Background Information. Carcinoma of the oesophagus is the sixth leading cause of cancer death in the western world and is associated with a 5-year survival of less than 15%. Recent evidence suggests that stromal—epithelial interactions are fundamental in carcinogenesis. The advent of co-culture techniques permits the investigation of cross-talk between the stroma and epithelium in a physiological setting. We have characterized a histologically representative oesophageal organotypic model and have used it to compare the most commonly used squamous oesophageal cell line, HET-1A, with primary oesophageal squamous cells for use in studies of the oesophageal epithelium in vitro.
Results. When grown in an organotypic culture with normal fibroblasts, the oesophageal carcinoma cell lines OE21 (squamous) and OE19 (adenocarcinoma) morphologically resembled the tumour of origin with evidence of stromal invasion and mucus production, respectively. However, HET-1A cells, which were derived from normal squamous oesophageal cells, appeared dysplastic and failed to display evidence of squamous differentiation. By comparison with primary oesophageal epithelial cells, the HET-1A cells were highly proliferative and did not express the epithelial markers E-cadherin or CK5/6 (casein kinase 5/6), or the stratified epithelial marker ΔNp63, but did express the mesenchymal markers vimentin and N-cadherin.
Conclusion. Studies of epithelial carcinogenesis will benefit from culture systems which allow manipulation of the stromal and epithelial layers independently. We have developed an organotypic culture using primary oesophageal squamous cells and fibroblasts in which a stratified epithelium with a proliferative basal layer that stains strongly for ΔNp63 develops. This model will be suitable for the study of the molecular events in the development of Barrett's oesophagus. The most commonly used normal oesophageal squamous cell line, HET-1A, does not have the characteristics of normal oesophageal squamous cells and should not be used in models of the normal oesophageal epithelium. Until more representative cell lines are available, future studies in oesophageal cancer will be reliant on the availability and manipulation of primary tissue.
Carcinoma of the oesophagus is the sixth leading cause of cancer death in the western world and is associated with a 5-year survival of less than 15% (Koppert et al., 2005). The two most common oesophageal cancer types are ESCC (oesophageal squamous cell carcinoma) and EAC (oesophageal adenocarcinoma). Squamous cell carcinoma is the more common type of esophageal cancer worldwide, but the incidence of EAC is rising rapidly in Northern and Western Europe, which may reflect a population-wide increased exposure of the lower oesophagus to acid and bile related to obesity (Bosetti et al., 2008). Adenocarcinoma of the oesophagus is associated with the finding of a specialized intestinal-type columnar metaplasia in the distal oesophagus—BE (Barrett's oesophagus) (Fitzgerald, 2005; Flejou, 2005; Koppert et al., 2005; Gutierrez-Gonzalez and Wright, 2008). BE is thought to represent a protective response to chronic acid and bile injury. Patients with BE are at a 30–125-fold greater risk of developing EAC compared with the general population (Koppert et al., 2005). Given the high incidence of BE (1–1.5%) and the increasing incidence of EAC, novel biomarkers are urgently required to predict disease progression and to target therapy (Lord, 2003). Novel endoscopic therapies [including RF (radio frequency) ablation] may offer the potential for BE eradication, but, in a significant minority (∼20%), BE persists despite repeated endoscopic treatment (Rees et al., 2010; Shaheen et al., 2009). This suggests that a hitherto unknown mechanism is promoting the maintenance of BE and that this is likely to involve the epithelial microenvironment. The contribution of stromal signals to the epithelial differentiation of the distal oesophagus is not known and may be fundamental, not only to those patients in whom BE is not eradicated after RF ablation, but also to the initiation and maintenance of BE (Fitzgerald, 2006). Stromal gene expression signatures can differentiate between pre-invasive and invasive disease in gastro-intestinal cancers and can discriminate between normal oesophageal, BE and EAC tissues (Saadi et al., 2010). However, the stromal gene signature may be undetectable when whole tissue samples are analysed. This demonstrates that gene changes in the stroma can be independent of changes in the epithelium and are important in the disease process. Furthermore, it highlights the requirement for the use of experimental methods that allow manipulation of the stromal and epithelial layers independently in a physiologically representative setting.
The majority of studies to examine the molecular causes of BE have used either gene expression data from endoscopic biopsies or conventional tissue culture methods based on immortalized cell lines, such as the SV40T-antigen (simian-virus-40 large T-antigen)-immortalized oesophageal squamous cell line HET-1A, which is commonly used as a model of normal oesophageal squamous epithelium. The application of co-culture techniques permits a more physiological study of the behaviour of epithelial cells in vitro, allowing interactions between epithelial cells and their adjoining stroma and stromal cells to be studied. Organotypic culture has allowed investigation of the regulatory mechanisms controlling cell differentiation, tissue homoeostasis and tissue integrity (Chioni and Grose, 2008). Organotypic models have been used to investigate stromal—epithelial interactions in a variety of tumours, including skin, breast, prostate and ovary. Recent reports suggest that the generation of representative oesophageal organotypic models is feasible and furthermore likely to shed new light on the relationship between the stroma and epithelium in the development of EAC (Okawa et al., 2007; Green et al., 2010). A side-by-side analysis of the histology of oesophageal composites using four different scaffolds suggests that a matrix incorporating collagen-I is the optimum platform to study the interaction between squamous cells and fibroblasts (Green et al., 2010). We report the characterization of a histologically representative oesophageal organotypic model using a method documented to be a robust, accurate and reproducible technique for measuring tumour cell invasion in head and neck cancer (Nystrom et al., 2005). Furthermore, we show that primary oesophageal squamous cells form a stratified epithelium in organotypic culture. Importantly, we show that HET-1A cells fail to develop a stratified squamous epithelium in organotypic culture. We show that there are key molecular differences between HET-1A and primary oesophageal squamous cells, reinforcing the unsuitability of this cell line for use in in-vitro models of the normal oesophageal epithelium.
Generation of oesophageal organotypic cultures
Initial experiments focused on generating in-vitro models of the oesophagus in health and disease. The cell lines derived from normal oesophageal squamous epithelium (HET-1A), oesophageal squamous cell carcinoma (OE21) and oesophageal junctional adenocarcinoma (OE19) were each matured for 1 week in organotypic cultures containing the human fetal fibroblast cell line HFFF2 (Figures 1A–1C respectively). The two carcinoma cell lines showed characteristics expected based on their in-vivo growth. The OE21 cells showed disordered stratification with elongation and flattening of cells in the most superficial layers, in keeping with their squamous origin (Figure 1B). Invasion into the matrix was also clearly apparent. The OE19 cells grew as a neoplastic columnar-type epithelium with multiple mitotic figures and mucus production (Figure 1D). The HET-1A cells, in contrast, developed into a dysplastic epithelium with no evidence of stratification (Figure 1A), unlike the normal human oesophageal squamous epithelium (Figure 1E).
Comparison of primary oesophageal cells with HET-1A in organotypic culture
Because we found that the HET-1A cell line did not exhibit squamous morphology in organotypic culture, we compared HET-1A with primary oesophageal squamous cells prepared from oesophageal resections. We found subtle morphological differences between the cell lines when they were grown in a culture flask. The HET-1A cells grew as thin, elongated cells with cellular projections (Figure 2A), different from the flattened and typically epithelial appearance of primary oesophageal keratinocytes (NOK210, a representative example of cells taken from three different oesophageal resections) (Figure 2B). When grown for 2 weeks in organotypic culture containing primary oesophageal fibroblasts (NOF2310), NOK210 cells developed a characteristic squamous morphology, with a basal cell layer (Figure 2D, arrow) and stratification; the uppermost cell layer appeared flattened with elongated nuclei. In comparison, the HET-1A cells displayed a dysplastic phenotype with a disordered epithelial layer, hyperchromatic nuclei, vacuole formation (Figure 2C, arrow) and no evidence of squamous differentiation. To determine whether or not the abnormal morphology of HET-1A cells in organotypic culture is inherent to the cells themselves or influenced by the cells in the matrix, we examined the morphology of HET-1A cells grown in the presence of the previously characterized normal primary oesophageal fibroblast isolate (F1) (Figure 2E) (Saadi et al., 2010) and again found no evidence of squamous differentiation. Molecular analysis was therefore undertaken in organotypic models containing NOF cells. The HET-1A cells were highly proliferative, as demonstrated by increased immunohistochemical staining for Ki67 throughout the epithelial layer (Figure 3A). This contrasts strikingly with the occasional proliferative cell restricted to the basal cell compartment in the primary oesophageal epithelial layer (Figure 3B). Despite high levels of proliferation in HET-1A, the thickness of the epithelium is not vastly different from that of the NOK210 cultures. This may be due to a concomitant increase in apoptosis in HET-1A cells, as demonstrated by the presence of cleaved caspase 3 in multiple nuclei (Figure 3C). Furthermore, HET-1A cells failed to express the squamous epithelial cytokeratins CK5/6 (Figure 3D), unlike the primary oesophageal keratinocytes (Figure 3E). The integrity of stratified epithelium is maintained in part by their intercellular connections, characterized by the presence of E-cadherin (Zhu and Watt, 1996; Hines et al., 1999; Vaughan et al., 2009). In organotypic culture, the NOK210 cells stained for E-cadherin, but the HET-1A cells did not (Figures 3F and 3G). To confirm this finding, the expression of E-cadherin mRNA was then measured by RT—PCR (reverse transcription—PCR) in the HET-1A cells and five oesophageal squamous biopsies. The results in Figure 3(H) show E-cadherin expression in the normal biopsies but not in HET-1A. In contrast, HET-1A expressed the mesenchymal markers vimentin and N-cadherin, whereas the normal squamous biopsies did not (results not shown).
Lack of expression of p63 in HET-1A
The demonstration that HET-1A cells in organotypic culture did not display an epithelial morphology or molecular phenotype led us to compare the expression of p63 isoforms in HET-1A cells and NOK. Whereas NOK210 expressed ΔNp63 isoforms, at the mRNA and protein level this was not observed in HET-1A (Figures 4A and 4B). Expression of ΔNp63 was strong in the basal cell layer of NOK210 organotypic cultures, but absent in comparable HET-1A organotypic cultures (Figure 4C). In a representative biopsy of normal human oesophagus, p63 nuclear expression was clearly demonstrated, with the strongest staining located in the basal layers (Figure 4D, left hand panel). Under higher magnification (Figure 4D, right-hand panel), p63 localization in the superficial mucosal layers is comparable with ΔNp63 in NOK210 organptypic cultures (Figure 4C).
Recent evidence suggests that stromal—epithelial interactions are fundamental to the establishment and progression of epithelial cancers. Because of this, robust and representative models of the normal oesophagus are required to investigate the stromal and epithelial factors that are important in these cancers, such as the establishment of BE and its transformation to EAC. Additionally, models of advanced disease will facilitate the study of invasion and metastasis, the hallmarks of a malignant tumour. The use of primary cells in organotypic culture presents practical problems in terms of cell availability, cell senescence and genetic manipulation. For this reason, established cell lines are generally favoured in the expectation that they will ensure a consistent, reproducible, robust and representative model of the native oesophageal epithelium. In studies of oesophageal cancer, the SV40T antigen-immortalized oesophageal squamous cell line HET-1A is commonly used as a model normal squamous epithelial cell. We show that this cell line, grown in organotypic culture, results in a dysplastic epithelial layer, unlike primary squamous epithelial cells. Further characterization of HET-1A shows that, unlike the primary cells, it is highly proliferative and expresses mesenchymal and not oesophageal epithelial markers.
Our data are consistent with a recent report that HET-1A cells grown on a porcine oesophageal matrix results in a hyper-proliferative, undifferentiated epithelium which does not express the squamous markers CK4, CK14 and involucrin (Green et al., 2010). Whereas Green et al. (2010) focused on developing the most representative oesophageal model by comparing a range of stromal scaffolds, we have attempted to further understand the difference in morphology between primary tissue and HET-1A and suggest ΔNp63 expression as a possible determinant of squamous differentiation. We have used a method to generate organotypic cultures that is significantly different from that used by Green et al. (2010). Furthermore, using our protocol, we found that HET-1A cells fail to express the squamous epithelial markers CK5/6 and fail to express a key determinant of epithelial integrity, E-cadherin, but do express the mesenchymal markers N-cadherin and vimentin.
The p53 protein family member p63 is vital for squamous epithelial development and maintenance (Barbieri and Pietenpol, 2006). A reduction in p63 expression as a result of chronic inflammation in the pre-malignant condition of the oral mucosa, OLP (oral lichen planus), is associated with decreased E-cadherin expression (Ebrahimi et al., 2008). The expression of involucrin, cytokeratins and cadherins can be regulated by ΔNp63 isoforms in other squamous tissues (De Laurenzi et al., 2000; Barbieri et al., 2006; Ebrahimi et al., 2008; Fukushima et al., 2009). An appropriate expression of p63 is required for squamous mucosal development and integrity, including in the oesophagus. In p63−/− mice, there is a striking loss of all stratified squamous epithelia (Mills et al., 1999). Furthermore, ectopic expression of p63 in single-layered lung epithelium results in a shift to a stratified squamous epithelium (Koster et al., 2004). Numerous studies have shown that at the protein level ΔNp63 is the predominant, if not the only, isoform of p63 expressed in epithelial cells (reviewed in Barbieri and Pietenpol, 2006). Moreover, p63 may define the difference between epithelial cells and stromal cells at the interface between these populations (Barbieri and Pietenpol, 2006). In keeping with this hypothesis, we have shown that ΔNp63 is absent from HET-1A, but intense staining is observed in the basal cell population of NOK cultures adjacent to the matrix containing fibroblasts, in concordance with p63 expression in human biopsies. Exposure of oesophageal squamous cells to acid and bile salts, as occurs in GORD (gastro-oesophageal reflux disease), has been shown to down-regulate p63 expression in vitro (Roman et al., 2007) and this correlates with the finding of reduced p63 expression in the chronic inflammatory condition OLP (Ebrahimi et al., 2008). It is possible that repression of p63 expression is required to allow the transition from a squamous to columnar phenotype in the distal oesophagus in response to chronic inflammation caused by GORD.
Our findings have implications for the study of the origin of the columnar epithelial cells in BE. Any model that aims to investigate the initiation of BE will require epithelial cells that retain determinants of squamous differentiation and the ability to respond to stromal signals. We have demonstrated that HET-1A cells lack p63 expression, confirming their unsuitability as a model of normal oesophageal epithelium.
The majority of recent studies of Barrett's transformation have concentrated on the gain of determinants of columnar differentiation (e.g. CDX1 and CDX2). The data presented here suggest that, along with the acquisition of intestinal markers, there may be a key role for the loss of determinants of a stratified squamous epithelium in Barrett's initiation and maintenance (e.g. p63). Therefore, experiments describing the genetic manipulation of HET-1A to induce a columnar phenotype should be viewed with caution, and further work will be required to determine the regulatory events in the oesophageal epithelium during Barrett's transformation. Organotypic models that incorporate morphologically and biochemically representative epithelial cells will be required for such studies, and our results suggest the use of primary tissue wherever possible.
Materials and methods
Primary cells and cell lines
Primary oesophageal keratinocytes (NOK210) and fibroblasts (NOF2310) were established from normal human oesophagus. The study was approved by the Southampton and South West Hampshire Research Ethics Committee (Oesophagus: molecular, cellular and immunological assessment. LREC number: 09/H0504/66. Principal Investigator: Mr T.J. Underwood).
Normal oesophageal tissue was harvested from an oesophagectomy specimen and transported in Hank's Balanced Salt Solution (Invitrogen, Carlsbad, CA, U.S.A.). The tissue was washed twice in PBS and left in PBS supplemented with 250 ng/ml amphotericin B (Invitrogen) until further treatment. For keratinocyte extraction, the tissue was incubated with 3 units/ml Dispase type II (Roche Molecular Biochemicals, Mannheim, Germany) at 4°C overnight. The epithelium was dissected and incubated with 0.05% trypsin/EDTA (Invitrogen) for 30 min. The undigested tissue remnant was removed and BEGM (bronchial epithelial growth medium; Lonza) supplemented with 250 ng/ml amphotericin B was added to the cell suspension which was then centrifuged. The pellet was resuspended in the same medium plus 10−10 M cholera toxin (Sigma—Aldrich, St. Louis, MO, U.S.A.) and cells were cultured at 37°C in a humidified atmosphere with 95% O2 and 5% CO2. For the fibroblast isolation, epithelial tissue was cut into 2 mm3 pieces and a single piece was placed in the well of a six-well plate. The explants were cultured in DMEM (Dulbecco's modified Eagle's medium; Invitrogen) supplemented with 10% FBS (foetal bovine serum; Autogen Bioclear), 100 units/ml penicillin, 100 μg/ml streptomycin, 250 ng/ml amphotericin B and 292 μg/ml L-glutamine (Invitrogen) at 37°C in a humidified atmosphere with 10% CO2.
The HET-1A and the OE21 cell line were obtained from the A.T.C.C. (Manassas, VA, U.S.A.). F1 primary fibroblasts were a gift from Dr Rebecca Fitzgerald (Hutchinson/MRC Research Centre, Cambridge, U.K.). HET-1A is a non-tumourigenic SV40T-transformed squamous oesophageal cell line (Stoner et al., 1991). The HET-1A cell line was cultured in BEGM at 37°C and 5% CO2. OE21 and OE19 cells were cultured in RPMI 1640 supplemented with 10% FBS (Autogen Bioclear), 100 units/ml penicillin, 100 μg/ml streptomycin, 250 ng/ml amphotericin B and 292 μg/ml L-glutamine (Invitrogen) at 37°C in a humidified atmosphere with 5% CO2.
Cells were examined at room temperature (20°C) in BEGM and photographed using an Olympus CC12 digital camera and analysed with Cell 2.6 Imaging software (Olympus UK Ltd).
Organotypic cultures were prepared as previously described (Nystrom et al., 2005). The organotypic cultures were cultured for 14 days, with a change of medium every 2 days (BEGM for NOK210 or RPMI 1640/10% FCS for OE21 and OE19). The gels were then bisected and fixed in formalin overnight, followed by 24 h in 70% ethanol.
Histology and immunohistochemistry
Gels were embedded in paraffin wax and 4 μm-thick sections were stained with H&E (haematoxylin and eosin) or Alcian Blue/PAS. Sections (4 μm) were also immunostained with antibodies to CK5/6, Ki67 (Dako), E-cadherin and p63 (Leica Microsystems), cleaved caspase 3 (Asp175) (Cell Signaling Technology) or ΔNp63 (Nylander et al., 2002). The pretreatment regimens and antibody retrieval were performed according to the manufacturer's specifications, or as previously described for ΔNp63 (Nylander et al., 2002). The dilutions used were: 1:100 for CK5/6, 1:150 for Ki67, 1:50 for E-cadherin, 1:25 for p63, 1:200 for cleaved caspase 3 and 1:20000 for ΔNp63. Sections were viewed under a Nikon Eclipse E600 (Nikon) microscope with Nikon Plan objective lenses ×20/0.5, ×10/0.3, ×40/0.75 and ×100/1.3 and recorded digitally on a Nikon Coolpix 4500 Camera (Nikon).
The cells were washed with PBS, scraped into PBS, and pelleted by centrifugation. They were then lysed by incubation for 15 min at 4°C in 2 vol. of urea lysis buffer (7 M urea, 0.05% Triton X-100, 25 mM NaCl, 20 mM Hepes, pH 7.6 and 100 mM dithiothreitol). Lysates were clarified by centrifugation at 13000 g for 10 min. Protein was quantified using a Bio-Rad protein assay reagent. Proteins were resolved by SDS/PAGE and transferred to Hybond-ECL membranes (GE Healthcare).
A mouse monoclonal anti-p63 clone 4A4, 1:200 (Santa Cruz) and a rabbit polyclonal anti-β-actin clone 20–33, 1:5000 (Sigma) were used for immunoblotting. Blocking and antibody incubations were done in 3% low-fat milk in PBS/0.025% Tween 20, and washes were in PBS/0.1% Tween 20. The horseradish-peroxidase-labelled secondary antibody was detected with Supersignal (Pierce), and gel images were collected and analysed on a Fluor-S MultiImager (Bio-Rad) equipped with Quantity One software (Bio-Rad).
The total RNA was isolated from cells or oesophageal tissue using the RNeasy Mini Kit with on-column DNase treatment using the RNase-Free DNase Set (Qiagen) as per the manufacturer's instructions. First strand cDNA synthesis was conducted using M-MLV (Promega) or SuperScript II (Invitrogen) reverse transcriptase according to manufacturers' protocols. The different isoforms of p63 were amplified using the following primers: ΔNp63: FW (forward)-5′-GGAAAACAATGCCCAGACTC-3′ and RV (reverse)-5′-AGAGAGCATCGAAGGTGGAG-3′ (product: 185 bp; TAp63: FW-5′-TTTGAAACTTCACGGTGTGC-3′ and RV-5′-GCTGGAAAACCTCTGGACTG-3′ (product: 163 bp); p63α: FW-5′-GAGGTTGGGCTGTTCATCAT-3′ and RV-5′-GAGGAGAATTCGTGGAGCTG-3′ (product: 174 bp); p63β: FW-5'-GAAACGTACAGGCAACAGCA-3′ and RV-5′-CAGACTTGCCAGATCCTGA-3′ (product: 371 bp); p63γ: FW-5′-GGGCCGTGAGACTTATGAAA-3′ and RV-5′-ATTCCTGAAGCAGGCTGAAA-3′ (product: 157 bp). E-cadherin was amplified using: FW-5′-CCCACCACGTACAAGGGTC-3′ and RV-5-CTGGGGTATTGGGGGCATC-3′ (product: 93 bp). Expression of β-actin and GAPDH (glyceraldehyde-3-phosphate dehydrogenase) were examined to determine the integrity of the RNA samples. β-Actin: FW-5′-CTCAGGAGGAGCAATGATCTTG-3′ and RV-5′-CTGGGCATGGAGTCCTGTGG-3′ (product: 204 bp); GAPDH: FW-5′-ACCCAGAAGACTGTGGATGG-3′ and RV-5-CAGTGAGCTTCCCGTTCAG-3′ (product: 138 bp). The Go Taq Polymerase (Promega) was used in the RT—PCR.
To determine the relative gene expression of E-cadherin, real-time QPCR (quantitative PCR) was performed in triplicate using the QuantiTect SYBR Green PCR Kit (Qiagen) in a Rotorgene 3000 apparatus (Corbett Research, Sydney, NSW, Australia). Data were collected and analysed using the Rotorgene application software v6 (Corbett Life Science). Relative expression levels were determined using the comparative quantification feature of the software, and normalized to GAPDH. The PCR products were electrophoresed on 1.5% (w/v) agarose gels and stained with ethidium bromide to confirm expected product sizes.
Timothy Underwood, Mathieu Derouet and Jeremy Blaydes were responsible for the concept and design of the research, carrying out the experimental work, data analysis and interpretation and writing of this manuscript. John Primrose assisted with concept and design and the interpretation of data. Michael White performed the experiments on p63 expression. Fergus Noble was responsible for concept development, ethical approval and the collection of human tissue with Timothy Underwood. Karwan Moutasim generated organotypic models with Mathieu Derouet, all of which were reviewed by Gareth Thomas. Eric Smith and Paul Drew performed experiments on E-cadherin, N-cadherin and vimentin expression. All authors were involved in writing the article.
We thank P. J. Coates (Centre for Oncology and Molecular Medicine, Division of Medical Sciences, University of Dundee, Ninewells Hospital and Medical School, Dundee, DD1 9SY, U.K.) for immunohistochemical analysis of organotypic models for p63 expression. We also thank Dr Rebecca Fitzgerald for the F1 primary fibroblast cell line and her continued help and support.