Open Full Page

Vol. 9, 2567–2575, July 2003 Clinical Cancer Research 2567

Tight Junction Proteins Claudin-3 and Claudin-4 Are Frequently Overexpressed in Ovarian Cancer but Not in Ovarian Cystadenomas

Leticia B. A. Rangel,1 Rachana Agarwal,1 Results: Although expressed at low levels in some nor- Theresa D’Souza, Ellen S. Pizer, Piero L. Alo`, mal human tissues, including the ovary, CLDN3 and CLDN4 are highly up-regulated in epithelial ovarian cancers of all Wayne D. Lancaster, Lucie Gregoire, subtypes. Immunohistochemical analyses using our ovarian Donald R. Schwartz, Kathleen R. Cho, and tissue array confirmed the high level of expression of clau- Patrice J. Morin2 din-3 and claudin-4 in the majority of ovarian carcinomas, Laboratory of Cellular and Molecular Biology, Gerontology Research including many tumors exhibiting cytoplasmic staining. Center, National Institute on Aging, Baltimore, Maryland 21224 Ovarian cystadenoma did not frequently overexpress these [L. B. A. R., R. A., T. D., P. J. M.]; Department of Basic and Clinical proteins, suggesting that the expression of these proteins is Pharmacology, Federal University of Rio de Janeiro, Rio de Janeiro, associated with malignancy. In ovarian cancer cell lines, Brazil [L. B. A. R.]; Dipartimento di Medicina Sperimentale e Patologia, Universita`di Roma, Rome, Italy [P. L. A.]; Department of claudin-3 and claudin-4 expression was not associated with Obstetrics and Gynecology, Wayne State University School of functional TJs as measured by transepithelial electrical re- Medicine, Detroit, Michigan 48201 [W. D. L., L. G.]; Department of sistance. Pathology, The University of Michigan Medical School, Ann Arbor, Conclusions: These results show that CLDN3 and Michigan 48109 [D. R. S., K. R. C.]; and Department of Pathology, CLDN4 are frequently up-regulated in ovarian tumors and The Johns Hopkins Medical Institutions, Baltimore, Maryland 21287 [E. S. P., P. J. M.] cell lines and may represent novel markers for this disease. Overexpression of these genes in ovarian cancer also suggests interesting scenarios for the involvement of TJ in ABSTRACT tumorigenesis. A better knowledge of the mechanisms un- Purpose: Claudin proteins represent a large family of derlying ovarian tumorigenesis will likely result in the de- integral membrane proteins crucial for tight junction (TJ) velopment of novel approaches for the diagnosis and ther- formation and function. Claudins have been shown to be apy of this deadly disease. up-regulated in various cancers and have been suggested as possible biomarkers and targets for cancer therapy. Because INTRODUCTION claudin-3 and claudin-4 have been proposed to be expressed Ovarian cancer is one of the leading causes of cancer in epithelial ovarian cancer, we have performed a detailed deaths in women due to difficulties in both diagnosis and analysis of CLDN3 and CLDN4 expression in a panel of therapy. Despite recent efforts in the search for molecular mech- ovarian tumors of various subtypes and cell lines. We also anisms responsible for the development of this cancer, the investigated whether high expression of claudin-3 and clau- pathways important for the initiation and development of ovar- din-4 was associated with TJ function in ovarian cancer ian malignant transformation have remained elusive. We and cells. others have recently used SAGE3 and cDNA microarrays to Experimental Design: RNA was obtained from a panel identify profiles of gene expression in ovarian carcinoma and of 39 microdissected epithelial ovarian tumors of various normal ovarian epithelium (1–5). These studies have yielded histological subtypes for real-time reverse transcription- many candidate target genes for detection and therapy. Our PCR analysis. In addition, a total of 70 cases of ovarian SAGE study of ovarian cancer recently identified genes encod- carcinomas, ovarian cysts, and normal ovarian epithelium ing TJ proteins claudin-3 and claudin-4 as two of the most from a tissue array were analyzed by immunohistochemis- highly up-regulated genes in ovarian cancer (1), but this con- try. Finally, a panel of cell lines was used for Western clusion was based on a limited number of ovarian samples, and analysis of claudin expression and TJ permeability studies. the details of this overexpression are unclear. TJs are critical structures for the maintenance of cellular polarity, as well as for the establishment of a permeability barrier for paracellular transport in epithelial and endothelial cells (6–8). The two major integral membrane constituents of Received 10/14/02; revised 2/5/03; accepted 2/10/03. The costs of publication of this article were defrayed in part by the TJs are occludin and claudin proteins. Whereas no occludin- payment of page charges. This article must therefore be hereby marked related genes have yet been identified, claudins constitute a advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 These authors contributed equally to this work. 2 To whom requests for reprints should be addressed, at Laboratory of Cellular and Molecular Biology, Gerontology Research Center, National 3 The abbreviations used are: SAGE, serial analysis of gene expression; Institute on Aging, NIH, 5600 Nathan Shock Drive, Baltimore, MD TJ, tight junction; TER, transepithelial electrical resistance; RT-PCR, 21224. E-mail: [email protected]. reverse transcription-PCR; MDCK, Madin-Darby canine kidney.

Fig. 1 SAGEmap analysis of CLDN3 and CLDN4 expression in cancer. Tags for CLDN3 (ctcgcgctgg) and CLDN4 (atcgtggcgg) were used to identify gene expression patterns in the indicated normal (N) and tumor (T) tissues. The novel set of Genie tools was used for analysis Fig. 2 Northern analysis of CLDN3 and CLDN4 in normal and malig- (28). To allow for meaningful comparisons between the various librar- nant ovarian tissues and cell lines. RNA samples for normal tissue (A), ies, tag frequency was normalized to tags per 200,000. primary tumors (B), or cell lines (C) were electrophoresed and probed using human CLDN3 and CLDN4 probes. All blots shown result from one experiment and were exposed to a PhosphorImager screen for 24 h.

family of at least 20 proteins. Normal cells typically express multiple claudin proteins, but some family members exhibit highly tissue-specific expression patterns (6). The exact function claudin overexpression and cancer initiation or progression is of claudin proteins within TJs is still unclear, but they appear to unclear. be important in TJ formation and function. Indeed, expression of This study represents the first detailed analysis of CLDN3 CLDN1 and CLDN2 is sufficient to induce the formation of TJs and CLDN4 expression in ovarian cancer. We show for the first in fibroblast cells (9). Claudin proteins are also likely involved time that these proteins are frequently elevated in all primary in signaling. This function is suggested by several lines of epithelial ovarian cancer subtypes and in many ovarian cancer evidence, including the fact that claudin proteins contain, at cell lines but not in nonmalignant ovarian cystadenomas. Al- their COOH termini, PDZ binding domains (10) that may recruit though normally found at the membrane, these proteins are various proteins involved in signaling. frequently mislocalized in cancer, suggesting abnormal process- The high degree of cellular organization typically observed ing and that their role in malignancy may be unrelated to known in normal differentiated tissues is often lost in cancer. Indeed, TJ functions. Indeed, claudin protein increase is not associated tumor cells frequently exhibit deficiencies in TJ function, as with increased TJ tightness as measured by TER in various well as decreased differentiation and cell polarity (11, 12). Loss ovarian cancer cell lines. Whereas the functional significance of of TJ integrity may be particularly important to allow the claudin overexpression in ovarian carcinoma is unclear, these diffusion of nutrients and other factors necessary for the survival proteins clearly represent a novel marker for ovarian cancer and and growth of the tumor cells (13). In addition, decreased may become a target for therapy or diagnosis of this disease. polarity and differentiation may be important for the metastatic phenotype, where individual cells must leave the primary site MATERIALS AND METHODS and enter the blood vessels to reach distant sites (14, 15). Cell Lines and Tissue Samples. Ovarian cancer lines Finally, the destruction of functional TJs in cancer may have a A2780, 2008, OVCAR-2, OVCAR-3, OVCAR-4, OVCAR-5, role in growth control. For example, in Drosophila, mutations in and OVCA432 were cultured in McCoy’s 5A growth medium many tumor suppressor genes lead to loss of cell polarity and supplemented with 10% fetal bovine serum and antibiotics (100 overproliferation of the epithelia (16). Because of the similarity units/ml penicillin and 100 ␮g/ml streptomycin). HOSE-B, an between the vertebrate and Drosophila epithelia, mammalian ovarian surface epithelial cell line immortalized with E6 and E7 cells are likely to require cytoarchitectural cues for cell growth (23), was cultivated in RPMI 1640 supplemented with 10% FCS control as well. and 300 ␮g/ml G418. SV40-immortalized cystadenoma line Whereas expression of TJ proteins such as occludin and ML3 (24) was kindly provided by Dr. Louis Dubeau (Los claudin-1 has been found to be decreased in cancer (17–19), Angeles, CA) and cultivated in MEM (Life Technologies, Inc., expression of some claudin family members is highly elevated Gaithersburg, MD) supplemented with 10% fetal bovine serum in various human cancers (1, 20–22). The relationship between and antibiotics as described above. Frozen primary colon tumors

Fig. 3 Real-time RT PCR analysis of CLDN3 and CLDN4 expression. The Y axis represents the fold induction relative to HOSE-B expression. The X axis represents each sample tested for CLDN3 (A) and CLDN4 (B) expression. The first fifteen bars represent normal primary tissues labeled as follows: 1, brain; 2, heart; 3, lung; 4, stomach; 5, liver; 6, colon; 7, kidney; 8, breast; 9, ovary; 10, uterus; 11, placenta; 12, cervix; 13, skeletal muscle; 14, spleen; 15, leukocyte. Bar 16 represents cystadenoma line ML3. The following 39 bars represent individual microdissected ovarian primary tumors of various subtypes, as indicated below the graph (CC, clear cell; Endo, endometrioid; Muc, mucinous).

were generously provided by Dr. Bert Vogelstein (Baltimore, University of Rome “La Sapienza” from 1983 to 1998 as well as MD). The primary brain tumor was a gift of Dr. Greg Riggins patients from Johns Hopkins School of Medicine were studied. (Durham, NC), and ovarian tumors were obtained through the All samples were collected anonymously according to Institu- Collaborative Human Tissue Network, Gynecologic Oncology tional Review Board guidelines. No patients received chemo- Group (Children’s Hospital, Columbus, OH). In addition, 35 therapy before surgery. Histopathological data included tumor snap-frozen ovarian carcinomas of various histological types stage according to International Federation of Gynecologists were obtained from the University of Michigan, Department of and Obstetricians (FIGO) classification and histological sub- Pathology. These tumors were manually microdissected using type. With the guidance of a pathologist (E. S. P.), representa- H&E-stained frozen sections as dissection guides. The charac- tive core tissue biopsies (2 mm in diameter) were taken from teristics of these specimens have been published previously archival paraffin-embedded primary ovarian tumors and seeded (25). RNA was prepared from cell lines and primary tumors as in a new recipient paraffin block. The tissue Ovarray consisted described previously (1) or purchased from Geneka Biotechnol- of two blocks of 60 cases each. For the experiments reported ogy Inc. (Montreal, Canada). RNA from normal tissue was here, 70 cases were analyzed: 3 normal ovary samples; 8 mu- purchased from Invitrogen (Carlsbad, CA). cinous cystadenomas; 7 serous cystadenomas; 40 serous adeno- Tissue Array and Immunohistochemistry. The protein carcinomas; 7 endometrioid adenocarcinomas; and 5 clear cell expression of claudin-3 and claudin-4 in ovarian tissue was adenocarcinomas (kidney tissue punches were also included as assessed by immunohistochemical staining using a tissue array internal control). Five-␮m-thick sections were cut from the constructed in our laboratory, the Ovarray. Ovarian tumors Ovarray, deparaffinized, and dehydrated. Immunohistochemical obtained from ovarian cancer patients surgically treated at the staining for claudin-3 and claudin-4 was performed using a

described previously (25). The primers for CLDN3 (forward, 5Ј-gtccgtccgtccgtccg; reverse, 5Ј-gcccagcacggccagc) and GAPDH (forward, 5Ј-gaaggtgaaggtcggagtc; reverse, 5Ј-gaagat- ggtgatgggatttc) were designed to cross intron-exon boundaries to distinguish PCR products generated from genomic versus cDNA template. CLDN4 does not contain introns, and the real-time RT-PCR was performed by the poly(A) cDNA- specific RT-PCR method (26) using appropriate primers (forward, 5Ј-ggacagcttcacccttgg; reverse, 5Ј-tttttttttttttttcctgtgca). In addition to GAPDH, RNase 18S was used for normalization, and the primers were purchased from Ambion (Austin, TX). Each PCR reaction was optimized to ensure that a single band of the appropriate length was amplified and that no bands corresponding to genomic DNA amplification or primer-dimer pairs were present. The PCR cycling conditions were performed Fig. 4 Ovarian cancer cell lines express claudin proteins. Western blot for all samples as follows: 50°C, 2 min for AmpErase UNG analysis of the various cell lines was performed with the indicated anti-claudin antibodies. The majority of the cancer cell lines studied incubation; 95°C, 10 min for AmpliTaq Gold activation; and 40 exhibit expression of claudin-3 and claudin-4 protein, which are absent cycles for the melting (95°C, 15 s) and annealing/extension from normal ovarian surface epithelium. Claudin-1 is included as a (60°C for 1 min) steps. PCR reactions for each template were control. done in duplicate in 96-well plates. The comparative CT method (PE Applied Biosystems) was used to determine gene expression in each sample relative to the value observed in the nonmalignant HOSE-B, using GAPDH and 18S ribosomal RNA as streptavidin peroxidase procedure. Primary rabbit polyclonal normalization controls. antibodies against claudin-3 and claudin-4 were kindly provided Immunoblotting. Confluent cell cultures were washed by Drs. M. Furuse and S. Tsukita (Kyoto University, Kyoto, with HBSS (Invitrogen), and whole cell lysates were made using Japan). Antigen-bound primary antibody was detected using lysis buffer [62.5 mM Tris-HCl (pH 6.8), 10% glycerol, and 2% standard avidin-biotin immunoperoxidase complex (Dako SDS]. Protein concentration was determined using the BCA Corp., Carpinteria, CA). Cases with less than 10% staining in assay kit (Pierce, Rockford, IL). Twenty ␮g of total proteins tumor cells were considered negative for claudin expression. were separated by 10–20% SDS-PAGE (Tris-glycine gels; In- The positive cases were classified as follows regarding the vitrogen) and transferred to polyvinylidene difluoride mem- intensity of claudin-3 and claudin-4 protein expression using a branes (Millipore, Bedford, MA). The membranes were blocked semiquantitative method: (a) ϩ, medium to weak staining; and with 5% nonfat dry milk, washed in Tris-buffered saline with (b) ϩϩ, medium to intense staining. Subcellular localization 0.05% Tween-20 (v/v) and probed with the primary antibody (membrane or cytoplasm) was also noted. Negative controls, in [anti-claudin-1, 1:200; anti-claudin-3, 1:200; anti-claudin-4, which the primary antibodies were absent, were processed in 1:250 (Zymed, South San Francisco, CA)], washed and incu- parallel, and no positive staining was observed. bated in horseradish peroxidase-conjugated secondary antibody Northern Blotting. Ten ␮g of total RNA isolated from (antimouse or antirabbit IgG, 1:10,000; Amersham Biosciences normal tissue, cancer cell lines, and primary tumors were size- Corp., Piscataway, NJ). For detection, enhanced chemilumines- fractionated on 1% agarose/12.5% formaldehyde gels, trans- cence was carried out using the enhanced chemiluminescence ferred by capillarity onto nylon membranes, and cross-linked by kit (ECL; Amersham Biosciences Corp.). UV irradiation and by incubation of the membranes at 80°C for TER Measurements. Cells were plated at a density of 2 min. Blots were prehybridized and hybridized with DNA 1 ϫ 105 cells/well in 12-well Transwell filters. TER was mea- probes at 65°C in hybridization solution containing 1% BSA, sured using a Millicell-ERS epithelial V-ohmmeter (World Pre- 7% SDS, 0.25 M phosphate buffer, and 1 mM EDTA. After cision Instruments, New Haven, CT). The TER values were hybridization, blots were washed six times under high strin- calculated by subtracting the blank values from the sample gency conditions (40 mM phosphate buffer and 1% SDS at values and normalized to the growth area of the monolayer. 65°C) and exposed to a PhosphorImager screen for 24 h. Human CLDN3 and CLDN4 were amplified by PCR and cloned into the pCiNeo vector using the EcoRI and SalI restriction sites. The RESULTS probes were obtained by isolation of the EcoRI/SalI fragment SAGEmap Mining Shows CLDN3 and CLDN4 Elevated and labeled with [␣-32P]dATP by random-primed labeling in Multiple Tissues. SAGEmap was recently created as pub- (Roche Molecular Biochemicals, Indianapolis, IN). lic database to provide a central location for SAGE data archival Real-Time RT-PCR. One ␮g of total RNA from the and retrieval (27). SAGEmap therefore allows virtual compari- various tissues and cell lines was used to generate cDNA using sons of gene expression in various normal and malignant tissues Taqman Reverse Transcription Reagents (PE Applied Biosys- of practically any gene of interest. Using the recently developed tems, Foster City, CA). The SYBR Green I assay and the web tools SAGE Genie (28), we evaluated CLDN3 and CLDN4 GeneAmp 5700 Sequence Detection System (PE Applied Bio- expression in six common malignancies for which data are systems) were used for detecting real-time PCR products as available for both the cancer and its normal counterpart. As

Fig. 5 Immunohistochemical examination of ovarian tissues present on the tissue Ovarray. Representative staining with anti-claudin-3 and anti-claudin-4 antibodies of normal ovarian surface epithelium and serous, endometrioid, and clear cell ovarian carcinomas is shown. Staining at the membrane and cytoplasm of cancer is evident. Specimens are counterstained with hematoxylin. Negative controls (without primary antibody) do not exhibit staining.

shown in Fig. 1, CLDN3 was elevated in all of the cancer types blotting (Fig. 2, A and B). It is important to note that whereas the evaluated except colon cancer, in which it was decreased. SAGE data were obtained from populations enriched in epithe- CLDN4 was elevated in ovarian, gastric, and pancreatic tumors lial cells, the Northern analysis was performed from RNA and decreased in colon cancer. Very few CLDN4 tags were obtained from bulk tissue. It is possible that the high level of detected in breast or brain tissues. CLDN3 and CLDN4 thus expression observed in normal colon epithelial cells by SAGE is appeared to be frequently up-regulated in cancer, except in the effectively diluted by the presence of other cells in the Northern colon, where high levels were present in normal tissues. experiment. CLDN3 and CLDN4 expression was also examined CLDN3 and CLDN4 Transcript Levels Are Elevated in in a series of cell lines to assess whether expression was retained Ovarian Cancer Tissues. To validate and extend these find- in cell lines and to identify potentially useful models to study the ings, we performed Northern analysis on a series of normal function of these proteins. Colon cancer cell lines HT-29 and tissues, primary tumors, and cell lines. CLDN3 and CLDN4 SW480, ovarian cancer cell lines OVCA432 and OVCA433, were not generally expressed in normal human tissues and were and prostate cancer line PC3 exhibited expression of both found at significant levels only in colon (Fig. 2A). CLDN3 was CLDN3 and CLDN4 (Fig. 2C). Breast cancer cell line MCF-7 also detected at low levels in normal breast and uterus. Although expressed CLDN3 only. The other cell lines studied did not absent in the brain tumor used as control, CLDN3 and CLDN4 express either transcript. were present in all of the ovarian and colon tumors analyzed To get highly sensitive measurements of claudin expres- (Fig. 2B). In contrast to what was observed by SAGE, expression in normal tissues and microdissected ovarian tumors of sion of these transcripts appeared elevated in colon tumors various subtypes, we developed a real-time RT-PCR assay. Both compared with normal colon tissues when analyzed by Northern CLDN3 and CLDN4 were found to be highly up-regulated in all

Table 1 Summary of claudin-3 and claudin-4 expression in ovarian tumors using the tissue array Percentage of cases with staininga Claudin-3 Claudin-4 Specimens Mb Cb MϩCb MCMϩC Normal ovary (n ϭ 3) 0 0 0 0 0 33 (ϩϩ)0 (ϩϩ)0 (ϩϩ)0 (ϩϩ)0 (ϩϩ)0 (ϩϩ)0 (ϩ)0 (ϩ)0 (ϩ)0 (ϩ)0 (ϩ)0 (ϩ)33 Serous (n ϭ 40) 0 15 70 18 3 43 (ϩϩ)0 (ϩϩ)5 (ϩϩ)47 (ϩϩ)0 (ϩϩ)0 (ϩϩ)5 (ϩ)0 (ϩ)10 (ϩ)23 (ϩ)18 (ϩ)3 (ϩ)38 Endometrioid (n ϭ 7) 0 0 86 14 0 57 (ϩϩ)0 (ϩϩ)0 (ϩϩ)57 (ϩϩ)14 (ϩϩ)0 (ϩϩ)0 (ϩ)0 (ϩ)0 (ϩ)29 (ϩ)0 (ϩ)0 (ϩ)57 Clear cell (n ϭ 5) 0206020040 (ϩϩ)0 (ϩϩ)0 (ϩϩ)60 (ϩϩ)0 (ϩϩ)0 (ϩϩ)0 (ϩ)0 (ϩ)20 (ϩ)0 (ϩ)20 (ϩ)0 (ϩ)40 a Number represent overall staining positivity. Staining is then stratified according to staining intensity: (ϩ), weak to medium staining; (ϩϩ), medium to intense staining. b M, membrane staining; C, cytoplasmic staining; MϩC, both cytoplasmic and membrane staining.

four major ovarian cancer subtypes [serous, mucinous, clear Table 1). This observation was applicable to all three sub- cell, and endometrioid (Fig. 3, A and B)] compared with types of ovarian cancers evaluated. Interestingly and unex- HOSE-B, a culture of ovarian surface epithelial cells. These pectedly, the majority of the samples exhibited a combination results are derived from 23 serous, 6 clear cell, 6 endometrioid, of cytoplasmic and membrane staining (Fig. 5; Table 1). and 4 mucinous carcinoma specimens. Real-time RT-PCR data Claudin-3 was never found exclusively at the membrane in confirmed our findings that CLDN3 and CLDN4 were not highly any of the samples, but it was found exclusively in the expressed in normal tissue, although some level of CLDN3 cytoplasm of 15% of the serous cases. Approximately 14– expression was detectable in lung, colon, breast, uterus, and 20% of the ovarian tumors of all subtypes exhibited claudin-4 placenta. CLDN4 expression was extremely low in most tissues staining exclusively at the membrane. Only 3% of serous examined, except in the lung, colon, and a few other tissues, cases had exclusive cytoplasmic staining of claudin-4. No where it was detectable at low levels. Importantly, CLDN3 and expression of claudin-3 was found in normal ovarian surface CLDN4 were also found at extremely low levels in ML3, an epithelium, whereas only one of three samples containing this ovarian cystadenoma cell line (Fig. 3, A and B, bar 16). normal ovarian tissue exhibited low levels of claudin-4. Claudin-3 and Claudin-4 Proteins Are Elevated in Overall, the experiments with the tissue array revealed that, Ovarian Cancer and Can Be Located at the Membrane or in depending on the subtype, 80– 84% of the ovarian cancers the Cytoplasm. To show that CLDN3 and CLDN4 were ele- were positive for claudin-3, whereas 60–71% were positive vated at the protein level, immunoblotting experiments were for claudin-4. Only 28–37% of the cystadenomas were performed. Claudin-3 and claudin-4 were readily detectable by weakly positive for claudin-3, and 12–14% were weakly immunoblotting at approximately M 22,000 (Fig. 4). Although r positive for claudin-4 (Fig. 6), indicating that overexpression claudin-3 and claudin-4 were completely absent on the nonmalignant surface epithelium sample HOSE-B, the amounts of of these proteins is associated with malignancy. This obser- claudin-3 and claudin-4 were elevated in most ovarian cancer vation is consistent with the RT-PCR results showing low cell lines, particularly in OVCAR-2. OVCA432 exhibited high levels of CLDN3 and CLDN4 in the cystadenoma cell line levels of claudin-3 and lower levels of claudin-4, consistent with ML3 (Fig. 3). the results obtained by Northern blotting (Fig. 2). Also consist- Claudin-3 and Claudin-4 Protein Expression Is Not ent with the Northern analysis, these proteins were not Associated with TJ Integrity. One major role of the TJs is to expressed in A2780. In contrast to claudin-3 and claudin-4, provide a seal between adjacent compartments, and the tightness claudin-1 was expressed in HOSE-B and found to be both of TJ is typically assessed by measuring TER across a mono- up-regulated and down-regulated in ovarian cancer cell lines. layer of confluent cells. To determine whether claudin-3 and This demonstrates that the expression patterns observed for claudin-4 expression was associated with TJ function, we de- claudin-3 and claudin-4 are specific for these claudin family termined TER for the various cell lines used in this study. members and do not represent a general regulation of all claudin Generally, ovarian cancer cell lines did not exhibit TER regard- proteins in ovarian cancer. less of claudin-3 and claudin-4 status (Table 2). This is best To extend these results to primary ovarian tissues, we exemplified by comparing OVCAR-3 and OVCAR-4, two cell constructed an ovarian tissue array (the tissue Ovarray) to lines with similar claudin expression (Fig. 4) but with very study claudin-3 and claudin-4 expression in ovarian tumors. different permeability (Table 2). OVCAR-4 and OVCAR-2 Claudin-3 and claudin-4 were generally found to be ex- exhibited TER, but the levels were low compared with MDCK, pressed in the majority of ovarian cancers examined (Fig. 5; a widely used model for TJ formation and function.

Table 2 Claudin expression and TER Cell lines CLDN3a CLDN4a TER (ohms.cm2) HOSE-B ϪϪ Ϫ 2008 ϩϩ Ϫ A2780 ϪϪ Ϫ OVCA432 ϩϩ ϩ Ϫ OVCAR-2 ϩϩϩ ϩϩϩ 52 OVCAR-3 ϩϩ ϩϩ Ϫ OVCAR-4 ϩϩ ϩϩ 269 OVCAR-5 Ϫϩ Ϫ MDCK Ϫϩϩ 2260 a Relative amount of protein detected by immunoblotting: Ϫ,no protein detected; ϩ, low amounts of proteins detected; and ϩϩ, large amounts of protein detected.

teins was rare in ovarian cystadenomas, clearly associating the presence of these proteins with malignancy. There is a possibility that claudin overexpression represents an epiphenomenon unrelated to the process of tumorigenesis. In this case, claudins may still represent useful clinical targets for drug delivery or for diagnosis. Because we and others find claudin-3 and claudin-4 up-regulated so frequently in various cancers, and considering the high selective pressure for advantageous gene expression within a tumor, we favor the hypothesis that claudin-3 and claudin-4 overexpression is important for ovarian cancer development. In this case, the role of claudin overexpression in ovarian cancer would be limited to two possibilities. First, claudin-3 and claudin-4, two proteins not normally present in ovarian tissues, may, when overexpressed, interfere with normal TJ formation and function. This dominant negative hypothesis is particularly interesting because it has previously been observed that overexpressing CLDN2 in MDCK cells reduces the tightness of TJ in these cells (32). It is possible that the exact ratios of various claudin proteins determine the permeability of TJs. Fig. 6 Immunohistochemical examination of ovarian cystadenomas. Representative staining with anti-claudin-3 and anti-claudin-4 antibod- Another possibility is that claudin-3 and claudin-4 may be ies of serous and mucinous cysadenoma is shown. Little or no staining required for signaling through important survival or prolifera- is observed. Specimens are counterstained with hematoxylin. Negative tive pathways in ovarian cancer, regardless of their role in TJ controls (without primary antibody) do not exhibit staining. permeability. To prove a functional role in ovarian tumorigenesis and to differentiate between these different possibilities, it will be necessary to generate stable transfectant models of normal and transformed ovarian cells overexpressing various DISCUSSION wild-type and dominant negative claudin constructs. These ex- TJs are crucial structures for normal epithelial cell home- periments are currently under way. ostasis. These structures appear to be an impediment to the It is intriguing that many ovarian tumor samples exhibit unrestricted growth and invasion phenotype observed in cancer. claudin-3 and claudin-4 staining in the cytoplasm. Aberrant For example, decrease or loss of TJ formation has been observed activation of specific pathways in ovarian cancer may be re- in colon cancer (12), hepatocellular carcinoma (29), and thyroid sponsible for claudin mislocalization and inappropriate signal- cancer (30). Several hypotheses have been suggested, but these ing or disruption of TJs. Interestingly, it has been observed that changes are believed to be crucial for the increased accessibility claudin proteins can be found in the cytoplasm as a result of of the tumor cells to nutrient (13), increased cell proliferation mitogen-activated protein kinase pathway or protein kinase C (16), and the increased motility associated with metastasis (14, activation (33, 34). These pathways will warrant further atten- 15). Down-regulation of occludin and other TJ proteins has been tion for their role in reorganizing TJs in ovarian cancer. It is suggested as a mechanism for TJ dismantling in cancer. important to point out that TJ disruption by itself may be In this context, it is intriguing to find claudin proteins responsible for claudin mislocalization. highly up-regulated in cancer because these proteins are gener- Analysis of CLDN3 and CLDN4 in various tumors using ally implicated in TJ formation and function (6, 10, 31). In this the SAGEmap resources suggests that overexpression of these study, we find claudin-3 and claudin-4 frequently up-regulated proteins may be a common phenomenon in many cancers. This in ovarian cancers of all subtypes. Up-regulation of these pro- is not surprising because epithelial cells face similar barriers to

uncontrolled proliferation. What are the roles of claudin-3 and 9. Tsukita, S., and Furuse, M. Pores in the wall: claudins constitute tight claudin-4 in cancer development? It is likely that a better un- junction strands containing aqueous pores. J. Cell Biol., 149: 13–16, derstanding of the roles of these claudins in normal tissue will 2000. help answer this question. More specifically, the colon, the only 10. Morita, K., Furuse, M., Fujimoto, K., and Tsukita, S. Claudin multigene family encoding four-transmembrane domain protein compo- normal tissue showing appreciable levels of CLDN3 and nents of tight junction strands. Proc. Natl. Acad. Sci. USA, 96: 511–516, CLDN4 expression, may provide an invaluable model for the 1999. study of these proteins. 11. Weinstein, R. S., Merk, F. B., and Alroy, J. The structure and Elucidation of the roles of CLDN3 and CLDN4 in ovarian and function of intercellular junctions in cancer. Adv. Cancer Res., 23: other cancers might provide new opportunity for therapy. Mecha- 23–89, 1976. nism-based targeting of the claudin pathway may inhibit cell pro- 12. Soler, A. P., Miller, R. D., Laughlin, K. V., Carp, N. Z., Klurfeld, liferation and metastasis. Although the mechanism relevant to D. M., and Mullin, J. M. Increased tight junctional permeability is claudin overexpression in ovarian cancer is still being investigated, associated with the development of colon cancer. Carcinogenesis (Lond.), 20: 1425–1431, 1999. this report unequivocally identifies claudin-3 and claudin-4 as 13. Mullin, J. M. Potential interplay between luminal growth factors biomarkers in ovarian cancer. This is particularly relevant, consid- and increased tight junction permeability in epithelial carcinogenesis. J. ering the fact that there have been multiple reports suggesting that Exp. Zool., 279: 484–489, 1997. the targeting of prostate and pancreatic cancer cells expressing 14. Ren, J., Hamada, J., Takeichi, N., Fujikawa, S., and Kobayashi, H. CLDN3 and CLDN4 with Clostridium perfringens enterotoxin, a Ultrastructural differences in junctional intercellular communication bacterial toxin that specifically binds these transmembrane proteins between highly and weakly metastatic clones derived from rat mammary (35, 36), may provide novel strategies to fight tumors resistant to carcinoma. Cancer Res., 50: 358–362, 1990. conventional therapy (22, 37). Considering the difficulties associ- 15. Martin, T. A., and Jiang, W. G. Tight junctions and their role in ated with ovarian cancer therapy, it is important to aggressively cancer metastasis. Histol. Histopathol., 16: 1183–1195, 2001. pursue alternative treatment strategies for this disease. 16. Wodarz, A. Tumor suppressors: linking cell polarity and growth control. Curr. Biol., 10: R624ϪR626, 2000. 17. Li, D., and Mrsny, R. J. Oncogenic Raf-1 disrupts epithelial tight ACKNOWLEDGMENTS junctions via downregulation of occludin. J. Cell Biol., 148: 791–800, 2000. We thank Dr. Thomas C. Hamilton for cell lines OVCAR-3, 18. Kimura, Y., Shiozaki, H., Hirao, M., Maeno, Y., Doki, Y., Inoue, OVCAR-4, and OVCAR-5. We are grateful to Dr. Robert C. Bast for M., Monden, T., Ando-Akatsuka, Y., Furuse, M., Tsukita, S., and cell line OVCA432 and to Dr. Michael Birrer for providing cell line Monden, M. Expression of occludin, tight-junction-associated protein, 2008. We also thank Dr. Michael Sutters for helpful comments on the in human digestive tract. Am. J. Pathol., 151: 45–54, 1997. manuscript and Cheryl Sherman-Baust for help with the real-time RT- 19. Kramer, F., White, K., Kubbies, M., Swisshelm, K., and Weber, PCR experiments. B. H. Genomic organization of claudin-1 and its assessment in heredi- tary and sporadic breast cancer. Hum. Genet., 107: 249–256, 2000. 20. Nacht, M., Ferguson, A. T., Zhang, W., Petroziello, J. M., Cook, REFERENCES B. P., Gao, Y. H., Maguire, S., Riley, D., Coppola, G., Landes, G. M., 1. Hough, C. D., Sherman-Baust, C. A., Pizer, E. S., Montz, F. J., Im, Madden, S. L., and Sukumar, S. Combining serial analysis of gene D. D., Rosenshein, N. B., Cho, K. R., Riggins, G. J., and Morin, P. J. expression and array technologies to identify genes differentially ex- Large-scale serial analysis of gene expression reveals genes differen- pressed in breast cancer. Cancer Res., 59: 5464–5470, 1999. tially expressed in ovarian cancer. Cancer Res., 60: 6281–6287, 2000. 21. Miwa, N., Furuse, M., Tsukita, S., Niikawa, N., Nakamura, Y., and 2. Schummer, M., Ng, V. L. V., Baumgarner, R. E., Nelson, P. S., Furukawa, Y. Involvement of claudin-1 in the ␤-catenin/Tcf signaling Schummer, B., Bednarski, D. W., Hassell, L., Baldwin, R. L., Karlan, pathway and its frequent upregulation in human colorectal cancers. B. Y., and Hood, L. Comparative hybridization of an array of 21,500 Oncol. Res., 12: 469–476, 2001. ovarian cDNAs for the discovery of genes overexpressed in ovarian 22. Long, H., Crean, C. D., Lee, W. H., Cummings, O. W., and Gabig, carcinomas. Gene (Amst.), 238: 375–385, 1999. T. G. Expression of Clostridium perfringens enterotoxin receptors clau- 3. Shridhar, V., Lee, J., Pandita, A., Iturria, S., Avula, R., Staub, J., din-3 and claudin-4 in prostate cancer epithelium. Cancer Res., 61: Morrissey, M., Calhoun, E., Sen, A., Kalli, K., Keeney, G., Roche, P., 7878–7881, 2001. Cliby, W., Lu, K., Schmandt, R., Mills, G. B., Bast, R. C., Jr., James, 23. Gregoire, L., Munkarah, A., Rabah, R., Morris, R. T., and Lan- C. D., Couch, F. J., Hartmann, L. C., Lillie, J., and Smith, D. I. Genetic caster, W. D. Organotypic culture of human ovarian surface epithelial analysis of early- versus late-stage ovarian tumors. Cancer Res., 61: cells: a potential model for ovarian carcinogenesis. In Vitro Cell Dev. 5895–5904, 2001. Biol. Anim., 34: 636–639, 1998. 4. Ismail, R. S., Baldwin, R. L., Fang, J., Browning, D., Karlan, B. Y., 24. Luo, M. P., Gomperts, B., Imren, S., DeClerck, Y. A., Ito, M., Gasson, J. C., and Chang, D. D. Differential gene expression between Velicescu, M., Felix, J. C., and Dubeau, L. Establishment of long-term normal and tumor-derived ovarian epithelial cells. Cancer Res., 60: in vitro cultures of human ovarian cystadenomas and LMP tumors and 6744–6749, 2000. examination of their spectrum of expression of matrix-degrading pro- 5. Sawiris, G. P., Sherman-Baust, C. A., Becker, K. G., Cheadle, C., teinases. Gynecol. Oncol., 67: 277–284, 1997. Teichberg, D., and Morin, P. J. Development of a highly specialized 25. Hough, C. D., Cho, K. R., Zonderman, A. B., Schwartz, D. R., and cDNA array for the study and diagnosis of epithelial ovarian cancer. Morin, P. J. Coordinately up-regulated genes in ovarian cancer. Cancer Cancer Res., 62: 2923–2928, 2002. Res., 61: 3869–3876, 2001. 6. Tsukita, S., Furuse, M., and Itoh, M. Multifunctional strands in tight 26. Folz, R. J., and Nepluev, I. Poly(A) cDNA-specific (PACS) RT- junctions. Nat. Rev. Mol. Cell. Biol., 2: 285–293, 2001. PCR: a quantitative method for the measurement of any poly(A)- 7. Anderson, J. M. Molecular structure of tight junctions and their role containing mRNA not affected by contaminating genomic DNA. Bio- in epithelial transport. News Physiol. Sci., 16: 126–130, 2001. techniques, 29: 762–768, 2000. 8. Zahraoui, A., Louvard, D., and Galli, T. Tight junction, a platform 27. Lal, A., Lash, A. E., Altschul, A. F., Velculescu, V. E., Zhang, L., for trafficking and signaling protein complexes. J. Cell Biol., 151: McLendon, R. E., Marra, M. A., Prange, C., Morin, P. J., Papadopoulos, F31ϪF36, 2000. N., B., V., Kinzler, K. W., Strausberg, R. L., and Riggins, G. J. A public

database for quantitative gene expression analysis in human cancers. 33. Lippoldt, A., Jansson, A., Kniesel, U., Andbjer, B., Andersson, A., Cancer Res., 59: 5403–5407, 1999. Wolburg, H., Fuxe, K., and Haller, H. Phorbol ester induced changes in 28. Boon, K., Oso´rio, E. C., Greenhut, S. F., Schaefer, C. F., Shoe- tight and adherens junctions in the choroid plexus epithelium and in the maker, J., Polyak, K., Morin, P. J., Buetow, K., Strausberg, R. L., de ependyma. Brain Res., 854: 197–206, 2000. Souza, S. J., and Riggins, G. J. An anatomy of normal and malignant 34. Chen, Y., Lu, Q., Schneeberger, E. E., and Goodenough, D. A. gene expression. Proc. Natl. Acad. Sci. USA, 99: 11287–11292, 2002. Restoration of tight junction structure and barrier function by down- 29. Swift, J. G., Mukherjee, T. M., and Rowland, R. Intercellular junctions regulation of the mitogen-activated protein kinase pathway in ras- in hepatocellular carcinoma. J. Submicrosc. Cytol., 15: 799–810, 1983. transformed Madin-Darby canine kidney cells. Mol. Biol. Cell, 11: 30. Cochand-Priollet, B., Raison, D., Molinie, V., Guillausseau, P. J., 849–862, 2000. Wassef, M., and Bouchaud, C. Altered gap and tight junctions in human 35. Fujita, K., Katahira, J., Horiguchi, Y., Sonoda, N., Furuse, M., and thyroid oncocytic tumors: a study of 8 cases by freeze-fracture. Ultra- Tsukita, S. Clostridium perfringens enterotoxin binds to the second struct. Pathol., 22: 413–420, 1998. extracellular loop of claudin-3, a tight junction integral membrane 31. Furuse, M., Fujita, K., Hiiragi, T., Fujimoto, K., and Tsukita, S. protein. FEBS Lett., 476: 258–261, 2000. Claudin-1 and -2: novel integral membrane proteins localizing at tight 36. McClane, B. A. Clostridium perfringens enterotoxin and intestinal junctions with no sequence similarity to occludin. J. Cell Biol., 141: 1539–1550, 1998. tight junctions. Trends Microbiol., 8: 145–146, 2000. 32. Furuse, M., Furuse, K., Sasaki, H., and Tsukita, S. Conversion of 37. Michl, P., Buchholz, M., Rolke, M., Kunsch, S., Lohr, M., Mc- zonulae occludentes from tight to leaky strand type by introducing Clane, B., Tsukita, S., Leder, G., Adler, G., and Gress, T. M. Claudin-4: claudin-2 into Madin-Darby canine kidney I cells. J. Cell Biol., 153: a new target for pancreatic cancer treatment using Clostridium perfrin- 263–272, 2001. gens enterotoxin. Gastroenterology, 121: 678–684, 2001.

Downloaded from clincancerres.aacrjournals.org on September 27, 2021. © 2003 American Association for Cancer Research. Tight Junction Proteins Claudin-3 and Claudin-4 Are Frequently Overexpressed in Ovarian Cancer but Not in Ovarian Cystadenomas

Leticia B. A. Rangel, Rachana Agarwal, Theresa D'Souza, et al.

Clin Cancer Res 2003;9:2567-2575.

Updated version Access the most recent version of this article at: http://clincancerres.aacrjournals.org/content/9/7/2567

Cited articles This article cites 35 articles, 16 of which you can access for free at: http://clincancerres.aacrjournals.org/content/9/7/2567.full#ref-list-1

Citing articles This article has been cited by 31 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/9/7/2567.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/9/7/2567. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.