Gene-Expression Profiling Reveals Distinct Expression Patterns For

Gene-expression proﬁling reveals distinct expression patterns for Classic versus Variant Merkel cell phenotypes and new classiﬁer genes to distinguish Merkel cell from small-cell lung carcinoma

Mireille Van Gele1, Glen M Boyle2, Anthony L Cook3,5, Jo Vandesompele1, Tom Boonefaes4, Pieter Rottiers4, Nadine Van Roy1, Anne De Paepe1, Peter G Parsons2, J Helen Leonard5 and Frank Speleman*,1

1Center for Medical Genetics, Ghent University Hospital, Ghent B-9000, Belgium; 2Melanoma Genomics Group, Queensland Institute of Medical Research, Brisbane, Queensland 4006, Australia; 3Institute for Molecular Biosciences, University of Queensland, Brisbane, Queensland 4072, Australia; 4Department of Molecular Biomedical Research, Flanders Interuniversity Institute for Biotechnology and University of Ghent, Ghent B-9000, Belgium; and 5Queensland Radium Institute Research Unit, Queensland Institute of Medical Research, Brisbane, Queensland 4006, Australia

Merkel cell carcinoma (MCC) is a rare aggressive skin Introduction tumor which shares histopathological and genetic features with small-cell lung carcinoma (SCLC), both are of Merkel cell carcinoma (MCC) is a rare aggressive skin neuroendocrine origin. Comparable to SCLC, MCC cell tumor and is assumed to arise from normal Merkel cells. lines are classified into two different biochemical sub- Merkel cells are neuroendocrine in origin, expressing groups designated as ‘Classic’ and ‘Variant’. With the aim markers such as neuron-specific enolase and bombesin. to identify typical gene-expression signatures associated These cells are located in the basal layer of the with these phenotypically different MCC cell lines epidermis, where they often function as slow-acting subgroups and to search for differentially expressed genes mechanoreceptors (Halata et al., 2003). MCC mostly between MCC and SCLC, we used cDNA arrays to affects elderly people and occurs predominantly on the profile 10 MCC cell lines and four SCLC cell lines. Using sun-exposed areas of the skin, suggesting UV exposure significance analysis of microarrays, we defined a set of 76 in its etiology (Miller and Rabkin, 1999; Van Gele et al., differentially expressed genes that allowed unequivocal 2000; Popp et al., 2002). identification of Classic and Variant MCC subgroups. We In a previous study, we have determined by compara- assume that the differential expression levels of some of tive genomic hybridization (CGH) the patterns of these genes reflect, analogous to SCLC, the different genomic imbalances which occur in MCC (Van Gele biological and clinical properties of Classic and Variant et al., 1998). Interestingly, the observed under- and over- MCC phenotypes. Therefore, they may serve as useful representations of partial chromosomal regions were prognostic markers and potential targets for the develop- quite similar to those observed in small-cell lung ment of new therapeutic interventions specific for each carcinoma (SCLC) (Ried et al., 1994; Levin et al., subgroup. Moreover, our analysis identified 17 powerful 1995; Petersen et al., 1997; Van Gele et al., 1998). Both classifier genes capable of discriminating MCC from are neuroendocrine tumors with remarkable histopatho- SCLC. Real-time quantitative RT–PCR analysis of these logical similarities, that is, small round cells often genes on 26 additional MCC and SCLC samples containing dense core granules and expressing several confirmed their diagnostic classification potential, opening identical immunohistochemical markers (Ratner et al., opportunities for new investigations into these aggressive 1993; Metz et al., 1998; Schmidt et al., 1998; Goessling cancers. et al., 2002). Therefore, the possibility exists that they ONCOGENOMICS Oncogene (2004) 23, 2732–2742. doi:10.1038/sj.onc.1207421 arise from a common cell lineage. Further support for Published online 2 February 2004 this hypothesis comes from the fact that cell lines derived from MCC and SCLC resemble each other in Keywords: Merkel cell carcinoma; small-cell lung carci- their biochemical, morphological and growth character- noma; differential expression profiling; differential istics. Similar to SCLC, MCC cell lines are classified into diagnosis; Classic/Variant two groups: ‘Classic’ and ‘Variant’ defined on their biochemical markers and neurosecretory granule status, which are further subdivided into four subtypes (I–IV) based on morphology, colony shape and aggregation (Carney et al., 1985; Gazdar et al., 1985; Leonard et al., *Correspondence: F Speleman; E-mail: [email protected] 1993, 1995a, 2002; Leonard and Bell, 1997). At present, Received 26 August 2003; revised 17 November 2003; accepted 2 little is known about the genes associated with these December 2003 characteristics. Gene-expression profiling of MCC and SCLC MV Gele et al 2733 In contrast to MCC, numerous molecular genetic studies have been performed on SCLC which have contributed to the understanding of SCLC pathogenesis (Fong et al., 1999; Minna et al., 2002). In addition, histochemical markers and differentially expressed genes distinguishing Classic from Variant SCLC cell lines have been identified and could lead to an improved understanding of the underlying genetic basis responsible for the biological and clinical heterogeneity among small- cell lung cancers (Broers et al., 1985; Zhang et al., 2000). In order to obtain further insights into the complex and heterogeneous molecular pathogenesis of MCC, we determined the gene-expression profiles of 10 MCC cell lines and four SCLC cell lines using Atlas cDNA arrays containing 1891 unique genes involved in many cellular functions. This study offers potential insights into the genes and signalling pathways involved in MCC and SCLC, a prerequisite for the development of new rational therapeutic interventions, which could lead to an improved patient survival or even complete remis- sion. Furthermore, we identified genes not previously implicated in these cancers, whose expression enabled discrimination between MCC and SCLC and may therefore aid in the differential diagnosis of cases where existing markers such as cytokeratin 20 are unable to differentiate between these two neuroendocrine cancers.

Results

Validation of atlas cDNA array data The expression level of 1891 genes was measured by the use of Atlas Human and Human Cancer 1.2 arrays for 10 MCC and four SCLC cell lines. After primary data analysis and normalization, the data of 1083 genes which had an expression value above background in at least six of the analysed samples were used for Figure 1 (a) Scaled-down representation of the hierarchical hierarchical clustering. Of these 1083 genes, 206 genes cluster diagram of 1083 selected genes and 10 MCC cell lines and were present on both the Human and Human Cancer four SCLC cell lines. A row in the cluster indicates expression of a 1.2 arrays and a high correlation was found between the specific gene across all the 14 samples. A column indicates the sample in which the gene is expressed. The color scale (Expression expression levels of these common genes for each sample Index) shown at the bottom (À3to þ 3 in log base 2 units) indicates (mean Spearman rank correlation coefficient ¼ 79.3%). that the relative expression level of the gene is greater, less than or Reliability and reproducibility of the array gene-expres- equal to the geometric mean expression across all 14 samples, sion data were further supported by (a) comparison of respectively. (b) Dendrogram representing similarities in the the array gene-expression levels and real-time RT–PCR expression patterns between experimental samples from Figure 1a. The symbols below the sample names reflect the different expression levels for 25 selected genes (mean Spearman (sub)groups of tumor cell lines (circle: Variant MCC; asterix: rank correlation coefficient ¼ 76.1%) (see below), (b) a Classic MCC; square: SCLC). (c) Real-time-based hierarchical highest degree of similarity evidenced by hierarchical cluster analysis of MCC cell lines and tumor samples for nine clustering for expression patterns of cell lines MCC14/1 genes, which can distinguish between the Classic and Variant MCC subtypes (asterix: Classic MCC cell lines; square: MCC tumors; and MCC14/2, derived from the same tumor (see circle: Variant MCC cell lines. (d) Average linkage hierarchical Figure 1b) and (c) confirmation of the differential gene cluster analysis of MCC and SCLC cell lines and tumor samples for expression for ASCL1 in SCLC versus MCC, as 14 genes identified by SAM as able to differentiate between MCC reported in the literature (see Discussion). (1, asterix) and SCLC (2, square) and quantified by real-time PCR. Same color scale as above for both figures, but expressed in log 10 base units Hierarchical clustering analysis of MCC and SCLC cell lines gene-expression levels of the 1083 preselected genes. Hierarchical clustering was used to identify similarities Figure 1a shows the complete cluster diagram. The in gene-expression patterns between MCC and SCLC dendrogram (Figure 1b) summarizes the degree of cell lines. Clustering of the 14 samples was based on the similarity in gene expression among the 14 analysed

Oncogene Gene-expression profiling of MCC and SCLC MV Gele et al 2734 samples. Two major subgroups were observed. Except median ‘false discovery rate’ (FDR) of 2.5%, which for MCC cell line UISO, group 1 contained all Variant means that there are about six false positives on average. MCC cell lines (circle:MCC26, MCC13, MCC14/2 and Highly differentially expressed significant genes were MCC14/1 (see Table 1)), indicating that these samples further selected if a differential expression pattern resemble each other in their gene-expression patterns. (4two-fold difference) was present in at least four of Group 2 included UISO, and all Classic MCC cell lines the five Classic MCC cell lines as compared to the (asterix:MKL-2, MKL-1, T95-45, MCC6 and MCC5) Variant ones and vice versa. In all, 46 genes with a which were mixed with all the four SCLC cell lines relative elevated expression in MCC Classic cell lines (square), of which GLC4 was the only Variant one (see and 30 genes with an increased expression level in the Table 1). From this analysis, we concluded that Variant MCC cell lines were identified (see Table 2). clustering of MCC cell lines into different subgroups Hierarchical cluster analysis of these 76 genes clearly predominantly coincided with the Classic or Variant classified the MCC cell lines in their respective phenotypes of the respective cell lines. However, phenotypic groups (data not shown). Genes with kinase hierarchical clustering of the 1083 preselected genes activities but also genes encoding for ligand and voltage- could not unequivocally separate MCC cell lines from gated ion channels, neuromediators, GDP/GTP exchan- SCLC cell lines. This does, however, further support a gers and signal-transduction receptors were seen at putative ontogenetic relationship for both tumors. higher expression levels in Classic MCC cell lines relative to Variant ones. Genes with a higher expression Identification of differentially expressed genes in Classic in Variant cell lines compared to Classic ones included versus Variant MCC cell lines genes involved in cell cycle control and proliferation (see Table 2). Unsupervised analysis of array data enables coherent patterns of gene expression to be identified, but provides Identification of differentially expressed genes in MCC little information about the statistical significance. versus SCLC Therefore, we decided to exploit a supervised strategy in order to identify a specific set of genes whose Hierarchical clustering analysis performed in this study expression pattern could discriminate Classic versus showed a high degree of similarity between MCC and Variant MCC cell lines. To this purpose, Classic (n ¼ 5) SCLC. We were interested, however, to search for genes and Variant (n ¼ 5) MCC cell lines were predefined as which could distinguish between the two tumor types. the two sample groups. Subsequently, a two-class SAM Therefore, the same supervised strategy as outlined analysis on the log transformed data matrix containing above was applied to identify genes differentially 1365 genes (see Material and methods) was performed. expressed in MCC versus SCLC cell lines. Due to the A cutoff value delta, depending on an arbitrary false- small gene-expression level variances between MCC and positive rate, was chosen to identify significantly SCLC, as a result of their similarity, we had to choose a differentially expressed genes. For this analysis, a delta rather low delta (D ¼ 0.33878) in order to include as value of 0.90572 was used. This led to the identification many genes as possible with significant but sometimes of a total of 239 differentially expressed genes with a small expression differences in MCC versus SCLC and

Table 1 Characteristics of the MCC and SCLC cell lines used for cDNA array hybridization Cell line Morphological type Colony shape Colony aggregation Classiﬁcationa

MCC cell lines MCC5 I 3-db Tight Classic MCC6 I 3-d Tight Classic MCC13 IV Flat NAc Variant MCC14/1 IV Flat NA Variant MCC14/2 IV Flat NA Variant MCC26 IV Flat NA Variant UISO IV Flat NA Variant MKL-1 III 2-d Loose Classic MKL-2 III 2-d Loose Classic T95-45 II 3-d Loose Classic

SCLC cell lines NCI-H69 II 3-d Loose Classic NCI-H146 II 3-d Loose Classic COR-L88 III 2-d Loose Classic GLC4 III 3-d Loose Variant

aClassic MCC cell lines express neuroendocrine markers including neuron-speciﬁc enolase and Chromogranin A, and contain neurosecretory granules (Leonard et al., 1993). Variant MCC cell lines have a selective loss of neuroendocrine markers including Chromogranin A, and do not contain neurosecretory granules, as evidenced by electron microscopy (Leonard et al., 1995a). Classic SCLC cell lines express levels of L-dopa decarboxylase and bombesin while Variant ones have undetectable levels of L-dopa decarboxylase and bombesin (Carney et al., 1985; Gazdar et al., 1985). b3-d; three-dimensional clusters, 2-d; two-dimensional clusters. cNot applicable, adherent growing cell lines.

Oncogene Gene-expression proﬁling of MCC and SCLC MV Gele et al 2735 Table 2 List of Classic and Variant MCC classiﬁer genes Spota GB Accb Symbol Gene/protein identity Fold chc Chrom locd Core P-value

Classic speciﬁc genes C_E09m D78345 IGHG3 Membrane-bound and secreted immunoglobulin 100 14q32.33 ND ND gamma heavy chain C_E08k U78095 SPINT2 Kunitz-type serine protease inhibitor 2 50 19q13.1 83.5 3.80E-04 C_D02h U96136 CTNND2 Delta catenin 38.46 5p15.2 ND ND H_D01d X53179 CHRNB2 Cholinergic receptor nicotinic beta polipeptide 2 25 1q21.3 ND ND H_D08e J05252 PCSK2 Neuroendocrine convertase 2 20 20p11.2 ND ND H_D07h L19761 SNAP25 25-kDa synaptosomal-associated protein 20 20p12–11.2 ND ND H_F06k X51405 CPE Carboxypeptidase H 11.11 4q32.3 ND ND H_B13d D10924 CXCR4 CXC chemokine receptor type 4 11.11 2q21 ND ND C_E09h AF003521 JAG2 Jagged homolog 2 11.11 14q32 ND ND C_A10n U72649 BTG2 NGF-inducible anti-proliferative protein PC3 10 1q32 93.4 1.03EÀ06 H_A13n U07139 CACNB3 Dihydropyridine-sensitive L-type calcium 9.09 12q13 ND ND channel beta-3 subunit H_B05f X74979 DDR1 Epithelial discoidin domain receptor 1 8.33 6p21.3 ND ND C_F09a U86759 NTN2L Netrin-2 8.33 16p13.3 ND ND H_D01g Y00757 SGNE1 Secretory granule endocrine protein I 8.33 15q13–14 ND ND C_B10h L42374 PPP2R5B Protein phosphatase 2A B56-beta 8.33 11q12 ND ND C_A04m X80343 CDK5R1 Cyclin-dependent kinase 5 activator 7.69 17q11.2 ND ND C_D09j AF011466 EDG4 G protein-coupled receptor EDG4 7.69 19p12 ND ND C_B13d L07597 RPS6KA1 Ribosomal protein S6 kinase II alpha 1 7.69 3 ND ND C_C12k U66197 FGF12 FHF-1 7.14 3q28 ND ND H_C05b L05500 ADCY1 Adenylate cyclase type I 7.14 7p13–12 ND ND H_B14a L14595 SLC1A4 Neutral amino-acid transporter A 6.67 2p15–13 ND ND H_C14e M23410 JUP Junction plakoglobin 6.67 17q21 94.5 1.12EÀ06 C_B02f D50925 PASK KIAA0135 6.67 2q37.3 ND ND C_C02l Y11416 TP73 p73 6.67 1p36.33 ND ND C_F13b L11931 SHMT1 Cytosolic serine hydroxylmethyltransferase 6.67 17p11.2 ND ND C_B08i U27193 DUSP8 Dual-speciﬁcity protein phosphatase 8 6.25 11p15.5 ND ND C_C02f L07540 RFC5 Replication factor C 36-kDa subunit 6.25 12q24.2–24.3 ND ND H_C12k L09561 POLE DNA polymerase II subunit A 6.25 12q24.3 ND ND C_F13n U96876 INSIG1 Insulin-induced protein 1 5.89 7q36 ND ND C_D04l Y08110 LR11 Low-density lipoprotein receptor-related 5.56 11q23.2–24.2 ND ND protein LR11 C_C03b U77352 MADD MAPkinase-activating death domain protein 5.27 11p11.2 ND ND H_D07g AF040255 DCX Doublecortin 5.26 Xq22.3–23 ND ND C_A11f AF006484 CDK2AP1 Putative oral tumor-suppressor protein 4.76 12q24.31 ND ND C_B13l D84064 HGS HRS 4.55 17q25 ND ND H_B03m X80907 PIK3R2 Phosphatidylinositol 3-kinase regulatory 4.55 19q13.2–13.4 ND ND beta subunit H_E12f U04810 TROAP Trophinin-associated protein 4.55 12p11.1 ND ND H_A05k U40343 CDKN2D Cyclin-dependent kinase 4 inhibitor 2D 4.35 19p13 ND ND C_C08e U86529 GSTZ1 Glutathione transferase zeta 1 4 14q24.3 ND ND H_C13k D38073 MCM3 MCM3 DNA replication licensing factor 4 6p12 ND ND H_C08g M93426 PTPRZ1 Protein-tyrosine phosphatase zeta 4 7q31.3 ND ND C_D11e U24497 PKD1 Autosomal dominant polycystic kidney 3.57 16p13.3 ND ND disease protein 1 H_C05e X70326 MLP MARCKS-like protein 3.33 1p34.3 75.8 2.67EÀ03 C_E05h X61157 ADRBK1 Beta-adrenergic receptor kinase 1 3.33 11q13 ND ND H_B01a X91906 CLCN5 Chloride channel protein 5 2.86 Xp11.23–11.22 ND ND H_A07n U69883 KCNN1 Calcium-activated potassium channel HSK1 2.7 19p13.1 ND ND H_A08i X60188 MAPK3 Mitogen-activated protein kinase 3 2.38 16p12–11.2 ND ND

Variant speciﬁc genes C_F04g X56134 VIM Vimentin 42.71 10p13 90.8 7.30EÀ06 H_A07d X16707 FOSL1 FOS-like antigen 1 25.05 11q13 80.8 8.39EÀ04 H_A05e M76125 AXL axl oncogene 21.14 19q13.1 90.3 9.56EÀ06 H_A03h X59798 CCND1 G1/S-speciﬁc cyclin D1 16.56 11q13 83.9 2.17EÀ04 H_F05n X03124 TIMP1 Tissue inhibitor of metalloproteinase 1 16.09 Xp11.3–11.23 ND ND H_C06h U28014 CASP4 Caspase 4 10.57 11q22.2–22.3 ND ND H_A11g X57766 MMP11 Matrix metalloproteinase 11 9.15 22q11.23 ND ND H_C14h M23254 CAPN2 M-type calcium-activated neutral proteinase 9.13 1q41–42 ND ND H_E03f M24069 CSDA Cold shock domain protein A 8.31 12p13.1 ND ND C_F09g M96322 AKAP12 Gravin 7.85 6q24–25 ND ND H_D10j Z36715 ELK3 ets domain protein elk-3 7.59 12q23 ND ND H_E09g M73780 ITGB8 Integrin beta 8 7.32 7p21.3 ND ND H_F03b M34664 HSPD1 Heat shock 60-kDa protein 7.19 2q33.1 ND ND C_F14a J03040 SPARC Secreted protein acidic and rich in cysteine 7.04 5q31.3–32 ND ND H_E13k X66945 FGFR1 Fibroblast growth factor receptor 1 6.78 8p11.2–11.1 ND ND C_D09m U90313 GSTTLp28 Glutathione-S-transferase-like protein 6.67 10q24.33 ND ND

Oncogene Gene-expression proﬁling of MCC and SCLC MV Gele et al 2736 Table 2 (continued ) Spota GB Accb Symbol Gene/protein identity Fold chc Chrom locd Core P-value

H_B12g U48959 MYLK Smooth muscle and nonmuscle myosin 6.18 3q21 ND ND light chain kinase C_B10c L31951 MAPK9 Mitogen-activated protein kinase 9 6.16 5q35 ND ND H_F02g U10117 SCYE1 Endothelial-monocyte activating 6.12 4q24 ND ND polypeptide II C_C10b U60520 CASP8 Caspase 8 5.74 2q33–34 ND ND C_E03i AF031385 CYR61 Cysteine-rich anigogenic inducer 61 4.97 1p31–22 ND ND H_B11e M59371 EPHA2 Ephrin type-A receptor 2 4.87 1p36 ND ND H_D08i M13667 PRNP Major prion protein 4.66 20pter-p12 ND ND H_F02l D00760 PSMA2 Proteasome subunit alpha type 2 4.23 7p13–12 ND ND H_A02d J04101 ETS1 ets1 proto-oncogene 3.93 11q23.3 ND ND C_A05i M25753 CCNB1 G2/mitotic-speciﬁc cyclin B1 3.83 5q12 ND ND H_A12c V00568 MYC myc proto-oncogene 3.67 8q24.12–24.13 94.7 1.00EÀ06 C_E11l M34671 CD59 CD59 glycoprotein 3.64 11p13 ND ND H_E13g D13866 CTNNA1 Alpha1 catenin 3.53 5q31 ND ND H_F07k X87212 CTSC Cathepsin C 3.33 11q14.1–14.3 ND ND

aSpot location of each gene on either the Atlas Human (H) or Cancer (C) arrays. bGenBank accession number. cFold change was calculated by dividing the mean expression level of all Classic MCC cell lines to the mean expression level of all Variant MCC cell lines for Classic specific genes (and vice versa for Variant specific genes). dChromosomal location of each gene. eExpression of a number of genes was confirmed by real-time quantitative RT–PCR. The Spearman rank correlation coefficient (%) was calculated between the array gene-expression levels and the real-time RT–PCR levels. ND, not done.

vice versa. This led to the identification of a total of 121 genes being identified in this analysis illustrated the high differentially expressed genes with a median FDR of degree of similarity at the RNA expression level between 27.9%, that is, 33.8 false-positive genes on average. In MCC and SCLC, and further supports a putative addition, the 121 SAM-identified genes were further ontogenetic relationship between both tumor types. selected as highly differentially expressed significant genes in MCC versus SCLC if a differential expression Validation of array gene-expression levels and pattern (4two-fold difference) was present in at least classification of additional MCC and SCLC cell lines/ four of the MCC cell lines as compared to SCLC or in at tumor samples by real-time RT–PCR analysis least three of the four SCLCs as compared to MCC. In this way, the use of a low delta value merged with a To verify the array gene-expression levels by an relatively high FDR was justified. Subsequently, 12 and independent and sensitive method, we performed real- initially seven genes showed higher and lower expression time quantitative RT–PCR on the same RNA samples levels respectively in MCC versus SCLC (see Table 3 and of the 14 cell lines used for filter array analysis. In all, 25 footnote f). Genes more highly expressed in MCC SAM selected genes were quantified. Of these, 16 genes included, for example, brain-derived neurotrophic factor, (ASCL1, GPX2, ID2, TFAP4, FLT1, IGFBP2, PRKCA, heat shock-related 70-kDa protein 2, neurogranin and ITGA3, BDNF, ILK, PRKR, CDC25B, EGR1, CHD2, intergrin alpha 3. In three of the four SCLC cell lines, MAP2K3 and HSPA2) were previously selected by their elevated levels of the neuronal differentiation marker ability to distinguish between MCC and SCLC (see ASCL1 (achaete-scute homolog 1), the basic helix– Table 3). The other nine genes selected arbitrarily from loop–helix transcription factor ID2 (inhibitor of DNA- the SAM list (SPINT2, MYC, AXL, CCND1, JUP, binding protein 2) and GPX2 (glutathione peroxidase- BTG2, FOSL1, VIM and MLP) had Classic versus related protein 2) were observed relative to those seen in Variant classification capability (see Table 2). In general, MCC cell lines. The lower expression level of ASCL1 in the quantitative RT-PCR data correlated very well with cell line GLC4 was of interest, as this was the only the array hybridization data (see Tables 2 and 3). Variant SCLC cell line examined. All others had a Figures 2a and b illustrate relative real-time RT–PCR Classic phenotype (see Table 1) consistent with ASCL1 data and array hybridization data for the genes ASCL1 being required for the neuroendocrine phenotype of and IGFBP2, with Spearman rank correlation coeffi- SCLC (Borges et al., 1997). A complete list of SAM cients of 70.5% (P-value ¼ 4.82EÀ03) and 92.1% genes differentially expressed between MCC and SCLC (P-value ¼ 2.98EÀ06), respectively. is given in Table 3. Two genes, FLT1 and EGR1 Average linkage hierarchical cluster analysis of 12 (Table 3, footnote f), were however excluded from MCC cell lines and 10 MCC tumors for nine SAM genes further analysis, as differences in array gene-expression with phenotypic classification potential and quantified levels were not confirmed by real-time quantitative RT– by real-time PCR resulted in separation of all (adherent) PCR analysis. The classification potential of the Variant MCC cell lines, all but one Classic MCC cell remaining 17 differentially expressed genes between line (MCC5) and nine of the 10 MCC tumors MCC and SCLC was visualized after re-clustering. Each (Figure 1c). These results illustrate that, even with a cell line clearly clustered into either the MCC or SCLC limited selected set of differential genes, a distinction subgroup (data not shown). The low number of classifier between Classic and Variant MCC cell lines can easily

Oncogene Gene-expression proﬁling of MCC and SCLC MV Gele et al 2737 Table 3 List of MCC and SCLC classiﬁer genes Spota GB Accb Symbol Gene/protein identity Fold chc Chrom locd Core P-value

MCC speciﬁc genes H_F09f M61176 BDNF Brain-derived neurotrophic factor 16.67 11p13 92.9 4.61EÀ06 H_F04a L26336 HSPA2 Heat shock-related 70-kDa protein 2 11.11 14q24.1 95.1 2.04EÀ06 H_C06c Y09689 NRGN Neurogranin 7.69 11q24 ND ND H_E06i M59911 ITGA3 Integrin alpha 3 7.14 17q21.32 85.9 8.20EÀ05 H_B03i M22199 PRKCA Protein kinase C alpha polypeptide 6.25 17q22–23.2 75.9 1.67EÀ03 H_C06j M35663 PRKR Interferon-inducible RNA-dependent protein kinase 4.17 2p22–21 89.9 1.24EÀ05 H_B12h L36719 MAP2K3 Mitogen-activated protein kinase kinase 3 3.7 17q11.2 80.7 4.91EÀ04 H_A01k D84212 STK6 Aurora-related kinase 1 3.57 20q13.2–13.3 ND ND H_E06h M34064 CDH2 Cadherin 2 2.94 18q11.2 83.3 2.17EÀ04 H_A08j D88435 GAK Cyclin G-associated kinase 2.86 4p16 ND ND H_B03g U40282 ILK Integrin-linked kinase 2.5 11p15.5–15.4 60.9 2.09EÀ02 C_A11l M81934 CDC25B Cell division cycle 25 homolog B 2.44 20p13 82.9 2.51EÀ04

SCLC specific genesf H_D01h L08424 ASCL1 Achaete-scute homolog 1 70.72 12q22–23 70.5 4.82EÀ03 H_F06a X53463 GPX2 Glutathione peroxidase-related protein 2 10.43 14q24.1 83.6 1.33EÀ03 H_D11m M97796 ID2 Inhibitor of DNA binding 2 protein 5.06 2p25 87.7 3.83EÀ05 C_A08f M35410 IGFBP2 Insulin-like growth factor-binding protein 2 3.33 2q33–34 92.1 2.98EÀ06 H_E04d S73885 TFAP4 AP4 basic helix–loop–helix DNA-binding protein 2.68 16p13 78.9 7.95EÀ04 aSpot location of each gene on either the Atlas Human (H) or Cancer (C) arrays. bGenBank accession number. cFold change was calculated by dividing the mean expression level of all MCC cell lines to the mean expression level of all SCLC cell lines for MCC-specific genes (and vice versa for SCLC specific genes). dChromosomal location of each gene. eExpression of a number of genes was confirmed by real-time quantitative RT–PCR. The Spearman rank correlation coefficient (%) was calculated between the array gene-expression levels and the real-time RT–PCR levels. ND, not done. fInitially identified SAM genes FLT1 and EGR1 were excluded for further analysis, as their array gene-expression levels were not confirmed by real-time RT–PCR analysis (Cor ¼À31.0%; P-value ¼ 2.81EÀ01 and Cor ¼À14.2%; P-value ¼ 6.26EÀ01, respectively). be made. Originally, the grouping of Classic and Variant MKL-1 (Figure 1c). Although PCR-based hierarchical phenotypes for MCCs was cell line based (Leonard et al., clustering of the nine selected SAM genes analysed in 1993; Leonard et al., 1995a, 2002; Leonard and Bell, this study could not classify MCC19 as a separate third 1997). However, we have observed concordant results biological MCC subgroup, as suggested by HATH1 between tumor samples and their respective cell lines in reactivity (Leonard et al., 2002), quantitative PCR several genomic deletion analyses (Leonard and Hayard, analysis of a larger panel of differentially expressed 1997; Leonard et al., 2000; Van Gele et al., 2000; Cook genes followed by clustering with an extended number et al., 2001) and in immunohistochemical studies. In of Variant MCC suspension cell lines may enable this to particular, the transcription factor HATH1, shown to be occur. expressed in normal Merkel cells and in Classic MCC In order to confirm the classification strength of the cell lines was also expressed only in those biopsies from SAM selected genes (MCC versus SCLC), real-time RT– which Classic MCC cell lines, were derived (Leonard PCR analysis was extended to two further MCC cell et al., 2002). In the present study, almost all MCC lines, 10 MCC tumor samples, 12 additional SCLC cell tumors clustered in one group which showed a higher lines and two SCLC tumors for ASCL1, GPX2, ID2, degree of similarity with the Classic cell lines compared TFAP4, IGFBP2, PRKCA, ITGA3, BDNF, ILK, PRKR to the Variant cell lines, although they shared some CDC25B, CHD2, MAP2K3 and HSPA2. Average gene-expression features common to both cell line linkage hierarchical cluster analysis based on the groups. Given the concordance previously seen between Spearman rank correlation coefficient as a similarity tumors and their respective cell lines, these groupings measure showed two major clusters (Figure 1d). Except are likely to have clinical significance and only further for one SCLC cell line (NCI-N464), cluster 1 contained studies on additional tumors would determine if those all MCC cell lines and MCC tumor samples. Cluster 2 examined could be typed as having a ‘Classic’ consisted of all the remaining SCLC cell lines and SCLC phenotype. tumors. Real-time PCR-based gene-expression profiling It should be mentioned that cell line MCC19 was therefore resulted in an almost perfect classification of recently thought to be a Variant MCC suspension cell the different tumors or cell lines into their respective line based on the lack of HATH1 expression (Leonard tumor groups. In addition, the cell lines MCC14/1 and et al., 2002). However, MCC19 still expresses, similar to MCC14/2 established from the same tumor sample Classic MCC suspension cell lines, the transcription remained clustered next to each other in a subgroup factor Brn-3c (Leonard et al., 2002) and Chromogranin with other Variant MCC cell lines. The cell lines MKL-1 A, a neuroendocrine marker. Therefore, it is probably and MKL-1 clone 2, also derived from a same patient, not surprising that MCC19 is outlying the Variant were found in different subclusters. The MKL-1 clone 2 cluster group, but shows instead a high degree of has, however, been grown separately from MKL-1 for a similarity with Classic MCC suspension cell lines such as long time, and has clonally evolved, apart from

Oncogene Gene-expression profiling of MCC and SCLC MV Gele et al 2738 with their Classic or Variant phenotypes. On the other hand, this analysis could not distinguish between MCC and SCLC, emphasizing their biological/genetic relationship. Therefore, we adopted a supervised data-mining strategy, that is, two-class SAM analysis, in order to identify (a) phenotypic classifier genes which allow to separate Classic from Variant MCC cell lines and (b) diagnostic classifier genes which may aid in the differential diagnosis of MCC and SCLC. This led to the identification of 76 highly differentially expressed significant genes, of which 46 showed higher expression in the Classic cell lines and 30 were more highly expressed in the Variant MCC cell lines. A subset of genes with higher levels of expression in the Classic cell lines are involved in signal-transduction pathways, and could lead to increased cell growth when over- expressed. This is exemplified by genes such as MAPK3 and MADD involved in the mitogen-activated protein (MAP) kinase pathway, and genes such as PI3-K p85 beta in the phosphatidylinositol 3-kinase (PIK3) pathway. In addition, Classic cell lines showed higher levels of expression of genes coding for neuromediators (SGNE1) and neurotransmittors (PCSK2), and proteins involved in neuronal development such as doublecortin (DCX) and MARCKS-like protein (MLP). This is in keeping with the observed neuroendocrine and more differentiated phenotypes associated with the Classic MCC cell lines. Ligand and voltage-gated ion channels and receptors essential for neurotransmission were also expressed in the Classic cell lines. Some of these ion Figure 2 Histogram comparing relative real-time PCR expression channels are known to play a role during mechanical levels (gray bars) and array hybridization levels (white bars) of (a) stimulation of normal Merkel cell receptors (Baumann ASCL1 and (b)ofIGFBP2 in MCC and SCLC cell lines (ordinate, et al., 2000; Tazaki et al., 2000). Variant MCC cell lines log10 space). The normalized expression level of each gene was could have lost expression of some of these ion channels. divided by its geometric mean across all 14 samples. The Spearman rank correlation coefficient (Sp r) for both genes is indicated on the Their specific function in MCC tumor cells, however, histogram has yet to be elucidated. Genes with higher expression in Variant cell lines were involved in cell cycle control (CCND1 and CCNB1) and becoming karyotypically unrelated to MKL-1 (Rosen cell proliferation (HSP60, MMP11, MAPK9, FOSL1, et al., 1987; Van Gele et al., 2002). AXL, MYC and ETS1). Some of these genes may The value of using cell lines for gene-expression correlate with the shorter doubling time and aggressive profiling and validation of differentially expressed genes nature of the Variant MCC cell lines, as illustrated by for tumor types such as SCLC where tumor material is their higher cloning efficiency in soft agar and their very difficult to obtain in sufficient amount or numbers reduced sensitivity to radiation (Leonard et al., 1995b). has recently been demonstrated in a global gene In addition, we observed high expression of vimentin,a expression analysis by Pedersen et al. (2003). Therefore, mesenchymal marker, together with FOSL1 (alias we believe that, in this study, the use of a limited number FRA1, FOS-related antigen 1). A tight correlation of of available SCLC tumors combined with a larger vimentin and FOSL1 expression was also recently found number of SCLC cells in culture for validation of SCLC in highly invasive breast cancer cell lines, pointing to a classifier genes is justified and reliable. possible role in tumor progression and enhanced cell migration of these cancer cells (Zajchowski et al., 2001). These two genes could be significant prognostic markers Discussion for the more aggressive MCC Variant types. Increased expression of vimentin was also observed by immuno- In this study, expression profiling of 10 MCC and four histochemical studies in Variant SCLC cell lines (Broers SCLC cell lines was performed through analysis of 1891 et al., 1985, 1986), and as a result of a suppression unique genes. Hierarchical clustering was used in a first subtractive hybridization experiment comparing a Clas- general attempt to assess the classification power of the sic to a Variant SCLC cell line (Zhang et al., 2000). obtained expression data set. Cluster analysis of 1083 These observations could point at a similar mechanism preselected genes allowed the MCC cell lines to of tumor progression or metastatic properties between segregate into two different subgroups mainly associated MCC and SCLC Variant phenotypes.

Oncogene Gene-expression profiling of MCC and SCLC MV Gele et al 2739 To extend and validate the information from the (Botchkarev et al., 1999). Higher expression in MCC as SAM analysis data set, nine genes able to differentiate compared to SCLC could contribute, in patients, to an between Classic and Variant cell lines were arbitrarily increase in the numbers of Merkel cells typically selected from the list and quantified by real-time PCR in observed in MCC tumors (Moll et al., 1996). 12 MCC cell lines and 10 other MCC tumors. Cluster Consequently, if BDNF could be downregulated in analysis of the PCR results confirmed the separation of MCC patients, this might have an antiproliferative the phenotypically different MCC cell lines, and effect. This hypothesis warrants further investigation. illustrates that, even with a limited selected set of Expression of CDC25B was recently observed in SCLC differential genes, a distinction between Classic and cell lines (Pedersen et al., 2003), and we showed now Variant MCC cell lines can be made. even higher levels of expression in MCC, suggesting that In addition, comparison of our results found for a possible upregulation of CDC25B in MCC may occur. MCC with recently published gene-expression profiles of None of the other 10 genes were previously shown to be SCLC cell lines by Pedersen et al. (2003) revealed a implicated in MCC and SCLC biology. One striking number of identical genes expressed in the Classic cell finding was the differential expression of the alpha lines, but not in the Variant ones of both tumor types. subunit of PKC (PRKCA) in MCC compared to SCLC These included the neuroendocrine gene SGNE1 and cell lines. Protein kinase C is a key protein involved in neuronal markers doublecortin and MACMARKS or the regulation of cell growth and activation of the MAP MLP. These genes could be used as markers for kinase pathway (Buchner, 2000), and this finding is in distinction between Classic and Variant phenotypes of keeping with the observed expression for MAP2K3. both tumor types. Moreover, the nine selected SAM Both overexpression and downregulation of PRKCA genes able to distinguish between Classic and Variant have been previously observed in different human tumor MCC cell lines could also be used as classifiers for types and correlated with malignant transformation and Classic and Variant SCLC cell lines (personal observa- proliferative activity of PRKCA (Benzil et al., 1992; tion). Two of these nine genes, vimentin (see also above) Scaglione-Sewell et al., 1998). PRKCA could thus be and SPINT2, were previously found to be differentially involved both in MCC and SCLC, albeit through a expressed between a Classic and more aggressive different mechanism in each of these tumor types. Variant SCLC cell line after a suppression subtractive Further investigation of this gene and other MCC and hybridization experiment (Zhang et al., 2000). Interest- SCLC classifier genes such as ITGA3, HSPA2, CDH2, ingly, SCLC tumors derived from Variant cell lines are NRGN, GAK, PRKR and STK6 should elucidate their more aggressive and patients have a worse prognosis role in MCC and/or SCLC biology. (Gazdar et al., 1985). Comparable to SCLC, lack of Five of the 17 SAM-identified MCC and SCLC expression of Classic marker genes described here or classifier genes had a higher expression level in SCLC as overexpression of Variant MCC classifier genes could compared to MCC. Our data confirmed the previously indicate a subset of more aggressive MCCs, for which reported expression of the neuroendocrine differentia- more intensive treatment and closer follow-up may be tion marker ASCL1 in (Classic) SCLCs and its lack of warranted in a similar way to our recent results for Brn- expression in MCC cell lines (Bhattacharjee et al., 2001; 3c/HATH1 expression (Leonard et al., 2002). Real-time Garber et al., 2001; Leonard et al., 2002; Pedersen et al., PCR analysis of 10 MCC tumors did not separate them 2003). The ID2 basic helix–loop–helix transcription into distinct subgroups. However, the number of Classic factor demonstrated higher levels of transcripts in versus Variant classifier genes analysed by real-time Classic SCLCs compared to MCCs. Gene-expression PCR in this study was limited to nine genes, and profile analysis of small-cell lung cancer cells by extension of this panel and also increased numbers of Pedersen et al. (2003) also detected expression of ID2 patients, for which survival data and treatment proce- in SCLC. ID2 plays a role in cell proliferation and dures are also available, might be beneficial. In addition, differentiation and is able to disrupt the antiproliferative future investigations of genes or disregulated pathways effects of retinoblastoma family members (Iavarone involved in Classic and Variant MCC and/or SCLC et al., 1994). It is possible that disruption of the RB1 cell lines could lead to potential targets for the pathway through increased expression of ID2 could be development of new therapeutic strategies specific for an important mechanism in neuroendocrine SCLCs each (sub)group. which may not occur in MCC. For the three remaining A second goal of the study was to identify genes which genes (GPX2, IGFBP2 and TFAP4) expressed in SCLC were able to distinguish MCC from SCLC. This led to cell lines but not in MCC, no previous involvement in the identification of 17 classifier genes whose gene- SCLC has been described. Further investigation of these expression levels showed significant differential gene genes should clarify their role in SCLC or MCC biology. expression between MCC and SCLC samples. In all, 12 Our gene-expression profiling and clustering with only of these genes showed a higher expression in MCC as 17 MCC and SCLC classifier genes identified through compared to SCLC. Of particular interest was brain- SAM analysis was extended through real-time RT–PCR derived neurotrophic factor (BDNF), which is known to on the original cell lines, as well as additional cell lines stimulate the mechanotransducing properties of normal and tumor samples. Our results showed that the selected Merkel cells (Carroll et al., 1998). In addition, over- genes were able to effectively cluster the samples, expression of BDNF in murine skin was shown to be providing an additional and simple test to differentiate associated with an increase in Merkel cell number between MCC and SCLC.

Oncogene Gene-expression profiling of MCC and SCLC MV Gele et al 2740 In conclusion, we generated a gene expression-based according to the Clontech Atlas cDNA Expression Arrays classification of at least two biological and possibly User Manual. Purification of the probe, hybridization and clinically distinct subgroups of MCC. Interestingly, washes were performed by following the manufacturer’s some of the differentially expressed genes are typically instructions. Each RNA sample was simultaneously hybridized classifying Classic and Variant phenotypes of MCC as to both filters in the same hybridization bottle. After the washes, membranes were exposed for one to three nights to well as SCLC. Further investigation could result in a phosphoimager plates and scanned with a PhosphoImager more selective therapeutic treatment applicable for both System using ImageQuaNT (Molecular Dynamics – tumor types and improvement of patient outcome. In Amersham Biosciences). addition, we demonstrated for the first time a high degree of similarity at the gene-expression level between Analysis of cDNA arrays MCC and SCLC. Furthermore, we were able to identify a subset of genes by supervised analysis, which may be The scanned gel images were converted to a 16-bit tagged image file format. Signal intensities were quantified using the helpful in the differential diagnosis of MCC and SCLC. VisualGrid software version 2.1 (GPC Biotech). The Arra- Our study also serves as the first step towards a further yAn2 software (T. Boonefaes, P. Rottiers, and J. Grooten. detailed study of differentially expressed genes involved ArrayAn2: optimized algorithms for primary data analysis of in cell proliferation, signal transduction and neuro- cDNA arrays, manuscript in preparation) was used for further transmission, finally leading to more insight into the primary data analysis. In short, the spot intensities were complex and heterogeneous biology of MCC and corrected for the local background signal intensity, followed SCLC. by a spot quality-control step to exclude spots influenced by intense signals of adjacent spots. The detection limit for expression values above background was calculated based on the variation of the local background intensity. Constitutive Materials and methods genes were selected (50% of the spots showing the lowest coefficient of variation over all arrays) and used for Cell lines normalization. Merkel cell carcinoma cell lines MCC5, MCC6, MCC13, MCC14/1, MCC14/2 and MCC26 were established at the Expression data analysis Queensland Radium Institute Research Unit, Queensland, Genes with an expression value above the background level in Australia and have been described in detail previously by at least six of the analysed samples were selected for further Leonard et al. (1993, 1995a). MCC cell line UISO was analysis. This resulted in a total of 1083 genes, of which 412 described by Ronan et al. (1993) and MKL-1 by Rosen et al. were common genes (i.e. 206 genes were present on both arrays (1987). MKL-2 was established at the Robert H Lurie (Human and Human Cancer 1.2). Cluster and Treeview Comprehensive Cancer Center, IL, USA and reported by software were used for unsupervised hierarchical clustering Van Gele et al. (2002). T95-45 was established at the Center for and visualization of the data (Eisen et al., 1998). Prior to Medical Genetics, Ghent, Belgium. Most cell lines were clustering, genes were mean centered and the expression data previously analysed by CGH and/or multiplex fluorescence matrix was log transformed (base 2). Subsequently, complete in situ hybridization (Van Gele et al., 1998, 2002). Small-cell linkage clustering using Spearman rank correlation coefficient lung carcinoma cell lines NCI-H69 and NCI-H146 were as similarity metric was performed to the samples and genes. obtained from the American Type Culture Collection, COR- The complete expression data matrix is available as a tab L88 was a gift from Dr PTwentyman, Cambridge, UK and delimited file from the authors on request. GLC4 was a gift from Dr Marc Maliepaard, Amsterdam, The We used the Significance Analysis of Microarrays or SAM Netherlands. The morphological characteristics, growth beha- algorithm (Tusher et al., 2001), which allows supervised vior and subgroup classification (Classic versus Variant) of the identification of significantly differentially expressed genes cell lines are summarized in Table 1. All cell lines were grown between predefined sample groups. In order to include less to 70% confluency in RPMI 1640 (Invitrogen) supplemented representative genes for the SAM analysis, the filter threshold with antibiotics, 10% fetal calf serum and 1% L-glutamine. was lowered by including genes expressed above background 2 Cells from 10 tissue culture flasks (75 cm ) were pelleted, quick in at least four of the analysed samples (1365 genes). frozen in liquid N2 and stored at À801C. Real-time quantitative RT–PCR cDNA array hybridization In all, 25 SAM identified genes were quantified by real-time Total RNA from cell lines was extracted using Proteinase K quantitative RT–PCR on the same 14 RNA samples as for and phenol/chloroform (Sigma), followed by a sodium acetate array hybridizations. In order to validate the array gene- precipitation in ethanol (MCC13, MCC14/1, MCC14/2 and expression data, the normalized array and real-time PCR data MCC26) or the Atlas Pure Total RNA Labelling System were each mean centered for the genes. The Spearman rank (Clontech – BD Biosciences). The resuspended RNA was correlation coefficient was then calculated between the array subsequently DNase I (Roche) (2 U/50 mg) treated. The quality gene-expression levels and real-time PCR expression levels for and integrity of the Dnase-treated RNA were checked by each gene using the Statistical Package for the Social Sciences ethidium bromide agarose gel electrophoresis. Expression (SPSS) Version 11.0 software. Primer sequences for all 25 analysis was performed using the Atlas Human 1.2 (7850-1) genes were designed with Primer Express 1.0 software (Applied and Atlas Human Cancer 1.2 (7851-1) nylon arrays (Clontech Biosystems) using the default TaqMan parameters, with – BD Biosciences). Both filters contained 1176 genes, of which modified minimum amplicon length requirements (75 bp). 461 were present on both arrays. For each sample, 12.5 mgof The primer sequences are submitted in a public database total RNA was used in the cDNA probe synthesis with (RTPrimerDB) for real-time PCR primers and probes (Pattyn [a-32P]dATP (NEN Life Science Products) and performed et al., 2003) (gene: primer-ID; ASCL1: 373, GPX2: 346, ID2:

Oncogene Gene-expression profiling of MCC and SCLC MV Gele et al 2741 102, IGFBP2: 349, FLT1: 348, TFAP4: 347, EGR1: 389, standard curve method (serial dilutions of a cDNA mixture BDNF: 352, HSPA2: 390, ITGA3: 351, PRKCA: 350, PRKR: containing two SCLC and two MCC samples) or the 354, MAP2K3: 388, CDH2: 387, ILK: 353, CDC25B: 355, comparative Ct method was used for quantification. PCR VIM: 356, FOSL1: 357, AXL: 360, CCND1: 87, MYC: 18, reagents were obtained from Eurogentec as SYBR Green I SPINT2: 358, BTG2: 361, JUP: 359, MLP: 362). In order to mastermixes, and used according to the manufacturer’s confirm the classification potential of the above-mentioned instructions. PCR reactions were run on an ABI Prism 5700 genes, real-time PCR analysis was extended to two further Sequence Detection System (Applied Biosystems). To correct MCC cell lines, 10 MCC tumors, 12 additional SCLC cell lines for differences in RNA quantities and cDNA synthesis and two SCLC tumors. Tumor samples (MCCT1, T2, T3, efficiency, relative gene-expression levels were normalized SCLCT1 and T2) were collected at the Department of using the geometric mean of five housekeeping genes (UBC, Dermatology or Pathology, Ghent University Hospital, HPRT1, GAPD, TBP and HMBS) according to Vandesom- Ghent, Belgium, and tumor samples MCCT4, T5, T6, T7, pele et al. (2002b). In order to perform hierarchical clustering, T8, T9 and T10 were collected at the Department of a real-time-based expression matrix was created by dividing Pathology, University Hospital, Leuven, Belgium. Tumor the normalized gene-expression level of each gene by its biopsies were homogenized with an Ultra-Turrax T25 (IKA- geometric mean across all samples, and data were subsequently Werke) in 2 ml lysis buffer (Qiagen). Total RNA of biopsies log transformed (base 10). was extracted using the RNeasy Midi Kit (Qiagen), according to the manufacturer’s instructions. Total RNA from MCC cell line MCC19 (Type II, suspension cell line with three- Acknowledgements dimensional loose colonies thought to be Variant (Leonard This work was supported by GOA Grant 12051397, FWO et al., 2002)) was isolated at the Queensland Radium Institute Grant G.0028.00 and the Queensland Cancer Fund and the Research Unit using Total RNA Isolation Reagent (Applied Queensland Radium Institute Research Unit. Anthony L Biotechnologies) from cells in exponential growth. The RNA Cook is supported by a University of Queensland Mid Year of SCLC cell lines NCI-H446, POVD and AFL was a gift from Scholarship. Jo Vandesompele is sponsored by VEO-grant Dr G Sozzi (Milan, Italy), RNA of GLC1, GLC7, GLC28, 011V1302. Nadine Van Roy is a postdoctoral researcher of the GLC36 and GLC45 was kindly provided by Dr K Kok Fund for Scientific Research, Flanders. This paper presents the (Groningen, The Netherlands) and RNA of NCI-H60, NCI- research results of the Belgian program of Interuniversity Poles H82, NCI-H250, NCI-N464 and MCC cell line MKL-1 of attraction initiated by the Belgian State, Prime Minister’s (subclone 2) (Type III, suspension cell line with two- Office, Science Policy Programming. The scientific responsi- dimensional loose colonies classified as Classic) was a gift bility is assumed by us. We would like to thank Drs MM from H Salwen (IL, USA). All RNAs were quantified using the Hughes, VM Hinkley, RW Allison, W Cockborn, TJ Harris Ribogreen reagent (Molecular Probes) on a TD-360 fluorom- and O Williams for their support in collecting the Queensland eter (Turner Designs). The relative gene-expression levels were MCC specimens, from which the MCC cell lines were determined using an optimized two-step SYBR green I RT– established, and H Salwen for providing MCC cell line PCR assay, as described by Vandesompele et al. (2002a). The MKL-2.

References

Baumann KI, Senok SS, Chan E and Yung WH. (2000). Cook AL, Pollock PM, Welch J, Walsh MD, Bowman RV, Merkel Cells, Merkel Cell Carcinoma and Neurobiology of Baumann KC, Hayward NK and Leonard JH. (2001). Int. J. the Skin, Suzuki H and Ono T (eds) Elsevier Science B.V.: Cancer, 93, 361–367. Tokyo, pp 73–81. Eisen MB, Spellman PT, Brown PO and Botstein D. (1998). Benzil DL, Finkelstein SD, Epstein MH and Finch PW. Proc. Natl. Acad. Sci. USA, 95, 14863–14868. (1992). Cancer Res., 52, 2951–2956. Fong KM, Sekido Y and Minna JD. (1999). J. Thorac. Bhattacharjee A, Richards WG, Staunton J, Li C, Monti S, Cardiovasc. Surg., 118, 1136–1152. Vasa P, Ladd C, Beheshti J, Bueno R, Gillette M, Loda M, Garber ME, Troyanskaya OG, Schluens K, Petersen S, Weber G, Mark EJ, Lander ES, Wong W, Johnson BE, Thaesler Z, Pacyna-Gengelbach M, van de Rijn M, Rosen Golub TR, Sugarbaker DJ and Meyerson M. (2001). Proc. GD, Perou CM, Whyte RI, Altman RB, Brown PO, Botstein Natl. Acad. Sci. USA, 98, 13790–13795. D and Petersen I. (2001). Proc. Natl. Acad. Sci. USA, 98, Borges M, Linnoila RI, van de Velde HJ, Chen H, Nelkin BD, 13784–13789. Mabry M, Baylin SB and Ball DW. (1997). Nature, 386, Gazdar AF, Carney DN, Nau MM and Minna JD. (1985). 852–855. Cancer Res., 45, 2924–2930. Botchkarev VA, Kief S, Paus R and Moll I. (1999). J. Invest. Goessling W, McKee PH and Mayer RJ. (2002). J. Clin. Dermatol., 113, 691–692. Oncol., 20, 588–598. Broers JL, Carney DN, de Ley L, Vooijs GPand Ramaekers Halata Z, Grim M and Bauman KI. (2003). Anat. Rec., 271A, FC. (1985). Proc. Natl. Acad. Sci. USA, 82, 4409–4413. 225–239. Broers JL, Carney DN, Klein Rot M, Schaart G, Lane EB, Iavarone A, Garg P, Lasorella A, Hsu J and Israel MA. (1994). Vooijs GPand Ramaekers FC. (1986). J. Cell Sci., 83, Genes Dev., 8, 1270–1284. 37–60. Leonard JH and Bell JR. (1997). Australas. J. Dermatol., 38, Buchner K. (2000). J. Cancer Res. Clin. Oncol., 126, 1–11. 91–98. Carney DN, Gazdar AF, Bepler G, Guccion JG, Marangos Leonard JH, Bell JR and Kearsley JH. (1993). Int. J. Cancer, PJ, Moody TW, Zweig MH and Minna JD. (1985). Cancer 55, 803–810. Res., 45, 2913–2923. Leonard JH, Cook AL, Nancarrow D, Hayward N, Van Gele Carroll P, Lewin GR, Koltzenburg M, Toyka KV and M, Van Roy N and Speleman F. (2000). Cancer Detect. Thoenen H. (1998). Nat. Neurosci., 1, 42–46. Prev., 24, 620–627.

Oncogene Gene-expression proﬁling of MCC and SCLC MV Gele et al 2742 Leonard JH, Cook AL, Van Gele M, Boyle GM, Inglis KJ, Ronan SG, Green AD, Shilkaitis A, Huang TS and Das Gupta Speleman F and Sturm RA. (2002). Int. J. Cancer, 101, 103– TK. (1993). J. Am. Acad. Dermatol., 29, 715–722. 110. Rosen ST, Gould VE, Salwen HR, Herst CV, Le Beau MM, Leonard JH, Dash P, Holland P, Kearsley JH and Bell JR. Lee I, Bauer K, Marder RJ, Andersen R, Kies MS, Moll R, (1995a). Int. J. Cancer, 60, 100–107. Franke WW and Radosevich JA. (1987). Lab. Invest., 56, Leonard JH and Hayard N. (1997). Genes Chromosomes 302–312. Cancer, 20, 93–97. Scaglione-Sewell B, Abraham C, Bissonnette M, Skarosi SF, Leonard JH, Ramsay JR, Kearsley JH and Birrell GW. Hart J, Davidson NO, Wali RK, Davis BH, Sitrin M and (1995b). Int. J. Radiat. Oncol. Biol. Phys., 32, 1401–1407. Brasitus TA. (1998). Cancer Res., 58, 1074–1081. Levin NA, Brzoska PM, Warnock ML, Gray JW and Schmidt U, Muller U, Metz KA and Leder LD. (1998). Am. J. Christman MF. (1995). Genes Chromosomes Cancer, 13, Dermatopathol., 20, 346–351. 175–185. Tazaki M, Tazaki Y, Inoue T and Shimono M. (2000). Merkel Metz KA, Jacob M, Schmidt U, Steuhl KPand Leder LD. Cells, Merkel Cell Carcinoma and Neurobiology of the Skin, (1998). Graefe’s Arch. Clin. Exp. Ophthalmol., 236, Suzuki H and Ono T (eds) Elsevier Science B.V.: Tokyo, pp. 561–566. 63–71. Miller RW and Rabkin CS. (1999). Cancer Epidemiol. Tusher VG, Tibshirani R and Chu G. (2001). Proc. Natl. Acad. Biomarkers Prev., 8, 153–158. Sci. USA, 98, 5116–5121. Minna JD, Roth JA and Gazdar AF. (2002). Cancer Cell, 1, Vandesompele J, De Paepe A and Speleman F. (2002a). Anal. 49–52. Biochem., 303, 95–98. Moll I, Gillardon F, Waltering S, Schmelz M and Moll R. Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N, (1996). J. Cutan. Pathol., 23, 109–117. De Paepe A and Speleman F. (2002b). Genome Biol., 3, Pattyn F, Speleman F, De Paepe A and Vandesompele research0034.001–research0034.0011. J. (2003). Nucleic Acids Res., 31, 122–123. Van Gele M, Kaghad M, Leonard JH, Van Roy N, Naeyaert Petersen I, Langreck H, Wolf G, Schwendel A, Psille R, Vogt JM, Geerts ML, Van Belle S, Cocquyt V, Bridge J, Sciot R, P, Reichel MB, Ried T and Dietel M. (1997). Br. J. Cancer, De Wolf-Peeters C, De Paepe A, Caput D and Speleman F. 75, 79–86. (2000). Br. J. Cancer, 82, 823–826. Pedersen N, Mortensen S, Sorensen SB, Pedersen MW, Van Gele M, Leonard JH, Van Limbergen H, Van Roy N, Rieneck K, Bovin LF and Poulsen HS. (2003). Cancer Van Belle S, Cocquyt V, De Paepe A, Salwen H and Res., 63, 1943–1953. Speleman F. (2002). Int. J. Cancer, 101, 137–145. Popp S, Waltering S, Herbst C, Moll I and Boukamp P. Van Gele M, Speleman F, Vandesompele J, Van Roy N and (2002). Int. J. Cancer, 99, 352–360. Leonard JH. (1998). Cancer Res., 58, 1503–1508. Ratner D, Nelson BR, Brown MD and Johnson TM. (1993). Zajchowski DA, Bartholdi MF, Gong Y, Webster L, Liu HL, J. Am. Acad. Dermatol., 29, 143–156. Munishkin A, Beauheim C, Harvey S, Ethier SPand Ried T, Petersen I, Holtgreve-Grez H, Speicher MR, Schrock Johnson PH. (2001). Cancer Res., 61, 5168–5178. E, du Manoir S and Cremer T. (1994). Cancer Res., 54, Zhang L, Cilley RE and Chinoy MR. (2000). J. Surg. Res., 93, 1801–1806. 108–119.

Oncogene