Therapeutic antitumor efficacy of monoclonal antibody against Claudin‐4 for pancreatic and ovarian cancers

Abstract

Claudin‐4 (CLDN4) is a tetraspanin transmembrane protein of tight junction structure and is highly expressed in pancreatic and ovarian cancers. In this study, we aimed to generate an anti‐Claudin‐4 monoclonal antibody (mAb) and evaluate its antitumor efficacy in vitro and in vivo. To isolate specific mAb, we generated CLDN3, 4, 5, 6, and 9, expressing Chinese hamster ovary (CHO) cells, and then used them as positive and negative targets through cell‐based screening. As a result, we succeeded in isolating KM3900 (IgG2a), which specifically bound to CLDN4, from BXSB mice immunized with pancreatic cancer cells. Immunoprecipitation and flow cytometry analysis revealed that KM3900 recognized the conformational structure and bound to extracellular loop 2 of CLDN4. Furthermore, binding of KM3900 was detected on CLDN4‐expressing pancreatic and ovarian cancer cells, but not on negative cells. Next, we made the mouse–human chimeric IgG1 (KM3934) and evaluated its antitumor efficacy. KM3934 induced dose‐dependent antibody‐dependent cellular cytotoxicity and complement‐dependent cytotoxicity in vitro, and significantly inhibited tumor growth in MCAS or CFPAC‐1 xenograft SCID mice in vivo (P < 0.05). These results suggest that mAb therapy against CLDN4 is promising for pancreatic and ovarian cancers. (Cancer Sci 2009; 100: 1623–1630)

Pancreatic cancer is one of the most difficult cancers to treat, and in 2006, in the USA, more than 30 000 people died with the disease. Treatment options including surgery, radiation therapy, and chemotherapy may extend survival and/or relieve symptoms, but prognosis is very poor, with a 5‐year survival rate of about 5%, and complete remission extremely rare.^{( 1 )} Furthermore, ovarian cancer now causes more deaths than any other affecting the female reproductive system. And in 2006, an estimated 20 180 new cases and 15 310 deaths were reported in the USA.^{( 1 )} Both cancers therefore require novel treatment strategies.

Since establishment of monoclonal antibody (mAb) technology, attempts have been made to use its high‐level specificity and affinity for treatment of cancer, and it is now established as an effective treatment. Thus far, nine mAbs for hematological and solid cancer therapy, targeting CD20, CD33, CD52, HER2, EGFR, and VEGF, have been approved in the USA, and over 100 more mAb products are in clinical trials.^{( 2 )} In an oncology setting, human immunoglobulin (Ig) G1 isotype is commonly used, because it can mediate multiple effector functions, including antibody‐dependent cellular cytotoxicity (ADCC) and complement‐dependent cytotoxicity (CDC). Recent studies suggest that these effector functions are important to the therapeutic mechanism of effective antibodies.^{( 3} , ⁴ , ⁵ , ⁶ , ⁷ , ^{8 )}

Human Claudin‐4 (CLDN4) is a tetraspanin transmembrane protein consisting of 209 amino acids (aa) and belongs to a CLDN family^{( 9} , ^{10 )} of 24 members.^{( 11 )} The protein structure is thought to consist of cytoplasmic N‐ and C‐termini, four transmembrane domains and two extracellular loops on 28–76 aa (EL1) and 141–159 aa (EL2) from the N‐terminus, respectively.^{( 12 )} CLDN members form homo‐ or hetero‐binding between adjacent endothelial or epithelial cells, and generate a tight junction, a key construct for epithelial barrier and defense functions.^{( 11 )} In addition, CLDN4, along with CLDN3, also functions as a receptor for Clostridium perfringens enterotoxin (CPE).^{( 13 )} Interestingly, immunohistochemical analysis and differential expression studies between normal and cancerous tissue indicate that CLDN4 expression is highly detectable in a variety of carcinomas,^{( 14} , ¹⁵ , ^{16 )} including pancreatic^{( 17} , ¹⁸ , ¹⁹ , ^{20 )} and ovarian cancers.^{( 21} , ²² , ²³ , ²⁴ , ^{25 )} Furthermore, CLDN4 protein is detectable in normal breast, prostate, bladder, and gastrointestinal mucosa, although the expression is substantially less intense than that seen in cancer tissue samples.^{( 18 )} Augmentation of CLDN4 expression is observed in conjunction with tumor progression and high expression is an indicator of poor prognosis in gastric^{( 26 )} and endometrial cancers.^{( 27 )} These may indicate involvement of CLDN4 in carcinogenesis, but conflicting results were obtained from the in vitro functional analysis using culture cells.^{( 28} , ²⁹ , ³⁰ , ^{31 )} Therefore, although the role of CLDN4 in cancerous change or malignancy remains unclear, its expression and involvement in malignant carcinomas suggest that CLDN4 is a promising target for antibody therapy based on effector functions.

However, it is considered difficult to generate the antibody by conventional immunization approaches in mice or rabbits using peptide antigens, because there is a high conservation of CLDN4 in various species.^{( 10 )} To avoid this problem, Offner et al. immunized chickens with synthetic peptides of extracellular domains of human CLDN4, and succeeded in the generation of chicken polyclonal antibody, which bound to CLDN4‐expressing cells.^{( 12 )}

In our own study, we selected a human pancreatic cancer cell line, Capan‐2, reported as having high level expression of CLDN4,^{( 19 )} and immunized BXSB mice. Using intact cells as an immunogen may directly stimulate immune cells to produce antibodies against cell surface antigens.^{( 32 )} The procedure may thus avoid the suppression mechanism of autoimmune response in mice. Furthermore, we made CLDN3, 4, 5, 6, and 9 expressing Chinese hamster ovary (CHO) cells (CLDN/CHO), and utilized these as positive and negative targets through cell‐based screening. CLDN3, 5, 6, and 9 possess >70% extracellular domain homology compared with CLDN4,^{( 12} , ^{16 )} so it is possible to deselect cross‐reactive antibodies against these CLDNs. Consequently, we succeeded in isolation of a murine mAb, KM3900 (IgG2a), which bound to CLDN4, but did not bind to CLDN3, 5, 6, and 9. Additionally, we produced mouse–human chimeric IgG1 of KM3900 (KM3934) by the CHO/DG44 cell line and evaluated its antitumor activity against pancreatic and ovarian cancers in vitro and in vivo.

Materials and Methods

Mice and cell lines.  BXSB mice and C.B‐17/lcr‐scid Jcl (SCID) mice were purchased from Japan SLC (Shizuoka, Japan) and CLEA Japan (Tokyo, Japan), respectively. These mice were maintained under pathogen‐free conditions. All experiments were performed in conformity with institutional guidelines and in compliance with national laws and policies.

Chinese hamster ovary (CHO) cell line DG44 was kindly provided by Dr Lawrence Chasin (Columbia University). Murine myeloma cell line P3‐U1 (ATCC CRL‐1597); human ovarian cancer cell lines, MCAS (JCRB 0240), OVCAR‐3 (ATCC HTB‐161), OVISE (JCRB 1043), Caov‐3 (ATCC HTB‐75), OV‐90 (ATCC CRL‐11732), ES‐2 (ATCC CRL‐1978), and TOV‐112D (ATCC CRL‐11731); and human pancreas cancer cell lines, Capan‐2 (ATCC HTB‐80), CFPAC‐1 (ATCC CRL‐1918), AsPC‐1 (ATCC CRL‐1682), KP‐3 L (JCRB 0178), HPAF‐II (ATCC CRL‐1997), HPAC (ATCC CRL‐2119), BxPC‐3 (ATCC CRL‐1687), PANC‐1 (ATCC CRL‐1469), and MIA PaCa‐2 (ATCC CRL‐1420) were purchased from ATCC (Manassas, VA, USA) or the Japanese Collection of Research Bioresources (JCRB) (Osaka, Japan).

Establishment of CLDN‐expressing CHO cells.  Human CLDN3 (Genebank accession no. XM_057967), CLDN4 (Genebank accession no. XM_057966), CLDN5, (Genebank accession no. XM_009839), CLDN6 (Genebank accession no. XM_012518), CLDN9 (Genebank accession no. NM_020982), and domain exchanged CLDN (refer to Fig. 3) cDNAs were generated by polymerase chain reaction (PCR). Next, myc/his tag sequence (5'‐gaacaaaaactcatctcagaagaggatctgaatatgcataccggtcatcatcaccatcaccat‐3') was connected to the 3' end of CLDN ORF by second‐round PCR. The myc/his‐tagged cDNA (CLDN‐myc/his) was cloned into pKANTEX93 vector^{( 33 )} and introduced into DG44 cells via electroporation. G418‐resistant clones were selected and a high‐producing cell was isolated by single cell cloning.

Isolation of a murine anti‐CLDN4 mAb.  Six‐week‐old‐female BXSB mice (Japan SLC) were immunized four times with 1 × 10⁷ Capan‐2 cells with Bordetella pertussis adjuvant. The spleen was removed 3 days after final injection of the antigen, and 2 × 10⁸ splenocytes were fused with 2 × 10⁷ P3‐U1 in the presence of polyethylene glycol 1000 (Junsei, Tokyo, Japan). Cultured hybridoma cells in wells showing anti‐CLDN4 antibody activity were screened with the 8200 cellular detection system (Applied Biosystems, Tokyo, Japan), using CLDN4/CHO as a target cell. After cloning twice with limited dilution, a stable clone was obtained (KM3900). The immunoglobulin class and subclass of KM3900 (IgG2a) were determined with antimouse isotype‐specific antibodies (Zymed, San Francisco, CA, USA).

Production and purification of mouse–human chimeric anti‐CLDN4 IgG1.  The heavy‐ and light‐chain variable region cDNAs from hybridoma cells producing KM3900 were isolated by PCR and cloned into pKANTEX93 vector^{( 33 )} for production of mouse–human chimeric IgG1 antibody. The vector was then introduced into DG44 cells via electroporation and the transfected cell was grown in serum‐free EX‐CELL301 (JRH Bioscience, Lenexa, KS, USA). Chimeric KM3900 antibody (KM3934) was then purified from the supernatant using Prosep‐A column (Nihon Millipore, Tokyo, Japan).

Flow cytometry (FCM).  For the analysis of KM3900 binding to cell surface molecules, cells were detached with 0.02% EDTA solution (Nacalai Tesque, Kyoto, Japan) and stained with 5 µg/mL of KM3900 or control mouse IgG2a (Dako, Tokyo, Japan).

To detect the transfected CLDN‐myc/his expression, cells pretreated with 70% ethanol (EtOH) were stained with 1 µg/mL of anti‐myc antibody (MBL, Nagoya, Japan), anti‐his antibody (Qiagen, Tokyo, Japan), or control mouse IgG1 (Dako).

The reactivity was detected by FITC‐conjugated goat antimouse Ig antibody (Dako). Stained cells were then analyzed using an EPICS XL‐MCL FCM (Beckman Coulter, Tokyo, Japan).

Fluorescent immunostaining.  Cells were cultured on Lab‐Tek chamber slides (Nalge Nunc International, Rochester, NY, USA) and then fixed with 3% paraformaldehyde. After washing with PBS, the cells were treated with 0.1% Triton X‐100, blocked with 1% BSA in PBS, and reacted with 2 µg/mL of anti‐myc antibody (MBL). The reactivity was detected by FITC‐conjugated goat antimouse Ig antibody (Dako). Nuclei were stained with TOTO3 (Invitrogen, Tokyo, Japan). Stained cells were then analyzed with a confocal laser scanning microscope (Nikon ECRIPS C1; Nikon, Tokyo, Japan).

Reverse transcription–polymerase chain reaction (RT‐PCR).  Total RNA was isolated from the cultured cells using RNAiso Plus (Takara, Otsu, Japan), and then was used to synthesize cDNA with PrimeScript II (Takara). The cDNA was subjected to PCR with sequence‐specific primers (Supporting Information Table S1). Amplified DNA products were separated on a 2% agarose gel and visualized by staining the gel with ethidium bromide.

Western blotting and immunoprecipitation.  Cells were lysed in lysis buffer containing 50 mM Tis‐HCl (pH 7.2), 150 mM NaCl, 2 mM MgCl₂, 2 mM CaCl₂, 0.1% NaN₃, 5 µM PMSF, 100 µM DTT, 1% Triton X‐100, 50 mM N‐ethylmaleimide, and 1 mg/mL leupeptin. Equal amounts of cell lysates (50 µg/lane) measured using a Protein Assay Kit (Bio‐Rad, Hercules, CA, USA) were separated SDS‐PAGE and electrophoretically transferred onto a PVDF membrane. The membrane was blocked with 10% BSA and incubated with 1 µg/mL 3E2C1 (Zymed) or anti‐actin antibody (Sigma‐Aldrich, Tokyo, Japan). After washing, the membrane was exposed to horseradish peroxidase–conjugated antimouse IgG (Zymed) and the bands were visualized by ECL western blotting detection reagents (GE Healthcare, Buckinghamshire, UK).

To evaluate antibody binding to CLDN4 protein under non‐denaturing conditions, cell lysates were immunoprecipitated with control antibody KM511^{( 34 )} or KM3900, and the precipitates were then subjected to the treatment described above.

Antibody‐dependent cellular cytotoxicity (ADCC).  PBMC were separated from peripheral blood of healthy donors using Lymphoprep (Axis‐Shield, Oslo, Norway) and used as effector cells. Aliquots of the target cells (1 × 10⁴ cells/well) and effector cells (2 × 10⁵ cells/well, effector/target ratio is 20/1) were put into 96‐well plates and incubated with various concentrations of KM3934 for 4 h at 37°C. After centrifugation, the released lactate dehydrogenase (LDH) in the supernatant was detected with the LDH detection kit (Wako, Osaka, Japan). Percentage‐specific lysis was calculated from sample counts according to the formula:

% cytotoxicity = 100 × (E–S)/(M–S)

Where E is experimental release (count in the supernatant from target cells incubated with antibody and effector cells), S is spontaneous release (count in the supernatant from target cells incubated with medium alone), and M is the maximum release (count released from target cells lysed with Triton X‐100).

Blood donors were randomly selected from healthy volunteers registered at Kyowa Hakko Kirin. All donors gave written informed consent prior to participation.

Complement‐dependent cytotoxicity (CDC).  Target cells (5 × 10⁴) were incubated with various concentrations of KM3934 and human serum (Sigma‐Aldrich) as the source of complement at a dilution of 1:6 in supplemented RPMI‐1640 medium for 2 h at 37°C in 96‐well flat‐bottomed plates. After incubation, the cell proliferation reagent WST‐1 (Roche Diagnostics, Tokyo, Japan) was added and the plates were further incubated for 4 h to detect the live cells. Absorbance (450–650 nm) of the formazan dye produced by metabolically active cells of each well was detected on the Emax plate reader (Molecular Devices, Union City, CA, USA). Cytotoxicity was calculated using the following formula:

% cytotoxicity = 100 ×[1 – (E–S)/(M–S)]

Where E is experimental absorbance (cells were incubated with antibody and complement), S is spontaneous absorbance (medium and complement alone), and M is the maximum absorbance (cells were incubated with medium and complement alone).

In vivo antitumor activity.  MCAS (1 × 10⁶) or CFPAC‐1 (5 × 10⁶) cells were injected s.c. into 5–6‐week‐old male SCID mice. After injection, KM3934 (0.1, 1, 10 mg/kg) or 10 mg/kg human IgG (Sigma‐Aldrich) was injected i.p. twice a week for 3–4 weeks. Tumor volume was calculated using the following formula:

Tumor volume (mm³) = 0.5 × (major diameter) × (minor diameter)²

Statistical analysis.  Results are expressed as means ± SD. The statistical significance of differential findings between the experimental groups of animals was determined by two‐tailed unpaired t‐test.

Results

Isolation of mAb against CLDN4.  To generate anti‐CLDN4 antibodies, BXSB mice were immunized with a pancreatic cancer cell line, Capan‐2, and hybridoma cells were made by fusion of murine splenocytes and myeloma cells. For the selection of mAb against CLDN4, we generated CLDN4/CHO cells and used as a positive target through cell base screening with the Applied Biosystems 8200 cellular detection system. The CLDN4 cDNA was directly tagged by myc/his epitope at the C‐terminus region, so transfected gene expression was evaluated by anti‐tag antibodies. When anti‐tag antibodies were incubated with cells pretreated with 70% EtOH, the binding was detected in CLDN4/CHO, but not in vector‐transfected CHO (Vector/CHO) with FCM (Fig. 1a). Furthermore, immunohistochemical analysis with anti‐myc antibody showed that CLDN4‐myc/his protein localized not only in the cytoplasm, but also on plasma membrane (Fig. 1a). The appropriate expression of CLDN4‐myc/his was also determined by RT‐PCR with sequence‐specific primers (Fig. 1b) and western blotting (data not shown).

Figure 1.

KM3900 binds to Claudin‐4 (CLDN4), but not to CLDN3, 5, 6, and 9. CLDN‐expressing CHO cells (CLDN3/CHO, CLDN4/CHO, CLDN5/CHO, CLDN6/CHO, and CLDN9/CHO) and vector‐transfected CHO (Vector/CHO) were generated. The expression CLDNs were tagged myc/his sequence at the C‐terminus, so the expression was evaluated by anti‐myc or his antibodies. Control mouse IgG1 (gray histogram) and anti‐myc or anti‐his antibody (open histogram) reacted to CLDN4/CHO or Vector/CHO pretreated with 70% ethanol. The binding was detected by FITC‐conjugated antimouse antibody (a, left). Furthermore, CLDN4/CHO and Vector/CHO were subjected to fluorescence immunohistochemical analysis using anti‐myc antibody and the staining was visualized by confocal laser scanning microscope (white bar, 40 µm). The green staining represents the binding of anti‐myc antibody, which was detected with FITC‐conjugated antimouse antibody. The blue staining represents the cell nuclei staining with TOTO3 (a, right). Control mouse IgG2a (gray histogram) or KM3900 (open histogram) reacted to CLDN3, 4, 5, 6, 9, and Vector/CHO cells and the binding was detected by FITC‐conjugated antimouse antibody (b). The mRNA expressions of transfected genes were detected by RT‐PCR with sequence‐specific primers (b). Reverse transcriptase (–) was used as a negative control (RT–).

From the screening, several mAbs were selected for binding to CLDN4/CHO, but not to Vector/CHO. Next, we evaluated the cross reactivity of these candidates against CLDN3, 5, 6, and 9/CHO cells. These CLDNs were also tagged by myc/his epitope at the C‐terminus region, so we confirmed the appropriate expressions of transfected genes by the same procedure to establish CLDN4/CHO (Fig. 1b). Through the screening process, several mAbs were deselected by cross reactivity. Finally, we succeeded in isolating KM3900, which bound to CLDN4/CHO, but not to CLDN3, 5, 6, 9, and Vector/CHO (Fig. 1b).

KM3900 recognized conformational structure and bound to EL2 of CLDN4.  To further evaluate binding of KM3900 to CLDN4, cell lysates were prepared from pancreatic cancer cells (Capan‐2 and HPAF‐II) and CLDN/CHO cells, and subjected to immunoprecipitation analysis. As a result, the reaction of KM3900 specifically precipitated CLDN4 (22 kDa) or CLDN4‐myc/his (25 kDa), but not control antibody KM511 (Fig. 2). Conversely, we failed to detect the binding of KM3900 in western blotting using KM3900 as a primary antibody (data not shown). These results show that KM3900 recognized the conformational structure of CLDN4.

Figure 2.

Immunoprecipitation of KM3900 to Claudin‐4 (CLDN4)‐expressing cells. Immunoprecipitation (IP) was conducted by control KM511 or KM3900 in cell lysates, and then the precipitates were subjected to western blotting (WB) using 3E2C1 to detect CLDN4.

Next, we aimed to determine which extracellular domain of CLDN4 was recognized by KM3900, and thereby generated artificial CLDNs in which the domains were exchanged between CLDN4 and 6.

CLDN4/6/6 (CLDN4/EL1 + CLDN6/EL2 + CLDN6/C‐terminus cytoplasmic domain), CLDN6/4/4 (CLDN6/EL1 + CLDN4/EL2 + CLDN4/C‐terminus cytoplasmic domain), and CLDN4/4/6 (CLDN4/EL1&EL2 + CLDN6/C‐teminus cytoplasmic domain) tagged myc/his sequence at the C‐terminus were constructed and generated the CHO cells (see Fig. 3 for details).

Figure 3.

KM3900 recognizes the EL2 domain of Claudin‐4 (CLDN4). Domain‐exchanged CLDNs (CLDN4/4/6, CLDN4/6/6, and CLDN6/4/4) were constructed and the expression CHO cells were generated. The exchange position in the amino acid (aa) sequence of CLDNs is described briefly (a–c). KM3900 (bold line) or control mouse IgG2a (dotted line) reacted to detached cells and binding was detected by FITC‐conjugated antimouse antibody.

Appropriate expression of transfection genes was confirmed by the same procedure described above (data not shown). Interestingly, KM3900 bound to CLDN6/4/4 and 4/4/6, but not to CLDN4/6/6 with FCM analysis (Fig. 3). Therefore, we believe that KM3900 recognizes the EL2 domain of CLDN4.

Binding of KM3900 against pancreatic and ovarian cancer cells.  FCM analysis of pancreatic and ovarian cancer cells was conducted. Nine pancreatic cancer cells (Capan‐2, CFPAC‐1, AsPC‐1, KP‐3L, HPAF‐II, HPAC, BxPC‐3, PANC‐1, and MIA PaCa‐2) and seven ovarian cancer cells (MCAS, OVCAR‐3, OVISE, Caov‐3, OV‐90, ES‐2, and TOV‐112D) were detached with 0.02% EDTA and reacted with KM3900. In pancreatic cancers, KM3900 bound to all nine cells, in which CLDN4 expression was determined by western blotting using 3E2C1, which recognized the C‐terminus of CLDN4 localized in cytoplasm (Fig. 4). Additionally, the results showed a positive relationship between the binding intensities of KM3900 and the expression level of CLDN4. In ovarian cancers, the reactivity of KM3900 showed only on CLDN4‐expressing ovarian cells (MCAS, OVCAR‐3, OVISE, and Caov‐3), but not on negative cells (OV‐90, ES‐2, and TOV‐112D) (Fig. 5). Moreover, when we evaluated binding of KM3900 against human peripheral blood cells (white blood cells, red blood cells, and platelets), the binding was not detected (data not shown). Although this analysis utilized only a limited number of cells, these results also suggest that KM3900 possesses specific reactivity against CLDN4.

Figure 4.

Flow cytometry analysis of KM3900 against pancreatic cancer cells. Cells were reacted with 5 µg/mL of KM3900 (open histogram) or control mouse IgG2a (gray histogram), and binding was detected by FITC‐conjugated antimouse antibody (a). Cell lysates derived from pancreatic cancer cells were separated with SDS‐PAGE and transferred to PVDF membrane. The bands show anti‐CLDN4 (3E2C1) or anti‐actin antibody reactions (b).

Figure 5.

Flow cytometry analysis of KM3900 against ovarian cancer cells. Cells were reacted with 5 µg/mL of KM3900 (open histogram) or control mouse IgG2a (gray histogram), and binding was detected by FITC‐conjugated antimouse antibody (a). Cell lysates derived from ovarian cancer cells were separated with SDS‐PAGE and transferred to PVDF membrane. The bands show anti‐CLDN4 (3E2C1) or anti‐actin antibody reactions (b).

Production and evaluation of antitumor activity of KM3934, mouse–human chimeric IgG1 of KM3900.  Using PCR, we cloned the variable regions of heavy and light chains from KM3900 hybridoma cells, and constructed the expression vector of mouse–human chimeric IgG1 with human IgG1 Fc domain (KM3934). CHO/DG44 was used for production of KM3934 and purified with Protein‐A column. The elution of more than 90% purity of intact IgG was confirmed by SDS‐PAGE. Specificity and affinity against CLDN4‐expressing cells were almost equivalent to murine KM3900 (data not shown).

We evaluated the human PBMC‐mediated ADCC of KM3934 against CLDN4/CHO, Vector/CHO, MCAS, and CFPAC‐1. The effector/target ratio was 20:1 and cytotoxicity was measured by released LDH activity from apoptotic cells by ADCC. As a result, antibody dose‐dependent ADCC activity was detected with a potency of 10–100 ng/mL, but not detected in Vector/CHO (Fig. 6a–d). When CDC activity was evaluated using human complement, activity against CLDN4/CHO was detected dose‐dependently, but not against Vector/CHO (Fig. 6e,f). CDC activity against MCAS and CFPAC‐1 were not detected (data not shown).

Figure 6.

Antibody‐dependent cellular cytotoxicity (ADCC) and complement‐dependent cytotoxicity (CDC) activities of KM3934. ADCC: Human PBMCs were purified from healthy donors and used as effector cells. Target cells, Vector/CHO (a), CLDN4/CHO (b), MCAS (c), and CFPAC‐1 (d), incubated with effector cells (Effector/Target = 20/1) and KM3934. The released lactate dehydrogenase was measured and cytotoxicity (%) was determined (N = 3). CDC: Target cells, Vector/CHO (e) and CLDN4/CHO (f), were incubated with human complement (16.7%) and KM3934. The amount of viable cells were measured by WST‐1 assay and cytotoxicity (%) was determined (N = 3).

To further evaluate the therapeutic potential of KM3934, we examined in vivo antitumor activity against MCAS‐ or CFPAC‐1‐inoculated SCID mice. After inoculation of tumor cells, antibody injection of KM3934 (0.1, 1, and 10 mg/kg) or human IgG control (10 mg/kg) was conducted i.p. twice a week. In MCAS xenograft mice, the KM3934 treatment significantly inhibited tumor growth in a dose‐dependent manner (P < 0.05) (Fig. 7a). The treated/control (T/C) ratios at day 21 of relative mean tumor volume were: 0.48 for 0.1 mg/kg, 0.08 for 1 mg/kg, and 0.07 for 10 mg/kg of KM3934. Meanwhile, significant inhibition of tumor growth in CFPAC‐1‐inoculated mice was produced by treatment of 10 mg/kg of KM3934 (P < 0.05) (Fig. 7b). The T/C indicated 0.36 at day 38.

Figure 7.

In vivo antitumor activity of KM3934. MCAS (1 × 10⁶) or CFPAC‐1 (5 × 10⁶) cells were injected s.c. into SCID mice (day 0). After inoculation, antibody injection of KM3934 (0.1 mg/kg, white square; 1 mg/kg, white triangle; 10 mg/kg, white circle) or control of human IgG (10 mg/kg, black circle) was made i.p. twice a week from day 0 (N = 5). Statistical significance was P < 0.05 (*) or P < 0.01 (**) versus control human IgG treatment.

Discussion

In this study, we succeeded in isolating anti‐CLDN4 mAb, which recognized the extracellular domain of CLDN4, and we evaluated its antitumor efficacy against ovarian and pancreatic cancer cells. As far as we know, this is the first report of generation of a mAb that recognizes the extracellular domain of CLDN4. Furthermore, the mouse–human chimeric mAb induced ADCC and CDC activities in vitro, and significantly inhibited xenograft tumor growth in vivo. These results suggest that mAb therapy targeting CLDN4 is a promising therapeutic approach to pancreatic and ovarian cancers.

To generate mAb, we immunized BXSB mice with an intact pancreatic cancer cell expressing CLDN4, and carried out cell‐based screening using CLDN4/CHO. Furthermore, we generated CLDN3, 5, 6, and 9/CHO cells and evaluated the cross reactivity of anti‐CLDN4 mAb candidates against them. We believe these processes taken together are important to isolate a mAb specific to CLDN4. Evaluation of cross reactivity, in particular, plays an important role in success, because several anti‐CLDN4 mAb candidates were deselected by cross reactivity against CLDN3, 5, 6, and/or 9 (data not shown). It is known that these CLDNs express in a variety of normal tissues, including the lung, kidney, pancreas, and colon.^{( 10} , ^{16 )} Therefore, cross reactivity may be critical to application of mAb against CLDN4.

KM3900 showed specific binding to the EL2 domain of CLDN4, but not to EL1. The binding property may account for specific binding to CLDN4, since the EL1 domain of CLDN4 is very similar (>80%) to the EL1 domains of CLDN3, 5, 6, and 9, while the EL2 domain of CLDN4 is less similar (40–80%) to the EL2 domains of these CLDNs. Therefore, the EL2 domain may be an acceptable epitope for specific binding against CLDN4.

To evaluate the therapeutic potential of anti‐CLDN4 mAb, we generated KM3934, mouse–human chimeric IgG1 of KM3900 with the human Fc domain, and measured the effector functions. As a result, dose‐dependent ADCC activities were detected against CLDN4/CHO, MCAS, and CFPAC‐1 in the method using human PBMC as effector cells. Recent studies indicate that ADCC is an important therapeutic mechanism of mAb therapy for cancer immunotherapy, not only in hematological cancers, but also in solid tumors.^{( 3} , ⁴ , ⁵ , ^{6 )} Although maximum cytotoxicity of KM3934 was not strong in our study (8–35%), the ADCC potential indicates the therapeutic potential of anti‐CLDN4 mAb for cancer. On the other hand, CDC activity was only detected against CLDN4/CHO, but not MCAS and CFPAC‐1 (data not shown). Further study is needed, but several factors, including antigen expression amount and complement inhibitors such as CD46, CD55, and CD59 on tumor cells, may account for the difference.^{( 35} , ^{36 )}

As direct killing or anti‐proliferation activity of KM3934 against MCAS and CFPAC‐1 was not shown in our study (data not shown), the effector function of KM3934 is probably a major contributor of in vivo antitumor activity against xenograft in SCID mice. However, we cannot rule out the possibility that an unidentified mechanism exists in addition to ADCC in the in vivo activity of KM3934, because of the role of CLDN4 in cancer progression. Agarwal et al. reported that overexpression of CLDN4 induced activation of matrix metalloproteinase‐2 and increased motility and invasiveness in human ovarian surface epithelial cells.^{( 31 )} On the other hand, several studies showed that overexpression of CLDN4 decreases motility, invasiveness, or anchorage in pancreatic^{( 28 )} and gastric cancer cells,^{( 29 )} and suppression of CLDN4 increases motility in colorectal cancer cells.^{( 30 )} The role of CLDN4 in cancer cell biology remains unclear, but the apparent contradiction suggests that CLDN4 possesses cell‐type specific effects.

As for targeting CLDN4, CPE and C‐CPE, the binding domains to CLDN3 and/or CLDN4, have recently been examined as potential cancer therapeutic agents. CPE is a single polypeptide of 319 aa bound to CLDN3 and/or CLDN4, and causes cytolysis in a dose‐dependent manner.^{( 19} , ^{37 )} In vivo efficacy of CPE treatment has been investigated against several types of cancer by local administration and a number of animal studies show some success.^{( 19} , ²³ , ³⁷ , ³⁸ , ^{39 )} On the other hand, C‐CPE is the C‐terminal 30 aa of CPE and is insufficient to induce cytolysis. Interestingly, C‐CPE is internalized following binding through endocytosis,^{( 40 )} so toxin‐conjugated C‐CPE was generated and showed cytotoxic activity in breast cancer cells.^{( 41 )} In addition, C‐CPE may have potential to enhance standard chemotherapy, because binding of C‐CPE results in removal of binding CLDNs from the cell surface, thereby increasing paracellular permeability.^{( 15 )} Compared with these CPE‐related agents, anti‐CLDN4 mAb may possess different pharmacological and therapeutic properties. Therefore, we believe that mAb therapy targeting CLDN4 is a valuable option for therapeutic or diagnostic application in cancer therapy.

In addition to pancreatic and ovarian cancers, CLDN4 is overexpressed in a variety of carcinomas, including uterus, colorectal, breast, and prostate cancers.^{( 14} , ¹⁵ , ^{16 )} Although further examination is needed, mAb therapy targeting CLDN4 may be potentially useful against these carcinomas. On the other hand, since CLDN4 is expressed in a variety of normal tissues including the lung and kidney,^{( 10 )} its effects need careful evaluation.

Abbreviations

CHO	Chinese hamster ovary
EGFR	epidermal growth factor receptor
HER2	human epidermal growth factor receptor 2
PBMC	peripheral blood mononuclear cell
VEGF	vascular endothelial growth factor

Disclosure Statement

The authors have been employed by Kyowa Hakko Kirin.

Supporting information

Table S1. Primers used for detection of Claudin‐4 (CLDN) gene expression.

Please note: Wiley‐Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

Supporting info item