Oncogene (2002) 21, 475 ± 478 ã 2002 Nature Publishing Group All rights reserved 0950 ± 9232/02 $25.00 www.nature.com/onc
SHORT REPORTS
Global gene expression pro®ling in Barrett's esophagus and esophageal cancer: a comparative analysis using cDNA microarrays FM Selaru1,3, T Zou1,3, Y Xu1,3, V Shustova1,3, J Yin1, Y Mori1, F Sato1, S Wang1, A Olaru1, D Shibata2, BD Greenwald1, MJ Krasna2, JM Abraham1 and SJ Meltzer*,1 1
Department of Medicine, Division of Gastroenterology, Greenebaum Cancer Center, University of Maryland School of Medicine, Baltimore VA Hospital, Baltimore, Maryland, MD 21201, USA; 2Department of Surgery, Division of Surgical Oncology, Greenebaum Cancer Center, University of Maryland School of Medicine, Baltimore VA Hospital, Baltimore, Maryland, MD 21201, USA
In order to identify and contrast global gene expression pro®les de®ning the premalignant syndrome, Barrett's esophagus, as well as frank esophageal cancer, we utilized cDNA microarray technology in conjunction with bioinformatics tools. We hybridized microarrays, each containing 8000 cDNA clones, to RNAs extracted from 13 esophageal surgical or endoscopic biopsy specimens (seven Barrett's metaplasias and six esophageal carcinomas). Hierarchical cluster analysis was performed on these results and displayed using a color-coded graphic representation (Treeview). The esophageal samples clustered naturally into two principal groups, each possessing unique global gene expression pro®les. After retrieving histologic reports for these tissues, we found that one main cluster contained all seven Barrett's samples, while the remaining principal cluster comprised the six esophageal cancers. The cancers also clustered according to histopathological subtype. Thus, squamous cell carcinomas (SCCAs) constituted one group, adenocarcinomas (ADCAs) clustered separately, and one signet-ring carcinoma was in its own cluster, distinct from the ADCA cluster. We conclude that cDNA microarrays and bioinformatics show promise in the classi®cation of esophageal malignant and premalignant diseases, and that these methods can be applied to small biopsy samples. Oncogene (2002) 21, 475 ± 478. DOI: 10.1038/sj/onc/ 1205111 Keywords: esophaegeal cancer; Barrett's esophagus; cDNA microassays; bioinformatics; cancer diagnosis Barrett's esophagus, the replacement of the normal squamous esophageal epithelium by a metaplastic columnar lining, is a premalignant condition caused by chronic gastroesophageal re¯ux (Ransford and
*Correspondence: SJ Meltzer, University of Maryland, Room N3W62, 22 S. Greene St., Baltimore, MD 21201, USA; E-mail:
[email protected] 3 These authors contributed equally to this work. Received 6 September 2001; revised 22 October 2001; accepted 30 October 2001
Jankowski, 2000). This condition predisposes to the development of esophageal cancer (adeno-carcinoma). In Western countries, esophageal cancer usually presents at an advanced stage and undergoes a rapidly fatal course (Daly et al., 2000). Approximately 75% of untreated patients with esophageal cancer die within 1 year of diagnosis (Parker et al., 1952). Five-year survival rates for esophageal carcinoma are approximately 25 ± 30% (Blot and McLaughlin, 1999). These rates apply to both SCCA, which is related to heavy tobacco use, and ADCA, which is associated with chronic gastroesophageal re¯ux and the premalignant lesion, Barrett's esophagus. It is hoped that novel biomarkers and diagnostic tools will lead to future improvements in early detection and survival in this disease. Recent advances in biotechnology have revolutionized the high-throughput analysis of gene expression. In particular, the development of cDNA microarray technologies has made it possible to analyse the expression of thousands of genes simultaneously. cDNA microarray technology, in combination with bioinformatics analyses, promises more accurate disease classi®cation, earlier detection, and higher eciency in the ®eld of cancer diagnosis (Moch et al., 1999; Wang et al., 1999; Elek et al., 2000). Studies have been performed in the human cancers acute leukemia, lymphoma, breast cancer, and others (Chu et al., 1998; Spellman et al., 1998; Eisen and Brown, 1999; Iyer et al., 1999; Perou et al., 1999, 2000; Alizadeh et al., 2000). These studies demonstrate the potential of a modern molecular taxonomy based on the statistical power of large phenotypic datasets (Bassett et al., 1999). Indeed, it is now possible to classify diseases or lesions based on a comprehensive characterization of their phenotypes (Perou et al., 1999, 2000; Alizadeh et al., 1998). This characterization takes advantage of the belief that phenotype is, in fact, the set of all genes whose expression is up- or downregulated, particularly when contrasting sets representing discrete lesions or diseases. Although direct analysis of cDNA microarrays generates precise ratios of expression, it is somewhat dicult to interpret this data. The bioinformatics program Cluster (Eisen et al., 1998) generates hierarchical and K-means clusters from tab-delimited text
Gene expression profiles distinguish esophageal lesions FM Selaru et al
476
®les containing expression ratios comparing two RNA species' expression levels. In addition, the program Treeview (Eisen et al., 1998) generates two-dimensional graphic representations of gene expression in red and green colors. These software programs make analyses of microarray data much more feasible and appear frequently in the current literature (Chu et al., 1998; Spellman et al., 1998; Eisen and Brown, 1999; Iyer et al., 1999; Perou et al., 1999, 2000; Alizadeh et al., 2000). The current scheme of histologic classi®cation, based on somewhat subjective, nonquantitative evaluations of limited visual microscopic histologic data, may convey limited information regarding lesion biology. We hypothesized that objective, quantitative analyses of large molecular genetic data sets are likely to classify these lesions to more accurately re¯ect their underlying biologies. Therefore, we sought to de®ne and contrast Barrett's esophagus and esophageal cancers based on their global gene expression patterns. Our data suggest that cDNA microarray technology and bioinformatics can accurately discriminate between discrete but closely related esophageal neoplastic stages. Scanned images of hybridized slides were created using a cDNA microarray scanner (GenePix 4000A, Axon Instruments, Foster City, CA, USA). We used the software program GenePix 3.0 (Axon) to analyse microarray images (see Methods: Figure 1). The companion software programs Cluster and Treeview, devised by Eisen et al. (1998), were used for all bioinformatics calculations (see Methods). Visual interpretation using Cluster and TreeView Results obtained by using Cluster and TreeView are displayed in the form of a clustergram (Figure 2). This Figure shows all 13 samples analysed (seven Barrett's and six esophageal tumors), based on the 8000 clones on the microarrays. Green areas represent clones underexpressed in patient tissues relative to the
Figure 1 Microarray scan of one hybridized slide. Red dots represent clones overexpressed in the tissue sample compared to the reference probe; green dots are clones underexpressed in probe relative to reference; and yellow dots represent clones for which the expression ratio in tissue vs. reference probe is close to one. Patients with a presumed diagnosis of Barrett's esophagus or esophageal carcinoma were enrolled in the study, with informed consent obtained prior to enrollment under a protocol approved by the University of Maryland/Baltimore VA Hospital Institutional Review Board (IRB). Cancerous specimens were obtained from surgical resections and grossly dissected free of nontumor components prior to snap-freezing. Four of the Barrett's metaplastic specimens were also obtained at surgery, but the remaining three were obtained at endoscopy by endoscopic forceps biopsy: in this latter case, histology was determined by obtaining adjacent endoscopic biopsy specimens for histopathologic examination. The reference probe aRNA was produced from a mixture containing equimolar aliquots of RNA from the cancerderived cell lines Hct116, HT29, CaCo-2, Hct15 (colorectal), HTB114 (leukemia), MCF-7 (breast), HeLa (cervical), and AGS (gastric). Total RNA was extracted from freshly frozen tissues by standard organic methods (Chomczynski and Sacchi, 1987) and Oncogene
ampli®ed as previously described (Van Gelder et al., 1990; Luo et al., 1999). For each two-way comparison, 3 ± 6 mg of aRNA prepared from reference cells or esophageal tissue were labeled by incorporating Cy3- or Cy5-labeled dCTP using random primers and Superscript reverse transcriptase. In the preparation of microarray clones, we followed protocols obtained from the National Cancer Institute-Advanced Technology Center (NCIATC: David Petersen). The 95% nonredundant, sequenceveri®ed, periodically annotation-updated IMAGE clone cDNA library prepared by the Lawrence Livermore Laboratories was used as a source of clones (Research Genetics, Huntsville, AL, USA). We prepared lysine-coated slides derived from the NCIATC and Stanford University protocols (http://www.microarrays.org/protocols.html). The 8000 clones were printed using eight pins in a 32-pin print head (Majer Precision Engineering, Tempe, AZ, USA) on a GeneMachines Omnigrid Arrayer (GeneMachines, Oxnard, CA, USA). The printed slides were UV-crosslinked, post-treated with succinic anhydride to reduce background, and subjected to hybridization. The hybridized slides were scanned using a GenePix 4000A dual-laser slide scanning system (Axon Instruments, Foster City, CA, USA) at wavelengths corresponding to each probe's unique ¯uorescence
Gene expression profiles distinguish esophageal lesions FM Selaru et al
477
reference probe (see Methods); red regions represent clones overexpressed in tissues. Each esophageal lesion RNA (red, Cy5-labeled) was co-hybridized on a separate microarray slide with the reference RNA (green, Cy3-labeled). Each lesion is represented by its own column and branch of the dendrogram (branched tree). Patients fall clearly into two main groups, de®ned by both the dendrogram and the sharply contrasting color patterns. These two principal clusters exhibit almost complementary gene expression pro®les: i.e., genes which are `red' (overexpressed) in the ®rst major cluster tend to be `green' (underexpressed) in the second major cluster, and vice versa. Cluster analysis thus discerned a clear dierence between these two groups without prior histopathological input. Esophageal cancer and Barrett's esophagus have distinct gene expression profiles Histopathologic correlations revealed that all six samples in the ®rst cluster (Figure 2) comprised frank carcinomas. The second cluster contained all seven Barrett's metaplastic samples. Cluster analysis within the tumor group The left main branch in Figure 2 depicts the esophageal cancer cluster, along with sample names. Three subgroups were clearly delineated within this cluster: E885T and E871T, both esophageal SCCAs; E865, a signet-ring carcinoma; and three ADCAs (E887T, E878T, E891T). Thus, within the tumor cluster, there were three subclusters: the SCCA cluster, the ADCA cluster, and the signet-ring carcinoma cluster. Cluster analysis within the Barrett's group All seven Barrett's samples clustered more closely with each other than with the esophageal tumors. In contrast to the esophageal carcinoma cluster, there were no clearly de®ned subgroups within the Barrett's cluster (Figure 2). Integrative comments The 13 esophageal samples in this study were accurately categorized , based solely on genomic data, as belonging to one of the following categories: ADCA, SCCA, signetring adenocarcinoma, or Barrett's esophagus. All of the Barrett's samples were contained in a single cluster. Within this group, in contrast to the esophageal cancers, dierences between samples were not sucient to generate further subclustering (Figure 2). Stated dierently, all of the Barrett's metaplasias clustered quite closely to each other, suggesting a somewhat uniform global gene expression pattern within this group. However, ADCAs clustered more closely with SCCAs than with Barrett's samples, implying that ADCAs are more similar to other esophageal cancer subtypes than to their precursor lesion, Barrett's metaplasia. This ®nding suggests that the global gene expression pro®le changes
Figure 2 Treeview display (clustergram) of cluster analysis of all 13 esophageal samples. This dendrogram shows that esophageal samples clustered into two principal groups. Histopathologic data revealed that the left group was composed exclusively of esophageal cancers, while the right group consisted of all seven Barrett's esophagus samples. Data imported from GenePix were manipulated and clustered, using established algorithms implemented in the software program Cluster (http://rana.lbl.gov/) (Eisen et al., 1998; Eisen and Brown, 1999). Average linkage clustering with centered correlation was used. TreeView software (ibid.) generated visual representations of the clusters
fundamentally during neoplastic progression of Barrett's esophagus to adenocarcinoma. Moreover, our data imply that the accuracy of cDNA microarrays is comparable to that of histologic examination in diagnosing esophageal neoplastic and preneoplastic lesions. A second point to be gleaned from this research is that microarray technology can be successfully applied to small, limited biopsy specimens, greatly extending the potential clinical utility Oncogene
Gene expression profiles distinguish esophageal lesions FM Selaru et al
478
of this approach. Finally, these ®ndings suggest that global gene expression patterns may themselves help de®ne or serve as biomarkers of subtly distinct esophageal histopathologic entities.
Acknowledgments NIH grants CA85069, DK47717, CA95323, CA77057 and the Medical Research Oce, Department of Veterans Aairs.
References Alizadeh A, Eisen M, Botstein D, Brown PO and Staudt LM. (1998). J. Clin. Immunol., 18, 373 ± 379. 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 Jr J, Lu L, Lewis DB, Tibshirani R, Sherlock G, Chan WC, Greiner TC, Weisenburger DD, Armitage JO, Warnke R, Staudt LM, Levy R, Wilson W, Glever M, Byrd J, Botstein D and Brown PO. (2000). Nature, 403, 503 ± 511. Bassett Jr DE, Eisen MB and Boguski MS. (1999). Nat. Genet., 21(1 Suppl): 51 ± 55. Blot WJ and McLaughlin JK. (1999). Semin. Oncol., 26(Suppl 15), 2 ± 8. Chomczynski P and Sacchi N. (1987). Anal. Biochem., 162, 156 ± 159. Chu S, DeRisi J, Eisen M, Mulholland J, Botstein D, Brown PO and Herskowitz I. (1998). Science 282, 699 ± 705. Erratum published Science, 282, 1421. Daly JM, Fry WA, Little AG, Winchester DP, McKee RF, Stewart AK and Fremgen AM. (2000). J. Am. Coll. Surg., 190, 562 ± 572. Eisen MB and Brown PO. (1999). Methods Enzymol., 303, 179 ± 205. Eisen MB, Spellman PT, Brown PO and Botstein D. (1998). Proc. Natl. Acad. Sci. USA, 95, 14863 ± 14868. Elek J, Park KH and Narayanan R. (2000). In Vivo 14, 173 ± 182. Iyer VR, Eisen MB, Ross DT, Schuler G, Moore T, Lee JCF, Trent JM, Staudt LM, Hudson Jr J, Boguski MS, Lashkari D, Shalon D, Botstein D and Brown PO. (1999). Science, 283, 83 ± 87.
Oncogene
Luo L, Salunga RC, Guo H, Bittner A, Joy KC, Galindo JE, Xiao H, Rogers KE, Wan JS, Jackson MR and Erlander MG. (1999). Nat. Med., 5, 117 ± 122. Moch H, Schraml P, Bubendorf L, Mirlacher M, Kononen J, Gasser T, Mihatsch MJ, Kallioniemi OP and Sauter G. (1999). Am. J. Pathol., 154, 981 ± 986. Parker EF, Hanna CB and Postlethwait RW. (1952). Ann. Surg., 135, 697. Perou CM, Jerey SS, van de Rijn M, Rees CA, Eisen MB, Ross DT, Pergamenschikov A, Williams CF, Zhu SX, Lee JC, Lashkari D, Shalon D, Brown PO and Botstein D. (1999). Proc. Natl. Acad. Sci. USA, 96, 9212 ± 9217. Perou CM, Sorlie T, Eisen MB, van de Rijn M, Jerey 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. Ransford R and Jankowski J. (2000). Acta Gastroenterol. Belg., 63, 18 ± 21. Spellman PT, Sherlock G, Zhang MQ, Iyer VR, Anders K, Eisen MB, Brown PO, Botstein D and Futcher B. (1998). Mol. Biol. Cell., 9, 3273 ± 3297. Van Gelder RN, von Zastrow ME, Yool A, Dement WC, Barchas JD and Eberwine JH. (1990). Proc. Natl. Acad. Sci. USA, 87, 1663 ± 1667. Wang K, Gan L, Jeery E, Gayle M, Gown AM, Skelly M, Nelson PS and Ng WV. (1999). Gene, 229, 101 ± 108.
