Proteomic Screening of a Cell Line Model of Esophageal Carcinogenesis Identifies Cathepsin D and Aldo-Keto Reductase 1C2 and 1B10 Dysregulation in Barrett’s Esophagus and Esophageal Adenocarcinoma Jean Breton,†,‡,§ Matthew C. Gage,† Alastair W. Hay,† Jeffrey N. Keen,| Christopher P. Wild,† Clare Donnellan,† John B. C. Findlay,| and Laura J. Hardie*,† Molecular Epidemiology Unit, Centre for Epidemiology and Biostatistics, Leeds Institute of Genetics, Health and Therapeutics, Clarendon Way, University of Leeds, Leeds LS2 9JT, United Kingdom, Laboratoire “Lésions des Acides Nucléiques”, Service de Chimie Inorganique et Biologique UMR E3 CEA, Université Joseph Fourier, 17 rue des Martyrs, 38 054 Grenoble cedex 9, France, UFR des Sciences Pharmaceutiques de Grenoble, Université Joseph Fourier, Domaine de la Merci, Bâtiment Jean Roget, 38 706 La Tronche cedex, France, and Institute for Membrane and Systems Biology and Leeds Institute of Genetics, Health and Therapeutics, University of Leeds, Faculty of Biological Sciences, Leeds LS2 9JT, United Kingdom Received November 23, 2007
Esophageal adenocarcinoma (EA) incidence is increasing rapidly and is associated with a poor prognosis. Identifying biomarkers of disease development and progression would be invaluable tools to inform clinical practice. Two-dimensional polyacrylamide gel electrophoresis was used to screen 10 esophageal cell lines representing distinct stages in the development of esophageal cancer. Thirty-three proteins were identified by MALDI-TOF-MS which demonstrated differences in expression across the cell lines. Western blotting and qRT-PCR confirmed increased cathepsin D and aldo-keto reductases 1C2 and 1B10 expression in metaplastic and dysplastic cell lines. Expression of these proteins was further assessed in esophageal epithelium from patients with nonerosive (NERD) and erosive gastro-esophageal reflux disease, Barrett’s esophagus (BE) and EA. When compared with normal epithelium of NERD patients, (i) cathepsin D mRNA levels demonstrated a stepwise increase in expression (p < 0.05) in erosive, metaplastic and EA tissue; (ii) AKR1B10 expression increased (p < 0.05) 3- and 9-fold in erosive and Barrett’s epithelium, respectively; and (iii) AKR1C2 levels increased (p < 0.05) in erosive and Barrett’s epithelium, but were reduced (p < 0.05) in EA. These proteins may contribute to disease development via effects on apoptosis, transport of bile acids and retinoid metabolism and should be considered as candidates for further mechanistic and clinical investigations. Keywords: Barrett’s • esophagus • cathepsin D • AKR1C2 • AKR1B10
Introduction Esophageal cancers can be classified in two main histological categories. The most common worldwide is squamous cell carcinoma (SCC). In western Europe, SCC is mainly attributed to alcohol and tobacco consumption and there is a decreasing incidence in some countries.1 However, the incidence of the second type, esophageal adenocarcinoma (EA), has risen * Corresponding author: Dr. Laura J. Hardie, Molecular Epidemiology Unit, Centre for Epidemiology and Biostatistics, Leeds Institute of Genetics, Health and Therapeutics, University of Leeds, Leeds LS2 9JT, United Kingdom. Tel: +44 (0) 113 343 7769. Fax: +44 (0) 113 343 6602. E-mail:
[email protected]. † Centre for Epidemiology and Biostatistics, Leeds Institute of Genetics, Health and Therapeutics, Clarendon Way, University of Leeds. ‡ Laboratoire “Lésions des Acides Nucléiques”, CEA-Université Joseph Fourier. § UFR des Sciences Pharmaceutiques de Grenoble, Université Joseph Fourier. | Institute for Membrane and Systems Biology and Leeds Institute of Genetics, Health and Therapeutics, University of Leeds. 10.1021/pr7007835 CCC: $40.75
2008 American Chemical Society
dramatically over the last 30 years in Western countries including the U.K.2 EA often arises on a background of Barrett’s esophagus (BE), a metaplastic condition stimulated by gastroesophageal reflux disease.3 This preneoplastic lesion is characterized by the replacement of the normal squamous epithelium of the esophagus by a columnar, intestinal-type which can progress to dysplasia and EA itself. Despite 5 year survival rates which do not exceed 20%,4 the carcinogenic process leading to EA remains poorly described. However, certain molecular changes which arise during the disease process may have value as biomarkers of EA risk in BE surveillance programs, serve as targets for chemoprevention, or predict treatment response and prognostic outcome in EA cases. Transcriptome and proteome wide analyses of esophageal cells and tissues are promising tools to identify such candidates.5 cDNA microarrays permit identification of genes, panels of genes or wider transcriptional profiles6–9 which are involved in the disease process. However, candidates identified through Journal of Proteome Research 2008, 7, 1953–1962 1953 Published on Web 04/09/2008
research articles this approach may not be amenable to standard immunodiagnostic methods such as immunohistochemistry (solid tissues) or ELISA (biological fluids) and as such may be difficult to translate into the clinical setting. Assessing protein profiles directly via proteomic approaches offers a complementary scheme to transcriptomics for biomarker discovery. Soldes et al.10 demonstrated decreased expression of the stress response protein Hsp27 in Barrett’s metaplasia using two-dimensional polyacrylamide gel electrophoresis (2D-PAGE). The authors proposed that low levels of this cytoprotective agent may increase the sensitivity of Barrett’s epithelium to oxidative stress, thereby promoting its progression toward dysplasia and cancer. In a rat surgical model of EA,11,12 overexpression of leukotriene A4 hydrolase (LTA4H) was detected in EA tissues compared with normal epithelium via 2D-PAGE. Overexpression was confirmed in human samples using immunohistochemistry; metaplastic, dysplastic and EA tissues demonstrated stronger LTA4H staining intensity than squamous epithelium. As LTA4H is involved in the synthesis of the inflammatory mediator leukotriene B4 and can be inhibited by pharmacological agents, LTA4H has been proposed as a potential new target for chemoprevention. Finally, direct proteomic analysis of human esophageal biopsies has recently revealed upregulation of Rho GDP-dissociation inhibitor 2, alpha enolase, lamin A/C and nucleoside diphosphate kinase in esophageal adenocarcinomas and highlighted overexpression of the p53 inhibitor, anterior gradient-2 protein in Barrett’s epithelium.13,14 To identify novel pathways linked to esophageal carcinogenesis, we have used 2D-PAGE to compare a panel of 10 cell lines representing different stages of the disease process. In addition to benign squamous epithelial and cancerous cell lines, the study uniquely includes comparison of the proteome of several metaplastic and dysplastic esophageal cell lines. A panel of proteins which exhibit altered expression across these cell lines is described, including upregulation of cathepsin D and aldo-keto reductases 1C2 and 1B10 in metaplastic and some EA cell lines. To assess the possible significance of cathepsin D and aldo-keto reductases 1C2 and 1B10 in relation to the true disease process, expression of these genes has been assessed in a large series of esophageal biopsies collected from patients with nonerosive reflux disease (NERD), erosive reflux disease (ERD), BE and EA. The results validate the cell line model approach and highlight dysregulation of cathepsin D and aldo-keto reductases 1C2 and 1B10 in EA development for the first time.
Experimental Procedures Cell Culture. The 10 human esophageal cell lines compared in this study were grown at 37 °C in 75 cm2 flasks and 5% CO2. HET-1A is a noncancer derived, SV40 immortalized squamous epithelial cell line.15 It was provided by Dr. C. C. Harris (NCI, Bethesda, MD) and was cultured using the airway epithelial cell growth medium (Promocell) in a flask previously coated with the FNC coating mix (Stratech). The CP-A hTRT cell line was isolated from biopsies of nondysplastic Barrett’s metaplasia, whereas the CP-C hTRT and CP-D hTRT cells originated from Barrett’s derived high-grade dysplasia collected from different patients.16,17 These 3 cell lines were kindly provided by Dr. Peter S. Rabinovitch (University of Washington, Seattle, WA). Their lifespan has been extended by transducing the human catalytic subunit of telomerase reverse transcriptase (hTRT). These cell lines were cultured in 1954
Journal of Proteome Research • Vol. 7, No. 5, 2008
Breton et al. MCDB153 (Sigma) adjusted to pH 7.14 and supplemented with 0.4 µg/mL hydrocortisone (Sigma), 20 ng/mL epidermal growth factor (Sigma), 125 µg/mL bovine pituitary extract (GIBCO), 10-10 M cholera toxin (Sigma), 20 µg/mL adenine (Sigma), 585 µg/mL glutamine (Sigma), 100 U/mL penicillin-streptomycin (GIBCO), 250 ng/mL amphotericin B (GIBCO), 1% InsulinTransferin-Selenium (GIBCO) and 5% fetal bovine serum (FBS, GIBCO). The FLO-1 SEG-1, BIC-1 and OE-33 cell lines originated from primary esophageal adenocarcinomas. FLO-1, SEG-1 and BIC-1 cell lines (kindly provided by Dr. David G. Beer, University of Michigan),10 were cultured in 90% DMEM (GIBCO)/10% FBS. The OE-33 cell line was cultured in 90% RPMI (GIBCO)/10% FBS. The KYSE-30 and OE-21 cell lines were derived from esophageal squamous cell carcinomas (SCC)18 and obtained from the ECACC (Salisbury, U.K.). The KYSE-30 cells were grown in RPMI 1640/Ham F12 (GIBCO) (1:1) supplemented with 5% FBS. The OE-21 cells were grown in 90% RPMI/10% FBS. Cell Lysis and Two-Dimensional Electrophoresis. Cells were grown until 80–90% confluent. For harvesting, cells were washed three times with 25 mL of ice-cold PBS, a fourth time with 25 mL of ice-cold sucrose (0.25 M) and drained thoroughly. Cells were detached and lysed with 1 mL of lysis buffer that included 7 M urea, 2 M thiourea, 4% (w/v) CHAPS, 64 mM DTT, 0.8% (w/v) Pharmalyte pH 3–10 (GE Healthcare), 1 mg/mL Pefabloc (Merck), and complete mini protease inhibitor (Roche, 1 tablet/3 mL of solution). The lysate was scraped off, transferred to an ultracentrifuge tube, and held at room temperature for 30 min before centrifugation at 42 000g for 1 h at 15 °C. Protein extracts were stored at -80 °C. The protein concentration of lysates was determined using the Bradford assay (Bio-Rad). Each lysate (100 µg of total protein) was mixed with DeStreak rehydration solution (GE Healthcare) to a final volume of 0.5 mL, and applied to a 24 cm pH3–10 linear IPG strip (Bio-Rad) overnight. IEF was carried out using the Protean IEF system (Bio-Rad) with the following program: (1) 250 V, 12 h; (2) 10 000 V, 2.5 h; (3) 60 000 V/h; (4) 500 V until equilibration. The first equilibration step was carried out by washing the strips for 15 min in a buffer containing 20 mg/mL DTT, 6 M urea, 0.05 M Tris-HCl (pH 8.8), 2% (w/v) SDS, and 20% (v/v) glycerol. The second equilibration step (15 min) used the same buffer but with iodoacetamide 25 mg/mL replacing DTT. For the second dimension, strips were transferred onto SDSPAGE gels containing 12% (w/v) acrylamide, prepared with a multicasting chamber (Bio-Rad). After introduction of a protein ladder and sealing with 1% (w/v) agarose (mixed with bromophenol blue), migration was carried out at a constant voltage of 200 V in the Protean Plus DoDecaCell tank (Bio-Rad) cooled by water flow maintained at 10 °C. Gels were incubated for at least 1 h in 40% (v/v) methanol, 10% (v/v) acetic acid and washed three times with 30% (v/v) ethanol. A sensitization stage was carried out for 1.5 min in a solution of sodium thiosulphate pentahydrate (350 mg/L) before 3 water washes and 20 min silver staining in 0.2% (w/v) silver nitrate/0.02% (v/v) formaldehyde (37% w/v). Gels were washed three times in water before incubation in the developing solution (30 g/L sodium carbonate, 5 mg/L sodium thiosulphate, and 0.05% (v/v) formaldehyde (37% w/v)). The development was stopped with 2 water washes and a 0.5% (w/ v) glycine stopping solution bath for 5 min. Gels were transferred for 30 min in water and washed two more times before
research articles
Proteome of a Cell Line Model of Esophageal Carcinogenesis
Figure 1. Representative 2D gels obtained from the SEG-1 and CP-A hTRT cell lines. Gel areas subjected to analysis across cell lines are delineated by a dotted line. The location of proteins demonstrating differences across cell lines corresponds to numbers in Table 2. Table 1. Groups of Cell Lines Constituted for the Systematic Comparison of 2D-E Gels Using the PD-Quest Softwarea HET-1A
Non affected squamous epithelium Barrett’s Metaplasia Barrett’s Dysplasia EA SCC Barrett’s Metaplasia + dysplasia Barrett’s Dysplasia + EA Barrett’s Metaplasia + dysplasia + EA Cancerous Noncancerous a
CP-A hTRT
CP-C hTRT
CP-D hTRT
X
X
BIC
FLO
SEG
OE-33
X
X
X
X
KYSE-30
OE-21
X
X
X
X
X X
X
X
X
X X X
X X X
X
X
X
X X X
X X X
X X X
X X X
EA, esophageal adenocarcinoma; SCC, squamous cell carcinoma.
scanning. This silver staining procedure is compatible with mass spectrometry analysis. Image Analysis. Gels were scanned using a Bio-Rad GS-800 calibrated densitometer and the images analyzed using PDQuest software version 7.0.1 (Bio-Rad) within the restricted areas indicated in Figure 1. Other sections of the gels did not provide sufficient resolution for reliable analyses. The cell lines were categorized in relation to the disease process in order to carry out the comparisons (Table 1). Using the PDQuest software, we compared all groups with each other and selected spots either over- or underexpressed more than 4-fold. Each series of 10 gels was analyzed twice with the software. In addition, spots (previously selected or not by the automated analysis) were checked individually by eye in order to avoid possible errors by the software. Identification of Proteins Using Mass Spectrometry and Peptide Mass Fingerprinting. Spots selected by the previous steps were excised and transferred into a 96-well plate manually or using the ProteomeWorks Spot Cutter robot (Bio-Rad). The excised pieces of gel were processed automatically by a MassPREP workstation (Waters) performing the following stages: washing/destaining (1:1 mixture of 30 mM potassium ferricyanide and 100 mM sodium thiosulphate), reduction (DTT 10 mM, ammonium bicarbonate 100 mM), alkylation (iodoacetamide 55 mM, ammonium bicarbonate 100 mM), acetonitrile dehydration, digestion (trypsin 6 ng/µL), extraction from the gel (1% (v/v) formic acid, 2% (v/v) acetonitrile) and finally spotting of the peptides and matrix (R-cyano-4-hydroxycinnamic acid, 2 mg/mL resuspended in 50% (v/v) acetonitrile,
0.1% (v/v) trifluoroacetic acid) on the MALDI plate (Waters). A M@LDI L/R MALDI-TOF mass spectrometer (Waters) was used to obtain the peptide mass spectra. Internal calibration was performed using a trypsin autodigestion peak (m/z 2211.105). Monoisotopic peak masses presenting m/z ratios from 900 to 2800 were submitted to the Mascot (Matrix Science Ltd.) search engine in order to identify proteins, using the following parameters: database, Swiss-Prot v53.0; taxonomy, human (16686 sequences); variable modifications, carbamidomethyl (C), propionamide (C) and oxidation (M); peptide mass tolerance, typically 200 ppm; 1 missed cleavage allowed; exclusion of commonly observed trypsin autolysis fragments and any contaminant keratin peptides (Supporting Information). Reproducibility. After obtaining the first set of protein candidates using the procedures described above, the complete protocol, from cell revival and culture, to mass spectrometry and peptide mass fingerprinting, was repeated. Western Blot Confirmation of Cathepsin D Expression Profile (Cell Lines). To our knowledge, antibodies specific to AKR1C2 and AKR1B10 are not commercially available and would be difficult to generate because of the extensive homology between AKR1C family members.19 For this reason, confirmation of 2D-PAGE results by Western blotting was confined to cathepsin D. Six micrograms of protein (contained in the lysis buffer described above) lysate was diluted 1× in Laemmli reducing buffer (62.5 mM Tris-HCl, 2% (w/v) SDS, 25% (v/v) glycerol and 2.5 mg/mL DTT). Lysate from the A431 cell line (Upstate, Chandlers Ford, U.K.), known to express cathepsin D, was used as a positive control. Proteins were then separated Journal of Proteome Research • Vol. 7, No. 5, 2008 1955
research articles at 100 V on a 10% (w/v) SDS-PAGE ReadyGel (Bio-Rad). To assess any discrepancies in protein loading, a replicate gel was run in parallel. This gel was stained with Coomassie blue, scanned and analyzed using Quantity-One software (Bio-Rad). Proteins from the first gel were transferred to a PVDF membrane and subsequently blocked for 1 h with a 5% (w/v) milk solution (in TBS Tween-20). Blots were incubated for 1 h with an anti-cathepsin D mouse monoclonal antibody (BC011, 1/5000, Calbiochem) in TBST/1% (w/v) milk powder. After washing, membranes were incubated for another hour with the secondary antibody (HRP-conjugated goat anti-mouse IgG, 1/2000, Pierce) in TBST/1% (w/v) milk powder. Signals were visualized using the SuperSignal West Dura extended duration substrate kit (Pierce). The Western blot was performed on two independent occasions to confirm results. Quantitative RT-PCR for AKR1C2, AKR1B10, and Cathepsin D (Cell Lines). All cell lines were revived and cultured in the same conditions used during the proteomics experiments. RNA was extracted using the Qiagen RNeasy Midi Kit according to the manufacturer’s instructions. A DNase digestion step was performed using the Qiagen RNase free DNase set. RNA quality was checked by OD measurement (260/280 nm ratio) using a NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies, Montchanin, DE) and by 2% (w/v) agarose gel electrophoresis. Two micrograms of RNA was reverse-transcribed using the ThermoScript RT-PCR system (Invitrogen). Primer concentrations were optimized and relative amplification efficiencies of housekeeping (HKG) and target genes were assessed as appropriate. HMBS was chosen as the HKG for the cell line analyses, after comparison with the expression of 3 other HKG candidates (TUBA3, RPS9, 18S rRNA). HMBS revealed the least variation in gene expression across cell lines (Ct standard deviation
