Differential gene expression profiles of radioresistant oesophageal cancer cell lines established by continuous fractionated irradiation

Cells and cell culture

The human oesophageal squamous cancer cell lines TE-2, TE-5, TE-9, and TE-13 were obtained from the Cell Resource Center for Biomedical Research Institute of Development, Aging and Cancer (Tohoku University, Sendai, Japan) (Nishihira et al, 1993). Cell lines KYSE170 and KYSE180 were kindly provided by Dr Shimada (Kyoto University, Kyoto, Japan) (Shimada et al, 1992). Cells were cultured in RPMI-1640 (Life Technologies, Grand Island, NY, USA) with 100 U ml⁻¹ of penicillin. 100 μg ml⁻¹ of streptomycin, and 10% fetal bovine serum, and incubated at 37°C in 5% CO₂/95% air.

Establishment of radioresistant cell lines

The cells were first grown to approximately 50% confluence in vented 25-cm² culture flasks. Cells were treated with 2 Gy of X-ray irradiation (120 kV and 3.5 mA for 4 min with a 0.5-mm A1 filter) using an X-ray generator (M-150 WE: SOFTEX Co., Tokyo, Japan) and then returned to the incubator. When they reached approximately 90% confluence, the cells were trypsinised and subcultured into new flasks. When they reached approximately 50% confluence, the cells were again irradiated (second fraction). The fractionated irradiations were continued until the total doses reached 60 Gy. The parental cells were subjected to identical trypsinisation, replating, and culture conditions but were not irradiated. For all assays on irradiated cells, there was at least a 2-week interval between the last 2 Gy FIR and the experiment.

Assay for radiosensitivity

Cell survival after X-irradiation was measured by clonogenic assay. The cells (10⁵ cells per dish) were incubated at 37°C for 48 h in the growth medium and were irradiated at different doses ranging from 0 to 10 Gy. The cells were harvested with EDTA–trypsin immediately after irradiation. Subsequently, a constant number of cells were plated in each 60-mm tissue culture dish (Iwaki Glass, Chiba, Japan), and incubated at 37°C for 10–14 days (three plates in each radiation dose). After fixation with formalin and staining with 0.1% crystal violet, colonies consisting of 50 cells or more were counted under a light microscope, and the surviving fraction was determined.

All survival curves represent at least three independent experiments. The data were fitted to the single hit, multitarget formula, where survival (S) is related to dose (D) by the expression S=1−(1−e^−DD₀)ⁿ≒ne^−DD₀. D₀ was used as a parameter to indicate the amount of irradiation required to reduce the survival fraction to approximately 0.37 from the survival curve.

Flow cytometry

Cells were analysed during exponential growth. They were trypsinised, rinsed in PBS, and fixed in 70% ethanol. Before analysis, cells were spun out of ethanol and resuspended in 1.6 ml PBS, 0.2 ml propidium iodide (PI) 400 μg ml⁻¹ (Sigma Chemical Co., Poole, UK), and 0.2 ml RNase 1 mg ml⁻¹ (Sigma). After incubation for 30 min at 37°C, the cells were spun down and resuspended in PBS. The cells were analysed by a FACScan flow cytometer (Becton Dickinson, San Jose, CA, USA) for DNA content using PI and a photomultiplier masked with a 610 nm long-pass filter. The data were analysed with a software program (Cell Quest; Becton Dickinson) with dead cells gated out using pulse processing. Two independent experiments were performed for each data set.

RNA extraction

Total RNA was extracted from each cell line with the RNeasy Midi Kit (QIAGEN) according to the manufacturer's protocol. The extracted RNA was treated with RNase-Free DNase set (QIAGEN) to remove any contamination of genomic DNA according to the manufacture's protocol.

Microarray design, production, and hybridisation

A 20 μg portion of total RNA extracted from each radioresistant subline was labelled with Cy3, and RNA from each parent cell line was labelled with Cy5 as described elsewhere (Sakakura et al, 2002). We used the RIKEN human cDNA microarray, which contains 21168 genes. Hybridisation, washing, and scanning were carried out according to a published method (Sakakura et al, 2002). These slides were scanned on a ScanArray 5000 confocal laser scanner, and the images were analysed using IMAGENE (Bio Discovery, Los Angeles, CA, USA).

Analysis of the data

To improve the accuracy of the data, we did the experiment twice, labelling the same RNA template in two separate reactions. Data were normalised to the reference standard by subtracting (in log space) the median observed value if other than zero. We used only data points that were reproducible. To this end, we developed a filtering program, PRIM (Kaota et al, 2001). After filtering was finished, we compared the results of the two experiments by calculating a Pearson's correlation coefficient. If the coefficient was equal to or greater than 0.7, we used the data in subsequent analysis. Ratio values from duplicate experiments were averaged, log-transformed (base 2), and stored in a table. Two-fold differences in up- or downregulated expression were used to identify altered genes. Hierarchical clustering and dendrogram figures were generated using Cluster and Tree-View software (http://rana.standford.edu).

Quantitative RT–PCR

A quantitative RT–PCR assay was used to follow-up on microarray data. A 4 μg portion of total RNA was used for creation of single-stranded cDNA using the SuperScript Preamplification System for First Strand cDNA Synthesis (Life Technologies Inc., Rockville, MD, USA). The cDNA was diluted and quantitatively equalised for PCR amplification. The TaqMan PCR method using a Gene Amp 5700 Sequence Detection System (Applied Biosystems, Foster City, CA, USA) was performed according to the manufacturer's instructions. We used the primers and probes provided as Assay-on-Demand Gene Expression Products (Applied Biosystems). We used beta-actin as an internal control gene to normalise the expression rate of each gene. The details have been described elsewhere (Sakakura et al, 2004). Each experiment was carried out three times and the average was calculated.

Clinical specimens

Eight patients with advanced squamous cell carcinoma of the oesophagus (T3 or T4) treated with preoperative radiation therapy at Kyoto Prefectural University of Medicine were subjected to preliminary clinical analysis. Tissue samples were obtained from the tumour site for an initial diagnostic biopsy, and 10–40 μg of total RNA was extracted, and cDNA was synthesised from 4μg of total RNA as described. Preoperative irradiation was performed with a dose of 2 Gy per fraction to a total dose of 40 Gy. The effects of radiation therapy in the resected specimens were determined using the histopathologic criteria of the Japanese Society for Esophageal Disease. Grade 0 represents no radiotherapy effect; grade 1, slightly effective, with viable cells recognised as more than one-third cancer cells; grade 2, moderately effective, with less than one-third cancer cells; and grade 3, markedly effective, with no viable cancer cells. We compared the expression rates of the selected genes for group A (patients whose histological grade after radiotherapy was either 0 or 1) with those for group B (patients whose histological grade after radiotherapy was either 2 or 3). A statistical analysis was performed using the Mann–Whitney U-test. The results with P-values of less than 0.05 were considered to be statistically significant. All experiments with clinical samples were conducted under institutional guidelines from the Ministry of Health and Welfare.

Establishment of cell lines resistant to irradiation

The human oesophageal cancer cell lines were treated with FIR, and all cell lines survived. The cell populations surviving FIR were analysed for their radiosensitivity using a clonogenic assay to assess survival after a single dose of radiation. The survival curve parameters for all cell lines are shown in Table 1

Cell cycle analysis

Cells in S phase tend to be more radiosensitive than cells in G1, and cells in G2–M are the most sensitive of all (Sinclair and Morton, 1966). Therefore, we examined the cell cycle distribution of cell lines using flow cytometry. The results in Table 2

Gene selection from cDNA microarray data analysis

We performed global expression analysis of 21168 genes using a high-density cDNA microarray and compared the four profiles of parent lines and radioresistant sublines. We identified 47 genes as differentially expressed in at least three of the four radioresistant sublines. In all, 19 genes were upregulated (Table 3

To verify the expression of genes identified in microarray experiments, real-time quantitative RT–PCR was performed using the same RNA as that used in the microarray analysis. We found that the results were in good agreement with those from the microarray data. Typical data for COX-2 and BIRC2 are shown in Figure 1

Figure 1

Quantitative RT–PCR analysis of the genes selected for cDNA microarray data validation. (A) Relative expression rate of COX-2 was normalised to beta-actin. (B) Relative expression rate of BIRC2 was normalised to beta-actin. Data represent mean±standard deviation (n=3).

Effects of radiotherapy and expression of selected genes in clinical specimens

In eight cases of squamous cell carcinoma of the oesophagus, COX-2 and BIRC2 mRNA expressions of pretreatment biopsy samples were assayed by quantitative RT–PCR. Although the expression rates of COX-2 for group A were not significantly higher than those for group B (Figure 2A

Figure 2

Effect of radiotherapy and expression rate of COX-2 (A) and BIRC2 (B) in clinical specimens. Expression rates were normalised to beta-actin. The grade of radiotherapy effect was determined based on the histopathological criteria of the Japanese Society for Esophageal Diseases. Grade 0 (•), ineffective; Grade 1 (○), slightly effective; Grade 2 (▴), moderately effective; and Grade 3 (▵), markedly effective. Group A included patients whose histological grade after radiotherapy was either 0 or 1, and group B included patients whose grade was either 2 or 3. The statistical analysis was performed using the Mann–Whitney U-test.

Genes related to apoptosis and cell cycle

Prostaglandin-endoperoxide synthase (PTSG), also know as cyclooxygenase (COX), is the key enzyme in prostaglandin biosynthesis, and acts both as a dioxygenase and as a peroxidase. There are two isozymes of COX: a constitutive COX-1 and an inducible COX-2, which differ in their regulation of expression and tissue distribution. COX-2 is expressed in a limited number of cell types and regulated by specific stimulatory events, suggesting that it is responsible for the prostanoid biosynthesis involved in inflammation and mitogenesis. Recent studies have highlighted the relevance of COX-2 in human carcinogenesis (Eberhart et al, 1994; Ristimaki et al, 1997; Hida et al, 1998; Koga et al, 1999; Tucker et al, 1999). In oesophageal cancer, increased levels of COX-2 have also been reported (Hwang et al, 1998; Wilson et al, 1998; Katja et al, 1999; Shamma et al, 2000). Although the role of COX-2 in oesophageal cancer is not clear, it is possible that COX-2 is involved in the regulation of cell survival and maintenance of growth because COX-2 inhibitors are known to induce apoptosis and inhibit growth in oesophageal carcinoma cells (Katja et al, 1999). Furthermore, Pyo et al (2001) showed that selective COX-2 inhibitors enhanced the radiosensitivity of rat intestinal epithelial cells, which were stably transfected with COX-2 cDNA, and Nakata et al (2004) reported similar findings with human A431 epidermoid carcinoma. Our data suggest that COX-2 may play an important role in radioresistance induced by FIR, and COX-2 inhibitors may prevent tumour cells from acquiring radioresistance. Furthermore, in our preliminary clinical study, the expression rate of COX-2 tended to reflect the effect of radiotherapy (Figure 2A, P=0.053). Additional investigation will be necessary to confirm this hypothesis.

BIRC2, also known as cIAP1, is a gene that encodes an antiapoptotic molecule in the IAP family: oesophageal squamous cell carcinoma cell lines overexpressing this gene were resistant to apoptosis induced by chemotherapeutic reagents (Imoto et al, 2001). Furthermore, Imoto et al (2002) reported that the expression of cIAP1 was correlated with resistance of cervical cancers to radiotherapy. In our study, the expression rates of BIRC2 were significantly higher in patients whose histological grade after radiotherapy was either 0 or 1 than those patients whose histological grade was either 2 or 3 (Figure 2B, P=0.025). These data also suggest that we may predict the effect of radiotherapy using the expression rate of BIRC2.

Upregulation of ICEBERG, a caspase-1 inhibitor, and downregulation of caspase 6 (CASP6) are extremely intriguing. ICEBERG proteins are intracellular regulators of caspase-1 activation and could play a role in the regulation of IL-1β secretion and NF-κB activation during the proinflammatory cytokine response (Humke et al, 2000; Druilhe et al, 2001), NF-κB is a well-defined radiation-responsive transcription factor (Thanos and Maniatis, 1995). Furthermore, NF-κB is associated with the cell cycle, and we found that cyclin A2 (CCNA2) was downregulated. However, genes such as p53, cyclin D1, and NF-κB did not differ in our study, and the cell cycles of radioresistant sublines were similar to that of parent cell lines (Table 2). Another mechanism of cell cycle regulation may exist.

Genes related to growth and metabolism

CD73, also termed 5′-nucleotidase, is a signalling pathway metabolite in the immune response of lymphocytes. This ectoenzyme catalyses the conversion of purine and pyrimidine ribo- and deoxyribonucleoside monophosphates (AMP, GMP, IMP) and leads to an elevation of the corresponding nucleosides (adenosine) in the extracellular space, perhaps modulating neuronal signalling and vascular perfusion, CD73 has also been called a cellular motility factor. There is an increasing amount of evidence for a modulatory role of PKC-mediated CD73 activity in ischaemia, regeneration and repair, glioma cell proliferation, and tissue invasion. Ludwig et al (1999) reported that CD73 expression was correlated with expression of protein kinase C (PKC) and epidermal growth factor receptor (EGFR). Protein kinase C is a known tumour metabolite in several proliferation-promoting pathways of EGF receptor signalling. Furthermore, Galmarini et al (2002) reported that CD73-positive acute myeloid leukaemia (AML) patients had worse disease-free survival and overall survival than negative patients, and suggested that CD73 expression might be a mechanism of resistance to cytarabine. To our knowledge, there have been no reports of an association between radioresistance and CD73.

Genes related to cell adhesion and motility

In our study, E-cadherin (CDH1), P-cadherin (CDH3), and protocadherin 9 (PCDH9) were downregulated in radioresistant sublines. However, Akimoto et al (1998) reported that E-cadherin expression was increased significantly in human squamous cell carcinoma after irradiation. This difference might be due to fractionated irradiation and acquisition of radioresistance. Some studies have shown that reduced E-cadherin expression was correlated with a high incidence of cancer metastasis and invasion (Doki et al, 1993; Kadowaki et al, 1994), and radiation enhanced the ability to form metastases (Sheldon and Fowler, 1976). Our results also suggest that the tumour cells that acquired radioresistance may have higher invasive ability and metastatic potential. Recently, an association between E-cadherin and COX-2 has been reported (Tsujii and Dubois, 1995), and Noda et al (2002) found that COX-2 inhibitors affect the transcription of E-cadherin.

Gene expression changes related to apoptosis, cell cycle, growth metabolism, and cell adhesion are likely to contribute to the complexity of radioresistance. It is unlikely that this phenomenon can be explained by altered expression of a single gene: whether this combination of gene expression changes leads to radioresistance of oesophageal cancers, however, is a matter to be investigated further. Identification of such genes could lead to new therapeutic targets.