Main

Lung cancer is the leading cause of cancer mortality in the Western countries. Four major histological subtypes have been described: squamous cell carcinoma, adenocarcinoma, small cell and large cell carcinomas.1 Squamous cell carcinomas and adenocarcinomas account for over a half of all lung cancers, and are believed to originate, respectively, from squamous cell dysplasia/carcinoma in situ,1, 2 and atypical adenomatous hyperplasia3 of lung epithelial cells. However, the precise histogenetical origin of these carcinomas is unknown: it could include pluripotent stem cells or differentiated cells and whether a common pluripotent progenitor cell exists is questioned by different authors. Immunohistochemistry with antibodies such as p63, TTF-1 and keratins 5/6/7/14/17, CD15 and thrombomodulin is currently used to distinguish pulmonary squamous cell carcinomas from adenocarcinomas.4, 5, 6

Epithelial cells are characterized by tight junctions, which seal their apical regions in a gasket-like manner, thus preventing the free diffusion of solutes (including growth factors and pathogens) across extracellular luminal and basal compartments. Recent studies have shown that some tight junction proteins are involved in the regulation of epithelial proliferation and differentiation (reviewed by Matter et al7). Tight junctions comprise transmembrane proteins and cytoplasmic proteins. Among tight junction membrane proteins, the claudin (CLDN) family includes over 20 members, which show tissue-specific expression, and are crucial for the tight junction barrier function. Several studies have shown an altered expression of CLDNs in human tumors, including tumors of the gastrointestinal tract,8, 9 ovary,10 pancreas,11 breast,12, 13 epidermis,14 liver,15, 16 uterus17 and prostate.18 Little information is available on the expression of cytoplasmic tight junction proteins in human cancers,19, 20, 21, 22 and only one study has been published about the immunohistochemical localization of a tight junction protein (occludin (OCLN)) in lung tumors.23

Considering the role of tight junctions in epithelial cell function, the involvement of tight junction proteins in the regulation of epithelial proliferation, and their potential usefulness as novel tools in cancer diagnosis, prognosis and treatment,24, 25 it is essential to understand in detail how tumor development correlates to changes in tight junction protein expression. To address this issue, we used immunocytochemistry and the quantitative reverse transcription-polymerase chain reaction (qRT-PCR) to examine the expression of several cytoplasmic and transmembrane proteins of tight junctions in human lung squamous cell carcinomas and adenocarcinomas. Our results provide new diagnostic tools for the differential diagnosis between these tumors and suggest new insights about tumor differentiation.

Materials and methods

Tissue Collection

The study material consisted of 90 carcinomas of the lung (41 adenocarcinomas and 49 squamous cell carcinomas) obtained from the archives of the Department of Pathology at the Geneva University Hospital, in accordance with approved protocols. Among the patients, 31 were females and 59 were males, with age ranging from 30 to 85 years (median 66).

For histopathological and immunohistochemical analysis, samples were fixed in 10% buffered formalin and embedded in paraffin. Tissues showed no macroscopic evidence of necrosis or inflammation. For RNA isolation, freshly resected adenocarcinoma (11 cases), squamous carcinoma (16 cases) and normal tissue (either parenchyma or cells scraped from the lumen of bronchi or both) were stored in RNA-later (Qiagen) at −80°C.

Immunohistochemistry

Sections (5 μm) adjacent to those used for histopathologic examination were processed for immunohistochemistry according to the manufacturer's recommendations with the following antibodies: CLDN-1 (polyclonal JAD.8, Invitrogen); CLDN-4 (clone 3E2C1, Invitrogen); CLDN-5 (clone 4C3C2, Invitrogen); ZO-1 (clone 1, Transduction Laboratories); cingulin (CGN) (clone 3G1H5, Zymed/Invitrogen). The immune reactions were carried out with the steam heat-induced epitope retrieval system, and revealed with the immunoperoxidase technique, using the Envision Kit (DakoCytomation AG, Denmark). The percentage of immunostained carcinomatous cells within the tumor was scored. For statistical purposes, cases were considered as negative when <25% of the cells were stained.

Real-Time qRT-PCR

To prepare RNA, tissues were homogenized on ice in RLT buffer (Qiagen) using a Polytron-Aggregate homogenizer, and RNA was isolated using the RNeasy mini kit (Qiagen). cDNA was synthesized from 1.5 μg total RNA, using the iScript cDNA synthesis kit (Biorad).

To quantitate transcript levels, SYBR-Green-based qRT-PCR (Applied Biosystems) and a 96-channel iCycler optical system (BioRad) were used. Triplicate samples contained 2.5 μM of gene-specific primer sets, 20 ng of cDNA, 10 nM fluorescein dye (BioRad) and 50% volume of PCR master mix. Primer sets (Table 1 Supplementary Material) for JAM-1, OCLN, CLDN-1 to -7, cingulin, GEF-H1, PAR-3, ZO-1, ZO-2 and ZO-3 were designed using Primer Express software. Primer specificity was checked by analyzing PCR products on 4% agarose gels, and by analyzing melting curves obtained by heating from 60 to 95°C, at a rate of 0.5°C per 10 s. Cycling conditions were: one cycle at 95°C for 10 min, 40 cycles of amplification (each 95°C for 15 s and 60°C for 45 s). To analyze the fluorescence signal, a threshold cycle (Ct) was determined, using the exponential growth phase and the baseline signal from fluorescence vs cycle number plots. Cts were normalized to the level of expression of housekeeping genes. Following data analysis with GeNorm v3.2 (http://medgen.ugent.be/~jvdesomp/genorm) software, the genes, succinate dehydrogenase complex, subunit A, flavoprotein (Fp) (SDHA, NM_004168), ubiquitin (UBC, M26880) and tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, zeta polypeptide (YWHAZ, NM_003406), were found to be the most stable internal control genes, and were chosen for the normalization of experimental values, using the geometric averaging method.26

Table 1 Expression of claudin-1, -4, -5, ZO-1 and cingulin in normal and neoplastic human lung tissues
Full size table

Statistical Analysis

Associations between clinicopathologic parameters and CLDN-1 and CLDN-5 expression were analyzed by the χ2 test. P<0.05 was considered statistically significant.

The statistical significance of the difference in transcript levels between cancerous and control tissues was assessed with the two-tailed one-sample t-test, using Prism software (GraphPad Software Inc.).

Results

Immunolocalization of CLDN-1, CLDN-4, CLDN-5, ZO-1 and CGN in Lung Adenocarcinomas and Squamous Cell Carcinomas

Sections of normal bronchial tissue, lung parenchyma, squamous cell carcinoma and adenocarcinoma were labeled with antibodies against three members of the CLDN family of transmembrane tight junction proteins (CLDN-1, CLDN-4, CLDN-5), and two cytoplasmic tight junction proteins (ZO-1 and CGN) (Figure 1 and Table 1). Intense membrane labeling for CLDN-1 was detected in the basal cells of normal bronchial epithelium (double arrows in Figure 1a), whereas cylindrical cells displayed weak or no labeling along their lateral borders (arrowhead in Figure 1a), suggesting that differentiation of basal cells into differentiated cylindrical cells is associated with a downregulation of CLDN-1 expression. Cells undergoing squamous cell dysplasia showed strong CLDN-1 immunolabeling (arrow in Figure 1a), whereas CLDN-5 and CGN labeling were undetectable in these cells (not shown). Conversely, CLDN-1 labeling was not detected in pneumocytes of normal parenchyma (arrowhead in Figure 1b). No CLDN-1 labeling was detected in adenocarcinomas (arrowhead in Figure 1c). Conversely, squamous cell carcinomas showed intense CLDN-1 immunolabeling, which was distributed all around the cell peripheries (arrow in Figure 1d). In summary, strong CLDN-1 immunolabeling was detected in normal basal bronchial epithelial cells and in squamous cell carcinomas, whereas no staining was detected in pneumocytes and adenocarcinoma tissue, as well as fibroblasts, endothelial cells and lymphocytes of peripheral parenchyma (Table 1).

Figure 1
figure 1

Immunolocalization of CLDN-1 (a–d), CLDN-4 (e–h), CLDN-5 (i–l), ZO-1 (m–p) and CGN (q–t) in normal human bronchial epithelium (a, e, i, m, q), normal parenchyma (b, f, j, n, r), adenocarcinoma (c, g, k, o, s) and squamous cell carcinoma (d, h, l, p, t). Arrows indicate positive staining, and arrowheads indicate negative staining. In the case of normal parenchyma, either pneumocytes (b, f, r) or blood vessels (j, n) are shown. Sections were observed at original magnifications between × 10 and × 40. Note the lack of staining of adenocarcinoma by CLDN-1 (c), and of squamous cell carcinoma by CLDN-5 (l) and CGN (t).

Full size image

CLDN-4 labeling was detected in the basal cells of the bronchial epithelium, along the lateral membrane of cylindrical cells (arrows in Figure 1e), and in pneumocytes (arrow in Figure 1f). Strong staining was detected both in adenocarcinomas (arrow in Figure 1g) and squamous cell carcinomas (arrow in Figure 1f). Endothelial cells, lymphocytes and fibroblasts were negative (Table 1).

The expression and localization of CLDN-5 were clearly distinct from those of CLDN-1 and CLDN-4. In normal bronchial epithelium, cylindrical cells showed strong labeling along lateral borders (arrow in Figure 1i), whereas basal cells were negative (arrowhead in Figure 1i). Endothelial cells (arrow in Figure 1j) and pneumocytes of the peripheral parenchyma were positive (Table 1). Strong membrane labeling of CLDN-5 was detected in adenocarcinomas (arrow in Figure 1k). In contrast, squamous cell carcinomas showed no CLDN-5 labeling (arrowhead in Figure 1l), whereas neighboring blood vessels were positive (arrow in Figure 1l). In summary, CLDN-5 was detected in cylindrical cells and adenocarcinomas, but not in basal cells and squamous cell carcinomas (Table 1).

In normal bronchial epithelium, the immunolocalization of the cytoplasmic tight junction proteins ZO-1 (Figure 1m) and CGN (Figure 1q) was similar, with the accumulation of labeling in a honeycomb-like pattern at the apicolateral border of epithelial cells, facing the lumen, corresponding to tight junctions (arrows in Figure 1m and q). Basal cells were negative for CGN and ZO-1 (arrowheads in Figure 1m and q). Endothelial cells of blood vessels within the parenchyma were positive for ZO-1 (arrow in Figure 1n), and negative for CGN (Table 1), whereas pneumocytes were positive for CGN (arrow in Figure 1r) and ZO-1 (Table 1). In adenocarcinomas, ZO-1 and CGN immunolabeling was detected in the glandular component, and was distributed in the region of the cells facing residual true or virtual lumens (arrows in Figure 1o and s). Squamous cell carcinomas showed diffuse ZO-1 staining, which was more intense at the border between tumor tissue and stroma (arrow in Figure 1p). In contrast, CGN immunolabeling was reduced in squamous cell carcinoma (arrowhead in Figure 1t), but occasionally detected in poorly differentiated cases in superficial neoplastic cells facing a virtual lumen. In summary, both ZO-1 and CGN were detected in cylindrical cells, pneumocytes and adenocarcinomas, and ZO-1 was also detected in endothelial cells and squamous cell carcinomas (Table 1).

The expression of CLDN-1 and CLDN-5 was correlated to the clinicopathological characteristics of the lung carcinomas. A statistically significant correlation was found between CLDN-1 expression and squamous cell carcinoma (Table 2), and between CLDN-5 expression and adenocarcinomas (Table 3). However, there was no significant correlation between the staining for these markers and parameters such as pathologic T stage, N stage and histologic differentiation (Table 2 and Table 3).

Table 2 Relationship between claudin-1 protein expression and clinicopathological characteristics of lung carcinomas
Full size table
Table 3 Relationship between claudin-5 protein expression and clinicopathological characteristics of lung carcinomas
Full size table

Changes in Transcript Levels for Tight Junction Proteins in Lung Squamous Cell Carcinomas and Adenocarcinomas

To confirm and extend the observations on the altered patterns of tight junction protein expression in lung squamous cell carcinomas and adenocarcinomas, transcript levels for 14 different tight junction-associated proteins, including the ones analyzed by immunocytochemistry, were measured by qRT-PCR (Table 4). Transcript levels were expressed as percentages, taking as 100%, the transcript levels of control normal tissue obtained from the same patient. Either parenchyma or scraped cells from normal bronchi, or both were used as a control for each sample of tumor tissue (Table 4).

Table 4 Quantitative analysis of transcript levels for tight junction proteins by qRT-PCR
Full size table

Squamous cell carcinomas showed statistically significant decreases in mRNA levels (vs bronchial cells) for JAM-1 (which was decreased by ∼45%) OCLN (∼67%), CLDN-3 (∼95%), CLDN-4 (∼65%), CLDN-7 (∼76%), CGN (∼69%), ZO-2 (∼71%) and ZO-3 (∼97%). When parenchyma was used as a control, squamous cell carcinomas showed a significant decrease in mRNA levels for OCLN (∼83%), CLDN-5 (∼84%), CGN (∼76%), GEF-H1 (∼69%), ZO-1 (∼76%), ZO-2 (∼70%) and ZO-3 (∼60%). The only proteins for which a statistically significant increase in mRNA levels was found in squamous cell carcinoma (vs parenchyma) were CLDN-1 (∼7-fold increase) and CLDN-5 (∼8-fold increase) (vs bronchial cells). Although the increase in CLDN-1 transcript levels was correlated to an increased protein detection by immunocytochemistry, no CLDN-5 was detected in squamous cell carcinoma by antibody staining (Figure 1). Increases were also observed in the transcript levels of CLDN-1, CLDN-2 and GEF-H1, but they were not statistically significant in our sample pool (Table 4).

For adenocarcinomas, the relative levels of mRNA of tight junction proteins were somewhat different (except for JAM-1, OCLN, CGN, PAR-3 and ZO-2), depending on whether parenchyma or scraped bronchial cells were used as a control. As the results with bronchial cells were in better agreement with the results obtained by immunocytochemistry, bronchial cells may represent the most appropriate normal control tissue. Notably, when using bronchial cells as a control, we observed a statistically significant decrease in the transcript levels for CLDN-1 (∼58%), CLDN-3 (∼71%), CLDN-4 (∼47%), CLDN-7 (∼57%), ZO-2 (∼72%) and ZO-3 (∼82%). A statistically significant increase was observed for GEF-H1 mRNA (∼2-fold), whereas expression of CGN and ZO-1 was similar between adenocarcinoma and bronchial cells (Table 4). In agreement with the immunocytochemistry results, CLDN-5 mRNA levels were increased (∼33-fold) compared with bronchial cells, but, surprisingly, were decreased (∼87%) compared with parenchyma. This latter discrepancy could be due to the abundance of CLDN-5 transcript in highly vascularized parenchyma. The increase in CLDN-4 mRNA (∼2.6-fold) compared with parenchyma was correlated with the immunocytochemistry observations, whereas the increase in CLDN-1 mRNA was not, suggesting that CLDN-1 mRNA translation may be repressed in adenocarcinomas.

Discussion

This study provides the first analysis of the expression of a large panel of tight junction proteins in human lung adenocarcinomas and squamous cell carcinomas, using both immunocytochemistry and qRT-PCR. We demonstrate that CLDN-1, CLDN-5 and CGN are differently expressed in these two tumors, and their localization in normal bronchial tissue suggests hypotheses on the histogenesis of these tumors. Furthermore, we show that expression of several other tight junction proteins is altered in human lung tumors.

CLDN-1 is expressed in a large number of tissues, and plays a crucial role in the barrier function of skin.27 Here we showed that CLDN-1 expression is high in lung squamous cell carcinomas (both by protein detection and transcript quantification), but reduced or undetected in adenocarcinomas. Interestingly, an increase in CLDN-1 expression was observed in keratinized regions of squamous cell carcinoma of the epidermis,14 and in colon cancers,28 whereas a decrease or loss of expression was reported in breast cancer cell lines29, 30 and tumors.31 These findings suggest that a squamous cell tumor phenotype is more likely to be correlated with high CLDN-1 expression, whereas tumors with glandular components are more likely to lose CLDN-1 expression. In addition, our observation that in normal bronchial epithelium CLDN-1 is abundant in basal cells, but scarce or absent in cylindrical cells and pneumocytes raises the hypothesis that squamous cell carcinomas arise either from uncontrolled proliferation and lack of normal differentiation of the basal cells of the bronchial epithelium or from the reversion of differentiated cells to this phenotype. As CLDN-1 has been implicated in epithelial-mesenchymal transition, possibly through the modulation of the β-catenin/Tcf signaling pathway,28 it is possible that CLDN-1 dysregulation may be part of the molecular pathway that leads to the development of squamous cell carcinomas.

The pattern of CLDN-5 expression in normal bronchial epithelium and in tumors was essentially opposite to that of CLDN-1, with high expression in normal cylindrical cells, pneumocytes and adenocarcinomas, and low or undetectable expression in basal cells and squamous cell carcinomas. These results indicate not only that CLDN-5 could be used as an additional diagnostic tool but also that the differentiation pathway of adenocarcinomas is distinct of that of squamous cell carcinomas. Little information is available on CLDN-5 expression in human cancers, apart from a reported decrease in hepatocellular and renal carcinomas.32 Previous studies indicated that CLDN-5 is specifically expressed in endothelial cells.33, 34 We confirmed CLDN-5 expression in blood vessels, and we also demonstrated its expression in bronchial cells and pneumocytes. The high level of CLDN-5 expression in endothelia may explain why mRNA levels of CLDN-5 were decreased in adenocarcinomas (which contain mostly epithelial cells), when compared with lung parenchyma (where abundant blood vessels are present) (Table 4). On the other hand, CLDN-5 mRNA levels were high in adenocarcinomas, when compared with bronchial cells (Table 4). In squamous cell carcinomas, CLDN-5 mRNA appeared to be abundant (only when compared with scraped bronchial cells), despite the lack of CLDN-5 immunoreactivity in these tumors. This discrepancy could be explained by a downregulation of CLDN-5 expression at the translational level. As different results were obtained in some cases depending on whether parenchyma or scraped bronchial cells were used as a control tissue, results using qRT-PCR should be interpreted with caution. A careful choice of the control normal tissue, as well as the use of multiple internal control housekeeping genes is necessary. Results should be confirmed by immunocytochemistry, which provides essential information about the cellular and subcellular distribution of proteins. In this regard, it is noteworthy that in lung normal and tumor epithelial cells CLDN-1, -4 and -5 were not restricted to apicolateral junctions, as CGN and ZO-1, but were detected throughout the membrane surface, raising the hypothesis that CLDNs may play additional roles, besides their canonical role in the tight junction barrier.

qRT-PCR indicated a decreased expression of CLDN-3 and CLDN-7 both in adenocarcinomas and squamous cell carcinomas, when compared with bronchial cells. CLDN-4 mRNA was decreased in squamous cell carcinomas, but the decrease in mRNA (Table 4) did not correlate with a decreased protein expression (Figure 1). This could be due to an increased mRNA or protein stability. CLDN-3 and CLDN-4 are upregulated in ovarian and uterine cancers,10, 17, 35, 36 and CLDN-4 expression has been exploited for therapeutic purposes.24 CLDN-7 expression is decreased in breast carcinomas,13, 37 but upregulated in gastric carcinomas.9 With regards to other tight junction membrane proteins, we observed a statistically significant decrease in JAM-1 transcript levels in squamous cell carcinoma, when compared with normal bronchial cells, and in OCLN mRNA in squamous cell carcinoma. This latter result is in agreement with one previous immunohistochemistry study of OCLN expression in lung carcinomas.23

When analyzing the expression of cytoplasmic tight junction proteins, we observed significant decreases in the expression of ZO-2 and ZO-3 in both squamous cell carcinomas and adenocarcinomas, when compared with bronchial cells, and ZO-2 also when compared with parenchyma. However, the expression of the related protein ZO-1 was maintained. CGN was expressed in lung adenocarcinomas, but was significantly reduced in squamous cell carcinomas, and present only in superficial neoplastic cells facing a virtual lumen. These observations, together with our previous report on high CGN expression in colon adenocarcinomas,20 indicate that CGN could be used as an additional diagnostic tool to distinguish adenocarcinomas from other types of epithelial cancers, and to identify glandular differentiation in poorly differentiated carcinomas. The mRNA levels for the RhoA activator GEF-H1 were decreased in squamous cell carcinomas, when compared with parenchyma, and increased in adenocarcinomas, when compared with bronchial cells. Interestingly, GEF-H1 upregulation is induced by mutant p53, and leads to increased proliferation.38

Taken together, our results show that the composition of the tight junction multiprotein complex is considerably altered following cancer development in lungs. In a few instances, dramatic protein upregulation was observed (most notably, CLDN-1 in squamous cell carcinomas and CLDN-5 in adenocarcinomas), and, more often, a decrease in expression was detected. Although it is unclear how these changes affect the structure and function of tight junctions in carcinomas, it is reasonable to hypothesize that a downregulation of the adhesive function of tight junctions may lead to increased invasion and metastasis. In addition, downregulation of some tight junction proteins is known to increase epithelial proliferation in invertebrate39 and vertebrate model systems.40, 41, 42 The disruption of the tight junction multiprotein complex may also affect the establishment and maintenance of the polarized phenotype. Although it remains unclear if and how the modified expression of any given specific tight junction protein is associated with cancer pathogenesis, our study suggests that it may be useful to understand cancer histogenesis. Further work should be carried out to understand how molecular abnormalities in the expression of tight junction proteins could contribute to tumorigenesis, and to expand their use as tools for cancer diagnosis, prognosis and treatment.