Fax: (713) 745-1163
Gene-expression profiling in Chinese patients with colon cancer by coupling experimental and bioinformatic genomewide gene-expression analyses
Identification and validation of IFITM3 as a biomarker of early colon carcinogenesis
Article first published online: 9 MAY 2008
Copyright © 2008 American Cancer Society
Volume 113, Issue 2, pages 266–275, 15 July 2008
How to Cite
Fan, J., Peng, Z., Zhou, C., Qiu, G., Tang, H., Sun, Y., Wang, X., Li, Q., Le, X. and Xie, K. (2008), Gene-expression profiling in Chinese patients with colon cancer by coupling experimental and bioinformatic genomewide gene-expression analyses. Cancer, 113: 266–275. doi: 10.1002/cncr.23551
- Issue published online: 8 JUL 2008
- Article first published online: 9 MAY 2008
- Manuscript Accepted: 3 MAR 2008
- Manuscript Revised: 7 FEB 2008
- Manuscript Received: 28 DEC 2007
- Key Basic Research Project of the Science and Technology Commission of Shanghai Municipality. Grant Number: 05JC14029
- Outstanding Medical Academic Leader of Shanghai Municipality. Grant Number: LJ06024
- Natural Science Foundation of the People's Republic of China. Grant Number: 30470977
- Research Scholar
- American Cancer Society. Grant Number: CSM-106640
- National Cancer Institute. Grant Number: 1R01-CA093829
- National Institutes of Health
- colon tumor
Expression microarrays are widely used for investigating the nature and extent of global gene-expression changes in human cancer. Accurate genomewide gene-expression profiles have not been conducted in colon tumor and normal colon tissue specimens obtained from Chinese patients.
In the present study a pure population of colon cancer and normal colon cells was obtained and the global gene-expression differences were compared in the 2 cell types using combined experimental and bioinformatic approaches. Various categories of genes that were differentially expressed in those 2 types of cells were identified, including a novel candidate tumor marker, IFITM3.
Elevated IFITM3 expression in colon cancer cells was first confirmed using quantitative real-time polymerase chain reaction. IFITM3 protein expression in human colon cancer specimens was further analyzed using both tissue microarray and standard tissue sections by immunostaining analyses. It was found that there was a significant IFITM3 increase in adenoma as compared with that in normal colon tissue.
The data suggest that IFITM3 plays an important role in early colon cancer development. Cancer 2008. © 2008 American Cancer Society.
Colon cancer is 1 of the most common tumors worldwide, and in China in particular the incidence of colon cancer is steadily increasing because of a recent increase in changes in lifestyle and nutritional habits.1 Although activated oncogenes and inactivated tumor suppressor genes are known to be involved in important biological changes in the cells responsible for the development of colon cancer at the various stages of progression, much about the underlying etiopathogenic mechanisms related to mechanisms of human colon cancer development and progression remain to be fully elucidated. Genomewide gene-expression analysis of normal colon cells and corresponding colon cancer cells is crucial to understanding how these oncogenes and tumor suppressor genes alter the complex cellular molecular context of colon cancer and thus drive colon tumor progression.2
Microarray analysis is the most widely used methodology for gene-expression analysis. In recent years researchers have used microarrays to identify genes and pathways key to colon carcinogenesis as well as new biomarkers and target genes for prognosis for, diagnosis of, and therapy for colon cancer.3–6 Despite the enormous and well-validated efficiency of this technology in colon cancer research, further improvement is needed. First, heterogeneity of surgical tissue specimens may compromise accurate interpretation of gene-expression measurement, and in turn significantly confound downstream applications. To solve this problem, laser capture microdissection (LCM) can be used to obtain purified cell populations from the most heterogeneous tissue specimens and thus helps achieve precise molecular identification of specific cell types after gene-expression analysis.7 Second, the typical result of a microarray analysis is a list of tens or even hundreds of genes that are differentially regulated under the study conditions. The common task faced by researchers in microarray analysis is to use such gene lists to gain a better understanding of the biological phenomena involved in tumor development and progression. Gene Set Enrichment Analysis (GSEA) is a computational method developed to search databases for groups of functionally related genes that are overexpressed or underexpressed across a list of genes and ranked by their differential levels of expression in microarray experiments. Software programs such as Onto-Express (OE), Pathway-Express (PE), and FatiGO are examples of GSEA methods (Intelligent Systems and Bioinformatics Laboratory, Wayne State University, Detroit, Mich).8–10 Finally, prioritization of putative cancer markers is a critical step for further functional validation, but is often challenging because of the sheer number of genes under consideration. Serial analysis of gene expression (SAGE) is a method that is used for transcriptome research. Combined analysis using microarray and SAGE digital gene expression displayer (DGED) is feasible for quick, accurate prioritization of tumor-associated genes for functional analysis in patients with colon cancer.11
To effectively identify tumor markers and novel tumor suppressor/oncogenes, we sought to simultaneously improve tissue specimen preparation using the LCM technique before whole genome gene-expression profiling. We used GSEA to gain functional insight into a set of colon tumor-associated genes and selected a prioritized set of these genes for experimental validation by incorporating digital validation and filtration using SAGE DGED. Our results illustrate the accuracy and efficiency of this comprehensive transcriptome profiling protocol as exemplified by a rapid detection and validation of the candidate tumor marker IFITM3 in human colon cancer of Chinese patients.
MATERIALS AND METHODS
Human Tissue Specimens and Patient Information
For genomewide expression profiling, colon tumor and normal colon tissue specimens were obtained at the time of surgery from patients with colon cancer who underwent radical colectomy at Shanghai Jiaotong University Affiliated First People's Hospital in Shanghai (Table 1). Grossly visible normal and cancerous portions of the specimens were snap-frozen in liquid nitrogen and stored at −70°C. Hematoxylin and eosin-stained frozen sections of these specimens were examined by 2 pathologists. Tumor staging for the specimens was carried out using the sixth edition of the American Joint Committee on Cancer (AJCC) staging system.12 For immunostaining of IFITM3 protein, we used human colon cancer tissue specimens preserved in the Colon Cancer Tissue Bank at Shanghai Jiaotong University Affiliated First People's Hospital. Primary colon cancer in these patients was diagnosed and treated at the hospital from 2001 to 2003. We selected 83 cases to represent all of the stages and histological types of malignant colon cancer. Specifically, the patients consisted of 40 (48.2%) men and 43 (51.8%) women; 18 (21.7%) over age 60 and 65 (78.3%) under age of 60 with a mean age of 65.1 years; 28 (33.7%) rectum, 17 (20.5%) sigmoid colon, 27 (32.5%) ascending colon, 5 (6.0%) descending colon, and 6 (7.2%) transverse colon; 17 (20.5%) well differentiation, 43 (51.8%) moderate differentiation, and 23 (27.7%) poor differentiation; 12 (14.5%) grade 1, 46 (55.4%) grade 2, and 25 (30.1%) grade 3; 42 (50.6%) nodal positive and 41 (49.4%) nodal negative; and 63 (79.9%) negative distant metastasis and 20 (20.1%) positive distant metastasis. Moreover, 83 cases of matched normal colon tissues and 17 cases of adenoma tissues were also used in this study. All tissues were obtained with patient consent for the present study and all laboratory work involving human specimens was approved by the Institutional Review Board of and performed at Shanghai First People's Hospital. None of them underwent preoperative chemotherapy and/or radiation therapy. In addition, a tissue microarray used for immunostaining analysis of IFITM3 protein expression was purchased from Shanghai Outdo Biotech (Shanghai, China). The array consisted of 22 colon adenocarcinoma, 20 colon adenoma, and 8 normal colon mucosa tissues.
|Sex||Age, y||No. of specimens||AJCC stage||Grade|
|Nodal status||Distant metastasis|
LCM, RNA Extraction, and T7-Based Amplification
Eight-micrometer-thick sections were cut from each of the frozen tumor and normal tissue specimens. These sections were immediately fixed in cold (4°C) 100% ethanol for 5 minutes and then immersed in hematoxylin and RNase-free water for 10 seconds and xylene for 1 minute. Approximately 3 mm2 malignant and normal cells in each sample were captured using the Veritas LC/LCM system (Arcturus Engineering, Mountain View, Calif), and the total RNA was extracted using an RNeasy Micro Kit (Qiagen, Hilden, Germany) (Fig. 1). Colon cancer cell RNA and normal colon cell RNA were pooled and assayed using a 2100 bioanalyzer (Agilent Technologies, Palo Alto, Calif). One hundred nanograms of both RNA pools were amplified using the MessageAmp II amplified RNA (aRNA) Amplification Kit (Ambion, Austin, Tex). Two rounds of RNA amplification were performed to obtain enough aRNA for the microarray experiments described below.
Microarray Hybridization, Scanning, and Acquisition of Data
Synthesis of cDNA and preparation of biotin-labeled cRNA were performed using a 1-cycle cDNA synthesis kit and the GeneChip IVT Labeling Kit (Affymetrix, Santa Clara, Calif). A hybridization solution was used to prehybridize Human Genome U133 Plus 2.0 Arrays (Affymetrix) for 10 minutes at 45°C and 60 rpm. The hybridization solution was removed and replaced with a hybridization solution containing fragmented cRNA. The arrays were hybridized for 16 hours at 45°C and 60 rpm. The arrays were subsequently washed (GeneChip Fluidics Station 450; Affymetrix), stained with streptavidin-phycoerythrin, and scanned using the GeneChip Scanner 3000 7G (Affymetrix).
Gene-Expression Value Calculation
A scanned image of the cDNA microarray was analyzed using the GeneChip operating software program. Gene-expression values (signal, detection, and detection P-value) and relative gene-expression values (signal ratio and change) in cDNA microarray were calculated according to the Affymetrix Statistical Algorithms Descriptions Document. In the present report, 2 independent microarrays were performed for each pair of tumor and normal colon cells. R1 and R2 were the ratio values of gene expression in cancer cells versus normal cells from the first and second microarray, respectively. An adjusted ratio value (R) was calculated according to the formula: R = (R1 * R2) ˆ 0.5. If R for a gene was ≥2.0 or ≤0.5, this gene was then defined as a differentially expressed gene.
Gene Set Enrichment Analysis (GSEA)
The cDNA microarray data were annotated according to rudimental molecular function categories according to HG-U133_Plus_2_Annot_go (Shanghai Biochip, Shanghai, China). The gene-expression profiles of the colon tumor and normal colon tissue specimens and differentially expressed transcripts as detected by cDNA microarray were grouped according to 16 categories. The level of gene enrichment in each category was calculated as follows:
*The numbers were calculated according to the gene-expression profiles of the normal colon tissue specimens.
To establish which biological processes and pathways were overrepresented in our list of significantly up-regulated and down-regulated transcripts, GSEA was performed using OE and PE, 2 publicly available software programs used for functional interpretation of genome-scale data. OE constructed functional profiles for the molecular function categories for these transcripts and calculated the statistical significance values for each category. PE built a list of pathways for annotating the resulting data by associating our transcript list with Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways.
Comparison of the cDNA Microarray Data With SAGE DGED Results
The SAGE Genie Website13 has highly intuitive visual displays of human and murine gene expression based on a unique analytical process that reliably matches SAGE tags with each other. SAGE DGED distinguishes significant differences in gene-expression profiles in 2 pools of SAGE libraries. Differences in SAGE tag expression in normal colon tissue (NC-1 and NC-2) and colon tumor tissue (Tu-98 and Tu-102) SAGE libraries were analyzed using SAGE DGED. This method uses Bayesian analysis to compute the probability that sampling error accounts for the more frequent occurrence of a tag in 1 SAGE library pool than in the other. An expression factor of 2 and a significance filter of 0.05 were selected as the criteria for our Bayesian analysis. A list of 53 up-regulated genes and 132 down-regulated genes was reported as a result. The gene-expression trends in these 185 genes were compared using 2 platforms: LCM microarray analysis and SAGE DGED.
Validation of LCM cDNA Microarray Data By Using Real-Time Polymerase Chain Reaction
Six genes expressed at different ratios in normal colon and colon cancer cells were analyzed using quantitative polymerase chain reaction (PCR) with the ABI Prism 7000 sequence detection system (Applied Biosystems, Foster City, Calif) and the SYBR Premix Ex Taq Kit (Takara Bio, Shiga, Japan); glyceraldehyde-3-phosphate dehydrogenase was used as an internal control. The primer sequences were as follows: TGFBI: forward, 5′-CACATCTACACGTGGCTT GGA-3′; reverse, 5′-ACACACCATGGCTCTGTCACA-3′; IFITM3: forward, 5′-TCCCACGTACTCCAACTTCCA-3′; reverse, 5′-AGCACCAGAAACACGTGCACT-3′; NME1: forward, 5′-AGGATTCCGCCTTGTTGGT-3′; reverse, 5′-GGCCCTGAGTGCATGTATTTC-3′; FCGBP: forward, 5′-GACCAGACCAATGATTGGCA-3′; reverse, 5′-GAAAT CTTCGTTCCTGGTGGAT-3′; TSPAN1: forward, 5′-AAGCAAAAGGCTCACGACCA-3′; reverse, 5′-AATTCCAGCTGCCACACCA-3′; glyceraldehyde-3-phosphate dehydrogenase: forward, 5′-TGACTTCAACAGCGACA CCCA-3′; reverse, 5′-CACCCTGTTGCTGTAGCCAAA-3′. aRNA (3 μg) was reverse-transcribed from cDNA. Real-time reverse-transcribed (RT)-PCR was performed in a 25 μL total reaction mixture: 2 μL of cDNA as a template, 12.5 μL of SYBR Green, 0.5 μL of ROX dye, and 0.5 μL of the forward and reverse primers (5 pmol each). The thermal cycling conditions were 95°C for 2 minutes followed by 95°C for 15 seconds and 59°C for 1 minute, which was repeated for 40 cycles. Real-time PCR was repeated 3 times for each gene.
Tissue sections 4 μm thick were processed for detection of IFITM3 using a mouse monoclonal antibody H00010410-M01 (1:1000; Abnova, Taipei, China) and a secondary goat antimouse IgG (Dako EnVision Peroxidase Mouse, code K4000; Dako, Carpinteria, Calif). The sections were then counterstained with Mayer hematoxylin. Two independent investigators scored the sections without knowledge of the patient outcome (double-blinded). A mean value of 2 independent scores was calculated in the present study. Depending on the percentage of IFITM3-positive cells and staining intensity, IFITM3-positive staining was classified into 3 groups: negative/weak positive, moderate positive, and strong positive.
The 2-tailed χ2 test was performed to determine the significance of the relationship among the covariates. Each experiment was performed independently at least twice with similar results; 1 representative experiment is presented. P-values <0.05 were deemed significant.
Generation and Characterization of Gene-Expression Profiling
The microarray data for the pooled colon tumor specimens indicated a percentage present call of 33.3% and a mean gene-expression signal of 169.2. We detected 18,025 transcripts representing 15,276 genes in this microarray analysis. The microarray data for the pooled normal colon tissue specimens indicated a percentage present call of 32.5% and an average a mean gene-expression signal of 168.1. We detected 17,795 transcripts representing 14,939 genes in the microarray analysis used for gene-expression profiling of the colon tumor specimens.
Of 54,675 gene transcripts analyzed, 4300 were differentially expressed in colon cancer and normal colon cells. Of these 4300 transcripts, expression of 2125 and 2175 of them were up-regulated and down-regulated, respectively, in colon cancer cells. We classified individual transcripts in the normal colon tissue gene-expression profile and differentially expressed transcript sets according to 16 functional categories (Fig. 1). The categories for gene-enrichment levels greater than 1.1 included antioxidant_activity (1.44), signal_transducer_activity (1.39), apoptosis_or_tumor (1.23), enzyme_regulator_activity (1.20), and cell cycle (1.18).
Functional Categorization and Pathway Analysis of cDNA Microarray Data
We analyzed the association of the differentially expressed transcripts in the microarray data with Gene Ontology (GO) molecular identifications using the OE program. We observed a list of 1102 transcripts associated with GO molecular function terms and that the differentially expressed transcripts were statistically significantly enriched in 14 GO function categories (Table 2). We also analyzed the association of the differentially expressed transcripts in the microarray data with KEGG pathways using the PE program. We observed a total of 53 KEGG pathways and that the differentially expressed transcripts were statistically significantly enriched in 12 KEGG pathways (Table 3).
|Tag sequence||Gene||Tag||Signal||Signal log ratio||Description|
|GCCACCCCCT||H19*||56||0||4867.7||210.1||4.8||Imprinted maternally expressed untranslated mRNA|
|GTGTGTTTGT||TGFBI*||32||0||14547.5||928.9||3.6||Transforming growth factor-β-induced, 68 kDa|
|GACTGCGTGC||TBRG4||30||4||1200.8||117.6||3.2||Transforming growth factor-β regulator 4|
|ACTGGGTCTA||NME1*||50||12||5183.1||1312.0||2.0||Nonmetastatic cells 1, protein (NM23A) expressed in|
|TGAAATAAAA||NPM1*||24||2||3015.6||2350.3||1.8||Nucleophosmin (nucleolar phosphoprotein B23, numatrin)|
|ACCTGTATCC||IFITM3*||61||8||34489.0||8097.4||1.1||Interferon-induced transmembrane protein 3|
|CTGATGGCAG||LARP1||18||0||8051.5||3576.7||1.1||La ribonucleoprotein domain family, member 1|
|GGGGTCAGGG||PYGB||245||64||7294.9||3999.1||1.1||Phosphorylase, glycogen; brain|
|CACCACCACC||IFRD2||24||1||2916.9||1252.6||1.0||Interferon-related developmental regulator 2|
|Tag sequence||Gene||Tag||Signal||Signal log ratio||Description|
|ATTTCAAGAT||CA2*||1||56||18.1||13548.4||−9.1||Carbonic anhydrase II|
|CTGGCAAAGG||MS4A12||0||36||261.9||27161.9||−6.8||Membrane-spanning 4-domains, subfamily A, member 12|
|ATACTCCACT||GUCA2A*||13||141||329.4||20803.6||−5.9||Guanylate cyclase activator 2A (guanylin)|
|CCTTCAAATC||CA1*||1||46||610.8||32944.8||−5.6||Carbonic anhydrase I|
|CTTATGGTCC||DHRS9*||1||47||263.4||24609.4||−5.5||Dehydrogenase/reductase (SDR family) member 9|
|TGCTCCTACC||FCGBP*||39||253||769.1||39996.3||−5.2||Fc fragment of IgG binding protein|
|GTCATCACCA||GUCA2B*||0||57||183.0||8918.2||−5.0||Guanylate cyclase activator 2B (uroguanylin)|
|GTTCAAGATG||VSIG2||1||18||53.9||2477.7||−4.6||V-set and immunoglobulin domain containing 2|
Confirmation of Individual Gene Expression in the cDNA Microarray Data
Traditionally, this is done using a tedious combination of searches of the literature and several public databases. For our study, we searched the GeneCards database and the PubMed Website for information on the top 20 up-regulated and down-regulated genes in colon cancer cells. GeneCards is an integrated database of human genes that includes automatically mined genomic, proteomic, and transcriptomic data as well as disease relationship, single-nucleotide polymorphism, gene-expression, and gene-function information. We obtained general information about the aliases of the genes of these up-regulated and down-regulated genes from the GeneCards database and searched the literature for information relevant to these genes using the PubMed site. We found reports of differential expression of these genes in colon cancer cases (25% [10 of 40]) and of differential expression of a nonredundant subset of these genes in other cancer types (33% [13 of 40]).
For high-throughout validation of our cDNA microarray data, we used 2 platforms—LCM microarray and SAGE DGED—to compare the LCM microarray data for the 185 genes screened using SAGE DGED. Of the 132 down-regulated genes, 102 were down-regulated in colon cancer cells according to our microarray experiments. In comparison, of the 53 up-regulated genes, 17 were up-regulated in colon cancer cells according to our microarray experiments. Thus, the level of expression of 64% (119 of 185) differentially expressed genes screened using SAGE DGED was validated by the analysis of the microarray data. The top 10 up- and down-regulated genes in colon cancer cells are presented.
Real-Time PCR Confirmation
To validate the cDNA microarray data, we performed real-time PCR analysis to examine the expression of 5 genes differentially expressed at various levels: 3 with up-regulated expression and 2 with down-regulated expression. The aRNA used in our microarray experiments was used for this real-time PCR analysis. For all of the genes examined the trends of relative expression of their transcripts in the microarray and real-time PCR analyses were comparable (Fig. 2A).
Expression of IFITM3 Protein
Recently, researchers identified IFITM as a new molecular marker for human colorectal carcinoma14 and found that IFITM mRNA was expressed in sporadic colon tumor specimens and colon cancer cell lines but not in normal mucosa tissue.15 These important findings prompted us to determine whether IFITM3 protein is differentially expressed in human colon cancer. To confirm differential gene expression as identified according to LCM microarray and SAGE DGED analysis, we initially performed immunohistochemical analysis IFITM3 protein expression using a tissue microarray that consisted of 22 colon adenocarcinoma, 20 colon adenoma, and 8 normal colon mucosa tissues. We found IFITM3 protein-positive staining primarily in the cytoplasm of the colon cancer and normal colon cells (Fig. 2B). Also, we observed strongly or moderately IFITM3 protein-positive staining in 55% (12 of 22) of colon carcinoma specimens, 20% (4 of 20) of colon adenoma specimens, and no normal colon tissue specimens (0 of 8); the differences in staining in these 3 groups were significant (P = .006, χ2 test).
Furthermore, we extended IFITM3 immunostaining analysis using traditional tissue sections of 83 human colon cancer specimens and 83 matched normal colon tissue specimens, and 17 colon adenoma specimens. IFITM3 expression was significantly higher in colon adenoma and carcinoma than that in normal tissues (P = .006 and 0.000, respectively, χ2 test), whereas there was no significantly difference of IFITM3 expression in adenoma and carcinoma (P = .245, χ2 test) (Fig. 2C). Our data suggested that an increased IFITM3 expression appears to be an early event in colon carcinogenesis.
In the present study we obtained a pure population of colon cancer and normal colon cells and compared the global gene-expression differences in the 2 cell types using combined experimental and bioinformatic approaches. We identified various categories of genes that were differentially expressed in those 2 types of cells, including a novel candidate tumor marker IFITM3. Elevated IFITM3 expression in colon cancer cells were confirmed using quantitative PCR, tissue microarray, and traditional immunostaining analyses. There was a significant IFITM3 increase in adenoma as compared with that in normal colon tissue, suggesting that IFITM3 plays an important role in early colon cancer development.
Expression microarrays are widely used for investigating the nature and extent of global gene-expression changes in human cancer cases. However, accurate genomewide gene-expression profiles have not been conducted in colon tumor and normal colon tissue specimens obtained from Chinese patients. To better understand the transcriptome changes in Chinese colon cancer, we constructed gene-expression profiles for human colon cancer using genomewide microarray analysis combined with LCM and bioinformatic approaches. Most of the molecular function categories and pathways for the enriched gene sets in this study were consistent with what was reported previously and reflected the prevailing characteristics of the molecular context of development of colon cancer.16 Specifically, in our study the transcriptome changes in the categories and pathways intimately connected to these alterations included signal transduction, cell cycle, apoptosis, cell proliferation, cell division, axon guidance pathway, proteolysis, and angiogenesis.17–22
The differentially expressed transcripts were significantly enriched in the molecular function categories of lipid metabolism, carbohydrate metabolism, protein biosynthesis, transport, and enzyme regulator activity, indicating the presence of extensive metabolic dysregulation in colon cancer cells, which could lead to malignant progression.23 Other significantly enriched gene sets in our study are reported including antioxidant activity,24 immune response,25 Gap junctions,26 ubiquitin-mediated proteolysis,27 the transforming growth factor-β signaling pathway,28 actin cytoskeleton regulation,29 and the insulin signaling pathway30 in the development of colon cancer. Therefore, molecular function categories of a significantly enriched gene set in our study was highly consistent with a majority of the malignant characteristics of colon cancer, suggesting the reliability of our data. Interestingly, some of the molecular function categories of these gene sets, such as leukocyte transendothelial migration, prion disease, and SNARE interactions in vesicular transport were difficult to interpret using existing cancer theories. Further bioinformatic and functional analyses of these gene sets would provide insight into colon tumorigenesis.
Identifying differentially expressed genes using microarray data for further study is a challenging task.31, 32 In the present study, we identified candidate genes that differentially expressed in normal and cancerous colon cells by comparing our own data with the data available from the literature and with that obtained from the SAGE Genie Website. The previously reported differential expression of 25% (10 of 40) of the top 20 up-regulated genes and top 20 down-regulated genes in colon cancer suggested that LCM microarray is an effective method for screening the tumor-associated genes in colon cancer cells. Other researchers confirmed the differential expression of 33% (13 of 40) of these genes in other cancer types. These genes are also potential tumor-associated genes in colon cancer. Recently, researchers confirmed that combined multiple high-throughput analysis is an effective strategy for cancer gene identification. We selected a set of 185 genes screened using SAGE DGED for comparison with the microarray data. We found that the expression of 64% (119 of 185) of these genes matched the gene-expression trend for these genes in the microarray data. This result was acceptable considering that the overall agreement between the results obtained using LCM microarray and SAGE DGED is modest.33
Finally, we identified IFITM3 as a potential biomarker by using the LCM microarray and SAGE DGED platforms. IFITM3 gene encodes a cell-surface protein, which is a potential serum tumor marker for colon cancer. cDNA microarray analysis combined with tissue microarray analysis is a rapid 2-step screening approach for identification of differentially expressed genes in colon cancer cells.34–36 Recently, researchers identified IFITM as a new molecular marker for human colorectal carcinoma14 and found that IFITM mRNA was expressed in sporadic colon tumor specimens and colon cancer cell lines but not in normal mucosa tissue.15 We observed that IFITM3 mRNA expression was up-regulated in colon adenoma and adenocarcinoma cells using LCM microarray/SAGE DGED and real-time PCR analysis. Differential IFITM3 protein expression in normal and cancerous colon tissues was confirmed by immunostaining using tissue microarray as well as traditional tissue sections. IFITM3 protein expression was progressively elevated in the normal-adenoma-carcinoma sequence of colon cancer development. Elevated expression of IFITM3 protein was particularly evident in adenoma as compared with that in normal colon tissues, suggesting that IFITM3 expression occurs during the early phase of colon cancer development and progression. Interestingly, higher IFITM1 expression is correlated with improved survival in chronic myeloid leukemia patients,37 while the IFITM1 gene is suggested as a novel critical biomarker of drug response in esophageal squamous cell carcinoma.38 We are currently investigating whether IFITM3 protein expression in colon cancer is associated with tumor characteristics, including tumor grade/stage, tumor location, differentiation status, and response to downstream therapy.
In summary, combined LCM-assisted tissue-specimen preparation and cDNA microarray and SAGE DGED analysis was an accurate means of rapid identification and validation of tumor-associated genes in colon cancer, an effective method of prioritizing candidate markers for further screening using tissue microarray analysis, and an essential step toward global interpretation of genomewide tumor-associated gene expression data. Therefore, gene-expression profiling by coupling experimental and bioinformatic genomewide gene-expression analyses will greatly facilitate the identification and validation of biomarkers and novel functional genes of colon carcinogenesis. Our identification of IFITM3 as a biomarker in early colon carcinogenesis illustrated the accuracy and efficiency of these combined approaches. Further investigation is warranted to determine the regulation and role of IFITM3 in colon cancer development and progression.
The authors thank Gerard Quinones for assistance in the preparation of the article and Don Norwood for editorial comments. This work was supported in part by the Key Basic Research Project of the Science and Technology Commission of Shanghai Municipality (05JC14029), the Program for Outstanding Medical Academic Leader of Shanghai Municipality (LJ06024), the Natural Science Foundation of the People's Republic of China (30470977) (to Z.P.), and Research Scholar Grant CSM-106640 from the American Cancer Society and Grant 1R01-CA093829 from the National Cancer Institute, National Institutes of Health (to K.X.).
- 7Comparison of RNA amplification methods and chip platforms for microarray analysis of samples processed by laser capture microdissection. J Cell Biochem. 2007 (in press)., , , et al.
- 10http://vortex.cs.wayne.edu/Projects.html. Intelligent Systems and Bioinformatics Laboratory, Computer Science Department, Wayne State University, Detroit MI, April 17, 2007.
- 13http://cgap.nci.nih.gov/SAGE. SAGE Genie, National Cancer Institute (NCI), April 17, 2007.
- 23Enzymology of cancer cells. N Engl J Med. 1977; 486: 492–497..