Development of a Novel CLDN18.2-directed Monoclonal Antibody and Antibody–Drug Conjugate for Treatment of CLDN18.2-Positive Cancers

This article is featured in Selected Articles from This Issue, p. 1363

Claudin 18 (CLDN18) is a member of the claudin family of tight junction proteins that play a critical role in maintaining cell-to-cell adhesion and regulating transport between adjacent cells (1). This family of 23 proteins is differentially expressed in a variety of epithelial and endothelial cells throughout the body (1, 2). CLDN18 was originally thought to be expressed in lung and gastric tissue. It has since been shown to exist in two distinct isoforms; CLDN18.1 and CLDN18.2 (3, 4). CLDN18.1 expression is restricted to normal lung tissue and is also highly expressed in lung adenocarcinomas (5). CLDN18.2 expression is present in normal gastric mucosa cells but otherwise undetectable in any other normal adult cell type either at the transcript or protein level (5). Conversely, CLDN18.2 is frequently detected in malignancies of the stomach, gastroesophageal junction (GEJ), and pancreas (4, 6, 7). In normal tissues, CLDN18.2 protein is localized to tight junctions; however, alterations in cell polarity that occur during malignant transformation are thought to cause CLDN18.2 to be localized throughout the surface of cancer cells (4). Gastric cancer is the third most common cause of cancer-related death worldwide with a median overall survival of approximately 12 months after first-line chemotherapy combinations (8). Response rates in pancreatic cancer are even lower, despite several multiagent chemotherapy regimens, median survival for patients with advanced pancreatic ductal adenocarcinoma (PDAC) remain at less than 12 months (9). The presence of the CLDN18.2 antigen on the surface of gastric/GEJ and pancreas cancer cells provides a window to develop novel therapeutics for these difficult to treat cancers.

The presence of two extracellular loops on the cell surface make CLDN18.2 an attractive target for the development of treatments that utilize these antigens either to generate a therapeutic immune response or specifically deliver cytotoxic chemotherpuetic agents to cells expressing CLDN18.2. While extracellular loop 2 of the protein is highly conserved between the two CLDN18 isoforms (4), a series of unique amino acids on loop 1 allow for development of isoform specific antibodies. A number of CLDN18.2-directed therapeutics are now in clinical development. Currently leading the field is zolbetuximab (IMAB362, Astellas Pharma), a chimeric CLDN18.2 mAb, that has demonstrated significant improvements in progression–free survival (PFS) and overall survival (OS) in combination with mFOLFOX6 [modified folinic acid (or levofolinate), fluorouracil, and oxaliplatin regimen] or capecitabine and oxaliplatin (CAPOX) versus either chemotherapy regimen alone in patients with advanced unresectable or metastatic gastric or GEJ adenocarcinomas that express high levels of CLDN18.2 (10, 11). CLDN18.2-directed chimeric antigen receptor-T (CAR-T) therapies have also entered first in human trials, again with promising signs of efficacy, but with non-target related toxicities such as cytokine-release syndrome that continue to accompany this therapeutic modality (12).

Novel antibody–drug conjugates (ADC) directed against CLDN18.2 have also entered clinical testing, with encouraging responses observed thus far in phase I clinical studies (13, 14). ADCs present an additional option for navigating the therapeutic window afforded by this target. Advances in the field of ADC technology allowing for more stable delivery of a cytotoxic payload have contributed to renewed interest in this therapeutic modality and have resulted in several recent ADC approvals in solid tumors (15–18). Development of high affinity therapeutic mAbs and ADCs for CLDN18.2 may provide additional benefits beyond those observed with zolbetuximab.

In this study, we present data on development and characterization of two novel CLDN18.2-directed therapeutics. A fully humanized, high affinity CLDN18.2-specific mAb and CLDN18.2-specific ADC that offer the potential to be best in class. We evaluate the efficacy of these two molecules in a panel of cell line xenografts (CDX) and patient-derived xenografts (PDX) of gastric and pancreas cancer. In addition, arrays of human tumor tissues were assessed for CLDN18.2 expression using a CLDN18.2 selective IHC antibody to determine incidence and potential patient populations that might benefit from these CLDN18.2-directed therapies.

Mouse mAbs to CLDN18.2 were generated at the Antibody Technology shared resource center, Fred Hutch Cancer Research Center (FHCRC, Seattle, WA) using 20-week-old female Balb-C, CB6F1 and CD-1 mice (Charles River) immunized with peptides spanning loop 1 extracellular domain (ECD) of CLDN18.2 (New England Peptide) and NIH-3T3 cells expressing full-length CLDN18.2. Splenocytes were isolated from high-titer mice and fused with the P3×63 myeloma cell line. The resultant hybridomas were then screened for CLDN18.2 peptide binding, binding to HEK293T cells transfected to overexpress CLDN18.2 and nonengineered cancer cell lines that normally express CLDN18.2. Positive hybridomas were subcloned and antibody sequencing was performed using RNA extracted from hybridoma cells (RNeasy Mini Kit, Qiagen). cDNA was synthesized by using High-Capacity cDNA Reverse Transcription Kit (ThermoFisher Scientific). Antibody sequencing was performed by Laragen Inc, the antibody variable region was annotated with a service from IMGT (http://www.imgt.org/). Antibody sequences were then expressed in ExpiCHO Expression Systems (ThermoFisher Scientific) and purified using Protein A/G Plus Agarose (Thermo Fisher Scientific). The resultant mouse mAbs were screened by flow cytometry for selective binding to CLDN18.2-expressing cells. Antibodies showing selective in vivo efficacy in CLDN18.2⁺ CDX were then selected for humanization (19). The fully humanized immunoglobulin G1 (IgG1) mAb, CLDN18.2–307-mAb, originally developed from the mouse mAb clone (#307), was selected as the lead therapeutic candidate (20). CLDN18.2-307-mAb was formulated in a 29.8 mg/mL in histidine buffer [20 mmol/L histidine, 8% sucrose (w/v), pH 6.0]. The humanized CLDN18.2-307-ADC and Hu-IgG control-ADC were generated by conjugation of CLDN18.2-307-mAb or nontargeting Hu-IgG control antibody with monomethyl auristatin E (MMAE) via a protease labile linker, MC-VC-PAB, to achieve an average drug-to-antibody ratio (DAR) of 4 as described in Supplementary Materials and Methods. Zolbetuximab (IMAB362) was purchased from MedChemExpress (catalog no. HY-P99058).

Human cancer cell lines were maintained in appropriate culture media (e.g., RPMI1640, DMEM, L-15) supplemented with 10%–15% heat-inactivated FBS, 2 mmol l-glutamine and 1% penicillin-G/streptomycin/fungizone solution (PSF, Irvine Scientific). RNA-Sequencing analysis was performed by BGI Americas using the Illumina platform on a panel of 618 cancer cell lines (21). Cell lines were passaged at least three times post thawing before assay, and not passaged more than 12 times during these studies. CLDN18 RNA levels were expressed as Log (FKPM). To generate CLDN18.2- and CLDN18.1-overexpressed cell lines, expression vectors containing the appropriate nucleotide sequence (OriGene Technologies, Inc.) were transduced into either HEK293T or NIH-3T3 cells and selected for survival in medium containing puromycin (1 μg/mL). Each vector also expressed a GFP marker sequence. Positive cells were subcloned to obtain a stable, uniform, clonal population of CLDN18.2-overexpressing cells. Artificially engineered cell lines were maintained in appropriate culture media. Cells were routinely assessed for Mycoplasma contamination using a multiplex PCR method (22), and cell line authentication by short tandem repeat profiling using GenePrint-10 System (Promega).

For indirect flow, cells were incubated with 10 μg/mL primary CLDN18.2 followed by incubation with 6 μg/mL fluorochrome-labeled secondary antibody (BioLegend, Alexa Fluor 647 anti-human IgG Fc). FL4-H values acquired by BD Accuri C6 software were reported. For direct flow, cells were incubated with titrated CLDN18.2 antibodies at starting concentrations of either 10.24 or 5 μg/mL, serially diluted 1 to 2 for a total of 5 or 9 dilution points. After incubation, cells were washed and stained with fluorochrome-labeled secondary antibody (BioLegend, Alexa Fluor 647 anti-human IgG Fc) at 0.5 μL/reaction. For a negative control, cells were incubated with the secondary antibody alone. RL1-H values were acquired and analyzed by iQue flow cytometry (Sartorius). Induction of cellular apoptosis and cell viability assays were conducted as described in Supplementary Materials and Methods. Antibody equilibrium dissociation constants (K_D values) to CLDN18.2 on cell surface were measured using a KinExA 4000 (Sapidyne Instruments Inc.) as described in Supplementary Materials and Methods.

For cell imaging, CLDN18.2 antibodies were primarily labeled with Alexa Fluor 647 (red color) Antibody Labeling Kit (Thermo Fisher Scientific). Cell nuclei were stained with DNA dye Hoechst 33342 (blue; Abcam) at 1 μmol/L final concentration. Yellow color organelles show colocalization of lysosome (green) and CLDN18.2-CLDN18.2-307-ADC (red). Internalization asays were conducted as described in Supplementary Materials and Methods.

The procedures for the antibody-dependent cell-mediated cytotoxicity (ADCC) and antibody-dependent cellular phagocytosis (ADCP) assays are described in Supplementary Materials and Methods.

Tissue sections (4 μm) from formalin-fixed, paraffin-embedded (FFPE) CDX, PDX, and human normal and cancer tissue microarrays (TMA; US Biomax) were used in IHC analyses for CLDN18.2 (Abcam, catalog no. ab222512), Ki-67 antibody (Cell Signaling Technology, catalog no. 9027) or wide-spectrum cytokeratin (pan-CK; Abcam, catalog no. 9377) expression. Signal was detected using EnVision kit HRP-conjugated anti-rabbit secondary antibody (Abcam, catalog no. ab214880). Slides were then counterstained with hematoxylin. CLDN18.2 expression was estimated across all tissue samples using a combined metric assessing percent positive cells in the malignant cell population as well as the intensity of membrane staining. CDXs were scored for membranous CLDN18.2 staining as either 0 (negative), 1 (patchy), 2 (moderate and uniform), or 3 (intense and uniform). Slides were also imaged using Aperio ScanScope AT (Leica Biosystems) for quantification of percent positive cells with moderate to high intensity membranous staining using Definiens Tissue Studio software (Definiens).

CDX and PDX models were established in 6-week-old CD-1 athymic nude mice (Charles River Laboratories). For CDX studies, cell lines were injected subcutaneously at 1.0 × 10⁷ viable cells per mouse in a 50% Matrigel/media mix. When tumors reached an average size of 150–250 mm³, mice were randomized into treatment groups (6–8 mice/group). PDX material was obtained from Pancreas Transplant Center at UCLA as dispersed, cryopreserved cells. Patients consented to use of their tissue for research under Institutional Review Board Protocol # 11–002112. Cells were thawed and injected for in vivo passage in mice until sufficient numbers were established to randomize into treatment groups (4–8 mice/group). Tumor xenografts were measured with calipers 3 times/week, and tumor volumes were determined by multiplying height, width, and length. CLDN18.2–307-mAb and nontargeting Hu-IgG1 control antibody were dosed intravenously by tail vein injection once weekly. Mice were treated weekly for up to 7 weeks and followed postdosing for up to 4 weeks. CLDN18.2–307-ADC was dosed intravenously once weekly by tail vein injection for 3 cycles or for 9 weekly cycles in the case of the XWR3 study. HUPT4 and PATU8998S CDX studies were also treated with Hu-IgG control-ADC intravenously once weekly for three cycles. Data were analyzed using StudyLog software from StudyDirector. For tissue acquisition studies, mice were euthanized at specific time-points and excised tumors stored as FFPE tissue. All animal work was conducted in accordance with and overseen by UCLA's Institutional Animal Care and Use Committee (IACUC). Statistical differences between treatment groups at specific time points were performed using a one-tailed unpaired Student t test. Differences between groups were considered statistically significant at P < 0.05. All statistics were calculated in Microsoft Excel.

Data generated in this study are available within the article and its Supplementary Data files or from the corresponding author upon reasonable request.

CLDN18.2–307-mAb demonstrated selective binding to HEK293T cells engineered to overexpress CLDN18.2 (HEK293T-CLDN18.2) as well as to human cancer cell lines known to express high levels of CLDN18 mRNA. Conversely, it showed no binding to parental HEK293T cells, HEK293T cells overexpressing CLDN18.1 (HEK293T-CLDN18.1) or a CLDN18 mRNA-negative melanoma cell line (Fig. 1A and B; Supplementary Table S1). Distinct binding peaks for CLDN18.2–307-mAb and zolbetuximab were observed in both HEK293T-CLDN18.2 cells and human cancer cell lines that express CLDN18.2 (Fig. 1A and B). Further analyses by direct flow cytometry, using primary labeled antibodies across a range of 10 different concentrations, showed distinct binding curves for CLDN18.2-307-mAb and zolbetuximab in HEK293T-CLDN18.2 cells (Fig. 1C). The higher binding of CLDN18.2-307-mAb over zolbetuximab is even more pronounced in natural expressing HUPT4 and SNU601 human cancer cell lines (Fig. 1D). The binding affinities (K_D) of the two antibodies were measured using the KinExA platform. CLDN18.2-307-mAb (K_D; 0.639 nmol/L, CI 0.336 – 1.14 nmol/L) had a 4-fold higher binding affinity when compared with zolbetuximab (K_D; 2.57 nmol/L, CI 1.33 – 4.68 nmol/L) in HEK293T-CLDN18.2 cells. Moreover, in HUPT4 pancreatic cancer cells that naturally express CLDN18.2, a >1,000-fold higher affinity was observed for CLDN18.2-307-mAb (K_D; 0.551 nmol/L, CI 0.263 – 0.996 nmol/L) over zolbetuximab (K_D; 728.53 nmol/L, CI 332.27 – 2.77 nmol/L; Supplementary Table S2).

Treatment of the CLDN18.2⁺ HUPT4 cells with CLDN18.2-307-mAb induced dose-dependent ADCC activity compared with IgG1 isotype control treatment of the same cells. Each of the three assays using different peripheral blood mononuclear cell (PBMC) donors presented EC₅₀ values ranging from 0.0001 to 0.0061 μg/mL (Fig. 1E; Supplementary Table S3). Treatment of pHrodo Green AM (pHG)-labeled HUPT4 cells with the CLDN18.2–307-mAb in the presence of in vitro–derived human monocytes demonstrated the induction of ADCP by the CLDN18.2-mAb. Dose-dependent induction of phagocytosis, ranging between 5% and 12%, was observed for each of the macrophage donors (Fig. 1F; Supplementary Table S3).

The efficacy of CLDN18.2-307-mAb was assessed in CDXs of CLDN18.2⁺ and CLDN18.2⁻ cancers. CDXs were scored by IHC as either 0 (negative), 1+, 2+, or 3+ for membranous CLDN18.2 staining (Fig. 2A), using a mAb shown to be selective for CLDN18.2 (Supplementary Fig. S1). Treatment with the antibody induced significant tumor growth inhibition (TGI) in two CLDN18.2⁺ CDX models of pancreas cancer (HUPT4, 110.2% TGI and PATU8998S, 58.4% TGI) and in one model of gastric cancer (SNU601, 66.7% TGI). No significant impact on xenograft growth was observed in CLDN18.2-negative (CLDN18.2⁻) M202 melanoma CDX (Fig. 2C; Supplementary Table S4). In the three CDX models tested, CLDN18.2-307-mAb demonstrated equal or superior efficacy to zolbetuximab. In the HUPT4 pancreas cancer CDX, the antibody induced superior growth inhibition when compared with zolbetuximab (110.2% TGI vs. 26.3% TGI; Fig. 2B; Supplementary Table S4). All mAbs were well tolerated in all in vivo studies as determined by body weight measurements (Supplementary Table S4). Although it should be noted that CLDN18.2-307-mAb does not cross-react with mouse CLDN18.2.

The contribution of fc effector function to the in vivo efficacy of CLDN18.2-307-mAb was assessed in HUPT4 xenografts using NK cell–deficient NSG (NOD.CB17-Prkdcscid/NCrCrl) mice. Treatment with CLDN18.2-307-mAb induced only marginal inhibition in xenograft tumor growth during the first week of dosing, after which, efficacy was lost. (Fig. 2D; Supplementary Table S4). This is in contrast with the complete responses observed in NK cell–competent CD-1 nude mice (Fig. 2B, left). These data indicate that in animals, ADCC makes a significant contribution to the mechanism of action of CLDN18.2-307-mAb.

The K_D of CLDN18.2-307-ADC was confirmed to be in the same range as the CLDN18.2 mAb in cells that either naturally or are engineered to overexpress CLDN18.2 (Supplementary Table S2). The ADC also maintains selectivity for CLDN18.2 over CLDN18.1 (Supplementary Fig. S2). Both the mAb and ADC are readily internalized at a similar rate upon binding to CLDN18.2 cell surface protein on HUPT4 pancreatic cancer cells (Fig. 3A). Strong cell surface staining was observed in cells at the earliest time point imaged posttreatment with fluorochrome primary labeled CLDN18.2–307-mAb or CLDN18.2-307-ADC. Internalization of the CLDN18.2 protein (red) and localization to the lysosome (green/yellow) were observed at 4, 24, and 96 hours posttreatment (Fig. 3A).

The potential for CLDN18.2-307-ADC to selectively target and kill CLDN18.2+ cancer cells was assessed in human cancer cell lines using cell viability and apoptosis induction assays. Treatment with CLDN18.2-307-ADC induced a dose-dependent reduction in the number of viable cells in pancreas (HUPT4, IC₅₀ = 0.051 μg/mL) and gastric cell lines (SNU601, IC₅₀ = 0.014 μg/mL; Fig. 3B). Selective induction of cell death/apoptosis was also observed in CLDN18.2⁺ cells lines at 48 and 72 hours posttreatment with the ADC (Supplementary Fig. S3). A nontargeting control ADC and mAb had no impact on proliferation or survival of these cell lines at concentrations below 12.5μg/mL (Fig. 3B; Supplementary Fig. S3). CLDN18.2-307-ADC showed no activity against the CLDN18.2⁻ (M202, IC₅₀ >12.5μg/mL) cells (Fig. 3B; Supplementary Fig. S3).

Treatment with CLDN18.2-307-ADC induced complete tumor regressions in two separate human CLDN18.2⁺ pancreatic xenografts (HUPT4 and PATU8998S). Response to CLDN18.2-307-ADC was distinct from that of the nontargeting hu-IgG control-ADC (Supplementary Fig. S4). No xenograft regrowth was observed for a period of up to 7 weeks postdosing in CLDN18.2-307-ADC–treated mice (Fig. 4A). These results were also reproduced in CLDN18.2⁺ gastric cancer xenografts (SNU601). In contrast, no significant inhibition of tumor growth was observed in a CLDN18.2⁻ melanoma xenograft model (Fig. 4A; Supplementary Table S5).

IHC analysis of tissues collected from HUPT4 pancreas cancer xenografts following treatment with CLDN18.2-307-ADC showed progressive loss of CLDN18.2⁺ cells and significant reduction in expression of the Ki-67 cell proliferation marker and the pan-CK human epithelial cell marker (Fig. 4B and C). Decreases in CLDN18.2 staining were first observed at 3 days following the first dose. Complete loss of CLDN18.2⁺ cells was observed by day 17 (post third dose). The concomitant loss of pan-CK and Ki67 staining in these samples is consistent with loss of xenograft tumor burden in these mice (Fig. 4B and C). Xenograft tissue collected following the third dose of the ADC showed complete loss of human epithelial cells (pan-CK staining) in two of the three samples collected and greatly reduced cell numbers in the third (Fig. 4B and C). To further investigate the potential of CLDN18.2-307-ADC to eliminate xenograft tumor tissue, mice bearing the same CLDN18.2⁺ HUPT xenografts, were treated with a single dose of CLDN18.2-307-ADC and followed for xenograft tumor regrowth. Treatment resulted in a complete loss of xenograft tumor burden at approximately 2 weeks postdosing. No xenograft tumor regrowth was observed during a follow up period of a further 4 weeks (Fig. 4D). Necropsy at end of study, found no trace of xenograft tumor tissue present in these mice. The lack of xenograft tumor regrowth in these animals underscores the potential of CLDN18.2-307-ADC to kill both the CLDN18.2⁺ and any residual CLDN18.2⁻ populations in tumor tissue.

IHC analysis of a panel of 23 human pancreatic cancer PDXs, demonstrated high positive staining (2+/3+) in 6 tissues with another 6 showing lower positive staining (1+; Supplementary Fig. S5). In total, 12 of 23 (52.2%) of the pancreatic cancer PDXs showed some positive staining for CLDN18.2 membrane protein. CLDN18.2-307-ADC efficacy was assessed in a subgroup of these PDXs representing high, low, and negative CLDN18.2 staining (Fig. 5A–C). Significant xenograft tumor regressions were observed in 4 of 4 PDX models with high (2+/3+) CLDN18.2 staining. These responses were sustained without signs of regrowth over a period of 3–6 weeks postdosing (Fig. 5D). Responses to the ADC were also observed in a CLDN18.2 low (1+) model but responses took longer to develop. Four cycles of dosing were required for 1+ positive PDXs to begin to achieve responses similar to that observed in CLDN18.2-high–positive models (Fig. 5E). No significant impact on tumor growth was observed in two different CLDN18.2-negative (0) PDX models (Fig. 5F). For comparison purposes, CLDN18.2 expression in the pancreas PDXs was reevaluated according to the inclusion criteria for zolbetuximab phase III clinical studies (≥75% of tumor cells showing moderate to strong membranous staining; refs. 10, 11). According to these analyses, only one of the PDXs used in these efficacy studies would be classed as a high expressor (XWR23; 81.5% positive cells). However, complete responses were observed in four other xenograft models that have low to moderate CLDN18.2 expression according to these criteria; XWR195 (26.7%), XWR6 (39.5%), XWR8 (46.2%), and XWR3 (15.0%). These data further support our findings that the CLDN18.2-ADC selectively kills CLDN18.2⁺ human pancreas cancer cells and can also achieve meaningful efficacy in CLDN18.2 low–positive cancers. Similar to the CLDN18.2 mAb, the ADC was well tolerated in mice with no dose-limiting toxicities as determined by body weight loss (Supplementary Table S5).

To determine potential patient populations that could benefit from a CLDN18.2-based therapy, we used our CLDN18.2-specific IHC assay to stain >1,000 human cancer tissues across a range of different cancer types (Table 1; Supplementary Fig. S6). In a panel of 166 gastric adenocarcinomas, 95 (58%) show positive staining for CLDN18.2 at some level (≥1+) with 25% showing high-positive (2+/3+) staining. GEJ adenocarcinomas also stained positive for CLDN18.2 in 45 of 75 (60%) cases, with 20 of these showing high positive (2+/3+) staining. In a panel of 354 pancreatic adenocarcinomas, CLDN18.2 staining was detected in 20% (72/354), with strong (2+/3+) staining in 8% of cases (27/354). Outside of gastric and pancreas cancers, the frequencies of CLDN18.2 positivity dropped significantly, with approximately 2% of NSCLC and 3% of colorectal cancer adenocarcinomas staining positive (Table 1). In a panel of 103 normal human tissues representing 29 different organs and tissues, specific staining for CLDN18.2 was only detected in normal stomach tissue (Supplementary Fig. S7).

Gastric, GEJ, and pancreas cancers are diseases of significant unmet medical need. The low survival rates associated with these malignancies are attributed to late-stage disease presentation and high frequencies of resistance to first- and second-line therapies. Overall survival following progression on currently approved therapies remains at less than 12 months for most patients (8, 9). As such the development of novel therapeutic strategies for these cancers is critical. The CLDN18.2 protein is specifically expressed in significant subsets of these cancers and presents an opportunity to use this cell surface antigen as a target to recruit immune responses and/or deliver toxic payloads to the cancer cells. Herein we describe the development and characterization of a novel mAb and ADC that specifically target CLDN18.2.

CLDN18.2-307-mAb is a fully humanized mAb that binds selectively to CLDN18.2 without binding the CLDN18.1 isoform. The selectivity of the mAb for the CLDN18.2 isoform eliminates the possibility for unwanted toxicities from binding to the CLDN18.1 isoform expressed in healthy lung tissue (5). CLDN18.2–307-mAb binds with high affinity to cells that are either engineered to express or natively express CLDN18.2. Our in vitro and in vivo studies show that fc-effector function is a major contributor to the mechanism of action of CLDN18.2-307-mAb. Coculture experiments with human donor PBMCs, show selective and dose-dependent induction of ADCC in CLDN18.2⁺ human cancer cells in response to treatment with the CLDN18.2-307-mAb. Further analysis showed dose-dependent stimulation of phagocytosis by CLDN18.2-307-mAb when monocytes isolated from human donors are cultured with CLDN18.2⁺ cancer cells. These data are supported by xenograft studies that show efficacy of CLDN18.2-307-mAb is lost in NK cell–deficient NSG mice that have limited capacity to mount an ADCC response. This proposed mechanism of action of CLDN18.2-307-mAb is in line with published pharmacodynamic data from clinical studies using zolbetuximab that associate immune effector stimulation with activity (23). Collectively, these data support the hypothesis that although CLDN18.2 may not be a driver of cancer progression, the presence of this antigen on the surface of cancers cells is sufficient for an effective pharmaceutical intervention strategy.

In CD-1 nude mice, treatment with CLDN18.2-307-mAb induces significant TGI of CLDN18.2⁺ human pancreas and gastric CDXs. Our studies show CLDN18.2–307-mAb has equal or superior efficacy to zolbetuximab, a chimeric CLDN18.2 mAb completing phase III clinical development, across multiple models. This increased efficacy is possibly due to a higher binding affinity for CLDN18.2 over zolbetuximab. Both antibodies have binding affinities within range of one another in cell lines artificially engineered to overexpress extremely high levels CLDN18.2; however, in cell lines that natively express CLDN18.2, the CLDN18.2-307-mAb has a >1,000-fold affinity for CLDN18.2 over zolbetuximab. It is possible that when CLDN18.2 is expressed in its natural configuration, zolbetuximab has a lower affinity for the protein. We believe our strategy to include cell lines that naturally express CLDN18.2 in the initial screens for candidate antibodies, increased the possibilty of identifying mAbs with high affinity to this target as expressed in its tertiary structure in actual (nonengineered) human cancer cells. Data reported from recent phase III clinical studies have demonstrated that the addition of zolbetuximab to standard of care (SOC) in metastatic gastric and GEJ adenocarcinomas induces significant improvements in PFS and OS over SOC alone (10, 11). In patients that express high levels of CLDN18.2 (≥75% of tumor cells showing moderate to strong membranous staining), a median PFS of 10.61 months was observed in the zolbetuximab containing treatment group versus 8.67 months in the control group (11). These data are consistent with phase II clinical studies that show efficacy of zolbetuximab either as a single agent or in combination with SOC in patients with high levels of expression of CLDN18.2 (24, 25). In patients with CLDN18.2⁺ rates of 40%–69% in cells, no benefit was observed with addition of zolbetuximab to EOX (24). We believe that a higher affinity antibody could be more effective in targeting CLDN18.2, opening the potential for targeting CLDN18.2 in moderate- to low-expressing cancers and ultimately expanding the patient population that may benefit from this therapy. The fully humanized nature of CLDN18.2-307-mAb may also offer advantages over the chimeric antibody, such as reduced risk of antibody immunogenicity.

Clinical studies with zolbetuximab have shown that it is possible to therapeutically target CLDN18.2 safely with superior efficacy compared with SOC therapies in CLDN18.2⁺ upper gastrointestinal cancers. Despite expression of this target in normal GI tissue, adverse events were limited to grade 2–3 toxicities such as nausea/vomiting in patients treated with zolbetuximab (10, 11). Preclinical studies have shown that in normal gastric mucosa, CLDN18.2 protein expression is restricted to the tight junction area between adjacent cell membranes where it is less accessible to antibody binding (4). These factors are encouraging for the development of more potent therapeutic modalities against this target. Currently there are multiple CLDN18.2-directed bispecific antibodies and CAR T-cell therapies in preclinical and clinical development (26–28). Interim analysis of a study of 37 patients treated with a CLDN18.2-selective CAR T-cell therapy, reported that treatment of CLDN18.2⁺ cancers induced an overall response rate of 48.6%. However, 100% of these patients experienced grade 3 or higher hematologic toxicity, and 95% of patients experienced some level of cytokine release syndrome (CRS; ref. 12). Toxicities such as these are unlikely to be target-related. Instead they are part of their mechanism of action indicating that although clinical efficacy can be achieved, safely dosing this class of therapeutics remains a concern. A number of CLDN18.2-directed ADCs are currently in either preclinical or early-stage clinical development (13, 14, 29). Independent phase I dose escalation studies of two novel CLDN18-ADCs, CMG901 and SYSA1801, have both reported encouraging signs of efficacy and tolerability in CLDN18.2⁺ patients (13, 14). ADCs represent a promising therapeutic strategy for navigating clinical efficacy and safety. Thanks to the development newer linker strategies that control DAR, and use of more effective payloads, ADCs are once again at the forefront of targeted therapy development for cancer. Recent clinical approvals of novel ADCs in breast, gastric, urothelial, and head and neck cancers demonstrate how this therapeutic modality can be safely and effectively dosed in human patients with cancer (15–18, 30). We believe through careful selection of target antigens, such as CLDN18.2, novel ADCs can be developed to treat challenging malignancies such as gastric and pancreas cancer.

In this study, we generated a high-affinity, specific, fully humanized mAb that was also used to develop a CLDN18.2-selective ADC. Selectivity for CLDN18.2 over CLDN18.1 was confirmed with the ADC, as was internalization upon binding to the cell surface target. CLDN18.2-307-ADC selectively inhibits the proliferation and viability of CLDN18.2⁺ gastric and pancreatic cancer cells in vitro and induces rapid and sustained regression of CLDN18.2⁺ gastric and pancreas cancer CDXs and PDXs. Response to treatment was associated with a complete loss of CLDN18.2⁺ cells in xenograft tumor tissue and a significant reduction in cell proliferation (Ki67) and human epithelial (pan-CK) cell markers. The use of a protease cleavable linker (MC-VC-PAB) in the construct of CLDN18.2-307-ADC allows for potential bystander activity against adjacent CLDN18.2⁻ cells in tumor tissue. This was evidenced by the fact that treatment CLDN18.2-307-ADC induces complete and sustained loss of xenograft tumor burden in treated animals. Importantly, tumor regressions were not observed in CLDN18.2⁻ xenografts underscoring its specificity for target-positive cancers.

To investigate the potential for CLDN18.2-307-ADC to provide benefit in a broader population of CLDN18.2-expressing cancers, we screened a panel of pancreas cancer PDXs for response to CLDN18.2-307-ADC. Complete and sustained tumor regressions were observed in 100% of CLDN18.2-high (2+/3+) expressing PDXs. Complete responses were also observed in CLDN18.2-low (1+) pancreas PDXs, which included samples with CLDN18.2⁺ staining in as few as 15% of tumor cells, indicating that the CLDN18.2 ADC may be effective in patient populations that have failed to show significant benefit to zolbetuximab therapy (24). The use of a high-affinity antibody and a cleavable linker are likely contributing to the efficacy of CLDN18.2-307-ADC that is observed in these CLDN18.2 low–expressing models. Furthermore, these data provide encouragement that a CLDN18.2-ADC may be able to make an impact on patient outcomes in pancreas cancer where CLDN18.2-directed therapies such as the mAb, zolbetuximab, have yet to report efficacy.

We assessed CLDN18.2 expression in a broad panel of human cancer TMAs using a CLDN18.2-selective antibody. CLDN18.2⁺ staining was identified in 58% of gastric adenocarcinomas, 60% of GEJ adenocarcinomas, and 20% of pancreas adenocarcinomas. It is our hope that a high affinity CLDN18.2 mAb and ADC can be used to reach the broad population of CLDN18.2⁺ cancers. CLDN18.2-307-mAb and CLDN18.2-307-ADC have completed IND enabling toxicity studies with acceptable pharmacokinetic and toxicity profiles, which has enabled initiation of clinical studies. Published clinical data with an existing chimeric CLDN18.2 mAb (zolbetuximab) are encouraging that an acceptable safety profile is possible when targeting this protein. Given the expression profile of CLDN18.2, it is anticipated that the toxicities associated with dosing CLDN18.2-307-ADC will be consistent with that observed with other clinically approved MMAE-conjugated ADCs (18, 31).

In summary, we have developed two novel molecules that could be used for the treatment of gastric and pancreatic cancers. CLDN18.2-307-mAb is a potentially best in class therapeutic monoclonal that has been shown preclinically to inhibit proliferation of CLDN18.2⁺ pancreas and gastric cancers via the induction of ADCC. CLDN18.2-307-ADC selectively kills CLDN18.2⁺ cancer cells and is also effective in targeting cancers with lower levels of CLDN18.2. Because of the independent mechanisms of action of both molecules, we believe that there is room in the clinical landscape for development of both modalities. CLDN18.2-307-mAb and CLDN18.2-307-ADC are now in separate phase I dose escalation studies for the treatment of CLDN18.2⁺ cancers (NCT05159440 & NCT05156866).

N.A. O'Brien reports other support from TORL Biotherapuetics during the conduct of the study; in addition, N.A. O'Brien has a patent for Patent issued to TORL Biotherpuetics. M.S.J. McDermott reports other support from TORL Biotherapeutics LLC during the conduct of the study; other support from 1200 Pharma LLC outside the submitted work; in addition, M.S.J. McDermott has a patent for Claudin18 Antibodies and Methods of Treating Cancer licensed to TORL Biotherapeutics LCC. J. Zhang reports other support from TORL Biotherapeutics during the conduct of the study; other support from TORL Biotherapeutics outside the submitted work. M. Lu reports other support from TORL Biotherapeutics during the conduct of the study; other support from TORL Biotherapeutics outside the submitted work. B. Hoffstrom reports other support from TORL Biotherapeutics during the conduct of the study; other support from TORL Biotherapeutics outside the submitted work. T. Luo reports other support from TORL Biotherapuetics during the conduct of the study; and I am a stock holder in TORL Bio which is directly associated with this work and then stockholder in 1200Pharma, which is not associated with this work. R. Ayala reports other support from TORL Biotherapeutics and other support from 1200 Pharma outside the submitted work. K. Chau reports other support from Torl Biotherapeutics outside the submitted work. A.M. Madrid reports other support from TORL Biotherapeutics and other support from 1200 Pharma outside the submitted work. T.R. Donahue reports other support from Trethera Corporation outside the submitted work. D.J. Slamon reports non-financial support and other support from BioMarin, grants, non-financial support, and other support from Pfizer, grants, personal fees, and non-financial support from Novartis, personal fees from Eli Lilly, other support from Amgen, other support from Seattle Genetics, other support from 1200 Pharma, and other support from TORL BioTherapeutics outside the submitted work. No disclosures were reported by the other authors.

N.A. O'Brien: Conceptualization, data curation, investigation, methodology, writing–original draft. M.S.J. McDermott: Conceptualization, data curation, investigation, writing–review and editing. J. Zhang: Data curation, investigation, methodology, writing–review and editing. K.W. Gong: Data curation, investigation, methodology, writing–review and editing. M. Lu: Data curation, investigation, methodology, writing–review and editing. B. Hoffstrom: Data curation, investigation, methodology, writing–review and editing. T. Luo: Data curation, investigation, methodology. R. Ayala: Data curation, investigation, methodology. K. Chau: Data curation, investigation, methodology, writing–review and editing. M. Liang: Data curation, investigation, methodology, writing–review and editing. A.M. Madrid: Data curation, investigation, writing–review and editing. T.R. Donahue: Resources, writing–review and editing. J.A. Glaspy: Conceptualization, resources, writing–review and editing. L. Presta: Software, formal analysis, Investigation, Writing–review and editing. D.J. Slamon: Conceptualization, resources, investigation, writing–review and editing.

This study was funded in part through a Sponsored Research Agreement between UCLA and TORL Biotherapeutics LLC.