ANTICANCER RESEARCH 24: 2627-2632 (2004)
Effect of Confluence State and Passaging on Global Cancer Gene Expression Pattern in Oral Carcinoma Cell Lines MERJA RUUTU1, BO JOHANSSON1, REIDAR GRENMAN2, KARI SYRJÄNEN3 and STINA SYRJÄNEN1 1Department
of Oral Pathology and Oral Radiology and MediCity Research Laboratory, Faculty of Medicine, University of Turku, Lemminkäisenkatu 2, 20520 Turku; 2Department of Otorhinolaryngology – Head and Neck Surgery and Medical Biochemistry, Faculty of Medicine, University of Turku, Kiinanmyllynkatu 6, 20520 Turku, Finland; 3Unit of Cytopathology, National Centre of Epidemiology, Surveillance and Promotion of Health, Istituto Superiore di Sanità (ISS), Viale Regina Elena, 299, 00161, Rome, Italy
Abstract. Background: After establishment of a cell culture, inhomogenous growth and cell selection takes place at early passages. Reproducing the same confluence state among experiments might be difficult. Materials and Methods: Here we used cDNA arrays to compare the global gene expression pattern of two oral cancer cell lines at 80%, 100% and superconfluence stages. Also, the stability of the global gene expression pattern during culture was assessed in two cell lines at passages 10 and 53 (early and late passage). An intraclass correlation coefficient test was used for comparisons. Results: The consistency between the different confluence states was almost perfect (> 0.8) and substantial (0.6-0.8). Also the consistency between early and late passages was almost perfect in the cell lines used. Conclusion: Our results indicate that neither the state of confluence nor passage have significant effects on the global gene expression profile, despite the variability in expression levels of individual genes. DNA microarray technology has been introduced as a new screening tool to detect novel biomarkers in oral carcinoma (1-3). Instead of oral carcinoma tissue, tumor-derived cell lines can be used to study the molecular events in oral carcinogenesis (4-6) and they are also suitable for microarray studies (7-10). In most studies, the cell lines are grown as monolayers on plastic support, which offers some advantages over organ maintenance. The ability to cultivate large numbers of cells is
important in array studies, where approximately 5-20 Ìg of RNA is needed per assay. A major disadvantage is that cells dedifferentiate in culture. Furthermore, the cells are usually harvested as sub-confluent cultures for assays, and it is nearly impossible to standardize the confluence state at harvest. Another problem, not adequately addressed, is the heterogeneity of keratinocytes at early passages after the establishment of cultures from the original tumors (11). There is evidence suggesting that the phenotype of the carcinoma cells depends on the culture conditions (12,13), the early passage cells being more heterogeneous due to the ongoing selection of immortal tumor cells. Todate, little attention has been focused on the standardization of the materials used in the array studies. No systematic data is available concerning the effect of confluency or passaging on gene expression profiles in cell lines. The aim of this study was to assess the stability of the gene expression profile of oral carcinoma cell lines at three different confluence states; 1) sub-confluent, 2) confluent and 3) super-confluent states. In addition, the effect of passaging on overall gene expression was studied. In this study, a cDNA array was used to investigate the effects of cultivation specifically on the expression of cancer-related genes in carcinoma cell lines. Microarray techniques can give information about whether cells and genes are globally affected by cultivation conditions, as well as showing significant changes in individual genes.
Materials and Methods Correspondence to: Prof. Stina Syrjänen, DDS, PhD, Department of Oral Pathology and Oral Radiology, MediCity Research Laboratory, Faculty of Medicine, University of Turku, Lemminkäisenkatu 2, 20520 Turku, Finland. Tel: +358-2-3338349, Fax: 358-2-3338399, e-mail:
[email protected] Key Words: Confluence, oral carcinoma cell lines, gene expression.
0250-7005/2004 $2.00+.40
Cell lines. Three cell lines established from oral squamous cell carcinomas (14) were cultured in Dulbecco’s minimal essential medium (D-MEM, Paisley, UK), supplemented with 1% non essential amino acids, 2 mM L-glutamine, 50 mg/ml streptomycin, 100 U/ml penicillin and 10% fetal calf serum, in 75-mm3 bottles. For the confluence studies, the UT-SSC-10 and UT-SCC-37 cell lines were used. The cell line data is given in Table I. The cell lines
2627
ANTICANCER RESEARCH 24: 2627-2632 (2004) Table I. Characterization of the cell lines, original tumors and the patients. Cell line and passage
Sex and age
Origin
UT-SCC 10 p8
Male, 62 yrs
Tongue
UT-SCC 37 p10, p 53
Female, 61 yrs
Buccal mucosa
UT-SCC 54B p10, p53
Female, 58 yrs
Buccal mucosa
Clinical status
TNM and histological grade
Smoking history
Survival
T1N0M0, G II
25 years
Died in 12 months
Oral lichen planus
T2N0M0, G I
40 years
Disease-free
Oral lichen planus
T2N0M0 G I
No
Disease-free
were harvested at 80% confluence, 100% confluence and superconfluence. To reach super-confluence, the cells were grown for an additional 6 days after confluence. The medium was changed every second day for the two first days, and then every day until the super-confluent state was reached. Both cell lines had the same growth properties. The effect of passaging was studied by using two cell lines, UTSCC 37 and UT-SCC 54B at passage 10 and passage 53. The characteristics of these cell lines are given in Table I. The origin and behavior of the tumors were similar. Also, the overall global gene expression profiles of these cell lines were similar at early passages. The cells were harvested when grown to 80% confluence. RNA extraction. Total RNA was isolated from the cells by CsClgradient centrifugation. Cells were first lysed with GIT buffer (4 M guanidine isothiocyanate, 1 M sodium-citrate, beta-mercaptoethanol 0.7%, sodium-laurylsarcosine 0.5% pH 7.0) and the cell suspension was mixed with 5.7 M caesium chloride (with 2.5% 1 M sodium citrate) and ultra-centrifuged (30,000 rpm, 21 hours, +20ÆC). RNA was collected from the bottom of the ultracentrifuge tube, impurities were extracted with phenol:chloroform (1:1) and the RNA was precipitated with ethanol. The ultra-centrifugation procedure results in a very pure DNA-free RNA (A260/280 1.96-2.0). cDNA array analysis. An aliquot of 5 Ìg of total RNA was used in the array analysis, performed according to Clontech’s Human Cancer 1.2 cDNA expression array kit manual (Clontech, Palo Alto, CA, USA). Total RNA was reverse-transcribed with MMLV reverse transcriptase and labelled with ·-32P ATP (Amersham Pharmacia Biotech Inc., Piscataway, NJ, USA). Labelled cDNAs were purified with column chromatography and hybridized overnight to sequences on nylon membranes, representing 1.176 cancer-related genes and nine housekeeping genes, according to the manufacturer. The reagents documented in Clontech’s array kit were used for all the hybridization and washing steps. The membranes were phosphoimaged by a Fuji BAS 5000 Image Analysis System (FUJIFILM Medical Systems, USA) and the intensities of spots were analyzed with Clontech’s AtlasImage 2.0 software. Gene expression data were normalized against those of the housekeeping gene GAPDH (liver glyceraldehyde 3-phosphate dehydrogenase). Other housekeeping genes were at a constant level in all arrays. The Cluster and TreeView programs by Michael Eisen from Stanford University, USA (15) were used to analyze the
2628
data. In the cluster analysis (hierarchical clustering), all gene values were normalized against the data obtained at the sub-confluent state. In our previous studies, the reproducibility of the Clontech’ s cDNA array kit showed correlation coefficients over 0.8 (16). Statistical analysis of agreement and reproducibility. The consistency (agreement, reproducibility) in the global gene expression (for 1,185 genes), under three different conditions of confluence and in two different passages, was calculated for the two cell lines (UTSSC-10 and UT-SCC-37) separately, using the reliability analysis for the intraclass correlation coefficient (ICC) (17). The ICC results (with 95% CI) were interpreted using the average measure intraclass correlation, thus excluding the assumption that there are some confluence-gene expression interactions that are uncontrollable by the model (18).
Results and Discussion Consistency and reproducibility of global gene expression. Our basic hypothesis was that confluence state as well as number of passages significantly affects global gene expression patterns. To test this hypothesis, we used statistical methods to assess the global changes in cancer-related gene expression patterns. The test for ICC (17) was chosen. Originally introduced to assess inter-rater agreement, this ANOVA-type model was applied here by using the expression scores of each individual gene as the "responses", and the different confluence states (and passages) as the "raters". As shown in Table II, the consistency between the different confluence states of UT-SCC 10 was almost perfect (> 0.8). In UT-SCC 37, the consistency between the different confluence states was slightly less but still substantial (0.6-0.8). Using the absolute agreement model, ICC between the tree confluence states was substantial in UT-SCC 37 and almost perfect in UT-SSC 10. The agreement was lowest, but still moderate (0.4-0.6), between sub-confluent and confluent sates of UT-SCC 37. Table III summarizes the effect of passaging on the global gene expression of 1,185 genes of the UT-DEC 37
Ruutu et al: Effect of Confluence and Passaging on Global Cancer Gene Expression
Table II. The effect of confluence states on the global gene expression (1,185 genes) of UT-DEC 10 and UT-DEC37 cell lines calculated for both consistency and absolute agreement..
Table III. The effect of different passages on the global gene expression (1,185 genes) of UT-DEC 37 and UT-DEC54 cell lines calculated for consistency and absolute agreement.
Cell lines/ ICC test model
Cell lines/ICC test
Consistency: UT-SCC 37
UT-SCC 10
Comparisons Intraclass Correlation Coefficient (ICC) (95% CI) SUB/ SUB/ CONF/ ALL CONF. SUPER SUPER STATES
0.632 0.643 0.835 0.802 (0.588-0.672) (0.600-0.682) (0.815-0.853) (0.781-0.820) 0.858 0.902 0.802 0.898 (0.841-0.873) (0.890-0.912) (0.778-0.823) (0.888-0.908)
Comparisons Intraclass Correlation Coefficient (ICC) (95%CI) Passage 10 versus Passage 53
Consistency: UT-SCC-37 UT-SCC-54 Absolute Agreement: UT-SCC 37 UT-SCC- 54
0.858 (0.841-0.873) 0.888 (0.874-0.900) 0.713 (-0.174-0.892) 0.670 (-0.241-0.891)
Absolute Agreement: UT-SCC 37 0.536 0.643 0.758 0.746 (0.091-0.728) (0.600-0.681) (0.300-0.883) (0.575-0.834)
*ICC values: >0.8 almost perfect; 0.6-0.8 substantial; 0.4-0.6 moderate; 0.2-0.4 low; 0.8 almost perfect; 0.6-0.8 substantial; 0.4-0.6 moderate; 0.2-0.4 low;
