Background
A growing understanding of the heterogeneous nature of breast cancer has stemmed primarily from gene expression analysis studies, and more recently, integrated analysis of copy number and exome sequencing [
1]. This has led to a redefinition of breast cancer subsets [
1]. This new classification of breast cancer subtypes, focused on 10 genetically distinct groups, confirmed the prevalence of four previously identified molecular subtypes (luminal A, luminal B, HER2 +ve and the basal-like) [
1]. Whereas the luminal A and B subtypes are characterized by their epithelial phenotypes, hormone sensitivity (estrogen receptor positive, ER+/ progesterone receptor positive, PR+), mildly invasive capacity and relatively good clinical outcome, the HER2+ and basal-like breast cancer (BLBC) subtypes are characterized by their mesenchymal phenotype, insensitivity to hormonal therapy (ER-ve; PR-ve), enhanced invasiveness and metastatic capacity [
2] and poor clinical outcome [
3‐
7].
The claudins belong to a family of tight junction (TJ) proteins (24 identified to date), that are crucial for the organization of epithelial cell polarity [
8]. They contribute to the trans-epithelial barrier that controls the transport of ions and small molecules. They are also considered essential for the overall maintenance of the differentiated state of epithelial cells [
9,
10]. The claudins share a very distinct transmembrane topology: each family member is predicted to possess four transmembrane domains with intracellular amino and carboxyl-termini in the cytoplasm and two extracellular loops [
11,
12]. The expression pattern of the claudins is usually tissue specific; however, most tissues express multiple claudins that can interact in either a homotypic or heterotypic fashion to form the TJ strand. As well, the exact combination of claudin proteins within a given tissue determines the selectivity, strength and tightness of the TJ [
11]. The claudins are also capable of recruiting signaling proteins, thereby regulating various cellular processes including cell growth, differentiation and tumorigenesis [
13,
14].
Claudin 1, the first member of this family to be identified, forms the backbone of the TJ strands and is crucial for the epidermal barrier function [
15]. In cancer, an absence of, or defects in tight junctions have been associated with the development of the neoplastic phenotype. Although long suspected to play an active role in tumorigenesis, only recently have a number of studies demonstrated that claudin 1 directly participates in the progression of several cancers including melanomas [
16], oral squamous cell carcinomas [
17] and colon cancers [
18].
Studies from our laboratory [
19] and others [
20‐
22] point toward a putative tumor suppressor role of claudin 1 in breast cancer as it is frequently down regulated in human invasive breast cancer and its absence or the down regulation of its expression is associated with poor prognosis [
23]. We have however, also found high claudin 1 and claudin 4 protein expression associated with the BLBC subtype [
19]. The BLBCs correspond to a subgroup of breast cancers that are poorly characterized and thus, mostly insensitive to most classical therapeutic strategies. Although a large cohort of human invasive breast cancers (350 samples) was examined in this earlier study, these tumors were of mixed pathological lesions (ductal, lobular, medullary, papillary, metaplastic), and of these, only 18 were of the BLBC subtype. As such, the clinical relevance of claudin 1 expression to the BLBCs could not be fully addressed.
The present study was carried out to determine whether the observed significant association between claudin 1 and the BLBC subtype could be clinically relevant. Specifically, we wanted to address whether there was an association between high levels of claudin 1 and disease recurrence and patient survival. However, since generally <15% of breast cancers are basal-like [
24], the construction of a BLBC enriched tissue microarray (TMA) warranted the screening of a large number of tissue specimens. Thus, our strategy was to first pre-select tumors that were ER-ve and PR-ve (previously carried out by the ligand binding assay) and then identify those tumors that exhibited HER2 negativity as well as EGFR or CK5/6 positivity by immunohistochemistry (IHC). Seventy-nine out of 151 tumors were confirmed to be “basal-like” in our basal-like enriched TMA. Additionally,
in vitro studies were carried out to examine whether claudin 1 had a direct functional role in human breast cancer. For these studies we used the human breast cancer cell line, BT-20 which is both phenotypically basal-like [
25,
26] and endogenously expresses high levels of this protein. Altogether this study provides evidence that claudin 1 identifies a specific subgroup of BLBC patients. We also demonstrate that claudin 1 could directly contribute to breast cancer progression.
Methods
Tissue microarrays
All invasive breast cancers used in the present study were obtained from the Manitoba Breast Tumour Bank (MBTB, University of Manitoba), which operates with the approval from the Faculty of Medicine, University of Manitoba, Research Ethics Board. As well the studies reported in this manuscript have been performed with the approval of the Bannatyne Campus, University of Manitoba, Research Ethics Board. Collection, handling and histo-pathological assessment of tumor tissues have been previously described [
27,
28]. The breast cancer tissue microarray (TMA) was constructed by the MBTB using a cohort of 151 breast tumor samples, which were determined to be estrogen receptor negative (ER-ve), progesterone receptor negative (PR-ve) by the ligand binding assay (ER-ve <3 fmol/mg protein, PR-ve <10 fmol/mg protein). Further, using a strict criteria for the basal-like subtype (ER-ve, PR-ve, HER2-ve and EGFR and/or CK5/6 +ve), 79 tumors were identified by IHC as having the BLBC phenotype. The remaining 72 tumors were designated as “non-basal”. The clinico-pathological characteristics of the patient cohorts were provided by the MBTB and used for statistical analyses.
Immunohistochemical analysis of TMAs
IHC was performed as described previously on the BLBC enriched TMA [
28]. Briefly, serial sections (5 μm) of the TMAs were stained with rabbit polyclonal antibodies to claudin 1 at a dilution of 1:150 (Life Technologies Inc., Burlington, ON, Canada), or claudin 4 at a dilution of 1:1200 (Abcam, Toronto, ON, Canada). The paraffin-embedded tissue sections were processed using an automated Discovery Staining Module, Ventana System (Tucson, AR, USA). Tissues were processed and incubated for 60 minutes with the primary antibody and 30 minutes with the secondary antibody following standard protocol. Validation of claudin 1 and claudin 4 antibodies has also been described previously [
19]. Antibodies to CK5/6 (D5/16B4, Life Technologies Inc.), EGFR (3C6, Ventana Systems), and HER2 (Cb11, NovaCastra, Concord, ON, Canada) were used as previously detailed [
28]. The TMA consisted of a total of 151 human invasive breast tumor biopsies, however only those tumors from which we were able to retrieve interpretable data (intact, unfolded tumor sections) were considered for our analysis. The IHC data, compiled into the database maintained by the MBTB, was made available for correlation analyses and other statistical comparisons [
27,
29].
Quantification and cut-off selection
Positive staining was assessed by light microscopy. A semi-quantitative assessment was used. Both staining intensity (scale 0–3) and the percentage of positive cells (0-100%) were multiplied to generate an H score ranging from 0–300, as previously described [
27,
28]. TMA staining was evaluated independently by two investigators AB and CP. Where discordance (i.e. different scores given by different investigators) was found, cases were re-evaluated commonly and a consensus reached. Only tumor biopsies whose ER/PR status was determined by both ligand-binding assay (ER-ve <3 fmol/mg protein, PR-ve <10 fmol/mg protein), and by IHC (ER-ve/PR-ve <10% positive cells) were considered as negative in this study. Primary categorical analysis was carried out as follows: positivity for CK5/6 and EGFR was set as ≥10% of cells staining, and for HER2, tumor cores that showed membrane-staining intensity of 2 or 3 were considered positive.
Human breast cancer cell lines and cell culture
The HBC cell line BT-20 was obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). Cells were cultured in Eagle’s Minimum Essential Medium (EMEM, Hyclone Laboratories Inc., Logan UT, USA) with 10% fetal bovine serum (PAA Laboratories Inc. Etobicoke, ON, Canada) supplemented with 100 units/mL penicillin, 100 mg/mL streptomycin, and 1mM pyruvate. Cells were grown at 37°C in an atmosphere of 95% air and 5% CO2.
Generation of stable claudin 1 knockdown clonal cell lines
BT-20 cells were stably transfected with a SureSilencing shRNA control sequence plasmid (SA Biosciences Corporation, Frederick, MD, USA), and two different shRNA sequences (sequence 3 and 4; SA Biosciences) specific for the claudin 1 gene using Lipofectamine 2000 (Life Technologies Inc,). Single clones were selected using Hygromycin B (Life Technologies, Inc.), and knockdown of claudin 1 was confirmed by Western blot analysis.
Subcellular fractionation
BT-20 cells were grown to 80% confluency and subcellular fractions were isolated using the ProteoExtract® Subcellular Proteome Extraction Kit (S-PEK, Calbiochem, Billerica, MA, USA) according to the manufacturer’s instructions. Protein fractions were subjected to acetone precipitation and pellets were reconstituted in sample isolation buffer (50 mM Tris-Cl pH 6.8: 5% SDS: 5mM β-glycerophosphate, containing complete mini protease inhibitor cocktail, Roche Diagnostics, Mississauga, ON, Canada). The mini BCA assay (ThermoScientific, Ottawa, ON, Canada) was used to determine the protein concentration of each fraction, prior to equal loading in 15% SDS-polyacrylamide electrophoresis gel and Western blotting.
Wound healing/migration assay
BT-20 cells were grown to full confluency on 6-well plates and a scratch was made through the cell monolayer using a pipette tip. After washing twice with PBS, fresh tissue culture medium was added and photographs (ScopePhoto 3.0, ScopeTek DCM130 microscope camera) of wounded areas were taken in a time-dependent manner up to 18 hours after making the scratch. Measurements of the wound area were evaluated using the Image-J program (National Institutes of Health).
Western blot analysis
Cells were lysed in an isolation buffer (50 mM Tris-Cl pH 6.8: 5% SDS: 5mM β-glycerophosphate, containing a complete mini protease inhibitor cocktail, Roche Diagnostics, Laval, QC, Canada) and mixed 3:1 with 4X sodium dodecyl sulfate (SDS) buffer [(500 mM Tris, pH 6.8), 40% glycerol, 8% SDS, 0.04% (w/v) bromophenol blue and 0.4M dithiothreitol (DTT)]. The samples were boiled for 5 min. at 100°C and electrophoresed in 15% SDS-polyacrylamide electrophoresis gel. Proteins were transferred to nitrocellulose, membranes were blocked in 5% non-fat milk in Tris-buffered saline with 0.05% Tween-20 (TBS-T) for 1 hr. Membranes were then incubated overnight at 4°C with primary antibodies (claudin 1, Life Technologies Inc.; β-actin, Abcam) diluted 1:1000, and 1:5000 respectively in blocking solution. Subsequently, the membranes were washed with TBS-T (three times 10 min.) and incubated with goat anti-rabbit or goat anti-mouse immunoglobulin G horseradish peroxidase conjugate (1:10000; Bio-Rad Laboratories Inc.) for 1 hr. at room temperature. The membrane was washed with TBS-T (three times 10 min.) and developed with Pico chemiluminescence substrate (Pierce Biotechnology, Rockford, IL, USA).
Fluorescent microscopy
For immunofluorescence staining, BT-20 cells were cultured on glass cover slips and fixed with 100% methanol for 20 min at -20°C. Cover slips were then rinsed with PBS and the cells were permeabilized with 0.2% Tween-20 in PBS for 5 min., followed by three 20 min. washes with PBS. After blocking with 1% BSA in PBS for one hour at room temperature, cells were incubated with the claudin 1 rabbit primary antibody (Life Technologies Inc., dilution 1:50) overnight at 4°C in a humid chamber. The cells were washed three times for 10 min. with PBS and incubated with secondary anti-rabbit antibody conjugated with Cy3 (dilution 1:100) for one hour at room temperature. Cells were washed again with PBS, incubated with 4′, 6-diamidino-2-phenylindole-dihydrochloride (DAPI) and mounted in FluorSave (Calbiochem).
Real-time PCR arrays
Cells were grown in EMEM in 6-well plates until 75-85% confluent and directly lysed by adding 350 uL Buffer RTL Plus from the RNeasy RNA extraction kit (Qiagen Sciences, Mississauga, ON, Canada). Equal amounts of RNA from two control clones were pooled and compared in triplicate with RNA from two claudin 1 knockdown clones. RNA (1μg/reaction) was reverse transcribed using the RT
2 First Strand Kit (SA Biosciences Corporation). cDNA samples (25ng) were applied to each real-time PCR reaction on the human EMT RT
2 Profiler PCR array (SA Biosciences Corporation) containing 84 key genes that change their expression during EMT. Real time PCR was carried out using the iCycler (BioRad Laboratories). The cycle profile consisted of denaturation at 95°C for 10 min., followed by 40 cycles of 95°C for 15 secs. and 60°C for 1 min. The iCycler iQ Optical System Software Version 3.0a (BioRad Resource Guide) was used to determine the cycle threshold (C
T) for each reaction. Data was analyzed using the web-based PCR Array Data Analysis Software (SA Biosciences Corporation;
http://www.sabiosciences.com/pcrarraydataanalysis.php). Five housekeeping genes were used as controls.
Statistical analysis
Analysis was carried out as previously described [
27,
28], using SAS 9.2 (SAS, Cary, NC) statistical software. The Wilcoxon Two Sample test and the Kruskal-Wallis test were used to interrogate claudin l levels in tumor subtypes and tumors from different age groups of patients. Associations between claudin 1 and other clinical-pathological variables were tested using contingency methods (continuity adjusted Chi-Square was used for node, age and size data; Exact Linear Association was used for grade). Linear regression analyses with claudin 1 levels as dependent were also carried out. Univariate survival analyses were performed using Cox regression to generate Kaplan-Meier curves. Overall survival (OS) was defined as the time from initial surgery to the date of death attributable to breast cancer only. Recurrence time was defined as the time from initial surgery to the date of clinically documented local or distant disease recurrence. Analysis of Variance (ANOVA) followed by Bonferroni’s Multiple Comparison Test were used to assess differences in migration rates in the wound healing assays.
Discussion
Based on the observation that claudin 1 is down regulated or absent in invasive HBC [
19‐
22], and that an absence of claudin 1 was shown to correlate with poor prognosis and shorter patient survival time [
23], it has been speculated that claudin 1 could be a putative tumor suppressor in breast cancer. However, these studies, including those from our laboratory, were carried out on breast tumors of mixed pathological lesions. Moreover, when the breast cancers were grouped according to ER status, we observed that not only was the frequency of claudin 1 expression significantly higher in the ER-ve cancers but that a higher level of the protein was also associated with the BLBC subtype; the latter has recently been confirmed by a report by Lu et al., [
32] as well as our present study. Additionally, in The Cancer Genome Atlas (TCGA) breast carcinoma provisional dataset, RNAseq analysis has shown claudin 1 to be up regulated in 17/81 (21%) of basal-like tumors compared with 2/324 (<1%) of luminal A/B cases [
33]. Since BLBCs are usually mesenchymal in phenotype and high claudin 1 is generally associated with epithelial phenotype, this result was unexpected. However high endogenous claudin 1 levels have also been observed in HBC cell lines as in the case of the BT-20 cell line and several other basal-like cell lines such as HCC1143, and HCC1937 [
34]. It is possible that in these breast cancer cells, claudin 1 has a different function.
An important finding of the present study was the significant association between claudin 1 and patient age. BLBC derived from women over 55 years of age were more likely to exhibit high claudin 1 expression. The significance of this observation is not known, but it is plausible that increased claudin 1 levels in these women may be related to decreased hormonal levels generally associated with the post-menopausal stage in a woman's life. As we have previously shown, there is a positive association between claudin 1 expression and ER-ve breast cancers [
19]. Thus, the relationship between estrogen and claudin 1 warrants further examination.
The present study also reveals a significant positive relationship between claudin 1 and claudin 4. However, interestingly, no significant association between claudin 4 and patient age was established suggesting that claudin 1 may have a unique role independent of claudin 4.
We also observed that mislocalization of claudin 1 to the cytoplasm was a frequent occurrence in BLBC. Such mislocalization of claudin 1 in the cytoplasm is not unique to breast cancer, as indeed there have been several recent reports of claudin 1 mislocalization in the cytoplasm, and in some cases, the nucleus, in a number of other cancers including melanomas, colon, and oral squamous and colon cancer [
11,
16‐
18,
35]. In these cancers, claudin 1 mislocalization was shown to increase the invasiveness of the cancer cells [
11,
16‐
18,
35]. This observation leads us to speculate that it is possible that cytoplasmic claudin 1 may have a different function from membranous claudin 1, as mislocalization of a number of membrane and subcellular proteins to the cytoplasm in some studies has been shown to impart tumorigenicity [
36‐
40].
We showed that stable shRNA knockdown of claudin 1 in BT-20 HBC cells resulted in a subsequent decrease in cell migration and motility. Claudin 1 knockdown also resulted in a significant up regulation of the expression of EMT related genes, SERPINE 1 (plasminogen activator inhibitor type 1, PAI1) and secreted phosphoprotein 1 (SSP1; also known as osteopontin) that have been shown to suppress cancer cell migration. In previous reports, SERPINE 1 was shown to inhibit cell migration during wound healing by blocking integrin from binding to vitronectin [
41]. Vitronectin enhances the migration of cells and is required for cell motility [
41]. Conversely, SERPINE 1 is also thought to have a role other than a protease inhibitor as it has been shown to decrease the adhesive strength of cells to their substratum. SERPINE 1 is also regulated by a variety of hormones and cytokines [
42]. This would be important if in older women, the up regulation of claudin 1 is related to their hormonal status, in particular, the lower estrogen level that is associated with the post-menopausal state. Another gene that was highly up regulated when claudin 1 was suppressed was SSP1. SSP1 is a phosphorylated glycoprotein secreted by several cell types, including those involved in bone turnover and is associated with bone metastasis in cancer [
43‐
45]. It is also secreted by cells of the immune system and is believed to be an early marker for breast cancer [
46]. The significant up regulation of these molecules in response to claudin 1 knockdown suggests that claudin 1 may be a regulator of genes associated with cancer progression and metastasis.
At the same time, we observed the down regulation of expression in another group of genes thought to be important for maintaining the EMT phenotype; TCF4, SNAIL2, FOXC2 and CALD1. SNAIL 2, a transcription factor and an important marker of EMT, has been shown to repress both E-cadherin, a master programmer of EMT [
47], and claudin 1 [
48‐
52]. TCF4, which belongs to the β-catenin pathway, is a member of the Zeb family of transcription factors. It has been suggested that claudin 1 is a targeted gene of β-catenin. Miwa et al. [
53] reported that in squamous cell carcinoma, TCF4 and β-catenin complexes bound TCF4 binding elements at two sites in the 5′ flanking region of the claudin 1 gene and that the binding promoted transcription of claudin 1. As well, SSP1, whose expression is significantly up regulated when claudin 1 is inhibited in this cell line, is a downstream target for TCF4 [
54]. TCF4 can act as a promoter or repressor of HBC progression by regulating SSP1 [
44,
54]. FOXC2 (forkhead box C2), another gene whose expression is significantly down regulated, is a sonic hedgehog (SHH) signaling molecule [
55]. Elevated levels of FOXC2 protein have recently been shown to be significantly associated with the BLBC phenotype and with poor disease free survival [
55]. Interestingly, SNAIL2, TCF4 and FOXC2 have been identified as part of the E-cadherin repressor interactome in EMT [
56] and are involved in many relationships regulating each other in a hierarchical pattern. In this general pathway, it is believed that SNAIL 2 is initially induced, leading to the activation of TCF4 and FOXC2. Also, knocking-down claudin 1 strongly increased the expression of the BMP7 gene, which belongs to one of the largest sub-families of transforming growth factor beta (TGFβ) [
57]. TGFβ, itself another important EMT molecule, has a dual role during tumor progression; initially as a suppressor, and then as a promoter. BMP7 is also known to display a number of diverse behaviors with regards to cell proliferation, cell migration, invasion and apoptosis in breast cancer cell lines, primary tumors as well as xenografts [
30,
31,
58‐
60]. Thus, the influence of claudin 1 on these signaling pathways in the BT-20 HBC cells hints at the complexity of its involvement in cellular processes and tumorigenesis [
13,
14].
The effect of claudin 1 on cell migration was dose dependent (although not statistically significant). We observed that the rate of migration of clone 3, a clone in which claudin 1 was almost completely knocked down, was slower (p<0.01 when compared with control cells) compared to the other clonal line, clone 4 (p<0.05).
Our earlier studies indicated that tumors which displayed the basal-like phenotype more frequently expressed claudin 1, and were also more likely to express higher levels of claudin 1. Many of these tumors also displayed mislocalization of claudin 1 to the cytoplasm, suggesting that the role of claudin 1 in the breast cancer cell is influenced not only by its level but by its location as well.
Altogether, our studies show that high claudin 1 protein levels are significantly associated with a particular group of older BLBC patients. In this regard, claudin 1 has the potential to serve as a marker for a subset of patients within the BLBC phenotype and in so doing may facilitate more personalized management of this disease. We also show in vitro that in basal-like HBC cells, claudin 1 inhibition results in decreased cell migration. Therefore, the expression of high claudin 1 levels in the BLBC subtype, particularly in women over 55 years of age suggests that these patients may warrant more aggressive treatment as their breast cancer may be more migratory resulting in a tendency to move away from the primary location.
Competing interest
The authors have no known conflicts of interests either financial or personal between themselves and others that might bias the work.
Authors’ contributions
YM, designed, and supervised the study. YM and AAB co-wrote the manuscript. AAB, and CP carried out H scoring evaluation of the TMAs. XM generated stable clonal cell lines. XM, DM, KD and SC carried out functional assays in the human breast cancer cell lines. KD was also responsible for the subcellular fractionation assays. EL and LCM contributed intellectually to many aspects of the study, and EL reviewed and edited the manuscript. All authors read and approved the final manuscript.