Accumulating data now suggest that ZO-1, once delocalized from tight junctions, could be implicated in the regulation of tumor-promoting genes. Because of their major implication in different steps of tumor progression, we investigated here the influence of ZO-1 on chemokines expression in breast cancer cells. Using GeneArray analysis to compare chemokine mRNA expression in breast tumor cells transfected with a siRNA against ZO-1, we identified CXCL-8IL-8 as a major potential target of ZO-1 signaling, being strongly downregulated following ZO-1 siRNA transfection. Examining further the relationship between ZO-1 and interleukin-8 (CXCL8/IL-8), we first showed that CXCL8/IL-8 expression correlates with a relocalization of ZO-1 in several breast cancer cell lines. Moreover, CXCL8/IL-8 is downregulated in invasive BT549 cells transfected with three different ZO-1 siRNA and overexpressed in noninvasive BT20 and SKBR3 cells transfected with vectors expressing ZO-1. We also provide evidence for an activation of the CXCL8/IL-8 promoter by ZO-1. Finally, we show that the regulation of CXCL8/IL-8 by ZO-1 is independent of the β-catenin pathway. Our results thus clearly show an implication of ZO-1 in CXCL8/IL-8 regulation. Because of the major implications of CXCL8/IL-8 in tumor invasion, such a regulation could play an important role in breast cancer progression. Mol Cancer Res; 10(1); 121–32. ©2011 AACR.
During the metastatic progression of epithelial cancer cells, intercellular adhesion complexes are reorganized, and their molecular components are downregulated and/or relocalized. The reorganization of E-cadherin/catenin complexes and, more particularly, the involvement of the β-catenin pathway in the regulation of tumor-associated genes has been largely documented during epithelial tumor cell invasion (1, 2). Studies are now also accumulating that show the implication of tight junction (TJ) molecules in tumor progression (3, 4).
In normal epithelia, TJs are constituted of transmembrane proteins (occludin, claudins, junctional adhesion molecules) linked to the actin network via cytoplasmic molecules, including those of the zonula occludens (ZO) family: ZO-1, -2, and -3. ZO-1 has been the first described member of the ZOs and is considered as a central scaffolding protein regulating TJ organization. ZO-1 indeed interacts with other TJ molecules through its conserved structural domains: 3 PDZ (PSD95/DLG/ZO-1) domains, 1 src homology 3 (SH3) domain, and 1 guanylate kinase homologous (GK) domain (5–7). For instance, ZO-1 binds transmembrane molecules of the TJ, including occludin, via its GK domain and several claudins via the first PDZ domain. ZO-1 also interacts with cytoplasmic molecules of the TJ such as ZO-2 or ZO-3 via the second PDZ domain. In addition, ZO-1 binds actin via its carboxyl terminal region and an indirect binding of ZO-1 with the cytoskeleton also occurs through peripherial junctional proteins including cortactin, α-actin, AF6/Afadin, or cingulin. Finally, the SH3 domain of ZO-1 binds several signaling proteins such as serine protein kinase (ZAK), the Y-box transcription factor ZONAB (ZO-1–associated nucleic acid-binding protein), the heat shock protein Apg2 and the G proteins Gα12.
In carcinomas, ZO-1 is generally considered as a tumor suppressor. Several studies have thus reported that a reduction of ZO-1 expression correlates with tumor dedifferentiation and progression in breast and colorectal carcinomas (8). Nevertheless, ZO-1 has also been found overexpressed in different types of cancers, including pancreatic cancers, gastric cancers, and melanoma (9–11) More particularly, a loss of membrane-associated localization and a cytoplasmic distribution of ZO-1 has been reported in breast carcinomas (8). Also, in vitro, migrating epithelial Madin-Darby canine kidney cells (MDCK) cells as well as fibroblasts under proliferative and promigratory conditions have been reported to display a ZO-1 nuclear localization (12, 13). It has thus been suggested that, as described for β-catenin, ZO-1 can shuttle between the plasma membrane and the cytoplasm or nucleus and exert specific roles in these different subcellular compartments (12, 14, 15). Accordingly, nuclear exporting (NES) and nuclear localization (NLS) signals have been identified on ZO-1 (14). Also, ZO-1 is today considered as a member of the NACos family (nuclear and adhesion complex components) described as adhesion molecules which can also be found in the nucleus (16). However, data reporting nuclear ZO-1 are scarce and a functional role of nuclear ZO-1 has not been clearly established. A relocalization of ZO-1 out of cell–cell contacts and an increased cytoplasmic level of ZO-1 has been implicated in the proinvasive role of ZO-1 rather than a nuclear shuttling. Supporting this concept, it has been shown that a form of ZO-1 which lacks the GK domain, consequently displaying altered binding to occludin and impaired recruitment to the plasma membrane, accumulates in the cytoplasm and is able to confer tumorigenicity to transfected MDCK cells injected into nude mice (17). Modulating ZO-1 expression by cDNA or siRNA transfections, we and others also showed the tumor invasion–promoting activity of ZO-1 in breast and melanoma tumor cell lines, respectively (11, 18). On the basis of these results, the concept has emerged that, in addition to its structural role in TJ organization, ZO-1, once delocalized from TJs, may be involved in intracellular signaling pathways and in gene expression favoring tumor progression.
Chemokines are proinflammatory cytokines that can be divided into 4 classes depending on the number of amino acids separating the first 2 conserved cysteines: C, CC, CXC, and CX3C chemokines. Chemokines were initially defined as soluble factors regulating directional leukocyte migration during inflammation process but which are now also regarded as potent regulators of cancer development (19, 20). Interleukin-8 (CXCL8/IL-8) has been the first identified chemokine and belongs to the CXC chemokine family. CXCL8/IL-8 activate a large number of intracellular signaling pathways (serine/threonine kinases, tyrosine kinases, or Rho-GTPases) through G protein coupled receptors (CXCR1 and CXCR2) and exert many biological effects on different cell types of the tumor microenvironment (21). It has been shown to attract, among others, neutrophils and basophils. CXCL8/IL-8 has also largely been implicated as a positive regulator of angiogenesis, a particular role shared by all other CXC chemokines displaying an ELR motif (22, 23). CXCL8/IL-8 has thus been shown to promote angiogenesis via its interactions with its 2 receptors, CXCR1 and CXCR2, on human umbilical vein endothelial cells and on HMECs (24–26). In agreement with its tumor-promoting role, CXCL8/IL-8 is found overexpressed in many different types including breast carcinoma (27, 28). CXCL8/IL-8 expression also more precisely correlates with the pathologic stage in prostate cancer (29), with a high tumor grade and metastatic potential in breast tumors (30, 31) and with an increased vascularization in gastric carcinoma (32) and breast carcinoma (31).
If the implication of chemokines, and particularly CXCL8/IL-8, in tumor growth and invasion is now well established, the mechanisms contributing to their overexpression by epithelial tumor cells are still poorly understood. In a previous study, we showed that a reorganization of adherens junctions and the activation of the β-catenin pathway regulate chemokine expression (33). In this study, we have examined the influence of ZO-1 on chemokines expression in breast cancer cell lines. We thus identified CXCL8/IL-8 as a major potential target of ZO-1 signaling which could be involved in breast tumor progression.
Materials and Methods
Cell lines and culture conditions
All human mammary epithelial cells used were obtained from the American Type Culture Collection. Human breast carcinoma cell lines (MCF7, BT20, SKBR3, MDA-MB-231, BT549, and Hs578T) were grown in Dulbecco's Modified Eagle's Medium (DMEM; Invitrogen) supplemented with 10% fetal calf serum (FCS), 2 mmol/L of l-glutamine, and 100 UI/mL of penicillin and streptomycin (Invitrogen). THP-1 monocytic human cells were grown in RPMI medium supplemented with 20% FCS, 2 mmol/L of l-glutamine, and 100 UI/mL of penicillin and streptomycin. MCF10A cells were grown in a 1:3 (v/v) mixture of HAM F12 and DMEM, supplemented with 20 μg/mL of adenine, 5 μg/mL of insulin, 0.5 μg/mL of hydrocortisone, 2 ng/mL of epidermal growth factor (EGF), 5 μg/mL of transferrin, 1.5 ng/mL of triiodothyronine, 10% FCS, and 100 UI/mL of penicillin and streptomycin. MCF10A cells were treated with 5 ng/mL TGF-β (R&D systems) and 10 ng/mL EGF (Sigma).
THP-1 cells (105) were suspended in 200 μL of serum-free RPMI medium supplemented with 0.5% bovine serum albumin (BSA) and placed in the upper compartment of a 8 μmol/L transwell (6.5-mm diameter; Costar). The lower compartment was filled with 600 μL of 48-hour conditioned serum-free RPMI medium of BT549 cells transfected either with a scrambled control siRNA or an anti–ZO-1 siRNA. After 7 hours of incubation at 37°C, THP-1 cells that had migrated in the lower chamber of the transwell were counted with a cell counter (Cedex; Roche Diagnostics).
Modified Boyden chamber invasion assay
The in vitro invasive properties of cells were assessed using a Matrigel-coated modified Boyden chamber assay. A total of 105 cells were placed in the upper compartment of the invasion chamber (BD BioCoat Matrigel Invasion Chamber; BD Biosciences). The chambers were incubated for 8 hours at 37°C. The filters were then fixed in methanol and stained with hematoxylin. Quantification of the invasion assay was done by counting the number of cells at the lower surface of the filters (30 fields at 40-fold magnification).
Transfection of siRNA
Nineteen nucleotide (nt)-specific sequences were selected in the coding sequence of ZO-1 (Genebank accession number: L14837) and of β-catenin (Genebank accession number: NM_003257) to generate 21-nt sense and 21-nt antisense strands of the type (19N)TT (N, any nucleotide). The sense and antisense strands were then annealed to obtain duplexes with identical 3′ overhangs. The sequences were submitted to a BLAST search against the human genome to ensure the specificity of the siRNA to the targeted sequence. Corresponding scrambled duplexes which do not recognize any sequence in the human genome were used as control. The 19-nt–specific sequences for the 3 ZO-1 siRNA and for the β-catenin siRNA are as follow: ZO-1 si1 5′-GUUAUACGAGCGAUCUCAU-3′, ZO-1 si2 5′-GGAGGAAACAGCUAUAUGG-3′, ZO-1 si3 5′-GACGAGAUAAUCCUCAUUU-3′, β-catenin si 5′-CTCAACTGAAAGCCGTTTC-3′. For transfection of the siRNA duplexes, 75 × 103 cells were plated in 6-well plates in 2 mL per well of culture medium. At 24 hours after plating, the cells were transfected by phosphate calcium precipitation by adding in each well 200 μL of a mixture containing 20 nmol/L of the siRNA duplexes, 140 mmol/L NaCl, 0.75 mmol/L Na2HPO4, 6 mmol/L glucose, 5 mmol/L KCl, 25 mmol/L HEPES, and 125 mmol/L CaCl2. Sixteen hours after transfection, the cells were extensively washed with PBS and incubated for 48 hours in culture medium, before they were harvested for Gene Array, reverse transcriptase PCR (RT-PCR) analyses, and Western blotting. Conditioned media were also collected for ELISA analyses.
The relative mRNA expression of human chemokines and chemokines receptors was analyzed with GEArray Q serie n° HS005.2 (SA Biosciences) according to the manufacturer's protocol. Total RNA from BT549 cells transfected with ZO-1 scrambled or ZO-1 siRNA was extracted using the High Pure RNA Isolation Kit (Roche Diagnostics). An amount of 2 μg RNA was reverse transcribed and double-stranded cDNA probes were generated by biotin-16-dUTP incorporation using the AmpoLabeling-LPR Kit (SA Biosciences), according to the manufacturer's instructions. The cDNA probes were denatured at 94°C for 2 minutes. The membranes were prehybridized in GEAhyb Hybridization Solution (SA Biosciences), supplemented with heat-denatured salmon sperm DNA (Sigma) for 4 hours at 60°C. Labeled cDNA probes were hybridized overnight at 60°C with continuous agitation. Following repetitive washing in salt sodium citrate buffer containing SDS, membranes were incubated with alkaline phosphatase–conjugated streptavidin. Signals were detected using the CDP-Star chemiluminescent substrate (SA Biosciences) and subsequent exposure to X-ray films. After densitometry scanning on a Fluor-S MultiImager (Bio-Rad), spots were converted into numerical data using GEArray Expression Analysis Suite software (SA Biosciences), and values were normalized to the signal of the housekeeping gene β-actin.
The human CXCL8/IL-8 promoter constructs have been previously described (34) and were kindly provided by Professor M.A. Buendia (Pasteur Institute, Paris, France). They contain either a 193-bp fragment, a 193-bp fragment mutated in the β-catenin/TCF-binding site or a 173-bp fragment lacking the β-catenin/TCF-binding site of the 5′ upstream regulatory region of the CXCL8/IL-8 gene cloned in the pGL3 firefly luciferase reporter plasmid (Promega). The TOP-FLASH and FOP-FLASH reporter plasmids, containing 3 wild-type (TOP-FLASH), or mutated (FOP-FLASH) copies of the β-catenin/TCF-binding site upstream of a minimal herpesvirus thymidine kinase promoter driving the firefly luciferase expression, were kindly provided by Professor H.C. Clevers (University Hospital, Utrecht, The Netherlands; ref. 35). The expression vectors encoding wild-type ZO-1 (ZO-1 WT) or the NH2-terminal fragment of ZO-1 comprising the PDZ domains 1–3 (ZO-1 PDZ) have been characterized previously (17).
Transient transfections were done with Fugene-6 transfection reagent (Roche Diagnostics) on 1.2 × 105 BT20 cells and 1.5 × 105 SKBR3 cells plated in 6-well plates. At 24 hours after plating, transfection was carried on as recommended by the manufacturer by adding, in each well, a mixture containing 100 μL of serum-free medium, 3 μL of Fugene-6, and 2 μg of the ZO-1 WT or ZO-1 PDZ expression vector. Controls were generated by transfecting the cells with 2 μg the corresponding empty vector pLNCX. The cells were harvested 48 hours after transfection for RT-PCR or Western blotting analyzes. Conditioned media were also collected for ELISA analyses.
For double transient transfection experiments with ZO-1 expression vectors and β-catenin siRNA, SKBR3 cells were subsequently transfected with the siRNA and then with the expression vectors. Sixteen hours after transfection of the siRNA, the cells were washed twice with PBS, then transiently transfected with the expression vectors for 48 hours as detailed above.
Luciferase reporter assays
Transient transfection experiments were carried out using Fugene-6 transfection reagent on 5 × 104 BT20 and SKBR3 cells seeded in a 24-well plate half an hour before the addition of the DNA/Fugene mixture. Each well was incubated with a mixture containing 20 μL of serum-free DMEM, 0.6 μL of Fugene-6, 0.2 μg of the firefly promoter–luciferase reporter constructs (either CXCL8/IL-8 promoter, TOP-FLASH, or FOP-FLASH reporter construct), 0.2 μg of the expression vectors (ZO-1 WT, ZO-1 PDZ, or β-catenin and TCF4 expression vectors, or the corresponding empty vector) and 0.8 ng of the Renilla luciferase expression vector phRG-TK (Promega). At 24 hours after transfection, the cells were lysed in 100 μL of passive lysis buffer and the luciferase activity was determined with a luminometer using the Dual Luciferase Assay System (Promega) on 20 μL of lysate. For each experiment, the firefly luciferase activity was normalized to the activity of the Renilla luciferase used as internal control. The results were expressed as fold induction determined by normalizing each firefly luciferase value to the Renilla luciferase internal control value and by dividing these normalized values with the mean normalized value of the corresponding reporter construct transfected with the empty expression vector. To assess the β-catenin/TCF/LEF activities, the normalized values obtained with the FOP-FLASH reporter plasmid were subtracted from the normalized values obtained with the TOP-FLASH reporter plasmid. The resulting values were then divided with the mean normalized value of the condition transfected with the empty expression vector. Each experiment was carried out at least 3 times in triplicate. Data were expressed as means ± SEM. A one-sample t test was done and a P value less than 0.05 was considered significant.
Total RNA was extracted using the High Pure RNA Isolation Kit (Roche Diagnostics). Ten nanogram of RNA was then reverse transcribed into cDNA and amplified using the GeneAmp Thermostable rTth enzyme (Applied Biosystems). The reverse and forward primers for CXCL8/IL-8, ZO-1, MMP-14, vimentin, β-catenin, and 28S were designed as follows: CXCL8/IL-8 primers (forward 5′-GAATTCTCAGCCCTCTTCAAAAAC-3′ and reverse 5′-GCCAAGGAGTGCTAAAGAACTTAG-3′), ZO-1 primers (forward 5′-GCAGCTAGCCAGTGTACAGTATAC-3′ and reverse 5′-GCCTCAGAAATCCAGCTTCACGAA-3′), MMP-14 primers (forward 5′-GGATACCCAATGCCCATTGGCCA-3′ and reverse 5′-CCATTGGGCATCCAGAAGAGAGC-3′), vimentin primers (forward 5′-GCGTGACGTACGTCAGCAATATGA-3′ and reverse 5′-GTTCCAGGGACTCATTGGTTCCTT-3′), β-catenin primers (forward 5′-CGAGGACCAGGTGGTGGTTAATAA-3′ and reverse 5′-CAGAATCCACTGGTGAACCAAGCA-3′) and 28S (forward 5′-GATTCTGACTTAGAGGCGTTCAGT-3′ and reverse 5′-GTTCACCCACTAATAGGGAACGTGA-3′). For quantification, the values obtained for the RNA of interest were normalized to the values obtained for 28S amplification in 3 independent transfection experiments. Results were expressed as fold induction determined by dividing the normalized value of a given stimulated condition with the normalized value of the corresponding control condition. Data are expressed as means ± SEM. A one-sample t test was done and a P value less than 0.05 was considered significant.
Detection of CXCL8/IL-8 by ELISA
CXCL8/IL-8 secretion was measured by ELISA in culture media of the different breast cancer cell lines or of ZO-1 siRNA-transfected BT549 cells. The CXCL8/IL-8 ELISA was done according to the manufacturer's instructions (DuoSet human IL-8 R&D Systems). Results are expressed in pg/mg of protein dosed in cell extracts and as means ± SEM of 3 independent experiments.
In conditioned media of the ZO-1–transfected BT20 and SKBR3 cells, CXCL8/IL-8 secretion was measured using a chemiluminescent ELISA assay according to the manufacturer's instructions (Quantiglo human IL-8; R&D Systems). Results were expressed as fold induction determined by dividing the amount of CXCL8/IL-8 detected in ZO-1 siRNA condition with the amount detected in the corresponding scrambled siRNA control. Data were expressed as means ± SEM of 3 independent experiments. A one-sample t test was done and a P value less than 0.05 was considered significant.
Western blotting analysis
Total protein extracts were made in Tris 50 mmol/L pH7.4, NaCl 150 mmol/L, Nonidet P-40 1%, Triton 1%, Deoxycholic acid 1%, SDS 0.1%, and Iodoacetamid 5 mmol/L supplemented with a cocktail of proteases inhibitors (Roche Diagnostics). Subcellular fractions were extracted with the Proteoextract subcellular proteome extraction kit according to the manufacturer's instructions (Merck). Twenty microgram of total protein samples or 6 μg of subcellular fractions were separated by electrophoresis on 10% SDS-polyacrylamide gels and then transferred on a polyvinylidene difluoride membrane (PerkinElmer). Immunoblotting was done with a monoclonal anti ZO-1 antibody (ZO1-1A12; Zymed Laboratories) followed by an incubation with a horseradish peroxidase (HRP)-conjugated goat anti-mouse secondary antibody (Dako). Signals were visualized using the Western Lightning Chemiluminescence Reagent Kit (Pierce). For loading normalization, the blots were then incubated, after extensive washes, with a rabbit anti-actin (Sigma) or a mouse anti-GAPDH (glyceraldehyde-3-phosphate dehydrogenase) antibody, used as a cytosolic marker (Chemicon) or with a goat anti-nibrin antibody, used as a nuclear marker (Santa Cruz Biotechnology) followed by a HRP-conjugated rabbit anti-goat or goat anti-mouse secondary antibody (Dako). Signals were detected with the Western Lightning Chemiluminescence Reagent Kit (Pierce).
Confluent monolayers of cells cultured on glass coverslips were fixed with methanol for 10 minutes at −20°C. The coverslips were then saturated for 10 minutes with 3% BSA in PBS. Monolayers were next (after intermediate washes in PBS) incubated for 1 hour with the monoclonal antibody to ZO-1 and then with a polyclonal goat anti-mouse antibody for 1 hour. The coverslips were then mounted with Aqua Poly/Mount antifading solution (Polysciences) onto glass slides and observed under a Zeiss fluorescence microscope.
Screening of chemokine expression by gene array
To investigate a potential regulation of chemokines by ZO-1, we first used a chemokine Gene Array GEArray Q serie (SA Biosciences) to screen and compare the mRNA expression profile of chemokines in invasive breast tumor BT549 cells transfected with a ZO-1 siRNA and a corresponding scrambled control siRNA. From this screening, CXCL8/IL-8 emerged as a strongly downregulated gene following ZO-1 siRNA transfection (Fig. 1A). To confirm the results obtained with the GEArray analysis using ZO-1 si1, showing that ZO-1 regulates CXCL8/IL-8 expression, we transfected BT549 cells with 2 other ZO-1 siRNA sequences (ZO-1 si2 and ZO-1 si3). As shown in Fig. 1B, the 3 siRNA sequences efficiently decreased ZO-1 protein levels. We concomitantly observed that CXCL8/IL-8 mRNA (Fig. 1C) and protein (Fig. 1D) expression was clearly downregulated in BT549 tumor cells treated with the 3 ZO-1 siRNA. A functional consequence of such chemokine regulation by ZO-1 was shown using a transmigration assay with the monocytic THP-1, previously shown to display a chemotactic response to various chemokine including CXCL8/IL-8 (36). THP-1 cells migrated less toward conditioned medium of ZO-1 siRNA BT549 cells than toward conditioned medium from ZO-1 scrambled control transfectants (Fig. 1E).
CXCL8/IL-8 is expressed by invasive breast cancer cells
To further explore a potential relationship between ZO-1 and CXCL8/IL-8 expression, we examined their expression in a set of breast tumor cell lines displaying various degrees of invasiveness. The invasive phenotype of these breast cancer cells, assessed by a modified Boyden chamber assay, has been extensively characterized previously (37) and confirmed in this study (Fig. 2A). Examining CXCL8/IL-8 mRNA expression by RT-PCR analysis, we observed that invasive cells (MDA-MB-231, BT549, and Hs578T) expressed high levels of CXCL8/IL-8 compared with noninvasive cell lines (MCF7, BT20, and SKBR3; Fig. 2B). CXCL8/IL-8 mRNA in invasive cell lines was also correlated to the expression of other proinvasive markers such as vimentin or matrix metalloproteinase (MMP-14), previously shown to be target genes of ZO-1 signaling (Fig. 2B; refs. 17, 18). This higher expression of CXCL8/IL-8 in invasive breast cancer cell lines was confirmed at the protein level by ELISA analysis (Fig. 2C). Looking at ZO-1 expression by Western blotting, we found it was expressed in all the cell lines (Fig. 2D) as multiple bands, likely corresponding to different isoforms or degradation products. By immunofluorescence analysis, however, we observed that ZO-1 subcellular localization differs from one cell line to another. In contrast to invasive cell lines, noninvasive cell lines rapidly form clusters of cells. In these noninvasive cell lines, ZO-1 labeling thus varies with cell density and was mainly found at the cell membrane in intercellular contacts of cell clusters. In contrast, in invasive breast tumor cell lines expressing CXCL8/IL-8, which do not form clusters, ZO-1 staining was mostly cytoplasmic independently of cell confluency. This is illustrated for BT549 (representative of invasive cells) and BT20 cells (representative of noninvasive cells) in Fig. 2E.
To correlate further CXCL8/IL-8 expression and ZO-1 relocalization, we used the MCF10A cells, which were previously shown to undergo alterations of TJs following long-term TGF-β treatment (38). Here, using a combined treatment of EGF and TGF-β, we observed an induction of CXCL8/IL-8, both at the protein and RNA level, which coincided with a diminution of the typical honeycomb membrane staining of ZO-1 (Fig. 3).
In these different cell systems, a cytoplasmic localization and/or a diminution of membrane-associated localization of ZO-1 is thus associated with enhanced CXCL8/IL-8 expression, supporting our observations that ZO-1 siRNA transfection in invasive BT549 downregulates CXCL8/IL-8.
CXCL8/IL-8 expression is regulated by ZO-1
To complement our results obtained with ZO-1 siRNA, showing a downregulation of CXCL8/IL-8 by ZO-1 siRNA, we inversely transiently transfected noninvasive BT20 and SKBR3 breast cancer cells with expression vectors encoding the full-length ZO-1 (ZO-1 WT), the NH2-terminal fragment of ZO-1 comprising the PDZ domains 1–3 (ZO-1 PDZ) that is unable to bind occludin and thus presents a diminished ability to localize at the TJ complexes or with the empty vector as control. These transfections enhanced ZO-1 expression both at the protein level, as shown by immunofluorescence (Fig. 4A), and at the RNA level by RT-PCR (Fig. 4B). Using RT-PCR analysis, we observed a significant induction of CXCL8/IL-8 mRNA after transfection of either ZO-1 full length or the PDZ construct in both cell lines (Fig. 4B). Indeed, we observed a 2.10 (± 0.21) and 2.18 (± 0.22) fold induction of CXCL8/IL-8 mRNA expression, respectively, after ZO-1 WT and ZO-1 PDZ transfection of BT20 cells compared with the empty vector. Similarly, the transfection of SKBR3 cells with the ZO-1 WT and ZO-1 PDZ plasmids lead to a significant induction of CXCL8/IL-8 expression (1.96 ± 0.26 and 2.23 ± 0.07 fold respectively). A significant accumulation of CXCL8/IL-8 protein was accordingly detected by ELISA in the conditioned media of BT20 and SKBR3, particularly after ZO-1 PDZ transfection (Fig. 4C). Because ZO-1 could have a promigratory/invasive signaling role once delocalized from the cell membrane complexes, we examined the subcellular localization of ZO-1 in cells transiently transfected with ZO-1 WT or ZO-1 PDZ expression vectors. As shown by Western blotting analyses on subcellular protein extracts, an increase of the cytoplasmic pool, and also of the nuclear pool of the ZO-1 WT or the truncated ZO-1 PDZ protein, was observed in the BT20 and the SKBR3 cells following the transfection of the corresponding expression vectors (Fig. 5).
CXCL8/IL-8 promoter is activated by ZO-1
Previous studies have shown that transfection of ZO-1 to enhance its cytoplasmic level and activate the β-catenin pathway (17, 18). Moreover, Levy and colleagues have shown that CXCL8/IL-8 is regulated by the β-catenin pathway through a specific TCF/LEF-binding site (34). We thus further investigated whether the β-catenin pathway is implicated in the regulation of CXCL8/IL-8 by ZO-1. We used a luciferase reporter plasmid containing a 193-bp fragment of the CXCL8/IL-8 promoter as well as the corresponding construct in which the β-catenin/TCF-binding site has been mutated to inhibit the binding of β-catenin/TCF complexes (193 pro mut). A truncated form of 173 bp that lacks the β-catenin/TCF-binding site was also used. These promoter constructs have been previously described (34). Cotransfection experiments were carried out in BT20 and SKBR3 cells using the 193 bp, the mutated 193 bp, or the 173 bp human CXCL8/IL-8 promoter reporter plasmids and expression vectors encoding ZO-1 WT, ZO-1 PDZ, or the empty vector (pLNCX) as control. In BT20 and SKBR3 cells, cotransfection of the 193-bp CXCL8/IL-8 promoter construct with ZO-1 WT and ZO-1 PDZ led to the induction of the promoter activity by 3.9 ± 0.4 and 5.5 ± 0.6 fold and 2.5 ±0.6 and 4 ± 0.8 fold, respectively. The 193-bp mutated and the 173-bp reporter constructs were also significantly activated following the transfection of the ZO-1 WT or the PDZ expression vectors (Fig. 6). This suggests that the β-catenin pathway is not implicated in the regulation of CXCL8/IL-8 by ZO-1.
CXCL8/IL-8 regulation by ZO-1 does not implicate the β-catenin pathway
To confirm this lack of implication of the β-catenin pathway in CXCL8/IL-8 regulation by ZO-1, we cotransfected BT20 and SKBR3 cells with the TOP-FLASH or FOP-FLASH reporter plasmid and the ZO-1 WT or ZO-1 PDZ expression vectors. As shown in Fig. 7A, the activity of the TOP-FLASH/FOP-FLASH reporter system was accordingly not enhanced following the transfection of the ZO-1 WT or PDZ expression vectors.
Furthermore, we have also analyzed CXCL8/IL-8 regulation by ZO-1 following β-catenin RNAi inhibition in SKBR3 cells. We thus analyzed CXCL8/IL-8 expression in SKBR3 cells successively transfected with a siRNA against β-catenin (or a control siRNA), then with the expression vector for ZO-1 cDNA (or the vector control). As shown in Fig. 7B, the transfection of the β-catenin siRNA did not modify CXCL8/IL-8 induction by ZO-1. Confirming our luciferase assays, these data also suggest that β-catenin is not involved in CXCL8/IL-8 regulation by ZO-1.
In this study, we identified CXCL8/IL-8 as a potential target of ZO-1 regulation. We first showed that CXCL8/IL-8 is highly expressed only in invasive breast cancer cells or in EGF/TGF-β–treated MCF10A cells, in which ZO-1 does not localize at cell–cell contacts but, rather, displays a cytoplasmic staining. We further documented a downregulation of CXCL8/IL-8 in invasive breast tumor cells transfected with ZO-1 siRNA. Inversely, transfection of ZO-1 expression vectors in noninvasive cells induced CXCL8/IL-8 expression and activated its promoter.
Although this is the first report correlating ZO-1 localization and CXCL8/IL-8 expression, independent studies have associated CXCL8/IL-8 expression with high invasive properties. De Larco and colleagues have shown that CXCL8/IL-8 is produced at high levels in MDA-MB-231 and MDA-MB-435 compared with MCF7 and T47D cells (27) and is associated with their high metastatic potential. Freund and colleagues have also reported that CXCL8/IL-8 mRNA expression is 80-fold higher in MDA-MB-231 than in MCF7 cells (39). Furthermore, our data are in agreement with studies reporting CXCL8/IL-8 expression in a variety of human cancer cell lines, including renal carcinoma cells (40), melanoma cells (26, 41, 42), breast cancer cells (43), prostate cancer cells (44), colon carcinoma cells (45), and bladder cancer cells (46). Also, these studies associated CXCL8/IL-8 expression and the metastatic potential of the cancer cells. Moreover, CXCL8/IL-8 has been associated with a poor prognosis of breast carcinoma (28, 31, 43), gastric carcinoma (47), melanoma (48), and nasopharyngeal carcinoma (49).
On the other hand, our data are in agreement with the literature showing a relocalization of ZO-1 associated with high migratory/invasive potential of epithelial cells (8, 12, 50). However, beyond establishing a correlation between CXCL8/IL-8 expression and the cytoplasmic localization of ZO-1, we have functionally involved ZO-1 in the regulation of CXCL8/IL-8 expression. Indeed, we have shown that CXCL8/IL-8 expression is downregulated in invasive BT549 cells transfected with 3 different ZO-1 siRNA and upregulated in BT20 and SKBR3 cells transfected with ZO-1 cDNA. Only a few studies have documented gene regulation by ZO-1. Balda and colleagues described ZO-1 as a regulator of ErbB-2 expression and cell proliferation in breast cancer cells (51). The induction of an epithelial-to-mesenchymal transition associated with a decrease of epithelial gene expression and an increase in the expression of several mesenchymal genes has also been reported in MDCK cells transfected with ZO-1 expression vectors and correlated to increased invasiveness (17). We recently reported an induction of MMP-14 (membrane type 1-MMP) expression by ZO-1 in breast cancer cells (18). Taken together, these results therefore suggest that, once delocalized from the junctional membrane adhesion complexes, ZO-1 can contribute to signaling events leading to the activation of the transcription of tumor-promoting genes such as CXCL8/IL-8.
The molecular mechanisms involved in such gene regulation by ZO-1 remain, however, unclear. ZO-1 can interact with several molecular partners, which could directly or indirectly contribute to its ability to regulate gene expression. For example, the regulation of ErbB-2 by ZO-1 involves a direct interaction between ZO-1 and the Y-box transcription factor ZONAB at the cell membrane, inducing a membrane sequestration of this factor, which is a repressor of the ErbB-2 promoter (51). Recently, a sequestration of ubinuclein (ubn-1) by ZO-1 has also been described which could impair the nuclear activity of this NACos, also known to interact with the transcription factor AP-1 (52, 53). In these studies, however, ZO-1 was rather shown to control gene expression by sequestering nuclear factors at the cell membrane. In contrast, the data which have established a proinvasive signaling role of ZO-1 strongly implicate the relocalization of ZO-1 out of cell–cell contacts and an increased cytoplasmic level, but not particularly, a nuclear shuttling. Thus, an activation of the β-catenin/TCF/LEF transcription pathway has been documented in MDCK cells and in breast tumor cells following transfection of ZO-1 expression vectors which led to enhanced cytoplasmic levels of ZO-1 (17, 18). We and others have reported that chemokines could also be the targets of the β-catenin/TCF pathway. This has been shown for CXCL8/IL-8 by Levy and colleagues in human hepatoma cells (34), by Masckauchán and colleagues in human endothelial cells (54), and by our group for MCP-1/CCL2 in breast tumor cell lines (33). In the light of these studies, we have also examined a potential implication of the β-catenin pathway in the regulation of CXCL8/IL-8 by ZO-1. However, our present data suggest that the β-catenin pathway is not implicated in the regulation of CXCL8/IL-8 by ZO-1. A β-catenin-independent regulatory mechanism is thus rather likely to be involved in CXCL8/IL-8 regulation by ZO-1. Along these lines, the human CXCL8/IL-8 promoter can be activated by a complex cooperation of other transcription factors such as NFkB, AP-1, PEA3, and c/EBP in human breast cancer cells (39, 55, 56) and a potential direct or indirect interaction between ZO-1, and these transcriptional pathways in different cell compartments (both in the cytoplasm and in the nucleus) cannot be excluded and is under investigation.
In conclusion, we have identified CXCL8/IL-8 as a potential target of ZO-1 signaling in breast tumor cells. Considering the important implication of CXCL8/IL-8 and the reorganization of TJs during tumorigenesis, such a regulation of CXCL8/IL-8 by ZO-1 might play an important role in breast cancer progression.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
This work was supported by grants from the «Fonds National de la Recherche Scientifique» (F.N.R.S., Belgium), the «Fondation contre le Cancer» (Belgium), the «PHC-Tournesol» (Belgium), the European Framework Programme 7 “MICROENVIMET”, the «Fonds de la Recherche Scientifique Médicale», the «Fonds spéciaux de la Recherche» (University of Liège), the «Centre Anticancéreux près l'Université de Liège», the «Fonds Léon Frédéricq» (University of Liège), the «Région Champagne-Ardenne», the «Ligue Contre le Cancer», the Lions Club of Soissons. A. Brysse is a research fellow from the “Télévie” and C. Gilles is a senior research associate from the F.N.R.S.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Received April 26, 2011.
- Revision received October 19, 2011.
- Accepted October 28, 2011.
- ©2011 American Association for Cancer Research.