- Email: [email protected]
Association between gene expression profile and tumor invasion in oral squamous cell carcinoma
Cancer Genetics and Cytogenetics 154 (2004) 27–35
Association between gene expression profile and tumor invasion in oral squamous cell carcinoma Gokce A. Torunera,b,c,1, Celal Ulgera,b,c,1, Mualla Alkana,b, Anthony T. Galanted, Joseph Rinaggioe, Randall Wilkf, Bin Tiang, Patricia Soteropoulosa,d, Meera R. Hameedh, Marvin N. Schwalba,b, James J. Dermodya,b,* a
.
Department of Microbiology and Molecular Genetics, University of Medicine and Dentistry of New Jersey (UMDNJ)–New Jersey Medical School, 185 South Orange Avenue, MSB-F659, Newark, NJ 07103 b Center for Human and Molecular Genetics, UMDNJ–New Jersey Medical School, Newark, NJ 07103 c David Jurist Research Center, Hackensack University Medical Center, Hackensack, NJ 07601 d Center for Applied Genomics, International Center for Public Health, Newark, NJ 07103 e Department of Diagnostic Sciences, UMDNJ–New Jersey Dental School, Newark, NJ 07103 f Department of Oral and Maxillofacial Surgery, UMDNJ–New Jersey Dental School, Newark, NJ 07103 g Department of Biochemistry and Molecular Biology, UMDNJ–New Jersey Medical School, Newark, NJ 07103 h Division of Surgical Pathology, Department of Pathology and Laboratory Medicine, UMDNJ–New Jersey Medical School, Newark, NJ 07103 Received 26 November 2003; received in revised form 6 January 2004; accepted 28 January 2004
Abstract
There are limited studies attempting to correlate the expression changes in oral squamous cell carcinoma with clinically relevant variables. We determined the gene expression profile of 16 tumor and 4 normal tissues from 16 patients by means of Affymetrix Hu133A GeneChips. The hybridized RNA was isolated from cells obtained with laser capture microdissection, then was amplified and labeled using T7 polymerase-based in vitro transcription. The expression of 53 genes was found to differ significantly (33 upregulated, 20 downregulated) in normal versus tumor tissues under two independent statistical methods. The expression changes in four selected genes (LGALS1, MMP1, LAGY, and KRT4) were confirmed with reverse transcriptase polymerase chain reaction. Two-dimensional hierarchical clustering of the 53 genes resulted in the samples clustering according to the extent of tumor infiltration: normal epithelial tissue, tumors less than or equal to 4 cm in dimension, and tumors more than 4 cm in dimension (P ⫽ 0.0014). The same pattern of clustering was also observed for the 20 downregulated genes. We did not observe any associations with lymph node metastasis (P ⫽ 0.097). 쑖 2004 Elsevier Inc. All rights reserved.
1. Introduction Oral squamous cell carcinoma (OSCC) is the sixth most common type of carcinoma worldwide. It has been estimated that 30,000 new cases and 7800 OSCC-related deaths occur in the United States annually [1]. The incidence rates are higher in males and blacks. Tobacco products and alcohol are major etiological factors, and the rate of consumption of these products is believed to partially account for the difference in incidence rate according to race and sex [2]. Because OSCC is a major health problem, understanding the molecular basis of this neoplasm is of great importance.
1 These authors contributed equally to this work. * Corresponding author. Tel.: (973) 972-5567; fax: (973) 972-3783. E-mail address: [email protected] (J.J. Dermody).
0165-4608/04/$ – see front matter 쑖 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.cancergencyto.2004.01.026
In the past, the lack of fine extraction techniques of tumor cells from the specimen and the absence of high-throughput technologies for analysis hindered efforts to determine the exact nature of the molecular changes. Technological breakthroughs, however, including laser capture microdissection (LCM) and microarray technology, have fundamentally altered cancer research. LCM enables the isolation of a pure population of cancer or normal cells from the tissue, and microarrays make the simultaneous analysis of thousands of genes possible. Although OSCC is a common cancer, there are few microarray-based studies conducted on these tumors [3–9]. Although all of these previous studies provided insights into gene expression changes in these tumors, only two studies demonstrated an association between gene expression changes and a clinically relevant variable such as patient survival [4] or lymph node metastasis [8].
28
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
Table 1 Characteristics of the OSCC patients, with TNM staging
2. Materials and methods
Patient code
Primary site
Age
Sex
TNM
2.1. Tissue collection
OSCC-2 OSCC-4 OSCC-5 OSCC-6 OSCC-7 OSCC-8 OSCC-9 OSCC-10 OSCC-11 OSCC-12 OSCC-13 OSCC-14 OSCC-15 OSCC-16 OSCC-17 OSCC-18
Floor of mouth Floor of mouth Tongue Vestibule Floor of mouth Tongue Tongue Tongue Tongue Maxilla Floor of mouth Floor of mouth Floor of mouth Mandible Tongue Floor of mouth
58 60 64 72 48 77 66 64 32 46 54 60 51 50 47 52
F M M M M M M M M M M F M F M M
T2N0M0 T4N1M0 T1N2bM0 T2N0M0 T4N1M0 T2N0M0 T2N0M0 T4N2bM0 T3N2bM0 T4N0M0 T4N0M0 T4N2bM0 T2N0M0 T4N0M0 T4N3M0 T4N2M0
Samples were derived from surgically resected tissue from 16 patients with OSCC (Table 1). Informed consent was obtained from patients according to the regulations of the University of Medicine and Dentistry of New Jersey. Tissue samples from 16 tumor and 4 normal tissues of patients were immediately frozen in liquid nitrogen after surgery. Paired control nontumor tissues from patients OSCC-2, OSCC-9, OSCC-14, and OSCC-18 were obtained from a clinically unaffected site and were histologically normal. Frozen tissue was placed in a cryomold with OCT embedding medium on dry ice for 1 minute. The cryomold was filled with embedding medium and covered with dry ice placed at ⫺80⬚C until the cutting period (http://dir.nichd.nih. gov/lcm/LCMTAP.htm). Embedded frozen tissue specimens were cut in a series of 7-µm-thick sections onto plain uncoated glass slides at ⫺20⬚C. The sections were stained with fast hematoxylin–eosin staining in RNase-free recipients and solutions (http://dir.nichd.nih.gov/lcm/LCMTAP.htm#Staining). After staining the tissue, tumor and normal tissues were diagnosed by the pathologist, then LCM was performed. The tumors were staged according to 2003 cancer staging guidelines issued by the American Joint Committee on Cancer. About 120,000 pure tumor and normal epithelial cells were harvested from cryosections using PixCell II
Abbreviation: TNM, tumor–node–metastasis.
We performed microarray analysis using RNA from OSCC cells obtained with LCM from 16 patients and compared these results with 4 control squamous cell epithelium samples. We identified expression profiles of differentially expressed genes between normal and tumor tissues and found that tumor invasiveness of a tumor is highly correlated with its gene expression profile.
Fig. 1. LCM from hematoxylin–eosin stained OSCC cryosection. (A) Section before microdissection and (B) after removal of water and prior to LCM. (C) Tumor island lifted from the surrounding tissue and (D) LCM-captured tumor island.
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
29
Table 2 The upregulated genes in OSCC Gene
GenBank accession no.
Score (d)
Fold change
Cancer association
C3 LGALS1
NM_000064 NM_002305
1.8 1.43
62.42 31.76
No Yes
MMP1
XM_006270
1.50
30.19
Yes
IGLJ3 UBD IGKC G1P2 TPM1 IFI44 KIAA0746 DZIP1 C1orf29 FYB C1S UCP2 G1P3 BF SCAP2 SKIL TLR2 ADA PAPSS2 IGFBP7
M87790.1 NM_006398 M63438.1 NM_005101 Z24727.1 BE049439 AB018289 NM_014934 NM_006820 AF198052.1 M18767.1 U94592.1 NM_022873 NM_001710 NM_003930 BF725121 NM_003264 X02189 AW299958 NM_001553
1.59 1.74 1.87 1.57 1.62 2.30 2.79 1.68 1.89 1.61 1.94 1.65 1.98 2.12 2.38 2.5 1.93 1.65 1.5 1.93
19.92 17.36 16.80 14.11 13.77 11.99 10.32 9.78 9.54 8.00 7.91 7.65 6.43 6.36 6.31 6.14 6.03 5.43 5.4 5.36
No Yes No No Yes No No No No No No No No No No Yes No No No Yes
BIGM103 FLJ22344 IFITM1 NRG1 CHK ALDH5A1 APOL1 CA2
AB040120.1 NM_02471 NM_003641 NM_013959 NM_001277 AL031230 AF323540.1 M36532.1
1.94 1.80 1.88 1.52 1.86 1.97 1.46 1.56
5.22 5.16 5.15 4.77 4.47 4.37 4.36 4.13
No No Yes Yes Yes No No Yes
TSPAN-2 APOL2
BF129969 BC004395.1
2.00 1.68
4.08 4.00
No No
References van den Brule et al., 2003 [20]; Hittelet et al., 2003 [21] Seiki et al., 2003 [19]
Lee et al., 2003 [22]
Raval et al., 2003 [23]
Zhang et al., 2003 [24]
Watson et al., 2002 [25]; Landberg et al., 2001 [26]
Huang et al., 2001 [27] Tsai et al., 2003 [28] Bougeret et al., 2001 [29]
Chiang et al., 2002 [30]; Bekku et. al., 2000 [31]
OSCC association
References
No Yes
Choufani et al., 1999 [12]
Yes
Alevizos et al., 2001 [3]; Nagata et al., 2003 [8]
No No No No No No No No No No No No No No No No No No No No No No No No No No No No No No
The d score is based on the ratio of gene expression change to standard deviation in the data for that gene.
LCM system (Arcturus Engineering, Mountain View, CA) (Fig. 1) [10].
performed to produce biotin-labeled cRNA from the doublestranded cDNA using the Enzo BioArray high yield RNA transcript labeling system (Enzo, Farmingdale, NY).
2.2. Total RNA isolation and target sample generation LCM-isolated tissues were homogenized and total RNA was extracted using the RNeasy kit (Qiagen, Valencia, CA) according to the manufacturer’s procedure (http://www1. qiagen.com/literature/qiagennews/0401/1018052_QNews42001 _p3_5.pdf). The Affymetrix (Santa Clara, CA) small-sample target labeling protocol was used for amplifying and labeling RNA (http://www.cgrb.orst.edu/CSL/downloads/smallsample.pdf). The SuperScript Choice system (Invitrogen Life Technologies, Rockville, MD) was used to synthesize double-stranded cDNA from LCM-derived isolated RNA. After collecting the cDNA, the Ambion MEGAscript T7 kit (Ambion, Austin, TX) was used for one round of linear amplification. A second in vitro transcription reaction was
2.3. Hybridization cRNA (15 µg) was fragmented and added to a hybridization mixture. Expression profiles were created using the HGU133A chip (Affymetrix), which contains ~22,280 known human transcripts and ESTs. These transcripts are represented using probe pairs consisting of 25-mer oligonucleotides. Hybridization was performed overnight at 45⬚C for 16 hours using the GeneChip Hybridization Oven 640 (Affymetrix). Washing and staining (streptavidin–phycoerythrin) was done using EukGE-WS2v3 protocol in the GeneChip Fluidics Station 400 (Affymetrix). Images were acquired using the Affymetrix GeneArray scanner.
30
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
Table 3 List of the downregulated genes in OSCC Gene
GenBank accession no.
Score (d)
Fold change
Cancer association
P11 C1orf10 MAL TGM3
NM_006025 NM_016190 NM_002371 NM_003245
⫺5.89 ⫺7.01 ⫺7.43 ⫺5.26
61.49 53.12 45.99 26.84
No Yes Yes No
Xu et al., 2000 [32] Mimori et al., 2003 [33]
No No Yes Yes
KRT4
X07695.1
⫺5.47
17.08
Yes
Hourihan et al., 2003 [34]
Yes
DSG1
NM_001942
⫺5.33
16.11
Yes
No
LAGY SCEL NICE-1 SPRR3
AB059408.1 NM_003843 NM_019060 NM_005416
⫺6.34 ⫺4.87 ⫺5.36 ⫺4.85
13.71 11.83 11.12 10.10
Yes Yes No Yes
Nakashima et al., 2002 [35]; Lip et. al., 2001 [36] Chen et al., 2003 [37] Huang et al., 2001 [27]
GPX3 PEG3 HPGD FLJ22671 NPR3 CLDN17 CYP3A5P2 PPP1R3C ID4 CEACAM6
AW149846 AL042588 AL574184 NM_02486 AI628360 NM_012131 X90579.1 N26005 NM_001546.1 M18728.1
⫺6.89 ⫺4.00 ⫺5.14 ⫺7.58 ⫺6.24 ⫺6.32 ⫺4.24 ⫺4.82 ⫺3.66 ⫺3.83
8.71 8.58 7.75 6.93 6.57 6.05 5.37 5.22 4.92 4.43
Yes Yes Yes No No No No No Yes Yes
2.4. Data analysis 2.4.1. Scaling and normalization The initial analysis of generated Affymetrix data was performed with their MAS 5.0 software. The chip data were scaled to 500, then exported as text files to be loaded into the GeneSpring 6.0 program (Silicon Genetics, Redwood City, CA). The prescaled data were normalized by using the normalization steps “per chip to 50th percentile” and “per gene to median” as recommended by the manufacturer. 2.4.2. Determination of gene expression changes Genes having “absent” calls in all of the samples were excluded from the analysis. The remaining genes were analyzed with two methods: fold change analysis and significance analysis. Only the expression changes in the genes detected by both methods were considered significant. 2.4.3. Fold change analysis The data set consisted of 4 reference and 16 tumor samples. The overall analysis was composed of 4 crosses, because there were 4 normal samples. Each cross was made up of 16 comparisons, because there were 16 tumor samples. During each cross, the level of gene expression in the normal tissue was compared with each of the 16 tumors (therefore 16 comparison are made). Fourfold or greater expression changes in at least 15 of 16 comparisons were scored as significant expression changes for that gene in one cross. To
References
Zucchini et al., 2001 [38]; Chen et al., 2000 [39] Hough et al., 2001 [40] Maegawa et al., 2001 [41] Kawamata et al., 2003 [42]
Lasorella et al., 2001 [43] Jantscheff et al., 2003 [44]
OSCC association
No Yes No No
References
Mendez et al., 2002 [7] Gonzalez et al., 2003 [5]; Sok et al., 2003 [9] Alevizos et al., 2001 [3]; Mendez et al., 2002 [7]; Nagata et al., 2003 [8]; Sok et. al., 2003 [9]
Sok et al., 2003 [9]
No No No No No No No No No No
be in the final list based on fold change analysis, a gene should show significant expression change in 4 of 4 crosses. 2.4.4. Significance analysis The difference of expression changes between the reference and tumor samples were determined using the Significance of Analysis of Microarrays (SAM) program [11]. The input values for SAM plot calculator for ∆ and fold change parameters are 0.5 and 4, respectively. The calculated median false discovery rate for these input values was 8.2%. 2.4.5. Hierarchical clustering Two-dimensional hierarchical clustering of the genes (with Pearson correlation) showing significant expression changes (shown in Tables 2 and 3) was performed using the GeneSpring 6.0 program. 2.5. Confirmatory RT-PCR assay The expression levels were measured with real-time quantitative reverse transcriptase polymerase chain reaction (RT-PCR) using the LightCycler (Roche Diagnostics, Indianapolis, IN). The RT was performed with the QuantiTect SYBR Green RT-PCR kit (Qiagen). Reaction conditions were those recommended by the manufacturer. One microliter of RNA was used in a 20-µL final reaction volume.
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
31
Table 4 Primers used for RT-PCR Gene name
Accession no.
Primer sequence
Amplicon size
GAPDH
AF261085
149 bp
LGALS1
NM_002305
MMP1
XM_006270
LAGY
NM_139212
KRT4
X07695
F: 5′-GTGAAGGTCGGAGTCAACG-3′ R: 5′-TGGGTGGAATCATATTGGAAC-3′ F: 5′-CGAGTGCGAGGCGAGGTG-3′ R: 5′-GAAGTCACCGTCAGCTGC-3′ F: 5′-CCTTCTACCCGGAAGTTGAG-3′ R: 5′-TCCGTGTAGCACATTCTGTC-3′ F: 5′-TGCTAGCTGTCCTGCTGT-3′ R: 5′-GTTTCTGTCTTCTGGCCC-3′ F: 5′-GCAGCTAGATACCTTGGGCAA-3′ R: 5′-CTTCATACTTAGTCTTGAAGTCCTCCAC-3′
For LightCycler RT-PCR, Relative Quantification Software 1.01 (Roche, Mannheim, Germany) was used to measure relative expression of the four target genes (MMP1, LGALS1, KRT4, LAGY) in different RNA samples within individual experiments using the constitutively expressed GAPDH gene as an external standard (reference gene) (Table 4). Universal human reference RNA (Stratagene, La Jolla, CA) was used as a calibrator RNA in the assay. The correlation coefficients for RT-PCR and Affymetrix experiments were determined with Spearman’s rank order correlation.
327 bp 158 bp 456 bp 95 bp
only the 20 downregulated genes, we observed the same pattern of clustering (Fig. 3); however, the clustering pattern was significantly different when we performed the analysis just for the 33 genes showing increased level of expression (data not shown).
3. Results 3.1. Gene expression profiling Gene expression profiling of patients listed in Table 1 was determined with the Affymetrix Hu-133A arrays (22,500 genes). Raw data were scaled and normalized as already indicated. Significant gene expression changes were determined independently with two different statistical methods. Fold change analysis using GeneSpring 6.0 detected 91 genes that were at least fourfold up- or downregulated. Significance analysis using the SAM program resulted in 327 genes with significant expression changes. The 53 genes common to both methods were further analyzed (Tables 2 and 3). The genes in the common list were analyzed with a twodimensional hierarchical clustering algorithm, to observe the extent of the similarities in gene expression according to specimen type and specific genes. Clustering analysis revealed that 4 normal epithelial tissue samples, 10 tumors with T3–T4 stage (size > 4 cm), and 5 tumors with T1–T2 stage (size ⭐ 4 cm) were clustered together. The exception to this clustering accuracy was case OSCC-9; this less invasive tumor fell into the more aggressive group, along with larger tumors (Fig. 2). It was the only case in which we observed a discrepancy between the classifications based on gene expression profile and histological examination of the tumor. Although we observed a strong association between gene expression and the extent of tumor infiltration (P ⫽ 0.0014) (Table 5), we did not observe a similar association between gene expression profile and lymph node metastasis (P ⫽ 0.097). When we performed the analysis on
Fig. 2. Two-dimensional hierarchical clustering using gene expression of fourfold upregulated and fourfold downregulated genes in 20 samples (16 tumor + 4 normal). Samples with suffix T are tumors; with suffix N, normal tissues. Branches are colored accordingly: normal tissues, red; tumors less than or equal to 4 cm in dimension, green; tumors greater than 4 cm in dimension, purple.
32
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
Table 5 Distribution of the OSCC specimens according to observed (based on pathological staging) and predicted (based on clustering of 53 genes) stage of the tumor Predicted stage, no. of specimens Observed stage
T1 or T2
T3 or T4
Significance
T1 or T2 T3 or T4
5 0
1 10
P ⫽ 0.0014a
a
Calculated using Fisher’s exact test.
To determine the biological relevance of the genes identified in our study, we conducted a PubMed search on these genes (Tables 2 and 3). Of the 53 genes identified, 23 (43.4%) had previously been shown to be associated with cancers and 7 (13.2%) were reported to have an expression change in OSCC. Among the 30 genes upregulated in OSCC, 10 (30.3%) had previously been shown to be associated with cancers, including 3 (9.1%) associated with OSCC. Of the 20 genes downregulated in OSCC, 13 (65.0%) had previously been shown to be associated with cancers, including 4 (20%) associated with OSCC. 3.2. Validation with RT-PCR Two upregulated genes (LGALS1, MMP1) and two downregulated genes (LAGY, KRT4) were also analyzed with realtime quantitative RT-PCR using LightCycler in 10 samples (2 normal, 8 tumor). The expression levels of the genes detected with RT-PCR and Affymetrix arrays were highly correlated: LGALS1, r ⫽ 0.89; MMP1, r ⫽ 0.96; LAGY, r ⫽ 0.79; and KRT4, r ⫽ 0.78.
4. Discussion Setting appropriate standards for determining the significance of gene expression changes is a major challenge. We chose fourfold or greater differences in gene expression between normal and tumor samples as our basic criterion for a significant change and analyzed the data with two different methods (i.e., fold change analysis and significance analysis) independently. Only the genes identified with both of these methods were scored as significant genes. The drawback of using such stringent criteria is the loss of sensitivity. We believe, however, that obtaining high specificity (i.e., avoiding false positives) is more important than the loss of sensitivity. Experimental data for all genes both before and after the normalization in GeneSpring 6.0 will be available as supplementary data on the Internet (http://www.ncbi.nih.gov/ geo/). Under both methods, 53 genes showed a significant expression change. When we applied two-dimensional hierarchical clustering to that data set, the gene expression data were clustered according to extent of infiltration of the tumor. Normal epithelial tissues, tumors at the T1–T2 stage, and tumors at the T3–T4 were clustered together, excepting only specimen OSCC-9. Based on our data, the gene expression
Fig. 3. Two-dimensional hierarchical clustering using gene expression of only the fourfold downregulated genes in 20 samples (16 tumor ⫹ 4 normal). Samples with suffix T are tumors; with suffix N, normal tissues. Branches are colored accordingly: normal tissues, red; tumors less than or equal to 4 cm in dimension, green; tumors greater than 4 cm in dimension, purple.
profile is associated with the depth of tumor invasion rather than the infiltration of lymph nodes. We did not observe a significant association between lymph node metastasis and gene expression profile, as others have reported [8]. When we analyzed only the genes downregulated in tumor tissue (n ⫽ 20), the pattern of clustering was the same as the whole data set (n ⫽ 53). We think that this is due to differences in the strength of the detected association, which is indicated by differences in d values on Tables 2 and 3. The absolute d values are much higher for the downregulated genes than for the upregulated genes. This observation by itself points to the importance of conducting a significance analysis in addition to fold change analysis. We also believe that our observation has possible clinical ramifications, in that it potentially enables the subclassification of the tumor based on RT-PCR of 20 critical genes, rather than using the whole chip with 22,500 genes. Nonetheless, additional experiments must be conducted on a larger sample size before this can become of clinical use.
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
We were surprised to detect an association based on the extent of the infiltration of the tumor, given that the criteria for both fold change analysis and significance analysis were based on tumor versus normal comparisons, rather than intratumor comparisons such as superficial versus infiltrative tumor. It is evident that we are identifying the genes whose level of expression has a quantitative effect on the progression of tumor invasion. Six of the genes in the 53gene data set had previously been shown to be associated with OSCC: LGALS1 [12], MMP1 [3,8], SCEL [9], KRT4 [3,7–9], TGM [5,9], and MAL [7].
Table 6 List of genes with expression change, classified according to their function based on GO ontology terms Function
Gene
Expression status
Cell adhesion
CLDN17 DSG1 LGALS1 CEACAM6 SCEL SKIL SPRR3 KRT4 IGFBP7 TSPAN-2 PEG3 BF FYB G1P2 G1P3 IGKC TLR2 IGLJ3 IFI44 MMP1 APOL1 APOL2 CYP3A5P2 ALDH5A1 CA2 HPGD GPX3 ADA PAPSS2 BIGM103 UCP2 TGM3 UBD C1S P11 ID4 C3 IFITM1 MAL NRG1 SCAP2 PPP1R3C NPR3 CHK
⫺ ⫺ ⫹ ⫺ ⫺ ⫹ ⫺ ⫺ ⫹ ⫹ ⫺ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫺ ⫺ ⫹ ⫹ ⫹ ⫹ ⫺ ⫹ ⫹ ⫺ ⫺ ⫹ ⫹ ⫺ ⫹ ⫹ ⫺ ⫺ ⫹
Cell differentiation
Cell proliferation
Immune response
Invasive growth Lipid metabolism
Metabolism
Nucleotide metabolism Metal ion transport Mitochondrial transport Protein modification Proteolysis; peptidolysis Regulation of transcription Signal transduction
Abbreviations: ⫺, downregulated; ⫹, upregulated.
33
We also classified the genes according to the functional properties based on GO Ontology terms (http://www.geneontology.org) (Table 6). Most of the genes had roles in cell proliferation, cell adhesion, cell differentiation, transcription regulation, or signal transduction. These observations are not unexpected, as molecules having these types of functions are known to be critical in cancer development. We made two additional observations. The first regards the ERBB2 pathway, which has been shown to be affected in OSCC pathogenesis [13]. The NRG1 gene, which was upregulated in our specimens, encodes some alternatively spliced isoforms known as heregulins. The CHK gene, also upregulated, is a molecule that phosphorylates the c-terminal of src-kinases. After heregulin stimulation, the ERBB2 receptor physically associates with the CHK; this association allows the CHK to phosphorylate the src-kinases downstream of the pathway [14]. The other biologically relevant observation regards the invasiveness of the tumor. The stratified squamous epithelium in the oral cavity is a leakproof barrier system separating the cavity and the connective tissue under the epithelium. One reason why all these well-differentiated cells are held together as a group is that the presence of molecules that enable cells to form complexes such as tight junctions and desmosomes. During pathogenesis of the cancer, in addition to uncontrolled proliferation, the cells dedifferentiate, disaggregate, and infiltrate the inner connective tissue. Our findings of downregulation of the KRT4, CLDN17, DSG1, and SCEL genes and upregulation of the MMP1 gene are consistent with these processes. KRT4 and SCEL are well-known differentiation markers [15,16]. Their downregulation is an indication of dedifferentiation of epithelial cells. CLDN17 is one of the molecules that constitutes the tight junction [17], and desmoglin-1 encoded by DSG1 is a desmosome component [18]. It is reasonable that downregulation of these molecules is detrimental to the integrity of these complexes responsible for cell–cell attachment. The upregulation of the MMP1 gene, encoding collagenase 1, facilitates the infiltration of the tumor cells by physically destroying connective tissue [19]. In conclusion, we have detected a strong correlation between the gene expression profile in tumors and the depth of the invasion. The genes we detected have not only strong statistical significance, but also sound biological relevance.
Acknowledgments G.A. Toruner and C. Ulger were supported in part by The Tomorrow’s Children’s Fund.
References [1] Greenlee RT, Hill-Harmon MB, Murray T, Thun M. Cancer statistics, 2001. CA Cancer J Clin 2001;51:15–36.
34
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
[2] Canto MT, Devesa SS. Oral cavity and pharynx cancer incidence rates in the United States, 1975–1998. Oral Oncol 2002;38:610–7. [3] Alevizos I, Mahadevappa M, Zhang X, Ohyama H, Kohno Y, Posner M, Gallagher GT, Varvares M, Cohen D, Kim D, Kent R, Donoff RB, Todd R, Yung CM, Warrington JA, Wong DT. Oral cancer in vivo gene expression profiling assisted by laser capture microdissection and microarray analysis. Oncogene 2001;20:6196–204. [4] Belbin TJ, Singh B, Barber I, Socci N, Wenig B, Smith R, Prystowsky MB, Childs G. Molecular classification of head and neck squamous cell carcinoma using cDNA microarrays. Cancer Res 2002;62:1184– 90. [5] Gonzalez HE, Gujrati M, Frederick M, Henderson Y, Arumugam J, Spring PW, Mitsudo K, Kim HW, Clayman GL. Identification of 9 genes differentially expressed in head and neck squamous cell carcinoma. Arch Otolaryngol Head Neck Surg 2003;129:754–9. [6] Hwang D, Alevizos I, Schmitt WA, Misra J, Ohyama H, Todd R, Mahadevappa M, Warrington JA, Stephanopoulos G, Wong DT. Genomic dissection for characterization of cancerous oral epithelium tissues using transcription profiling. Oral Oncol 2003;39:259–68. [7] Mendez E, Cheng C, Farwell DG, Ricks S, Agoff SN, Futran ND, Weymuller EA Jr, Maronian NC, Zhao LP, Chen C. Transcriptional expression profiles of oral squamous cell carcinomas. Cancer 2002;95: 1482–94. [8] Nagata M, Fujita H, Ida H, Hoshina H, Inoue T, Seki Y, Ohnishi M, Ohyama T, Shingaki S, Kaji M, Saku T, Takagi R. Identification of potential biomarkers of lymph node metastasis in oral squamous cell carcinoma by cDNA microarray analysis. Int J Cancer 2003;106: 683–9. [9] Sok JC, Kuriakose MA, Mahajan VB, Pearlman AN, DeLacure MD, Chen FA. Tissue-specific gene expression of head and neck squamous cell carcinoma in vivo by complementary DNA microarray analysis. Arch Otolaryngol Head Neck Surg 2003;129:760–70. [10] Ohyama H, Zhang X, Kohno Y, Alevizos I, Posner M, Wong DT, Todd R. Laser capture microdissection-generated target sample for high-density oligonucleotide array hybridization. Biotechniques 2000; 29:530–6. [11] Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. [Erratum: Proc Natl Acad Sci U S A 2001;98:10515.] Proc Natl Acad Sci U S A 2001;98: 5116–21. [12] Choufani G, Nagy N, Saussez S, Marchant H, Bisschop P, Burchert M, Danguy A, Louryan S, Salmon I, Gabius HJ, Kiss R, Hassid S. The levels of expression of galectin-1, galectin-3, and the ThomsenFriedenreich antigen and their binding sites decrease as clinical aggressiveness increases in head and neck cancers. Cancer 1999;86:2353–63. [13] Khan AJ, King BL, Smith BD, Smith GL, DiGiovanna MP, Carter D, Haffty BG. Characterization of the HER-2/neu oncogene by immunohistochemical and fluorescence in situ hybridization analysis in oral and oropharyngeal squamous cell carcinoma. Clin Cancer Res 2002; 8:540–8. [14] Zrihan-Licht S, Deng B, Yarden Y, McShan G, Keydar I, Avraham H. Csk homologous kinase, a novel signaling molecule, directly associates with the activated ErbB-2 receptor in breast cancer cells and inhibits their proliferation. J Biol Chem 1998;273:4065–72. [15] Fuchs E. Keratins as biochemical markers of epithelial differentiation. Trends Genet 1988;4:277–81. [16] Kvedar JC, Manabe M, Phillips SB, Ross BS, Baden HP. Characterization of sciellin, a precursor to the cornified envelope of human keratinocytes. Differentiation 1992;49:195–204. [17] Brandner JM, Kief S, Grund C, Rendl M, Houdek P, Kuhn C, Tschachler E, Franke WW, Moll I. Organization and formation of the tight junction system in human epidermis and cultured keratinocytes. Eur J Cell Biol 2002;81:253–63. [18] Kowalczyk AP, Stappenbeck TS, Parry DA, Palka HL, Virata ML, Bornslaeger EA, Nilles LA, Green KJ. Structure and function of desmosomal transmembrane core and plaque molecules. Biophys Chem 1994;50:97–112.
[19] Seiki M. Membrane-type 1 matrix metalloproteinase: a key enzyme for tumor invasion. Cancer Lett 2003;194:1–11. [20] van den Brule F, Califice S, Garnier F, Fernandez PL, Berchuck A, Castronovo V. Galectin-1 accumulation in the ovary carcinoma peritumoral stroma is induced by ovary carcinoma cells and affects both cancer cell proliferation and adhesion to laminin-1 and fibronectin. Lab Invest 2003;83:377–86. [21] Hittelet A, Legendre H, Nagy N, Bronckart Y, Pector JC, Salmon I, Yeaton P, Gabius HJ, Kiss R, Camby I. Upregulation of galectins-1 and -3 in human colon cancer and their role in regulating cell migration. Int J Cancer 2003;103:370–9. [22] Lee CG, Ren J, Cheong IS, Ban KH, Ooi LL, Yong Tan S, Kan A, Nuchprayoon I, Jin R, Lee KH, Choti M, Lee LA. Expression of the FAT10 gene is highly upregulated in hepatocellular carcinoma and other gastrointestinal and gynecological cancers. Oncogene 2003;22: 2592–603. [23] Raval GN, Bharadwaj S, Levine EA, Willingham MC, Geary RL, Kute T, Prasad GL. Loss of expression of tropomyosin-1, a novel class II tumor suppressor that induces anoikis, in primary breast tumors. Oncogene 2003;22:6194–203. [24] Zhang F, Lundin M, Ristimaki A, Heikkila P, Lundin J, Isola J, Joensuu H, Laiho M. Ski-related novel protein N (SnoN), a negative controller of transforming growth factor-β signaling, is a prognostic marker in estrogen receptor-positive breast carcinomas. Cancer Res 2003;63:5005–10. [25] Watson MA, Gutmann DH, Peterson K, Chicoine MR, KleinschmidtDeMasters BK, Brown HG, Perry A. Molecular characterization of human meningiomas by gene expression profiling using high-density oligonucleotide microarrays. Am J Pathol 2002;161:665–72. [26] Landberg G, Ostlund H, Nielsen NH, Roos G, Emdin S, Burger AM, Seth A. Downregulation of the potential suppressor gene IGFBPrP1 in human breast cancer is associated with inactivation of the retinoblastoma protein, cyclin E overexpression and increased proliferation in estrogen receptor negative tumors. Oncogene 2001;20:3497– 505. [27] Huang Y, Prasad M, Lemon WJ, Hampel H, Wright FA, Kornacker K, LiVolsi V, Frankel W, Kloos RT, Eng C, Pellegata NS, de la Chapelle A. Gene expression in papillary thyroid carcinoma reveals highly consistent profiles. Proc Natl Acad Sci U S A 2001;98:15044–9. [28] Tsai MS, Shamon-Taylor LA, Mehmi I, Tang CK, Lupu R. Blockage of heregulin expression inhibits tumorigenicity and metastasis of breast cancer. Oncogene 2003;22:761–8. [29] Bougeret C, Jiang S, Keydar I, Avraham H. Functional analysis of Csk and CHK kinases in breast cancer cells. J Biol Chem 2001;276: 33711–20. [30] Chiang WL, Chu SC, Yang SS, Li MC, Lai JC, Yang SF, Chiou HL, Hsieh YS. The aberrant expression of cytosolic carbonic anhydrase and its clinical significance in human non-small cell lung cancer. Cancer Lett 2002;188:199–205. [31] Bekku S, Mochizuki H, Yamamoto T, Ueno H, Takayama E, Tadakuma T. Expression of carbonic anhydrase I or II and correlation to clinical aspects of colorectal cancer. Hepatogastroenterology 2000;47: 998–1001. [32] Xu Z, Wang MR, Xu X, Cai Y, Han YL, Wu KM, Wang J, Chen BS, Wang XQ, Wu M. Novel human esophagus-specific gene c1orf10: cDNA cloning, gene structure, and frequent loss of expression in esophageal cancer. Genomics 2000;69:322–30. [33] Mimori K, Shiraishi T, Mashino K, Sonoda H, Yamashita K, Yoshinaga K, Masuda T, Utsunomiya T, Alonso MA, Inoue H, Mori M. MAL gene expression in esophageal cancer suppresses motility, invasion and tumorigenicity and enhances apoptosis through the Fas pathway. Oncogene 2003;22:3463–71. [34] Hourihan RN, O’Sullivan GC, Morgan JG. Transcriptional gene expression profiles of oesophageal adenocarcinoma and normal oesophageal tissues. Anticancer Res 2003;23:161–5. [35] Nakashima S, Natsugoe S, Matsumoto M, Kijima F, Miyazono F, Ishigami S, Baba M, Takao S, Aikou T. Biological properties of biopsy
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
[36]
[37]
[38]
[39]
specimens are useful for predicting lymph node micrometastasis in esophageal carcinoma. Anticancer Res 2002;22:2951–6. Li G, Schaider H, Satyamoorthy K, Hanakawa Y, Hashimoto K, Herlyn M. Downregulation of E-cadherin and desmoglein 1 by autocrine hepatocyte growth factor during melanoma development. Oncogene 2001;20:8125–35. Chen Y, Petersen S, Pacyna-Gengelbach M, Pietas A, Petersen I. Identification of a novel homeobox-containing gene, LAGY, which is downregulated in lung cancer. Oncology 2003;64:450–8. Zucchini C, Biolchi A, Strippoli P, Solmi R, Rosati G, Del Governatore M, Milano E, Ugolini G, Salfi N, Farina A, Caira A, Zanotti S, Carinci P, Valvassori L. Expression profile of epidermal differentiation complex genes in normal and anal cancer cells. Int J Oncol 2001; 19:1133–41. Chen BS, Wang MR, Cai Y, Xu X, Xu ZX, Han YL, Wu M. Decreased expression of SPRR3 in Chinese human oesophageal cancer. Carcinogenesis 2000;21:2147–50.
35
[40] Hough CD, Cho KR, Zonderman AB, Schwartz DR, Morin PJ. Coordinately up-regulated genes in ovarian cancer. Cancer Res 2001;61: 3869–76. [41] Maegawa S, Yoshioka H, Itaba N, Kubota N, Nishihara S, Shirayoshi Y, Nanba E, Oshimura M. Epigenetic silencing of PEG3 gene expression in human glioma cell lines. Mol Carcinog 2001;31:1–9. [42] Kawamata H, Furihata T, Omotehara F, Sakai T, Horiuchi H, Shinagawa Y, Imura J, Ohkura Y, Tachibana M, Kubota K, Terano A, Fujimori T. Identification of genes differentially expressed in a newly isolated human metastasizing esophageal cancer cell line, T.Tn-AT1, by cDNA microarray. Cancer Sci 2003;94:699–706. [43] Lasorella A, Uo T, Iavarone A. Id proteins at the cross-road of development and cancer. Oncogene 2001;20:8326–33. [44] Jantscheff P, Terracciano L, Lowy A, Glatz-Krieger K, Grunert F, Micheel B, Brummer J, Laffer U, Metzger U, Herrmann R, Rochlitz C. Expression of CEACAM6 in resectable colorectal cancer: a factor of independent prognostic significance. J Clin Oncol 2003;21:3638–46.
Association between gene expression profile and tumor invasion in oral squamous cell carcinoma Gokce A. Torunera,b,c,1, Celal Ulgera,b,c,1, Mualla Alkana,b, Anthony T. Galanted, Joseph Rinaggioe, Randall Wilkf, Bin Tiang, Patricia Soteropoulosa,d, Meera R. Hameedh, Marvin N. Schwalba,b, James J. Dermodya,b,* a
.
Department of Microbiology and Molecular Genetics, University of Medicine and Dentistry of New Jersey (UMDNJ)–New Jersey Medical School, 185 South Orange Avenue, MSB-F659, Newark, NJ 07103 b Center for Human and Molecular Genetics, UMDNJ–New Jersey Medical School, Newark, NJ 07103 c David Jurist Research Center, Hackensack University Medical Center, Hackensack, NJ 07601 d Center for Applied Genomics, International Center for Public Health, Newark, NJ 07103 e Department of Diagnostic Sciences, UMDNJ–New Jersey Dental School, Newark, NJ 07103 f Department of Oral and Maxillofacial Surgery, UMDNJ–New Jersey Dental School, Newark, NJ 07103 g Department of Biochemistry and Molecular Biology, UMDNJ–New Jersey Medical School, Newark, NJ 07103 h Division of Surgical Pathology, Department of Pathology and Laboratory Medicine, UMDNJ–New Jersey Medical School, Newark, NJ 07103 Received 26 November 2003; received in revised form 6 January 2004; accepted 28 January 2004
Abstract
There are limited studies attempting to correlate the expression changes in oral squamous cell carcinoma with clinically relevant variables. We determined the gene expression profile of 16 tumor and 4 normal tissues from 16 patients by means of Affymetrix Hu133A GeneChips. The hybridized RNA was isolated from cells obtained with laser capture microdissection, then was amplified and labeled using T7 polymerase-based in vitro transcription. The expression of 53 genes was found to differ significantly (33 upregulated, 20 downregulated) in normal versus tumor tissues under two independent statistical methods. The expression changes in four selected genes (LGALS1, MMP1, LAGY, and KRT4) were confirmed with reverse transcriptase polymerase chain reaction. Two-dimensional hierarchical clustering of the 53 genes resulted in the samples clustering according to the extent of tumor infiltration: normal epithelial tissue, tumors less than or equal to 4 cm in dimension, and tumors more than 4 cm in dimension (P ⫽ 0.0014). The same pattern of clustering was also observed for the 20 downregulated genes. We did not observe any associations with lymph node metastasis (P ⫽ 0.097). 쑖 2004 Elsevier Inc. All rights reserved.
1. Introduction Oral squamous cell carcinoma (OSCC) is the sixth most common type of carcinoma worldwide. It has been estimated that 30,000 new cases and 7800 OSCC-related deaths occur in the United States annually [1]. The incidence rates are higher in males and blacks. Tobacco products and alcohol are major etiological factors, and the rate of consumption of these products is believed to partially account for the difference in incidence rate according to race and sex [2]. Because OSCC is a major health problem, understanding the molecular basis of this neoplasm is of great importance.
1 These authors contributed equally to this work. * Corresponding author. Tel.: (973) 972-5567; fax: (973) 972-3783. E-mail address: [email protected] (J.J. Dermody).
0165-4608/04/$ – see front matter 쑖 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.cancergencyto.2004.01.026
In the past, the lack of fine extraction techniques of tumor cells from the specimen and the absence of high-throughput technologies for analysis hindered efforts to determine the exact nature of the molecular changes. Technological breakthroughs, however, including laser capture microdissection (LCM) and microarray technology, have fundamentally altered cancer research. LCM enables the isolation of a pure population of cancer or normal cells from the tissue, and microarrays make the simultaneous analysis of thousands of genes possible. Although OSCC is a common cancer, there are few microarray-based studies conducted on these tumors [3–9]. Although all of these previous studies provided insights into gene expression changes in these tumors, only two studies demonstrated an association between gene expression changes and a clinically relevant variable such as patient survival [4] or lymph node metastasis [8].
28
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
Table 1 Characteristics of the OSCC patients, with TNM staging
2. Materials and methods
Patient code
Primary site
Age
Sex
TNM
2.1. Tissue collection
OSCC-2 OSCC-4 OSCC-5 OSCC-6 OSCC-7 OSCC-8 OSCC-9 OSCC-10 OSCC-11 OSCC-12 OSCC-13 OSCC-14 OSCC-15 OSCC-16 OSCC-17 OSCC-18
Floor of mouth Floor of mouth Tongue Vestibule Floor of mouth Tongue Tongue Tongue Tongue Maxilla Floor of mouth Floor of mouth Floor of mouth Mandible Tongue Floor of mouth
58 60 64 72 48 77 66 64 32 46 54 60 51 50 47 52
F M M M M M M M M M M F M F M M
T2N0M0 T4N1M0 T1N2bM0 T2N0M0 T4N1M0 T2N0M0 T2N0M0 T4N2bM0 T3N2bM0 T4N0M0 T4N0M0 T4N2bM0 T2N0M0 T4N0M0 T4N3M0 T4N2M0
Samples were derived from surgically resected tissue from 16 patients with OSCC (Table 1). Informed consent was obtained from patients according to the regulations of the University of Medicine and Dentistry of New Jersey. Tissue samples from 16 tumor and 4 normal tissues of patients were immediately frozen in liquid nitrogen after surgery. Paired control nontumor tissues from patients OSCC-2, OSCC-9, OSCC-14, and OSCC-18 were obtained from a clinically unaffected site and were histologically normal. Frozen tissue was placed in a cryomold with OCT embedding medium on dry ice for 1 minute. The cryomold was filled with embedding medium and covered with dry ice placed at ⫺80⬚C until the cutting period (http://dir.nichd.nih. gov/lcm/LCMTAP.htm). Embedded frozen tissue specimens were cut in a series of 7-µm-thick sections onto plain uncoated glass slides at ⫺20⬚C. The sections were stained with fast hematoxylin–eosin staining in RNase-free recipients and solutions (http://dir.nichd.nih.gov/lcm/LCMTAP.htm#Staining). After staining the tissue, tumor and normal tissues were diagnosed by the pathologist, then LCM was performed. The tumors were staged according to 2003 cancer staging guidelines issued by the American Joint Committee on Cancer. About 120,000 pure tumor and normal epithelial cells were harvested from cryosections using PixCell II
Abbreviation: TNM, tumor–node–metastasis.
We performed microarray analysis using RNA from OSCC cells obtained with LCM from 16 patients and compared these results with 4 control squamous cell epithelium samples. We identified expression profiles of differentially expressed genes between normal and tumor tissues and found that tumor invasiveness of a tumor is highly correlated with its gene expression profile.
Fig. 1. LCM from hematoxylin–eosin stained OSCC cryosection. (A) Section before microdissection and (B) after removal of water and prior to LCM. (C) Tumor island lifted from the surrounding tissue and (D) LCM-captured tumor island.
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
29
Table 2 The upregulated genes in OSCC Gene
GenBank accession no.
Score (d)
Fold change
Cancer association
C3 LGALS1
NM_000064 NM_002305
1.8 1.43
62.42 31.76
No Yes
MMP1
XM_006270
1.50
30.19
Yes
IGLJ3 UBD IGKC G1P2 TPM1 IFI44 KIAA0746 DZIP1 C1orf29 FYB C1S UCP2 G1P3 BF SCAP2 SKIL TLR2 ADA PAPSS2 IGFBP7
M87790.1 NM_006398 M63438.1 NM_005101 Z24727.1 BE049439 AB018289 NM_014934 NM_006820 AF198052.1 M18767.1 U94592.1 NM_022873 NM_001710 NM_003930 BF725121 NM_003264 X02189 AW299958 NM_001553
1.59 1.74 1.87 1.57 1.62 2.30 2.79 1.68 1.89 1.61 1.94 1.65 1.98 2.12 2.38 2.5 1.93 1.65 1.5 1.93
19.92 17.36 16.80 14.11 13.77 11.99 10.32 9.78 9.54 8.00 7.91 7.65 6.43 6.36 6.31 6.14 6.03 5.43 5.4 5.36
No Yes No No Yes No No No No No No No No No No Yes No No No Yes
BIGM103 FLJ22344 IFITM1 NRG1 CHK ALDH5A1 APOL1 CA2
AB040120.1 NM_02471 NM_003641 NM_013959 NM_001277 AL031230 AF323540.1 M36532.1
1.94 1.80 1.88 1.52 1.86 1.97 1.46 1.56
5.22 5.16 5.15 4.77 4.47 4.37 4.36 4.13
No No Yes Yes Yes No No Yes
TSPAN-2 APOL2
BF129969 BC004395.1
2.00 1.68
4.08 4.00
No No
References van den Brule et al., 2003 [20]; Hittelet et al., 2003 [21] Seiki et al., 2003 [19]
Lee et al., 2003 [22]
Raval et al., 2003 [23]
Zhang et al., 2003 [24]
Watson et al., 2002 [25]; Landberg et al., 2001 [26]
Huang et al., 2001 [27] Tsai et al., 2003 [28] Bougeret et al., 2001 [29]
Chiang et al., 2002 [30]; Bekku et. al., 2000 [31]
OSCC association
References
No Yes
Choufani et al., 1999 [12]
Yes
Alevizos et al., 2001 [3]; Nagata et al., 2003 [8]
No No No No No No No No No No No No No No No No No No No No No No No No No No No No No No
The d score is based on the ratio of gene expression change to standard deviation in the data for that gene.
LCM system (Arcturus Engineering, Mountain View, CA) (Fig. 1) [10].
performed to produce biotin-labeled cRNA from the doublestranded cDNA using the Enzo BioArray high yield RNA transcript labeling system (Enzo, Farmingdale, NY).
2.2. Total RNA isolation and target sample generation LCM-isolated tissues were homogenized and total RNA was extracted using the RNeasy kit (Qiagen, Valencia, CA) according to the manufacturer’s procedure (http://www1. qiagen.com/literature/qiagennews/0401/1018052_QNews42001 _p3_5.pdf). The Affymetrix (Santa Clara, CA) small-sample target labeling protocol was used for amplifying and labeling RNA (http://www.cgrb.orst.edu/CSL/downloads/smallsample.pdf). The SuperScript Choice system (Invitrogen Life Technologies, Rockville, MD) was used to synthesize double-stranded cDNA from LCM-derived isolated RNA. After collecting the cDNA, the Ambion MEGAscript T7 kit (Ambion, Austin, TX) was used for one round of linear amplification. A second in vitro transcription reaction was
2.3. Hybridization cRNA (15 µg) was fragmented and added to a hybridization mixture. Expression profiles were created using the HGU133A chip (Affymetrix), which contains ~22,280 known human transcripts and ESTs. These transcripts are represented using probe pairs consisting of 25-mer oligonucleotides. Hybridization was performed overnight at 45⬚C for 16 hours using the GeneChip Hybridization Oven 640 (Affymetrix). Washing and staining (streptavidin–phycoerythrin) was done using EukGE-WS2v3 protocol in the GeneChip Fluidics Station 400 (Affymetrix). Images were acquired using the Affymetrix GeneArray scanner.
30
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
Table 3 List of the downregulated genes in OSCC Gene
GenBank accession no.
Score (d)
Fold change
Cancer association
P11 C1orf10 MAL TGM3
NM_006025 NM_016190 NM_002371 NM_003245
⫺5.89 ⫺7.01 ⫺7.43 ⫺5.26
61.49 53.12 45.99 26.84
No Yes Yes No
Xu et al., 2000 [32] Mimori et al., 2003 [33]
No No Yes Yes
KRT4
X07695.1
⫺5.47
17.08
Yes
Hourihan et al., 2003 [34]
Yes
DSG1
NM_001942
⫺5.33
16.11
Yes
No
LAGY SCEL NICE-1 SPRR3
AB059408.1 NM_003843 NM_019060 NM_005416
⫺6.34 ⫺4.87 ⫺5.36 ⫺4.85
13.71 11.83 11.12 10.10
Yes Yes No Yes
Nakashima et al., 2002 [35]; Lip et. al., 2001 [36] Chen et al., 2003 [37] Huang et al., 2001 [27]
GPX3 PEG3 HPGD FLJ22671 NPR3 CLDN17 CYP3A5P2 PPP1R3C ID4 CEACAM6
AW149846 AL042588 AL574184 NM_02486 AI628360 NM_012131 X90579.1 N26005 NM_001546.1 M18728.1
⫺6.89 ⫺4.00 ⫺5.14 ⫺7.58 ⫺6.24 ⫺6.32 ⫺4.24 ⫺4.82 ⫺3.66 ⫺3.83
8.71 8.58 7.75 6.93 6.57 6.05 5.37 5.22 4.92 4.43
Yes Yes Yes No No No No No Yes Yes
2.4. Data analysis 2.4.1. Scaling and normalization The initial analysis of generated Affymetrix data was performed with their MAS 5.0 software. The chip data were scaled to 500, then exported as text files to be loaded into the GeneSpring 6.0 program (Silicon Genetics, Redwood City, CA). The prescaled data were normalized by using the normalization steps “per chip to 50th percentile” and “per gene to median” as recommended by the manufacturer. 2.4.2. Determination of gene expression changes Genes having “absent” calls in all of the samples were excluded from the analysis. The remaining genes were analyzed with two methods: fold change analysis and significance analysis. Only the expression changes in the genes detected by both methods were considered significant. 2.4.3. Fold change analysis The data set consisted of 4 reference and 16 tumor samples. The overall analysis was composed of 4 crosses, because there were 4 normal samples. Each cross was made up of 16 comparisons, because there were 16 tumor samples. During each cross, the level of gene expression in the normal tissue was compared with each of the 16 tumors (therefore 16 comparison are made). Fourfold or greater expression changes in at least 15 of 16 comparisons were scored as significant expression changes for that gene in one cross. To
References
Zucchini et al., 2001 [38]; Chen et al., 2000 [39] Hough et al., 2001 [40] Maegawa et al., 2001 [41] Kawamata et al., 2003 [42]
Lasorella et al., 2001 [43] Jantscheff et al., 2003 [44]
OSCC association
No Yes No No
References
Mendez et al., 2002 [7] Gonzalez et al., 2003 [5]; Sok et al., 2003 [9] Alevizos et al., 2001 [3]; Mendez et al., 2002 [7]; Nagata et al., 2003 [8]; Sok et. al., 2003 [9]
Sok et al., 2003 [9]
No No No No No No No No No No
be in the final list based on fold change analysis, a gene should show significant expression change in 4 of 4 crosses. 2.4.4. Significance analysis The difference of expression changes between the reference and tumor samples were determined using the Significance of Analysis of Microarrays (SAM) program [11]. The input values for SAM plot calculator for ∆ and fold change parameters are 0.5 and 4, respectively. The calculated median false discovery rate for these input values was 8.2%. 2.4.5. Hierarchical clustering Two-dimensional hierarchical clustering of the genes (with Pearson correlation) showing significant expression changes (shown in Tables 2 and 3) was performed using the GeneSpring 6.0 program. 2.5. Confirmatory RT-PCR assay The expression levels were measured with real-time quantitative reverse transcriptase polymerase chain reaction (RT-PCR) using the LightCycler (Roche Diagnostics, Indianapolis, IN). The RT was performed with the QuantiTect SYBR Green RT-PCR kit (Qiagen). Reaction conditions were those recommended by the manufacturer. One microliter of RNA was used in a 20-µL final reaction volume.
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
31
Table 4 Primers used for RT-PCR Gene name
Accession no.
Primer sequence
Amplicon size
GAPDH
AF261085
149 bp
LGALS1
NM_002305
MMP1
XM_006270
LAGY
NM_139212
KRT4
X07695
F: 5′-GTGAAGGTCGGAGTCAACG-3′ R: 5′-TGGGTGGAATCATATTGGAAC-3′ F: 5′-CGAGTGCGAGGCGAGGTG-3′ R: 5′-GAAGTCACCGTCAGCTGC-3′ F: 5′-CCTTCTACCCGGAAGTTGAG-3′ R: 5′-TCCGTGTAGCACATTCTGTC-3′ F: 5′-TGCTAGCTGTCCTGCTGT-3′ R: 5′-GTTTCTGTCTTCTGGCCC-3′ F: 5′-GCAGCTAGATACCTTGGGCAA-3′ R: 5′-CTTCATACTTAGTCTTGAAGTCCTCCAC-3′
For LightCycler RT-PCR, Relative Quantification Software 1.01 (Roche, Mannheim, Germany) was used to measure relative expression of the four target genes (MMP1, LGALS1, KRT4, LAGY) in different RNA samples within individual experiments using the constitutively expressed GAPDH gene as an external standard (reference gene) (Table 4). Universal human reference RNA (Stratagene, La Jolla, CA) was used as a calibrator RNA in the assay. The correlation coefficients for RT-PCR and Affymetrix experiments were determined with Spearman’s rank order correlation.
327 bp 158 bp 456 bp 95 bp
only the 20 downregulated genes, we observed the same pattern of clustering (Fig. 3); however, the clustering pattern was significantly different when we performed the analysis just for the 33 genes showing increased level of expression (data not shown).
3. Results 3.1. Gene expression profiling Gene expression profiling of patients listed in Table 1 was determined with the Affymetrix Hu-133A arrays (22,500 genes). Raw data were scaled and normalized as already indicated. Significant gene expression changes were determined independently with two different statistical methods. Fold change analysis using GeneSpring 6.0 detected 91 genes that were at least fourfold up- or downregulated. Significance analysis using the SAM program resulted in 327 genes with significant expression changes. The 53 genes common to both methods were further analyzed (Tables 2 and 3). The genes in the common list were analyzed with a twodimensional hierarchical clustering algorithm, to observe the extent of the similarities in gene expression according to specimen type and specific genes. Clustering analysis revealed that 4 normal epithelial tissue samples, 10 tumors with T3–T4 stage (size > 4 cm), and 5 tumors with T1–T2 stage (size ⭐ 4 cm) were clustered together. The exception to this clustering accuracy was case OSCC-9; this less invasive tumor fell into the more aggressive group, along with larger tumors (Fig. 2). It was the only case in which we observed a discrepancy between the classifications based on gene expression profile and histological examination of the tumor. Although we observed a strong association between gene expression and the extent of tumor infiltration (P ⫽ 0.0014) (Table 5), we did not observe a similar association between gene expression profile and lymph node metastasis (P ⫽ 0.097). When we performed the analysis on
Fig. 2. Two-dimensional hierarchical clustering using gene expression of fourfold upregulated and fourfold downregulated genes in 20 samples (16 tumor + 4 normal). Samples with suffix T are tumors; with suffix N, normal tissues. Branches are colored accordingly: normal tissues, red; tumors less than or equal to 4 cm in dimension, green; tumors greater than 4 cm in dimension, purple.
32
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
Table 5 Distribution of the OSCC specimens according to observed (based on pathological staging) and predicted (based on clustering of 53 genes) stage of the tumor Predicted stage, no. of specimens Observed stage
T1 or T2
T3 or T4
Significance
T1 or T2 T3 or T4
5 0
1 10
P ⫽ 0.0014a
a
Calculated using Fisher’s exact test.
To determine the biological relevance of the genes identified in our study, we conducted a PubMed search on these genes (Tables 2 and 3). Of the 53 genes identified, 23 (43.4%) had previously been shown to be associated with cancers and 7 (13.2%) were reported to have an expression change in OSCC. Among the 30 genes upregulated in OSCC, 10 (30.3%) had previously been shown to be associated with cancers, including 3 (9.1%) associated with OSCC. Of the 20 genes downregulated in OSCC, 13 (65.0%) had previously been shown to be associated with cancers, including 4 (20%) associated with OSCC. 3.2. Validation with RT-PCR Two upregulated genes (LGALS1, MMP1) and two downregulated genes (LAGY, KRT4) were also analyzed with realtime quantitative RT-PCR using LightCycler in 10 samples (2 normal, 8 tumor). The expression levels of the genes detected with RT-PCR and Affymetrix arrays were highly correlated: LGALS1, r ⫽ 0.89; MMP1, r ⫽ 0.96; LAGY, r ⫽ 0.79; and KRT4, r ⫽ 0.78.
4. Discussion Setting appropriate standards for determining the significance of gene expression changes is a major challenge. We chose fourfold or greater differences in gene expression between normal and tumor samples as our basic criterion for a significant change and analyzed the data with two different methods (i.e., fold change analysis and significance analysis) independently. Only the genes identified with both of these methods were scored as significant genes. The drawback of using such stringent criteria is the loss of sensitivity. We believe, however, that obtaining high specificity (i.e., avoiding false positives) is more important than the loss of sensitivity. Experimental data for all genes both before and after the normalization in GeneSpring 6.0 will be available as supplementary data on the Internet (http://www.ncbi.nih.gov/ geo/). Under both methods, 53 genes showed a significant expression change. When we applied two-dimensional hierarchical clustering to that data set, the gene expression data were clustered according to extent of infiltration of the tumor. Normal epithelial tissues, tumors at the T1–T2 stage, and tumors at the T3–T4 were clustered together, excepting only specimen OSCC-9. Based on our data, the gene expression
Fig. 3. Two-dimensional hierarchical clustering using gene expression of only the fourfold downregulated genes in 20 samples (16 tumor ⫹ 4 normal). Samples with suffix T are tumors; with suffix N, normal tissues. Branches are colored accordingly: normal tissues, red; tumors less than or equal to 4 cm in dimension, green; tumors greater than 4 cm in dimension, purple.
profile is associated with the depth of tumor invasion rather than the infiltration of lymph nodes. We did not observe a significant association between lymph node metastasis and gene expression profile, as others have reported [8]. When we analyzed only the genes downregulated in tumor tissue (n ⫽ 20), the pattern of clustering was the same as the whole data set (n ⫽ 53). We think that this is due to differences in the strength of the detected association, which is indicated by differences in d values on Tables 2 and 3. The absolute d values are much higher for the downregulated genes than for the upregulated genes. This observation by itself points to the importance of conducting a significance analysis in addition to fold change analysis. We also believe that our observation has possible clinical ramifications, in that it potentially enables the subclassification of the tumor based on RT-PCR of 20 critical genes, rather than using the whole chip with 22,500 genes. Nonetheless, additional experiments must be conducted on a larger sample size before this can become of clinical use.
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
We were surprised to detect an association based on the extent of the infiltration of the tumor, given that the criteria for both fold change analysis and significance analysis were based on tumor versus normal comparisons, rather than intratumor comparisons such as superficial versus infiltrative tumor. It is evident that we are identifying the genes whose level of expression has a quantitative effect on the progression of tumor invasion. Six of the genes in the 53gene data set had previously been shown to be associated with OSCC: LGALS1 [12], MMP1 [3,8], SCEL [9], KRT4 [3,7–9], TGM [5,9], and MAL [7].
Table 6 List of genes with expression change, classified according to their function based on GO ontology terms Function
Gene
Expression status
Cell adhesion
CLDN17 DSG1 LGALS1 CEACAM6 SCEL SKIL SPRR3 KRT4 IGFBP7 TSPAN-2 PEG3 BF FYB G1P2 G1P3 IGKC TLR2 IGLJ3 IFI44 MMP1 APOL1 APOL2 CYP3A5P2 ALDH5A1 CA2 HPGD GPX3 ADA PAPSS2 BIGM103 UCP2 TGM3 UBD C1S P11 ID4 C3 IFITM1 MAL NRG1 SCAP2 PPP1R3C NPR3 CHK
⫺ ⫺ ⫹ ⫺ ⫺ ⫹ ⫺ ⫺ ⫹ ⫹ ⫺ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫹ ⫺ ⫺ ⫹ ⫹ ⫹ ⫹ ⫺ ⫹ ⫹ ⫺ ⫺ ⫹ ⫹ ⫺ ⫹ ⫹ ⫺ ⫺ ⫹
Cell differentiation
Cell proliferation
Immune response
Invasive growth Lipid metabolism
Metabolism
Nucleotide metabolism Metal ion transport Mitochondrial transport Protein modification Proteolysis; peptidolysis Regulation of transcription Signal transduction
Abbreviations: ⫺, downregulated; ⫹, upregulated.
33
We also classified the genes according to the functional properties based on GO Ontology terms (http://www.geneontology.org) (Table 6). Most of the genes had roles in cell proliferation, cell adhesion, cell differentiation, transcription regulation, or signal transduction. These observations are not unexpected, as molecules having these types of functions are known to be critical in cancer development. We made two additional observations. The first regards the ERBB2 pathway, which has been shown to be affected in OSCC pathogenesis [13]. The NRG1 gene, which was upregulated in our specimens, encodes some alternatively spliced isoforms known as heregulins. The CHK gene, also upregulated, is a molecule that phosphorylates the c-terminal of src-kinases. After heregulin stimulation, the ERBB2 receptor physically associates with the CHK; this association allows the CHK to phosphorylate the src-kinases downstream of the pathway [14]. The other biologically relevant observation regards the invasiveness of the tumor. The stratified squamous epithelium in the oral cavity is a leakproof barrier system separating the cavity and the connective tissue under the epithelium. One reason why all these well-differentiated cells are held together as a group is that the presence of molecules that enable cells to form complexes such as tight junctions and desmosomes. During pathogenesis of the cancer, in addition to uncontrolled proliferation, the cells dedifferentiate, disaggregate, and infiltrate the inner connective tissue. Our findings of downregulation of the KRT4, CLDN17, DSG1, and SCEL genes and upregulation of the MMP1 gene are consistent with these processes. KRT4 and SCEL are well-known differentiation markers [15,16]. Their downregulation is an indication of dedifferentiation of epithelial cells. CLDN17 is one of the molecules that constitutes the tight junction [17], and desmoglin-1 encoded by DSG1 is a desmosome component [18]. It is reasonable that downregulation of these molecules is detrimental to the integrity of these complexes responsible for cell–cell attachment. The upregulation of the MMP1 gene, encoding collagenase 1, facilitates the infiltration of the tumor cells by physically destroying connective tissue [19]. In conclusion, we have detected a strong correlation between the gene expression profile in tumors and the depth of the invasion. The genes we detected have not only strong statistical significance, but also sound biological relevance.
Acknowledgments G.A. Toruner and C. Ulger were supported in part by The Tomorrow’s Children’s Fund.
References [1] Greenlee RT, Hill-Harmon MB, Murray T, Thun M. Cancer statistics, 2001. CA Cancer J Clin 2001;51:15–36.
34
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
[2] Canto MT, Devesa SS. Oral cavity and pharynx cancer incidence rates in the United States, 1975–1998. Oral Oncol 2002;38:610–7. [3] Alevizos I, Mahadevappa M, Zhang X, Ohyama H, Kohno Y, Posner M, Gallagher GT, Varvares M, Cohen D, Kim D, Kent R, Donoff RB, Todd R, Yung CM, Warrington JA, Wong DT. Oral cancer in vivo gene expression profiling assisted by laser capture microdissection and microarray analysis. Oncogene 2001;20:6196–204. [4] Belbin TJ, Singh B, Barber I, Socci N, Wenig B, Smith R, Prystowsky MB, Childs G. Molecular classification of head and neck squamous cell carcinoma using cDNA microarrays. Cancer Res 2002;62:1184– 90. [5] Gonzalez HE, Gujrati M, Frederick M, Henderson Y, Arumugam J, Spring PW, Mitsudo K, Kim HW, Clayman GL. Identification of 9 genes differentially expressed in head and neck squamous cell carcinoma. Arch Otolaryngol Head Neck Surg 2003;129:754–9. [6] Hwang D, Alevizos I, Schmitt WA, Misra J, Ohyama H, Todd R, Mahadevappa M, Warrington JA, Stephanopoulos G, Wong DT. Genomic dissection for characterization of cancerous oral epithelium tissues using transcription profiling. Oral Oncol 2003;39:259–68. [7] Mendez E, Cheng C, Farwell DG, Ricks S, Agoff SN, Futran ND, Weymuller EA Jr, Maronian NC, Zhao LP, Chen C. Transcriptional expression profiles of oral squamous cell carcinomas. Cancer 2002;95: 1482–94. [8] Nagata M, Fujita H, Ida H, Hoshina H, Inoue T, Seki Y, Ohnishi M, Ohyama T, Shingaki S, Kaji M, Saku T, Takagi R. Identification of potential biomarkers of lymph node metastasis in oral squamous cell carcinoma by cDNA microarray analysis. Int J Cancer 2003;106: 683–9. [9] Sok JC, Kuriakose MA, Mahajan VB, Pearlman AN, DeLacure MD, Chen FA. Tissue-specific gene expression of head and neck squamous cell carcinoma in vivo by complementary DNA microarray analysis. Arch Otolaryngol Head Neck Surg 2003;129:760–70. [10] Ohyama H, Zhang X, Kohno Y, Alevizos I, Posner M, Wong DT, Todd R. Laser capture microdissection-generated target sample for high-density oligonucleotide array hybridization. Biotechniques 2000; 29:530–6. [11] Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. [Erratum: Proc Natl Acad Sci U S A 2001;98:10515.] Proc Natl Acad Sci U S A 2001;98: 5116–21. [12] Choufani G, Nagy N, Saussez S, Marchant H, Bisschop P, Burchert M, Danguy A, Louryan S, Salmon I, Gabius HJ, Kiss R, Hassid S. The levels of expression of galectin-1, galectin-3, and the ThomsenFriedenreich antigen and their binding sites decrease as clinical aggressiveness increases in head and neck cancers. Cancer 1999;86:2353–63. [13] Khan AJ, King BL, Smith BD, Smith GL, DiGiovanna MP, Carter D, Haffty BG. Characterization of the HER-2/neu oncogene by immunohistochemical and fluorescence in situ hybridization analysis in oral and oropharyngeal squamous cell carcinoma. Clin Cancer Res 2002; 8:540–8. [14] Zrihan-Licht S, Deng B, Yarden Y, McShan G, Keydar I, Avraham H. Csk homologous kinase, a novel signaling molecule, directly associates with the activated ErbB-2 receptor in breast cancer cells and inhibits their proliferation. J Biol Chem 1998;273:4065–72. [15] Fuchs E. Keratins as biochemical markers of epithelial differentiation. Trends Genet 1988;4:277–81. [16] Kvedar JC, Manabe M, Phillips SB, Ross BS, Baden HP. Characterization of sciellin, a precursor to the cornified envelope of human keratinocytes. Differentiation 1992;49:195–204. [17] Brandner JM, Kief S, Grund C, Rendl M, Houdek P, Kuhn C, Tschachler E, Franke WW, Moll I. Organization and formation of the tight junction system in human epidermis and cultured keratinocytes. Eur J Cell Biol 2002;81:253–63. [18] Kowalczyk AP, Stappenbeck TS, Parry DA, Palka HL, Virata ML, Bornslaeger EA, Nilles LA, Green KJ. Structure and function of desmosomal transmembrane core and plaque molecules. Biophys Chem 1994;50:97–112.
[19] Seiki M. Membrane-type 1 matrix metalloproteinase: a key enzyme for tumor invasion. Cancer Lett 2003;194:1–11. [20] van den Brule F, Califice S, Garnier F, Fernandez PL, Berchuck A, Castronovo V. Galectin-1 accumulation in the ovary carcinoma peritumoral stroma is induced by ovary carcinoma cells and affects both cancer cell proliferation and adhesion to laminin-1 and fibronectin. Lab Invest 2003;83:377–86. [21] Hittelet A, Legendre H, Nagy N, Bronckart Y, Pector JC, Salmon I, Yeaton P, Gabius HJ, Kiss R, Camby I. Upregulation of galectins-1 and -3 in human colon cancer and their role in regulating cell migration. Int J Cancer 2003;103:370–9. [22] Lee CG, Ren J, Cheong IS, Ban KH, Ooi LL, Yong Tan S, Kan A, Nuchprayoon I, Jin R, Lee KH, Choti M, Lee LA. Expression of the FAT10 gene is highly upregulated in hepatocellular carcinoma and other gastrointestinal and gynecological cancers. Oncogene 2003;22: 2592–603. [23] Raval GN, Bharadwaj S, Levine EA, Willingham MC, Geary RL, Kute T, Prasad GL. Loss of expression of tropomyosin-1, a novel class II tumor suppressor that induces anoikis, in primary breast tumors. Oncogene 2003;22:6194–203. [24] Zhang F, Lundin M, Ristimaki A, Heikkila P, Lundin J, Isola J, Joensuu H, Laiho M. Ski-related novel protein N (SnoN), a negative controller of transforming growth factor-β signaling, is a prognostic marker in estrogen receptor-positive breast carcinomas. Cancer Res 2003;63:5005–10. [25] Watson MA, Gutmann DH, Peterson K, Chicoine MR, KleinschmidtDeMasters BK, Brown HG, Perry A. Molecular characterization of human meningiomas by gene expression profiling using high-density oligonucleotide microarrays. Am J Pathol 2002;161:665–72. [26] Landberg G, Ostlund H, Nielsen NH, Roos G, Emdin S, Burger AM, Seth A. Downregulation of the potential suppressor gene IGFBPrP1 in human breast cancer is associated with inactivation of the retinoblastoma protein, cyclin E overexpression and increased proliferation in estrogen receptor negative tumors. Oncogene 2001;20:3497– 505. [27] Huang Y, Prasad M, Lemon WJ, Hampel H, Wright FA, Kornacker K, LiVolsi V, Frankel W, Kloos RT, Eng C, Pellegata NS, de la Chapelle A. Gene expression in papillary thyroid carcinoma reveals highly consistent profiles. Proc Natl Acad Sci U S A 2001;98:15044–9. [28] Tsai MS, Shamon-Taylor LA, Mehmi I, Tang CK, Lupu R. Blockage of heregulin expression inhibits tumorigenicity and metastasis of breast cancer. Oncogene 2003;22:761–8. [29] Bougeret C, Jiang S, Keydar I, Avraham H. Functional analysis of Csk and CHK kinases in breast cancer cells. J Biol Chem 2001;276: 33711–20. [30] Chiang WL, Chu SC, Yang SS, Li MC, Lai JC, Yang SF, Chiou HL, Hsieh YS. The aberrant expression of cytosolic carbonic anhydrase and its clinical significance in human non-small cell lung cancer. Cancer Lett 2002;188:199–205. [31] Bekku S, Mochizuki H, Yamamoto T, Ueno H, Takayama E, Tadakuma T. Expression of carbonic anhydrase I or II and correlation to clinical aspects of colorectal cancer. Hepatogastroenterology 2000;47: 998–1001. [32] Xu Z, Wang MR, Xu X, Cai Y, Han YL, Wu KM, Wang J, Chen BS, Wang XQ, Wu M. Novel human esophagus-specific gene c1orf10: cDNA cloning, gene structure, and frequent loss of expression in esophageal cancer. Genomics 2000;69:322–30. [33] Mimori K, Shiraishi T, Mashino K, Sonoda H, Yamashita K, Yoshinaga K, Masuda T, Utsunomiya T, Alonso MA, Inoue H, Mori M. MAL gene expression in esophageal cancer suppresses motility, invasion and tumorigenicity and enhances apoptosis through the Fas pathway. Oncogene 2003;22:3463–71. [34] Hourihan RN, O’Sullivan GC, Morgan JG. Transcriptional gene expression profiles of oesophageal adenocarcinoma and normal oesophageal tissues. Anticancer Res 2003;23:161–5. [35] Nakashima S, Natsugoe S, Matsumoto M, Kijima F, Miyazono F, Ishigami S, Baba M, Takao S, Aikou T. Biological properties of biopsy
G.A. Toruner et al. / Cancer Genetics and Cytogenetics 154 (2004) 27–35
[36]
[37]
[38]
[39]
specimens are useful for predicting lymph node micrometastasis in esophageal carcinoma. Anticancer Res 2002;22:2951–6. Li G, Schaider H, Satyamoorthy K, Hanakawa Y, Hashimoto K, Herlyn M. Downregulation of E-cadherin and desmoglein 1 by autocrine hepatocyte growth factor during melanoma development. Oncogene 2001;20:8125–35. Chen Y, Petersen S, Pacyna-Gengelbach M, Pietas A, Petersen I. Identification of a novel homeobox-containing gene, LAGY, which is downregulated in lung cancer. Oncology 2003;64:450–8. Zucchini C, Biolchi A, Strippoli P, Solmi R, Rosati G, Del Governatore M, Milano E, Ugolini G, Salfi N, Farina A, Caira A, Zanotti S, Carinci P, Valvassori L. Expression profile of epidermal differentiation complex genes in normal and anal cancer cells. Int J Oncol 2001; 19:1133–41. Chen BS, Wang MR, Cai Y, Xu X, Xu ZX, Han YL, Wu M. Decreased expression of SPRR3 in Chinese human oesophageal cancer. Carcinogenesis 2000;21:2147–50.
35
[40] Hough CD, Cho KR, Zonderman AB, Schwartz DR, Morin PJ. Coordinately up-regulated genes in ovarian cancer. Cancer Res 2001;61: 3869–76. [41] Maegawa S, Yoshioka H, Itaba N, Kubota N, Nishihara S, Shirayoshi Y, Nanba E, Oshimura M. Epigenetic silencing of PEG3 gene expression in human glioma cell lines. Mol Carcinog 2001;31:1–9. [42] Kawamata H, Furihata T, Omotehara F, Sakai T, Horiuchi H, Shinagawa Y, Imura J, Ohkura Y, Tachibana M, Kubota K, Terano A, Fujimori T. Identification of genes differentially expressed in a newly isolated human metastasizing esophageal cancer cell line, T.Tn-AT1, by cDNA microarray. Cancer Sci 2003;94:699–706. [43] Lasorella A, Uo T, Iavarone A. Id proteins at the cross-road of development and cancer. Oncogene 2001;20:8326–33. [44] Jantscheff P, Terracciano L, Lowy A, Glatz-Krieger K, Grunert F, Micheel B, Brummer J, Laffer U, Metzger U, Herrmann R, Rochlitz C. Expression of CEACAM6 in resectable colorectal cancer: a factor of independent prognostic significance. J Clin Oncol 2003;21:3638–46.