oncogenomics - Nature

14 downloads 254 Views 382KB Size Report
Gene expression profiles of pancreatic cancer and stromal desmoplasia ...... II reverse transcription kit (Gibco-BRL, Life Technologies, Paisley, Scotland). The.
Oncogene (2001) 20, 7437 ± 7446 ã 2001 Nature Publishing Group All rights reserved 0950 ± 9232/01 $15.00 www.nature.com/onc

Gene expression pro®les of pancreatic cancer and stromal desmoplasia Tatjana Crnogorac-Jurcevic1,5, Evangelis Efthimiou1,5, Paola Capelli3, Ekaterina Blaveri1, Antonella Baron3, Benoit Terris1, Melanie Jones1, Kerry Tyson2, Claudio Bassi4, Aldo Scarpa3 and Nicholas R Lemoine*,1 Imperial Cancer Research Fund Molecular Oncology Unit, Imperial College School of Medicine at Hammersmith Campus, London, UK; 2Oxford Glycosciences Proteome Discovery, Abingdon, UK; 3Department of Pathology, UniversitaÁ di Verona, Italy; 4 Department of Surgery, UniversitaÁ di Verona, Italy

Gene expression studies were undertaken in normal pancreas and pancreatic adenocarcinomas to determine new candidate genes that can potentially be used as markers of the disease. The characteristic desmoplastic stromal reaction of pancreatic adenocarcinoma greatly hampers expression studies in this tumour type, and usually necessitates time-consuming tissue microdissection for enrichment of the tumour cell population. We show that ®ne needle aspiration of cancer provides a fast and ecient way of obtaining samples highly enriched in tumour cells with sucient yields of RNA. Using Atlas cancer cDNA arrays with 588 cancer-related genes, we describe gene expression pro®les of normal pancreas, bulk pancreatic tumour tissues and pancreatic tumour aspirates containing more than 95% tumour cells. Analysis of bulk tissue specimens revealed di€erentially expressed genes belonging predominantly to the stromal component of the tumour. This contrasted with the results obtained from tumour-cell enriched samples. Several genes already described in pancreatic cancer (caspase 8, TIMP1, CD9, IL-13) were also di€erentially expressed in our study. Furthermore, we found dysregulated expression of genes not previously associated with pancreatic adenocarcinoma, such as Rac 1, GLG1, NEDD5, RPL-13a, RPS9 and members of the Wnt5A gene family. In summary, we present a panel of genes newly identi®ed in the pathogenesis of pancreatic adenocarcinoma and demonstrate that ®ne needle aspirates of the tumour mass are a convenient source of material for gene expression studies in tumours accompanied by desmoplastic reactions. Oncogene (2001) 20, 7437 ± 7446. Keywords: pancreatic cancer; microarray; expression pro®ling; desmoplasia

*Correspondence: NR Lemoine, ICRF Molecular Oncology Unit, Department of Cancer Medicine, Imperial College School of Medicine, Hammersmith Campus, London W12 0NN, UK; E-mail: [email protected] 5 The ®rst two authors contributed equally to this work Received 13 August 2001; revised 28 August 2001; accepted 28 August 2001

Introduction Large scale gene expression pro®ling using oligonucleotide or cDNA arrays has become one of the most fruitful methods for characterizing physiological and pathological processes. It has been used in a variety of applications, from annotation of the human genome to tumour classi®cation that might predict prognosis and response to treatment (Alizadeh et al., 2000; Golub et al., 1999; Perou et al., 2000; Ross et al., 2000; Shoemaker et al., 2001). Pancreatic adenocarcinoma poses particular problems for gene expression studies. This malignancy is characterized by a dense desmoplastic stromal component, with neoplastic cells often representing a minor population in many parts of the tumour mass. To date, two di€erent approaches have been used to compensate for the dilution of the neoplastic element. One previous study (Gress et al., 1996) used hybridization to ®lters carrying cDNA clones derived from pancreatic cancer cell lines in an attempt to restrict the expression pro®le to genes more likely to be derived from the malignant epithelial component of the tumour. The other used a mixture of RNAs from chronic pancreatitis and health pancreas as the driver population in representational di€erence analysis (Gress et al., 1997). Recently, the laser-capture microdissection technique for procuring individual components of complex tissues has been designed (Emmert-Buck et al., 1996; Sirivatanauksorn et al., 1999), but has proven to be labour-intensive and time-consuming. In addition, the amount of material generated by microdissection is limited and, without ampli®cation procedures, usually not sucient for gene expression studies. In this study we have therefore explored an alternative approach for enrichment of neoplastic cells by using ®ne needle aspiration from the tumour mass. We show that it is a practical and convenient approach for generating suitable source material with high quality and quantity of nucleic acids for downstream molecular analysis. Adenocarcinoma of the exocrine pancreas is the most common type of pancreatic malignancy. It often presents symptoms only when it is already in an advanced state and curative resection is no longer possible. Detailed knowledge of genes involved in the growth of pancreatic adenocarcinoma might enable

ONCOGENOMICS

1

Gene expression in pancreatic adenocarcinoma T Crnogorac-Jurcevic et al

7438

construction of molecular tests for earlier diagnosis (Lemoine, 1997) and identify new targets for therapy (Aspinall and Lemoine, 1999). We have therefore examined expression pro®les in both the enriched malignant epithelial component of the tumour and in the bulk cancerous tissue consisting predominantly of stromal elements in order to build up a more complete picture of the network of events involved in pancreatic tumorigenesis.

Results Analysis of bulk tumours including stromal components Gene expression in bulk adenocarcinoma samples was compared to that of normal pancreas. Di€erentially expressed genes that are common in hybridization with all six bulk specimens vs normal pancreas are listed in Table 1. The highest upregulation was displayed by collagen, vimentin and decorin, and the most marked downregulation by ribosomal protein L13a and interleukin 13. Analysis of samples selectively enriched for tumour component Enrichment of the samples with tumour cells was achieved by the ®ne needle aspiration technique described in Materials and methods, which produced samples containing around 95% cytokeratin-positive tumour cells (Figure 1). Comparative hybridization analysis of tumour cellenriched samples against normal tissue showed a di€erent pattern of gene expression between the aspirated cancer cells and the normal pancreas (Table 2, Figure 2), that was distinct from the one obtained when comparing bulk tumours with normal tissue (Table 1). Vimentin, decorin and CD9 showed no di€erential expression after enrichment of samples with tumour cells. Some genes that had di€erentials below the twofold cut-o€ in the analysis of bulk tissue

samples displayed higher up- or down-regulation in the analysis of the puri®ed cell population (Golgi membrane sialoglycoprotein (MG160/GLG1), matrix metalloproteinase 16, leukocyte interferon-inducible peptide). Several genes already known through candidate gene analysis to be upregulated in pancreatic cancer were con®rmed in our analysis, including caspase 8 and tissue inhibitor of matrix metalloproteinase 1 (TIMP1). Some genes showed similar di€erentials in the analysis of both whole-tissue homogenates and tumour cell aspirates, suggesting that their expression pattern is shared by both the neoplastic cells and their supporting stromal elements. Examples in this class include neural precursor cell expressed/developmentally downregulated 5 (NEDD5, more than twofold upregulated in cancer compared to normal) and interleukin-13 (IL-13, more than ®vefold downregulated). Hence by comparing the pro®le of gene expression in the material of mixed cellularity to that in cancer cells alone, it is also possible to de®ne the contribution of stromal elements. Genes belonging to the Wnt gene family, namely Wnt5A, secreted apoptosis-related proetin-1 (SARP1/hFRP2), dishevelled (DVL) and c-jun N terminal kinase 2 (JNK2) that were previously not associated with pancreatic adenocarcinomas showed deregulated expression as well (Table 2, Figure 2). Of 187 genes that were expressed in tumour cellenriched samples (i.e. intensity value higher than twice the background), 149 were found in common with the genes expressed in each of the bulk pancreatic cancer experiments. The full list of those commonly expressed genes that represent the expression pro®le of pancreatic adenocarcinoma is available as supplementary information on the WWW (http://www.lif.icnet.uk/axp/mphh/ Supplementary/index.html). In hybridizations with the normal donor pancreas, expression of a total of 164 genes was con®rmed in at least two experiments. Those genes comprising the gene expression pro®le of completely healthy pancreas can also be viewed as supplementary information at the same WWW location.

Table 1 Summary of genes consistently found to be di€erentially expressed in pancreatic cancer tissues compared to normal pancreas

Oncogene

Accession number

Up-regulated genes

Comparison1

Comparison 2

X14420 X56134 M14219 M38690 M29870 D63878

Procollagen 3 alpha 1 subunit Vimentin Decorin CD9 Rac 1 NEDD5

ratio ratio ratio ratio ratio ratio

diff. 54 ratio 2.7 diff. 548 ratio 10.7 diff. 55 ratio 4.6

Accession number X56932 L06801 M65199 L20861 M57730 U147971 X00351

Down-regulated genes RPL-13a Interleukin 13 Endothelin 2 Wnt5A Ephrin-A1 precursor Ribosomal protein S9 Beta actin

ratio 0.3 diff. 7505 diff. 7333 diff. 7635 diff. 7524 diff. 7333 diff. 7403

8.2 4.9 4.7 4.1 2.6 2.2

ratio 0.08 ratio 0.13 ratio 0.13 ratio 0.70 diff. 7165 ratio 0.3 diff. 7221

Gene expression in pancreatic adenocarcinoma T Crnogorac-Jurcevic et al

7439

Figure 1 (a) Histopathological appearance of a human pancreatic cancer showing a cancer cellularity of about 35%; (b) the ®ne needle aspirate obtained in this case, where cancer cells comprise more than 95% of the total cell population. (Hematoxylin and eosin staining, original magni®cation 620); (c) Immunocytochemical staining of cellular aspirate for cytokeratin 8/18; (d) SYBRGold-stained agarose gel of three RNA samples derived from the cellular aspirates

Table 2

Summary of genes di€erentially expressed in pancreatic cancer cells obtained by ®ne needle aspiration compared to normal pancreas

Accession number

Up-regulated genes

Ratio

U28811 X02492 U60520 D50477 BC001710 U16306 X03124 U46461 X78817 Z18951 X13916 X07282 D63878

Golgi membrane sialoglycoprotein GLG1 Leukocyte interferon-inducible peptide Caspase 8 Matrix metalloproteinase 16 precursor NIK serine/threonine protein kinase Versican Tissue inhibitor of matrix metalloproteases 1 Dishevelled homologue 1 Rho-GAP activating protein 4 Caveolin-1 alpha-2-macroglobulin receptor Retinoic acid receptor beta NEDD5

10.3 8.9 7.8 4.6 3 2.8 2.6 2.6 2.4 2.4 2.1 2.1 2

Accession number AF017986 L25851 LO6801 X6B203 J02703 U72661 X76132 X79483 U01134 L20471 M62880 X51521 X98085 M38090 M94151 M98776 L20861 L31951

Down-regulated genes SARP 1 (hFRP-2) Integrin alphaE Interleukin-13 Vascular endothelial growth factor receptor 3 Integrin beta3 Ninjurin-1 Tumour suppression protein DCC precursor Extracellular signal-regulated kinase 6 Vascular endothelial growth factor receptor 1 Basigin (emmprin) Integrin beta7 Cytovillin 2, ezrin Tenascin-R CD9 alpha-2 catenin Cytokeratin 1 Wnt5A c-jun N-terminal kinase 2

Ratio 0.13 0.16 0.18 0.22 0.23 0.24 0.25 0.25 0.27 0.27 0.37 0.38 0.41 0.43 0.47 0.48 0.52 Nc

Difference 2356 4284 752 485 389 563 1516 1481 231 162 391 116 78 Difference 7696 7400 71793 7369 7348 71212 7254 7510 7167 73805 71629 7564 7636 7756 7456 7626 71129 7599 Oncogene

Gene expression in pancreatic adenocarcinoma T Crnogorac-Jurcevic et al

7440

Figure 2 Representative Clontech human cancer Atlas array analysis. The arrows identify the positions of several genes di€erentially expressed in cancer aspirates vs normal pancreas

Validation of gene expression data We have validated di€erential gene expression by realtime quantitative RT ± PCR for two of the genes (Rac 1 and GLG1) and the other eight were con®rmed by data from publicly available SAGE studies and UniGene Digital Di€erential Display (DDD). Real-time quantitative RT ± PCR analysis of Rac 1 and GGL1 expression We have chosen to study in more detail the ras-related gene Rac 1, knowing the importance of ras and other small GTPases in pancreatic cancer and carcinogenesis in general. Rac 1 was expressed in all experiments with cancer samples and 2.7-fold up-regulated in the ®rst set of three carcinomas (see Table 1 and supplementary data). Quantitative data on Rac 1 expression in normal and malignant pancreatic tissues and pancreatic cancer cell lines are shown in Figure 3. In 14 of 20 (70%) cancer cases studied, Rac 1 was overexpressed more than twofold in comparison to the mean value for completely healthy pancreatic tissues (Npa1 and Npa2 in Figure 3). A slight increase in Rac 1 expression was noticed in sample 4N (matched non-neoplastic tissue to malignant sample Tm4). Histopathologic examination showed that this sample was complicated by a particularly marked chronic pancreatitis, which probably contributed to the Rac 1 overexpression. In eight Oncogene

tumours, Rac 1 was up-regulated from 10-fold (case Tm13) to almost 40-fold (case Tm2). Of note, weak expression of Rac 1 in tumour samples Tm15, Tm16 and Tm17, which constitute the second pool of three bulk specimens corresponds well to their low expression obtained in array hybridization. All pancreatic cancer cell lines displayed even higher Rac 1 expression levels, between 20 and 140 times the level of Rac 1 expression in normal pancreas (note separate scale in Figure 3). Real-time quantitative RT ± PCR data for the GLG1 gene are summarized in Figure 4. More than half of the tumour samples show more than twofold gene upregulation in comparison to the mean value for completely healthy pancreatic tissues Npa1 and Npa2, with the highest overexpression being more than 10fold. In the two cases where there were matched specimens from the same patient (Tm4/4N and Tm10/ 10N), GLG1 expression was more than threefold higher in the tumour tissues. All the six cell lines showed more than 10-fold upregulation of the GLG1 transcript. SAGE and DDD data Upregulation of decorin and collagen 3 alpha 1 was con®rmed when comparing four pooled pancreatic cancer libraries with ®ve pooled libraries from normal pancreas by Digital Di€erential Display [http://

Gene expression in pancreatic adenocarcinoma T Crnogorac-Jurcevic et al

7441

Figure 3 Real-time quantitative RT ± PCR analysis of Rac 1 gene expression. Data are presented as copies of mRNA molecules in normal pancreatic samples (NPa1, NPa2, 10N, 4N), pancreatic adenocarcinoma samples (Tm1 ± Tm20) and six pancreatic cancer cell lines

Figure 4 Real-time quantitative RT ± PCR analysis of GLG1 gene expression. Data are presented as copies of mRNA molecules in normal pancreatic samples (NPa1, NPa2, 10N, 4N), pancreatic adenocarcinoma samples (Tm1 ± Tm20) and six pancreatic cancer cell lines Oncogene

Gene expression in pancreatic adenocarcinoma T Crnogorac-Jurcevic et al

7442

www.ncbi.nlm.nih.gov/UniGene] (data not shown). Those two genes were also con®rmed as di€erentially expressed in SAGE data (Figure 5). In addition, we identi®ed downregulation of the ribosomal protein genes RPS9 and RPL-13a in cancer, when comparing two bulk pancreatic cancer SAGE libraries with SAGE libraries from normal duct epithelial cells (http:// www.ncbi.nlm.nih.gov/SAGE/sagexp.cgi). Di€erential gene expression of an additional four genes observed in our study was con®rmed with SAGE data (Figure 5). Namely, versican, Rho GTPase-activating protein 4 and caveolin-1 were only found expressed in cancer tissue and/or cell lines; and caspase 8 was found to be more highly expressed in pancreatic cell lines than in bulk cancer and normal ducts, which is in good agreement with our ®nding of its upregulation in only tumour cell-enriched samples (Table 2).

Discussion Fine needle aspiration of cancer for neoplastic cell enrichment We describe a convenient method of obtaining an enriched population of tumour cells from pancreatic adenocarcinoma, which is a tumour type characterized by a highly desmoplastic stromal reaction. Tissue aspiration is less time-consuming and more straightforward than microdissection techniques and provides RNA samples of good quality and quantity. This is essential for gene expression studies and critical in the analysis of the pancreas, which is an organ with high endogenous nuclease activity. Cell enrichment by aspiration is based on the general characteristic of epithelial tumour cells to have decreased adhesion, therefore allowing easy separation from the surrounding stroma. This technique could potentially be applied

to obtain tumour cell-enriched samples from other epithelial tumour types characterized by various levels of desmoplastic reaction. Differentially expressed genes in bulk neoplastic tissues The ®rst part of our study addressed genes expressed in bulk specimens of pancreatic cancer, in which the pro®les will re¯ect the mixed composition of the tissue sample. A number of transcripts were expressed in both (see supplementary data at http://www.lif.icnet. uk/axp/mphh/Supplementary/index.html), analogous to the situation described in a SAGE study of normal colon and colon carcinomas (Zhang et al., 1997). As bulk adenocarcinoma samples in general contain not more than 30 ± 40% tumour cells, those samples were expected to display genes predominantly involved in the development of desmoplasia. This phenomenon is an intense stromal ®brosis induced by activation of ®broblasts and pancreatic stellate cells, with increased production of matrix proteins and matrix degrading enzymes (Gress et al., 1998). An accentuated stromal reaction has been described as an independent prognostic factor in a variety of tumour types including breast (Cardone et al., 1997) and colorectal cancer (Halvorsen and Seim, 1989). Desmoplastic stroma typically consists of collagens, proteoglycans and glycoproteins. We describe the involvement of several genes that contribute to the desmoplastic reaction in pancreatic adenocarcinoma, namely collagens of types I and III, decorin and versican. Involvement of collagens has already been well documented (Gress et al., 1995; Ryu et al., 2001). The small, collagen-associated proteoglycan decorin plays an important role in the remodelling and maintenance of extracellular matrix (ECM) during in¯ammation, ®brosis and cancer growth through its ability to bind ECM constituents and growth factors

Figure 5 Virtual Northern of genes di€erentially expressed in SAGE libraries of normal pancreatic ductal cells (HX and H126), bulk pancreatic cancer tissues (Panc91 ± 6113 and Panc96 ± 6252) and pancreatic cancer cell lines (CAPAN1, CAPAN2, HS766T and PANC1). Col3a1: procollagen 3 alpha 1 subunit; RPS9: ribosomal protein S9; RPL-13a: ribosomal protein L-13a VER: versican; ARHGAP4: Rho-GAP activating protein 4; CAV1: caveolin-1; CASP8: caspase 8 Oncogene

Gene expression in pancreatic adenocarcinoma T Crnogorac-Jurcevic et al

like epidermal growth factors and TGFb, whose involvement in pancreatic adenocarcinoma has been well documented (Asakura et al., 1999; Lohr et al., 2001; Moscatello et al., 1998). Versican is another proteoglycan that was found upregulated in pancreatic cancer. Immunolocalization of both versican and decorin has been described in the peritumoural stroma of early-stage prostate cancer, another tumour type that typically displays a desmoplastic reaction (Ricciardelli et al., 1998). One additional gene that was found upregulated in bulk cancer tissue in comparison with normal cells was Rac 1. The involvement of various small GTPases has already been described in a number of epithelial tumours. In a recent study, RhoA overexpression was found in the majority of colon, breast and lung carcinomas, proving to be a more general phenomenon than abnormalities of p53, ras and cmyc, each of which were observed in not more than 30 ± 60% of tumours (Fritz et al., 1999). In pancreatic adenocarcinomas, higher expression levels of RhoC have been described and found signi®cantly correlated with poor prognosis (Suwa et al., 1998). We report that Rac 1 (ras-related C3 botulinum toxin substrate) is upregulated in around 70% of tumour samples examined (Table 1 and Figure 3). Rac 1 overexpression was therefore con®rmed both by a hybridization approach and by real-time quantitative RT ± PCR. In addition to increased levels of expression in malignant pancreatic tissue, Rac 1 expression was extremely high in all six pancreatic cancer cell lines examined (Figure 3). Rac 1 regulates various cell functions such as cell shape, cell motility and cytokinesis through reorganization of the actin cytoskeleton and cross-talk with cadherin-dependent intercellular adhesion (Takaishi et al., 1997). It acts downstream of ras and is critical to ras-induced transformation, but it is also an independent transforming agent. Fibroblasts expressing Rac 1 mutated to valine at position 12 show a marked increase in lamellipodia, increased and anchorageindependent cell proliferation and enhanced tumorigenicity in athymic nude mice (Qiu et al., 1995). In our series of 20 pancreatic adenocarcinoma cases and eight cell lines it appears that there is no connection between K-ras status and Rac 1 expression level (data not shown), which is consistent with a recent study of Rac 1 in breast carcinoma (Schnelzer et al., 2000). In contrast to breast cancer, K-ras activating mutations are present in the majority of pancreatic adenocarcinomas, and so it would be of interest to study the interaction of the two GTPases, as well the potentially independent growth promoting activity of Rac 1 in pancreatic carcinoma. Up-regulation in bulk samples but no di€erential expression in aspirates would argue for shared stromal and epithelial expression. At present, speci®c antibodies that distinguish Rac 1 and Rac 2 are not available (the Rac 1 and Rac 2 genes are 92% homologous and both transcripts were obtained in our study), and the exact location of Rac 1 expression remains uncertain.

The most marked downregulation in bulk cancer experiments was displayed by ribosomal protein L13a. A secondary role other than in protein biosynthesis has recently been described for a number of ribosomal proteins. RPL13a, for example, is involved in regulation of cell proliferation (inducing arrest of cell cycle in G2/M phase) and induces apoptosis in leukaemic U937 cells (Chen et al., 1998). Its downregulation shown here is consistent with a potentially similar function in pancreatic cancer cells. Our study revealed several genes whose involvement in pancreatic exocrine tumours has been previously reported: TIMP1, caspase 8, IL-13 and CD9. Their presence was invaluable as an independent con®rmation of the validity of data obtained by array hybridization. CD9 is a ubiquitously expressed tetraspanin that forms complexes with integrins, other tetraspanins and MHC antigens and is associated with small GTP-binding proteins (Berditchevski and Odintsova, 1999). In pancreatic cancer, reduced CD9 expression levels have been associated with high tumour grade and lymph node metastasis (Sho et al., 1998). In our study it was evident that whilst there are elevated levels of CD9 in the stroma (Table 1), there is reduced CD9 expression in the tumour cell population (Table 2). This again points to the necessity of monitoring gene expression pro®les in the context of a homogenous population of cells puri®ed from a complex tissue. Interleukin-13 is a pleiotropic cytokine produced by tumour-in®ltrating lymphocytes. Its receptor is present on the majority of cancer cells and cell lines (including pancreatic cancer), usually conveying an antiproliferative e€ect (Kornmann et al., 1999). Interleukin-13 was found downregulated in both bulk tumour samples and aspirates (Tables 1 and 2). TIMP1 and caspase 8 were obtained only after enrichment of samples with tumour cells (Bramhall et al., 1997; Virkajarvi et al., 1998; Zhang et al., 1997).

7443

Differentially expressed genes in cancer cells The most interesting genes that were found dysregulated in tumour cell-enriched samples were GLG1, leukocyte interferon-inducible peptide, NEDD5 and Wnt gene family members ± Wnt5A, SARP1/hFRP-2, dishevelled and JNK2. Most of them have not been previously associated with pancreatic malignancy. Golgi sialoglycoprotein MG-160, GLG1 (also called cysteine-rich ®broblast growth factor receptor) was the gene that displayed the highest di€erential in the analysis of puri®ed tumour cells compared to normal pancreas (more than 10-fold upregulation). Its overexpression was additionally con®rmed by quantitative real-time RT ± PCR (Figure 4). Its postulated involvement in mediating secretion of ®broblast growth factors makes it a plausible candidate given the important role of this family in pancreatic tumorigenesis (Kohl et al., 2000). NEDD5 protein homologue KIAA0158 is a cytoskeletal GTPase involved in cytokinesis and other cellular processes (Kinoshita et al., 1997). It was found Oncogene

Gene expression in pancreatic adenocarcinoma T Crnogorac-Jurcevic et al

7444

consistently upregulated in all our experiments and it would be interesting to investigate further the role of this gene in the process of tumorigenesis. Several members of the Wnt5A gene family appeared deregulated in pancreatic cancers, namely secreted apoptosis-related protein 1 (SARP1)/hFRP-2, dishevelled, JNK2 and Wnt5A itself, implicating the importance of this pathway in pancreatic cancer growth. In addition, SKY (spectral karyotype analysis) data from our laboratory show that deletion of the chromosomal region containing Wnt5A (13p14 ± 21) in six out of 20 (30%) pancreatic carcinoma cell lines, while the SARP1/hFRP-2 locus (4q) is deleted in eight out of 20 cell lines (40%) (Sirivatanauksorn et al., 2001). Involvement of the Wnt5A gene family has not been previously associated with pancreatic adenocarcinoma. Wnt5A was described as a tumour suppressor gene in uroepithelial malignancies, and aberrant expression was found in lung, breast and prostate carcinomas as well as in melanomas (Iozzo et al., 1995; Olson et al., 1997). SARP-1/hFRP-2 belongs to a family of secreted glycoproteins that modulate the signalling activity of Wnt proteins. It is highly expressed in normal pancreas (Hu et al., 1998). Upregulation of SARP-1/hFRP-2 is associated with growth inhibition, and down-regulation might contribute to malignancy. Lower expression of this gene was found in ®ve out of ®ve breast carcinomas, two out of four ovary tumours and three out of ®ve kidney carcinomas (Zhou et al., 1998). In our experiments, it was downregulated sevenfold relative to normal pancreas (Table 2). Dysregulated expression of dishevelled 1 and c-jun N-terminal kinase 2, which are components of the same pathway, was also observed and warrants further investigation of Wnt signalling in pancreatic tumorigenesis. The results described demonstrate the practical value of cDNA array technology in deciphering the molecular genetics of pancreatic adenocarcinoma. The gene expression pro®les of predominantly stromal elements and epithelial tumour components isolated by cell aspiration give an insight into molecular crosstalk between tumour cells and their microenvironment that contributes to the biological behaviour of pancreatic cancer. Although the study was performed on a limited number of samples, it has nevertheless highlighted several genes that are potential candidates for further investigation and may prove useful for improving diagnosis and therapy of this highly malignant tumour.

Materials and methods Tissue specimens, aspirates and cell lines Pancreatectomy specimens were obtained at the Department of Pathology, University of Verona, Italy with full ethical approval from the host institution. Snap-frozen samples from six pancreatic ductal adenocarcinomas prepared immediately following surgical resection Oncogene

were used to characterize gene expression pro®les in bulk neoplastic tissue. Cryostat sections were analysed to verify the neoplastic cellularity, which was typically 30 ± 40%. Three of those six bulk specimens were also used as a source for ®ne needle aspirations. Immediately after the surgery, the freshly resected tumour mass was aspirated using a sterile syringe with a 21G needle to obtain enriched populations of neoplastic cells. A total of ®ve aspirations were made in each case; they were pooled and quickly rinsed with phosphate-bu€ered saline. A small aliquot was used for cytological and immunocytochemical examination for epithelial cytokeratins (anticytokeratin 8/18 and CAM5.2 [anticytokeratin 7/8]; both from Research Monoclonal Antibodies, ICRF Clare Hall, Middlesex, UK). Only aspirates that contained more than 95% tumour cells were used in the study. The remainder of the aspirated material was processed immediately for RNA extraction. Real-time quantitative RT ± PCR was performed on a total of 20 adenocarcinoma cases, including the six used for the array studies (Tm15 ± Tm20) and six pancreatic cell lines. Mia PaCa-2, PANC-1 and BxPC-3 were obtained from the American Type Culture Collection (Rockville, MD, USA; PaTu 8988s and PaTu 8988t from Dr HF Kern (Department of Cell Biology, Philipps University, Marburg, Germany) and HPAF from Dr R Metzger, (University of N. Carolina, Durham, USA). All cell lines were cultured in RPMI-1640 medium (ICRF Media Production, Clare Hall, Middlesex, UK) supplemented with 10% heat-inactivated foetal calf serum (Gibco-BRL, Life Technologies, Paisley, UK). As the non-neoplastic areas of the pancreas in cancer cases are usually a€ected by chronic pancreatitis, a snap-frozen, normal pancreas not suitable for transplantation from an organ donor was used as the control in all phases of the study. A normal pancreatic RNA sample purchased from Clontech (Basingstoke, Hampshire, UK) was used as an additional control in real time RT ± PCR experiments. RNA preparation Total RNA from tissue specimens, cellular aspirates and cell lines was prepared using TRIZOL reagent (Gibco-BRL, Life Technologies Inc., Frederick, MA, USA) according to the manufacturer's protocol. The yield from the cellular aspirates, containing around 1 ± 26106 cells, was typically 2 ± 5 mg. The quality of RNA from each sample was veri®ed by electrophoresis through agarose gel and visualization with SYBR Gold dye (Molecular Bioprobes, Eugene, Oregon, USA). Probe labelling and hybridization The pro®les of gene expression were investigated using Human Cancer Atlas cDNA Expression Array membranes (Clontech, Palo Alto, CA, USA). The arrays comprise sequences of 588 annotated cancer-associated genes belonging to various functional groups and include negative control and housekeeping genes (http:/www.clontech.com/atlas/genelists/ index.shtml). To reduce the e€ect of biological heterogeneity of tumour samples in our studies, and obtain the pro®le of genes commonly expressed in pancreatic adenocarcinomas, each hybridization was performed with a pool of three samples. Three pairs of ®lters were used. One ®lter from each pair was hybridized with control cDNA prepared from the normal organ donor pancreas. Two ®lters (from ®lter pair 1 and 2) were hybridized with target cDNA prepared from two di€erent sets of three pooled adenocarcinomas, and the third

Gene expression in pancreatic adenocarcinoma T Crnogorac-Jurcevic et al

one (from ®lter pair 3) was probed with the cDNA synthesized from three pooled tumour aspirates. The membranes were hybridized according to the manufacturer's speci®cations. In each case, an aliquot of 2 mg of total RNA was reverse transcribed with [a-32P]dATP (10 mCi/ml, 3000 Ci/mmol; Amersham Pharmacia Biotech, UK) and the reagents supplied in the Atlas kit to synthesize 32P-labelled cDNA target. The membranes were hybridized overnight with 1.56106 c.p.m./ml of ExpressHyb hybridization solution (Clontech, Palo Alto, CA, USA) at 688C in a Hybaid roller bottle incubator. Membranes were washed twice with 26SSC/1% and twice with 0.16SSC/0.5% SDS at 688C for 30 min each. Filters were exposed for up to 3 days in phosphor storage cassettes and scanned in a phosphorimager (Molecular Dynamics, Sunnyvale, CA, USA). Data analysis Image analysis and quantitation was performed with AtlasImage 1.01a software (Clontech Laboratories, Palo Alto, CA, USA). After grid assignment, the intensity value was calculated for each gene; as the arrays used were doublespotted, intensity was calculated as the average of the total signal (pixel value obtained after phosphorimager scanning) from the left and right spots. After the subtraction of background, global normalization was performed in the following way: the normalization coecient was calculated by dividing the average values of all the genes from Array-1 (normal) with Array-2 (cancer) which was then used to multiply the intensity value of each gene. That gives the Adjusted Intensity. The ratio for each gene is then calculated by dividing the Array-2 Adjusted Intensity by the Adjusted Intensity of Array-1. A gene was considered to be expressed if the Intensity was more than twice the background level, and a ratio of twofold (0.5) was taken as the cut-o€ value for scoring upregulated (or downregulated) expression. When the Adjusted Intensity of a particular gene was less than twice the background on one of the arrays, the ratio could not be calculated as it would mean dividing by `0' (`Nc' in Table 2), therefore the Di€erence in Intensity values was calculated instead (Di€erence = Array-2 Adjusted Intensity ± Array-1 Adjusted Intensity). A Di€erence with a negative value represents a gene downregulated in cancer. As AtlasImage is designed for comparison of two arrays from each individual experiment and not for multi®lter comparisons, the data obtained from di€erent experiments were analysed in more detail using the GeneSpring program (Silicon Genetics). The 588 genes represented on the Atlas array were entered as a new `genome', and subclassi®ed into functional groups according to the list from Clontech (http://www.clontech.com/atlas/genelists/index.shtml). All expressed genes (intensities more than twice the background)

from individual experiments were compared and Venn diagrams created to enable us to visualise the intersections (commonly expressed genes) between di€erent experiments. The full lists of gene expression pro®les from both tumour and normal pancreas created in this way are available as supplementary data (http://www.lif.icnet.uk/axp/mphh/ Supplementary/index.html).

7445

Real-time quantitative RT ± PCR Real-time quantitative RT ± PCR was used as an independent method to con®rm up-regulation of two di€erentially expressed genes. Quantitative measurements of Rac 1 and GLG1 expression were obtained in 20 pancreatic cancer tissues, two matched non-neoplastic samples (coded as 4N and 10N), two normal pancreatic samples and six cell lines. cDNAs were synthesized from 1 to 5 mg total RNA using an oligo dT primer and the Superscript II reverse transcription kit (Gibco-BRL, Life Technologies, Paisley, Scotland). The cDNAs were column-puri®ed (Qiagen, Crawley, West Sussex, UK) and eluted at a concentration of 10 ng/ml. The primers used for PCR were as follows: Rac 1 sense, 5'-AG CCTTCGCACTCAATGCCAAC-3'; Rac 1 antisense, 5'TCGGCACAACAATGCAGTG TAG-3'; GLG1 sense, 5'CGCTCTTCTCAGAAGTGACACC-3'; GLG1 antisense, 5'GAGAGACTTGTCTACAGGCAGG-3'. These amplify a 249 and 190 bp fragment from the 3' untranslated region of the Rac 1 and GLG1 transcripts respectively. Reactions containing 10 ng cDNA, SYBR green sequence detection reagents (Applied Biosystems) and sense and antisense primers were assayed on an ABI7700 sequence detection system (Applied Biosystems). The PCR conditions were 1 cycle at 508C for 2 min, 1 cycle at 958C for 10 min, and 40 cycles of 958C for 15 s, 658C for 1 min. The accumulation of PCR product was measured in real time as the increase in SYBR green ¯uorescence, and the data were analysed using the Sequence Detector program v1.6.3 (Applied Biosystems). Standard curves relating initial template copy number to ¯uorescence and ampli®cation cycle were generated using the ampli®ed PCR product as a template, and were used to calculate Rac 1 (or GLG1) copy number in each sample.

Acknowledgements Supported by Imperial Cancer Research Fund; Special Trustees of Hammersmith Hospital; Associazione Italiana Ricerca sul Cancro (AIRC) to A Scarpa, Milan, Italy; Consorzio Studi Universitari di Verona, Italy; Ministero UniversitaÁ e Ricerca Scienti®ca e Tecnologica (MURST), Co®n MM06158571 ± Co®n 9906195987, Rome, Italy.

References Alizadeh AA, Eisen MB, Davis RE, Ma C, Lossos IS, Rosenwald A, Boldrick JC, Sabet H, Tran T, Yu X, Powell JI, Yang L, Marti GE, Moore T, Hudson J, Lu L, Lewis DB, Tibshirani R, Sherlock G, Chan WC, Greiner TC, Weisenburger DD, Armitage JO, Warnke R, Levy R, Wilson W, Grever MR, Byrd JC, Botstein D, Brown PO and Staudt LM. (2000). Nature, 403, 503 ± 511. Asakura S, Kato H, Fujino S, Konishi T, Tezuka N and Mori A. (1999). Hum. Pathol., 30, 195 ± 198. Aspinall RJ and Lemoine NR. (1999). Ann. Oncol., 10, 188 ± 192.

Berditchevski F and Odintsova E. (1999). J. Cell Biol., 146, 477 ± 492. Bramhall SR, Neoptolemos JP, Stamp GW and Lemoine NR. (1997). J. Pathol., 182, 347 ± 355. Cardone A, Tolino A, Zarcone R, Borruto Caracciolo G and Tartaglia E. (1997). Panminerva Med., 39, 174 ± 177. Chen FW, Davies JP and Ioannou YA. (1998). Mol. Genet. Metab., 64, 271 ± 282.

Oncogene

Gene expression in pancreatic adenocarcinoma T Crnogorac-Jurcevic et al

7446

Oncogene

Emmert-Buck MR, Bonner RF, Smith PD, Chuaqui RF, Zhuang Z, Goldstein SR, Weiss RA and Liotta LA. (1996). Science, 274, 988 ± 1001. Fritz G, Just I and Kaina B. (1999). Int. J. Cancer, 81, 682 ± 687. Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesirov JP, Coller H, Loh ML, Downing JR, Caligiuri MA, Bloom®eld CD and Lander ES. (1999). Science, 286, 531 ± 537. Gress TM, Menke A, Bachem M, Muller-Pillasch F, Ellenrieder V, Weidenbach H, Wagner M and Adler G. (1998). Digestion, 59, 625 ± 637. Gress TM, Muller-Pillasch F, Geng M, Zimmerhackl F, Zehetner G, Friess H, Buchler M, Adler G and Lehrach H. (1996). Oncogene, 13, 1819 ± 1830. Gress TM, Muller-Pillasch F, Lerch MM, Friess H, Buchler M and Adler G. (1995). Int. J. Cancer, 62, 407 ± 413. Gress TM, Wallrapp C, Frohme M, Muller-Pillasch F, Lacher U, Friess H, Buchler M, Adler G and Hoheisel JD. (1997). Genes Chromosomes Cancer, 19, 97 ± 103. Halvorsen TB and Seim E. (1989). J. Clin. Pathol., 42, 162 ± 166. Hu E, Zhu Y, Fredrickson T, Barnes M, Kelsell D, Beeley L and Brooks D. (1998). Biochem. Biophys. Res. Commun., 247, 287 ± 293. Iozzo RV, Eichstetter I and Danielson KG. (1995). Cancer Res., 55, 3495 ± 3499. Kinoshita M, Kumar S, Mizoguchi A, Ide C, Kinoshita A, Haraguchi T, Hiraoka Y and Noda M. (1997). Genes Dev., 11, 1535 ± 1547. Kohl R, Antoine M, Olwin BB, Dickson C and Kiefer P. (2000). J. Biol. Chem., 275, 15741 ± 15748. Kornmann M, Klee€ J, Debinski W and Korc M. (1999). Anticancer Res., 19, 125 ± 131. Lemoine NR. (1997). Digestion, 58, 550 ± 556. Lohr M, Schmidt C, Ringel J, Kluth M, Muller P, Nizze H and Jesnowski R. (2001). Cancer Res., 61, 550 ± 555. Moscatello D, Santra M, Mann DM, McQuillan DJ, Wong AJ and Iozzo RV. (1998). J. Clin. Invest., 101, 406 ± 412. Olson DJ, Gibo DM, Saggers G, Debinski W and Kumar R. (1997). Cell Growth Di€er., 8, 417 ± 423. Perou CM, Sorlie T, Eisen MB, van de Rijn M, Je€rey SS, Rees CA, Pollack JR, Ross DT, Johnsen H, Akslen LA, Fluge O, Pergamenschikov A, Williams C, Zhu SX, Lonning PE, Borresen-Dale AL, Brown PO and Botstein D. (2000). Nature, 406, 747 ± 752.

Qiu RG, Chen J, Kirn D, McCormick F and Symons M. (1995). Nature, 374, 457 ± 459. Ricciardelli C, Mayne K, Sykes PJ, Raymond WA, McCaul K, Marshall VR and Horsfall D. (1998). Clin. Cancer Res., 4, 963 ± 971. Ross DT, Scherf U, Eisen MB, Perou CM, Rees C, Spellman P, Iyer V, Je€rey SS, Van de Rijn M, Waltham M, Pergamenschikov A, Lee JC, Lashkari D, Shalon D, Myers TG, Weinstein JN, Botstein D and Brown PO. (2000). Nat. Genet., 24, 227 ± 235. Ryu B, Jones J, Hollingsworth MA, Hruban RH and Kern SE. (2001). Cancer Res., 61, 1833 ± 1838. Schnelzer A, Prechtel D, Knaus U, Dehne K, Gerhard M, Grae€ H, Harbeck N, Schmitt M and Lengyel E. (2000). Oncogene, 19, 3013 ± 3020. Sho M, Adachi M, Taki T, Hashida H, Konishi T, Huang CL, Ikeda N, Nakajima Y, Kanehiro H, Hisanaga M, Nakano H and Miyake M. (1998). Int. J. Cancer, 79, 509 ± 516. Shoemaker DD, Schadt EE, Armour CD, He YD, GarrettEngele P, McDonagh PD, Loerch PM, Leonardson A, Lum PY, Cavet G, Wu LF, Altschuler SJ, Edwards S, King J, Tsang JS, Schimmack G, Schelter JM, Koch J, Ziman M, Marton MJ, Li B, Cundi€ P, Ward T, Castle J, Krolewski M, Meyer MR, Mao M, Burchard J, Kidd MJ, Dai H, Phillips JW, Linsley PS, Stoughton R, Scherer S and Boguski MS. (2001). Nature, 409, 922 ± 927. Sirivatanauksorn V, Sirivatanauksorn Y, Gorman PA, Davidson JM, Sheer D, Moore PS, Scarpa A, Edwards PA and Lemoine NR. (2001). Int. J. Cancer, 91, 350 ± 358. Sirivatanauksorn Y, Drury R, Crnogorac-Jurcevic T, Sirivatanauksorn V and Lemoine NR. (1999). J. Pathol., 189, 150 ± 154. Suwa H, Ohshio G, Imamura T, Watanabe G, Arii S, Imamura M, Narumiya S, Hiai H and Fukumoto M. (1998). Br. J. Cancer, 77, 147 ± 152. Takaishi K, Sasaki T, Kotani H, Nishioka H and Takai Y. (1997). J. Cell Biol., 139, 1047 ± 1059. Virkajarvi N, Paakko P and Soini Y. (1998). Histopathology, 33, 432 ± 439. Zhang L, Zhou W, Velculescu VE, Kern SE, Hruban RH, Hamilton SR, Vogelstein B and Kinzler KW. (1997). Science, 276 1268 ± 1272. Zhou Z, Wang J, Han X, Zhou J and Linder S. (1998). Int. J. Cancer, 78, 95 ± 99.