HEPES (Gibco)). Minced tissue was added to prewarmed digestion solution, basal media supplemented with B-27 without vitamin A (Gibco), 1.25 mM NAC (Sigma‒Aldrich), 5% RSPO1-CM (Homemade), 2.5 mg/mL collagenase D (Roche), and 0.1 mg/mL DNase I (Sigma‒Aldrich). The minced tissues in the digestion solution were incubated using a shaking incubator at 37 °C for 30 min to 2 h (confirming the tissue conditions in the digestion solution every 30 min). After incubation, digested tissues were filtered through a 100 µm nylon cell strainer (SPL) with additional cold-basal media and spun down at 200 x g for 5 min at 4 °C, and the supernatant was discarded. If red blood cells remained in the pellet, the pellet was incubated in 3 mL of ACK buffer (Gibco) for 3 min at room temperature (RT). After washing, the cells were resuspended in 70% Matrigel (Corning, diluted basal media) and seeded in a prewarmed plate (seeding density: 1 × 10^5^ cells per well in a 24-well plate). After incubation for 30–60 min, prewarmed media was added, and the media was replaced every 2–3 days. The organoid media was based on basal media supplemented with B-27 without vitamin A, 1.25 mM NAC, 10 nM gastrin (Sigma‒Aldrich), 50 ng/mL human EGF (R&D Systems), 15% RSPO1-CM, 100 ng/mL human FGF10 (Peprotech), 50 ng/mL human HGF (Peprotech), 3 µM CHIR99021 (TOCRIS, Bristol, UK), 100 ng/mL human FGF7 (Peprotech), 2 µM A83-01 (TOCRIS), 0.5 nM Wnt Surrogate FC Fusion protein (IPA), 10 µM Y-27632 (TOCRIS), and 10 µM TRULI (TOCRIS).^62^ The grown HCC-PDOs were passaged at a 1:2 ratio once every 1–2 weeks as described in previous studies.^60,62^

Immunofluorescence of HCC-PDOs

Whole-mount organoid staining was conducted using a method described in Dekkers et al.^63^ Briefly, HCC-PDOs were harvested from Matrigel using cold-basal media or cold-cell recovery solution (Corning) and fixed in 4% PFA on ice for 30 min. After incubation, the fixed HCC-PDOs were washed using 1X DPBS (Gibco) and transferred to a confocal plate. Fixed HCC-PDOs were blocked with blocking solution (1X DPBS supplemented with 0.1% Triton X-100 (Sigma‒Aldrich) and 0.2% bovine serum albumin (BSA; Sigma‒Aldrich)) at 4 °C for 1 h. After blocking, HCC-PDOs were incubated with primary antibodies (GPC3, Invitrogen, 1:200) at 4 °C overnight in blocking solution. After incubation, the samples were washed using blocking solution 3 times and incubated with secondary antibodies (anti-mouse Alexa 488, Invitrogen, 1:1000) and DAPI (Invitrogen, 1:1000) overnight in blocking solution at 4 °C. Stained HCC-PDOs were washed 3 times and stored at 4 °C in DPBS before imaging. The samples were confirmed using a fluorescence microscope, DM IL LED (Leica, Wetzlar, Germany).

Flow cytometry

Cells were harvested as single-cell suspensions. After washing with PBS, single cells were stained with Zombie (BioLegend, San Diego, CA, USA) for live/dead staining and blocked with FcBlocker (BD Bioscience) in FACS buffer (0.09% sodium azide and 2% FBS in PBS). Single cells were then incubated with a fluorophore-conjugated antibody (CD45, CD11b, CD3, NK1.1, CD4, CD8, Ly6G, CD11c, MHCII, CD80, CD86; BioLegend) in FACS buffer at RT for 2 h. After staining, the fluorescently labeled cells were washed twice with FACS buffer and fixed in 2% PFA. Fluorescently labeled cells were acquired on a MA900 (NFEC-2024-10-300262; Sony Biotechnology Inc., San Jose, CA, USA) flow cytometer. The gating strategy is shown in Supplementary Figs. 1 and 16.

Animal study

All in vivo experiments were conducted according to the guidelines of the protocol approved by the Sungkyunkwan University Institutional Animal Care and Use Committee (SKKUIACUC2023-04-41-1; Republic of Korea). The mice were kept in a controlled environment with a regulated temperature and 12 h light-dark cycle. Seven-week-old C57BL/6 and nude mice were obtained from Orient Bio (Republic of Korea). Seven-week-old NOD-Prkdc^em1Baek^Il2rg^em1Baek^ (NSG) mice were obtained from JA Bio (Republic of Korea). Before the experimental procedures, the mice were acclimated for one week. To establish an orthotopic model of HCC, thioacetamide (TAA; 50 mg/kg, Sigma‒Aldrich) was intraperitoneally injected into the mice three times for 10 days to induce liver fibrosis. Subsequently, 2 × 10^7^ Hepa1c1c7 cells/mL were prepared by mixing equal volumes of cells and Matrigel (Corning, Corning, NY, USA). Twenty microliters of this mixture was then inoculated into the livers of C57BL/6 or nude mice. After 10 days, IRE was directly administered to the tumor region of the liver (voltage: 500 V, pulse duration: 100 µs, electrode spacing: 1 mm), and then control mNK or CAR-mNK cells (2 × 10^6^ cells/mouse) were injected intravenously.

Huh-7 cells (5 × 10^6^) were mixed with Matrigel (DPBS/Matrigel ratio = 1) and then inoculated subcutaneously into the right flank of NSG or nude mice. When the diameter of the subcutaneous tumors reached 7 mm, IRE was applied to the tumor (voltage: 500 V, pulse duration: 100 µs, electrode spacing: 5 mm). After IRE treatment, 1 × 10^7^ Con-NK92MI or CAR-NK92MI cells were intravenously injected into the tail vein. Animals were euthanized when tumors reached 1500 mm³ of calculated tumor volume.

Biochemical assessments and histological analysis

For histological and biochemical assessments, mice were euthanized at specific time points, after which organs and blood samples were collected. Blood samples were centrifuged at 4000 rpm for 20 min to separate serum. Key biochemical parameters, such as AST, ALT, TBIL, and LDH levels, were quantified using a Fujifilm DRI-CHEM NX500i automated chemistry analyzer (Tokyo, Japan). For histopathological examination, major organs were quickly excised, fixed in 4% PFA, and processed. Tissue samples were dehydrated, infiltrated, and embedded in Paraplast. Paraffin sections were cut using a microtome, stained with hematoxylin and eosin (H&E), and analyzed using a ScanScope CS2 system (Leica Biosystems, San Diego, CA, USA).

Immunofluorescence tissue imaging analysis

Paraffin sections were deparaffinized, rehydrated, and heat-treated in antigen retrieval buffer. After several PBS washes, sections were blocked with 5% BSA in PBS, incubated for 30 min, and incubated overnight with primary antibody (NKp46, ABclonal, 1:50) at 4 °C. After washing with PBS, the sections were incubated with HRP-conjugated secondary antibody at RT for 20 min (1:1000). After washing with PBS, signal amplification was conducted by TSA Multiplex Immunohistochemistry kits (TissueGnostics, Vienna, Austria). Immunofluorescence images were obtained using TissueFAXs i8 plus (TissueGnostics).

Antigen-specific CD8+ T-cell analysis

Single-cell suspensions of spleen from each mouse (5 × 10⁵ cells per well) were seeded in v-bottom 96-well plates and restimulated for 4 h with Hepa1c1c7 cell lysate (10 μg/ml). A cell stimulation cocktail containing protein transport inhibitors (Invitrogen) was added to the cells. For flow cytometry analysis, cells were fixed and permeabilized using the intracellular fixation & permeabilization buffer set (Invitrogen) according to the manufacturer’s protocol. Permeabilized cells were blocked using FcBlocker and stained with fluorescent-conjugated antibodies (CD45, CD3, CD11b, CD8, and IFN-γ) at RT for 2 h. Fluorescently labeled cells were acquired on a MA900 flow cytometer. The gating strategy is shown in Supplementary Fig. 19.

Statistical analysis

Statistical analyses were performed using GraphPad Prism V.10.5.0 software (GraphPad Software, Boston, MA, USA). Statistical significance was determined using an unpaired Student’s t test, two-way ANOVA, and log-rank test. Data are presented as the mean ± S.D. from at least three independent experiments, with P < 0.05 considered statistically significant.

Supplementary information

Supplementary Materials Western blot uncropped