Identification of Differentially Expressed Genes in Pancreatic Cancer Cells Using Cdna Microarray1

[CANCER RESEARCH 62, 2890–2896, May 15, 2002] Identification of Differentially Expressed Genes in Pancreatic Cancer Cells Using cDNA Microarray1

Haiyong Han, David J. Bearss, L. Walden Browne, Robert Calaluce, Raymond B. Nagle, and Daniel D. Von Hoff2 Arizona Cancer Center [H. H., D. J. B., R. C., R. B. N., D. D. V. H.] and Department of Pathology [W. B., R. C., R. B. N.], University of Arizona, Tucson, Arizona 85724

ABSTRACT pressor genes that are altered in pancreatic cancer include BRCA2 (10), ALK-5 (11), MKK4 (12), and STK11 (13). One oncogene that is To identify new diagnostic markers and drug targets for pancreatic commonly mutated in pancreatic cancer is K-ras.K-ras mutations cancer, we compared the gene expression patterns of pancreatic cancer have been found in 90% of the pancreatic cancers, with most of these cell lines growing in tissue culture with those of normal pancreas using cDNA microarray analysis. Fluorescently (cyanine 5) labeled cDNA being point mutations on codon 12 (7, 14). Some other cancer-related probes, made individually from mRNA samples of nine pancreatic cell genes such as Her-2/neu, COX-2, and VEGF have also been reported lines, were each combined with fluorescently (cyanine 3) labeled universal to be overexpressed in pancreatic cancer cells (15–17). Microsatellite reference mRNA. The mixed probes of each sample were then hybridized instability in cancer cells usually suggests a defective mismatch repair with 5760 cDNA arrays (5289 unique cDNA sequences) printed on indi- system (18). Yamamoto et al. (19) have reported recently that 26 of vidual microscope slides. Fluorescently (cyanine 5) labeled normal pan- 100 sporadic pancreatic tumors showed microsatellite instability, in- creas mRNA was also compared with the same universal reference mRNA dicating mutations in mismatch repair genes in those tumors. reference pool. The expression ratios of neoplastic versus normal pancreas The identification and characterization of these cancer-related cells were then calculated by multiplying the ratio of cancer versus the genes have increased our understanding of pancreatic cancer devel- universal reference mRNA and the ratio of the universal reference mRNA opment, but unfortunately the treatment of pancreatic cancer has not cell versus normal pancreas. For 5289 different genes interrogated by the arrays, 30 of them showed an expression ratio 2 SD from the mean in at advanced as much in the past 20 years. This is mainly attributable to least three of the nine pancreatic cell lines studied. To confirm the the lack of early diagnosis and effective chemotherapeutic treatments. expression profiles of these genes, quantitative reverse transcription-PCR To increase the survival rate of pancreatic cancer patients, better and Northern blot were carried out for 25 of the overexpressed genes. To tumor markers for diagnosis and new molecular targets for drug verify the overexpression in patient samples, two of the overexpressed development are desperately needed. genes, c-Myc and Rad51, were selected to undergo analysis by reverse Because the development and progression of pancreatic cancer is a transcription-PCR in frozen tumor tissues and by immunostaining in very complicated process, we hypothesized that many other genes, as paraffin-embedded tissue section microarrays. The results of these exper- yet undiscovered, are potential tumor markers or drug targets. How- iments are in agreement with the microarray data. Potential up-regulated ever, identifying these genes by conventional methods such as North- targets of note from this study include urokinase-type plasminogen acti- ern blots, differential display, and serial analysis of gene expression vator receptor, serine/threonine kinase 15, thioredoxin reductase, and CDC28 protein kinase 2, as well as several others. has been either labor intensive or nonsystematic (20, 21). Proteomics have provided some promise for massive protein analysis, but tech- niques involved are still in the stage of early development (22). In this INTRODUCTION study, we have used a high-density cDNA microarray technique to assess the gene expression levels in neoplastic versus normal pancre- The development and growth of cancer is a multistep process atic cells. This DNA microarray technique allows simultaneous com- including initiation, progression, invasion, and ultimately establish- parison of a large number of genes in two samples in a quantitative ment of metastatic disease. Each of these steps involves multiple and expeditious way (21). We used a cDNA microarray slide con- genetic alterations that give cancer cells a selective advantage over taining 5289 unique cDNA sequences and compared the mRNA levels normal cells (1). Similar to most tumor types, the development of of nine pancreatic cell lines with that of normal pancreas. We char- malignant adenocarcinoma of the pancreas is thought to be driven by acterized gene-expression profiles that molecularly discriminate pan- the accumulation of genetic alterations. Over the past decade, many creatic cancer cell lines from normal pancreas cells. Moreover, we studies involving pancreatic cancer have searched for cancer-causing identified genes that may be involved in pancreatic tumorigenesis as genes (2, 3). As a result, several cancer-related genes have been well as genes that are potential clinical biomarkers that may lead to an identified in pancreatic tumors. These genes can be categorized into improved early diagnosis for this disease. Several of these genes may three groups including: (a) tumor suppressor genes; (b) oncogenes; also constitute potential novel therapeutic targets. and (c) mismatch repair genes (2, 4, 5). DPC4, p53, and p16 are the three most frequently inactivated tumor suppressor genes in sporadic pancreatic adenocarcinomas (6). Accumulating data suggest that p53 MATERIALS AND METHODS function is inactivated in ϳ75% of pancreatic tumors, whereas p16 is lost in ϳ95% of all pancreatic cancers (7, 8). DPC4, also known as Cell Culture. Pancreatic cell lines AsPC-1, BxPC-3, Capan-1, CFPAC-1, SMAD4, is a tumor growth factor-␤ signaling pathway member and is HPAF II, MIA PaCa-2, PANC-1, and SU.86.86 were purchased from Amer- ican Type Tissue Culture Collection. Pancreatic cell line Mutj (UACC-462) inactivated in ϳ50% of all pancreatic cancers (9). Other tumor sup- was established at the Arizona Cancer Center (23). All pancreatic cell lines were cultured in RPMI 1640 supplemented with fetal bovine serum (10% for Received 9/21/01; accepted 3/7/02. AsPC-1, BxPC-3, HPAF II, MIA PaCa-2 PANC-1, and Mutj; 15% for CF- The costs of publication of this article were defrayed in part by the payment of page PAC-1 and SU.86.86; 20% for Capan-1), penicillin, and streptomycin. HeLa charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. cells were grown in DMEM supplemented with 10% fetal bovine serum. All 1 Supported in part by a grant from the National Foundation for Cancer Research and cells were harvested when they were about 80–90% confluent and were the Arizona Cancer Center. D. V. H. is a fellow of National Foundation for Cancer directly subjected to RNA isolation. Research. Microarray Sample Preparation and Hybridization. cDNA microarray 2 To whom requests for reprints should be addressed, at Arizona Cancer Center, University of Arizona, P. O. Box 245024, Tucson, AZ 85724-5024. Phone: (520) 626- slides used in this study were fabricated in the microarray core facilities at the 7925; Fax: (520) 626-6898; E-mail: [email protected]. Arizona Cancer Center (24). Briefly, each slide has 5760 spots divided into 2890

four blocks, with each containing eight identical ice plant genes from Mesem- RT-PCR. Two ␮g of total RNA isolated from pancreatic cancer cell pellets bryanthemum crystallinum and 23 different housekeeping genes as references or frozen pancreatic tumor tissues were used for reverse transcriptase reactions for data normalization. Each slide had 5289 unique human cDNA sequences. (20 ␮l in total volume), which were carried out using the Omniscript RT kit Poly(A)ϩ RNA was directly isolated from cell pellets using the FastTrack (Qiagen, Inc.), following the manufacturer’s protocol. The PCRs were then 2.0 kit (Invitrogen, Carlsbad, CA), following the instruction manual provided carried out by mixing 2 ␮l of reverse transcriptase reaction mixture, 5 ␮lof ϩ by the manufacturer. Normal pancreas poly(A)ϩ RNA was isolated from total 10ϫ PCR buffer containing 15 mM Mg2 ,1␮lof10mM deoxynucleotide RNA, which was purchased from Clontech Laboratories (Palo Alto, CA) using triphosphate mixture, 2.5 ␮lof5␮M PCR primer pair for specific gene, 1 ␮l the Oligotex Direct mRNA kit (Qiagen, Inc., Valencia, CA). This “normal of ␤-actin primer pair, 1 ␮lof␤-actin competimers (Ambion, Inc., Austin, ␮ ␮ ␮ pancreata” consisted of a pool of two tissue specimens donated by two male TX), 37 lofH2O, and 0.5 l of 5 units/ l Taq polymerase (Promega Corp., Caucasians 18 and 40 years of age. Labeling and purification of cDNA probes Madison, WI). The amplification cycle (94°C for 30 s; 56°C for 45 s; and 72°C were carried out using the MICROMAX direct cDNA microarray system for 1 min) was repeated 29 times. PCR primers for individual genes were ϳ (NEN Life Science Products, Boston, MA). Two to 4 ␮g of the poly(A)ϩ RNA designed to generate a DNA fragment 600 bp in length (if the mRNA itself samples were used for each labeling. Probes for each pancreatic cell line were is less than 600 bases, PCR products were generated in maximal length) using labeled with Cy5,3 and probes for HeLa cells were labeled with Cy3. For HeLa the Primer3 program (25). cell versus normal pancreas hybridization, a normal pancreas sample was Northern Blot. RNA electrophoresis and transferring to Zeta-Probe GT membranes (Bio-Rad, Hercules, CA) were performed as described previously labeled with Cy3, and a HeLa cell sample was labeled with Cy5. Purified (24). 32P-labeled probes were made from the agarose gel-purified RT-PCR cDNA probes were dried and dissolved in 15 ␮l of hybridization buffer products of each gene using the RadPrime DNA Labeling System (Invitrogen). (included in the MICROMAX direct cDNA microarray system kit). The probes The probe hybridization and stripping buffers and conditions were as provided were then denatured by heating at 95°C for 2 min and applied to the array area by the membrane manufacturer. Hybridized membranes were exposed to a of a predenatured microarray slide. The microarray slide was covered with a PhosporImager (Molecular Dynamics, Sunnyvale, CA), and signals were quan- 22 ϫ 22-cm slide coverslip and incubated in a HybChamber (GeneMachines, tified using the ImageQuant software. San Carlos, CA) at 62°C for overnight. On the second day, the slide was Pancreatic Tumor Tissue Array Construction and Immunohistochem- ϫ ϫ washed in 0.5 SSC, 0.01% SDS for 5 min; 0.06 SSC, 0.01% SDS for 5 istry. Morphologically representative areas of 42 archival cases of pancreatic ϫ min; and 0.06 SSC for 2 min. Finally, the slide was dried by spinning at tumors, 35 of which are documented ductal adenocarcinomas, from the Uni- ϫ 500 g for 1 min and scanned in a dual-laser (635 nm for red fluorescent Cy5 versity of Arizona Health Sciences Center and the Tucson Veterans Admin- and 532 nm for green fluorescent Cy3) microarray scanner (GenePix 4000; istration Medical Center, were selected from formalin-fixed tissue samples Axon Instruments, Foster City, CA). embedded in paraffin blocks. Two 1.5-mm-diameter cores/case were re- Data Normalization and Analysis. Fluorescence intensities for both dyes embedded in a tissue microarray using a tissue arrayer (Beecher Instruments, (Cy3 and Cy5) and local background subtracted values for individual spots Silver Spring, MD) according to a method described previously (26). Serial were obtained using the GenePix 4000 microarray scanner and accompanying sections of the paraffin-embedded pancreatic tissue array were deparaffinized software (Axon Instruments). The data were imported into Microsoft Excel and reacted with primary antibodies specific for c-Myc (clone 9E10.3; Neo- spreadsheets for analysis. Defective spots, ones that are substandard on the Markers, Fremont, CA) or Rad51 (Oncogene, Boston, MA). Before antibody scanned image or have negative background subtracted values, were first incubation, the slides were processed for antigen retrieval. This consisted of excluded. To minimize the effects of measurement variations introduced by microwaving the slides in citrate buffer (0.1 M, pH 6.0) in a pressure cooker for artificial sources during experiments, we only included the spots that had 25 min and then were left to cool. The slides were incubated with the antibody significant signals in both channels. The determination of this significance is for 1 h. Biotinylated antimouse/antirabbit secondary antibodies were applied, based on signal intensities of nonhomologous ice plant genes. Generally, if the followed by streptavidin-peroxidase complex (DAKO, Carpinteria, CA). Col- signal intensity of a spot is less than the average of ice plant spots, we ored products were produced using the diaminobenzidine substrate. Staining considered the signal as nonsignificant. In this analysis, we empirically deter- reactions were scored as diffuse or focal and were graded (from 0, negative to mined the significance cutoff for signal:background ratio as 1.4. In other 4ϩ, intensely positive) for both neoplasm and background stroma. words, a spot was excluded from further analysis if it had a signal:background ratio of Ͻ1.4 in both channels. For each spot, the median of ratios (the median RESULTS of the pixel-by-pixel ratios of pixel intensities that have the median background intensity subtracted) was used in subsequent analysis. Spots represent- Differential Gene Expression in Pancreatic Cancer Cells. We ing housekeeping genes were used to normalize the entire slide so that all compared the gene expression of nine pancreatic cell lines with slides could be compared directly. For each pancreatic cell line, at least two normal pancreas using a cDNA microarray containing 5289 unique hybridizations were carried out. The average of median ratios from replicates genes. To identify genes that exhibited the most significant and was calculated for each spot. consistent expression changes in pancreatic cancer cells compared To ensure that the exact same reference samples were used for all necessary with normal pancreas, we normalized the gene expression ratios as experiments, we used a HeLa cell mRNA pool as our universal reference for described in “Materials and Methods.” We first selected outlying all microarray hybridizations. In other words, the Cy5-labeled probes for each genes that have an average expression ratio Ͼ2.0 SD from the mean. pancreatic cell line were mixed with Cy3-labeled probes for HeLa cells and This 2.0 SD cutoff represents an ϳ95% confidence interval. All cell hybridized to one slide to obtain the ratio of pancreatic cell line versus HeLa lines except SU.86.86 had about 30–50 genes that survived this cutoff cell. On the other hand, Cy5-labeled HeLa cell probes were mixed with at either end (overexpression or underexpression). SU.86.86 had ϳ10 Cy3-labeled normal pancreas probes and hybridized to a slide to obtain the genes that survived this cutoff. We then examined the expression ratio of HeLa cell versus normal pancreas. Each slide was normalized by the housekeeping genes; therefore, errors caused by hybridization differences from levels of individual outliers across the nine pancreatic cell lines and slide to slide are minimized. The two ratios were then multiplied to generate chose those that exhibited a significant expression change (2.0 SD the ratio of pancreatic cancer cell line versus normal pancreas. Finally, the from the mean) in at least three of the nine cell lines as our “true” ratios were taken as log2 transformation, and the SDs of the mean were then differentially expressed genes. There were a total of 58 genes that met calculated from these log2 ratios for each cell line. We used 2.0 SD as our those analysis criteria for differential expression (30 genes for over- cutoff for the determination of expression outliners (see “Results”). expression and 28 genes for underexpression). They accounted for ϳ1.1% of the 5289 unique cDNA clones included in the microarrays. 3 The abbreviations used are: Cy5, cyanine 5; Cy3, cyanine 3; RT-PCR, reverse These genes are listed in Table 1 (overexpressed genes) and Table 2 transcription-PCR; uPAR, urokinase-type plasminogen activator receptor; NGAL, neu- (underexpressed genes). From a drug design perspective, the overex- trophil gelatinase-associated lipocalin; Rho-GDI, Rho GDP dissociation inhibitor-␤; HMG, high mobility group; PCNA, proliferating cell nuclear antigen; STK15, serine/ pressed genes represent greater potential as targets for the develop- threonine kinase; NCA, nonspecific cross-reacting antigen. ment of small molecule inhibitors, whereas the underexpressed genes 2891

Table 1 Genes significantly overexpressed in pancreatic cell linesa Function and accession number Gene name Transcription and translation W87741 v-myc avian myelocytomatosis viral oncogene homologue (c-Myc)b AA448261 HMG protein isoforms I and Y (HMG I and Y)b T89996 Fos-like antigen-1 (Fos-like) H27379 Transcription elongation factor A (SII), (TEFA SII) AA599116 Small nuclear ribonucleoprotein polypeptides B and B1 (Sm B/BЈ) R53421 Membrane-bound transcription factor protease (MBTF protease) R54097 Human translational initiation factor 2, subunit 2 (TIF2) Cell adhesion and migration AA455222 uPARb R33355 Type I transmembrane protein Fn14 (Fn14) Calcium binding H63077 Annexin I (lipocortin I)b AA464731 S100 calcium-binding protein A11 (calgizzarin)b DNA replication and mitosis AA450265 PCNAb R19158 STK15b AA025937 Primase, polypeptide 1 (49 kD) AA397813 CDC28 protein kinase 2 (CKS2) DNA repair W00895 Rad51 (Saccharomyces cerevisiae) homologueb Others R61674 Protein tyrosine phosphatase type IVA, member 1 (PRL-1) AA054073 NCAb AA453335 Thioredoxin reductase 1 (TrxR-1) AA401137 NGALb AA865265 Cytochrome c R52654 Cytochrome c-1 AA465386 Human Gu protein mRNA H26176 FER-1 (Caenorhabditis elegans)-like 3 (myoferlin) AA455969 Prion protein (p27-30) AA487634 Rho-GDI N59721 Trinucleotide repeat containing 3 R23055 Expressed sequence tags R35245 Expressed sequence tags R96941 Expressed sequence tags a Genes with an expression ratio Ͼ2.0 SD from the mean in at least three of the nine pancreatic cell lines were considered as significantly overexpressed. b Genes reported previously as being overexpressed in pancreatic cancer. are possible candidates for gene therapy or other functional replace- Many of these genes, such as annexin I (27, 28), c-Myc (v-myc avian ment treatment. In this study, we focused on the validation of over- myelocytomatosis viral oncogene homologue; Ref. 29), uPAR (30), expressed genes. HMG protein isoforms I and Y (31, 32), PCNA (33), and nonspecific The 30 genes whose expressions were up-regulated in pancreatic cross-reacting antigen (34), are known to be up-regulated in cancers. cancer cells represented a variety of functional groups (Table 1). Genes that have been reported previously as being overexpressed in

Table 2 Genes significantly underexpressed in pancreatic cell linesa Accession number Gene name AA399410 Signal transducer and activator of transcription 3 H85962 Mitogen-activated protein kinase kinase 7 N79230 Human MAC30 mRNA, 3Ј end W25035 Human FK-506 binding protein 8 H72027 Gelsolin precursor R16957 H-2K binding factor-2 AA485377 v-fos FBJ murine osteosarcoma viral oncogene homologue H99813 Glutathione S-transferase ␪ 1 R25020 Human APEG-1 mRNA AA480851 Claudin 10 AA486261 Signal sequence receptor, ␦ AA459051 Death-associated protein 1 N54420 Rho guanine exchange factor (GEF) 11 H50677 RNA binding motif protein 6 H59915 CD24 antigen R75819 FK506-binding protein 2 (13 kD) R20379 Eukaryotic translation elongation factor 2 AA460986 Human GPI-H mRNA AA452872 Human GCN5 (hGCN5) gene AA486653 CD81 antigen AA570216 Expressed sequence tags, moderately similar to AF078844 1 hqp0376 protein AA701502 Platelet-derived growth factor ␣ polypeptide AA496691 Dystroglycan 1 (dystrophin-associated glycoprotein 1) H94897 AD-017 protein AA013336 Human OS-9 precursor mRNA H06675 Human NADϩ-dependent succinate-semialdehyde dehydrogenase AA458472 Human MHC class II HLA-DR2-Dw12 AA063521 BCL2/adenovirus E1B 19 kD-interacting protein 3 a Genes with an expression ratio Ͻ2.0 SD from the mean in at least three of the nine pancreatic cell lines were considered significantly underexpressed. 2892

Table 3 Expression ratios of the up-regulated genes based on microarray analysis Expression ratio of cell line vs. normal pancreasa

Gene name Accession no. AsPC-1 BxPC-3 Capan-1 CFPAC-1 HPAF II Mia PaCa-2 Mutj Panc-1 Su8686 uPAR AA455222 8.6 3.2 2.2 10.3 0.9 6.6 6.6 NCA AA054073 38.3 40 11.4 3 PCNA AA450265 3 7.1 6.7 4.6 3.2 3.5 4.7 6.9 3.3 S100 calcium-binding protein A11 (calgizzarin) AA464731 4.9 5.6 9.5 4.5 3.3 3.7 Protein tyrosine phosphatase type IVA, member 1 R61674 6 6.4 6.5 4.4 NGAL AA401137 27.6 6 4.2 36 0.7 0.3 33.5 v-myc avian myelocytomatosis viral oncogene homologue W87741 5 7.3 4.6 5.9 4.2 2.8 3.4 Rad51 (Saccharomyces cerevisiae) homologue W00895 11.3 6.6 5.5 5 4.5 2.9 13.7 10 Annexin I (lipocortin I) H63077 5.1 10.7 11.7 3.8 18.3 23.3 9.5 2.3 12 HMG protein isoforms I and Y AA448261 12.1 6.3 4.3 5.3 14 4.6 11.5 Fos-like antigen-1 T89996 7.2 6.6 3.6 16.8 5.8 10.8 10.1 Transcription elongation factor A (SII), 1 H27379 5.3 2.5 3.6 3.8 2.4 2.1 2.4 6.2 3.4 Small nuclear ribonucleoprotein polypeptides B and B1 AA599116 3.9 5.4 4.8 2 6.7 2.6 2.8 3.7 1.7 Membrane-bound transcription factor protease, site 2 R53421 5.3 4.6 3.5 2.6 21.2 Human translational initiation factor 2, ␤ R54097 4.2 4.6 3.6 5.5 4.1 2.4 5.8 6 Type I transmembrane protein Fn14 R33355 12.1 8.6 9.5 10.4 1.3 8.1 Primase, polypeptide 1 (49 kD) AA025937 8.2 7.9 7.7 8.7 2.9 7.8 8.8 Thioredoxin reductase 1 AA453335 8.5 1.1 1.7 6.1 9.1 3.6 4.4 7.6 Cytochrome c AA865265 6.6 7.4 10 1.3 11.4 3.9 12.9 17.4 13.3 Cytochrome c-1 R52654 7.3 8.6 5.6 5.2 6.8 3 13.7 18.3 13.4 Prion protein (p27-30) AA455969 2.1 5.2 4.3 1.8 7.7 4.1 2.5 Fer-1 (Caenorhabditis elegans)-like 3 (myoferlin) H26176 17.1 11.2 9.1 9.3 7.7 7 6.9 9.2 Rho GDI AA487634 0.7 6.3 6.2 1.5 0.2 0.7 0.7 8 CDC28 protein kinase 2 AA397813 18.3 41 19.7 6 24.7 25.3 24.1 23.8 11.3 STK15 R19158 4.9 4.4 3.7 2.9 2.7 5.9 7.2 Human Gu protein mRNA AA465386 6.4 7 5.4 7.6 5.1 3.8 8.6 15.7 Trinucleotide repeat containing 3 N59721 23.7 2.8 1.5 17.3 27.8 Expressed sequence tags R23055 9.8 11.3 3.4 14.8 3.3 13.9 18.3 Expressed sequence tags R96941 3.5 4.4 5.6 2.5 3.1 5.2 5.3 Expressed sequence tags R35245 5.9 5.5 6.2 1.7 9.3 2 6.7 a Blank cells indicate that data points represented did not pass the microarray spot quality control during data analysis.

pancreatic cancer are noted in Table 1. There were also genes that are 3). Rho-GDI was overexpressed by 6–8-fold in BxPC-3, Capan-1, not reported previously, including prion protein, myoferlin, and pri- and SU.86.86 cell lines, whereas in AsPC-1, MIA PaCa-2, Mutj, and mase polypeptide 1. However, the identification of cytochrome c and PANC-1, it was down-regulated to 0.2–0.7-fold of normal pancreas Rad51 as up-regulated in pancreatic cancer cells is unexpected be- (Table 3). cause cytochrome c is a key proapoptotic molecule in mitochondria Our RT-PCR and Northern blot analysis tended to agree with the (35), and Rad51 is involved in DNA double-strand break repair (36). expression patterns generated by microarray data. As shown in Figs. Other pancreatic cancer-associated genes reported previously showed 1 and 2, RT-PCR and Northern results demonstrate an increase in some evidence of up-regulation in our microarray experiments but expression for most genes in all nine cell lines. Quantitation of the failed to meet our strict criterion of 2.0 SD of the mean in at least three RT-PCR and Northern blots showed that the fold of increase was in cell lines. the same range as expected from microarray data with the exception Table 2 lists the genes that were down-regulated in pancreatic of uPAR in the MIA PaCa-2 cell line (quantitation data not shown). cancer cells. A few of these have been reported previously to be The microarray data predicted that uPAR expression would be slightly down-regulated in cancers. For instance, down-regulation of gelsolin lower in MIA PaCa-2 cells than in normal pancreas (expression ratio, was found to correlate with progression to breast carcinoma (37), lack 0.9; Table 3). Both RT-PCR and Northern blot analysis, however, of glutathione S-transferase ␪ is considered to be related to increased showed increased expression in MIA PaCa-2 cells (4.3-fold in RT- risk of various cancers (38), and the death-associated proteins, a new PCR and 2.4-fold in Northern blot). Intriguingly, the expression class of proapoptotic molecules, have been shown to be tumor sup- patterns of NGAL and Rho-GDI are correlated very well with the pressive (39–41). Many of these genes, however, were either not microarray data. In both RT-PCR (Fig. 1) and Northern blots (Fig. 2), considered previously to be cancer-associated or not functionally NGAL is highly up-regulated in AsPC-1, Capan-1, CFPAC-1, HPAF characterized. II, and Su.86.86 but not in Mutj and MIA PaCa-2. Quantitation of the RT-PCR and Northern Blot Validation of Overexpressed Northern blots showed a slight decrease of NGAL expression in both Genes. To validate the microarray data, we further investigated the Mutj and MIA PaCa-2 cell lines (quantitation data not shown). The mRNA levels of the genes that exhibited up-regulated expression on expression pattern of Rho-GDI shown by RT-PCR and Northern blot cDNA microarray using quantitative RT-PCR and Northern blots. also matches the pattern shown by microarray analysis with up- Among the 30 up-regulated genes, 25 had at least some known regulation in BxPC-3, Capan-1, and SU.86.86 and down-regulation in functions and were selected for RT-PCR and Northern blot analysis. AsPC-1, MIA PaCa-2, Mutj, and PANC-1 (Figs. 1 and 2; quantitation Although the expression ratios calculated from microarray data varied of RT-PCR and Northern blot are not shown). Noticeably, the RT- in levels, most of the 25 genes explored by RT-PCR and Northern PCR and Northern blot agree for most of the genes. blots showed overexpression in all nine pancreatic cancer cell lines Expression of Rad51 and c-Myc in Patient Samples. To verify (Table 3). A few genes, however, such as NGAL and Rho-GDI, the gene overexpression in patient samples, we chose two genes exhibited very dramatic expression variations across the cell lines. identified by the cDNA microarray, c-Myc and Rad51, to undergo NGAL was up-regulated by Ͼ27-fold in three cell lines (AsPC-1, RT-PCR analysis using RNA isolated from frozen pancreatic tumor HPAF II, and SU.86.86). Its expression in MIA PaCa-2 and Mutj was tissues. Fig. 3 shows the RNA levels of c-Myc and Rad51 in eight down-regulated to 0.7 and 0.3 of normal pancreas, respectively (Table pancreatic tumor samples. c-Myc is highly overexpressed in patient 2893

pression of these genes was not unexpected because they are involved in the regulation of cell growth and proliferation. Cell adhesion and migration-related genes were another major group of genes overexpressed in pancreatic cancer cells. This group includes uPAR (42) and transmembrane protein Fn14 (43). Two calcium-binding proteins, S100A11 (calgizzatin) and annexin I, were also highly expressed in pancreatic cancer cells. Both genes have been reported previously to be up-regulated in tumor cells (44–46). Other groups of genes that were found up-regulated and have known functions include DNA replication and mitosis-related genes such as primase 1, PCNA, STK15, CDC28 protein kinase 2, and DNA repair gene Rad51. STK15 encodes a centrosome-associated kinase and is thought to be important for centrosome duplication and distri- bution (47). CDC28 protein kinase 2 (cyclin-dependent kinase subunit 2) binds to cyclin-dependent kinases, but its precise function is un- clear. Recently, its expression has been linked to human lymphoid cell proliferation (48). The identification of Rad51 is rather surprising because one would not expect a DNA repair gene to be overexpressed in cancer cells. It is possible, however, that deregulated or increased DNA repair results in genomic instability and therefore tumorigenesis. In fact, Maacke et al. (49, 50) reported that Rad51 is overexpressed in human pancreatic cancer, and the overexpression of wild-type Rad51 Fig. 1. RT-PCR validation of the 25 up-regulated genes. Top band in each panel, product of the testing gene as labeled on the left. Bottom band, ␤-actin as internal control. correlates with histological grading of invasive ductal breast cancer. The ␤-actin amplification was carried out in the same reaction tube for all testing genes except PRL-1, for which ␤-actin reaction was performed in a separate tube with the same amount of reverse transcriptase products as for PRL-1 reactions. DNA was resolved on 1% agarose gels and visualized by ethidium bromide staining. Lane 1, normal pancreas; lane 2, AxPC-1; lane 3, BxPC-3; lane 4, Capan-1; lane 5, CFPAC-1; lane 6, HPAF II; lane 7, MIA PaCa-2; lane 8, Mutj; lane 9, PANC-1; lane 10, SU.86.86. The gene abbreviations are listed in Table 1.

samples 6 and 8 and moderately overexpressed in patient samples 4 and 5 (Fig. 3, upper panel; quantitation not shown). Therefore, the overall overexpression rate for the c-Myc gene in tumor tissues is ϳ50%. Rad51 was overexpressed in three tissue samples (Fig. 3, patients 3, 4, and 8), one of which (patient 4) had Ͼ11-fold up- regulation (Fig. 3, lower panel; quantitation not shown). To examine whether the overexpression at the RNA level translated to overexpression at the protein level, we carried out immunohistochemical studies on c-Myc and Rad51 genes with a paraffin-embedded pancreatic tumor microtissue array. The tissue array contained 35 different pancreatic adenocarcinoma samples collected and archived Fig. 2. Northern blot validation of the 25 up-regulated genes. Ten ␮g of total RNA in our institutions (Fig. 4A). Fig. 4, B and C, showed some examples were used for each sample. Hybridizations of the genes were carried out on six blots by stripping off the probes after each hybridization. ␤-Actin was used to show the difference of positive staining of c-Myc and Rad51. The overall c-Myc-positive in RNA loading for different blots. The blot used to hybridize with uPAR and nonspecific staining (tumor score is at least 1ϩ over background stroma score; see cross-reacting antigen (NCA) probes had a different batch of total RNA from that of the “Materials and Methods”) frequency is 71.8%. About 15.6% of the other five blots. These other five blots were prepared at the same time using the same batch of total RNA and had the same ␤-actin hybridization pattern. Therefore, only three tumors showed very strong c-Myc staining (scored 3ϩ or over). ␤-actin hybridizations are shown. Lane 1, normal pancreas; lane 2, AxPC-1; lane 3, Rad51 has a positive staining frequency of 74.2%, with 12.9% of BxPC-3; lane 4, Capan-1; lane 5, CFPAC-1; lane 6, HPAF II; lane 7, MIA PaCa-2; ϩ lane 8, Mutj; lane 9, PANC-1; lane 10, SU.86.86. The gene abbreviations are listed tumors showing very strong staining (scored 3 or more). These in Table 1. results were consistent with the RNA expression data shown by RT-PCR (Fig. 3).

DISCUSSION In the present study, we used a cDNA microarray to explore the variations in gene expression between pancreatic cancer cells and normal pancreas on a genomic scale. Thirty genes were identified as significantly up-regulated in pancreatic cancer cells. These genes can be categorized into several groups on the basis of their known functions (Table 1). The first major group of genes was transcription or translation-related genes and included transcription factors such as Fig. 3. Expression of c-Myc and Rad51 in pancreatic tumor samples. Two ␮g of total RNA isolated from frozen patient tissues were reverse-transcribed and PCR amplified c-Myc, HMG protein isoforms I and Y, fos-like antigen-1 (31), and with c-Myc-orRad51-specific primers. ␤-Actin was amplified in the same reaction tube transcriptional or translational machinery-related genes. The overex- as c-Myc or Rad51 as internal control. 2894

Fig. 4. Immunostaining of c-Myc and Rad51 on pancreatic tumor tissue microarray. A, overview of the pancreatic tissue microarray. B, magnified view of a tissue microarray disk with positive c-Myc staining. C, magnified view of a tissue microarray disk with positive Rad51 staining.

RT-PR and tissue microarray immunohistochemistry studies in believe that the genes recorded in Table 1 represent a list of genes clinical tumor samples confirmed that c-Myc and Rad51 were over- potentially important to the tumorigenesis process in the pancreas. expressed in a majority of pancreatic cancers (Figs. 3 and 4). These Within the set of genes we identified, some may represent potential findings were important because only if the genes are overexpressed tumor markers or drug targets. At present, we are evaluating some of in clinical samples and at the protein level are they potential targets these genes as potential candidates for targeted drug development. for screening and therapeutic development. These results also indicate that findings in cell lines can be verified in tissue samples, and gene ACKNOWLEDGMENTS expression at the RNA level can be translated to expression at the protein level. Tissue microarray is a recently developed technology We thank Dr. George Watts and Mary Guzman for technical assistance on that enables the simultaneous examination of multiple histological microarray experiments and Kimiko Della Croce for help on cell culture. We sections at one time as compared with one section at a time for the also thank Dr. Ronald Schifman and the Tucson Veterans Administration conventional method (26). In this study, we constructed a pancreatic Medical Center for authorization to include five tumor specimens in the tissue tumor tissue microarray that contained 35 different adenocarcinoma microarray. tissue samples, each of which having two representative 1.5-mm disks from the different areas of the same paraffin-embedded section. Im- REFERENCES munohistochemical studies using the whole tumor sections demon- 1. Mauro, M. J., and Druker, B. J. STI571: a gene product-targeted therapy for leukemia. strated that the two 1.5-mm disks were highly representative of the Curr. Oncol. Rep., 3: 223–227, 2001. 2. Goggins, M., Kern, S. E., Offerhaus, J. A., and Hruban, R. H. Progress in cancer tissues from which they originated (data not shown). We are now genetics: lessons from pancreatic cancer. Ann. Oncol., 10 (Suppl. 4): 4–8, 1999. using the same pancreatic tissue microarray to verify the overexpres- 3. Manu, M., Buckels, J., and Bramhall, S. Molecular technology and pancreatic cancer. sion of other genes manifested in the cDNA microarray study. Br. J. Surg., 87: 840–853, 2000. 4. Kern, S. E. Molecular genetic alterations in ductal pancreatic adenocarcinomas. Med. Cancer cells from different tumor types share many features at both Clin. North Am., 84: 691–695, 2000. the cellular and the molecular levels. However, because of their nature 5. Sakorafas, G. H., Tsiotou, A. G., and Tsiotos, G. G. Molecular biology of pancreatic of increased genetic instability, human cancers are highly heteroge- cancer; oncogenes, tumour suppressor genes, growth factors, and their receptors from a clinical perspective. Cancer Treat. Rev., 26: 29–52, 2000. neous, even within a single tumor. To identify genes that are most 6. Hruban, R. H., Offerhaus, G. J. A., Kern, S. E., Goggins, M., Wilentz, R. E., and Yeo, consistently and frequently overexpressed during pancreatic tumori- C. J. Tumor-suppressor genes in pancreatic cancer. J. Hepato-Biliary-Pancreatic Surg., 5: 383–391, 1998. genesis, we used stringent criteria in our microarray data analysis as 7. Dong, M., Nio, Y., Tamura, K., Song, M. M., Guo, K. J., Guo, R. X., and Dong, Y. T. described above. Although we may have missed some genes, we Ki-ras point mutation and p53 expression in human pancreatic cancer: a comparative 2895

study among Chinese, Japanese, and Western patients. Cancer Epidemiol. Biomark. 29. Nesbit, C. E., Tersak, J. M., and Prochownik, E. V. MYC oncogenes and human Prev., 9: 279–284, 2000. neoplastic disease. Oncogene, 18: 3004–3016, 1999. 8. Hu, Y. X., Watanabe, H., Ohtsubo, K., Yamaguchi, Y., Ha, A., Okai, T., and Sawabu, 30. Andreasen, P. A., Egelund, R., and Petersen, H. H. The plasminogen activation N. Frequent loss of p16 expression and its correlation with clinicopathological system in tumor growth, invasion, and metastasis. Cell. Mol. Life Sci., 57: 25–40, parameters in pancreatic carcinoma. Clin. Cancer Res., 3: 1473–1477, 1997. 2000. 9. Iacobuzio-Donahue, C. A., Klimstra, D. S., Adsay, N. V., Wilentz, R. E., Argani, P., 31. Reeves, R., Edberg, D. D., and Li, Y. Architectural transcription factor HMGI(Y) Sohn, T. A., Yeo, C. J., Cameron, J. L., Kern, S. E., and Hruban, R. H. Dpc-4 protein promotes tumor progression and mesenchymal transition of human epithelial cells. is expressed in virtually all human intraductal papillary mucinous neoplasms of the Mol. Cell. Biol., 21: 575–594, 2001. pancreas: comparison with conventional ductal adenocarcinomas. Am. J. Pathol., 32. Giannini, G., Kim, C. J., Marcotullio, L. D., Manfioletti, G., Cardinali, B., Cerignoli, 157: 755–761, 2000. F., Ristori, E., Zani, M., Frati, L., Screpanti, I., and Guilino, A. Expression of the 10. Hruban, R. H., Petersen, G. M., Goggins, M., Tersmette, A. C., Offerhaus, G. J., HMGI(Y) gene products in human neuroblastic tumours correlates with differentia- Falatko, F., Yeo, C. J., and Kern, S. E. Familial pancreatic cancer. Ann. Oncol., 10 tion status. Br. J. Cancer, 83: 1503–1509, 2000. (Suppl. 4): 69–73, 1999. 33. Elias, J. M. Cell proliferation indexes: a biomarker in solid tumors. Biotech. Histo- 11. Goggins, M., Shekher, M., Turnacioglu, K., Yeo, C. J., Hruban, R. H., and Kern, S. E. chem., 72: 78–85, 1997. Genetic alterations of the transforming growth factor ␤ receptor genes in pancreatic 34. Kuroki, M., Matsushita, H., Matsumoto, H., Hirose, Y., Senba, T., and Yamamoto, T. and biliary adenocarcinomas. Cancer Res., 58: 5329–5332, 1998. Nonspecific cross-reacting antigen-50/90 (NCA-50/90) as a new tumor marker. 12. Su, G. H., Hilgers, W., Shekher, M. C., Tang, D. J., Yeo, C. J., Hruban, R. H., and Anticancer Res., 19: 5599–5606, 1999. Kern, S. E. Alterations in pancreatic, biliary, and breast carcinomas support MKK4 as 35. Gottlieb, R. A. Mitochondria: execution central. FEBS Lett., 482: 6–12, 2000. a genetically targeted tumor suppressor gene. Cancer Res., 58: 2339–2342, 1998. 13. Su, G. H., Hruban, R. H., Bansal, R. K., Bova, G. S., Tang, D. J., Shekher, M. C., 36. Shinohara, A., and Ogawa, T. Rad51/RecA protein families and the associated Westerman, A. M., Entius, M. M., Goggins, M., Yeo, C. J., and Kern, S. E. Germline proteins in eukaryotes. Mutat. Res., 435: 13–21, 1999. and somatic mutations of the STK11/LKB1 Peutz-Jeghers gene in pancreatic and 37. Winston, J. S., Asch, H. L., Zhang, P. J., Edge, S. B., Hyland, A., and Asch, B. B. biliary cancers. Am. J. Pathol., 154: 1835–1840, 1999. Downregulation of gelsolin correlates with the progression to breast carcinoma. 14. Blanck, H. M., Tolbert, P. E., and Hoppin, J. A. Patterns of genetic alterations in Breast Cancer Res. Treat., 65: 11–21, 2001. ␪ pancreatic cancer: a pooled analysis. Environ. Mol. Mutagen., 33: 111–122, 1999. 38. Landi, S. Mammalian class GST and differential susceptibility to carcinogens: a 15. Day, J. D., Digiuseppe, J. A., Yeo, C., Lai-Goldman, M., Anderson, S. M., Goodman, review. Mutat. Res., 463: 247–283, 2000. S. N., Kern, S. E., and Hruban, R. H. Immunohistochemical evaluation of HER-2/neu 39. Kissil, J. L., and Kimchi, A. Death-associated proteins: from gene identification to the expression in pancreatic adenocarcinoma and pancreatic intraepithelial neoplasms. analysis of their apoptotic and tumour suppressive functions. Mol. Med. Today, 4: Hum. Pathol., 27: 119–124, 1996. 268–274, 1998. 16. Yip-Schneider, M. T., Barnard, D. S., Billings, S. D., Cheng, L., Heilman, D. K., Lin, 40. Inbal, B., Shani, G., Cohen, O., Kissil, J. L., and Kimchi, A. Death-associated protein A., Marshall, S. J., Crowell, P. L., Marshall, M. S., and Sweeney, C. J. Cyclooxy- kinase-related protein 1, a novel serine/threonine kinase involved in apoptosis. Mol. genase-2 expression in human pancreatic adenocarcinomas. Carcinogenesis (Lond.), Cell. Biol., 20: 1044–1054, 2000. 21: 139–146, 2000. 41. Kogel, D., Prehn, J. H., and Scheidtmann, K. H. The DAP kinase family of pro- 17. von Marschall, Z., Cramer, T., Hocker, M., Burde, R., Plath, T., Schirner, M., apoptotic proteins: novel players in the apoptotic game. Bioessays, 23: 352–358, Heidenreich, R., Breier, G., Riecken, E. O., Wiedenmann, B., and Rosewicz, S. De 2001. novo expression of vascular endothelial growth factor in human pancreatic cancer: 42. Rabbani, S. A., and Xing, R. H. Role of urokinase (uPA) and its receptor (uPAR) in evidence for an autocrine mitogenic loop. Gastroenterology, 119: 1358–1372, 2000. invasion and metastasis of hormone-dependent malignancies. Int. J. Oncol., 12: 18. Peltomaki, P. DNA mismatch repair and cancer. Mutat. Res., 488: 77–85, 2001. 911–920, 1998. 19. Yamamoto, H., Itoh, F., Nakamura, H., Fukushima, H., Sasaki, S., Perucho, M., and 43. Meighan-Mantha, R. L., Hsu, D. K., Guo, Y., Brown, S. A., Feng, S. L., Peifley, Imai, K. Genetic and clinical features of human pancreatic ductal adenocarcinomas K. A., Alberts, G. F., Copeland, N. G., Gilbert, D. J., Jenkins, N. A., Richards, C. M., with widespread microsatellite instability. Cancer Res., 61: 3139–3144, 2001. and Winkles, J. A. The mitogen-inducible Fn14 gene encodes a type I transmembrane 20. Carulli, J. P., Artinger, M., Swain, P. M., Root, C. D., Chee, L., Tulig, C., Guerin, J., protein that modulates fibroblast adhesion and migration. J. Biol. Chem., 274: Osborne, M., Stein, G., Lian, J., and Lomedico, P. T. High throughput analysis of 33166–33176, 1999. differential gene expression. J. Cell. Biochem. Suppl., 30–31: 286–296, 1998. 44. Stulik, J., Koupilova, K., Osterreicher, J., Knizek, J., Macela, A., Bures, J., Jandik, P., 21. van Hal, N. L., Vorst, O., van Houwelingen, A. M., Kok, E. J., Peijnenburg, A., Langr, F., Dedic, K., and Jungblut, P. R. Protein abundance alterations in matched Aharoni, A., van Tunen, A. J., and Keijer, J. The application of DNA microarrays in sets of macroscopically normal colon mucosa and colorectal carcinoma. Electro- gene expression analysis. J. Biotechnol., 78: 271–280, 2000. phoresis, 20: 3638–3646, 1999. 22. Pandey, A., and Mann, M. Proteomics to study genes and genomes. Nature (Lond.), 45. Van Ginkel, P. R., Gee, R. L., Walker, T. M., Hu, D. N., Heizmann, C. W., and 405: 837–846, 2000. Polans, A. S. The identification and differential expression of calcium-binding pro- 23. Korc, M., Meltzer, P., and Trent, J. Enhanced expression of epidermal growth factor teins associated with ocular melanoma. Biochim. Biophys. Acta, 1448: 290–297, receptor correlates with alterations of chromosome 7 in human pancreatic cancer. 1998. Proc. Natl. Acad. Sci. USA, 83: 5141–5144, 1986. 46. Ahn, S. H., Sawada, H., Ro, J. Y., and Nicolson, G. L. Differential expression of 24. Calaluce, R., Kunkel, M. W., Watts, G. S., Schmelz, M., Hao, J., Barrera, J., annexin I in human mammary ductal epithelial cells in normal and benign and Gleason-Guzman, M., Isett, R., Fitchmun, M., Bowden, G. T., Cress, A. E., Futscher, B. W., and Nagle, R. B. Laminin-5-mediated gene expression in human prostate malignant breast tissues. Clin. Exp. Metastasis, 15: 151–156, 1997. carcinoma cells. Mol. Carcinog., 30: 119–129, 2001. 47. Zhou, H., Kuang, J., Zhong, L., Kuo, W. L., Gray, J. W., Sahin, A., Brinkley, B. R., 25. Rozen, S., and Skaletsky, H. J. Code available at http://www-genome.wi.mit.edu/ and Sen, S. Tumour amplified kinase STK15/BTAK induces centrosome amplifica- genome_software/other/primer3.html. Primer3, 1998. tion, aneuploidy and transformation. Nat. Genet., 20: 189–193, 1998. 26. Kononen, J., Bubendorf, L., Kallioniemi, A., Barlund, M., Schraml, P., Leighton, S., 48. Urbanowicz-Kachnowicz, I., Baghdassarian, N., Nakache, C., Gracia, D., Mekki, Y., Torhorst, J., Mihatsch, M. J., Sauter, G., and Kallioniemi, O. P. Tissue microarrays Bryon, P. A., and Ffrench, M. ckshs expression is linked to cell proliferation in for high-throughput molecular profiling of tumor specimens. Nat. Med., 4: 844–847, normal and malignant human lymphoid cells. Int. J. Cancer, 82: 98–104, 1999. 1998. 49. Maacke, H., Jost, K., Opitz, S., Miska, S., Yuan, Y., Hasselbach, L., Luttges, J., 27. Vaarala, M. H., Porvari, K., Kyllonen, A., and Vihko, P. Differentially expressed Kalthoff, H., and Sturzbecher, H. W. DNA repair and recombination factor Rad51 is genes in two LNCaP prostate cancer cell lines reflecting changes during prostate over-expressed in human pancreatic adenocarcinoma. Oncogene, 19: 2791–2795, cancer progression. Lab. Investig., 80: 1259–1268, 2000. 2000. 28. Emmert-Buck, M. R., Gillespie, J. W., Paweletz, C. P., Ornstein, D. K., Basrur, V., 50. Maacke, H., Opitz, S., Jost, K., Hamdorf, W., Henning, W., Kruger, S., Feller, A. C., Appella, E., Wang, Q. H., Huang, J., Hu, N., Taylor, P., and Petricoin, E. F., III. An Lopens, A., Diedrich, K., Schwinger, E., and Sturzbecher, H. W. Over-expression of approach to proteomic analysis of human tumors. Mol. Carcinog., 27: 158–165, wild-type Rad51 correlates with histological grading of invasive ductal breast cancer. 2000. Int. J. Cancer, 88: 907–913, 2000.

2896

Downloaded from cancerres.aacrjournals.org on September 24, 2021. © 2002 American Association for Cancer Research. Identification of Differentially Expressed Genes in Pancreatic Cancer Cells Using cDNA Microarray

Haiyong Han, David J. Bearss, L. Walden Browne, et al.

Cancer Res 2002;62:2890-2896.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/62/10/2890

Cited articles This article cites 47 articles, 9 of which you can access for free at: http://cancerres.aacrjournals.org/content/62/10/2890.full#ref-list-1

Citing articles This article has been cited by 51 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/62/10/2890.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/62/10/2890. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.