Early Detection and Diagnosis
Increased claudin-4 expression is associated with poor prognosis and high tumour grade in breast cancer
Article first published online: 19 NOV 2008
Copyright © 2009 Wiley-Liss, Inc.
International Journal of Cancer
Volume 124, Issue 9, pages 2088–2097, 1 May 2009
How to Cite
Lanigan, F., McKiernan, E., Brennan, D. J., Hegarty, S., Millikan, R. C., McBryan, J., Jirstrom, K., Landberg, G., Martin, F., Duffy, M. J. and Gallagher, W. M. (2009), Increased claudin-4 expression is associated with poor prognosis and high tumour grade in breast cancer. Int. J. Cancer, 124: 2088–2097. doi: 10.1002/ijc.24159
- Issue published online: 24 FEB 2009
- Article first published online: 19 NOV 2008
- Manuscript Accepted: 11 NOV 2008
- Manuscript Received: 30 AUG 2008
- Enterprise Ireland
- Irish Research Council for Science Engineering and Technology under the EMBARK initiative
- Cancer Research Ireland
- Science Foundation Ireland
- British Association for Cancer Research and the Health Research Board of Ireland
- Swedish Cancer Society
- Swegene/Wallenberg Consortium North
- Gunnar, Arvid and Elisabeth Nilsson Cancer Foundation
- Per-Eric and Ulla Schyberg Foundation
- Lund University Research Funds and Malmö University Hospital Research and Cancer Funds
- Programme for Third Level Institutions (PRTLI)
- breast cancer;
- tissue microarrays;
- prognostic biomarkers;
- tight junctions
- Top of page
- Material and methods
- Supporting Information
The role of intercellular tight junctions in breast epithelial cells is traditionally thought to be in maintaining polarity and barrier function. However, claudin-4, a tight junction protein, is overexpressed in breast tumour cells compared to normal epithelial cells, which generally corresponds to a loss in polarity. The aim of this study was to investigate the distribution and potential clinical value of claudin-4 in breast cancer, and to evaluate its usefulness as a prognostic and predictive biomarker. Expression of claudin-4 was initially examined by Western blot analysis in a cohort of 88 breast tumours, and was found to correlate positively with tumour grade and negatively with ER. Claudin-4 expression was then evaluated by immunohistochemistry in a larger cohort of 299 tumours represented on a tissue microarray. Claudin-4 expression correlated positively with tumour grade and Her2, and negatively with ER. High claudin-4 expression was also associated with worse breast cancer-specific survival (p = 0.003), recurrence-free survival (p = 0.025) and overall survival (p = 0.034). Multivariate analysis revealed that claudin-4 independently predicted survival in the entire cohort (HR 1.95; 95%CI 1.01–3.79; p = 0.047) and in the ER positive subgroup treated with adjuvant tamoxifen (HR 4.34; 95%CI 1.14–16.53; p = 0.032). This relationship between increased claudin-4 expression and adverse outcome was validated at the mRNA level in a DNA microarray dataset of 295 breast tumours. We conclude that high levels of claudin-4 protein are associated with adverse outcome in breast cancer patients, including the subgroup of patients treated with adjuvant tamoxifen. © 2008 Wiley-Liss, Inc.
During progression, breast cancer tissue becomes increasingly disorganised, with associated loss of defined ductal structure and reduced ability of mammary epithelial cells to polarise.1 These alterations are often accompanied by modified cell–cell and cell-matrix adhesions. One of the first adhesion proteins found to play a role in cancer progression was E-cadherin, the main transmembrane protein of the adherens cell–cell junctions, which is frequently lost during cancer progression.2 Tight junctions, the other main cell–cell adhesion complexes, are apically located and act to maintain cell polarity and control paracellular permeability, as well as creating a barrier between the apical and basolateral compartments of the plasma membrane.3 Integral to these junctions are members of the claudin family of transmembrane proteins.
The claudin family comprises 24 related members, whose expression is often tissue specific.4 These proteins have 4 transmembrane domains, and are considered to be the backbone of tight junctions. The extracellular loops of each claudin molecule bind to a claudin molecule on an adjoining cell at the so-called membrane ‘kissing points’. These structures, together with other integral proteins, such as occludin and junctional adhesion molecule (JAM), and a number of peripheral proteins, form the tight junction macromolecular complex.5 Tight junctions can vary in composition, permeability and ion specificity depending on the tissue and the state of differentiation.6
Although claudin proteins are primarily known for their cell adhesion function, they also interact with PDZ domain-containing molecules via a conserved YV domain at their C-terminus.3 Thus, claudin-4 has been shown to bind to the zonula occludens (ZO) proteins, cytoplasmic tight junction components which interact with numerous signalling pathways. Claudin-4 can also bind indirectly to actin filaments via ZO-1 and -2, which may allow the activity of tight junctions to affect cell polarity or motility.7–9 A large number of PDZ domain-containing proteins are now known to interact with claudins, and these are thought to be central to the intercellular signalling of tight junctions.5 Other molecules known to interact with claudin-4 include EphA210 and Protein Kinase Cϵ,11 both of which phosphorylate the cytoplasmic tail of claudin-4 and regulate its incorporation into tight junctions.
Several preliminary studies have been carried out on the expression of different claudins in various cancer types. Although loss of claudins has been reported in some cases, such as claudin-7 in breast cancer,12 several members of the claudin family have been found to be overexpressed in a wide variety of cancers. This includes claudin-4 in gastric cancer13, 14 and pancreatic cancer15; and claudins-3 and -4 in prostate cancer,16, 17 ovarian cancer,18 endometrial cancer19, 20 and breast cancer.21, 22 In breast cancer, previous studies on claudin-4 have been contradictory: one study found that claudin-4 expression was lost in Grade 1 invasive breast tumours, with increased expression detected in Grade 2 and 3 tumours22; others have found that claudin-4 is overexpressed in primary breast carcinomas compared to normal mammary epithelium.21 Both of these studies were carried out on small patient cohorts (n = 56 and n = 21, respectively). The only study carried out on a larger patient cohort (n = 184) found that claudin-4 was expressed in 92% of the breast carcinomas examined. Claudin-4 was found to have no association with tumour grade, type, ER or PR status, although it was strongly associated with expression of claudin-3.23 Claudin-4 may also be associated with the basal subtype of breast cancer, typified by ER, PR and Her2 negativity, and CK5/6 positivity. A recently published study which examined a cohort of 66 non-basal-like and 38 basal-like breast tumours found that expression of claudin-4 protein was increased in basal-like tumours.24 No previous study on claudin-4 expression in breast cancer has shown an association with survival.
Here, we attempt to clarify the role of claudin-4 in breast cancer by analysing its expression in two unrelated breast cancer patient cohorts using Western blot analysis and immunohistochemistry (IHC). We also investigate the expression of this gene at the mRNA level in a publicly available breast cancer dataset. In all three datasets analysed, we found a strong association between claudin-4 expression and adverse outcome in breast cancer.
Material and methods
- Top of page
- Material and methods
- Supporting Information
The first cohort of breast tissues investigated in this study by Western blot analysis (Cohort I) consisted of 88 primary breast carcinomas and 10 normal breast tissues, collected at St. Vincent's University Hospital, Dublin between 1994 and 2005. The median age was 60 years (range 30–81). Following surgical resection and pathological evaluation, breast cancer tissues were snap frozen in liquid nitrogen and stored at −80°C until use. Patient and tumour characteristics of this cohort are outlined in Table I.
|Claudin-4 negative (n = 8)||Claudin-4 positive (n = 80)||Median||p-value|
|≤50||3 (37.5)||22 (27.5)||0.2521||0.897|
|>50||5 (62.5)||58 (72.5)||0.3271|
|Ductal||7 (87.5)||74 (94.9)||0.2975||0.3171|
|Ductal & Lobular||1 (12.5)||1 (1.3)||0.0519|
|≤2cm||1 (14.3)||17 (22.7)||0.2675||0.622|
|>2cm||6 (85.7)||58 (77.3)||0.3047|
|Negative||2 (28.6)||37 (48.7)||0.3496||0.086|
|Positive||5 (71.4)||39 (51.3)||0.2306|
|I & II||7 (87.5)||27 (34.6)||0.1051||0.019|
|III||1 (12.5)||51 (65.4)||0.3288|
|Positive||7 (100)||58 (72.5)||0.2304|
The tissue microarray (TMA) used in this study was derived from a reference cohort of 512 consecutive invasive breast cancer cases diagnosed at the Department of Pathology, Malmo University Hospital, Malmo, Sweden, between 1988 and 1992 (Cohort II), and has been previously described.25 In brief, the median age was 65 years (range 27–96) and median follow-up time regarding disease-specific and overall survival was 11 years (range 0–17). Patients with recurrent disease and previous systemic therapies were excluded, as well as a number of misclassified ductal carcimona in situ (DCIS) cases. Two hundred and sixty-three patients were dead at the last follow-up (December 2004), 90 of which were classified as breast cancer-specific deaths. From the original cohort of 512 patients, samples were available from 299 patients for analysis of claudin-4 protein expression. These 299 samples had a higher proportion of tumours with low vascular endothelial growth factor (VEGF) (p = 0.043) and low vascular endothelial growth factor receptor 2 (VEGFR2) (p = 0.004) when compared to the 213 missing samples. No difference was seen in patient age (p = 0.502), tumour size (p = 0.775), grade (p = 0.751) or histological type (p = 0.324), nodal status (p = 0.471), ER (p = 0.712), PR (p = 0.921) or Her2 (p = 0.602) expression between available and unavailable samples. Patient and tumour characteristics of the available cohort are outlined in Table II. The study has been approved by the Ethics Committee at Lund University and Malmo University Hospital.
|Low claudin-4 (n = 157)||High claudin-4 (n = 142)||p-value|
|Median (range)||66 (35–96)||64 (35–91)||0.003|
|≤50||14 (8.9)||30 (21.1)|
|>50||143 (91.1)||112 (78.9)|
|Median (range)||16 (0–100)||17 (5–100)||0.508|
|≤2cm||102 (65)||87 (61.3)|
|>2cm||55 (35)||55 (38.7)|
|Ductal||100 (69.4)||105 (79.5)||0.0851|
|Lobular||27 (18.8)||12 (9.1)|
|Tubular||10 (6.9)||8 (6.1)|
|Medullary||2 (1.4)||5 (3.5)|
|Mucinous||5 (3.5)||2 (1.5)|
|Negative||87 (64.4)||80 (60.2)||0.468|
|Positive||48 (35.6)||53 (39.8)|
|I & II||113 (72.4)||84 (59.2)||0.016|
|III||43 (27.6)||58 (40.8)|
|Negative||16 (10.5)||29 (21)||0.014|
|Positive||136 (89.5)||109 (79)|
|Negative||42 (33.9)||48 (43.2)||0.140|
|Positive||82 (66.1)||63 (56.8)|
|Ki 67 (%)|
|<10%||65 (42.8)||49 (35.8)||0.224|
|>10%||87 (57.2)||88 (64.2)|
|Low (0–2+)||89 (78.8)||94 (86.2)||0.143|
|High (3)||24 (21.2)||15 (13.8)|
|Low (0–2+)||98 (89.1)||91 (86.7)||0.586|
|High (3)||12 (10.9)||14 (13.3)|
|Low (0–1)||89 (87.3)||79 (76.7)||0.049|
|High (2–3)||13 (12.7)||24 (23.3)|
|Cyclin D1 (%)|
|Low (0–25%)||51 (34)||57 (41.9)||0.168|
|High (>25%)||99 (66)||79 (58.1)|
Haematoxylin and eosin stains of all 512 tumours were re-evaluated and areas representative of invasive cancer were marked. The TMA was constructed using an automated tissue arrayer (MTA-27, Beecher Inc., WI). Two 0.6 mm tissue cores were extracted from each donor block and placed into a recipient block. The total number of cores per block was limited to ∼200 (100 patients), with a total of 5 blocks arrayed.
A panel of breast cell lines was selected to test antibody specificity. This included three ER-positive cell lines (MCF-7, T47D, BT474) and four ER-negative cell lines (MDA-MB-231, SKBR3, Hs578T and MCF10a). All cell lines were purchased from European Collection of Cell Cultures (Wiltshire, UK). The MCF-7, MDA-MB-231, SKBR3, BT474 and T47D cell lines were cultured in DMEM supplemented with 10% (w/v) foetal calf serum, 2 mM L-glutamine, 50 IU/ml penicillin, and 50 μg/ml streptomycin sulphate. Hs578t cells were also supplemented with 10 μg/ml insulin. MCF10a cells were grown in DMEM/F12 supplemented with 5% horse serum, 20 ng/ml EGF, 0.5 mg/ml hydrocortisone, 100 ng/ml cholera toxin, 10 μg/ml insulin, 50 IU/ml penicillin and 50 μg/ml streptomycin sulphate. Cells were maintained in humidified air with 5% CO2. Studies of protein expression were performed on cells approaching confluence. All cell lines were free of mycoplasma contamination.
Western blot analysis
For analysis of cell lines, protein was extracted from cells approaching confluence by the addition of radioimmunoprecipitation assay buffer (RIPA) followed by centrifugation at 16,000g for 20 min at 4°C. The supernatants were removed and the protein levels determined using the bicinchoninic acid (BCA) method (Pierce, IL). Samples containing 25 μg aliquots of protein were separated by SDS-PAGE on a 12% polyacrylamide gel under reducing conditions. Membranes were blocked in 5% non-fat milk for 1 hr at room temperature. Claudin-4 protein expression was detected using a mouse monoclonal anti-claudin-4 antibody (Zymed; clone 3E2C1) at a dilution of 1:1000 applied overnight at 4°C. Membranes were washed in TBS-T (0.1% Tween 20) and incubated for 1 hr with horseradish peroxidase-conjugated anti-mouse immunoglobulin (1:5000 dilution). The blots were again washed in TBS-T. HRP was detected using Enhanced Chemilunimescence plus (Amersham Biosciences, UK). Chemiluminescence was detected by autoradiography using X-ray film. Membranes were stripped and reprobed with anti-β-actin (1:5000 dilution; Abcam, UK) as a loading control.
For analysis of breast tissues by Western blotting (Cohort I), tissue samples were homogenised using a Mikro-Dismembrator (Braun Biotech T1 International, Germany). Protein was extracted from homogenised tissue samples using 50 mmol/l Tris-HCl (pH 7.4) containing protease inhibitor cocktail (Roche) and Triton X-100 (1%) under agitation at 4°C for 1 hr. Determination of protein concentration and Western blot procedure were carried out as above. Positivity of expression was determined based on the presence of a visible protein at the correct molecular weight. The intensity of protein expression observed was semi-quantified using the UVIBandMap programme (Windows application VIO.02), with normalisation of claudin-4 protein levels against β-actin. Specificity of the antibody reaction was confirmed by: (i) omission of the primary antibody and (ii) detection of a similar pattern of expression using a second commercially available primary antibody against claudin-4 (clone 6A43; US Biologicals, USA).
Cell pellet arrays
To confirm Western blotting results, a cell pellet array was constructed and IHC was performed on the same panel of breast cancer cell lines. Cells were trypsinised and fixed for 1 hr in 4% formalin, following which they were centrifuged and resuspended in 0.8% agarose. The gel plugs containing fixed tumour cells were then processed through gradient alcohols before being cleared in xylene and washed multiple times in molten paraffin. Once processed, cells were embedded in paraffin, and then arrayed in triplicate 0.6 mm cores using a manual tissue arrayer (MTA-1, Beecher Inc, WI). IHC was carried out on 5 μm sections.
Sections of cell pellet arrays or TMAs were deparaffinised in xylene and rehydrated in descending gradient alcohols. Heat-mediated antigen retrieval was performed using 10 mM sodium citrate buffer (pH 6.0) in a PT module (LabVision, UK) for 15 min at 95°C. The LabVision IHC kit (LabVision, UK) was used for staining. Endogenous peroxidase activity was blocked by incubation with 3% hydrogen peroxide for 10 min. Sections were blocked for 10 min in UV blocking agent and the anti-claudin-4 monoclonal antibody (Zymed, USA) was applied at a dilution of 1:1000 for 1 hr. Sections were washed in phosphate buffered saline with 0.1% Tween 20 (PBS-T), following which primary antibody enhancer was applied for 20 min, and sections were washed in PBS-T again. Sections were then incubated with HRP polymer for 15 min, washed in PBS-T and then developed for 10 min using diaminobenzidine (DAB) solution (LabVision, UK). All the incubations and washing stages were carried out at room temperature. The sections were counterstained in haematoxylin, dehydrated in alcohol and xylene and mounted using an automated coverslipper (Leica, Germany). As a negative control, the anti-claudin-4 antibody was substituted with an anti-mouse immunoglobulin G isotype control.
TMA sections (4 μm) were examined for claudin-4 protein expression using the same protocol as outlined earlier. TMA sections had been previously been stained in the Ventana Benchmark (Ventana Medical Systems Inc, USA) using prediluted antibodies to ER (clone 6F11, Ventana), PR (clone 16, Ventana) and Her2 (Pathway CB-USA 760-2694), or in the Dako Techmate 500 (Dako, Denmark) for Ki-67 (1:200, M7240, Dako), VEGF-A (1:400, A20, Santa-Cruz, CA, USA) and cyclin D1 (1:100, M7155, Dako).
Evaluation of immunohistochemical staining
Slides were scanned at 20× magnification using a ScanScope XT slide scanner (Aperio Technologies, CA). For manual scoring, tumour samples were evaluated by two independent observers, including one pathologist, and scored for claudin-4 expression on a scale of 0–3 where 0 is negative, 1 is weakly positive, 2 is medium positive and 3 is strongly positive (Fig. 1c). Only membranous staining was classified as positive. The mean value of both scores for each patient was used for statistical analysis. High claudin-4 was classified as tumours with a staining intensity of ≥2, and low claudin-4 was classified as tumours with a median intensity of <2.
For automated scoring, cores were analysed using a membrane algorithm (Aperio Technologies), designed to identify cell membranes following DAB staining. The average positive pixel intensity value for each core was used for statistical analysis. This was measured on a continuous scale from 0 (positive) to 255 (negative). For analysis, this continuous data was separated into three equal groups, with the groups expressing low and moderate claudin-4 being combined, as their survival curves overlapped, and compared to the group expressing high claudin-4. This resulted in two groups for analysis, separated at the 66th percentile.
Statistical analysis of DNA microarray data
To evaluate claudin-4 expression in a third patient group, relevant histopathological and clinical data relating to 295 patients with breast cancer26 was downloaded from a publicly available website (Rosetta Inpharmatics Inc., http://www.rii.com). For claudin analysis, the log ratios of gene expression values were extracted from the dataset and used without modification. Tumour samples were classified using a previously published method.27 First, samples were separated into three equally sized groups according to claudin-4 mRNA expression. The two groups with lower claudin-4 mRNA expression were then combined, and the survival curves of the resulting two groups were compared using the log rank test. The χ2 test and Fisher's exact test were used for relating claudin-4 mRNA levels to clinicopathological variables.
For Cohort I, the Mann–Whitney U test and the Kruskal–Wallis test were used to compare claudin-4 expression (as a continuous variable) to tumour characteristics and ER status. The χ2 test and Fisher's exact test were used to evaluate associations between claudin-4 expression and clinico-pathological characteristics in Cohort II, and the van de Vijver dataset. Kaplan-Meier plots were used for survival analysis, and the log rank test was used to compare curves separated according to claudin-4 expression. Cox proportional hazards regression was used to estimate proportional hazards ratios for claudin-4 and other clinicopathological variables, in both univariate and multivariate models. Spearman's Rho test was used to compare manual and automated scoring methods. All calculations were carried out using SPSS version 12 (SPSS Inc., Chicago, IL).
- Top of page
- Material and methods
- Supporting Information
Expression pattern of claudin-4 in breast cancer cell lines
The specificity of the claudin-4 antibody was initially validated by Western blot analysis on a panel of breast cancer cell lines. Claudin-4 migrated with an approximate molecular mass of 22 kDa, with high expression observed in MCF7, T47D and BT474 cells, low expression in SKBR3 and MDA-MB-231 cells, and no detectable expression in Hs578t or MCF10a cells (Fig. 1a). This expression pattern seems to closely follow expression of the estrogen receptor (ER). Specificity of the claudin-4 antibody was verified by performing IHC on the corresponding formalin-fixed, paraffin-embedded breast cancer cell lines. This immunohistochemical examination showed membranous expression of claudin-4 in all cell lines which were found to be positive by Western blot. T47D cells were subsequently used as a positive control and Hs578t cells as a negative control (Fig. 1b) for all further IHC studies.
Claudin-4 protein expression in breast carcinoma tissue determined by Western blot analysis
To establish if claudin-4 was expressed in breast cancers, Western blotting was carried out on a discrete panel of breast tumour samples (Cohort I). Of the 88 samples investigated, 80 (90.9%) displayed a specific protein migrating with an approximate molecular mass of 22 kDa. No significant correlation was found between claudin-4 expression and either tumour size, presence or absence of axillary node metastasis, histology type or patient age at time of diagnosis (Table I). Levels of claudin-4, however, were significantly higher in ER-negative tumours compared to ER-positive tumours (p = 0.006), and in high grade (Grade 3) versus low grade tumors (Grade 1 and 2) (p = 0.001) (Fig. 2).
Claudin-4 expression in breast tumours as determined by IHC
Claudin-4 expression was determined by IHC in 299 breast tumours present on a TMA (Cohort II). Staining was predominantly membranous, with some claudin-4 positive samples displaying a low level of cytoplasmic staining. Two hundred and ninety-nine of the original cohort of 512 tumours (58.4%) were available for analysis. Two hundred and thirteen tumours were not available for analysis due to core loss during sectioning and staining. To assess for any possible bias in the tumours available for analysis, clinico-pathological characteristics for the available (n = 299) and unavailable (n = 213) samples were compared. A similar distribution of most variables was seen between available and unavailable samples, apart from VEGF (p = 0.043) and VEGFR2 (p = 0.004), which were both lower in the available samples. Of the 299 tumours available, 279 (93.3%) expressed claudin-4. Staining was scored on a scale from 0 to 3 based on intensity of staining (Fig. 1c). High claudin-4 expression was defined as a staining intensity of ≥2, and low claudin-4 expression was defined as a staining intensity of <2. On the basis of this analysis, 142 tumours (47.5%) were classified as expressing high levels of claudin-4 and 157 tumours (52.5%) were classified as expressing low levels of claudin-4.
Correlation of claudin-4 expression with clinicopathological and immunohistochemical parameters
On the basis of the previously mentioned classification of claudin-4 expression, we investigated possible associations between claudin-4 expression and a variety of well-defined clinico-pathological variables in Cohort II (Table II). Claudin-4 expression correlated positively with tumour grade (p = 0.016) and Her2 expression (p = 0.049), while showing a negative association with ER expression (p = 0.014). High levels of claudin-4 were also found to be more prevalent in tumours from younger women (p = 0.003). Claudin-4 expression was found to be independent of tumour size or type, or lymph node metastasis.
Claudin-4 expression is associated with shorter patient survival
We proceeded to examine the relationship between claudin-4 expression and survival. In agreement with our findings showing that claudin-4 was correlated with a number of factors associated with aggressive disease, we found a significant correlation between high levels of claudin-4 expression and reduced breast cancer-specific survival (BCSS) (p = 0.003), recurrence-free survival (RFS) (p = 0.025) and overall survival (OS) (p = 0.034) (Fig. 3). The 10-year RFS for patients expressing low claudin-4 was 71%, compared to 58% for patients with high levels of claudin-4.
To compare the prognostic impact of claudin-4 with established factors, Cox regression analysis was carried out. Univariate analysis of BCSS demonstrated that high levels of claudin-4 expression were significantly associated with shortened BCSS (HR 2.24; 95% CI 1.30–3.88; p = 0.004), as well as tumour grade, nodal status and Ki-67. ER and younger age at diagnosis correlated with extended BCSS (Table III). Using multivariate analysis, claudin-4 remained a significant predictor of reduced BCSS (HR 1.95; 95% CI 1.01–3.79; p = 0.047) when adjusted for other well-established variables. Tumour grade (HR 3.96; 95% CI 1.79–8.79; p = 0.001) and nodal status (HR 3.48; 95% CI 1.76–6.88; p = <0.001) also remained predictive of reduced BCSS.
|Entire cohort (n = 299)||ER positive, tamoxifen treated patients (n = 87)|
|Prognostic Factor||HR||(95% CI)||p||HR||(95% CI)||p||HR||(95% CI)||p||HR||(95% CI)||p|
|Claudin-4 (high vs. low, ref)||2.24||(1.30–3.88)||0.004||1.95||(1.01–3.79)||0.047||3.85||(1.47–10.06)||0.006||4.34||(1.14 –16.53)||0.032|
|Age (>50 vs. ≤ 50, ref)||0.61||(0.38–0.98)||0.042||0.73||(0.35–1.52)||0.401||0.58||(0.18–1.93)||0.377||1.32||(0.20–8.80)||0.776|
|Tumour size (continuous)||1.01||(1.01–1.02)||0.001||1.00||(0.99–1.02)||0.554||1.02||(1.01–1.03)||0.001||1.01||(0.99–1.03)||0.235|
|Tumour grade (3 vs. 0–2, ref)||4.31||(2.81–6.63)||<0.001||3.96||(1.79–8.79)||0.001||3.20||(1.51–6.78)||0.002||3.49||(0.73–16.67)||0.118|
|Nodal status (pos vs. neg, ref)||5.35||(3.34–8.56)||<0.001||3.48||(1.76–6.88)||<0.001||3.18||(0.96–10.56)||0.059||2.00||(0.36–11.01)||0.428|
|ER (pos vs. neg, ref)||0.50||(0.30–0.83)||0.007||1.02||(0.47–2.18)||0.965||–||–||–||–||–||–|
|Ki–67 (>10% vs ≤10%, ref)||2.46||(1.46–4.14)||0.001||1.30||(0.56–3.03)||0.542||2.25||(0.96–5.27)||0.062||0.64||(0.13–3.19)||0.582|
|Her2 (2,3 vs. 0,1, ref)||1.75||(0.99–3.08)||0.053||1.33||(0.67–2.65)||0.418||1.06||(0.31–3.62)||0.926||1.37||(0.29–6.59)||0.693|
Subset analysis revealed that increased claudin-4 expression was associated with a reduced BCSS (p = 0.010) and RFS (p = 0.039) in the ER-positive group of patients (n = 243; Fig. 3). Further analysis of this group showed that high claudin-4 expression was associated with shorter BCSS in the ER-positive patients treated with adjuvant tamoxifen (n = 87; p = 0.003), but not in the untreated group (n = 98; p = 0.349). The number of ER-negative patients (n = 45) was too small for meaningful statistical analysis.
In order to further examine a possible role for claudin-4 in tamoxifen resistance, we carried out additional Cox regression analysis focussing on the ER positive, tamoxifen treated patients (n = 87). In a univariate model, claudin-4 was associated with a decreased BCSS (HR 3.85; 95% CI 1.47–10.06; p = 0.006), as well as tumour size and grade. In a multivariate model of this group of patients, which included all well-established markers, claudin-4 was an independent prognostic marker for BCSS (HR 4.34; 95% CI 1.14–16.53; p = 0.032) (Table III).
Automated scoring of claudin-4 immunohistochemical staining correlates well with manual scoring
In order to evaluate the accuracy of the manual scoring method used to score the TMA, we also performed automated analysis using a membranous algorithm (Aperio). An excellent correlation was seen between automated and manual scores (Spearman's Rho = 0.841; p < 0.001). The continuous data generated by the algorithm was separated into two groups using the 66th percentile as a cut-off. Using these groups, we looked at claudin-4 in relation to survival, and found that it correlated well with BCSS (p = 0.019) and RFS (p = 0.016) (Supp. Info. Fig. S1). We also examined correlations between with clinico-pathological and automated IHC data. In agreement with manual analysis, we found that high levels of claudin-4 as determined by image analysis were associated with increased tumour grade (p = <0.001), ER negativity (p = <0.001), younger age at diagnosis (p = 0.012), and Her2 expression (p = 0.002). Furthermore, high levels of claudin-4 as determined by image analysis were associated with PR negativity (p = 0.024), high Ki-67 expression (p = 0.009) and low Cyclin-D1 expression (p = 0.024) (Supp. Info. Table S1). These results suggest that automated image analysis may be a useful technique for examination of claudin-4 IHC.
Claudin-4 is associated with poor prognosis at the mRNA level in a publicly available breast cancer DNA microarray dataset
In order to validate these results at the mRNA level, we analysed claudin-4 expression and its association with survival in an independent DNA microarray dataset derived from 295 breast tumours, which has previously been described in detail.26 High expression of claudin-4 mRNA in this dataset was found to be associated with a shorter time to overall survival (p = 0.033), as well as a reduced metastasis-free survival (p = 0.010) (Fig. 4). Increased levels of claudin-4 mRNA were also associated with higher tumour grade (p = 0.050) and ER negativity (p < 0.001). When tumours were classified into molecular subtypes as defined by Sorlie et al.,28 those expressing high levels of claudin-4 were overrepresented in the basal subtype, and under-represented in the Luminal A subtype (p = 0.006). Increased claudin-4 mRNA expression was also found to be associated with the 70-gene prognostic signature derived by van 't Veer et al.29 (Supp. Info. Table S2). Association with tamoxifen response could not be determined within this cohort due to low numbers of ER-positive tamoxifen-treated patients.
- Top of page
- Material and methods
- Supporting Information
The aim of this study was to determine the prognostic significance of claudin-4 expression in breast cancer. Previous studies on claudin-4 in breast cancer have looked at much smaller cohorts, with limited clinical data, and no information on patient survival. In this study, we first examined a panel of breast cancer cell lines for claudin-4 expression by Western blot analysis, and found that expression seemed to be higher in ER-positive cell lines. We then looked at a small cohort of normal and tumour samples (Cohort I) by Western blot analysis. Expression of claudin-4 was significantly higher in primary breast carcinomas compared to normal breast tissue, in ER-negative tumours compared to ER-positive tumours, and in grade 3 tumours compared to Grade 1 and 2 tumours. No decrease was seen in Grade 1 tumour samples compared to normal, as has been previously reported.22 The conflicting results obtained from cell line models and clinical samples highlights the limitations of relying solely on cell line models. In fact, the more invasive cells examined in this study (Hs578t and MDA-MB-231), which were negative for claudin-4 expression, have been found based on their phenotype and invasiveness, to more closely resemble mesenchymal or stromal cells than epithelial cells, and thus may not be good models for examination of claudin-4 expression in the ER-negative context.30–32
We then proceeded to analyse claudin-4 expression in a larger cohort of 299 samples on a TMA (Cohort II), and demonstrated a conclusive link between increased claudin-4 expression and reduced patient survival. Within this cohort, 279 tumours (93.3%) showed claudin-4 positivity, similar to the levels of positivity reported by previous studies.22, 23 In a previous study with a comparable number of samples (n = 184), no association was found between claudin-4 expression and tumour grade or hormone receptor status.23 The reason for this may be that of 184 samples, 179 were classed as positive, and of these 179, 134 were classed as strongly positive, with the rest being classed as weakly or moderately positive. This lack of distribution of scores may be due to a disproportionately large number of high grade tumours in this cohort, or inadequate discrimination between levels of staining. Further studies on the expression of this protein in breast cancer have found an association with higher grade tumours; however, sample numbers were small, and hormone receptor status or survival status were not investigated in relation to claudin-4 expression.21, 22 Interestingly, a recent study found that claudin-4 expression was associated with the basal-like subtype of breast cancer,24 a finding which correlates with our analysis of a publicly available DNA microarray dataset,26 showing that claudin-4 mRNA expression is increased in basal-like breast tumours.
Further analysis of the 299 samples on the TMA revealed that claudin-4 expression was associated with predictors of poor prognosis. Despite an association with ER-negative tumours, which is typical of a poor prognostic marker, claudin-4 was found to be a significant predictor of BCSS and RFS within the ER positive cohort alone. Furthermore, in patients with ER-positive tumours who were treated with tamoxifen, high claudin-4 expression was an indicator of poor response to treatment, and may be a predictor of tamoxifen resistance. These findings were substantiated by a multivariate Cox regression analysis which showed claudin-4 to be an independent predictor of survival, both in the overall cohort, and more significantly, in the ER-positive patients who had received tamoxifen. No previous studies have shown an association between claudin-4 expression and tamoxifen response. It should be noted that patients in the tamoxifen treated group would have had a worse prognosis and were mostly node-positive, whereas patients in the untreated group were mostly node-negative. These variables were accounted for in the multivariate model of analysis mentioned previously, but a cohort receiving randomised tamoxifen treatment would be needed to verify this association. Nevertheless, in the tamoxifen-treated group of patients, claudin-4 is an excellent independent marker of poor prognosis.
With regards to the effect of claudin-4 expression on tamoxifen resistance, it may be that claudin-4 is involved in the mechanism of tamoxifen resistance, or it is possible that patients with lower levels of claudin-4 are intrinsically lower risk and will not benefit from tamoxifen treatment. Previous studies have shown a relationship between claudin-4 and ER: examination of an estrogen responsive gene database shows that claudin-4 has a putative ERE in its promoter,33 and claudin-4 has been shown to be downregulated at the mRNA level after estrogen treatment of a human breast cancer cell line.34, 35 It is, therefore, possible that tamoxifen treatment could induce an increase in claudin-4 expression. However, further investigation of this putative relationship is needed, using both clinical and functional approaches, in order to determine the role of claudin-4 in tamoxifen response.
Although it is clear that claudin-4 is overexpressed in breast tumours, and that this overexpression is predictive of poor outcome, the mechanism by which this is occurring is unclear. Several recent studies have examined the methylation status of claudin-4 in different tumour types, and it is possible that hypomethylation of the claudin-4 promoter may be responsible for its overexpression in breast cancer, as has been reported for ovarian cancer.36, 37 Further studies on claudin-4 found that it is upregulated in response to EGF treatment in Madin-Darby canine kidney cells.38, 39 EGF is commonly overexpressed in breast cancer, and an increase in EGF expression in breast tumours could drive claudin-4 expression. Functional studies in a number of models have also found that claudin-4 can be phosphorylated by a number of mechanisms, and this phosphorylation leads to reduced integration of claudin-4 into sites of cell–cell contact, and disruption of the barrier function. Two molecules shown to control this phosphorylation were EphA2, demonstrated in a colon carcinoma cell line,10 and PKCϵ, shown in an ovarian cancer cell line,11 both of which induced similar effects on barrier function, despite phosphorylating different amino acids on the carboxy terminus of claudin-4. These studies suggest that claudin-4 is involved in a complex signalling network, and more than one mechanism is likely to be responsible for its overexpression.
Indeed, this complexity in signalling may explain the cell-specific effects of claudin-4. Although it has been reported to be upregulated in a number of cancers, a study has shown that claudin-4 is downregulated in high-grade bladder cancers, and that this downregulation is induced by promoter hypermethylation.40 This is the complete opposite to what has been observed for claudin-4 in ovarian cancer.36 Even when claudin-4 is overexpressed, its effects vary considerably between cancer types. In ovarian cancer, overexpression of claudin-4 in an immortalised ovarian epithelial cell line was found to increase invasion, migration, cell survival and MMP-2 expression.41 Conversely, claudin-4 overexpression in a pancreatic cancer cell line led to reduced invasion, migration and anchorage-independent growth in vitro, and reduced metastasis in vivo, while having no effect on MMP-2 expression.42 These diverse effects may be dependent on the expression of other members of the claudin protein family, or on the signalling environment within the cell. The cell type is certainly important: studies on claudin-4 knockdown in MDCK and LLC-PK1 cells have shown that claudin-4 can act as a barrier to Na+ ions, or a channel for Cl− ions, depending on the cellular context. This may, in part, account for the wide variation of effects seen with claudin-4 overexpression. Certainly, these studies make it clear that the function of claudin-4 in breast cancer needs to be investigated fully and cannot be inferred from results in other tissue types.
Another question in relation to claudin-4 overexpression is whether or not it is associated with functional tight junctions. Although increased cytoplasmic localisation of overexpressed claudin-4 was observed in one study on ovarian cancer,18 other work has contradicted this,36 and no significant level of cytoplasmic staining was seen in this study, in line with previous studies in breast tumours.21–23 However, membranous localisation does not necessarily imply that claudin-4 is functioning correctly, and a study on claudin-4 overexpression in colorectal cancer found that tight junctions were disorganised and paracellular permeability was increased, despite claudin-4 being localised predominantly in the membrane.43
In view of the results of this study, and others mentioned above, it is likely that the traditional view of adhesion proteins being lost with progression of breast cancer will need to be revised in the case of these tight junction proteins. Although we have shown that claudin-4 is overexpressed in breast cancer and that this overexpression is predictive of shorter survival, the mechanism by which claudin-4 is exerting this effect is unclear. Interestingly, previous studies on mammary gland development in the mouse have shown that claudin-4 is highly upregulated in the mammary gland during pubertal development, a time of rapid cellular growth and invasion.44 Many parallels have been drawn between mammary gland development and breast cancer, due to the overlap of signalling pathways driving these processes,1 and it is possible that this model of development may be useful in elucidating the mechanism of claudin-4 function in growth and invasion.
As a tumour biomarker, claudin-4 is easily quantifiable by IHC and thus readily translatable to the clinic. Claudin-4 could well prove to be a prognostic biomarker for several types of epithelial tumours, as it has been shown to be associated with increased grade in a number of cancers, including prostate cancer,17 and associated with poor patient outcome in renal cell carcinoma,45 endometrial cancer20 and gastric adenocarcinoma.14 In particular, claudin-4 has been shown to independently predict survival for gastric adenocarcinoma.14 Moreover, a specific agent exists, known as clostridium perfringens enterotoxin (CPE), which can target claudin-3 and -4 and kill cells expressing these proteins. This toxin has been shown to be effective in treating breast tumour xenografts in mice,21 and intracranial treatment was found to inhibit tumour growth and increase survival in murine models of breast cancer metastasis to the brain.46 Studies in other cancer types show that CPE may be effective in tumour cells which are highly resistant to chemotherapy.47, 48 CPE is also effective in diminishing the barrier function of epithelial cells expressing claudin-3 or 4,49 which may prove useful in increasing uptake and effectiveness of cytotoxic drugs. If claudin-4 does prove to be a useful breast cancer biomarker, then the existence of a readymade target treatment would be invaluable for future therapies.
- Top of page
- Material and methods
- Supporting Information
Funding is acknowledged from Enterprise Ireland, the Irish Research Council for Science Engineering and Technology under the EMBARK initiative (for support of Ms. Fiona Lanigan's post-graduate studies), Cancer Research Ireland (for part-support of Dr. Brennan's post-graduate studies), Science Foundation Ireland, the British Association for Cancer Research and the Health Research Board of Ireland, the latter under the auspices of the ‘Breast Cancer Metastasis: Biomarkers and Functional Mediators’ research programme. The cross-national component of the project was facilitated by the Marie Curie Transfer of Knowledge Industry-Academia Partnership research programme, TargetBreast (www.targetbreast.com).
- Top of page
- Material and methods
- Supporting Information
- 24Expression of tight junction protein Claudin-4 in basal-like breast carcinomas. Pathol Oncol Res, in press., , , , , , .
- 36Claudin-4 overexpression in epithelial ovarian cancer is associated with hypomethylation and is a potential target for modulation of tight junction barrier function using a C-terminal fragment of Clostridium perfringens enterotoxin. Neoplasia 2007; 9: 304–14., , , , , , , , , , , , et al.
- Top of page
- Material and methods
- Supporting Information
Additional Supporting Information may be found in the online version of this article.
|IJC_24159_sm_suppfigure1.tif||1659K||Comparison between manual and automated scoring methods in Cohort II (a) Kaplan-Meier estimates of BCSS and (b) RFS using the automated scoring method, with the continuous data separated into two groups at the 66th percentile.|
|IJC_24159_sm_supptable1.doc||55K||Association of claudin-4 with clinicopathological parameters in Cohort II using the automated scoring method.|
|IJC_24159_sm_supptable2.doc||40K||Analysis of claudin-4 expression at the mRNA level within the van de Vijver dataset.|
Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.