Oncolytic adenovirus-mediated delivery of anti-CLDN18.2 IgG3 for the treatment of solid tumors

Abstract

Background

Claudin 18.2 (CLDN18.2) is a promising therapeutic target for gastric, pancreatic and ovarian cancers. The Food and Drug Administration-approved antibody, zolbetuximab (IgG1), induces tumor cell death via antibody-dependent cell-mediated cytotoxicity (ADCC), antibody-dependent cellular phagocytosis (ADCP) and complement-dependent cytotoxicity (CDC). While the IgG3 subclass offers superior ADCC, ADCP and CDC effector functions compared with IgG1, its clinical application has been limited by a short plasma half-life due to reduced binding to the neonatal Fc receptor.

Methods

To overcome this limitation, we engineered a conditionally replicating oncolytic adenovirus (OAd-aCLDN18.2) to deliver and locally express an anti-CLDN18.2 IgG3 antibody specifically within the tumor microenvironment. This strategy leverages the tumor-specific replication of the oncolytic virus (OVs) to achieve high local concentrations of the potent IgG3 antibody, minimizing systemic exposure and its associated half-life liability.

Results

OAd-aCLDN18.2 effectively delivered and expressed the functional anti-CLDN18.2 IgG3 antibody. This localized antibody secretion induced robust ADCC, ADCP and CDC effects against CLDN18.2-positive tumor cells. In vitro, the virus-delivered antibody potently mediated natural killer (NK) cell activation and cytotoxicity. In murine tumor models, treatment with OAd-aCLDN18.2 resulted in significantly superior tumor control compared with systemic administration of zolbetuximab (IgG1), without notable toxicity or changes in body weight. The enhanced efficacy was attributed to the localized, intratumoral production of the highly active IgG3 antibody, combined with the OV’s ability to remodel the tumor microenvironment and potentiate a stronger local NK cell response. Viral replication was largely confined to the tumor, indicating a favorable safety profile.

Conclusions

This study presents the first report of an OV armed with an anti-CLDN18.2 IgG3 antibody. By combining the potent effector functions of the IgG3 subclass with the localized delivery capabilities of an OV, OAd-aCLDN18.2 overcomes the traditional pharmacokinetic drawbacks of IgG3. This approach elicits robust immune-mediated antitumor responses and demonstrates superior efficacy over the current standard-of-care IgG1 antibody in preclinical models. Our findings support OAd-aCLDN18.2 as a novel and promising therapeutic strategy for CLDN18.2-positive gastric, pancreatic, and ovarian cancers.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• IgG3 antibodies mediate stronger antibody-dependent cell-mediated cytotoxicity (ADCC), antibody-dependent cellular phagocytosis (ADCP) and complement-dependent cytotoxicity (CDC) than IgG1 but have limited clinical utility due to their short plasma half-life caused by poor neonatal Fc receptor binding. Oncolytic viruses (OVs) can selectively replicate in tumors and serve as vectors for local delivery of therapeutic antibodies.

WHAT THIS STUDY ADDS

• This study demonstrates that an OV locally delivered anti-claudin 18.2 (CLDN18.2) IgG3 antibody achieves intratumoral concentrations, overcoming the half-life limitation of IgG3. The virus-armed IgG3 shows superior antitumor efficacy compared with zolbetuximab in preclinical models through enhanced ADCC, ADCP and CDC activity.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This work provides a clinically viable strategy to harness the potent effector functions of IgG3 antibodies by using OVs for intratumoral delivery. This approach may guide future development of antibody-armed OVs for CLDN18.2-positive tumors and other solid malignancies.

Background

Claudin 18.2 (CLDN18.2), a tight-junction-associated protein, is involved in the formation of tight junctions between cells and affects the permeability of paracellular cells.¹ CLDN18.2 is highly expressed in gastric, pancreatic and ovarian cancers and is associated with the initiation, progression, metastasis and prognosis of cancer.^{1 2} Given its restricted expression in normal tissues and frequent overexpression in tumors, CLDN18.2 has been recently recognized as a potential therapeutic target for gastric, pancreatic and ovarian cancers.^1–6 Currently, multiple clinical trials are underway to evaluate various CLDN18.2-targeted agents.^{1 6 7} Zolbetuximab is a human chimeric IgG1 monoclonal antibody targeting CLDN18.2 and is approved by the Food and Drug Administration (FDA) in 2024.^{3 8} It can bind to the extracellular domain of CLDN18.2 and induce tumor cell death through antibody-dependent cell-mediated cytotoxicity (ADCC), antibody-dependent cellular phagocytosis (ADCP) and complement-dependent cytotoxicity (CDC).^{3 8} At present, it has achieved certain efficacy in the clinical treatment of patients with gastroesophageal cancer.^{3 9 10}

Compared with the widely used IgG1, IgG3 antibodies have a longer hinge region, higher sensitivity to bind to Fc-gamma receptors (FcγRs) on immune cells, and can induce stronger ADCC and CDC effects than IgG1.^11–13 However, the R435 of IgG3 is different from the H435 of other IgG molecules, which affects its binding to the neonatal Fc receptor (FcRn), resulting in a shorter metabolic half-life in plasma, so it has not been clinically applied.^{11 12}

Oncolytic viruses (OVs) are tumor-specific virotherapy agents that replicate selectively in cancer cells while sparing normal tissues.^{14 15} Beyond their direct oncolytic effect, OV serves as versatile delivery platforms for therapeutic antibodies, enabling local production within the tumor.^16–19 This strategy circumvents the poor penetration of systemically administered antibodies into solid tumors and minimizes off-target toxicity, as viral replication—and thus antibody expression—is largely restricted to malignant lesions.¹⁵ Importantly, OV infection remodels the immunosuppressive tumor microenvironment by recruiting immune effector cells, reducing immunosuppressive populations, and decreasing inhibitory cytokines,^{20 21} thereby creating a favorable setting for antibody-mediated immunity. Collectively, these features maximize local antibody concentrations at the tumor site while reducing systemic exposure, offering a synergistic approach to enhance therapeutic efficacy.

In this study, we engineered an oncolytic adenovirus (OAd) that secretes a CLDN18.2-targeting IgG3 antibody on replication in cancer cells. We demonstrate that OAd-aCLDN18.2 delivers the CLDN18.2 antibody, which induces ADCC, ADCP and CDC. Compared with zolbetuximab, OAd-aCLDN18.2 elicits stronger immune responses and superior antitumor efficacy.

Methods

Cell culture

Human embryonic kidney (HEK) 293 cells and the human gastric adenocarcinoma cell line SGC-7901 were obtained from the Chinese Academy of Sciences, Shanghai Institutes for Biological Sciences (Shanghai, China). The poorly differentiated human mucinous adenocarcinoma cell line MGC-803 was purchased from Shanghai Fuheng Biotechnology (Shanghai, China). The human ovarian cancer cell lines OVCAR3 and human pancreatic cell line Panc-01 were purchased from Wuhan Procell Life Science and Technology (Shanghai, China). MKN45 and HeLa were purchased from BOSTER Biological Technology (Wuhan, China). These cells were maintained in Dulbecco’s Modified Eagle Medium (#12491015; Gibco) or Roswell Park Memorial Institute 1640 (#12633020; Gibco) supplemented with 10% fetal bovine serum (DT500S; DearyTech), and 1% penicillin-streptomycin in an incubator at 37°C with 5% CO₂.

Plasmid construction and virus preparation

Adenovirus type five plasmids were purchased from the MiaoLing Plasmid Platform. The OAd was constructed as we previously described.²² The viral E1A gene is under the control of the human telomerase reverse transcriptase (hTERT) promoter, which enables OAd to replicate selectively in tumor cells and induce tumor-specific oncolytic cell death.

The antibody sequence carried is selected to be connected to the light and heavy chain sequences of the anti-CLDN18.2 antibody using the T2A linker after the E1A sequence. The sequences of the variable domain of anti-CLDN18.2 antibody were based on an anti-human CLDN18.2 monoclonal antibody, zolbetuximab. To better promote the secretion of the antibody, secretion signal peptides are added to the front ends of the heavy and light chains, respectively. The amino acid sequence of the antibody was obtained based on the sequence of zolbetuximab, and amino acid sequences for secretion signal peptides were added at the front ends of the heavy and light chains, respectively. It was synthesized by GENEZIE (Suzhou, China).

Homologous recombination of OAd-aCLDN18.2 was performed with AdMax system. Ad genomic and shuttle plasmids were transfected into HEK293 cells to produce OAd-aCLDN18.2. Once an extensive cytopathic effect was observed, the virus was harvested from HEK293 cells using three freeze–thaw cycles. Viruses were purified via cesium chloride density gradient centrifugation and concentrated via dialysis bag activation using Dialysis Cassette (#66380; Thermo).

Supernatants containing OAd or OAd-aCLDN18.2 were generated by infecting HEK293 cells at a multiplicity of infection (MOI) of 0.5 for 48 hours and titrated with threefold serial dilutions ranging from 1:10×3¹ to 1:10×3¹² in 96-well plates for further assays.

Cell transfection

The plasmid was mixed with polyethylenimine of the same amount in Opti-MEM (#12492013, Gibco) for 20 min and then the mixture was added to the cell supernatant. The cells were collected 48 hours after transfection for total protein extraction.

Viral oncolysis and replication assay

For the viral oncolysis assay, MGC-803, SGC-7901, MKN45, OVCAR3 or Panc-01 cells were seeded into a 96-well-plate at a density of 2×10⁴ cells per well, and infected with serial MOI of OAd or OAd-aCLDN18.2. 4 days after infection, cell viability was examined with Cell Counting Kit-8 (#C6005; NCM Biotech) and the half-maximal inhibitory concentration (IC50) was calculated. For the viral replication assay, MKN45, SGC-7901 or OVCAR3 cells were seeded into a 12-well-plate at a density of 3×10⁵ cells per well. Cells were infected with OAd or OAd-aCLDN18.2 and then harvested at different time points (MOI=0.5 for MGC-803 cells; MOI=0.1 for SGC-7901 cells; MOI=1 for MKN45, OVCAR3 and Panc-01 cells). Total DNA was extracted using TIANamp Genomic DNA Kit (#DP304; TIANGEN). 20 ng total DNA was detected by quantitative PCR against the Ad5 E1A gene with specific primers (primer-F-5′-TTGGGTCCGGTTTCTATGCC-3′ and primer-R-5′-CACCCTCTTCATCCTCGTCG-3′; GENEWIZ). Fold change relative to 0 hours was calculated.

Western blot analysis

Total protein was isolated from cells or animal tissues using a radio immunoprecipitation assay buffer (#P0045; Beyotime Institute of Biotechnology). Protein was quantified using the Wanleibio BCA Protein Assay Kit (#WLA004; Wanleibio). Equal amounts of protein were separated via sodium dodecyl sulfate-polyacrylamide gel electrophoresis and electrophoretically transferred onto polyvinylidene fluoride membranes. To avoid non-specific binding, the membranes were blocked with 5% non-fat milk for 1 hour at room temperature, then incubated with anti-CLDN18 antibody (1:1,000; YM9290; Immunoway), anti-IgG3 antibody (1:1,000; ab1928; Abcam), antitumor necrosis factor (TNF)-α antibody (#YM8306; Immunoway), anti-interferon (IFN)-γ antibody (#YT2279; Immunoway), anti-IFN-γ antibody (#YT2279; Immunoway) or anti-granzyme B (GZB) antibody (#LS-C174385; Thermo) at 4°C overnight. After washing, the membranes were incubated with horseradish peroxidase (HRP)-conjugated Affinipure Goat Anti-Rabbit IgG (H+L) (1:5,000; #SA00001-2; Proteintech) for 1 hour at room temperature. The membrane was washed and then imaged in a pre-cooled chemiluminescence instrument.

Isolation of peripheral blood mononuclear cells

Peripheral blood mononuclear cells (PBMCs) were isolated from peripheral blood of healthy adults with PBMC Isolation Solution Kit (P8680; Solarbio). PBMCs isolation was performed according to the manufacturer’s instructions.

ELISA analysis

The antigen-binding activity of anti-CLDN18.2 antibodies and the expression levels of IgG3 were evaluated using a sandwich ELISA. Briefly, 96-well MaxiSorp plates (#0811K52420; Solarbio) were coated with 200 ng per well of recombinant human CLDN18.2 protein (#HY-P70701; MCE) or anti-human IgG3 capture antibody (#MA5-49520; Thermo Fisher) overnight at 4°C. A standard curve was constructed with serially diluted human IgG3 standard protein (#SAB5600212; Sigma-Aldrich). Supernatants from infected cells were added as test samples. Following incubation, 100 µL of detection antibody (10 µg/mL) was applied for 45 min at room temperature with mild agitation. After extensive washing, colorimetric detection was performed using 3,3’,5,5’-tetramethylbenzidine (#PR1200; Solarbio) substrate (5–30 min in the dark), and the reaction was stopped with stop solution. The absorbance was recorded at 450 nm using a microplate reader (Tecan Life Sciences; Männedorf; Switzerland), and sample IgG3 concentrations were calculated against the standard curve.

For cytokine secretion assay, tumor tissues were weighed and homogenized in ice-cold radioimmunoprecipitation assay (RIPA) lysis buffer (supplemented with protease inhibitors) (#R0020; Solarbio) at a ratio of 1:9 (w/v) using a tissue homogenizer. The homogenates were centrifuged at 10,000 g for 10 min at 4°C, and the supernatants were collected. Total protein concentration in the lysates was determined using a BCA protein assay kit. All lysates were then normalized to the same protein concentration for subsequent ELISA analysis with the mouse TNF-α (#KTE7015; Abbkine) or IFN-γ (#KTE7003; Abbkine) ELISA Kit.

ADCC assay

Primary human natural killer (NK) cells were stained with Pacific Blue-conjugated anti-human CD56 antibody (#362519; BioLegend) and isolated from PBMCs of healthy donors using a BD FACSMelody cell sorter (BD Biosciences). Purified CD56⁺ NK cells were used as effector cells in subsequent assays.

MGC-803, SGC-7901, OVCAR3, Panc-01 and MKN45 cells were labeled with 0.5 µmol/L 5-(and 6)-carboxyfluorescein diacetate succinimidyl ester (CFSE) and used as target cells. A total of 2×10⁴ target cells per well were seeded into 96-well plates. After 24 hours, 2×10⁵ PBMCs or isolated NK cells were added along with harvested viral supernatants at a dilution of 1:1,000 (or various dilutions). Following 48 hours of co-culture, images were acquired, and cells were then collected and stained with a Fixable Viability Kit to identify dead cells. The percentage of dead target cells and the expression level of CD69, an early activation marker of NK cells, were quantified by flow cytometry.

ADCP assay

Mouse bone marrow-derived macrophages (BMDMs) were isolated from the tibias and femurs of BALB/c mice by flushing the bone marrow cavities with culture medium using a 25 G needle. The resulting bone marrow cells were passed through a 70 µm nylon mesh (#352350; Falcon) and washed five times with phosphate buffered saline (PBS). Isolated BMDMs were seeded at a density of 1.5×10⁷ cells per 60 mm culture dish (#150462; Thermo Fisher) and cultured for 7 days in medium supplemented with murine macrophage colony-stimulating factor (#315-02; PeproTech), with half-medium changes on days 3 and 5. On day 7, cells were harvested by centrifugation at 400 g for 5 min at 4°C and stained with anti-mouse CD11b antibody (#101211; BioLegend) to identify macrophages.

For the phagocytosis assay, SGC-7901 cells were labeled with 0.5 µmol/L CFSE and co-cultured with BMDMs at an effector-to-target ratio of 1:2 (macrophages to tumor cells) in the presence of harvested viral supernatants at a dilution of 1:1,000 (or 1:1,500) for 4 hours at 37°C in a 5% CO₂ incubator. As a negative control, BMDMs were pretreated with cytochalasin D for 30 min. Phagocytic activity was quantified as the percentage of CD11b⁺CFSE⁺ macrophages among total CD11b⁺ macrophages, reflecting the proportion of macrophages that engulfed CFSE-labeled tumor cells.

CDC assay

MGC-803, SGC-7901, OVCAR3, MKN45 and Panc-01 cells were seeded into 96-well plates at a density of 2×10⁴ cells per well. After 24 hours, the culture medium was replaced with a mixture containing 25% normal human serum (NHS) from healthy donors and various dilutions of supernatants from OAd or OAd-aCLDN18.2-infected cells (at a dilution of 1:1,000 or various dilutions). Serum samples were heat-inactivated at 56°C for 30 min to eliminate endogenous complement activity and served as negative controls. The cells were then incubated for 4 hours at 37°C in a 5% CO₂ incubator. Subsequently, 10 µL of Cell Counting Kit-8 solution was added to each well, followed by incubation for an additional 40 min at 37°C in a 5% CO₂ incubator. Absorbance was measured at 450 nm using a microplate reader.

Animal experiments

Female BALB/c nude mice aged 4 weeks old were purchased from Charles River Laboratories (Beijing, China). All mice were maintained in a specific pathogen-free facility at China Medical University. Tumor cells (2×10⁶/mouse) in 0.1 mL of PBS were injected subcutaneously into the right inguinal region of nude mice. Tumor volume was calculated using the standard formula: tumor volume (mm³)=length×width²×0.5. Once the mean tumor volume reached approximately 50 mm³, tumor-bearing mice were randomly assigned to experimental groups (n=5 per group). PBS, OAd, and OAd-aCLDN18.2 were administered via intratumoral injection (1×10⁸ plaque forming unit (PFU) each), whereas zolbetuximab (10 mg/kg) was delivered via intraperitoneal administration. Each treatment was administered once every 3 days for a total of three times. Tumor growth and body weight were monitored by caliper measurement every 3 days. Tumors were harvested 23 or 30 days post-cell engraftment for subsequent analyses.

Immunohistochemical staining

Heart, liver, spleen, lung, kidney tumor and tissues were collected from BALB/c nude mice (obtained as described above) and immediately fixed in 4% paraformaldehyde. Following deparaffinization and rehydration, tissue sections were subjected to antigen retrieval by boiling in 0.01 M citrate buffer for 10 min. Endogenous peroxidase activity was blocked by incubating the sections with 3% hydrogen peroxide in methanol for 25 min. The sections were then sequentially incubated with primary monoclonal antibodies against GZB (1:3,000; #66481-1-Ig; Proteintech), IFN-γ (1:3,000; #YT2279; Immunoway), and TNF-α (1:3,000; #YM8306; Immunoway), followed by HRP-conjugated goat anti-mouse IgG (H+L) secondary antibody (1:200; #GB23301; Servicebio). Antibody binding was visualized using 0.05% 3,3-diaminobenzidine tetrachloride as the chromogenic substrate. Sections were counterstained with hematoxylin, and representative images were captured under an inverted phase-contrast microscope. The proportion of positively stained cells was quantified using ImageJ V.1.46r (National Institutes of Health).

Flow cytometry

The antibodies and fluorescent dyes were purchased from BioLegend as follows: Zombie NIR Fixable Viability Kit (#423105), FITC anti-mouse Ly-6C (#128005), PE anti-mouse CD45 (#147711), Pacific Blue anti-mouse CD45 (#103125), PE anti-human IgG Fc (#410708), Pacific Blue anti-mouse CD49b (#108918), PE/Cyanine7 anti-mouse Ly-6G (#127618), APC anti-mouse CD11b (#101212), FITC anti-mouse F4/80 (#123107), Pacific Blue anti-mouse CD86 (#105021), PE anti-mouse CD86 (#159203), APC anti-mouse CD206 (MMR) (#141707), APC anti-mouse CD69 (#104513), PE/Cyanine7 anti-mouse CD49b (#103517), Pacific Blue anti-human CD56 (#362519), Human TruStain FcX (#422301). CD16/CD32 (#MA5-29707) was purchased from Thermo Fisher.

To obtain single-cell suspensions, tumor tissues were minced into small pieces and incubated in lysis buffer containing 1 mg/mL collagenase IV (#abs47048003; Absin) and 10 U/mL DNase I (#M0303S; New England Biolabs) in Hank’s buffer (#abs9258; Absin) for 2 hours at 37°C. Both tumor and spleen tissues were then filtered through a 70 µm cell strainer (#352350; Falcon). For tumor tissues, fibrous debris was removed by centrifugation using a 30% Percoll gradient (#40501ES60; Yeasen). Red blood cells were lysed using 2 mL of red blood cell lysis buffer (#abs9241; Absin), and the remaining cells were washed and processed for subsequent flow cytometric staining.

For flow cytometry, cell suspensions were first stained with a Fixable Viability Kit (1:1,000) for 20 min to exclude dead cells, blocked with Fc block antibody for 15 min at 4°C prior to surface staining, followed by incubation with the appropriate anti-human or anti-mouse antibodies for 30 min at 4°C. Stained cells were acquired on a BD FACSymphony A1 flow cytometer (BD Biosciences), and data were analyzed using FlowJo software (TreeStar).

Cell populations in vivo study were identified as follows: NK: live/CD45⁺/CD49b⁺; polymorphonuclear myeloid-derived suppressor cells (PMN-MDSCs): live/CD45⁺/CD11b⁺/Ly6G⁺/Ly6C^Lo; monocytic MDSC: live/CD45⁺/CD11b⁺/Ly6G^Lo/Ly6C⁺; classically activated macrophages (M1 macrophages): live/CD45⁺/CD11b⁺/F4/80⁺/CD86⁺; alternatively activated macrophages (M2 macrophages): live/CD45⁺/CD11b⁺/F4/80⁺/CD206⁺.

Statistical analysis

Statistical analyses were performed using GraphPad Prism software (V.7; GraphPad Prism Software, La Jolla, California, USA). Differences were considered significant when p<0.05.

Results

Generation and characterization of anti-CLDN18.2-IgG3-armed oncolytic adenovirus (OAd-aCLDN18.2)

We generated an OAd encoding an anti-CLDN18.2 IgG3 antibody (OAd-aCLDN18.2). To construct OAd-aCLDN18.2, the anti-CLDN18.2 IgG3 expression cassette was inserted into the backbone of an OV, under the transcriptional control of an hTERT promoter, which drives the concurrent transcription of both the E1A gene and the anti-CLDN18.2 IgG3 transgene (figure 1A). The oncolytic viral potential (figure 1B) and replication capacity (figure 1C) of the recombinant OAd-aCLDN18.2 were comparable to those of unarmed OAd, and the IC50 was calculated (online supplemental figure S1A), indicating that insertion of the anti-CLDN18.2 transgene did not compromise the intrinsic oncolytic properties of the adenoviral backbone. High levels of soluble anti-CLDN18.2 IgG protein were produced and efficiently released from OAd-aCLDN18.2-infected tumor cells in vitro, in which heavy and light chains were detected by western blot (figure 1D, online supplemental figure S1B) and ELISA (figure 1E). To evaluate the binding of OV-delivered anti-CLDN18.2 to CLDN18.2, an ELISA was performed across a range of supernatant dilutions. The anti-CLDN18.2 antibodies exhibited concentration-dependent binding to CLDN18.2, confirming the integrity of the antigen-binding function (figure 1F). To evaluate the potential of CLDN18.2 as a specific target for clinical treatment, cell surface expression of CLDN18.2 of six CLDN18.2⁺ cell lines (MGC-803, SGC-7901, Panc-01, MKN45, OVCAR3, HeLa) was assessed by western blot and flow cytometry, and PBMCs were used as CLDN18.2⁻ control (figure 1G, online supplemental figure S1C,D).

Figure 1. Construction and validation of OAd-aCLDN18.2. (A) Schematic structure of OAd and OAd-aCLDN18.2. The recombinant oncolytic adenovirus OAd-aCLDN18.2 was constructed using the adenoviral ITR backbone, with the E1A and anti-CLDN18.2 sequence driven by the hTERT promoter. (B) Cytotoxic effects of OAd-aCLDN18.2. MGC-803, SGC-7901, MKN45, OVCAR3 and Panc-01 cells were infected with serial dilutions of OAd or OAd-aCLDN18.2. Cell viability was examined 4 days post-infection. (C) Viral replication of OAd-aCLDN18.2. MKN45, SGC-7901 and OVCAR3 cells were infected with OAd or OAd-aCLDN18.2 and harvested at the indicated time points. DNA was extracted, and qPCR was performed to detect E1A gene copy number. Fold change relative to 0 hours was calculated. MKN45, SGC-7901 and OVCAR3 cells were infected at an MOI of 1, 0.1 and 1. (D) Detection of anti-CLDN18.2 antibodies secreted by infected MGC-803 (left) and SGC-7901 (right). Cell culture supernatants were collected 48 hours post-infection and analyzed by western blot. (E) Kinetics of anti-CLDN18.2 antibody secretion. Supernatants collected from OAd-aCLDN18.2-infected MGC-803 and SGC-7901 cells at indicated time points were analyzed by ELISA. (F) Binding activity of secreted anti-CLDN18.2 antibody to human CLDN18.2. HEK293 cells were infected with OAd-aCLDN18.2 at an MOI of 0.5 for 48 hours, and the harvested supernatants were evaluated by ELISA at dilution ratios of 1:1, 1:5, and 1:10. (G) CLDN18.2 expression profiles in indicated cell lines. Expression of CLDN18.2 in PBMCs, MGC-803, SGC-7901, OVCAR3, Panc-01, MKN45, and HeLa cells was assessed by western blot. aCLDN18.2, anti-claudin 18.2; hTERT, human telomerase reverse transcriptase; ITR, inverted terminal repeat; MOI, multiplicity of infection; OAd, oncolytic adenovirus; PBMC, peripheral blood mononuclear cell; qPCR, quantitative PCR.

OAd-aCLDN18.2 mediates PBMC-mediated killing in vitro

ADCC mediated by NK cells is an Fc-dependent effector function of IgG that plays a critical role in antitumor therapy. This activity is primarily triggered by the IgG1 and IgG3 subclasses through binding to FcγRIIIa. To exclude the potential interference of viral particles on cytotoxicity activities, serially diluted supernatants from OAd or OAd-aCLDN18.2-infected cells were incubated with tumor cells, and the dilution factors without detectable cytotoxicity were selected for subsequent experiments (online supplemental figure S2A). To evaluate the effects of OV-delivered anti-CLDN18.2 IgG3 on NK cells, unstimulated human PBMCs, which contained 20.9% NK cells (online supplemental figure S2B), were used as effector cells and co-cultured with tumor target cells. We first added the supernatants from OAd-aCLDN18.2-infected cells to the co-cultures and assessed the ADCC effects mediated by OAd-aCLDN18.2 against various tumor cell lines (online supplemental figure S2C). Compared with OAd, OAd-aCLDN18.2 significantly enhanced the ADCC activity against SGC-7901, MGC-803, MKN45, Panc-01 and OVCAR3 (figure 2A–C). To further examine the impact of antibody concentration on the observed cytotoxicity, serial dilutions of the OAd or OAd-aCLDN18.2 infected supernatants were tested in the ADCC assay. The killing activity progressively declined with increasing dilution folds, demonstrating a dose-dependent response characteristic of the antibody-mediated effect (online supplemental figure S2D).

Figure 2. Effect of OAd-aCLDN18.2 on cancer cells in vitro. (A) Cytotoxicity assay. SGC-7901 and MGC-803 cells were co-cultured with PBMCs in the presence of supernatants from the OAd or OAd-aCLDN18.2-infected HEK293 cells (1:1,000). After 48 hours, cell viability was assessed using the CCK-8 assay. (B) PBMCs were cultured alone or co-cultured with the MGC-803 tumor cell line in the absence (Mock) or presence of supernatants of OAd or OAd-aCLDN18.2-infected cells for 48 hours. Representative brightfield images are shown. (C) Flow cytometry-based killing assay. MGC-803, SGC-7901, OVCAR3, Panc-01 or MKN45 cells were labeled with CFSE and then co-cultured with PBMCs in the presence of supernatants from OAd or OAd-aCLDN18.2-infected HEK293 cells; uninfected cells (Mock) served as a control. After 48 hours, representative flow cytometry plots (left) and cumulative data (right) are shown. Dead target cells were identified as CFSE⁺Zombie Dye⁺. The percentage of dead target cells was calculated as (CFSE⁺Zombie Dye⁺/CFSE⁺)×100%. Statistical significance was determined using an unpaired Student’s t-test (A) and ordinary one-way ANOVA (C) (n=3 per group). **p<0.01, ***p<0.001, and ****p<0.0001. Data are presented as the mean±SD. aCLDN18.2, anti-claudin 18.2; ANOVA, analysis of variance; CCK8, Cell Counting Kit-8; CFSE, 5-(and 6)-carboxyfluorescein diacetate succinimidyl ester; OAd, oncolytic adenovirus; PBMC, peripheral blood mononuclear cell; PBS, phosphate buffered saline.

OAd-aCLDN18.2 mediates ADCC, CDC and ADCP in vitro

To determine whether OAd-aCLDN18.2 affects NK cell antitumor activity, we co-cultured human NK cells with CFSE-labeled SGC-7901 and MGC-803 tumor cells in the presence of supernatants from OAd or OAd-aCLDN18.2-infected cells. NK cells were isolated from PBMCs using flow cytometry and their purity was assessed before and after sorting (online supplemental figure S2B). We then evaluated the ADCC activity of OAd-aCLDN18.2 against tumor cell lines. Compared with OAd, OAd-aCLDN18.2 significantly enhanced ADCC activity against both SGC-7901 and MGC-803 (figure 3A, online supplemental figure S3A). We also examined the expression of CD69, an early activation marker of NK cells, following co-culture with target cells (figure 3B). Notably, supernatants from OAd-aCLDN18.2-infected cells increased CD69 expression on NK cells, indicating enhanced NK cell activation.

Figure 3. OAd-aCLDN18.2 induces cytotoxicity against tumor cells. (A) ADCC assay. MGC-803 or SGC-7901 cells were labeled with CFSE. After 24 hours, NK cells were added and co-cultured in the absence (Mock) or presence of the supernatants from OAd or OAd-aCLDN18.2-infected HEK293 cells (1:1,000) for an additional 48 hours. Representative flow cytometry plots (left) and cumulative data (right) are shown. Dead target cells were identified as CFSE⁺Zombie Dye⁺. The percentage of dead target cells was calculated as (CFSE⁺Zombie Dye⁺/CFSE⁺)×100%. (B) NK cell activation. Cumulative flow cytometry data showing CD69 expression on NK cells following co-culture with target cells under the indicated conditions. (C) CDC assay. Tumor cells were seeded into 96-well plates at a density of 2×10⁴ cells per well. After 24 hours, PBMCs were added and co-cultured with the absence (Mock) or presence of the supernatants from OAd or OAd-aCLDN18.2-infected HEK293 cells (1:1,000) for an additional 48 hours. NHS was used as a source of complement. MGC-803, SGC-7901, Panc-01, MKN45, and OVCAR3 cells were used as target cells. The H.I. NHS served as a negative control. (D) ADCP assay. To evaluate the ability of OAd-aCLDN18.2 to promote phagocytosis by primary murine macrophages, supernatants from OAd or OAd-aCLDN18.2-infected HEK293 cells were co-incubated with SGC-7901 tumor cells and mouse BMDMs at dilutions of 1:1,000 (L) and 1:1,500 (H). And cytochalasin D (1 µg/mL) was added and served as the negative control. Phagocytic activity was quantified by flow cytometry (L: low dilution, H: high dilution). (E) Secretion of GZB, TNF-α, and IFN-γ in the co-culture supernatants was detected by western blot. (F) Densitometric quantification of the protein bands shown in (E) Statistical analysis was performed by ordinary one-way ANOVA (n=3). *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001. Data are presented as the mean±SD. ADCC, antibody-dependent cell-mediated cytotoxicity; ADCP, antibody-dependent cellular phagocytosis; aCLDN18.2, anti-claudin 18.2; ANOVA, analysis of variance; BMDMs, mouse bone marrow-derived macrophages; CBB, coomassie brilliant blue; CDC, complement-dependent cytotoxicity; CFSE, 5-(and 6)-carboxyfluorescein diacetate succinimidyl ester; GZB, granzyme B; H.I., heat-inactivated; IFN, interferon; NHS, normal human serum; NK, natural killer; ns, not significant; OAd, oncolytic adenovirus; PBMC, peripheral blood mononuclear cell; PBS, phosphate buffered saline; TNF, tumor necrosis factor.

Next, to investigate the sensitivity of the tumor cell lines to OAd-aCLDN18.2-mediated CDC, the target cells were co-cultured with NHS as a complement source, along with supernatants collected from OAd or OAd-aCLDN18.2-infected cells. The heat-inactivated NHS served as a negative control. We found that OAd-aCLDN18.2 enhanced CDC activity compared with OAd (figure 3C). Serial dilution of OAd-aCLDN18.2 supernatants in the CDC assay resulted in a progressive decrease of cytolytic activity, demonstrating a dose-dependent profile (online supplemental figure S3B).

Previous studies have demonstrated that human IgG can bind to mouse BMDMs due to the high homology of Fc receptors.²³ We next compared the ability of OAd-aCLDN18.2 to modulate the phagocytosis of SGC-7901 cells using murine macrophages. Flow cytometry analysis revealed that OAd-aCLDN18.2 induced a significantly higher level of BMDM-mediated phagocytosis against SGC-7901 cells compared with vehicle control and the killing activity declined with increasing dilution folds (figure 3D, online supplemental figure S3C). To evaluate the profile of cytokine secretion, western blot analysis was conducted to examine immune factors released by PBMCs in vitro. The results showed that OAd-aCLDN18.2 enhanced the secretion of these immune factors from PBMCs (figure 3E,F).

Therapeutic effects of OAd-aCLDN18.2 in xenograft model

After assessing the antitumor activity of OAd-aCLDN18.2 in vitro, we next evaluated its therapeutic efficacy in vivo. Although nude mice lack functional T cells, they retain a competent innate immune system, including NK cells, macrophages, and other myeloid cells, which serve as the primary effectors mediating ADCC, ADCP, and CDC—the key mechanisms underlying the antitumor activity of antibody-based therapies. Immunodeficient BALB/c nude mice were subcutaneously implanted with SGC-7901 cells and received either PBS, OAd, OAd-aCLDN18.2, or zolbetuximab (figure 4A). Both virus treatments and zolbetuximab delayed tumor growth. Notably, OAd-aCLDN18.2 was significantly more effective than the unarmed OAd in reducing both tumor volume and tumor weight. Furthermore, OAd-aCLDN18.2 demonstrated the most potent therapeutic effect among all treatment groups, including zolbetuximab (figure 4B–D). And there was no significant change in body weights between different groups (figure 4E).

Figure 4. Antitumor effect of OAd-aCLDN18.2 in SGC-7901 xenograft model. (A) Experimental treatment schedule for the SGC-7901 tumor-bearing immunodeficient mouse model (4 weeks old, n=5 per group). (B) Representative tumor images from each treatment group at the end of the study. (C–D) Tumor growth curves (C) and tumor weights (D) of SGC-7901-engrafted immunodeficient mice. (E) Summary data for body weight changes for all treatment groups. (F) Frequency of NK cells in tumor and spleen tissues. (G) CD69 expression on NK cells in the tumor and spleen. (H–C) Frequencies of M1/M2 macrophages and PMN-MDSCs in tumor (H) and spleen (I). (J) Immunohistochemical detection of GZB, IFN-γ, and TNF-α in tumor tissues. (K) ELISA quantification of IFN-γ and TNF-α levels in tumor tissues. (L) RT-qPCR of the major organs for all treatment groups. Representative RT-qPCR analysis of E1A copy numbers in major organs (heart, liver, spleen, kidney, and lung) and tumor tissues from mice in the PBS, OAd, and OAd-aCLDN18.2 treatment groups. Statistical analysis was performed by unpaired two-way ANOVA (C) and ordinary one-way ANOVA (D, F–I, K) (n=5; n=3 for panel L). *p<0.05, **p<0.01, ***p<0.001. Data are presented as the mean±SD. aCLDN18.2, anti-claudin 18.2; ANOVA, analysis of variance; GZB, granzyme B; IFN, interferon; M1 macrophages, classically activated macrophages; M2 macrophages, alternatively activated macrophages; NK, natural killer; ns, not significant; OAd, oncolytic adenovirus; PBS, ‌phosphate buffered saline; PMN-MDSC, polymorphonuclear myeloid-derived suppressor cells; RT-qPCR, reverse transcription-quantitative PCR; TNF, tumor necrosis factor.

To explore the mechanism underlying the antitumor activity of OAd-aCLDN18.2, we isolated tumor-infiltrating NK cells and splenocytes and analyzed the cells by flow cytometry (online supplemental figure S4A). As shown in figure 4F,G, compared with the PBS or OAd group, OAd-aCLDN18.2 treatment increased the proportion of NK cells within the tumor, but not in the spleen. Additionally, we observed elevated expression of the activation marker CD69 on tumor-infiltrating NK cells, whereas no obvious change was detected in splenic NK cells (figure 4F,G, online supplemental figure S4A,B). These findings suggest a tumor-localized, rather than systemic, activation of NK cells following OAd-aCLDN18.2 treatment.

We next examined several myeloid cell populations (online supplemental figure S4C–E), including M1/M2 macrophages and PMN-MDSCs. In contrast to the PBS control, treatment with OAd, OAd-aCLDN18.2, or zolbetuximab led to a significant increase in tumor-infiltrating M1 macrophages, whereas no such change was observed in the spleen. Notably, compared with OAd and zolbetuximab, OAd-aCLDN18.2 treatment reduced M2 macrophages and MDSCs within the tumor, with no corresponding alterations detected in splenic M2 or PMN-MDSCs populations (figure 4H,I). Furthermore, we assessed the expression of the effector cytokines IFN-γ, TNF-α and GZB in the tumor of SGC-7901 tumor-bearing immunodeficient mice using immunohistochemistry (figure 4J) and ELISA (figure 4K). As shown in figure 4J,K, significant differences were observed between the PBS group and each of the treatment groups, indicating that OAd-aCLDN18.2 enhances cytokine release.

To evaluate the in vivo safety profile of OAd-aCLDN18.2, we measured viral replication in mice across the different treatment groups. The OAd genome was detected in the liver, spleen, lung, and tumor tissues, with high-level expression observed exclusively in the tumor and only low-level expression detected in the other organs. No viral genome was detected in the heart or kidney (figure 4L). Collectively, these results indicate that the virus replicates primarily within tumor tissues with limited systemic distribution, further supporting that OAd-aCLDN18.2 enhances antitumor immunity while demonstrating a favorable safety profile in vivo.

To further evaluate the therapeutic efficacy of OAd-aCLDN18.2 in vivo, we established subcutaneous MKN45 gastric cancer xenografts in BALB/c nude mice (4 weeks old, n=5 per group) by inoculating 5×10⁶ tumor cells (figure 5A). Result shows that both virus-based treatments and zolbetuximab effectively delayed tumor growth. Notably, OAd-aCLDN18.2 exhibited significantly greater antitumor efficacy than the unarmed OAd, as evidenced by reductions in both tumor volume and tumor weight (figure 5B–D). Moreover, no significant differences in body weight were observed among the treatment groups, indicating an acceptable safety profile of the virus treatments (figure 5E).

Figure 5. Antitumor effect of OAd-aCLDN18.2 in MKN45 xenograft model. (A) Experimental treatment schedule for the MKN45 tumor-bearing immunodeficient mouse model. (B) Representative tumor images from each treatment group at the end of the study. (C–D) Tumor growth curves (C) and tumor weights (D) of MKN45-engrafted immunodeficient mice. (E) Summary data for body weight changes for all treatment groups. (F) Frequency of NK cells (left) and CD69 expression on NK cells (right) in the tumor. (G) Frequencies of PMN-MDSCs in tumor. Statistical analysis was performed by unpaired two-way ANOVA (C) and ordinary one-way ANOVA (D, F, G) (n=5). *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. Data are presented as the mean±SD. aCLDN18.2, anti-claudin 18.2; ANOVA, analysis of variance; NK, natural killer; ns, not significant; OAd, oncolytic adenovirus; PBS, ‌phosphate buffered saline; PMN-MDSC, polymorphonuclear myeloid-derived suppressor cells.

To elucidate the mechanism underlying the antitumor activity of OAd-aCLDN18.2, we analyzed tumor-infiltrating lymphocytes and immunosuppressive cell populations by flow cytometry. As shown in figure 5F, treatment with OAd-aCLDN18.2 markedly increased the proportion of tumor-infiltrating NK cells and elevated the expression of the activation marker CD69 on these cells, compared with the PBS and OAd groups. Additionally, OAd-aCLDN18.2 treatment significantly reduced the intratumoral frequency of MDSCs relative to both OAd and zolbetuximab groups (figure 5G). Collectively, these findings suggest that OAd-aCLDN18.2 promotes a tumor-localized activation of NK cells while alleviating PMN-MDSC-mediated immunosuppression, which may collectively contribute to its superior antitumor efficacy.

Taken together, these findings suggest that OAd-aCLDN18.2 promotes the infiltration of immune effector cells into the tumor while reducing the abundance of M2 macrophages and the infiltration of immunosuppressive cell populations, such as PMN-MDSCs. Consequently, OAd-aCLDN18.2 alleviates the immunosuppressive tumor microenvironment and enhances antibody-mediated antitumor immunity.

Discussion

In this study, we developed a conditionally replicating OAd engineered to continuously deliver an anti-CLDN18.2 IgG3 antibody (OAd-aCLDN18.2) and systematically evaluated its antitumor efficacy and underlying immune mechanisms. To our knowledge, this is the first report of an OV armed with a CLDN18.2-targeting IgG3 antibody, using viral delivery to overcome the pharmacokinetic limitations of the IgG3 subclass. Our findings demonstrate that OAd-aCLDN18.2 achieves robust and localized antitumor effects through enhanced ADCC, CDC, and ADCP, while favorably modulating the tumor immune microenvironment.

In vitro, OAd-aCLDN18.2 effectively mediated ADCC, CDC, and ADCP against CLDN18.2-positive tumor cells, accompanied by NK cell activation and enhanced secretion of multiple cytokines. Of note, although the anti-CLDN18.2 antibody is human-derived, it retains binding to murine Fc receptors due to high interspecies homology,¹² allowing us to evaluate ADCP using murine macrophages. These findings align with the known properties of IgG3: its extended hinge region confers exceptional flexibility for C1q binding, enabling significantly more potent complement activation than IgG1, while its high affinity for FcγRIIIa (CD16a) on NK cells drives robust ADCC.^{11 13 24} Although we used sorted CD56⁺ NK cells as effectors in our ADCC assays, the CD56⁺CD16⁺ subset is the principal mediator of ADCC via FcγRIIIa. Notably, the virus-delivered antibody retained these superior effector functions despite being produced locally rather than systemically administered, indicating that the OV platform does not compromise antibody quality or functionality. A limitation was that the antibody was evaluated as unpurified conditioned media and that purity, aggregation status, and precise binding affinity were not determined. Although IgG3 demonstrates the highest affinity binding to most FcγRs among all IgG subclasses, it has not been routinely selected as a therapeutic format due to its long hinge region and polymorphic nature.¹¹

Human and murine Fcγ receptor systems share substantial structural and functional homology, with approximately 60–70% sequence identity between orthologs.^{12 25 26} Human IgG subclasses bind to mouse FcγRs with relative affinities comparable to their human counterparts (IgG3>IgG1>IgG4>IgG2).²⁷ Importantly, human IgG3 exhibits high affinity for mouse FcγRIV, the functional ortholog of human FcγRIIIa, which mediates both ADCC and ADCP in murine effector cells.²⁸ Although the cellular distribution of FcγRs differs between species (mouse NK cells predominantly express FcγRIII, while macrophages express FcγRIV), this high conservation supports the use of murine models—particularly macrophages—to assess Fc-dependent effector functions of human antibodies.²⁹ This aligns with our observation that murine BMDMs mediated ADCP through engagement with human IgG3.

The in vivo antitumor efficacy of OAd-aCLDN18.2 was evaluated in immunodeficient BALB/c nude mice engrafted with SGC-7901 and MKN45 tumors. Although nude mice lack functional T cells, importantly, they retain a functional innate immune system, including NK cells, macrophages, and other myeloid cells, which are the primary effectors mediating ADCC, ADCP, and CDC—the key mechanisms we aimed to investigate. Another important consideration is that our human anti-CLDN18.2 IgG3 antibody can cross-react with murine Fc receptors owing to interspecies homology,¹² enabling functional engagement of innate immune cells (eg, NK cells and macrophages) in nude mice. Thus, despite the xenograft setting, the observed antitumor effects reflect genuine antibody-mediated ADCC, ADCP, and CDC activities. The absence of T cells in the nude mouse model imposes a limitation on interpreting TME remodeling. Our conclusions regarding changes in other cells should therefore be considered within this context. Future studies in immunocompetent models (such as syngeneic human CLDN18.2-expressing tumor models) are needed to fully characterize the adaptive immune responses.

A single intratumoral injection of OAd-aCLDN18.2 achieved significantly superior tumor growth suppression compared with both unarmed OAd and intraperitoneally administered zolbetuximab, the FDA-approved IgG1 antibody. This enhanced efficacy can be attributed to two complementary mechanisms. First, locally intratumoral production of IgG3 antibody from replicating viruses maintains persistent therapeutic concentrations at the tumor site, effectively circumventing the short plasma half-life (approximately 7 days for IgG3 vs 21–28 days for IgG1) that has historically precluded clinical development of IgG3-based therapies.^{11 12} Second, the OV itself contributes to antitumor immunity by inducing a local inflammatory response and remodeling the immunosuppressive tumor microenvironment.

Immune profiling of the tumor microenvironment revealed that OAd-aCLDN18.2 treatment significantly increased NK cell infiltration and activation, consistent with the potent ADCC-inducing capacity of IgG3. NK cells are the primary mediators of ADCC in vivo, and their recruitment to the tumor site is essential for optimal antibody therapeutic efficacy.^{30 31} Furthermore, we observed favorable modulation of the myeloid compartment, including changes in macrophage polarization, macrophage infiltration, and MDSC accumulation. These changes likely reflect the combined effects of local IgG3 production and virus-induced inflammatory responses, which together shift the tumor microenvironment from immunosuppressive to immunostimulatory.

An important safety observation is that viral replication remained largely restricted to tumor tissues, with no significant changes in body weight or other overt signs of toxicity. This tumor-selective replication is a hallmark of OVs and ensures that antibody production is confined to the tumor site, minimizing systemic exposure and off-target effects.¹⁵ Such localization confines both the efficacy and potential toxicity of the IgG3 antibody to the tumor, addressing concerns regarding its systemic immunogenicity and stability.

In conclusion, this study establishes a novel therapeutic paradigm that combines the superior effector functions of IgG3 antibodies with OV-mediated intratumoral delivery. By overcoming the pharmacokinetic barrier that has historically limited the clinical application of IgG3, this approach unlocks the full therapeutic potential of this antibody subclass. OAd-aCLDN18.2 represents a promising new strategy for CLDN18.2-positive gastric, pancreatic, and ovarian cancers, and this platform may be extended to other IgG3-based antibodies targeting different tumor-associated antigens.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.