Amivantamab induces immune-mediated cytotoxicity in mesothelioma via EGFR and MET engagement

Abstract

Introduction:

Diffuse pleural mesothelioma (DPM) is a lethal malignancy with no approved therapies directly targeting tumor cells. Based on reports that epidermal growth factor receptor (EGFR) and MET proto-oncogene (MET) are frequently expressed in DPM, we sought to systematically assess their co-expression and determine the preclinical activity of amivantamab, a bispecific antibody that targets EGFR and MET.

Methods:

We evaluated EGFR and MET expression in patient cohorts and cell lines using transcriptomic analysis, immunohistochemistry, and biochemical assays. The mechanistic actions of amivantamab, including receptor internalization, signaling blockade, and immune-mediated cytotoxicity, were assessed using in vitro co-culture systems with peripheral blood mononuclear cells (PBMCs) and natural killer (NK) cells. In vivo efficacy was evaluated in mesothelioma patient-derived xenograft (PDX) models using immunodeficient mice reconstituted with human NK cells.

Results:

Transcriptomic and single-cell RNA sequencing analyses of DPM patient cohorts revealed frequent co-expression of EGFR and MET, predominantly in malignant cells. Immunohistochemical analyses confirmed EGFR and MET protein expression across mesothelioma histologic subtypes. In vitro, amivantamab preferentially bound to DPM cells, inhibited ligand-induced EGFR and MET signaling, and promoted receptor internalization. Co-culture experiments demonstrated that amivantamab induced dose-dependent cytotoxicity via NK cell–mediated antibody-dependent cellular cytotoxicity (ADCC). In multiple mesothelioma PDX models, the combination of amivantamab and NK cells significantly reduced tumor growth without overt toxicity.

Conclusions:

Amivantamab demonstrates robust preclinical antitumor activity in mesothelioma primarily through innate immune–mediated cytotoxicity associated with EGFR and MET engagement. These findings support the clinical evaluation of bispecific EGFR/MET-targeting antibodies in mesothelioma.

Graphical Abstract

Introduction

Antibody-based therapies targeting surface receptors are a cornerstone of modern oncology¹. The single-chain variable fragments (scFvs) that recognize surface receptors can be reformatted to multiple IgG isotypes and their derivative antibody-drug conjugates, immunotoxins, and cellular therapies with chimeric antigen receptors². The immunoglobulin G1 (IgG1) isotype is one of the most commonly used antibody formats and can induce antibody-dependent cellular cytotoxicity, antibody-dependent cellular phagocytosis, and complement-dependent cytotoxicity^3–5.

Diffuse pleural mesothelioma (DPM) is a rare malignancy associated with poor survival outcomes^{6, 7}. Most cases are associated with asbestos exposure that often precedes diagnosis by decades, although radiation⁸ and rare molecular variants such as EWSR1-ATF1 fusions may also be involved^{9, 10}. The most significant clinical development in mesothelioma has been the approvals of immune checkpoint inhibitors that have improved overall survival compared to cytotoxic chemotherapy^11–14. The clinical benefit of dual immune checkpoint inhibition or chemo-immunotherapy is primarily observed in patients with the less common non-epithelioid histologic variants of mesothelioma^12–14. Therapies that improve survival outcomes for patients with the more common epithelioid histologic variant of mesothelioma are urgently needed. As these immune checkpoint inhibitors target and modulate T cell activity, there are currently no approved antibody-based therapies directly targeting mesothelioma. For these reasons, we explored the expression of targetable surface receptors in mesothelioma to identify clinically available antibody-based therapies that could potentially be repurposed for this disease. In light of prior reports of epidermal growth factor receptor (EGFR)^15–17 and MET proto-oncogene (MET)^{18, 19} expression in DPM, and the development of the bispecific antibody amivantamab that targets EGFR and MET, we sought to characterize the co-expression of these surface receptors and evaluate the preclinical activity and immune-mediated effects of amivantamab in DPM.

Materials and Methods

Experimental Design

The objective of this study was to evaluate the therapeutic efficacy and immune-mediated mechanisms of amivantamab, a bispecific EGFR/MET antibody, in diffuse pleural mesothelioma. We designed a study that integrated (i) bulk and single-cell transcriptomic analysis of clinical datasets, (ii) immunohistochemical evaluation of EGFR and MET protein expression in mesothelioma specimens, (iii) in vitro functional assays using a panel of nine mesothelial and mesothelioma cell lines to assess inhibition of signaling and antibody-dependent cellular cytotoxicity (ADCC), and (iv) in vivo efficacy studies using patient-derived xenograft (PDX) models reconstituted with human NK cells in NSG mice. Prespecified end points included tumor cell viability, signaling protein phosphorylation, and in vivo tumor growth inhibition (TGI).

Bioinformatics Analysis

To assess EGFR and MET expression in clinical mesothelioma samples, we analyzed public available RNA sequencing datasets. We obtained transcriptomic data from a DPM patient cohort (Bueno et al., EGAD00001001915, n = 209)²⁰ and the Cancer Genome Atlas Mesothelioma dataset (TCGA-MESO, n = 86). Raw sequencing reads were processed using the Mayo Analysis Pipeline for RNA-seq (MAP-R Seq)²¹ to generate gene-level counts. Expression values were then normalized to counts per million (CPM) using the edgeR Bioconductor pipeline. We calculated Pearson correlation coefficients (R) between EGFR and MET expression levels and plotted the relationships using GraphPad Prism 10.0.0. Exploratory subgroup analyses according to histologic subtype were also performed.

Single-cell RNA-seq

Processed single-cell RNA-seq data from pleural mesothelioma tumors (GSE190597) were downloaded as the authors’ fully processed RDS object. No additional QC filtering, normalization, or clustering was performed. To reproduce the tumor-only visualization shown in Fig. 1C of Giotti et al.²², the scANVI latent embeddings made available in the authors’ GitHub repository were imported and integrated with the object. Expression levels for EGFR and MET were extracted from the RNAseq assay data contained in the processed object, and plotted using the ScCustomize package in R.

Fig. 1. Tumor-restricted co-expression of epidermal growth factor receptor (EGFR) and MET proto-oncogene (MET) in clinical samples of mesothelioma.

(A) EGFR and MET messenger RNA (mRNA) expression in DPM patient samples from two independent cohorts (Bueno et al., n = 209; TCGA, n = 86). Expression values were normalized to counts per million (CPM) and are shown on a log₂ scale. Each point represents one tumor; colors indicate histological subtype (blue = epithelioid, red = sarcomatoid, purple = biphasic, green = not otherwise specified [NOS]). There is a significant positive correlation between EGFR and MET expression in both cohorts, consistent with frequent co-expression of these targets in DPM (see regression lines). (B) Uniform Manifold Approximation and Projection (UMAP) plots showing normalized single-cell expression of EGFR (left) and MET (right) across 84,526 primary tumor cells derived from 13 patients with DPM prior to treatment. Expression of both receptors is predominantly localized within malignant clusters. (C) Violin plots depicting log-normalized expression levels of EGFR (left) and MET (right) across annotated cell populations. Expression is shown as In (counts per 10,000 + 1). Malignant cells show markedly higher expression of both targets compared to non-malignant stromal or immune cells. (D) Joint density plot demonstrating co-expression of EGFR and MET at the single-cell level. EGFR⁺MET⁺ double-positive cells are highly enriched in the malignant compartment. (E) Scatter plot demonstrating the relationship between EGFR and MET protein expression quantified by immunohistochemistry (IHC) H-score in the institutional mesothelioma cohort (n = 27), including pleural, peritoneal, and pericardial cases. Each point represents one tumor specimen and is color-coded by histologic subtype. The observed positive correlation supports co-expression of these therapeutic targets at the protein level.

Cell cultures and reagents

The cell lines MET-5A, EM-Meso, H2373, H2452, H2461, H2596, Meso41, Meso95 and ONE58 were obtained from Tobias Peikert’s Lab (Mayo Clinic, Rochester MN, USA). MET-5A cells were cultured in Medium 199 (Medium 199, Earle’s Salts; #11150067, Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS; #10437–028, Thermo Fisher Scientific) and 1% Penicillin-Streptomycin solution (#15140–122, Thermo Fisher Scientific). Other cell lines were maintained in RPMI 1640 medium (10–040-CV, Corning) containing 10% FBS (#10437–028, Thermo Fisher Scientific) and 1% Penicillin-Streptomycin solution (#15140–122, Thermo Fisher Scientific). Cell lines were cultured under standard conditions. Authentication and mycoplasma testing were performed prior to experiments.

PDX Model Studies

All animal experiments were approved by the Mayo Clinic Institutional Animal Care and Use Committee (IACUC; protocol A00007907–24) and conducted in accordance with institutional guidelines. Patient-derived xenograft (PDX) experiments were performed to evaluate the in vivo antitumor activity of amivantamab. Female NOD/SCID/γc⁻/⁻ (NSG) mice (8–12 weeks old) were used for all studies. Because NSG mice lack functional T, B, and NK cells, human NK cells were co-implanted to enable immune effector–mediated cytotoxicity in vivo. This strategy was informed by our in vitro findings indicating that immune-mediated mechanisms substantially contribute to the antitumor activity of amivantamab in DPM.

For admixture xenografts, dissociated PDX tumor cells were injected subcutaneously into the flank either alone or together with ex vivo–expanded human NK cells. Mice were randomized into four groups: (i) tumor cells alone with PBS treatment, (ii) tumor cells alone treated with amivantamab, (iii) tumor cells plus NK cells with PBS treatment, and (iv) tumor cells plus NK cells treated with amivantamab.

Amivantamab was administered intraperitoneally twice weekly for three weeks beginning one day after implantation. Tumor size and body weight were measured twice weekly. Tumor volume was calculated as (1/2 × length × width²). Mice were euthanized if tumors reached predefined humane endpoints. Additional details regarding PDX model generation, NK cell preparation, and tumor growth inhibition analysis are provided in the Supplementary Methods.

Statistical Analysis

Statistical analyses were performed using GraphPad Prism 10.0.0. Data are expressed as mean ± SD for in vitro experiments and mean ± SEM for in vivo studies. Comparisons between two groups were performed using two-tailed unpaired t tests. For experiments with more than two groups, one-way ANOVA followed by Šidák multiple comparison test was used. All p values were based on two-sided hypotheses, with p < 0.05 considered statistically significant. Specific n values and statistical tests for each experiment are indicated in the respective figure legends.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files. The public datasets analyzed in this study are available from the following sources: the Bueno et al. cohort from the European Genome-phenome Archive (EGAD00001001915); The Cancer Genome Atlas Mesothelioma (TCGA-MESO) dataset from the National Cancer Institute’s Center for Cancer Genomics (https://www.cancer.gov/ccg/access-data); and the single-cell RNA-seq data from the Gene Expression Omnibus (GEO accession GSE190597). The patient-derived xenograft (PDX) models (177-R, 175-T and 263-T) can be requested through the NCI Patient-Derived Models Repository (PDMR), and all associated genomic, transcriptomic, and clinical data can be accessed via the NCI PDMR website (https://dctd.cancer.gov/drug-discovery-development/reagents-materials/pdmr).

The detailed methods and materials of the above experiments are presented in Supplementary Methods.

Results

EGFR and MET are co-expressed in mesothelioma

To characterize EGFR and MET expression in mesothelioma, we first analyzed two large, independent patient cohorts. Bulk RNA sequencing revealed that EGFR, MET or both were commonly expressed in DPM across all major histological subtypes, including epithelioid, biphasic, and sarcomatoid tumors (Fig. 1A). A significant, weak-to-moderate positive correlation between EGFR and MET expression levels was observed in both the Bueno et al. dataset (n = 209; R = 0.387, R² = 0.1497; p < 0.0001) and the TCGA-MESO cohort (n = 86; R = 0.494, R² = 0.2442; p < 0.0001). To further assess the cellular distribution of EGFR and MET expression in clinical specimens, we analyzed single-cell RNA sequencing (scRNA-seq) data from a recent study by Giotti et al.²², which profiled 84,526 primary tumor cells from 13 treatment-naïve patients with DPM (9 epithelioid, 2 sarcomatoid, and 2 biphasic). EGFR expression was predominantly localized to the malignant cell cluster, with limited expression also observed in fibroblast clusters. Similarly, MET expression was largely confined to the malignant compartment, with modest expression detected in endothelial and B cell clusters (Fig. 1B, C). Strikingly, joint expression density analysis revealed that cells co-expressing EGFR and MET were almost exclusively restricted to the malignant cluster, comprising the majority of this population (Fig. 1D). Together, these findings suggest that EGFR and MET are frequently co-expressed in mesothelioma and provide a biological rationale for dual EGFR/MET targeting with amivantamab.

To validate the transcriptomic findings at the protein level, 27 mesothelioma clinical specimens were assessed by immunohistochemistry (IHC) for EGFR and MET expression. The cohort included pleural (n = 17), peritoneal (n = 8), and pericardial (n = 2) mesothelioma specimens, comprising epithelioid (n = 16), biphasic (n = 2) and sarcomatoid (n = 9) histologies (Table S4). Correlation analysis showed a moderately strong positive relationship between EGFR and MET protein expression (n = 27; R = 0.627, R² = 0.3933, p = 0.0005), supporting frequent co-expression of these therapeutic targets at the protein level across mesothelioma histologic subtypes (Fig. 1E). Subtype-specific analyses supported this finding (Fig. S1A–C). Representative IHC staining demonstrating co-expression of EGFR and MET is shown in Fig. S2 and S3. Collectively, these findings support that EGFR and MET are expressed in mesothelioma.

Amivantamab Binds EGFR/MET-Expressing Mesothelioma Cells

We then sought to determine whether EGFR and MET proteins were expressed in DPM cell lines. For detailed mechanistic studies, H2373 and H2461 cell lines were selected to represent sarcomatoid and epithelioid histologic subtypes, respectively. Immunoblotting of cell lysates from multiple DPM lines demonstrated readily detectable levels of EGFR and MET bands, at levels higher than those observed in an immortalized mesothelial cell line (MET-5A) (Fig. 2A). To verify whether amivantamab binds DPM cells via EGFR or MET, we assessed amivantamab binding by flow cytometry and immunofluorescence. Amivantamab showed strong binding to the surface of DPM cell lines, resulting in a marked rightward shift in median fluorescence intensity (MFI) relative to isotype control antibody (Fig. 2B). Confocal microscopy further revealed co-localization of amivantamab with EGFR and MET on the cell membrane of DPM cells (Fig. 2C). Similar high-level binding and co-localization were consistently observed across a panel of additional DPM cell lines, confirming broad target availability (Fig. S4A and S4B). Together, these findings demonstrate that EGFR and MET are commonly co-expressed in DPM cell models and that amivantamab effectively engages its targets on the tumor cell surface.

Fig. 2. Amivantamab binds to and functionally inhibits EGFR and MET in mesothelioma cells through dual receptor blockade and internalization.

(A) Western blot analysis of EGFR and MET protein levels in mesothelioma cell lines vs. an immortalized mesothelial cell line (MET-5A). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) serves as a loading control. (B) Flow cytometry data of median fluorescence intensity (MFI) illustrating binding of amivantamab to the surface of mesothelial vs. mesothelioma cells. Treatment with amivantamab (3 μg/mL) causes a dramatic increase in cell-surface fluorescence compared to IgG1 control (3 μg/mL) in mesothelioma cells. ΔMFI = Amivantamab median MFI – IgG1 median MFI. (C) Confocal immunofluorescence images showing co-localization of amivantamab with EGFR and MET on mesothelioma cell membranes. H2373 cells were incubated with amivantamab (green), then fixed and stained for EGFR (red) and MET (blue); nuclei are counterstained with 4′,6-diamidino-2-phenylindole (DAPI) (cyan). (D) Western blots show phosphorylated EGFR, MET, AKT, and ERK1/2 in mesothelioma cells (H2373, H2461) stimulated with epidermal growth factor (EGF) and hepatocyte growth factor (HGF). Cells were pre-treated with amivantamab (300 μg/mL) for 1 hour, followed by stimulation with EGF (50 ng/mL for 5 minutes) or/and HGF (200 ng/mL for 15 minutes), either individually or simultaneously. GAPDH serves as a loading control. (E) Amivantamab induces rapid internalization of EGFR/MET in mesothelioma cells. Graph shows the increase in intracellular fluorescence over time for H2373, H2461, and MET-5A cells incubated with pH-sensitive dye–labeled amivantamab or immunoglobulin G1 (IgG1). Fluorescence (arbitrary units) reflects antibody–receptor complexes in acidic endosomes. Curves are mean ± standard deviation (SD) (n = 12). * p < 0.05, **** p < 0.0001.

Amivantamab Suppresses Oncogenic Signaling via Dual Receptor Blockade and Tumor-Selective Internalization of EGFR and MET

We next sought to determine the impact of amivantamab on EGFR and MET activity. First, we assessed whether amivantamab can inhibit EGFR and MET activation in DPM cell lines. In cell culture, H2373 and H2461 mesothelioma cells showed robust induction of phosphorylation of EGFR and MET upon stimulation with their cognate ligands epidermal growth factor (EGF) and hepatocyte growth factor (HGF), respectively, under control conditions. Phospho-EGFR levels were markedly reduced in amivantamab-treated cells despite EGF stimulation. Similarly, HGF-driven MET phosphorylation was suppressed by amivantamab. Concomitantly, downstream signaling through the phosphoinositide 3-kinase (PI3K) / AKT and mitogen-activated protein kinase (MAPK) / ERK pathways was attenuated. Importantly, simultaneous stimulation with both EGF and HGF also resulted in suppression of EGFR and MET downstream signaling by amivantamab, supporting its ability to inhibit dual receptor signaling under conditions of concurrent ligand activation (Fig. 2D). These findings indicate that amivantamab suppresses ligand-induced activation of both EGFR and MET in DPM cells and attenuate downstream signaling output.

We also found that amivantamab promotes receptor down-modulation on the surface of DPM cells. To quantify receptor internalization, DPM cell lines and MET-5A cells were incubated with amivantamab labeled using the Human FabFluor-pH Red reagent, which fluoresces following entry into acidic intracellular compartments. Real-time imaging and quantification were conducted using the IncuCyte S3-C2 live-cell analysis system. Fluorescence intensity significantly increased over time in amivantamab-treated DPM cells, consistent with robust internalization of EGFR–MET–amivantamab complexes into acidic endosomal vesicles (Fig. 2E). This effect was significantly higher in both H2373 (p < 0.0001) and H2461 (p = 0.0013) cell lines with amivantamab compared to IgG1 controls. In contrast, no significant internalization was detected in the immortalized mesothelial cell line MET-5A compared to IgG1 controls (p = 0.82). These findings indicate that amivantamab preferentially induces receptor internalization in mesothelioma cells compared with immortalized mesothelial cells, leading to down-modulation of surface EGFR and MET. This tumor-selective internalization likely contributes to the attenuation of downstream signaling pathways and is consistent with previous observations of amivantamab-mediated EGFR/MET receptor clearance.^{23, 24}

Immune Effector Cell Engagement Is Essential for Amivantamab-Mediated Cytotoxicity

We then sought to determine whether immune effector cells mediate amivantamab-induced cytotoxicity against DPM. DPM cell lines stably expressing mCherry fluorescent protein were used to quantify viable tumor cells by fluorescence intensity. As shown in Fig. 3A and B, amivantamab had minimal direct impact on DPM cell viability in the absence of immune cells. In contrast, when PBMCs were co-cultured with the tumor cells, amivantamab induced potent dose-dependent tumor cell killing. At the highest dose tested (10 μg/mL), amivantamab plus PBMCs significantly reduced tumor cell viability relative to untreated controls, with half maximal inhibitory concentration (IC₅₀) values of 1.63 μg/mL for H2373 and 2.43 μg/mL for H2461. In contrast, PBMCs with an isotype IgG1 had no significant effect. Notably, neither amivantamab nor PBMCs alone elicited cytotoxicity; only the combination was cytotoxic. These results demonstrate that amivantamab has minimal antitumor activity against DPM in the absence of immune effector cells, highlighting the critical role of immune engagement in mediating its cytotoxic effects in vitro.

Fig. 3. Amivantamab mediates potent immune cell–dependent in mesothelioma cells.

(A, B) Amivantamab-mediated cytotoxicity is dependent on peripheral blood mononuclear cells (PBMCs). Dose–response curves show the viability of H2373 (A) and H2461 (B) cells after 72 hours of treatment with an IgG1 control or amivantamab, in the presence or absence of PBMCs at an effector-to-target (E:T) ratio of 10:1. (C, D) Amivantamab enables potent natural killer (NK) cell-mediated cytotoxicity. Dose–response curves show the viability of H2373 (C) and H2461 (D) cells after 72 hours of treatment with an IgG1 control or amivantamab, in the presence or absence of NK cells (E:T ratio of 10:1). (E) Representative live-cell fluorescence images depicting the dose-dependent cytotoxic effect of amivantamab (Ami) on mCherry-expressing H2373 tumor cells co-cultured with human NK cells (10:1 E:T) at 72 hours. An IgG1 control is shown for comparison. (F) Amivantamab induces potent Fc gamma receptor (FcγR) engagement. The antibody-dependent cellular cytotoxicity (ADCC) reporter assay shows relative luminescence units (RLU) from co-cultures of effector Jurkat cells (expressing FcγRIIIa) with the indicated target cells (MET-5A, H2373, H2461) in the presence of IgG1 or amivantamab (Ami) at 0.01 μg/mL. Data in (A-D) and (F) are shown as mean ± SD (n = 6 replicates); **p < 0.01, ****p < 0.0001.

Next, we investigated the role of specific immune effector cells in mediating amivantamab-induced tumor cell killing by comparing the activity of isolated natural killer (NK) cells to that of PBMCs. NK cells were isolated from the same donors as the PBMCs and co-cultured with DPM cells at the same effector-to-target (E:T) ratio of 10:1 to ensure comparability.

In the presence of amivantamab, NK cells induced rapid and robust cytotoxicity against DPM cells, whereas NK cells with control IgG exerted minimal effects (Fig. 3C, D). Representative fluorescence images of the H2373 co-culture at 72 hours visually confirm this potent cytotoxic effect (Fig. 3E). Notably, a direct comparison of the dose-response curves revealed that the IC₅₀ values for NK cell–mediated killing (H2373: 25.27 ng/mL; H2461: 13.53 ng/mL) were substantially lower than those observed for PBMC-mediated killing (compare Fig. 3C, D with Fig. 3A, B). These results suggest that NK cell–driven ADCC contributes more potently to the in vitro antitumor effects of amivantamab compared to the activity mediated by the broader cellular populations within PBMCs.

To further quantify immune effector engagement, we performed an ADCC reporter assay using a reporter construct under nuclear factor of activated T cells (NFAT) control. In this system, amivantamab induced significantly higher luciferase activity compared with control IgG1 in all tested cell lines — MET-5A, H2373, and H2461 (Fig. 3F). Specifically, amivantamab induced a modest but significant increase in luciferase activity with the immortalized mesothelial cell line MET-5A (1.95-fold, p = 0.0053). In contrast, this effect was markedly greater in the malignant DPM cell lines H2373 (7.03-fold, p < 0.0001) and H2461 (10.89-fold, p < 0.0001). These data indicate that although amivantamab can induce FcγR engagement in both malignant and non-malignant mesothelial cells, the magnitude of ADCC activation is markedly greater in DPM cells than in the immortalized MET-5A model. This finding suggests that amivantamab preferentially enhances immune-mediated cytotoxicity in tumor cells, potentially reflecting higher receptor density or increased susceptibility to effector mechanisms in malignant versus nonmalignant mesothelial contexts.

Amivantamab exhibits potent, NK cell-dependent antitumor activity in patient-derived xenograft models

To evaluate the ADCC activity of amivantamab in vivo, we established reconstitution models using immunodeficient NSG mice. A preliminary study in one PDX model (263-T) confirmed that amivantamab monotherapy had no significant antitumor activity in the absence of co-implanted human immune cells (p = 0.996 vs. PBS). Based on this result and the goal of minimizing unnecessary animal use, subsequent in vivo studies were designed to focus on immune cell–dependent antitumor activity under NK cell–reconstituted conditions, and therefore did not include an amivantamab monotherapy arm.

We first evaluated therapeutic efficacy in a pleural mesothelioma PDX model (177-R, epithelioid histology), which was confirmed to co-express EGFR and MET (Fig. S6A, B). The in vivo study design is shown schematically in Fig. 4A. This model involved the co-implantation of PDX cells and ex vivo-expanded human NK cells into immunodeficient NSG mice. The combination of amivantamab and human NK cells significantly inhibited tumor growth compared to control groups receiving either PDX cells alone or PDX cells with NK cells (Fig. 4B). A one-way ANOVA confirmed a significant treatment effect (p = 0.015), with the amivantamab-treated group exhibiting a tumor growth inhibition (TGI) of 89.1% relative to the NK cell control group. This tumor suppression was durable, sustaining for at least 25 days following the final antibody dose. Importantly, no overt toxicity was observed, and body weight remained stable (Fig. 4C).

Fig. 4. Amivantamab, in combination with human NK cells, exhibits potent and well-tolerated antitumor activity in a pleural mesothelioma patient-derived xenograft (PDX) model.

(A) Schematic of the 177-R PDX model in NOD/SCID/γc^−/− (NSG) mice. (B) Tumor growth curves for an epithelioid mesothelioma 177-R PDX under different treatment conditions. Black = PDX alone (no NK cells) treated with PBS; Blue = PDX + NK cells treated with PBS; Red = PDX + NK cells treated with amivantamab (30 mg/kg, i.p., twice weekly). Points represent mean tumor volume (mm³) ± SEM (n = 5 mice per group). (C) Body weight curves for the corresponding treatment groups. Points represent mean body weight (g) ± SEM.

Statistical analyses of final tumor volumes (day 47) and body weights were performed using one-way ANOVA followed by Šidák’s multiple comparison test. ns, not significant; *, p < 0.05.

The activity of amivantamab was also assessed in two peritoneal mesothelioma PDX models. In the 175-T peritoneal model, amivantamab demonstrated significant antitumor activity (TGI of 68.9%; Fig. S6C), consistent with the pleural model findings. We further evaluated amivantamab in the 263-T model, which showed a marked response with a TGI of 75.7% on day 54 relative to the NK cell control group (Fig. S6E). Both peritoneal models (175-T and 263-T) were confirmed to co-express EGFR and MET and the treatments were well tolerated (Fig. S6A, B, D, F).

Combination with the IL-15 superagonist N-803 further enhances antitumor activity

To explore strategies to enhance the clinical value of amivantamab, we investigated its combination with the clinically approved IL-15 superagonist N-803. In vitro, the combination of amivantamab and N-803 demonstrated synergistic antitumor activity against H2373 mesothelioma cells across a broad range of concentrations (Fig. S7A). In contrast, the interaction in H2461 cells was more complex, showing antagonism at lower concentrations but synergy at higher concentrations, indicating heterogeneous interaction patterns across cell lines (Fig. S7B). We next evaluated this combination in the 177-R PDX model, which included an amivantamab monotherapy arm (NK+Ami). Both the NK+Ami and NK+N803+Ami groups demonstrated substantial tumor growth inhibition compared with control groups (Fig. S7C). Notably, the addition of N-803 further enhanced antitumor efficacy, with tumor growth inhibition (TGI) increasing from 89.0% in the NK+Ami group to approximately 100% in the NK+N803+Ami group. Furthermore, tumors became non-palpable in 4 of 5 mice treated with the NK+N803+Ami combination, indicating near-complete suppression of measurable tumor growth in most animals.

Although the difference in final tumor volumes between NK+Ami and NK+N803+Ami did not reach statistical significance (p = 0.8722, one-way ANOVA), this was likely due to the substantial baseline efficacy of amivantamab in this model, resulting in a potential floor effect that limits detection of incremental benefits. Importantly, the addition of N-803 did not result in increased toxicity as assessed by body weight monitoring (Fig. S7D).

Together, these findings indicate that combining amivantamab with N-803 further enhances antitumor activity in vivo and represents a feasible and potentially effective strategy to augment its therapeutic efficacy.

Discussion

Our work provides a strong preclinical rationale for the use of the EGFR/MET bispecific antibody amivantamab as a potential therapeutic strategy in mesothelioma. We first confirmed that EGFR and MET are commonly co-expressed in DPM patient tumors and cell lines. We then demonstrated that amivantamab exerts potent antitumor activity against DPM through dual mechanisms: direct inhibition of oncogenic signaling and engagement of immune effector cells. Importantly, our in vivo findings using a DPM PDX model revealed that amivantamab, in the presence of adoptively transferred human NK cells, induced significant and durable tumor regression, primarily driven by antibody-dependent cellular cytotoxicity (ADCC). Consistent with this, similar antitumor activity was also observed across two independent PDX models of peritoneal mesothelioma, suggesting broader applicability of our findings to the two most common sites of mesothelioma. These results suggest that amivantamab is a promising candidate for clinical evaluation in the treatment of mesothelioma.

Amivantamab has demonstrated clinical efficacy in EGFR-mutant NSCLC^26–30. Whereas amivantamab has been shown to be active through receptor-ligand blocking, receptor down-modulation and downstream signaling inhibition in EGFR-mutant NSCLC^{23, 25, 31}, the immune-mediated effects seem to be the dominant driver of amivantamab’s activity in mesothelioma. This finding differs from a pre-clinical investigation of wild-type EGFR NSCLC in which the activity of amivantamab was attributed to its disruption of amphiregulin interactions with EGFR³². In short, given the multiple mechanisms of action of amivantamab, its activity is likely highly context-dependent.

Prior studies have reported EGFR and MET activation in mesothelioma, with EGFR expression observed in a majority of tumors^33–35; however, clinical targeting of EGFR with tyrosine kinase inhibitors, alone^{16, 17} or in combination with vascular endothelial growth factor inhibitors, has been largely ineffective¹⁵.

Currently, there are no biomarker-selected therapies for the treatment of DPM. While immune checkpoint inhibitors have improved outcomes over chemotherapy, their benefit is primarily observed in non-epithelioid subtypes, and the role of PD-L1 as a predictive biomarker remains uncertain^{12–14, 36–38}. Therapies that improve survival for patients with the more common epithelioid histology are urgently needed. Other targeted therapies are also in development for mesothelioma, such as those harnessing collateral lethality with PRMT5 inhibition in the cases with loss of MTAP, or by disrupting the Hippo pathway^39–45.

In our in vivo DPM PDX model, we focused on the contribution of ADCC by supplementing human NK cells. Notably, the PDX model used in this study was derived from a patient with epithelioid mesothelioma, which is the most common histological subtype of DPM and often exhibits limited responsiveness to immune checkpoint inhibitors. This histologic context enhances the translational significance of our findings, as it highlights the therapeutic potential of amivantamab in a clinically relevant and underserved patient population. While this strategy enabled us to avoid graft-versus-host disease (GVHD) and allowed longer observation periods, it also introduced certain limitations⁴⁶. Due to the cytokine-deficient environment in NSG mice, NK cell viability is short-lived, and critical immune processes such as NK cell trafficking, monocyte-mediated trogocytosis, and T cell-mediated adaptive responses could not be evaluated. Nonetheless, the inclusion of appropriate control groups allowed us to isolate the antitumor effects of amivantamab and confirm its activity in this model.

In our in vitro PBMC co-culture experiments, human leukocyte antigen (HLA) matching was not performed, and PBMCs were obtained from unmatched donors. This raises the possibility of HLA mismatch-driven cytotoxicity confounding the evaluation of amivantamab’s specific effects. Nevertheless, all experiments included proper controls, and our findings consistently support a specific antitumor effect mediated by amivantamab. Future studies using HLA-matched PBMCs may allow for a cleaner assessment of amivantamab’s immune-mediated efficacy.

Given the absence of biomarker-directed therapies for DPM and the limited efficacy of immune checkpoint inhibitors in the epithelioid subtype, amivantamab offers a potential therapeutic avenue that directly targets tumor cells independent of PD-L1 status. Clinical evaluation in this patient population is warranted. In other cancers such as NSCLC and head and neck squamous cell carcinoma, combinations with epidermal growth factor receptor–tyrosine kinase inhibitors (EGFR-TKIs)²⁷ or PD-1 inhibitors⁴⁷ have shown clinical promise. Similar strategies may prove beneficial in mesothelioma as well.

IL-15 agonists may enhance ADCC through NK cell activation. Our preliminary data suggest potential synergy with the FDA-approved IL-15 superagonist N-803 for bladder cancer, although further studies are needed.

Finally, our findings suggest that the therapeutic potential of amivantamab extends beyond the pleural subtype. Potent antitumor activity was observed in two independent peritoneal mesothelioma PDX models that co-expressed EGFR and MET (Fig. S7A, B), supporting broader applicability across mesothelioma subtypes. Given the absence of biomarker-directed therapies for both pleural and peritoneal mesothelioma, and the limited efficacy of immune checkpoint inhibitors in the common epithelioid subtype, amivantamab represents a compelling therapeutic strategy that directly targets tumor cells independent of PD-L1 status. Antitumor activity across models with varying EGFR and MET expression suggests that response may not strictly depend on high target expression and warrants further investigation. Collectively, these data provide a strong rationale for advancing amivantamab into clinical trials in patients with EGFR/MET-expressing mesothelioma.