The first 2 authors contributed equally to this paper.
Cancer Cell Biology
Reexpression of the TJ protein CLDN1 induces apoptosis in breast tumor spheroids
Article first published online: 28 OCT 2003
Copyright © 2003 Wiley-Liss, Inc.
International Journal of Cancer
Volume 108, Issue 3, pages 374–383, 20 January 2004
How to Cite
Hoevel, T., Macek, R., Swisshelm, K. and Kubbies, M. (2004), Reexpression of the TJ protein CLDN1 induces apoptosis in breast tumor spheroids. Int. J. Cancer, 108: 374–383. doi: 10.1002/ijc.11571
- Issue published online: 21 NOV 2003
- Article first published online: 28 OCT 2003
- Manuscript Accepted: 8 AUG 2003
- Manuscript Revised: 21 JUN 2003
- Manuscript Received: 24 JAN 2003
- University of Washington Medical Center Breast program Tennis Tournament
- breast cancer;
- tight junction;
- spheroid culture
Members of the claudin family together with occludin are the major constituents of the tight junction (TJ) complex. The human homologue of the murine CLDN1, previously called SEMP1, was identified by differential expression analysis, and the CLDN1 mRNA was found to be downregulated or completely lost in human breast cancer cells in vitro. Retroviral-induced CLDN1 reexpression in breast cancer cells results in plasma membrane homing of the protein and reconstitution of paracellular flux inhibition, which is not dependent on the presence of occludin protein. In this report, we investigated the physiologic role of CLDN1 in CLDN1-transduced MDA-MB 361 breast tumor cells in adherent 2D and suspension 3D spheroid cell cultures. Retroviral-transduced bulk cultures were FACS-sorted to enrich for 100% CLDN1-positive clonal derivatives with similar expression levels of CLDN1 mRNA and protein. There was no difference in proliferation and cell death characteristics in 2D adherent cell cultures of CLDN1-positive compared to control CLDN1-negative and mock-transduced cell cultures. In contrast, the majority of the CLDN1-transduced derivatives displayed a significant elevation of apoptosis that became evident as early as 2 days after 3D spheroid culture onset. This elevated apoptosis was independent of the volume of established spheroids. The cellular immunofluorescence analysis of CLDN1 protein expression in transduced bulk cultures revealed a CLDN1-positive subfraction with a heterogeneous pattern of membrane and cytosolic immunostaining. In the clonal MDA-MB 361 CLDN1-positive cultures, we found that a more prominent cell membrane localization correlated with a pronounced increase of apoptosis in tumor spheroids. In parallel, inhibition of the paracellular flux rate was observed. These findings support a potential role of the TJ protein CLDN1 in restricting nutrient and growth factor supplies in breast cancer cells, and they indicate that the loss of the cell membrane localization of the TJ protein CLDN1 in carcinomas may be a crucial step during tumor progression. © 2003 Wiley-Liss, Inc.
Solid tumor development is complex and encompasses alterations in a variety of physiologic processes such as enhanced proliferative capacity, resistance toward apoptosis, metastatic potential and escape from immunologic detection. However, beside these diverse and multiple changes occurring in tumors, cellular accessibility of nutrients and growth factors must be present continuously. To ensure early neoplastic survival, the onset of neovascularization is one of the most significant hallmarks during tumor progression.1, 2 One rate-limiting step of growing tumors may be the diffusion of molecules into the paracellular space. Therefore, it has been hypothesized that the presence of intact functional tight junctions (TJs) in tumors of epithelial origin might limit the optimal supply of biomolecules necessary for cell growth and survival.3, 4 In fact, several reports provide evidence for the relevance of functional loss of TJs in developing human tumors in vivo. Diminished TJ formation has been reported for hepatocellular carcinoma,5 thyroid tumors6 and colon carcinoma.7 In experimental animal models, alterations of TJs have been observed in canine bladder carcinoma8 and murine mammary carcinoma.9 Furthermore, the loss of TJ-associated molecules such as ZO-1 has been correlated with tumor progression.10, 11 Finally, the relevance of TJ complexes in tumorigenesis is highlighted by the fact that loss of TJ-associated tumor suppressor genes also results in tumor formation in imaginal disks of Drosophila.12, 13
TJs constitute the barrier of paracellular flux in epithelia and control the permeation of molecules between the apical lumen and basolateral face. Identification of the multiple proteins that constitute the epithelia- or tissue-specific TJs have only recently been elucidated.14, 15, 16, 17, 18, 19 One of the first molecules associated with the TJ, occludin, was identified as a requisite integral protein.20 However, in the murine occludin knockout TJ-like structures were still present. The 4-transmembrane proteins claudin-1 and -2 were then shown to reconstitute the TJ.21 It has been suggested that claudin-1 and -2 might recruit occludin to TJs,22 and both claudins have been shown to interact with intracellular TJ-associated proteins from the membrane-associated guanylate kinase (MAGUK) family. Sequence and similarity analysis have revealed the existence of a claudin family with up to more than 20 proteins,15, 17, 23 but for only a few members of the claudin family has a specialized function been identified.24, 25, 26, 27, 28, 29, 30
In parallel with the identification of the murine claudin-1 and -2, we identified the human counterpart of the murine claudin-1, previously called SEMP1, in a differential display approach comparing isogenic exponentially growing, nontumorigenic and senescent human epithelial breast cells.26 The human claudin-1 (CLDN1), which shares a 98% amino acid homology to the murine claudin-1, is expressed in many human tissues containing epithelium, such as mammary gland. Of particular interest in this regard is that in most breast cancer cell lines, claudin-1 mRNA expression was absent or significantly downregulated.26 The loss of expression in breast cancer cell lines is not due to genetic mutations in the human CLDN1 gene, neither in sporadic nor in familial breast cancer compared to normal controls.31 By cellular immunofluorescence analysis of CLDN1 retroviral-transduced breast cancer cells, we detected homing of the protein at cell-cell contact sites in confluent cell layers even in the absence of occludin.32 Finally, as shown in MDCK cells,33 we confirmed the attenuation of paracellular flux due to exogenous CLDN1 expression in breast cancer cells in an occludin-negative background.32
We hypothesized that the diminished paracellular flux obtained in CLDN1-transduced breast cancer cells would result in impairment of proliferation. To evaluate this, we investigated cell proliferation and apoptosis in CLDN1-transduced MDA-MB 361 breast cancer cells cultured as a monolayer and as cell aggregates, or spheroids, in 3D suspension cultures. In this report, we show that retroviral-induced expression of CLDN1 in breast tumor cells does not alter cell proliferation or apoptosis rates in monolayer cultures. However, when the cells are cultured as tumor spheroids, significant apoptosis is induced in CLDN1-transduced MDA-MB 361 cells, supporting the hypothesis that the limited nutrient supply is an antitumorigenic property of functional TJs.3, 4
MATERIAL AND METHODS
Breast cancer cell lines T47-D and MDA-MB 361 were cultivated as monolayers or spheroids in RPMI 1640 supplemented with 10% FCS, 2 mM L-glutamine, 1 × nonessential amino acids and 1 mM sodium pyruvate. The amphotropic retrovirus-producing AM12 cells were cultured in DMEM basal medium with identical supplements as above. Monolayers cultures were cultivated in T75 cm2 flasks (Falcon). For the generation of MDA-MB 361 spheroids, the cells were seeded at 4 × 104 cells/ml into 24-well plates covered with poly-HEMA.34 Fresh culture medium was substituted twice a week.
The CLDN1 retrovirus was derived from a Fanconi group C retrovirus containing the novel l-NGFR cell surface marker as a reporter molecule driven by the 5′ LTR promotor to identify the transduced cells.35 The downstream-located SV40 promotor was substituted with a CMV promotor, which controls CLDN1 expression. The full-length human CLDN1 cDNA was cloned as the open reading frame (ORF). Infectious amphotropic retroviruses were obtained as described previously.32 To obtain high-titer supernatants, fresh medium was added to subconfluent AM12 cultures, and the virus containing supernatant used for breast cancer cell transduction was harvested 24 hr later.35
CLDN1-transduced breast cancer cells
Retroviral CLDN1 transduction of the MDA-MB 361 breast cancer cells was performed by incubation of a subconfluent culture for 24 hr with conditioned supernatant of the CLDN1 retroviruses produced by AM12 cells. From this heterogeneous CLDN1 expressing bulk culture, pure CLDN1-positive MDA-MB 361 clones were derived by FACS sorting of l-NGFR-positive cells using an anti-l-NGFR FITC-labeled antibody (FACS Vantage, Becton Dickinson, San Jose, CA). Mock-transduced cells were obtained via transduction with the vector harboring the l-NGFR reporter gene only.
Quantitative CLDN1 RT-PCR
The quantitative determination of CLDN1 mRNA expression by RT-PCR was performed using the LightCycler (Roche, Mannheim, Germany) as described previously.32 The cell samples were lysed under denaturing conditions with guanidine isothiocyanate and β-mercaptoethanol. Total RNA was isolated subsequently using a silica gel-based membrane for RNA binding (Qiagen, Chatsworth, CA). The quantity of CLDN1-specific mRNA was calculated from a standard curve. Template DNA for the CLDN1 RNA standard was produced by cloning of the ORF of CLDN1 into the pCR 2.1 TOPO vector (Invitrogen, La Jolla, CA) under the control of the T7 promotor. CLDN1 mRNA was transcribed using T7 RNA Polymerase (Roche) and subsequently purified using Roche RNA minispin columns after complete DNAse digestion (Roche). Concentration of RNA standards was determined using the Ribo Green RNA Quantification kit (Molecular Probes, Eugene, OR). The primer design for RT-PCR reactions was performed by selection of the optimal melting point, minimal primer dimer and internal secondary structures (sequences: p23quant 5′-ATGGCCAACGCGGGGC-3′ and p23quantrev 5′-ACGTAGTCTTTCCCGCTG-3′). The RT-PCR reactions were performed with 5 mM MgCl2.
Western blot analyses
Western blot analysis of CLDN1 was performed by standard procedures as described elsewhere.32 In brief, cells were lysed in RIPA buffer (Roche). The protein lysates were separated on a 14% Tris-glycine gel (Novex), and PVDF membranes blotted from protein gels were incubated with a monoclonal anti-CLDN1 antibody. The CLDN1-specific antibody that recognizes a C-terminal epitope was obtained by DNA immunization as described previously.32 The anti-CLDN1 signals were detected using standard chemiluminescence Western blotting kit protocols (Roche). The analyses of occludin and ZO-1 protein expression were performed as described above using 10% and 4% Tris-glycine gels, respectively, blotting the proteins onto nitrocellulose membranes. The anti-occludin (Zymed, San Francisco, CA) and anti-ZO-1 (Zymed) polyclonal antibodies were both used at 1:2,000 dilutions. The primary antibodies were detected by 0.02 U/ml of a secondary antimouse IgG-peroxidase (POD) antibody (Roche).
MDA-MB 361 cells were cultured on microscope slides coated with collagen I (Falcon, Lincoln Park, NJ) in standard medium to high cell densities. After removal of the supernatant, the cells were washed with cold PBS and fixed subsequently with cold 1% paraformalde-hyde/PBS for 1 hr at room temperature. Permeabilization was achieved by a 1-hr incubation in 0.1% Triton X100/PBS. Thereafter, indirect immunofluorescence staining was performed with a murine monoclonal antibody against the human CLDN1 followed by an incubation with antimouse F(ab)2 labeled with Alexa488. Microscopic images were recorded with a Xillix camera adapted onto a Zeiss Axiphot microscope (10× objective) using the Openlab image acquisition software (Improvision, Lexington, MA). The total areas of spheroids and the propidum iodide (PI)-positive regions within the spheroids were obtained by region of interest selection and pixel quantification. The microscopic calibration of pixels to μm was performed using a micrometer standard grid plate (Zeiss, Thornwood, NY). Approximation of the corresponding volumes was performed assuming an ideal sphere shape of the spheroids.
The apoptotic cell death of MDA-MB 361 cells was analyzed by the annexin V-based detection of the translocation of phosphatidylserine from the inner side of the plasma membrane to the outer membrane layer. The adherent or in suspension growing MDA-MB 361 cells were trypsinized, washed once with PBS and stained with annexin V-Fluos (Roche). Dead cells were identified by propidium iodide counterstaining. The multiparameter FACS analysis was performed with a FACScan flow cytometer (Becton Dickinson).
Cell cycle analysis
The MDA-MB 361 spheroids were trypsinized, pelleted and resuspended at 5 × 105 cells/ml in DNA-staining buffer (100 mM Tris, pH7.4, 154 mM NaCl, 1 mM CaCl2, 0.5 mM MgCl2, 0.2% BSA, 0.1% NP40, 10 μg/ml propidium iodide) containing 10 U/ml RNAse A. After a 30-min staining period at 4°C, the cell cycle distribution was analyzed using a FACScan flow cytometer (Becton Dickinson) applying peak versus width recording to exclude cell clumps.
Expression of TJ-associated proteins
To investigate the in vitro physiologic function of CLDN1 in breast cancer cells, the CLDN1-negative MDA-MB 361 cells were transduced with the CLDN1/l-NGFR retrovirus. As evident from the CLDN1 RT-PCR expression analysis shown in Table I, the untransduced parental cell line and the mock-transduced MDA-MB 361 cells were negative for CLDN1 mRNA expression. Despite the fact that some heterogeneity of mRNA expression levels exists between the different cell cultures, all CLDN1 retroviral-transduced cell cultures, bulk culture and clonal derivatives exhibit significant expression of CLDN1 mRNA. The CLDN1 mRNA expression ranges from 2.3 to 6.8 pg/μl cell fraction used in the qRT-PCR, which corresponds approximately to 0.009% and 0.027% CLDN1-specific mRNA out of the total cellular RNA for clones 4 and 2, respectively.
|MDA-MB 361 cell culture||CLDN1 mRNA (pg/μl)||% of total RNA|
|CLDN1-transduced clone 1||5.1||0.020|
|CLDN1-transduced clone 2||6.8||0.027|
|CLDN1-transduced clone 3||6.0||0.024|
|CLDN1-transduced clone 4||2.3||0.009|
|CLDN1-transduced clone 5||2.5||0.010|
|CLDN1-transduced clone 6||4.7||0.019|
We have shown recently that in subconfluent CLDN1-transduced MDA-MB 361 cells, despite high-level mRNA, no CLDN1 protein could be detected at the cytoplasmic membrane.32 Therefore, we investigated the expression of CLDN1 protein and the TJ-associated proteins occludin and ZO-1 by Western blot analysis (Fig. 1). The T47-D breast cancer cell line was used as control since it displays constitutive expression of 3 TJ-associated proteins, claudin-1, occludin and ZO1. Neither the untransduced parental nor the mock-transduced MDA-MB 361 cells express CLDN1 protein. The CLDN1-transduced MDA-MB 361 bulk culture exhibits a low level of CLDN1 protein similar to that seen in CLDN1-transduced MDA-MB 361 clones 1, 4 and 5. Relative levels of CLDN1 mRNA and protein are similar in most clones with the exception of clones 4 and 5, which display low mRNA levels. ZO-1 protein was found in T47-D controls, as well as in all MDA-MB 361 cell cultures. Relative ZO-1 expression levels appeared to be independent of differential CLDN1 protein levels (Fig. 1). Occludin was found only in T47-D control cells and was absent in all CLDN1-negative and CLDN1-transduced MDA-MB 361 cells.
Of utmost physiologic importance for the function of CLDN1 is plasma membrane homing. We have shown recently that CLDN1-negative breast cancer cells are still capable of CLDN1 plasma membrane localization following retroviral CLDN1 transduction.32 Since the quantitative mRNA and protein expression of CLDN1 vary slightly between different cell cultures and clones, we investigated the localization of CLDN1 protein by immunofluorescence analysis. Using a monoclonal antihuman CLDN1-specific antibody, we show that the mock-transduced MDA-MB 361 cells are negative for CLDN1 protein. In the MDA-MB 361 CLDN1-transduced bulk culture, about one-half of the cell population stains positively using the antibody, while the remaining half is negative for CLDN1 protein (Fig. 2). Moreover, Figure 2 reveals that in addition to the membrane staining, some CLDN1-positive cells show significant cytoplasmic labeling. In order to obtain cell lines with consistent CLDN1 expression, CLDN1-positive MDA-MB 361 clonal cultures were generated by FACS cloning. As evident in Figure 2, all clones show positive staining with the anti-CLDN1 antibody in most cells in comparison to the CLDN1-transduced bulk culture. However, there was heterogeneity in the levels of cytoplasmic versus plasma membrane labeling between the different clones. The MDA-MB 361 CLDN1-transduced clones 1, 2, 4 and 5 show some diffuse cytoplasmic punctuated staining in addition to membrane labeling similar to cells of the bulk culture. In contrast, clones 3 and 6 show a prominent and more intense membrane labeling indicative of cell populations with a high rate of CLDN1 membrane homing. This localization pattern is similar to confluent T47-D cells exhibiting membrane expression of CLDN1.32
Induction of apoptosis in monolayer and spheroid cultures
The possibility that cell proliferation and cell death were affected following introduction of CLDN1 into breast tumor cells was investigated, comparing the untransduced control, the mock-transduced and CLDN1-transduced bulk cultures of MDA-MB 361 cells. In adherent 2D cell cultures, flow cytometric analyses were performed for all 3 cell populations, and these studies revealed a similar proliferation rate and cell cycle distributions of the subconfluent cultures with about 25% of the cells in the S/G2M fraction. Typical microscopic images of the mock- and CLDN1-transduced monolayer cultures consisting of dense confluent and subconfluent areas are shown in Figure 3. In the mock-transduced MDA-MB 361 cells, only a few nuclei stain positive with PI (dead cells). Neither a significant alteration of the cellular 2D morphology nor an increase of the dead cell fraction was evident in the CLDN1-transduced population. Typically, between 5 and 15 PI-positive nuclei were present per image in both cell cultures representing 1–3% of dead cells in each of the populations. In addition, the evaluation of the sub-G1 peak fraction of the cell cycles indicative of cell death revealed no differences comparing the mock- and CLDN1-transduced cells cultures. The major significant difference identified was the paracellar flux kinetics. Studies of confluent monolayer cell cultures using a 40 kD POD protein revealed a significant 60–80% lower flux rate within the first 2 hr of the CLDN1-positive compared to the mock-transduced breast cancer cells (data not shown; see also Hoevel et al.32).
To investigate whether CLDN1 could influence apoptosis or cell death by necrosis, we established a 3D spheroid cell model for the different MDA-MB 361 cell cultures.36, 37 When seeded in suspension as described above, within 5 hr all MDA-MB 361 populations rapidly established small aggregates of 3 to 4 cells. In the untransduced, mock-transduced bulk cultures and all CLDN1 clones, these small cell clusters aggregated further and proliferated into spheroids of about 20 to 50 cells in diameter after 4–6 days in suspension culture. Figure 4 displays 3 different typical spheroids of each MDA-MB 361 mock- and CLDN1-transduced cells cultured for 6 days in culture. From this bright-field microscopic image, it is evident that the CLDN1-transduced cells generate slightly tighter spheroids compared to the mock-transduced cells. There was no evidence of differences in size of the spheroids at any time comparing controls (mock-transduced) and the CLDN1-transduced counterparts. However, PI staining of the dead cells within the spheroids revealed a significantly increased number of dead cells of the CLDN1-transduced spheroids compared to the mock-transduced control. The extensive PI staining is located primarily in the core region of the CLDN1-positive MDA-MB 361 spheroids (Fig. 4).
To investigate whether these PI-positive dead cells were the result of necrotic or apoptotic cell death, analysis of spheroid cells was performed as early as 4 days after culture onset applying spheroid dissociation by trypsinization and subsequent incubation with annexin V-Fluos and propidium iodide. Figure 5. shows that only a few percent became apoptotic in the 3D aggregates of the control culture (5.7%) and mock-transduced population (5.2%). Even in the CLDN1-transduced MDA-MB 361 bulk culture, there was no evidence of a significant elevation of an apoptotic fraction (Fig. 5, top right). However, in the CLDN1-transduced MDA-MB 361 clonal cell populations, there was a significant increase of apoptosis as evident from the cell populations in the lower and upper right quadrants of each panel (viable apoptotic and dead apoptotic), with percentages of total apoptotic cells ranging from 7.2% (clone 4) and 23.5% (clone 5) to 45.2% (clone 6). In 5 out of 6 clones, the apoptosis observed increased significantly, with clone 4 exhibiting the least increment. Figure 6 summarizes the quantitative changes after a 4-day spheroid culture. Except for clone 4, the induction of apoptosis was at least more than 2-fold and up to 7-fold in the CLDN1-positive cell culture clones 1, 2, 3, 5 and 6.
The induction of apoptosis was investigated in further detail in spheroid cultures for up to 24 days comparing the control cells with the MDA-MB 361 cells clone 6, which exhibits the highest apoptosis induction (Fig. 6). As shown in Figure 7, the untransduced (solid line) and the mock-transduced (dotted line) control cultures display a slight increase of apoptosis, ranging from 12% to 17% throughout the observation period. The most rapid and highest elevation of apoptotic cells was present again in MDA-MB 361 clone 6 cells. After an 8-day culture period, the percentage of apoptotic cells increased to 50% of the total population and reached a plateau phase with about 60–65% of apoptotic cells from day 10 through day 24. Although the cell aggregates of all populations reached similar spheroid sizes, dead cells were present in the inner layers of the spheroids as detected by propidium iodide staining (Fig. 4; CLDN1-positive cells).
Formally, the differences of the apoptotic cell fractions of mock- and CLDN1-transduced MDA-MB 361 spheroids could be due to differences of the spheroid sizes. Therefore, we quantitated the volume spheroid sizes and their corresponding apoptotic cell fraction present in the PI-positive areas. Figure 8 shows that with increasing spheroid size, the apoptotic areas increase in a near linear fashion. The corresponding correlation coefficients for the CLDN1 and mock-cell spheroids are 0.97 and 0.91, respectively. The slope of the CLDN1-positive MDA-MB 361 spheroids is approximately 3-fold higher compared to the mock-control (0.46 vs. 0.14), indicating a 3-fold increased apoptosis rate at any spheroid size. In addition, independent of the volume of the spheroids, the PI-positive fraction in the spheroid core was always higher for the CLDN1-transduced cells. This excludes the possibility that the increased apoptotic cell fraction in the CLDN1-positive spheroids might be a secondary phenomenon of larger spheroid sizes.
The evaluation of the proliferation rate of the mock- and claudin-1-transduced MDA-MB 361 cells by flow cytometric analysis revealed no significant difference of the S- and G2M subpopulations. After a 6-day spheroid culture, the G1, S and G2M cell fractions were 81.9%, 9.0% and 9.1% for the mock-transduced populations, and for the CLDN1-positive cells the numbers were 80.1%, 7.9% and 11.9%, respectively. Similar cell cycle distributions were found at later time points (data not shown).
We recently identified the human CLDN1 gene by a differential expression approach and found evidence for a significant loss of expression or downregulation in breast cancer cell lines.26 The loss of CLDN1 expression in breast cancer is likely not due to genetic mutations in the promotor region of CLDN1 or the gene itself.31 In addition, we have shown that after retroviral CLDN1 transduction of CLDN1-negative breast tumor cells, the cells subsequently maintain CLDN1 plasma membrane homing despite the absence of occludin.32 Evidence for the altered function of TJs in primary tumors may initially be examined by immunochemistry. It has been demonstrated, for example, that TJ function is impaired in solid tumors, including colon cancer, thyroid cancer and hepatocellular carcinomas.5, 6, 7 For colon cancer development, it has been shown that alteration of TJ permeability is an early event as the leakiness is evident in adenomatous polyps prior rise of colon cancer cells.7 In the case of human breast cancer cell lines with the absence of CLDN1, we hypothesized that loss of expression may result in increased paracellular flux in epithelial carcinogenesis,32 enabling uncontrolled and unlimited access of trophic factors to tumor cells.3, 4 To identify the physiologic function of CLDN1 in MDA-MB 361 breast tumor cells, we compared in vitro 2D models, mimicking single epithelial cell layers, with multilayer 3D spheroids.
We found minor variation of CLDN1 mRNA and protein expression in most CLDN1-transduced MDA-MB 361 cell clones grown to confluency (Table I and Fig. 2). When cultivated as 2D monolayers cultures, the reexpression of CLDN1 did not affect cell proliferation or apoptosis compared to the mock- and bulk cultures. However, among the CLDN1-positive clones, there was significant difference of CLDN1 homing into the cell membrane. Clones 3 and 6 exhibited the highest level of CLDN1 membrane localization compared to both membrane and cytoplasmic homing within the other clones, including the bulk culture (Fig. 2). The spheroids from clones with prominent membrane homing displayed a more compact 3D morphology compared to the CLDN1-negative counterparts (Fig. 4), which indicates a possible role of CLDN1 in intercellular adhesion.38
The different MDA-MB 361 cell cultures, however, exhibited striking difference in the induction of apoptosis when grown as 3D spheroids. In the CLDN1-transduced bulk culture that exhibits CLDN1-positive and -negative cells (Fig. 2), we detected no enhancement of apoptosis compared to the mock-transduced or untransduced MDA-MB 361 cultures (Fig. 5). The cell cultures displaying the highest and most homogeneous incorporation of CLDN1 into the cell membrane (clones 3 and 6; Fig. 2) showed the most striking level of apoptosis (Figs. 5 and 6), as early as 4 days in spheroid cultures. An induction of apoptosis was observed for 5 out of 6 clones; even clone 5, which had low CLDN1 mRNA and protein expression, exhibited a significant increase of cell death in spheroid culture (Fig. 6). Therefore, neither the quantitative expression of mRNA nor the protein level of CLDN1 is predictive for apoptosis induction, which indicates that the apoptosis function appears to be determined by the protein localization of CLDN1 in the cell membrane of the breast cancer cells. This could explain the paradoxical finding of CLDN1 overexpression in colon cancer compared to normal tissue, in which immunostaining with a polyclonal anti-CLDN1 antibody reveals both diffuse cytoplasmic and membrane staining.39 Our data obtained with breast cancer cells with different CLDN1 homing pattern offer evidence for the suggestion that aberrant subcellular localization or abnormal formation of tight junctions are even as important as downregulation or loss of CLDN1 gene expression.26, 32
Recently, it was shown that CLDN1 expression suppresses paracellular flux, as had been reported for the TJ protein occludin.32, 33 In this report, we provide additional evidence that the expression and membrane homing of exogenously supplied CLDN1, which inhibits the paracellular flux in MDA-MB 361 CLDN1 cells, is paralleled by an increased rate of apoptosis in 3D spheroid cultures (Figs. 5 and 7). The prominent role of CLDN1 in the induction of apoptosis and the inhibition of the paracellular flux in MDA-MB 361 breast cancer cells is highly significant as MDA-MB 361 cells lack another integral membrane protein, namely, occludin (Fig. 1). Our data indicate that expression and membrane homing of CLDN1 decreases in vitro molecule permeability and might therefore also impair nutrient and growth factor supply in spheroids as well as in solid tumors. It remains to be determined whether the TJ structures generated by CLDN1 are sufficient itself for the inhibition of the flux rates or if other TJ-associated proteins act as partners to reconstitute the TJ complex in MDA-MB 361 breast cancer cells. CLDN2 may possibly be excluded as a partner since it has been shown recently in MDCK cells that reexpression of CLDN2 in these CLDN1-positive cell surprisingly increase the paracellular flux rate,40 and we find that MDA-MB 361 cells fail to express CLDN2 protein (data not shown). In addition, in the proximal nephron in mouse kidney exhibiting a leaky epithelium, CLDN2 is selectively expressed.41
In this report, we provide evidence that loss of expression of CLDN1 or altered homing pattern from the membrane to the cytosol in breast cancer cells would be advantageous for their survival and growth in 3D spheroids. We suggest that this may also pertain to primary or metastatic breast tumors. The hypothesis of Mullin that the loss of normal TJ function may contribute to tumorigenesis is supported by our findings.3, 4 The maintenance of barrier function of TJs in developing tumors would inhibit the supply of nutrients, growth factors and other biomolecules to cells in the inner mass to such an extent that tumors may not be capable of growth beyond a critical size because of the limited diffusion of molecules, for example, by transcytosis or passive diffusion. This suggestion is underlined by the fact that tumor cells in vivo begin neovascularization at very small sizes to enable supply of all necessary biomolecules necessary for survival.1, 2 Figure 8 provides evidence for this hypothesis, wherein increased levels of apoptosis in the CLDN1-positive spheroids is proportional to the spheroid size and present at volumes as small as 10 × 106 μm.3.3 This corresponds to a sphere diameter of about 200 μm3 or 15–20 cells in length. Even within such small tumors, initiation of vascularization may already be observed due to the limitation of diffusion or transcytosis.1 In addition to vascularization, the survival of tumors might also be enhanced by the loss of membrane expression of tight junction molecules. Further evidence is provided by our CLDN1-transduced bulk culture, which contains a mosaic of CLDN1-positive and -negative cells. Thus, fewer functional CLDN1 TJs may be present in 2D monolayer as well as 3D spheroid cultures that enable a sufficient supply of biomolecules and nutrients for all cells with no sign of apoptosis.
In addition to the nutrient supply hypothesis, formally an apoptosis-related signaling pathway induced by CLDN1 cell-cell interaction might be postulated. However, we find CLDN1-positive apoptotic breast tumor cells exclusively in the spheroid core region but not on the outer cell layers (Fig. 4), and in monolayer 2D breast cancer cultures we find no enhancement of apoptosis in CLDN1-positive confluent monolayers compared to subconfluent cell cultures or their CLDN1-negative counterparts (Fig. 3). These data argue against the existence of a CLDN1-induced apoptosis pathway.
Little is known, however, about the underlying molecular mechanisms that control the expression and homing of claudins, especially in human tumor cells, and there may exist at least 2 independent pathways: one that regulates CLDN1 expression and the other that controls specific homing into the cell membrane. In neoplastic cells, this may not result in the creation of a full constellation of normal polarized mammary epithelium, but it is sufficient to induce apoptotic functions. The inhibition of the MEK1 pathway in MDCK cells gave rise to more prominent homing of CLDN1 in these constitutive CLDN1-positive cells,42 and hormone exposure,43, 44 the application of retinoids45 or cytokines46, 47 induced formation of TJs in established cell lines. It is also known that the protein kinase C48, 49, 50, 51 and the ras-raf-PI3K43, 52 signaling pathways might be involved in the quantitative regulation of TJ-associated molecules. Future evaluation of our breast cancer models will include the analysis of the pathways of specific activators and inhibitors in reconstitution of the TJ functions. The understanding of the defects of the molecular pathways involving CLDN1 might give rise to a new concept in the treatment of solid tumors in addition to the inhibition of vascularization by inducing cell death through the blockage of the supply of survival biomolecules of tumors in vivo.
Supported by a grant from the University of Washington Medical Center Breast program Tennis Tournament (to K.S.).
- 16Molecular physiology and pathophyiology of TJs: I, TJ structure and function—lessons from mutant animals and proteins. Gastroint Liver Physiol 2000; 279: 250–4., , .
- 41Claudin-2 is selectively expressed in proximal nephron in mouse kidney. Am J Physiol 2001; 281: 966–74., , .