Cdna Microarray Analysis of Gene Expression Profiles Associated with Cervical Cancer

Cancer Research and Treatment 2003;35(5):451-459 cDNA Microarray Analysis of Gene Expression Profiles Associated with Cervical Cancer

Joo Hee Yoon, M.D.1, Joon Mo Lee, M.D.1, Sung Eun Namkoong, M.D.1, Su Mi Bae, M.S.2, Yong-Wan Kim, Ph.D.2, Sei-Jun Han, M.D.3, Young Lae Cho, M.D.4, Gye Hyun Nam, M.D.5, Chong Kook Kim, M.D.6, Jeong-Sun Seo, M.D.7 and Woong Shick Ahn, M.D.1

1Department of Obstetrics and Gynecology, 2Catholic Research Institutes of Medical Science, College of Medicine, The Catholic University of Korea, Seoul; 3Department of Obstetrics and Gynecology, College of Medicine, Chosun University, Gwangju; 4Department of Obstetrics and Gynecology, Kyungpook National University, Daegu; 5Department of Obstetrics and Gynecology, College of Medicine, Soonchunhyang University, Seoul; 6College of Pharmacy, 7Department of Biochemistry and Molecular Biology, Seoul National University, Seoul, Korea

Purpose: The molecular pathology of cervical cancers were classified into 345 mutually dependent function sets, associated with human papillomavirus infection is pre- resulting in 611 cellular processes according to their GO. sently unclear. In an effort to clarify this issue, the gene The GO analysis showed that cervical carcinogenesis expression profiles and pathogenic cellular processes of underwent complete down-regulation of cell death, pro- cervical cancer lesions were investigated. tein biosynthesis and nucleic acid metabolism. The genes Materials and Methods: Cervical cancer biopsies were related to nucleic acid binding and structural molecule obtained from patients at the Department of Obstetrics activity were also significantly down-regulated. In con- and Gynecology, The Catholic University of Korea. The trast, significant up-regulation was shown in the skeletal disease status was assigned according to the Interna- development, immune response and extracellular activity. tional Federation of Gynecology and Obstetrics. The Conclusion: These data are suggestive of potentially tissue samples of 11 patients (invasive cancer stage Ib- significant pathogenetic cellular processes, and showed IIIa) were investigated by a cDNA microarray of 4,700 that the down-regulated functional profiling has an im- genes, hierarchical clustering and the Gene Ontology portant impact on the discovery of pathogenic pathways (GO). Total RNA from cervical cancer and non-lesional in cervical carcinogenesis. GO analysis can also overcome tissues were labeled with Cy5 and Cy3. The HaCaT hu- the complexity of the expression profiles of the cDNA man epithelial keratinocyte cell line was used as a ne- microarray via a cellular process level approach. Thereby, gative control cell. The stages of invasive cancer were a valuable prognostic candidate gene, with real relevance Ib to IIIb. All specimens were obtained by punch-biopsies to disease-specific pathogenesis, can be found at the and frozen in liquid nitrogen until required. cellular process levels. (Cancer Research and Treatment Results: 74 genes, showing more than a 2 fold differ- 2003;35:451-459) ence in their expressions, were identified in at least 8 ꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏ of the 11 patients. Of these genes, 33 were up-regulated Key Words: Cervical neoplasm, cDNA microarray, Gene and 41 were down-regulated. The gene expression profiles ontology

high risk group of HPVs (notably type 16 and 18) has been INTRODUCTION detected in the cervical intraepithelial neoplasms (CIN) I, II and in invasive cancers (1). However, no HPV DNA was detected in normal tissues adjacent to the CIN. After a high risk HPV Cervical carcinomas are mainly caused by infection with infection, two viral oncogenic proteins, E6 and E7, play a human papillomaviruses (HPVs) forming a high risk group. A critical role in the induction of cervical cancers, by interacting with p53 and pRB, inactivating these cellular regulatory Correspondence: Woong Shick Ahn, Department of Obstetrics and proteins, respectively (2). The two viral oncogenic proteins, E6 Gynecology, College of Medicine, The Catholic University of and E7, are commonly expressed in these carcinoma cells, and Korea, 505 Banpo-dong, Seocho-gu, Seoul 137-701, Korea. are required for maintaining cancer malignancy. It has been (Tel) 02-590-2409, (Fax) 02-599-4120, (E-mail) ahnws@cmc. reported, with the exception of cervical cancers, that most cuk.ac.kr cancer development frequently results from a p53 gene Received June 17, 2003, Accepted August 27, 2003 mutation (3). The p53 mutation has been detected in more than 451 452 Cancer Research and Treatment 2003;35(5)

50% of cancer cells, but with much less frequency in cervical this study. The stages of the invasive cancers were Ib to IIIb. cancer cell types (4). In most cervical cancers, however, the All specimens were obtained by punch-biopsies and frozen in function of the p53 is down regulated by the E6 protein of HPV liquid nitrogen until required. 16, whereby the E6 binds to the p53, resulting in the de- RNA was prepared from each specimen using Trizol reagent gradation of E6-p53 complexes through the ubiquitin pathway. (MRC. Cincinnati, OH), according the manufacturer's protocol. The viral E6 protein is also required for the continuous growth For the RT-PCR reaction, 50μg of total RNA was mixed with of HPV-immortalized cells. Similarly, the E7 protein of HPV 1μl of control mRNA (lambda bactriophage mRNA, 0.5μg/μl) 16 is also associated with the inactivation of the retinoblastoma and 1.5μg of oligo dT primer (1.5μg), to a final volume of tumor suppressor gene product. However, a HPV infection alone 20.5μl. The reaction mixture was heated for 5 min at 70oC, is not enough to trigger cervical cancers. Few patients infected and the denatured RNA reacted for 1 hr, at 42oC, in AMV with high risk HPV develop cervical cancers with a long buffer (low dT NTP, Cy3 or Cy5-dUTP, RNase inhibitor and incubation time, suggesting that additional factors, or cellular AMV reverse transcriptase). 10μl of 1 M NaOH was added events, might be required for progression to cervical cancers. to the reaction, for 10 min at 37oC, followed by the addition While it is becoming increasingly clear that there are wide of 25μl of 1 M Tris-HCl (pH 7.5). Probes were purified using variations in the efficiency of tumor therapies among different sephacryl S-100 columns. Ethanol-precipitated DNA probes tissue types, relatively little is known regarding the mechanism were solubilized in 15μl of hybridization buffer. by which a gene, or gene complex, is directly tumor-specific. 2) Probe hybridization Here, to quantitatively understand the multiple relationships between differentially regulated gene expressions and cervical Each gene on the array MAGIC KOGEN 4.7 K (Macrogen, carcinogenesis, the annotation project, directed by the Gene Inc, Seoul) is approximately 0.5∼5 kb cDNA, and includes a Ontology (GO) Consortium (http://www.geneontology.org), was control house keeping gene, such as GAPDH, β-actin or α- used (5). Despite the significance of cellular process analysis tubulin. (www.macrogen.co.kr) The probes were denatured for in tumor biology, the GO analysis has not been widely used 2 min at 95oC, and kept on ice until required. This was because of its complexity and rapidly evolving property. The followed by centrifugation for 10 min, at 14,000 rpm, with the GO analysis organizes the regulated genes into three separate subsequent drop wise addition of the probe supernatant to the ontologies, comprising of biological process, cellular com- DNA chip, which was then covered with a glass slip. The ponent and molecular function. The GO show a set of well- hybridization reaction was performed for 12∼16 hrs, in 3× defined terms and relationships, through which the role of a SSC, at 62oC. The reagents included in the array kit were used particular gene, gene product or gene-product group can be for all the hybridization and washing steps. After hybridization, interpreted in the pathogenesis. Thus, an advanced strategy for the slide was washed twice, for 30 min, in the 60oC washing the identification of preferential tumor-specific pathways would buffer (2×SSC and 0.2% SDS). The slide was then dried at be needed using the GO analysis. As of April 12, 2003, the the room temperature. GO contains 7,089 process, 5,374 function and 1,257 com- 3) Northern blotting ponent terms; with 10,493 GO term definitions. In this study, the identification of genes, expressed in mul- Plasmids of the clones of interest were prepared, and used tiple clinical specimens, were described. To identify the gene as PCR templates. All the PCR products underwent electro- expression profiles of cervical cancer, a 4.7 k cDNA microarray phoresis to verify the presence of the inserts. 20μg of each was used. The gene expression patterns, between cervical RNA sample were subjected to electrophoresis through 1.1% cancer and the corresponding normal cells were compared, and formaldehyde-agarose gels, and transferred to Hybond-N nylon 74 genes identified, which were up- or down-regulated more membranes (Amersham Pharmacia Biotech, Arlington Heights, than 2 fold, in the cervical cancer specimens. 33 genes were IL). 32P-dATP was incorporated into the first strand cDNAs, identified as being up-regulated, whereas 41 were down-regul- from total RNA, during the reverse transcription, using oligo ated, more than 2 fold, in at least 8 of the 11 clinical cervical (dT) as the primer. The hybridization was carried out overnight cancer specimens, suggesting that the GO analysis is descriptive at 68oC, in Rapid Hyb-buffer (Amersham Co. Arlington Heights, of tumorigenesis, and thus, a valuable tool for diagnostic and IL). After hybridization, the membrane was washed twice at therapeutic interventions in pathogenesis research. These results room temperature, in 2×SSC and 0.1% SDS, and then twice, suggest that further investigation of these genes is warranted at 65oC, in 0.1×SSC and 0.1% SDS. The membrane was for disclosing the molecular mechanisms of cervical cancer. radiophotographed at -70oC for 24∼48 hr. The identified genes might also be used for the specific 4) Scanning and data analyses construction of a cervical cancer cDNA microarray. The Cy3 and Cy5 fluorescence were scanned using a laser confocal microscope, and the images analyzed using an Ima- MATERIALS AND METHODS Gene v5.0 (BioDiscovery Ltd, Swansea, UK), to calibrate the relative ratios and confidence intervals for the significance 1) RNA and probe DNA preparation determinations. The gene expression data were normalized aga- The cervical cancer biopsies were obtained from patients at inst those of the housekeeping gene GAPDH, which includes the Department of Obstetrics and Gynecology, The Catholic 9 housekeeping genes at a constant level in all the arrays. The University of Korea. The disease states were assigned according fluorescence intensity was processed, and the data imported into to the International Federation of Gynecology and Obstetrics. an Excel (Microsoft) database, with the corresponding gene HaCaT keratinocyte cells were used as a negative control in names, for analysis. For each gene, the relative change in the Joo Hee Yoon, et al：Multifactorial Analysis in Function with Cervical Cancer 453

Table 1. Differential gene expression profile in cervical cancer ꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚ Gene name Description Fold change ꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏ IER5 immediate early response 5 89.3 S100A8 S100 calcium binding protein A8 (calgranulin A) 76.6 IGL@ immunoglobulin lambda locus 67.9 SPRR2A small proline-rich protein 2A 56.5 INPP5E inositol polyphosphate-5-phosphatase, 72 kDa 51.1 IGKC immunoglobulin kappa constant 46.9 PRG1 proteoglycan 1, secretory granule 34.8 MGC27165 hypothetical protein MGC27165 29.7 COL1A1 collagen, type I, alpha 1 28.2 HLA-DRB3 major histocompatibility complex, class II, DR beta 3 23.5 HLA-DRA major histocompatibility complex, class II, DR alpha 19.8 AEBP1 AE binding protein 1 13.9 SPARC secreted protein, acidic, cysteine-rich (osteonectin) 13.3 HLA-DRB5 major histocompatibility complex, class II, DR beta 5 11.0 APOL2 apolipoprotein L, 2 10.3 AQP3 aquaporin 3 8.9 A2M alpha-2-macroglobulin 8.6 SEPP1 selenoprotein P, plasma, 1 8.0 CES1 carboxylesterase 1 (monocyte/macrophage serine esterase 1) 7.7 OSR1 oxidative-stress responsive 1 6.8 HLA-C major histocompatibility complex, class I, C 6.2 C3 complement component 3 5.9 POP5 RNase MRP/RNase P protein-like 5.6 IGHG3 immunoglobulin heavy constant gamma 3 (G3m marker) 5.5 MKI67 antigen identified by monoclonal antibody Ki-67 4.5 NK4 natural killer cell transcript 4 4.3 TBC1D1 TBC1 (tre-2/USP6, BUB2, cdc16) domain family, member 1 4.0 BST2 bone marrow stromal cell antigen 2 3.9 RARRES3 retinoic acid receptor responder (tazarotene induced) 3 3.9 YIF1P Yip1p-interacting factor 3.5 CPE carboxypeptidase E 2.7 MAP3K11 mitogen-activated protein kinase kinase kinase 11 2.5 HLA-DQB1 major histocompatibility complex, class II, DQ beta 1 2.4 DAP3 death associated protein 3 -2.0 SPA17 sperm autoantigenic protein 17 -2.0 OAT ornithine aminotransferase (gyrate atrophy) -2.0 UFD1L ubiquitin fusion degradation 1-like -2.1 RTN3 reticulon 3 -2.1 NRM nurim (nuclear envelope membrane protein) -2.1 STAU staufen, RNA binding protein (Drosophila) -2.1 KPNB3 karyopherin (importin) beta 3 -2.2 SIP Siah-interacting protein -2.2 GARS glycyl-tRNA synthetase -2.2 C1QBP complement component 1, q subcomponent binding protein -2.2 PRO1580 hypothetical protein PRO1580 -2.3 XTP2 HBxAg transactivated protein 2 -2.3 LYPLA1 lysophospholipase I -2.4 NRAS neuroblastoma RAS viral (v-ras) oncogene homolog -2.4 ITGA3 integrin, alpha 3 (antigen CD49C, alpha 3 subunit of VLA-3 receptor) -2.4 UREB1 upstream regulatory element binding protein 1 -2.4 API5 apoptosis inhibitor 5 -2.5 KIAA1007 KIAA1007 protein -2.5 PCNP PEST-containing nuclear protein -2.5 FUS fusion, derived from t(12;16) malignant liposarcoma -2.5 HSPCA heat shock 90 kDa protein 1, alpha -2.6 BZW2 basic leucine zipper and W2 domains 2 -2.6 ARHGAP12 Rho GTPase activating protein 12 -2.7 ꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏ 454 Cancer Research and Treatment 2003;35(5)

Table 1. Continued ꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚ Gene name Description Fold change ꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏ C20orf30 chromosome 20 open reading frame 30 -2.7 HDLBP high density lipoprotein binding protein (vigilin) -2.8 MRPS2 mitochondrial ribosomal protein S2 -2.8 GTF2F2 general transcription factor IIF, polypeptide 2, 30 kDa -2.8 CSE1L CSE1 chromosome segregation 1-like (yeast) -2.8 ATIC 5-aminoimidazole-4-carboxamide ribonucleotide formyltransferase -2.9 LDHA lactate dehydrogenase A -3.0 APPBP1 amyloid beta precursor protein binding protein 1, 59 kDa -3.0 NCL nucleolin -3.0 TOP2A topoisomerase (DNA) II alpha 170 kDa -3.1 VPS35 vacuolar protein sorting 35 (yeast) -3.1 GRP58 glucose regulated protein, 58 kDa -3.2 APLP2 amyloid beta (A4) precursor-like protein 2 -3.2 SLC3A2 solute carrier family 3, member 2 -3.2 PA2G4 proliferation-associated 2G4, 38 kDa -3.3 TUBB2 tubulin, beta, 2 -3.8 ALDH7A1 aldehyde dehydrogenase 7 family, member A1 -4.1 ꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏ expression was the ratio of the median expression level of a total gene expression patterns, with the genes whose expression sample versus the negative control. Genes were excluded from correlated with cervical cancer searched. The analysis identified the analyses if their expression was negative or too smeared. a total of 74 genes whose expression values displayed sig- Genes showing a difference of at least 2 in their expression nificant association (positively or negatively regulated) with levels, resulting in a correlation coefficient of 0.89, were cervical cancer. Of these genes, 33 were up-regulated, and 41 selected for a function analysis. Hierarchical clustering (Gene down- regulated, in cervical cancer compared with the normal Cluster v2.11) and display programs (Tree View v1.50) were tissues. These results are presented in Table 1. The hierarchical also used in the analysis (http://rana.stanford.edu/software) (6). clustering, TreeView analysis and Gene ontology classification, In the cluster analysis (hierarchical clustering) all gene values of the differentially expressed and regulated genes in cervical were normalized against the control cell line. To classify the cancer, revealed that most of the genes were functionally gene profiles into their gene ontology, a cellular process analy- related to cell growth regulation, and cell and tissue function sis was carried out, as previously described (7). In brief, each and structure, as shown in Tables 2 and 3. The cellular pro- gene was first associated with its corresponding current curated cesses identified as being significant by the GO are listed in gene entry in the UniGene website (http://www.ncbi.nlm.nih. the Tables, and include the expression levels. There were gov), and annotated by integration with the information (as of several significant differences between the expression profiles April 3, 2003) on the Gene Ontology website (http://www.gene- of these functions in the cervical cancer compared with the ontology.org). Next, the gene was queried for its available Gene normal tissues. Ontology Code for its biological process, cellular component 1) Up-regulated gene expression profiles in cervical and molecular function. All the files, including the results of cancer specimens the microarray experiment and the Gene Ontology analysis, were downloaded from our anonymous FTP site: ftp://160.1.9. Table 1 shows the genes that were up-regulated more than 42/work/cervical_cancer/. 2 fold in 8 of the 11 patients. There were 33 genes determined as showing at least a 2 fold enhanced expression in the patients, compared to the HaCaT keratinocyte cells. These genes might RESULTS possibly be associated with the continuous growth of cancer cells in vivo. In Tables 2 and 3, these genes were classified To identify genes involved in the development of cervical in relation to the biological process they are involved in, ac- cancer, the transcripts from normal and cervical cancer groups cording to the criteria of the Gene Ontology Consortium. These were identified using the Macrogen Magic KOGEN 4.7 K include 6 significantly up-regulated genes that code for the array, as shown in Fig. 1. After obtaining the expression ratio proteins involved in development, 11 involved in immune for each gene in the cervical cancer and normal tissues from responses, 2 for response to oxidative stress, 6 for extracellular each array, the values were globally normalized and trans- activity and numerous other miscellaneous genes. In particular, formed, as described in the Materials and Methods section. The genes coding for proteins involved in immune functions, such mean expression values were then used to identify the dif- as MHC II, DR alpha and beta, MHC I C, Ig heavy constant ferentially expressed genes whose expression values gave gamma 3, bone marrow stromal cell antigen 2, Ig kappa reproducible results. The hierarchical clustering showed the constant, complement component 3 and natural killer cell trans- Joo Hee Yoon, et al：Multifactorial Analysis in Function with Cervical Cancer 455

A B

Fig. 1. Cervical cancer-associated gene expression data. Total RNA from cervical cancer and non-lesional tissues were labeled with Cy5 and Cy3, and hybridized to our 4.6 k cDNA microarray (A). The green spots represent the genes whose expressions were greater in the normal tissues than cervical cancer lesion. The red spots represent the genes whose expressions were greater in the cervical cancer lesions. The yellow spots represent the genes that were expressed almost equally in both tissue types. Spots with dust/specks, or similar artifacts, were flagged to avoid their inclusion in the analysis. All the data were mean centered and clustered, using GeneCluster and average linkage hierarchical clustering. The reproducibility of the array was confirmed by deliberately spotted replicate cDNAs that clustered closely together. The scatter plots are represented as log-log scales of the expression values (B, C). For each graph, only those transcripts present in both samples are considered. Technical variability was provided by the comparison of the expression levels of two different chip analyses performed with the same sample. The vast majority of the points were similar in each plot. A slight increase in the number of data points in (B) indicated changes in the patterns of gene expression attributed to chip-specific characteristics. cript 4, had significant induction of their expressions. significantly down-regulated. In particular, the expression of the genes coding for proteins involved in cell death and nucleic 2) Down-regulated gene expression profiles in cervical acid metabolism, such as death associated protein 3 and general cancer specimens transcription factor IIF, completely disappeared. The genes with at least a 2-fold down-regulation in the cervical cancer specimens were also analyzed. In Table 1, 41 DISCUSSION genes are shown to demonstrate at least a 2 fold down- regulated expression in the patients, compared to the HaCaT keratinocyte cells. These genes were also classified in the same A 4,700-element microarray was used to identify the up- or way. As shown in Tables 2 and 3, 3 genes coding for cell down-regulated genes in cervical tumors. In our study, how- death, 7 for development, 5 for protein biosynthesis and nucleic ever, there was variability in the signal intensities in microarray acid metabolism, 7 for nucleic acid binding activity, 4 for assay. Therefore, genes were selected that showed more than structural molecule activity, and other miscellaneous geneswere a 2 fold difference in their gene expression in at least 8 of the 456 Cancer Research and Treatment 2003;35(5)

Table 2. Differential biological process profile in cervical cancer Table 2. Contiued ꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚ ꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚ Cell communication Immune response Cell adhesion Signal transduction S100A8 76.6 AEBP1 13.9 HLA-DRB3 23.5 IGKC 46.9 NK4 4.3 C3 5.9 HLA-DRB3 23.5 ITGA3 -2.4 BST2 3.9 HLA-DRA 19.8 CPE 2.7 APOL2 10.3 MAP3K11 2.5 HLA-C 6.2 SPA17 -2.0 C3 5.9 NRAS -2.4 APPBP1 -3.0 IGHG3 5.5 GRP58 -3.2 NK4 4.3 APLP2 -3.2 BST2 3.9 HLA-DQB1 2.4 Cell death C1QBP -2.2 DAP3 -2.0 Apoptosis inhibitor ꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏ CSE1L -2.8 API5 -2.5 Cell growth and/or maintenance FUS -2.5 Transport 11 patients. This suggests that the genes identified in this study SLC3A2 -3.2 APOL2 10.3 should show consistent gene expression profiles in more than TUBB2 -3.8 AQP3 8.9 8 of the different cancer tissue specimens. This approach Cell proliferation/cell cycle A2M 8.6 resulted in the separation of a significant number of genes, and MKI67 4.5 STAU -2.1 allowed a comparative analysis of their regulatory behavior in BST2 3.9 KPNB3 -2.2 cervical cancer and normal tissues, based on their patterns of RARRES3 3.9 CSE1L -2.8 expression. This cutoff point was arbitrary, and there might be NRAS -2.4 HDLBP -2.8 more genes whose expressions are unique to a single specimen. CSE1L -2.8 VPS35 -3.1 PA2G4 -3.3 GRP58 -3.2 However, the identification of the genes regulated in common SLC3A2 -3.2 in cervical cancer tissues might shed light on the understanding of cervical cancer, and its progression at molecular levels. Development As usual, the gene expression profiles have been classified SPRR2A 56.5 into reasonable groups, using statistical data mining methods, APOL2 10.3 such as the hierarchical and K-Means clustering. It was, BST2 3.9 however, based on the statistical relationship between the gene SPA17 -2.0 Skeletal development expression values in irrespective of the cellular processes, COL1A1 28.2 which resulted in a biological description limit, as well as a AEBP1 13.9 reproducibility problem. Thus, in order to give a biological in- SPARC 13.3 terconnectivity between a gene and its function, the differ- UFD1L -2.1 entially regulated genes should be classified using the Gene Ontology, as a new alternative based on the cellular process Metabolism of the gene products. Correlations were sought between the APOL2 10.3 genes in the functional groups and their expression patterns. MAP3K11 2.5 OAT -2.0 The changes observed in these analyses provided important UFD1L -2.1 insights into cervical cancer-specific gene and pathogenic HDLBP -2.8 functional changes. The hierarchical clustering and gene SLC3A2 -3.2 ontology analysis of the differentially regulated genes in GRP58 -3.2 cervical cancer assigned most of the genes to 639 cellular processes, showing the key functions of various cellular Protein biosynthesis Proteolysis and peptidolysis activities involved in the development of cervical cancer. GARS -2.2 INPP5E 51.1 Northern blots were performed to confirm the gene expression MRPS2 -2.8 AEBP1 13.9 patterns. Several up- and down-regulated genes had their CPE 2.7 patterns obtained from the microarray confirmed, showing the UFD1L -2.1 consistency of the assays. (Data not shown) PA2G4 -3.3 Nucleic acid metabolism Response to oxidative stress The genes associated with protein synthesis and cellular GTF2F2 -2.8 SEPP1 8.0 metabolism are common in the differential gene expression ATIC -2.9 OSR1 6.8 analysis in tumor tissues, as compared to their normal counter- TOP2A -3.1 parts (8). Complete down-regulation of the genes coding for proteins involved in nucleic acid metabolism and protein ꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏ biosynthesis in the cervical cancer specimens was also observed. These included the transcription factors (GTF2F2), glycyl-tRNA synthetase (GARS) and ribosomal protein S2 Joo Hee Yoon, et al：Multifactorial Analysis in Function with Cervical Cancer 457

Table 3. Differential molecular function profile in cervical cancer (MRPS2). Furthermore, the genes coding for 13 of the 20 ꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚꠚ metabolic enzymes were also down-regulated in their expres- Binding activity sions in the cervical cancer specimens compared to the control S100A8 76.6 Protein binding cells, suggesting the importance of these gene products in the IGKC 46.9 INPP5E 51.1 developmental pathway of cervical cancer cells. In contrast, the SPARC 13.3 APOL2 10.3 expression of some genes, including mitogen-activated protein APOL2 10.3 A2M 8.6 kinase kinase kinase11 (MAP3K11), carboxypeptidase (CPE) SEPP1 8.0 KPNB3 -2.2 and carboxylesterase (CES1), were up-regulated in cervical C3 5.9 VPS35 -3.1 cancer specimens. Furthermore, the expression of other genes, HDLBP -2.8 such as calgranulin A (S100A8) and retinoic acid receptor responder (RARRES3), were also up-regulated, suggesting that Nucleic acid binding/transcription/translation these gene products might be involved in assuring proper cell AEBP1 13.9 growth in the progression to cancer. Interestingly, the activity STAU -2.1 of the response to oxidative stress involved in the metabolism FUS -2.5 GTF2F2 -2.8 was completely up-regulated, as shown in Table 2. HDLBP -2.8 It was of note that a proliferative marker, Ki-67, had an NCL -3.0 increased gene expression in our cervical cancer specimens. TOP2A -3.1 This was compatible with the previous finding that cells expres- APLP2 -3.2 sing Ki-67 increased as a function of the increasing cervical lesion grade, possibly as a proliferation marker of cervical Enzyme activity cancers (9). The consistency in the expression of Ki-67 between Hydrolase activity Peptidase activity the studies could be regarded as validation of the assays. INPP5E 51.1 INPP5E 51.1 The differential expression of several genes are functionally CES1 7.7 AEBP1 13.9 related to cell cycle regulation in cervical cancer, with the CPE 2.7 CPE 2.7 pathways leading to cell growth and apoptosis playing key roles ATIC -2.9 GRP58 -3.2 in the growth and regulation of cervical cancer. Several genes GRP58 -3.2 PA2G4 -3.3 are included in this category, such as caspase 10. The micro- PA2G4 -3.3 array data indicated that MKI67, BST2 and RARRES3 were Kinase activity Transferase activity overexpressed in cervical cancer, and the overexpression of OSR1 6.8 OSR1 6.8 MKI67 has been shown to result in the induction of prolifer- MAP3K11 2.5 MAP3K11 2.5 ation; however, its effect in this process appears to be depen- SPA17 -2.0 OAT -2.0 dent on the grade of the cancer lesion. Although our results ATIC -2.9 provide additional support for the alteration of a set of genes involved in cell cycle regulation and apoptosis playing a critical Enzyme inhibitor activity role in cervical cancer growth and regression, further investi- A2M 8.6 gation to confirm their expressions and actions is warranted. C3 5.9 The integrity of the cytoskeleton plays an important role in KPNB3 -2.2 the progression of the cell cycle, cell death and cell differ- APLP2 -3.2 entiation. An abnormal cytoskeleton is often observed in cancer cells. In this study, the gene expression profiles involved in the Structural molecule activity cytoskeleton were relatively down-regulated. Genes coding for SPRR2A 56.5 COL1A1 28.2 cellular structure proteins, such as cornified envelope (SPRR2A) DAP3 -2.0 and collagen (COL1A1), were highly up-regulated in cervical STAU -2.1 cancer. In contrast, microtubule associated protein (STAU), st- MRPS2 -2.8 ructural constituent of ribosome (MRPS2) and tubulin (TUBB2), TUBB2 -3.8 were down-regulated. In prostate cancer, suppression of the tumorigenicity of prostatic cancer cells by STAU has been re- Extracellular activity ported (10), suggesting that intermediate structural proteins S100A8 76.6 might be involved in cervical tumors. In our study, the expres- COL1A1 28.2 sion of the integrin α3 gene was down-regulated, suggesting SPARC 13.3 a deregulation of intermediate filament proteins in cervical CES1 7.7 cancers. Integrins are known to be cell surface proteins that C3 5.9 mediate cell to cell, and cell to extracellular matrix, interac- NK4 4.3 tions. Deregulation of the expression of the integrin gene has RTN3 -2.1 been commonly detected in various tumors, including cervical cancer (11∼14). In other studies, however, the expressions of Chaperone activity integrin β1/4 and α6 genes were decreased in human cervical HSPCA -2.6 cancers (12,14). It is likely that the expression of an integrin TUBB2 -3.8 subunit could be differentially expressed during the progression ꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏꠏ 458 Cancer Research and Treatment 2003;35(5) of cervical tumors. suggesting the importance of these pathways, and their re- Disruption of the apoptotic and cell cycle progression ciprocal interactions, in events leading to cervical cancer pathways has been implicated in abnormal cell growth and growth. carcinogenesis (15,16). The genes involved in cell death, such Our most interesting finding was the down-regulation of as DAP3 and CSE1L, were observed to be down-regulated in transcriptomes shown in the cell adhesion, cell motility, struc- the cervical cancer specimens. Death-associated protein (DAP3) tural molecule activity and response to external stimuli func- encodes a 28S subunit protein of the mitochondrial ribosomes tional activities, which are supposed to play important roles in that participate in apoptotic pathways, and are initiated by cervical carcinogenesis. Microarray data could be used in the tumor necrosis factor-alpha, Fas ligand and gamma interferon. future for tumor classification and in the determination of tumor Being a positive mediator of cell death, DAP3 may be a tumor progression based upon the gene expression and cellular pro- suppressor gene subjected to loss, or inactivation, in tumors cess patterns. Moreover, the results might be used to identify (17). In contrast, the genes involved in chaperone activity molecular pathways that may be involved in cervical tumori- (HSPCA and TUBB2) were down-regulated in the cervical genesis. cancer specimens. This was inconsistent with the previous finding that the syntheses of the 90-kDa (HSP90) heat-shock CONCLUSIONS proteins encoded by HSPCA are increased in malignantly transformed cells, and have been associated with tumor proliferation, metastasis and resistance to chemotherapeutic agents The gene expression changes, and their cellular functions, (18). Our results illustrate the divergence of gene expressions which might be involved in the progression to cervical cancer, in the cervical cancer specimens. were identified, and have the potentially be used as candidates The differential expression of immune response has been for further investigation into the pathobiology of cervical directly, and indirectly, associated with various types of tissue cancer. Also, it has implicated several novel genes not pre- carcinogenesis. Numerous genes coding for proteins involved viously studied in the Gene Ontology. Thereby, the expressions in immune functions were notably up-regulated in the cervical of the genes in this study clearly displayed either a common cancer specimens. In our cervical cancer patients, a modest, but or specific association with cervical cancer, reflecting evidence consistent, immune response against carcinogens was observed. that indicates the similarities and differences between the In several types of carcinoma, a number of genes responsible molecular environments in cervical carcinogenesis. This will for immune response were reported to be up-regulated (19). effectively lead to the identification of new tumor markers for The significance of the activation of immune related genes in use in tumor diagnosis and treatment. cervical cancer is suggested by the induction of their expressions. One possibility is that the induced expressions of these REFERENCES immune related genes might be related to the vital cell state feature of cervical cancer cells. It would be interesting to know whether the expression of immune-related genes correlates well 1. Takac I, Marin J, Gorisek B. Human papillomavirus 16 and with their protein expression levels. However, it is unlikely that 18 infection of the uterine cervix in women with different these immune related genes are expressed in cervical cancer grades of cervical intraepithelial neoplasia (CIN). Int J Gy- naecol Obstet 1998;61:269-273. cells from an epithelial cell origin. The precise role of these 2. Bae SM, Kim YW, Yoon JH, Yoo JY Seo YS, Nam SL, Lee immune proteins in cervical cancer should be further clarified. JM, Namkoong SE, Kim CK, Kim Y-W, Ahn WS. Cell- The regulation of extracellular matrix expression is vital to specific growth inhibition of human cervical cancer cell by tissue remodeling in normal and pathological processes leading recombinant adenovirus p53 in vitro and in vivo. Cancer Res to tissue carcinogenesis (20). Interestingly, the genes in this Treat 2003;35:181-190. cellular process were highly up-regulated. Among the genes in 3. Lee SJ, Namkoong SE, Lee WC, Sul JW, Jee SH, You YK, this category were calgranulin A, collagens and osteonectin. Lee JE, Park JS. Polymorphisms of p53, p21 and IRF-1 and The expression of collagens I that regulates its turnover has cervical cancer susceptibility in Korean women. Cancer Res been demonstrated in tumorigenesis (21). Calgranulin A is also Treat 2002;34:357-364. 4. Greenblatt MS, Bennett WP, Hollstein M, Harris CC. Muta- a key regulator of a number of cellular processes, such as the tions in the p53 tumor suppressor gene: clues to cancer progression and differentiation of the cell cycle. An altered etiology and molecular pathogenesis. Cancer Res 1994;54: expression of this protein has been associated with cystic 4855-4878. fibrosis. Linkage analyses have identified disease susceptibility 5. Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry regions of an inflammatory disorder, suggesting the upregul- JM, Davis AP, Dolinski K, Dwight SS, Eppig JT, Harris MA, ation of calgranulin A (22). Hill DP, Issel-Tarver L, Kasarskis A, Lewis S, Matese JC, Other differentially regulated genes in cervical cancer belong Richardson JE, Ringwald M, Rubin GM, Sherlock G. Gene to signal transduction pathways. The activated MAP3K11 re- ontology: tool for the unification of biology. Nat Genet 2000; sults in the activation of transcription factors and phosphoryl- 25:25-29. ation, such as MAPK8/JNK kinase, and functions as a positive 6. Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and display of genome-wide expression patterns. Proc Nat regulator of the JNK signaling pathway. This kinase can direc- Acad Sci USA 1998;95:14863-14868. tly phosphorylate, and activates IkappaB kinase alpha and beta, 7. Pletcher SD, Macdonald SJ, Marguerie R, Certa U, Stearns SC, and is found to be involved in the transcription activity of NF- Goldstein DB, Partridge L. Genome-wide transcript profiles in κB, which is mediated by Rho family GTPases and CDC42, aging and calorically restricted Drosophila melanogaster. Joo Hee Yoon, et al：Multifactorial Analysis in Function with Cervical Cancer 459

Current Biology 2002;12:712-723. 19:711-720. 8. Wang K, Gan L, Jeffery E, Gayle M, Gown AM, Skelly M, 16. Wyllie AH. Apoptosis and carcinogenesis. Eur J Cell Biol Nelson PS, Ng WV, Schummer M, Hood L, Mulligan J. 1997;73:189-197. Monitoring gene expression profile changes in ovarian carci- 17. Kissil JL, Deiss LP, Bayewitch M, Raveh T, Khaspekov G, nomas using cDNA microarray. Gene 1999;229:101-108. Kimchi A. Isolation of DAP3, a novel mediator of interferon- 9. Resnick M, Lester S, Tate JE, Sheets EE, Sparks C, Crum CP. gamma-induced cell death. J Biol Chem 1995;270:27932- Viral and histopathologic correlates of MN and MIB-1 expres- 27936. sion in cervical intraepithelial neoplasia. Hum Pathol 1996; 18. Korneeva I, Bongiovanni AM, Girotra M, Caputo TA, Witkin 27:234-239. SS. IgA antibodies to the 27-kDa heat-shock protein in the 10. Gao AC, Lou W, Ichikawa T, Denmeade SR, Barrett JC, Isaacs genital tracts of women with gynecologic cancers. Int J Cancer JT. Suppression of the tumorigenicity of prostatic cancer cells 2000;87:824-828. by gene(s) located on human chromosome 19p13.1-13.2. Pro- 19. Iizuka N, Oka M, Yamada-Okabe H, Mori N, Tamesa T, state 1999;38:46-54. Okada T, Takemoto N, Tangoku A, Hamada K, Nakayama H, 11. Varner JA, Cheresh DA. Integrins and cancer. Curr Opin Cell Miyamoto T, Uchimura S, Hamamoto Y. Comparison of gene Biol 1996;8:724-730. expression profiles between hepatitis B virus-and hepatitis C 12. Aplin JD, Dawson S, Seif MW. Abnormal expression of in- virus-infected hepatocellular carcinoma by oligonucleotide mi- tegrin alpha 6 beta 4 in cervical intraepithelial neoplasia. Br croarray data on the basis of a supervised learning method. J Cancer 1996;74:240-245. Cancer Res 2002;62:3939-3944. 13. Hodivala KJ, Pei XF, Liu QY, Jones PH, Rytina ER, Gilbert 20. Branton MH, Kopp JB. TGF-beta and fibrosis. Microbes Infect C, Singer A, Watt FM. Integrin expression and function in 1999;1:1349-1365. HPV 16-immortalised human keratinocytes in the presence or 21. Nowak RA. Fibroids: Pathophysiology and current medical absence of v-Ha-ras. Comparison with cervical intraepithelial treatment. Baillieres Best Pract Res Clin Obstet Gynaecol neoplasia. Oncogene 1994;9:943-948. 1999;13:223-238. 14. Shim CS, Zhang W, Rhee CH, Lee JH. Profiling of differ- 22. Semprini S, Capon F, Tacconelli A, Giardina E, Orecchia A, entially expressed genes in human primary cervical cancer by Mingarelli R, Gobello T, Zambruno G, Botta A, Fabrizi G, complementary DNA expression array. Clin Cancer Res 1998; Novelli G. Evidence for differential S100 gene over-expres- 4:3045-3050. sion in psoriatic patients from genetically heterogeneous pedi- 15. Metcalfe A, Streuli C. Epithelial apoptosis. Bioassays 1997; grees. Hum Genet 2002;111:310-313.