Preclinical Efficacy of the Antibody–Drug Conjugate CLDN6–23-ADC for the Treatment of CLDN6-Positive Solid Tumors

Gynecologic malignancies, including ovarian and endometrial cancer, significantly impact the health of women worldwide and add to the global burden of cancer. Endometrial cancer is the sixth most common malignancy in women and an estimated 65,950 cases will be diagnosed in 2022 in the United States. There are approximately 19,880 patients who will be diagnosed with ovarian cancer over that same time period and 12,810 are expected to die from their disease, making it the fifth leading cause of cancer-related death in women in the United States (1). Patients with ovarian cancer are most often treated with surgery and a platinum-based chemotherapy regimen. However, most will eventually relapse and develop relatively rapid resistance to subsequent alternative chemotherapy regimens (2, 3). The median progression-free survival of platinum-resistant ovarian cancers is between 3 to 4 months and median overall survival is generally less than a year (2). As such, ovarian cancer has the highest mortality rate among gynecologic malignancies (1, 4). The development of molecular targeted therapies has improved the prognosis for many other solid tumors in recent years; however, the ovarian cancer mortality rate has remained stagnant at approximately 40% since the 1980's (5). Endometrial cancer is rising in both incidence and mortality in the United States (6). For recurrent endometrial cancers, combination chemotherapy with carboplatin and paclitaxel is considered the standard of care, with a median progression-free survival of 13 months and overall survival of 37 months (7). Novel targeted therapeutic strategies are needed for both ovarian and endometrial cancer.

Claudin-6 (CLDN6) is a member of the claudin family of membrane proteins which consists of 27 distinct proteins characterized by four transmembrane domains and two extracellular loops. Various claudin (CLDN) family members are located at tight junctions between epithelial cells, where they are thought to play a critical role in barrier functions. More recently, CLDN6 has also been implicated in some critical intracellular signaling pathways including the YAP1–snail1 axis (8), the ASK1-p38/JNK MAPK secretory signaling pathway (9) and CLDN6 signaling has been shown to activate ERα transcriptional activity in a ligand-independent manner, thereby promoting tumor progression in endometrial cancer (10). CLDN6 is also highly expressed in undifferentiated mouse and human stem cells and is believed to contribute to the tumorigenic potential of human pluripotent stem cell–containing cultures (11, 12). There is a high degree of sequence homology between CLDN6 and other family members. CLDN6 and CLDN9 differ by only 3 amino acids in their extracellular domains (ECD). CLDN3 and CLDN4 are also closely related to CLDN6. Both CLDN3 and CLDN4 are expressed at high levels in multiple normal tissues including pancreas, salivary gland, kidney, adrenal gland, small intestine, colon, and thyroid (13). While there is widespread CLDN6 expression in cells and tissues involved in the early stages of embryonic and fetal development (14), in the adult CLDN6 expression is largely restricted to malignant tissues where it has been implicated in the initiation, progression, and metastasis of some cancers (15, 16). It is reported to be aberrantly expressed in cancers of the ovary, stomach, lung, endometrium, and cervix (17–24). Moreover, in both ovarian and endometrial cancer patients, high levels of CLDN6 expression in malignant tissue is an independent prognostic marker for poorer progression-free and overall survival (23, 25). The composite data demonstrating high cancer versus normal tissue expression in adults makes CLDN6 an attractive target for the development of antibody-based therapeutics.

Antibody–drug conjugates (ADC) are a class of therapeutics that couple cytotoxic payloads to antibodies targeting cancer-associated proteins with the goal of increasing antitumor efficacy while decreasing the systemic toxicity of the payloads. Historically, the clinical success of ADCs has been limited, due to multiple issues including insufficient differential expression of target proteins in cancer versus normal tissue (26), lack of effective ADC internalization (27), modulation of ADC activity by regulators of endolysosomal trafficking, particularly C18ORF8/RMC1 (28), poor drug–antibody ratios (DAR; ref. 29), ineffective cytotoxic payload drugs (30), or poor linker chemistry (31). Overcoming these issues has led to the recent approval of several ADCs including polatuzumab vedotin-piiq (Polivy) for diffuse large B-cell lymphoma, brentuximab vedotin (Adcetris) for Hodgkin lymphoma in combination with chemotherapy, enfortumab vedotin (Padcev) for urothelial cancer, trastuzumab deruxtecan (Enhertu) for HER2+ breast cancer, and sacituzumab govitecan (Trodelvy) for metastatic triple-negative breast cancer (32–36). To date, however, there are no ADCs approved for the treatment of ovarian cancer and only one ADC that had progressed into phase III testing where it failed to meet its primary endpoint (37).

In the current study, we describe the development and preclinical characterization of a novel ADC, CLDN6–23-ADC, comprised of a selective anti-CLDN6 monoclonal antibody linked to monomethyl auristatin E (MMAE) and demonstrate its potential as a treatment for patients whose tumors are CLDN6-positive (CLDN6+).

The Cancer Genome Atlas (TCGA) RNA sequencing (RNA-seq) datasets [presented as fragments per kilobase million (FPKM)], were downloaded from UCSC Xena (RRID:SCR_018938 https://xena.ucsc.edu/; ref. 38). The Genotype-Tissue Expression (GTEx) dataset of normal tissue was downloaded from www.gtexportal.org/home/datasets on May 12, 2022. RNA-seq analysis was performed by BGI using the Illumina platform on a panel of over 500 human cancer cell lines, representing 17 unique malignant histologies. These cell lines were not conditioned to be resistant to chemotherapy. The unique identity of each cell line was confirmed by short tandem repeat profiling and cells were routinely assessed for mycoplasma contamination using a multiplex PCR method and maintained in culture for a maximum of 15 passages. This allowed for selection of CLDN6-high expressing cell line and xenograft models for use in CLDN6–23-ADC efficacy testing.

Briefly, nucleotide sequences encoding human CLDN6, CLDN3, CLDN4, or CLDN9, purchased from OriGene Technologies, Inc. (Maryland, USA) were engineered into bicistronic vectors with a CMV promoter and an attenuated internal ribosome entry site (IRES) of encephalomyocarditis virus. The IRES was located between the gene of interest cDNA and puromycin cDNA. A woodchuck posttranscriptional regulatory element was located downstream of the puromycin cDNA. Each vector also expressed a GFP marker sequence. The expression vectors were virally transduced into either HEK293T or NIH-3T3 cells and selected for survival in medium containing puromycin (1 μg/mL). The positive cells were subcloned to obtain a stable, uniform, clonal population of overexpressing cells.

Mouse mAbs to CLDN6 were generated by immunizing 20-week-old female Balb-C and CD1 mice with peptides spanning loop 2 ECD of CLDN6 (New England Peptide, Massachusetts) and NIH3T3 cells expressing CLDN6 by the Antibody Technology shared resource at Fred Hutch Cancer Research Center under a services agreement. Splenocytes were isolated from high-titer mice and fused with the P3×63 myeloma cell line. The resultant hybridomas were then screened for CLDN6 peptide binding and binding to HEK293T-CLDN6 overexpressing cells. Positive hybridomas were subcloned and antibody sequencing was performed using RNA extracted from hybridoma cells by RNeasy Mini kit (Qiagen, Maryland); the cDNA was synthesized using High-Capacity cDNA Reverse Transcription Kits (Thermo Fisher Scientific, Massachusetts). Antibody Sequencing was performed by Laragen Inc (California); the antibody variable region was annotated with a service from IMGT (http://www.imgt.org/). Antibodies were purified using Protein A/G Plus Agarose (Thermo Fisher Scientific, catalog no. 20424). Twenty-one unique mouse monoclonal antibodies were identified, expressed with a mouse IgG2A constant region and screened by flow cytometry for binding to CLDN6, and selectivity for CLDN6 was determined by screening against other CLDN family members using cell line models artificially overexpressing related claudin family members (Supplementary Table S1). Antibodies showing in vivo efficacy in CLDN6+ models with limited or no efficacy in a CLDN6-negative (CLDN6−) model were subsequently humanized (39). The CLDN6–23-ADC was generated by conjugation of CLDN6–23-mAb with MMAE via a protease labile linker, MC-VC-PAB by WuXi Biologics under a contract services agreement.

Direct flow assays were performed using CLDN6 antibody labelled with Alexa Fluor 647 [ThermoFisher Scientific (catalog no. A20006)] and in-direct flow assays were performed using primary antibody and fluorochrome-labeled secondary antibody (BioLegend, Alexa Fluor 647 anti-human IgG Fc, 0.3 mg/mL stock, catalog no. 409320) using the BD Accuri C6 where FL4-H values acquired by BD Accuri C6 software were reported. Using labelled primary antibodies, cells were imaged with Keyence fluorescence microscope (BZ-X710) to confirm the cell surface staining in native CLDN6 expressing cells and/or co-localization of mGFP and CLDN6 staining in artificial overexpression cells.

Antibodies were primarily labeled with Alexa Fluor 647 Antibody Labeling Kit (Thermo Fisher Scientific, catalog no. A20006) and using Ibidi USA μ-SLIDE 4 Well chamber (Thermo Fisher Scientific, catalog no. NC0515977) cells were transfected with CellLight Lysosomes-GFP, BacMam 2.0 (Thermo Fisher Scientific, catalog no. C10507), and Hoechst 33342 DNA dye (Abcam, catalog no. ab228551) following the manufacturer's instructions. Cells were stained with 25 μg/mL of primary labelled antibody and a time course of images were taken until internalization was complete using the Keyence fluorescence microscope (BZ-X710).

Western blotting was performed as previously described (40), using antibody targeting CLDN6 (Cell Signaling Technology, Massachusetts, #62831), Beta-actin (Cell Signaling Technology, #3700), and HRP-conjugated goat anti-rabbit secondary antibody (Cell Signaling Technology, #7074). Signal visualization was obtained using enhanced chemiluminescent substrate (GE Amersham) and the Fluro Chem Q imaging system (Protein Simple, California).

Antibody-dependent cellular cytotoxicity (ADCC) reporter assays were performed according to the manufacturer's protocol (Promega, catalog no. G7010). Briefly, 1×10⁴ cells/well/100 μL target cells were cultured in flat-bottom, optical polymer based 96-well plates (Nunc, Thermo Fisher Scientific, catalog no. 152028). After 24 hours, CLDN6–23-mAb antibody or nonspecific human IgG as a control were added to each well at the final concentrations of 10 μg/mL followed by three-fold serial dilutions. Effector cells were added at a ratio of 7.5:1 and following a 6-hour incubation, Bio-Glo luciferase assay substrate reagent was added and relative luminescence unit (RLU) values were measured using a Varioskan LUX (Thermo Fisher Scientific). RLU versus Log₁₀ [antibody concentration] were graphed and EC₅₀ of antibody response were calculated using curve fitting software (https://www.aatbio.com/tools/ec50-calculator).

OVCA429 and M202 cells were seeded overnight at ∼3,000 cells/well into 96-well plates after 24 hours treated with CLDN6–23-ADC or human IgG1 ADC (50 μg/mL to 0.014 μg/mL) in duplicate. After incubation for ∼65 hours, cell viability was determined by measuring the luminescence using CellTiter-Glo Luminescent Cell Viability Assay (Promega, catalog no. G7572). Data were expressed as average percentage of the cell viability of untreated cells.

To evaluate efficacy, cell line xenograft models were established through injection of human cancer cell lines (1 × 10⁷ cells per mouse) in 6-week-old CD-1 athymic nude mice (Charles River Laboratories, RRID:IMSR_CRL:086). When tumors reached an average size of 150 to 300 mm³, mice were randomized into treatment groups of 5 to 8 mice per group. For treatment, CLDN6 antibodies were diluted in sterile saline and given at 10 mg/kg once every 4 days or once weekly by intravenous tail vein injection. CLDN6–23-ADC, diluted in sterile saline, was dosed at 2.5 to 5 mg/kg once weekly intravenously. For controls, mice were treated with 10 mg/kg of non-targeting Hu-IgG1 control antibody (produced in house) once weekly intravenously. Tumor xenografts were measured with calipers 3 times/week, and tumor volumes were determined in mm³ by multiplying height, width, and length. For tissue acquisition studies, xenograft tumor tissue was excised, and stored as formalin fixed, paraffin embedded (FFPE) material. Experiments using NOD/SCID gamma (NSG) mice (Charles River Laboratories, RRID:IMSR_CRL:394), were conducted using the same conditions as described above. All animal work was conducted under a protocol approved by Institutional Animal Care and Use Committee (IACUC) and the UCLA Animal Research Committee. Data was analyzed using StudyLog software from StudyDirector (San Francisco, CA). Results are presented as mean volumes for each group. Error bars represent the SEM. Statistical differences between mean tumor volumes were performed using two-way repeated measures ANOVA together with a post hoc pairwise t test for comparison between treatment groups with Bonferroni to adjust for multiple comparisons, using the rstatix package in R (https://cran.r-project.org/web/packages/rstatix/index.html). Differences between groups were considered statistically significant at P < 0.05.

The efficacy of CLDN6–23-ADC was assessed in firefly-luciferase-expressing patient-derived xenograft (PDX) models of CLDN6+ and CLDN6− ovarian cancers established at the Dana-Farber Cancer Institute, Boston, Massachusetts (41). Female NSG mice (8 to 10 weeks old, Jackson labs, RRID:IMSR_JAX:005557) were orthotopically implanted through intraperitoneal injection of 5 × 10⁶ cells per mouse. PDX tumor burden was monitored by measuring bioluminescence using an IVIS Lumina II, 15 minutes after intraperitoneal injection of 150 μL D-luciferin at 15 mg/mL as previously described (41). Mice were randomized into treatment groups of six per treatment arm based on bioluminescence measurements. Treatment was started 2 weeks postimplantation by intravenous tail vein injection once weekly for 5 weeks and tumor burden was measured weekly until study termination. Studies were conducted through IACUC-approved animal protocols in accordance with Harvard Medical School institutional guidelines.

Tissue microarrays (TMA) assembled from FFPE tissues were purchased from US Biomax, Inc., and were immunostained for CLDN6. US Biomax, Inc. catalog numbers for the purchased ovarian cancer arrays are OV6161, OV2001b endometrial cancers are BC09012b, EM1021a and normal tissue are BN1021. The two-step IHC procedure was performed using a mouse monoclonal primary antibody specific to CLDN6 protein (58–4B-2 produced from ref. 42). The secondary antibody used was a horseradish peroxidase labeled polymer conjugated to goat anti-mouse immunoglobulins (Agilent, #K400111–2). Heat-induced epitope retrieval was performed in a solution containing ethylenediaminetetraacetic acid buffered to pH9 and heated to 99°C. Tissue sections were counterstained with methyl green and DAB immunoprecipitate present in tumor cell membranes was interpreted as positive immunostaining. The relative intensity and frequency of immunostaining was recorded.

UMUC4 xenograft tumor tissue was removed on day 3 or day 11 from animals treated on day 0 and/or day 6 with a control IgG, naked antibody, or CLDN6–23-ADC. These tumor samples were processed for routine histology and stained with hematoxylin and eosin (H&E) for assessment of tumor histology. Microscopic evaluation of the stained tumors included the determination of the UMUC4 tumor cellularity as a percentage of the viable cellular components of each excised mass. Residual basement membrane matrix (Corning, #354234), adipocytes, murine endothelial cells, and connective tissues were the primary components identified in addition to the implanted UMUC4 tumor cells. The distribution of viable cells from each component was determined, and the percentage composed of viable UMUC4 tumor cells was recorded.

The data generated during the current study are available within the article and its Supplementary Data files or from the corresponding author on reasonable request.

An analysis of publicly available mRNA data from TCGA was performed to identify antigens as potential suitable targets for the development of antibody-based therapeutics. CLDN6 displayed a highly diverse expression pattern across the 30 cancer histologies comprising the dataset. While there was limited to no expression in most cancers, CLDN6 was highly expressed (Log FPKM ≥ 10) in 89.7% of testicular cancers, 45.7% of ovarian cancers, 31.5% of uterine carcinosarcomas, and 21.8% of endometrial cancers (Fig. 1A). High CLDN6 expression was also detected in a subset of non–small cell lung cancer (8.3%) and stomach cancer (6.2%). Conversely, the GTEx dataset of normal tissue reveals little or no detectable expression of CLDN6 in any of the multiple adult tissue types analyzed, suggesting that CLDN6 is a cancer-specific antigen (Supplementary Fig. S1). Multiple cancer cell line models of CLDN6 expression, including several ovarian cancer cell lines (Fig. 1B), were identified from an RNA-seq analysis of a panel consisting of over 500 human cancer cell lines, representing 17 unique malignant histologies, and confirmatory Western blotting analyses for CLDN6 were performed in cell lines displaying the full range of CLDN6 mRNA levels. High protein expression of CLDN6 was detected in multiple ovarian cancer cell lines including OVCA429, OAW28, and moderate expression in OV90 and PE014 cells. Very high expression of CLDN6 was also detected in ARK2 (endometrial) and UMUC4 (bladder) cancer cell lines while M202 (melanoma) and MCF7 (breast) cancer cell lines which were negative for CLDN6 mRNA, were also negative for CLDN6 protein (Fig. 1C).

The CLDN6 protein consists of 220 amino acids with two ECD loops. Sequence alignment with CLDN9 (a paralog of CLDN6) reveals that while loop 1 is almost identical, there is a two amino acid difference in the sequence of loop 2 (Supplementary Fig. S2). To generate specific anti-CLDN6 monoclonal antibodies that do not bind to CLDN3, CLDN4, or CLDN9, mice were immunized with peptides spanning loop 2 (ECD, aa 138–160) as well as NIH-3T3 cells engineered to express the full-length CLDN6 protein on the cell surface. The resultant hybridomas were screened for CLDN6 binding, subcloned, sequenced, tested for selective in vivo growth inhibitory activity in CLDN6+ and CLDN6− xenograft models (Supplementary Fig. S3A–S3C), and then humanized. Following humanization, antibody #23 (CLDN6–23-mAb) was identified as having several ideal properties including: optimal binding to CLDN6 on the cell membrane of multiple CLDN6+ cell lines (Fig. 2A) and low levels of aggregation or degradation during and after temperature and freeze/thaw stress testing (data not shown). In addition, CLDN6–23-mAb exhibits selectivity for CLDN6 over other CLDN family members, as indicated by high binding to both artificial and endogenous CLDN6-overexpressing cells compared with cells overexpressing CLDN9, CLDN3, or CLDN4 (Fig. 2B).

To determine the therapeutic potential and specificity of CLDN6–23-mAb, we evaluated its in vivo growth inhibitory activity against CLDN6+ and CLDN6− xenograft models. Treatment with CLDN6–23-mAb resulted in significant tumor growth inhibition in mice bearing bladder cancer UMUC4 (CLDN6+, high expression) xenografts compared with IgG control and this response was sustained until study termination (36 days; Fig. 3A). Efficacy was observed in additional CLDN6+ xenograft models treated with CLDN6–23-mAb. Significant tumor inhibition was observed in the small cell lung cancer H841 (CLDN6+, high expression) xenografts with more modest inhibition in ovarian OV90 xenografts which exhibit moderate CLDN6 expression (Fig. 3B and C). Mice bearing melanoma M202 xenografts that are negative for CLDN6, did not exhibit any growth inhibition in response to CLDN6–23-mAb treatment indicating no off-target effect of the antibody (Fig. 3D). CLDN6–23-mAb was well tolerated in all in vivo studies as determined by body weight measurements (Supplementary Table S2).

Clinical trials have demonstrated the importance of ADCC as a mechanism of action of several monoclonal antibodies (43). Using an in vitro ADCC reporter assay, we established that CLDN6–23-mAb induces ADCC in CLDN6+ ARK2 cells but not in CLDN6− M202 cells (Fig. 3E). The NSG mouse is a valuable tool for in vivo ADCC exploration studies as these mice are B-cell, T-cell, and natural killer (NK)-cell deficient and, therefore, incapable of mounting an effective ADCC-based response. CLDN6+ UMUC4 xenografts were established in NSG mice and following treatment with CLDN6–23-mAb modest tumor growth inhibition was measured. In contrast to the significant tumor growth inhibition observed with UMUC4 xenografts grown in NK-competent mice, there was no significant effect on their growth in NSG mice (Fig. 3A and F) suggesting that induction of ADCC is an important component mechanism of action for growth inhibition with CLDN6–23-mAb. In addition, fluorescence microscopy analyses revealed that CLDN6–23-mAb induced rapid internalization of the CLDN6/CLDN6–23-mAb complex in CLDN6+ cancer cells. Strong CLDN6 membrane staining was observed at time 0 in ARK2 cells (Fig. 3G) and robust antibody-mediated internalization of CLDN6 was observed after approximately 3 hours. Internalized CLDN6 was colocalized with lysosomal-associated membrane protein 1 (LAMP1) indicating that intracellular trafficking of CLDN6 to the lysosome occurs with CLDN6–23-mAb, supporting its potential for conjugation to a cytotoxic payload to generate an ADC.

CLDN6–23-ADC was generated by conjugating MMAE to the anti-CLDN6 antibody via a VC-PAB linker resulting in an ADC with a DAR of 4.1. Both the unconjugated CLDN6–23-mAb and CLDN6–23-ADC had similar antigen-binding kDs of 2.33 and 2.32 nmol/L, respectively, as determined by KinExA (Supplementary Table S3), and exhibited similar specificity for CLDN6+ versus CLDN6− cancer cell lines in flow cytometry analyses (Fig. 4A) as well as clear selectivity for CLDN6 versus other CLDN family members (Supplementary Fig. S4). These data indicated that the conjugation process did not adversely affect CLDN6 antigen recognition. Colocalization with LAMP1 was observed in CLDN6+ cancer cell lines following treatment with CLDN6–23-ADC in both ARK2 and OVCA429 CLDN6+ cancer cell lines (Fig. 4B) at a similar rate to that of the parent CLDN6–23-mAb (5 and 3 hours, respectively, in ARK2 cells). A dose-dependent in vitro antiproliferative response to CLDN6–23-ADC was observed in the CLDN6+ OVCA-429 cells (EC₅₀ = 15.96 nmol/L; Fig. 4C). In contrast, CLDN6–23-ADC treatment of CLDN6− M202 cells resulted in minimal effects, similar to that observed with a nonbinding control ADC containing an identical linker payload (human IgG ADC) in both cell lines (EC₅₀ > 333.33 nmol/L in all three cases). A similar dose-dependent induction of apoptosis/necrosis was observed, following 48 hours of CLDN6–23-ADC treatment in OVCA-429 cells, while minimal effects were observed with control ADC or M202 cells compared with untreated controls (Supplementary Fig. S5).

To assess the in vivo efficacy of CLDN6–23-ADC, mice bearing CLDN6+ UMUC4 xenografts were treated with either CLDN6–23-mAb or CLDN6–23-ADC once weekly for 3 weeks. A comparison of tumor growth at day 35 showed that treatment with the unconjugated antibody resulted in significant growth inhibition compared with IgG control, while CLDN6–23-ADC treatment led to significant tumor regressions compared with IgG control (Fig. 4D). The CLDN6–23-ADC regressions were sustained through an additional extended period of observation despite the mice having received only 3 doses of the ADC. A histopathologic assessment of xenograft tumor samples collected from a duplicate study showed that treatment with either CLDN6–23-mAb or CLDN6–23-ADC resulted in a decreased overall tumor size compared with the normal IgG-treated control animals. The amount of viable tumor cells was assessed in the tumor samples from each treatment condition. This analysis showed markedly diminished viable cells in both CLDN6–23-mAb or CLDN6–23-ADC–treated xenografts (Fig. 4E). At day 3 posttreatment, 83% of control tumors was composed of viable tumor cells whereas the tumor cellularity of CLDN6–23-mAb and CLDN6–23-ADC–treated tumors was 25% and 24%, respectively. However, this difference in tumor cellularity was even more pronounced at day 11 of treatment where control and CLDN6–23-mAb tumors contained 78% and 72% viable tumor cells, respectively, while CLDN6–23-ADC–treated tumors contained <1% viable tumor cells. In addition, in a separate study a single dose of 1.0 mg/kg of CLDN6–23-ADC induced similar efficacy as a 10 mg/kg dose of the naked CLDN6–23-mAb. As such, it is reasonable to suggest that CLDN6–23-mAb is at least 10-fold less potent than the CLDN6–23-ADC and thus ADCC, a proposed mechanism of action of CLDN6–23-mAb, is likely to contribute less than 10% (1/10) to the efficacy of CLDN6–23-ADC (Supplementary Table S4).

The efficacy of CLDN6–23-ADC was further investigated in CLDN6+ UMUC4 (bladder cancer), ARK2 (endometrial cancer), and OV90 (ovarian cancer) cell line xenografts. Xenografts were treated once weekly for a total of 3 weeks. Tumor volumes were measured twice weekly and a comparison between IgG control and ADC-treated arms was performed at the point which the IgG control arm was terminated due to high tumor burden: day 43, day 25, and day 32 for UMUC4, ARK2, and OV90, respectively. Significant tumor regressions were observed in UMUC4, ARK2, and OV90 xenografts following treatment with CLDN6–23-ADC (Fig. 5A–C; Supplementary Table S5). In all three CLDN6+ models, the tumor regressions resulting from treatment with CLDN6–23-ADC were continually assessed beyond the 21-day treatment period and were sustained until study termination at days 189, 166, or 123 days for the UMUC4, ARK2, and OV90 studies, respectively. Treatment with CLDN6–23-ADC had minimal impact on the growth of CLDN6− M202 xenografts (Supplementary Fig. S6). CLDN6–23-ADC was well tolerated in all in vivo studies as determined by body weight measurements (Supplementary Table S5). To further confirm the therapeutic potential of CLDN6–23-ADC in human ovarian cancers, a panel of ovarian PDXs was examined for CLDN6 expression. From this panel, two models were selected that were CLDN6+ (DF-149 and DF-181) and one that was CLDN6− (DF-20; Fig. 5D). These PDX models were established as orthotopic xenografts in NSG mice via intraperitoneal injection of PDX cells and treated with either CLDN6–23-mAb or CLDN6–23-ADC. Consistent with prior results, response to the CLDN6 monoclonal antibody in NSG mice was limited. In contrast, treatment with CLDN6–23-ADC resulted in a robust decrease in tumor volume in the CLDN6+ PDX models (Fig. 5E). This decrease was accompanied by prolongation of survival, with all CLDN6–23-ADC—treated animals surviving throughout the entire study observation period (100 days). Conversely, as the CLDN6− PDX (DF-20) study progressed, the CLDN6–23-ADC–treated PDX exhibited a brief initial response, followed by an increase in tumor volume similar to the CLDN6–23-mAb or to IgG control-treated tumors, and almost all animals either succumbed or had to be euthanized due to tumor progression by day 100 (Fig. 5F).

The mRNA expression of CLDN6 detected in ovarian and endometrial cancers in the TCGA dataset (Fig. 1A) suggests that significant proportions of patients with ovarian and endometrial cancer are likely to express high levels of CLDN6 protein and potentially benefit from a CLDN6-directed targeted therapy. To assess the frequency and relative expression of CLDN6 in ovarian and endometrial cancer, patient samples were examined using IHC analyses in human TMAs. Samples were scored as either negative, weakly positive, moderately positive, or strongly positive in terms of membrane immunostaining intensity for CLDN6 expression as well as percentage of tumor cells stained. Using this approach, the overall CLDN6 positivity was estimated to be 29% among all ovarian epithelial carcinomas and 11% among all endometrial cancers (Table 1; Supplementary Fig. S7). Analysis by ovarian cancer histologic subtype revealed the highest CLDN6 positivity rate (45%) was observed in high-grade (grade 2/3) serous ovarian carcinomas (Table 1). CLDN6 expression was also assessed by IHC in normal human tissues using a TMA consisting of samples from the breast, bone marrow, cerebrum, colon, esophagus, kidney, liver, lung, lymph node, ovary, pancreas, prostate, rectum, skin, small intestine, spleen, stomach, testis, and thymus. No positive immunostaining was observed in any of these normal human tissue samples (Supplementary Fig. S8). Together these data demonstrate that a clinically significant subset of ovarian and endometrial cancers express CLDN6.

Developing therapeutic compounds that selectively and potently inhibit cancerous tissue while sparing normal tissue has long been the goal of targeted therapeutic drug development in oncology. ADCs were developed as a way of delivering potent cytotoxic drugs using a targeted system (antibody binding to a target antigen), as a means of preventing systemic toxicity that occurs when using standard parenteral or oral routes of administration employed for most chemotherapy regimens. Several factors are critical for the development and clinical utility of an ADC including the specificity of the antibody, the degree of expression of the target protein in both cancer and normal tissues, and the internalization of the protein–drug complex (44). In silico analyses using databases of cancer and normal tissues cross-referenced against an in-house human cancer cell line panel, revealed overexpression of CLDN6 in ovarian, endometrial, and a subset of lung cancers compared with normal tissues. Moreover, CLDN6 expression has been associated with worse progression-free and overall survival rates in both ovarian and endometrial cancer patients (23, 25). This suggests that CLDN6 could be an ideal target for the generation of antibody-based therapies and may be particularly well suited for the development of an ADC.

Following an antibody immunization and screening campaign, the monoclonal antibody CLDN6–23-mAb was identified as a lead candidate therapeutic antibody due to high-affinity binding to CLDN6, specificity for CLDN6 versus related CLDN family members (CLDN3, 4, and 9), a high degree of internalization, and robust and highly selective efficacy in CLDN6+ xenograft models as compared with CLDN6− models. The efficacy of some monoclonal antibodies is proposed to be due to inhibition of a direct biologic affect, such as blocking an activation signal necessary for proliferation and/or progression. Alternatively mechanisms based on induction of an antitumor immune effect via mechanisms such as ADCC and CDC may be critical to response (45). The observed in vivo efficacy of CLDN6–23-mAb is likely due predominately to the induction of ADCC as the antitumor effects of the monoclonal antibody were significantly diminished when tested in an ADCC-incompetent NSG model. Although the monoclonal antibody may not have direct biologic activity against cancer cells, it serves as an ideal delivery system for a cytotoxic payload given its robust and selective binding to CLDN6 and subsequent internalization of the CLDN6/CLDN6–23-mAb complex.

When conjugated to MMAE, via a VC-PAB linker, the resultant ADC, CLDN6–23-ADC, exhibited a similar degree of high affinity binding to CLDN6; retained its selectivity versus CLDN3, 4, and 9; and exhibited a similar degree of internalization in CLDN6+ models as the parent CLDN6–23-mAb. MMAE was selected as a payload for ADC conjugation as its clinical efficacy and toxicity profiles have been well established with several approved ADCs (i.e., brentuximab vedotin, polatuzumab vedotin, enfortumab vedotin). Due to the absent expression of CLDN6 in normal tissue, clinical toxicities beyond those previously reported for MMAE-containing ADCs such as peripheral neuropathy, hematologic toxicities, hepatotoxicity, myelosuppression, ocular disorders, and skin reactions, are not expected (46, 47). In vivo studies confirmed that CLDN6–23-ADC inhibited the growth of CLDN6 expressing UMUC4 xenograft models with much greater efficacy than was observed with the unconjugated antibody, CLDN6–23-mAb. Histologic analysis of representative samples of CLDN6+ xenograft tumors showed robust early tumor cell killing in response to both CLDN6–23-mAb and CLDN6–23-ADC treatment. However, xenograft tumor progression was ultimately observed in the mAb-treated mice but not in the ADC-treated animals. Sustained tumor regressions were observed in multiple CLDN6+ xenograft models, including ovarian and endometrial cancers, but not in the CLDN6− xenograft model highlighting the selectively of the response profile. These regressions were consistently maintained far beyond the 21-day treatment period, suggesting that the responses to CLDN6–23-ADC are robust and durable. Similarly, mice bearing ovarian PDXs treated with CLDN6–23-ADC showed a clear survival benefit after 100 days compared with control-treated mice. CLDN6–23-ADC was well tolerated in all in vivo studies as determined by body weight measurements. In addition, Investigational New Drug (IND) nonclinical pharmacology and toxicology studies of CLDN6–23-ADC have been completed in mice and cynomolgus monkeys to support clinical dosing in humans. The results of these IND studies are consistent with that of other MMAE-containing ADCs and support the clinical development of CLDN6–23-ADC.

Prior trials of ADCs in ovarian cancer highlight the important role that appropriate patient selection plays in drug development. Mirvetuximab soravtansine, an anti-folate receptor (FRα) ADC, is the only ADC to date to be evaluated in a phase III trial in ovarian cancer. This trial, FORWARD 1, failed to meet its primary end point as phase III response rates were lower than anticipated compared with prior phase I/II data (24% versus 46% in the high FRα groups in these trials, respectively). This failure could potentially have been due to difficulties in reliably determining high FRα expression status in patients. Inaccurate testing can inadvertently allow for enrollment of patients with lower levels of FRα than required for the antitumor effect (37). This highlights the need for a robust, specific, and reproducible selection marker coupled with a reliable test that allows for accurate determination of the correct potential treatment population. We have validated an antibody for selective and sensitive IHC determination of CLDN6 expression and found an estimated CLDN6 positivity rate of 29% in ovarian epithelial cancers, 45% for grade 2/3 serous carcinomas, and an overall 11% rate in endometrial carcinoma samples, respectively, using cancer TMAs, corroborating earlier reports of CLDN6 overexpression in cancer (19, 23). Using IHC analyses on normal TMAs, we also confirmed that normal tissues do not express detectable CLDN6 levels suggesting a wide therapeutic window for clinical studies of an anti-CLDN6 therapeutic.

Recently, CLDN6 gained additional traction as a potential therapeutic target for cancer development based on preliminary results from a phase I study exploring a CLDN6 CAR-T (combined with an RNA vaccine) showing promising early efficacy with patients with CLDN6+ testicular and ovarian cancer (48). However, the durability of the response was limited, and as anticipated, 50% of patients exhibited signs of cytokine release syndrome. In addition, AMG 794, an anti-CD3/anti-CLDN6 bispecific antibody has entered phase I clinical trials for CLDN6+ cancer (NCT05317078). CLDN6 was also explored as a clinical target with ASP1650, a chimeric CLDN6 monoclonal antibody that showed limited evidence of antitumor activity with an ORR of 2.4% (1/41 patients with partial response; no complete response) in patients with ovarian cancer (49). There was no efficacy observed in the phase II testicular cancer trial where the ORR was 0% (0/13 patients with complete or partial response) and this led to early discontinuation of the study (50, 51). In the current preclinical study, we show that targeting CLDN6 with an ADC is more effective at inhibiting tumor growth than a humanized monoclonal antibody and anticipate improved clinical outcomes over what was seen with a chimeric monoclonal antibody such as ASP1650.

In summary, CLDN6–23-ADC is a promising therapeutic for CLDN6+ ovarian and endometrial cancers and has potential utility as a targeted therapy for other CLDN6+ tumors using IHC analyses as a selection criteria regardless of histology. On the basis of these promising preclinical efficacy results, a first in human trial to evaluate safety, tolerability, pharmacokinetics, and antitumor activity of CLDN6–23-ADC has been launched in patients with advanced cancer including ovarian, endometrial, and advanced solid tumors and is currently ongoing (NCT05103683).

M.S.J. McDermott reports other support from TORL Biotherapeutics LLC during the conduct of the study, as well as other support from 1200 Pharma LLC outside the submitted work; in addition, M.S.J. McDermott has a patent for Claudin-6 Antibodies and Drug Conjugates licensed to TORL Biotherapeutics LLC and a patent for Claudin-6 Bispecific Antibodies licensed to TORL Biotherapeutics LLC. N.A. O'Brien reports other support from TORL Biotherapuetics during the conduct of the study, as well as other support from 1200 Pharma outside the submitted work; in addition, N.A. O'Brien has patents pending relevant to work with CLDN6. B. Hoffstrom reports other support from TORL Biotherapeutics during the conduct of the study, as well as other support from TORL Biotherapeutics outside the submitted work. K. Gong is a shareholder in TORL Biotherapeutics LLC. M. Lu reports other support from TORL Biotherapeutics during the conduct of the study, as well as other support from TORL Biotherapeutics outside the submitted work. J. Zhang reports other support from TORL Biotherapeutics during the conduct of the study, as well as other support from TORL Biotherapeutics outside the submitted work. T. Luo reports other support from TORL Biotherapuetics during the conduct of the study, as well as other support from 1200 Pharma outside the submitted work; in addition, T. Luo has patents pending relevant to work with CLDN6. M. Liang reports other support from TORL Biotherapeutics during the conduct of the study; in addition, M. Liang has a patent for CLDN6 Diagnostic Antibody pending. W. Jia reports other support from TORL Biotherapeutics during the conduct of the study. J.J. Hong reports other support from TORL Biotherapeutics LLC during the conduct of the study. K. Chau reports other support from TORL Biotherapeutics LLC during the conduct of the study, as well as other support from 1200 Pharma outside the submitted work. M.F. Press reports grants from TORL Biotherapeutics LLC and Dr. Miriam & Sheldon G. Adelson Medical Research Foundation during the conduct of the study, as well as other support from Biocartis SA, Eli Lilly & Company, Zymeworks, Novartis Pharmaceuticals, AstraZeneca, and Angle outside the submitted work. D.J. Slamon reports nonfinancial support and other support from BioMarin; grants, nonfinancial support, and other support from Pfizer; grants, personal fees, and nonfinancial support from Novartis; personal fees from Eli Lilly; and other support from Amgen, Seattle Genetics, TORL Biotherapeautics, and 1200 Pharma outside the submitted work. No disclosures were reported by the other authors.

M.S.J. McDermott: Conceptualization, data curation, formal analysis, writing–original draft, writing–review and editing. N.A. O'Brien: Conceptualization, data curation, formal analysis, writing–review and editing. B. Hoffstrom: Data curation, formal analysis. K. Gong: Data curation, formal analysis. M. Lu: Data curation, formal analysis. J. Zhang: Data curation, formal analysis. T. Luo: Data curation. M. Liang: Data curation. W. Jia: Data curation. J.J. Hong: Data curation. K. Chau: Data curation. S. Davenport: Data curation, formal analysis. B. Xie: Data curation. M.F. Press: Data curation, formal analysis, writing–review and editing. R. Panayiotou: Data curation, formal analysis. A. Handly-Santana: Data curation, formal analysis. J.S. Brugge: Formal analysis. L. Presta: Formal analysis. J. Glaspy: Conceptualization. D.J. Slamon: Conceptualization, formal analysis, writing–review and editing.

This study was funded in part through a Sponsored Research Agreement between UCLA and TORL Biotherapeutics LLC. M.F. Press and D.J. Slamon are supported by a grant from the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation. M.F. Press is also supported by an endowed chair, the Harold E. Lee Chair for Cancer Research. We thank Pearl Moharil for technical assistance with the PDX drug studies.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.